diff --git a/ARTIFACTS.json b/ARTIFACTS.json new file mode 100644 index 0000000000000000000000000000000000000000..70f5e49a0c76a1a496bebbd4f8870c28f85e98e6 --- /dev/null +++ b/ARTIFACTS.json @@ -0,0 +1,4500 @@ +{ + "schema": "research-copy-artifacts.v1", + "repository": "Xnhyacinth/LongWorld-Research-Data", + "source_repository": "Xnhyacinth/LongWorld-Real-Workflows", + "source_revision": "c51bfb3d6f9fac2823a4beb57b2d3750292baabc", + "source_file_count": 374, + "original_history_preserved": true, + "inference_executed_for_this_publication": false, + "privacy_scope": "All source-file bytes, canonicalized text and all parsed candidate-bearing research records reviewed. Public source documents, code/specification fixtures, attribution and fictional example-domain contacts are preserved. Only approved operational manifest path prefixes and root navigation change.", + "files": [ + { + "path": ".gitattributes", + "source_git_blob": "f9e738f184fa84e3e4bd7ef4ba97e255935ae98c", + "source_lfs_sha256": null, + "source_sha256": "2c2791d2d83662860c1d83c96e3d3d8eb7da4f1ba1b40f62855549dcb813bde3", + "source_bytes": 3854, + "public_sha256": "2c2791d2d83662860c1d83c96e3d3d8eb7da4f1ba1b40f62855549dcb813bde3", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "05_promoted/eval.jsonl", + "source_git_blob": "7442f2e13046190f4a0776bd3c9cde07cb680e98", + "source_lfs_sha256": "cc60fb0c666c593a02a2103721942b2dbe036ef3563e53973aa854cb6885030f", + "source_sha256": "cc60fb0c666c593a02a2103721942b2dbe036ef3563e53973aa854cb6885030f", + "source_bytes": 48540761, + "public_sha256": "cc60fb0c666c593a02a2103721942b2dbe036ef3563e53973aa854cb6885030f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "05_promoted/quality_report.json", + "source_git_blob": "859b966d72776daef8433a3af1078bf728de6418", + "source_lfs_sha256": null, + "source_sha256": "54d2048ae1ab5b67af687235470cc7f25bc451b99eb09968e6c04f2b66015f3d", + "source_bytes": 2558, + "public_sha256": "54d2048ae1ab5b67af687235470cc7f25bc451b99eb09968e6c04f2b66015f3d", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "05_promoted/release_gate_pass.json", + "source_git_blob": "9cbfb7f6c39cd366d384490261360038d32b1985", + "source_lfs_sha256": null, + "source_sha256": "fbab5136b57920281a0917cc287efe17af87581f1bd2a6191a7c9aef44dc4c5b", + "source_bytes": 1064, + "public_sha256": "fbab5136b57920281a0917cc287efe17af87581f1bd2a6191a7c9aef44dc4c5b", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "05_promoted/train.jsonl", + "source_git_blob": "2fb736fe4491b9cb6e3489e0cc8369edd0f26ae8", + "source_lfs_sha256": "6e8dd2b8752b0c71445770119ee0f7295208154d8ae6248a91170d3ee0ce1fab", + "source_sha256": "6e8dd2b8752b0c71445770119ee0f7295208154d8ae6248a91170d3ee0ce1fab", + "source_bytes": 161851783, + "public_sha256": "6e8dd2b8752b0c71445770119ee0f7295208154d8ae6248a91170d3ee0ce1fab", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "06_training/B1.json", + "source_git_blob": "5f9d4505de57950c209f8aafabc047742c2565c4", + "source_lfs_sha256": "c396513770b07cf07fa6c1909bea461fe8c65ee49142ae16a6d73db461749f1b", + "source_sha256": "c396513770b07cf07fa6c1909bea461fe8c65ee49142ae16a6d73db461749f1b", + "source_bytes": 14210676, + "public_sha256": "c396513770b07cf07fa6c1909bea461fe8c65ee49142ae16a6d73db461749f1b", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "06_training/B1.meta.json", + "source_git_blob": "bf4ef587dcfb2220b7b9ebbb184753e46c69a282", + "source_lfs_sha256": null, + "source_sha256": "9f2bf4f256c0d3ec2d8bb85392116f25327955cadb77ff601748feeeb58f2dc8", + "source_bytes": 398, + "public_sha256": "9f2bf4f256c0d3ec2d8bb85392116f25327955cadb77ff601748feeeb58f2dc8", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "06_training/B3.json", + "source_git_blob": "e9dbc6a33cfd81a4397920ed124e1aefd6dae65c", + "source_lfs_sha256": "f78e4d906a5aeacc5dcccc324b7d95fbb5d85d0b20b787c21e285b8d806ead1f", + "source_sha256": "f78e4d906a5aeacc5dcccc324b7d95fbb5d85d0b20b787c21e285b8d806ead1f", + "source_bytes": 14114164, + "public_sha256": "f78e4d906a5aeacc5dcccc324b7d95fbb5d85d0b20b787c21e285b8d806ead1f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "06_training/B3.meta.json", + "source_git_blob": "bb635a682daaffd2cfb4a41a17128ea15a4cfaf4", + "source_lfs_sha256": null, + "source_sha256": "eeec97a1bce2cf5f376f0745bc396e6654d1e71e64dd47b1fc6930c76a967c38", + "source_bytes": 469, + "public_sha256": "eeec97a1bce2cf5f376f0745bc396e6654d1e71e64dd47b1fc6930c76a967c38", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "06_training/B5.json", + "source_git_blob": "34e6e68c21f8f740e669e7a2e790334c6f09ec86", + "source_lfs_sha256": "8520d823419cdbe4093a2e5f1b00541ca50e540a4dc93ff313257ccacffaa0f2", + "source_sha256": "8520d823419cdbe4093a2e5f1b00541ca50e540a4dc93ff313257ccacffaa0f2", + "source_bytes": 14163448, + "public_sha256": "8520d823419cdbe4093a2e5f1b00541ca50e540a4dc93ff313257ccacffaa0f2", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "06_training/B5.meta.json", + "source_git_blob": "18a8db6f208e3f3b2f5684df099cf6c4c92bc437", + "source_lfs_sha256": null, + "source_sha256": "b95b7a956a9ab828efd403c840656b656880a2d06a509c0d29d0fe94bc306668", + "source_bytes": 622, + "public_sha256": "b95b7a956a9ab828efd403c840656b656880a2d06a509c0d29d0fe94bc306668", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "06_training/B5w.datasets.yaml", + "source_git_blob": "ee8eaf418e5a61d26965d82f51d00f157612e4a7", + "source_lfs_sha256": null, + "source_sha256": "25c6f926040f5608b3f95529c5b05a667bd69d16f309e201a9a7aee79b390b33", + "source_bytes": 88, + "public_sha256": "25c6f926040f5608b3f95529c5b05a667bd69d16f309e201a9a7aee79b390b33", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "06_training/B5w.json", + "source_git_blob": "8f4eea723a9091c4ed3e53871c48d35abf4c8d0f", + "source_lfs_sha256": null, + "source_sha256": "7da381371e7ca74efdba413544fd25e3bbdbd505bcea1ed163712b23af8e96b4", + "source_bytes": 7057556, + "public_sha256": "7da381371e7ca74efdba413544fd25e3bbdbd505bcea1ed163712b23af8e96b4", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "06_training/B5w.meta.json", + "source_git_blob": "501ccbc42ece1fcf97007c22f9e038150fc6a729", + "source_lfs_sha256": null, + "source_sha256": "1053655e0bd0be595b7f8a2220e70fa8125cc3617ac5b757b2d30577e9da5a76", + "source_bytes": 621, + "public_sha256": "1053655e0bd0be595b7f8a2220e70fa8125cc3617ac5b757b2d30577e9da5a76", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "06_training/dataset_info.json", + "source_git_blob": "a63f667ea2aee1b69d36b64f80c1d3a7abef3150", + "source_lfs_sha256": null, + "source_sha256": "7bd596940721516a48c6e079950e66a9e0f06224a23f9157f1f98329d1e23220", + "source_bytes": 915, + "public_sha256": "7bd596940721516a48c6e079950e66a9e0f06224a23f9157f1f98329d1e23220", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "06_training/export_summary.json", + "source_git_blob": "b79f0d460106946a12476a22a3ab78f50e9b6480", + "source_lfs_sha256": null, + "source_sha256": "bb7fd8268896188ebcedeb8434428c38c12a5cb89403038475f8bc205c5b8347", + "source_bytes": 2853, + "public_sha256": "bb7fd8268896188ebcedeb8434428c38c12a5cb89403038475f8bc205c5b8347", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "06_training/training_export_manifest.json", + "source_git_blob": "019ad2fa143225e3c9c58fee496d257ef3ede204", + "source_lfs_sha256": null, + "source_sha256": "d1a852bfd667dac05d82dac04261ecae005fb86a2d83bbba80779a53639290c9", + "source_bytes": 2341, + "public_sha256": "d1a852bfd667dac05d82dac04261ecae005fb86a2d83bbba80779a53639290c9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "COMMITTED", + "source_git_blob": "a448823046d7461648a6632276b527a563da5641", + "source_lfs_sha256": null, + "source_sha256": "6790e26b41ef7385445d9201d74d7fc46dcef19a02c0cccdfa306e3a0e55e49b", + "source_bytes": 147, + "public_sha256": "6790e26b41ef7385445d9201d74d7fc46dcef19a02c0cccdfa306e3a0e55e49b", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "PUBLIC_BACKUP_STATUS_20260930.md", + "source_git_blob": "9b26379575220b7ffe26c7e5ab95664e86d4bdb0", + "source_lfs_sha256": null, + "source_sha256": "418228283dc4d53159b43df468a51a73a4419fbb0165017f592028a843e120a0", + "source_bytes": 4248, + "public_sha256": "418228283dc4d53159b43df468a51a73a4419fbb0165017f592028a843e120a0", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "README.md", + "source_git_blob": "b681b80c1d78c9f7d4bd26e4db1fbb444c9607a3", + "source_lfs_sha256": null, + "source_sha256": "8704e3ade19f126d66fe2a9340f8f0179e928c8c086cbf00700fd1af3af53e5d", + "source_bytes": 4082, + "public_sha256": "0c7c145994412365a3a387b6330e892f16f8c949c8472dcd069c1e07c49961a6", + "operation": "add_sanitized_file", + "audit_status": "passed", + "context_decision": "Public project navigation links retained; root research-product description clarified", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/eval.jsonl", + "source_git_blob": "216bada322a003ca12c637ef036fb86eceaabfee", + "source_lfs_sha256": null, + "source_sha256": "ad09745c2debc60340946b59439c40fc2f997ef387b4fec2d772f11d9c25ab4c", + "source_bytes": 7389938, + "public_sha256": "ad09745c2debc60340946b59439c40fc2f997ef387b4fec2d772f11d9c25ab4c", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/llamafactory/B1.json", + "source_git_blob": "cdc0396f1fb5a21a86da9c58c484e6d607ea5262", + "source_lfs_sha256": null, + "source_sha256": "50277efcd0d46b0b995dfa31ef5dd2dc63b5bd70025d58f68e6f52e3d8ff8096", + "source_bytes": 3840049, + "public_sha256": "50277efcd0d46b0b995dfa31ef5dd2dc63b5bd70025d58f68e6f52e3d8ff8096", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/llamafactory/B1.meta.json", + "source_git_blob": "0b3204038dc22e8f2feafda3aa5e4616e58e8784", + "source_lfs_sha256": null, + "source_sha256": "a2d21fbf69d3b06c39075d8d55cd3c914a9b04467eecb92b28b6a19e94b562ef", + "source_bytes": 398, + "public_sha256": "a2d21fbf69d3b06c39075d8d55cd3c914a9b04467eecb92b28b6a19e94b562ef", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/llamafactory/B3.json", + "source_git_blob": "6a3218682d591ad4ed31dd739165daa39df023b6", + "source_lfs_sha256": null, + "source_sha256": "baf23e84d5dd0f7818a6a5820a7b6ae9be28707cdf8df07dbe6c94b1b1ac392d", + "source_bytes": 4018299, + "public_sha256": "baf23e84d5dd0f7818a6a5820a7b6ae9be28707cdf8df07dbe6c94b1b1ac392d", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/llamafactory/B3.meta.json", + "source_git_blob": "5716469c8075148670aa480d873db3cad1edd8d5", + "source_lfs_sha256": null, + "source_sha256": "078359e3bb257b76c246e47ed0b14ec9550e7dddab46d199d3006373e0a65054", + "source_bytes": 458, + "public_sha256": "078359e3bb257b76c246e47ed0b14ec9550e7dddab46d199d3006373e0a65054", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/llamafactory/B5.json", + "source_git_blob": "75b26e2aced55fdf4f6c73b155ba3b63b0b0d02c", + "source_lfs_sha256": null, + "source_sha256": "991591c004843e50d24ac5cd5fc84132a2e6518f098a5d1258fa0a6f01f23546", + "source_bytes": 3838162, + "public_sha256": "991591c004843e50d24ac5cd5fc84132a2e6518f098a5d1258fa0a6f01f23546", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/llamafactory/B5.meta.json", + "source_git_blob": "2389e35adf558baae4e5128ea74bda6a95362430", + "source_lfs_sha256": null, + "source_sha256": "4b28e9d00641ec5938033af92e87301c6a52a4c8bcffd9ae78e2865dae6c9064", + "source_bytes": 599, + "public_sha256": "4b28e9d00641ec5938033af92e87301c6a52a4c8bcffd9ae78e2865dae6c9064", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/llamafactory/B5w.datasets.yaml", + "source_git_blob": "ee8eaf418e5a61d26965d82f51d00f157612e4a7", + "source_lfs_sha256": null, + "source_sha256": "25c6f926040f5608b3f95529c5b05a667bd69d16f309e201a9a7aee79b390b33", + "source_bytes": 88, + "public_sha256": "25c6f926040f5608b3f95529c5b05a667bd69d16f309e201a9a7aee79b390b33", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/llamafactory/B5w.json", + "source_git_blob": "8017f8560edfa90fcff10a5e3ae57502ed0c650c", + "source_lfs_sha256": null, + "source_sha256": "662e1c8f638266ef4b91ca6afe37efcc79e01e12980eb349aa6b15a642251efe", + "source_bytes": 1935722, + "public_sha256": "662e1c8f638266ef4b91ca6afe37efcc79e01e12980eb349aa6b15a642251efe", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/llamafactory/B5w.meta.json", + "source_git_blob": "fd8418180c063329b107ab5a54f2c77a6890c635", + "source_lfs_sha256": null, + "source_sha256": "8712dd9ff29c6d9465041bbf6ee48aafff0afed01caad61e383860456991e65e", + "source_bytes": 596, + "public_sha256": "8712dd9ff29c6d9465041bbf6ee48aafff0afed01caad61e383860456991e65e", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/llamafactory/dataset_info.json", + "source_git_blob": "a63f667ea2aee1b69d36b64f80c1d3a7abef3150", + "source_lfs_sha256": null, + "source_sha256": "7bd596940721516a48c6e079950e66a9e0f06224a23f9157f1f98329d1e23220", + "source_bytes": 915, + "public_sha256": "7bd596940721516a48c6e079950e66a9e0f06224a23f9157f1f98329d1e23220", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/llamafactory/export_summary.json", + "source_git_blob": "afbba16ae9825aaaa211de6f4a0e76b77ed217a2", + "source_lfs_sha256": null, + "source_sha256": "371d697d0e92cd0c0bf7a433a50b9394fec51ee413e87978b5f9090c8078e85e", + "source_bytes": 2785, + "public_sha256": "371d697d0e92cd0c0bf7a433a50b9394fec51ee413e87978b5f9090c8078e85e", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/llamafactory/training_export_manifest.json", + "source_git_blob": "37b3e3e535ffe03588f05db844a966f27a9b1b95", + "source_lfs_sha256": null, + "source_sha256": "940c5144a9f45bc77f68195ab7ec71a2174e330f6ced103d61e580e7efdc9bd5", + "source_bytes": 2433, + "public_sha256": "940c5144a9f45bc77f68195ab7ec71a2174e330f6ced103d61e580e7efdc9bd5", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/quality_report.json", + "source_git_blob": "062133b319268f9164b7fec4c3c3208076020dc1", + "source_lfs_sha256": null, + "source_sha256": "b400f7f1a70f072d4850f099f64b0cc580258aa58de31b384e6819c04d87dbad", + "source_bytes": 3802, + "public_sha256": "b400f7f1a70f072d4850f099f64b0cc580258aa58de31b384e6819c04d87dbad", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/release_gate_receipt.json", + "source_git_blob": "26c23e1ccac6beb41ba45120266c8171629c64f7", + "source_lfs_sha256": null, + "source_sha256": "3a14f29d4311983ac5300a6c78acbd83e73a9658eb9bf82cfe45fad7dfe226d8", + "source_bytes": 1803, + "public_sha256": "3a14f29d4311983ac5300a6c78acbd83e73a9658eb9bf82cfe45fad7dfe226d8", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p14-authentic-six-domain-probe-12-v1-promoted-v4/train.jsonl", + "source_git_blob": "8c3c1954c6b8fc3fde4dcebb4a937220dd39e1cd", + "source_lfs_sha256": "f3946b7edac88cc1d7dc5fe01a3430dd54924eaade0c781255ac961ce2af573d", + "source_sha256": "f3946b7edac88cc1d7dc5fe01a3430dd54924eaade0c781255ac961ce2af573d", + "source_bytes": 33094068, + "public_sha256": "f3946b7edac88cc1d7dc5fe01a3430dd54924eaade0c781255ac961ce2af573d", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "local-probe-train-ready/p15-authentic-128k-extension-probe-2-v1-promoted-v7/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p15-authentic-128k-extension-probe-2-v1-promoted-v7/llamafactory/B5.json", + "source_git_blob": "566bacb07f72e850379bf7e999686b87d2ad6b01", + "source_lfs_sha256": null, + "source_sha256": "30dcac47c46895bf38de621c18cbec2c06851ef050c045940be953e6b8e91049", + "source_bytes": 4053124, + "public_sha256": "30dcac47c46895bf38de621c18cbec2c06851ef050c045940be953e6b8e91049", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p15-authentic-128k-extension-probe-2-v1-promoted-v7/llamafactory/B5.meta.json", + "source_git_blob": "530a43e5aa082f951c73f09fb811db4600ea94fc", + "source_lfs_sha256": null, + "source_sha256": "f7e52e68ed0b8dc286fa489a9c5638f8a8ff85136caad7901ded7208fe5dd5ba", + "source_bytes": 513, + "public_sha256": "f7e52e68ed0b8dc286fa489a9c5638f8a8ff85136caad7901ded7208fe5dd5ba", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p15-authentic-128k-extension-probe-2-v1-promoted-v7/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p15-authentic-128k-extension-probe-2-v1-promoted-v7/llamafactory/export_summary.json", + "source_git_blob": "63527e32e126c31cfc4fd8732bf4829769f4e187", + "source_lfs_sha256": null, + "source_sha256": "630412424b7eee8fec5965626ac14e81cb7f4882501b969ee57b73d54ac8edd8", + "source_bytes": 705, + "public_sha256": "630412424b7eee8fec5965626ac14e81cb7f4882501b969ee57b73d54ac8edd8", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p15-authentic-128k-extension-probe-2-v1-promoted-v7/llamafactory/training_export_manifest.json", + "source_git_blob": "c35bb66af4f64b677f45e622a03618076128ac28", + "source_lfs_sha256": null, + "source_sha256": "86d66ee6b292ba039bb5970b622565952b5976b244c70cada4ee46c99b2c4e75", + "source_bytes": 1515, + "public_sha256": "86d66ee6b292ba039bb5970b622565952b5976b244c70cada4ee46c99b2c4e75", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p15-authentic-128k-extension-probe-2-v1-promoted-v7/quality_report.json", + "source_git_blob": "74918350e103cd9c6b76b6530691927c26a618da", + "source_lfs_sha256": null, + "source_sha256": "96f1c5b1110c00190b91940d89374196c2db5fd9c07ac82281e2843a0dc44283", + "source_bytes": 2531, + "public_sha256": "96f1c5b1110c00190b91940d89374196c2db5fd9c07ac82281e2843a0dc44283", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p15-authentic-128k-extension-probe-2-v1-promoted-v7/release_gate_receipt.json", + "source_git_blob": "7901c8ecd29c9141fdac0cf92edeb5ae0ac34a2c", + "source_lfs_sha256": null, + "source_sha256": "95c7ae0381728268e2962acab23acb4e08ca556dbe3704b5df139e68b92a91bf", + "source_bytes": 1804, + "public_sha256": "95c7ae0381728268e2962acab23acb4e08ca556dbe3704b5df139e68b92a91bf", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p15-authentic-128k-extension-probe-2-v1-promoted-v7/train.jsonl", + "source_git_blob": "ef0c8fac82e8136f35f59655a3effd12e1d875b4", + "source_lfs_sha256": "7045ac04b301c7819d98d3bda4379cb99baa3aecdc08700d035b6d5ad86df8d5", + "source_sha256": "7045ac04b301c7819d98d3bda4379cb99baa3aecdc08700d035b6d5ad86df8d5", + "source_bytes": 14745403, + "public_sha256": "7045ac04b301c7819d98d3bda4379cb99baa3aecdc08700d035b6d5ad86df8d5", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "local-probe-train-ready/p16-macro-bea-128k-extension-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p16-macro-bea-128k-extension-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "cd3f1d6604581f3465d92f72923fb1a94078908a", + "source_lfs_sha256": null, + "source_sha256": "6a6cd40b5dfcc0a709d7bfecedbbd42ca0caced567dd81ad146255279ef95107", + "source_bytes": 1870845, + "public_sha256": "6a6cd40b5dfcc0a709d7bfecedbbd42ca0caced567dd81ad146255279ef95107", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p16-macro-bea-128k-extension-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "b0f2a1a855d922e1437b7d99fbbc1eecd839d07f", + "source_lfs_sha256": null, + "source_sha256": "dea576e7d2a04a232f3d0c9eb0bb891fe7caa15a256bfc939bcffd87b92fc823", + "source_bytes": 684, + "public_sha256": "dea576e7d2a04a232f3d0c9eb0bb891fe7caa15a256bfc939bcffd87b92fc823", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p16-macro-bea-128k-extension-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p16-macro-bea-128k-extension-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "c244431d6944a915f376d536aee68f99412da702", + "source_lfs_sha256": null, + "source_sha256": "9dfb3d60daf6c5cfd64dc1833497179f47c5f3ba84c1145ae42463badf608a51", + "source_bytes": 908, + "public_sha256": "9dfb3d60daf6c5cfd64dc1833497179f47c5f3ba84c1145ae42463badf608a51", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p16-macro-bea-128k-extension-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "10855166410660e689769990e7dc0bbc55f5ada8", + "source_lfs_sha256": null, + "source_sha256": "61d7117c95643df317aba580427b9dc2afe4f28b67d36d43bf1742b49090a5e9", + "source_bytes": 1515, + "public_sha256": "61d7117c95643df317aba580427b9dc2afe4f28b67d36d43bf1742b49090a5e9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p16-macro-bea-128k-extension-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "34a544d88922b13ea339ed967bee9f9031dbd4a6", + "source_lfs_sha256": null, + "source_sha256": "021f2f7d077505cd9d9570ccbbfcd5c051ed544dbb01ecc12e40a3dadef79d08", + "source_bytes": 2741, + "public_sha256": "021f2f7d077505cd9d9570ccbbfcd5c051ed544dbb01ecc12e40a3dadef79d08", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p16-macro-bea-128k-extension-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "a4db4eb2fa4925ac9b8e11f5718d6d30e25af4a6", + "source_lfs_sha256": null, + "source_sha256": "b9b2e5ae2ce478e632d0f366945e86e02d07a48374a4abb7c4e0cb5623988ccc", + "source_bytes": 2487, + "public_sha256": "b9b2e5ae2ce478e632d0f366945e86e02d07a48374a4abb7c4e0cb5623988ccc", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p16-macro-bea-128k-extension-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "0942fb9dc9d8c773d50530975a1e77dc389d296f", + "source_lfs_sha256": null, + "source_sha256": "002b4f0b25d5340928eadbf5ac768a3f219f10ea59f8fae3045a559f207cecf7", + "source_bytes": 1804, + "public_sha256": "002b4f0b25d5340928eadbf5ac768a3f219f10ea59f8fae3045a559f207cecf7", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p16-macro-bea-128k-extension-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "1726e678788b8386758a81c2cd4157833e089a83", + "source_lfs_sha256": null, + "source_sha256": "1ac9c734a095157a009e1a2c6f714d34b126b9cb84c73b4a3a6a6cb265c3ec1f", + "source_bytes": 6654307, + "public_sha256": "1ac9c734a095157a009e1a2c6f714d34b126b9cb84c73b4a3a6a6cb265c3ec1f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-codeforge-128k-extension-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-codeforge-128k-extension-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "2203cf7b591f7376654059e14df2f7afbb7fb3b3", + "source_lfs_sha256": null, + "source_sha256": "88b7d292d3cb28db46ac51842aaec3b2fd9996b48bc0e7fdfbf9c8520c143749", + "source_bytes": 1766731, + "public_sha256": "88b7d292d3cb28db46ac51842aaec3b2fd9996b48bc0e7fdfbf9c8520c143749", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-codeforge-128k-extension-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "7312682a5ab3082aa22dbf76ebeb38c9c3ef4e91", + "source_lfs_sha256": null, + "source_sha256": "3030c9d97613f6059c53032029d3294ec2c0176120858b20051986dbe783594a", + "source_bytes": 509, + "public_sha256": "3030c9d97613f6059c53032029d3294ec2c0176120858b20051986dbe783594a", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-codeforge-128k-extension-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-codeforge-128k-extension-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "ec94821e9d2ee553c18894364ad7bcb2c1d21c8c", + "source_lfs_sha256": null, + "source_sha256": "f943396d111b702277513c5d8e7bafca609c5c7480267fcc947ec29de73d2da0", + "source_bytes": 701, + "public_sha256": "f943396d111b702277513c5d8e7bafca609c5c7480267fcc947ec29de73d2da0", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-codeforge-128k-extension-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "eee7e93df7fcf38212ef970b74436c837d2e46e7", + "source_lfs_sha256": null, + "source_sha256": "a38ef77fa4e482f830cb61e90faae2790ab11ed1533ce9128fc0b2bb87e1e361", + "source_bytes": 1515, + "public_sha256": "a38ef77fa4e482f830cb61e90faae2790ab11ed1533ce9128fc0b2bb87e1e361", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-codeforge-128k-extension-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "a9a843aca51cf56b9da2f13371a4a14c4f4d628e", + "source_lfs_sha256": null, + "source_sha256": "d27c40827f76a0e8a998ded8da201b3b9bc185c32833ad8cb9efb21ed5e5f48d", + "source_bytes": 2380, + "public_sha256": "d27c40827f76a0e8a998ded8da201b3b9bc185c32833ad8cb9efb21ed5e5f48d", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-codeforge-128k-extension-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "a9a843aca51cf56b9da2f13371a4a14c4f4d628e", + "source_lfs_sha256": null, + "source_sha256": "d27c40827f76a0e8a998ded8da201b3b9bc185c32833ad8cb9efb21ed5e5f48d", + "source_bytes": 2380, + "public_sha256": "d27c40827f76a0e8a998ded8da201b3b9bc185c32833ad8cb9efb21ed5e5f48d", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-codeforge-128k-extension-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "0d3d41e1c8d76e42b5042645ec3dffe3cf76de98", + "source_lfs_sha256": null, + "source_sha256": "e2efd70fdd3be960283aa0dc61869db486fe89e21c5114434e72d8ff5f4cc9ab", + "source_bytes": 1803, + "public_sha256": "e2efd70fdd3be960283aa0dc61869db486fe89e21c5114434e72d8ff5f4cc9ab", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-codeforge-128k-extension-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "a1a9739f75a508c3d48e04103f32d4032fc0f158", + "source_lfs_sha256": null, + "source_sha256": "3821133ea50eb4729368e728902aea1139102f4eab8e27dadfad4fb782ecdd9c", + "source_bytes": 5568448, + "public_sha256": "3821133ea50eb4729368e728902aea1139102f4eab8e27dadfad4fb782ecdd9c", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-finance-128k-extension-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-finance-128k-extension-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "e4faad93ab0b6260fba3deac038bf11b38fe33a7", + "source_lfs_sha256": null, + "source_sha256": "0e757a0ddababe55db088d5a8add796217eef7c61c4f0499b314b696dcc112c1", + "source_bytes": 2252551, + "public_sha256": "0e757a0ddababe55db088d5a8add796217eef7c61c4f0499b314b696dcc112c1", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-finance-128k-extension-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "d087621e39712b006cbba2d68e4571b488d7efd4", + "source_lfs_sha256": null, + "source_sha256": "b34aa97605eb3ad987a9355d7c392b30f3fe84cad226f785bd03d2078b9a1aa8", + "source_bytes": 684, + "public_sha256": "b34aa97605eb3ad987a9355d7c392b30f3fe84cad226f785bd03d2078b9a1aa8", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-finance-128k-extension-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-finance-128k-extension-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "64d6cfef32827b1ca2f12183a8e4cc9c29f7ef4b", + "source_lfs_sha256": null, + "source_sha256": "a4f934318e26c666dd95644012463a5263f6ad88c2d00d934137e593e786017c", + "source_bytes": 908, + "public_sha256": "a4f934318e26c666dd95644012463a5263f6ad88c2d00d934137e593e786017c", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-finance-128k-extension-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "358b1b0f79d4403bb51cacabd45f4019b24df012", + "source_lfs_sha256": null, + "source_sha256": "a63b73dda6332b9e63467ace3c72f85f9917ceaf6cf9135a8bf1fdd7752e7733", + "source_bytes": 1513, + "public_sha256": "a63b73dda6332b9e63467ace3c72f85f9917ceaf6cf9135a8bf1fdd7752e7733", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-finance-128k-extension-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "3e28fcd5e382645a16127cb1420de5fba685e610", + "source_lfs_sha256": null, + "source_sha256": "dca7bd9f16bd42e13caaed17e0dfa8c7a34f02168cfa7059646f0246fbd48cd1", + "source_bytes": 2462, + "public_sha256": "dca7bd9f16bd42e13caaed17e0dfa8c7a34f02168cfa7059646f0246fbd48cd1", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-finance-128k-extension-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "3e28fcd5e382645a16127cb1420de5fba685e610", + "source_lfs_sha256": null, + "source_sha256": "dca7bd9f16bd42e13caaed17e0dfa8c7a34f02168cfa7059646f0246fbd48cd1", + "source_bytes": 2462, + "public_sha256": "dca7bd9f16bd42e13caaed17e0dfa8c7a34f02168cfa7059646f0246fbd48cd1", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-finance-128k-extension-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "bd4f75f15f8c1b26223dff71d48e10670430e4bc", + "source_lfs_sha256": null, + "source_sha256": "bb3d6f5011bbf2030792da74611c3c7b2fb0e359664db0b6dda283d53ae5bf6f", + "source_bytes": 1802, + "public_sha256": "bb3d6f5011bbf2030792da74611c3c7b2fb0e359664db0b6dda283d53ae5bf6f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p17-finance-128k-extension-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "52ac73b511d855fb3d793a5c1c557a8ebc513342", + "source_lfs_sha256": null, + "source_sha256": "325766e9a165deba8725092dae9589ae5b7103402622795bee58e7825d1c8707", + "source_bytes": 7293956, + "public_sha256": "325766e9a165deba8725092dae9589ae5b7103402622795bee58e7825d1c8707", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p40-ietf-oauth-semantic-growth-probe-1-v1-promoted-v14/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p40-ietf-oauth-semantic-growth-probe-1-v1-promoted-v14/llamafactory/B5.json", + "source_git_blob": "38fc1a7698b092706536656deeb2dcd110b52d3c", + "source_lfs_sha256": null, + "source_sha256": "25ef1898f605d331d1239b308a0c979f5b8a2ffe6a58aeafbdadd59f6e2d4f6d", + "source_bytes": 3038077, + "public_sha256": "25ef1898f605d331d1239b308a0c979f5b8a2ffe6a58aeafbdadd59f6e2d4f6d", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p40-ietf-oauth-semantic-growth-probe-1-v1-promoted-v14/llamafactory/B5.meta.json", + "source_git_blob": "802117f44a85f242a1915508db0d242d4b31f8f3", + "source_lfs_sha256": null, + "source_sha256": "552f1a340aaeb5dd481389476f7335904fa099ea8c3884850c470c3907ff352c", + "source_bytes": 595, + "public_sha256": "552f1a340aaeb5dd481389476f7335904fa099ea8c3884850c470c3907ff352c", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p40-ietf-oauth-semantic-growth-probe-1-v1-promoted-v14/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p40-ietf-oauth-semantic-growth-probe-1-v1-promoted-v14/llamafactory/export_summary.json", + "source_git_blob": "490df6c55e460ef5ef3748c49e89ebc391f2b7d1", + "source_lfs_sha256": null, + "source_sha256": "e5da294ab917e30fee97ebe4dbb122a36b409c57971d6ac7016880ccc3b8c534", + "source_bytes": 803, + "public_sha256": "e5da294ab917e30fee97ebe4dbb122a36b409c57971d6ac7016880ccc3b8c534", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p40-ietf-oauth-semantic-growth-probe-1-v1-promoted-v14/llamafactory/training_export_manifest.json", + "source_git_blob": "5d0fc1ca2d7507c29c8dfd4c45d4228da1337966", + "source_lfs_sha256": null, + "source_sha256": "6ffbc7527b92c857ceaa2d6dddddc28444ef591413bb259f7ce04f9f831d2b7a", + "source_bytes": 1517, + "public_sha256": "6ffbc7527b92c857ceaa2d6dddddc28444ef591413bb259f7ce04f9f831d2b7a", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p40-ietf-oauth-semantic-growth-probe-1-v1-promoted-v14/quality_report.json", + "source_git_blob": "3f2edb03feff786d2e630d68bcf542f5391faa52", + "source_lfs_sha256": null, + "source_sha256": "c689cbbbe122289a06d2c74eabcca371828d8a00e1d95920b80d52766dd89345", + "source_bytes": 2437, + "public_sha256": "c689cbbbe122289a06d2c74eabcca371828d8a00e1d95920b80d52766dd89345", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p40-ietf-oauth-semantic-growth-probe-1-v1-promoted-v14/release_gate_receipt.json", + "source_git_blob": "650cecc7159c600dcebe4ba90a5bd4dccc87bf7d", + "source_lfs_sha256": null, + "source_sha256": "678c6342df6c9cb4deac9312390832e7b089bba4ec377b17298071a4b4e01bc2", + "source_bytes": 1805, + "public_sha256": "678c6342df6c9cb4deac9312390832e7b089bba4ec377b17298071a4b4e01bc2", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p40-ietf-oauth-semantic-growth-probe-1-v1-promoted-v14/train.jsonl", + "source_git_blob": "1e9f93642642cb94c1b758a582f9101f6916978d", + "source_lfs_sha256": "23efd8d5a55f4c5a06aa48ed7b3bde4a27a916241215c2b494e49ac07f19514a", + "source_sha256": "23efd8d5a55f4c5a06aa48ed7b3bde4a27a916241215c2b494e49ac07f19514a", + "source_bytes": 18003027, + "public_sha256": "23efd8d5a55f4c5a06aa48ed7b3bde4a27a916241215c2b494e49ac07f19514a", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "9035d23aa7917fe67e086e39e19e834ff3c316fb", + "source_lfs_sha256": null, + "source_sha256": "47162556569da4d38758692cc90787f688cae2d80d1a1d28037bc51eeac7ddd4", + "source_bytes": 1410, + "public_sha256": "47162556569da4d38758692cc90787f688cae2d80d1a1d28037bc51eeac7ddd4", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "f2bbbacad1bd486fa3ae91bdb690e88f57c1974c", + "source_lfs_sha256": null, + "source_sha256": "1d93f7362a979189d51e70891af2d187c21f0ea2e0beb9ef1e57c0abd81d163d", + "source_bytes": 1749174, + "public_sha256": "1d93f7362a979189d51e70891af2d187c21f0ea2e0beb9ef1e57c0abd81d163d", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "c4659bbb160db778534d264dd891b822ea039634", + "source_lfs_sha256": null, + "source_sha256": "a08f05a03faf98eea8dd8009fadeb82ec7d68c3016e92b1ffa2859a50d8c08e1", + "source_bytes": 684, + "public_sha256": "a08f05a03faf98eea8dd8009fadeb82ec7d68c3016e92b1ffa2859a50d8c08e1", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "887947a90bd438623c8b858a58b07a4b7dfe3db7", + "source_lfs_sha256": null, + "source_sha256": "1532739f77f3298eccd04669937ed2c15249e4e1f1a1744aecd1351f79334ef8", + "source_bytes": 908, + "public_sha256": "1532739f77f3298eccd04669937ed2c15249e4e1f1a1744aecd1351f79334ef8", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "d5e695d0f847d70af35aafc12afdebdd15d6f8bf", + "source_lfs_sha256": null, + "source_sha256": "d962b260ac68d43a86f8099b85f24c2ee0fce8add412bd969d11bdd28f35c299", + "source_bytes": 1513, + "public_sha256": "d962b260ac68d43a86f8099b85f24c2ee0fce8add412bd969d11bdd28f35c299", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/promoted_128k.jsonl", + "source_git_blob": "82fa94c517bf8c1a5b6ae04128560b40ec03db9d", + "source_lfs_sha256": null, + "source_sha256": "96c2ec7adfdf7ff8a038d16c970a65994cb514f931f3ff80839d8a521dddda07", + "source_bytes": 4651906, + "public_sha256": "96c2ec7adfdf7ff8a038d16c970a65994cb514f931f3ff80839d8a521dddda07", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/promoted_16k.jsonl", + "source_git_blob": "0f308d21011102797d908d90e16145b86746e069", + "source_lfs_sha256": null, + "source_sha256": "0e34194d4dc23794b830f6f5105a008e37da41a676bcd7579920257e22cafb20", + "source_bytes": 496760, + "public_sha256": "0e34194d4dc23794b830f6f5105a008e37da41a676bcd7579920257e22cafb20", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/promoted_32k.jsonl", + "source_git_blob": "a1227fd57542808690deb0610242947e383332ce", + "source_lfs_sha256": null, + "source_sha256": "44764b52e045ef2073f2c3ecd64bbf35de6523914cc5d94903d5ae0d2fa55ff1", + "source_bytes": 898494, + "public_sha256": "44764b52e045ef2073f2c3ecd64bbf35de6523914cc5d94903d5ae0d2fa55ff1", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/promoted_64k.jsonl", + "source_git_blob": "cdf4e4e37bc0dacbfe013563898a807167618092", + "source_lfs_sha256": null, + "source_sha256": "5c19f39c42d13cdf7d70701a26d275f071cf97b1cb2332d063157cd7fc62fd75", + "source_bytes": 1663462, + "public_sha256": "5c19f39c42d13cdf7d70701a26d275f071cf97b1cb2332d063157cd7fc62fd75", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "782e7d3b900fe1daa46ee5c290bda8a1f893467d", + "source_lfs_sha256": null, + "source_sha256": "1f9813dc18086ebbc23d3bb5a6b2fb8503953189e53305b2795ca091ab4d7817", + "source_bytes": 2455, + "public_sha256": "1f9813dc18086ebbc23d3bb5a6b2fb8503953189e53305b2795ca091ab4d7817", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "c367d2cb75817e2736bd7415641dd1431c68ba7b", + "source_lfs_sha256": null, + "source_sha256": "973144ffee20b315bb91a4e385039b8d5c45e5aac198d96f046a0167614569ca", + "source_bytes": 1802, + "public_sha256": "973144ffee20b315bb91a4e385039b8d5c45e5aac198d96f046a0167614569ca", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "69e8a946f5a7e3b08a1c78d78acdb747bf5553f5", + "source_lfs_sha256": null, + "source_sha256": "5956b4854dfce0107c3ee27af42d8a025d0530463ca8271900be1bd6898f97d7", + "source_bytes": 5916, + "public_sha256": "5956b4854dfce0107c3ee27af42d8a025d0530463ca8271900be1bd6898f97d7", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "b3a95b18c1ac8e5d6bf4ebb31fd57dce493dfc6a", + "source_lfs_sha256": null, + "source_sha256": "825d6c126068a6e35843d861fc509f9b818db1135212d3e241a4a3d7339c71f8", + "source_bytes": 7710622, + "public_sha256": "825d6c126068a6e35843d861fc509f9b818db1135212d3e241a4a3d7339c71f8", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "bb640effaa5a0ac93de41802f1d42cd5a8cfd828", + "source_lfs_sha256": null, + "source_sha256": "a348ca2cfceb5ca67f8d6ad32e06d04781c9f8d138192be6ec46652c59689f13", + "source_bytes": 2322398, + "public_sha256": "a348ca2cfceb5ca67f8d6ad32e06d04781c9f8d138192be6ec46652c59689f13", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p55-finance-alphabet-asset-breakdown-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "fe04c3bb297f659fc278bf90854ec3db0802be78", + "source_lfs_sha256": null, + "source_sha256": "8dc9dd94e78bce2b7103c721d8d6e8bd0897629bc7169c5b18c26bbdfdb8090d", + "source_bytes": 5611844, + "public_sha256": "8dc9dd94e78bce2b7103c721d8d6e8bd0897629bc7169c5b18c26bbdfdb8090d", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "9c1ebcb139656af7b3e40ccdea3c3d798dc5b994", + "source_lfs_sha256": null, + "source_sha256": "d4cbfc780f756da5e3d39c5b71e62f1390c8948d5e86e5838c1b1122aa8c95cd", + "source_bytes": 1521, + "public_sha256": "d4cbfc780f756da5e3d39c5b71e62f1390c8948d5e86e5838c1b1122aa8c95cd", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "6b10229532d09675a0c58da79712854c99a1a93d", + "source_lfs_sha256": null, + "source_sha256": "7da4d1d6f14a4eeffa8463205fe081116f7ef47d7ae2e95d2e238aa90e26b446", + "source_bytes": 1405, + "public_sha256": "7da4d1d6f14a4eeffa8463205fe081116f7ef47d7ae2e95d2e238aa90e26b446", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "34a0d8444b0209c1cc064207bae53be261fbec34", + "source_lfs_sha256": null, + "source_sha256": "32ad84b3d2d41658a049b8be6b1017bfd85bf6e84854dc9170e6dfe100272435", + "source_bytes": 1032188, + "public_sha256": "32ad84b3d2d41658a049b8be6b1017bfd85bf6e84854dc9170e6dfe100272435", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "87876aac143a0ead479df6c8bd12da40b6c0da5e", + "source_lfs_sha256": null, + "source_sha256": "91338a5757b55b0fa63122cb59e2acf009048df3974180c80667994463cfa7ca", + "source_bytes": 423, + "public_sha256": "91338a5757b55b0fa63122cb59e2acf009048df3974180c80667994463cfa7ca", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "b98dd9ef25d37b52cf54d862ba53f53cc84038ea", + "source_lfs_sha256": null, + "source_sha256": "a40b07c19ffb5d9bf2a6f36936e0de6f4f18d1eb2ec8d749f164ca275a1114c2", + "source_bytes": 599, + "public_sha256": "a40b07c19ffb5d9bf2a6f36936e0de6f4f18d1eb2ec8d749f164ca275a1114c2", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "89a6740bb69d1a2cfefa1c5a29be71ea7c20d521", + "source_lfs_sha256": null, + "source_sha256": "d028091341183c9fa1371c31ee2aff57b1a02de4bf66779814a8e92b0b4d8762", + "source_bytes": 1510, + "public_sha256": "d028091341183c9fa1371c31ee2aff57b1a02de4bf66779814a8e92b0b4d8762", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "fd3a3d2f013be521fecd55138ed7feaa517c6ca8", + "source_lfs_sha256": null, + "source_sha256": "605a483bf3ff9b4601e2471bdb4e5cbf6c52e1aacea684cfecd18f7549947484", + "source_bytes": 2658, + "public_sha256": "605a483bf3ff9b4601e2471bdb4e5cbf6c52e1aacea684cfecd18f7549947484", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "48e809713429ac0b5d261debfc2e0c7580b135d2", + "source_lfs_sha256": null, + "source_sha256": "40c8d8719a061ada8d81553b4ef19bc97df70ca8a9273f22b8b471c4ba7ec92b", + "source_bytes": 2416, + "public_sha256": "40c8d8719a061ada8d81553b4ef19bc97df70ca8a9273f22b8b471c4ba7ec92b", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "61ddb518c5eb054a18cd88188b540b9ea8aef98c", + "source_lfs_sha256": null, + "source_sha256": "f83a4da31ea20cbeaeb8ae826be9ba3b007bb58e43c351031975c7f6bc27be8e", + "source_bytes": 1798, + "public_sha256": "f83a4da31ea20cbeaeb8ae826be9ba3b007bb58e43c351031975c7f6bc27be8e", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "ce2030d8e7e588ea4452c3815cb792c3b4f92e4d", + "source_lfs_sha256": null, + "source_sha256": "345bf258d3cbeda48a085056a9b515e42d0c49d213038379258ad0f75c3639d8", + "source_bytes": 3122, + "public_sha256": "345bf258d3cbeda48a085056a9b515e42d0c49d213038379258ad0f75c3639d8", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "06395d1e6b3c8b4aae5b04bcdba911e21715a269", + "source_lfs_sha256": null, + "source_sha256": "04d2bc83412aae8d500d4bdc669aae07c1f0902fcf868ae24933ace68c708226", + "source_bytes": 2858, + "public_sha256": "04d2bc83412aae8d500d4bdc669aae07c1f0902fcf868ae24933ace68c708226", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "34a50d8934fa808eb319e5a84fde143f0c9c8289", + "source_lfs_sha256": null, + "source_sha256": "fc6dc26a30932bf1a568d235f9c4a9b5a3cf9e9d2129003d62ee39c707d2c1ba", + "source_bytes": 3814375, + "public_sha256": "fc6dc26a30932bf1a568d235f9c4a9b5a3cf9e9d2129003d62ee39c707d2c1ba", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "6236cd78b24bf4d261da3dc62ba131eff5be6d01", + "source_lfs_sha256": null, + "source_sha256": "c1d9fbcfde80891f4ed4eeed62fffed28f580fd0eb210682edbee80477479f22", + "source_bytes": 825757, + "public_sha256": "c1d9fbcfde80891f4ed4eeed62fffed28f580fd0eb210682edbee80477479f22", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-amazon-128k-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "123c473e00b4445e41ffe681c816df577b623078", + "source_lfs_sha256": null, + "source_sha256": "8aa16e4d53ff7e18e15c54d081df0d89ad9cd0a8fa0bbbdc847984dcbc5ac92a", + "source_bytes": 3040937, + "public_sha256": "8aa16e4d53ff7e18e15c54d081df0d89ad9cd0a8fa0bbbdc847984dcbc5ac92a", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "a02300d8ce9d5ef4f6d64f5383c18442919b44cd", + "source_lfs_sha256": null, + "source_sha256": "4e399458156f85e9a1ba010b1f823effd64fd24187eb530dda3593e06bc64843", + "source_bytes": 1519, + "public_sha256": "4e399458156f85e9a1ba010b1f823effd64fd24187eb530dda3593e06bc64843", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "d08a4405f5ef1bae86ea58ce6226e922fab82c3b", + "source_lfs_sha256": null, + "source_sha256": "a8061dd23bda9b52a3ef5867d90979529283d212683712cfa2ae5b03038f2287", + "source_bytes": 1403, + "public_sha256": "a8061dd23bda9b52a3ef5867d90979529283d212683712cfa2ae5b03038f2287", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "585cf36fc136f21aaf7db4a5d6e7b3a19bfde269", + "source_lfs_sha256": null, + "source_sha256": "a7b43d02303379ed4122f409473443e23fcce0ab6b03baba91713c8865fb46d0", + "source_bytes": 940205, + "public_sha256": "a7b43d02303379ed4122f409473443e23fcce0ab6b03baba91713c8865fb46d0", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "b4ac7ea58aab4aa12a81e048731cfa4b5f7fa0fd", + "source_lfs_sha256": null, + "source_sha256": "a961f4d8a951d3d2312c97fb085055b64de1b9be6c49fea2cde22f71e2727356", + "source_bytes": 423, + "public_sha256": "a961f4d8a951d3d2312c97fb085055b64de1b9be6c49fea2cde22f71e2727356", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "8cffa6a0806f739329462db3ab24eede2b0cedac", + "source_lfs_sha256": null, + "source_sha256": "1cbdccc5f69ff5dca360746c35640a5733adb90b8f129bd8913389849a5be451", + "source_bytes": 599, + "public_sha256": "1cbdccc5f69ff5dca360746c35640a5733adb90b8f129bd8913389849a5be451", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "31ee5b4e4b88b0853ac8cf35d6ce43cce439a91a", + "source_lfs_sha256": null, + "source_sha256": "e4336634a25f3d9bd5d4d513d509a1dd96404cc23ed8fa794ab5390f067957a2", + "source_bytes": 1507, + "public_sha256": "e4336634a25f3d9bd5d4d513d509a1dd96404cc23ed8fa794ab5390f067957a2", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "9680cd8ef5f784e7e573b39920094b57e2624091", + "source_lfs_sha256": null, + "source_sha256": "bc486b0fedb62d0242d492dd079870b9dfb0dd0258a2fa893a54025815991197", + "source_bytes": 2654, + "public_sha256": "bc486b0fedb62d0242d492dd079870b9dfb0dd0258a2fa893a54025815991197", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "d4025a3a69cc5569caa6258b8f28ffbdf52a68dc", + "source_lfs_sha256": null, + "source_sha256": "137e4c35d5b5979736bc020ec273c4fe0ac0fd28bc391f7353d442f5567d6e7e", + "source_bytes": 2412, + "public_sha256": "137e4c35d5b5979736bc020ec273c4fe0ac0fd28bc391f7353d442f5567d6e7e", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "1f8b32bec48065668c6987071b20ca5f8172466a", + "source_lfs_sha256": null, + "source_sha256": "735562b7d11abdb1eb4c7a30b97930115ccb2b812e75ea24e0071ce184be2fa0", + "source_bytes": 1796, + "public_sha256": "735562b7d11abdb1eb4c7a30b97930115ccb2b812e75ea24e0071ce184be2fa0", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "3aa7e5e6573704a0ada07e116ddfe207a950e4da", + "source_lfs_sha256": null, + "source_sha256": "52bc09aa19331c5c8ebfbd90d053669e59555882816292009011ffdd80169c05", + "source_bytes": 3110, + "public_sha256": "52bc09aa19331c5c8ebfbd90d053669e59555882816292009011ffdd80169c05", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "3f6201768d7ff1cb0468baee52b4478f66b7a4bd", + "source_lfs_sha256": null, + "source_sha256": "c6fc50e68b8e537c308c02b06d91a30d97c0c12aef4f4214869b8a054aa4ab09", + "source_bytes": 2846, + "public_sha256": "c6fc50e68b8e537c308c02b06d91a30d97c0c12aef4f4214869b8a054aa4ab09", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "930adf94db997d3fc70b4d7a390bdcee9d5585a8", + "source_lfs_sha256": null, + "source_sha256": "782d33bfd21a219fabdbe9a00dc8188d951c999bded93176d13149667f1f22fe", + "source_bytes": 3483344, + "public_sha256": "782d33bfd21a219fabdbe9a00dc8188d951c999bded93176d13149667f1f22fe", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "5e5b49f2cb40503f9cbbef8781690955c2f0e993", + "source_lfs_sha256": null, + "source_sha256": "82b15f572f6fcc8e5c1edeac1eea7243cfba476bb7c98ded216648a33d155648", + "source_bytes": 598972, + "public_sha256": "82b15f572f6fcc8e5c1edeac1eea7243cfba476bb7c98ded216648a33d155648", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-meta-128k-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "e32a177770e71245ef7db284a3dff7a51cea0f25", + "source_lfs_sha256": null, + "source_sha256": "f1401a12d94b715c6647d788ab6088541cfa64e9f6f420874b04de52b18addd0", + "source_bytes": 2927205, + "public_sha256": "f1401a12d94b715c6647d788ab6088541cfa64e9f6f420874b04de52b18addd0", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "3df8d1c8c4c3aabe71d60e00c6d4672fb96aa505", + "source_lfs_sha256": null, + "source_sha256": "bfea888cdf928bd16548d819403f4e41c13103f944445205da8daa170825e86d", + "source_bytes": 1526, + "public_sha256": "bfea888cdf928bd16548d819403f4e41c13103f944445205da8daa170825e86d", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "15318ec53e25a322c3bfb61a9c100eae7c3f8c81", + "source_lfs_sha256": null, + "source_sha256": "245cf4ed0cba076c94469f70515b3bb3ad18102cf8550b84979afd0b43546191", + "source_bytes": 1410, + "public_sha256": "245cf4ed0cba076c94469f70515b3bb3ad18102cf8550b84979afd0b43546191", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "8626a4b15fa4bd09433aebdcb419d182ee8c53b2", + "source_lfs_sha256": null, + "source_sha256": "acc79c3eca84624da5114019bb4d7f3a4cdc664ceab7a50ab17e214d454e270f", + "source_bytes": 1783605, + "public_sha256": "acc79c3eca84624da5114019bb4d7f3a4cdc664ceab7a50ab17e214d454e270f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "94d6a1e2b53038a474ab4bb0f0998a9c780895e7", + "source_lfs_sha256": null, + "source_sha256": "80dcf92c828c8b2a407dc319537b05fd27d055187870d714d941b511f5f27eb2", + "source_bytes": 684, + "public_sha256": "80dcf92c828c8b2a407dc319537b05fd27d055187870d714d941b511f5f27eb2", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "dfbc2ad84eed1fe804dc75fd671e4fad81a752d0", + "source_lfs_sha256": null, + "source_sha256": "69c510011a48c88a6fb9942869d3f80dc77c7ffb4d251383df3b4a651f099532", + "source_bytes": 908, + "public_sha256": "69c510011a48c88a6fb9942869d3f80dc77c7ffb4d251383df3b4a651f099532", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "87c77dcbe35f8e0ede975b72aa5394fa6cb437c4", + "source_lfs_sha256": null, + "source_sha256": "32fbab183e6d362fe81eaab01e4d970239d45704bb6ed137cc11c4f647bf3a65", + "source_bytes": 1513, + "public_sha256": "32fbab183e6d362fe81eaab01e4d970239d45704bb6ed137cc11c4f647bf3a65", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/promoted_128k.jsonl", + "source_git_blob": "355021d9ddb80e17f17de37dd562e8fd0e948eda", + "source_lfs_sha256": null, + "source_sha256": "a3991d7a57520892ca145577a99458f1d5369d803661ad39a2c7e8cc21c1be25", + "source_bytes": 5108892, + "public_sha256": "a3991d7a57520892ca145577a99458f1d5369d803661ad39a2c7e8cc21c1be25", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/promoted_16k.jsonl", + "source_git_blob": "b6b88c16e05045ff14e52c7f2b8faa3f2d804a53", + "source_lfs_sha256": null, + "source_sha256": "cca27ae46d73eda0c5abc04549f62590924e46d511e89e3a5eb02cb92b18f7d9", + "source_bytes": 503356, + "public_sha256": "cca27ae46d73eda0c5abc04549f62590924e46d511e89e3a5eb02cb92b18f7d9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/promoted_32k.jsonl", + "source_git_blob": "73a30210c9438e44e244505276be2a6b0f5b045c", + "source_lfs_sha256": null, + "source_sha256": "c6bd1fa618e22987b8c0135338893f71093ef82a2ba4075e4d20c357b2803105", + "source_bytes": 912147, + "public_sha256": "c6bd1fa618e22987b8c0135338893f71093ef82a2ba4075e4d20c357b2803105", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/promoted_64k.jsonl", + "source_git_blob": "7f696aebd421e94e20b108cf2d240d4b6924cd6d", + "source_lfs_sha256": null, + "source_sha256": "1ed24264df151dc0e89b75e8fb7e627a5b5f467de1fdd42d3709d5b8009917e6", + "source_bytes": 1688382, + "public_sha256": "1ed24264df151dc0e89b75e8fb7e627a5b5f467de1fdd42d3709d5b8009917e6", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "209ec34431d515a82e4b882b48f396a53296da75", + "source_lfs_sha256": null, + "source_sha256": "b92fce463c8e27131536174d5385df569fb5c118adfa51e48af0c6bd4f2e85a1", + "source_bytes": 2710, + "public_sha256": "b92fce463c8e27131536174d5385df569fb5c118adfa51e48af0c6bd4f2e85a1", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "5fa2155d620b00680ee14c0db01fb7b6d6e87c1c", + "source_lfs_sha256": null, + "source_sha256": "b0b1382daec50f3aed3d189bc393ba8f6639c5e06fe21870f7572c9c4eea08cf", + "source_bytes": 2456, + "public_sha256": "b0b1382daec50f3aed3d189bc393ba8f6639c5e06fe21870f7572c9c4eea08cf", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "fe4d36cf55b828fbb3b4e36b7436ddd0cea3218b", + "source_lfs_sha256": null, + "source_sha256": "c45178745464b6437b833deecfdd36caffd618eadc248dfdbdf2f554f8e3b9d1", + "source_bytes": 1802, + "public_sha256": "c45178745464b6437b833deecfdd36caffd618eadc248dfdbdf2f554f8e3b9d1", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "b209e3632d69dfd58b3bfa5f615f91eb94684be4", + "source_lfs_sha256": null, + "source_sha256": "7be0a0b8ea7593be8230e9cb18d6dff90228be7f50cd01280a552f0e35f5d0c9", + "source_bytes": 6293, + "public_sha256": "7be0a0b8ea7593be8230e9cb18d6dff90228be7f50cd01280a552f0e35f5d0c9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "89d81685b44d32cb096aeb43052be1ec9c6a3498", + "source_lfs_sha256": null, + "source_sha256": "c666c816bf9923a2a3e638da1d4c62472208615be0db69d4b5b6c3d345dc1669", + "source_bytes": 5921, + "public_sha256": "c666c816bf9923a2a3e638da1d4c62472208615be0db69d4b5b6c3d345dc1669", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "aa65afe33685d3b99dd63b84aa12e2320991d063", + "source_lfs_sha256": null, + "source_sha256": "da21319a1eab41e0425f2af2452a266981af91ca23ef26807dcff8ee0754cfbd", + "source_bytes": 8212777, + "public_sha256": "da21319a1eab41e0425f2af2452a266981af91ca23ef26807dcff8ee0754cfbd", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "696af75ed94d8873b0113b1687208cf07467f0b2", + "source_lfs_sha256": null, + "source_sha256": "3bf4506bf70a540b550c56c267a4aa131e34659fb47761917002bbb898b400d5", + "source_bytes": 2684896, + "public_sha256": "3bf4506bf70a540b550c56c267a4aa131e34659fb47761917002bbb898b400d5", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-asset-trajectory-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "f041542f944d3940e3ce58c4ce381256477f345f", + "source_lfs_sha256": null, + "source_sha256": "f5ee277836db68bdbf009a0ee4c2812be302dc6c6c16e781516519627da61e35", + "source_bytes": 5676208, + "public_sha256": "f5ee277836db68bdbf009a0ee4c2812be302dc6c6c16e781516519627da61e35", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "d260f8665f85016edee4d89be603c50eba8fb793", + "source_lfs_sha256": null, + "source_sha256": "a3363644967fc274c73ca4586a2fe25db15d62f21337db60876e07bbb519aaf9", + "source_bytes": 1535, + "public_sha256": "a3363644967fc274c73ca4586a2fe25db15d62f21337db60876e07bbb519aaf9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "598cadb54648fbf5ccc07b1d3c04e84356655148", + "source_lfs_sha256": null, + "source_sha256": "6e653f604b698db8a38c97b5842e79afe6bd3e90af82a291b442916c8dade997", + "source_bytes": 1419, + "public_sha256": "6e653f604b698db8a38c97b5842e79afe6bd3e90af82a291b442916c8dade997", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "24f0c09f2d26558b61d8fd78e0aeeff37059316f", + "source_lfs_sha256": null, + "source_sha256": "a8af5d5b92de6dcf3f0783c5a80901b47b48ee1753d33c8b6b1731ab7ea92d2f", + "source_bytes": 241725, + "public_sha256": "a8af5d5b92de6dcf3f0783c5a80901b47b48ee1753d33c8b6b1731ab7ea92d2f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "f201e83536bb09b5615911de6930b09dff3f064f", + "source_lfs_sha256": null, + "source_sha256": "235062838cce040a692d93449e1acf9f3296cba48cf6fee8bb58685d6d68f8f6", + "source_bytes": 418, + "public_sha256": "235062838cce040a692d93449e1acf9f3296cba48cf6fee8bb58685d6d68f8f6", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "a922d715713d1cf15080cf4dcd40b05767089ce3", + "source_lfs_sha256": null, + "source_sha256": "bdcd1a3fc7d0ba725aa6a8857339586b4d7430dbe185f8a923cb00fd4bcca1f9", + "source_bytes": 594, + "public_sha256": "bdcd1a3fc7d0ba725aa6a8857339586b4d7430dbe185f8a923cb00fd4bcca1f9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "9f832ea63d3705cf0544909c09e9143d7b9dd4aa", + "source_lfs_sha256": null, + "source_sha256": "e25d58dd76f4f2b2e0fc274696b751f3445c2a809a56cbd16fc9283df090f85f", + "source_bytes": 1523, + "public_sha256": "e25d58dd76f4f2b2e0fc274696b751f3445c2a809a56cbd16fc9283df090f85f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "c92db6bb45616a356e6dcf95d63c37f0bbf8804e", + "source_lfs_sha256": null, + "source_sha256": "1ff982fae2cd242c7ab3c10b1d83f7fa386265ec523a5f4b46f9a8ef054c98bb", + "source_bytes": 2654, + "public_sha256": "1ff982fae2cd242c7ab3c10b1d83f7fa386265ec523a5f4b46f9a8ef054c98bb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "cf95bedd9580b33233ea8106b85161ff93273067", + "source_lfs_sha256": null, + "source_sha256": "4448ec7f3c7f77f20fd3177cfb0df3f91e21937603501d1a3145d0f35b81d6bb", + "source_bytes": 2412, + "public_sha256": "4448ec7f3c7f77f20fd3177cfb0df3f91e21937603501d1a3145d0f35b81d6bb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "4d1edc3be5b163766b2a5de45e11f66056e16052", + "source_lfs_sha256": null, + "source_sha256": "233e60bb3ab80e284103ed174e5757af4db741d14887f4ec21bcb4a7ec973c13", + "source_bytes": 1812, + "public_sha256": "233e60bb3ab80e284103ed174e5757af4db741d14887f4ec21bcb4a7ec973c13", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "a3a3d6ba2ad9b353e936faf56f09f78a6b208089", + "source_lfs_sha256": null, + "source_sha256": "773c9d501c215ce22a0e9048b574773803268b663a89905e2be0cfbe68693b13", + "source_bytes": 3171, + "public_sha256": "773c9d501c215ce22a0e9048b574773803268b663a89905e2be0cfbe68693b13", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "60e39696d5a991fa4a2386c724719613640acd6c", + "source_lfs_sha256": null, + "source_sha256": "dea1a41879ad7ffa7e7fa58d1f1b00fda9aefc3cf9ca668581dc38e5850f172b", + "source_bytes": 2907, + "public_sha256": "dea1a41879ad7ffa7e7fa58d1f1b00fda9aefc3cf9ca668581dc38e5850f172b", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "c6354f9b6c4f2e76db0e35b956dd74b888834092", + "source_lfs_sha256": null, + "source_sha256": "25126784adf5b12da904d5a7b678ef451eddbea1cf3dcaed6b9e1e20b52a251f", + "source_bytes": 1175714, + "public_sha256": "25126784adf5b12da904d5a7b678ef451eddbea1cf3dcaed6b9e1e20b52a251f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "0606903f3c93a52bd4fa7ea79a9d65341e6d0748", + "source_lfs_sha256": null, + "source_sha256": "8a111cf8952bb62d7f61e6f99401f7a143fc08389c04f3608f185705d846b4ee", + "source_bytes": 364851, + "public_sha256": "8a111cf8952bb62d7f61e6f99401f7a143fc08389c04f3608f185705d846b4ee", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-micron-dual-partition-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "7d338e11c6f3414f893d3f67344de2eb09a7222d", + "source_lfs_sha256": null, + "source_sha256": "eea0550544e6fd468b84e24715c7842b4e1beaa1ccc664a67faabaef6ed988ff", + "source_bytes": 857094, + "public_sha256": "eea0550544e6fd468b84e24715c7842b4e1beaa1ccc664a67faabaef6ed988ff", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "5f254b6231b1972044b8f86cc16ead7f433aa5ce", + "source_lfs_sha256": null, + "source_sha256": "70d07ec6a92205e6cb39d24a99f3b12d6c0fa2dbff8064bc6a2e23c0c250cc7f", + "source_bytes": 1425, + "public_sha256": "70d07ec6a92205e6cb39d24a99f3b12d6c0fa2dbff8064bc6a2e23c0c250cc7f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "5f254b6231b1972044b8f86cc16ead7f433aa5ce", + "source_lfs_sha256": null, + "source_sha256": "70d07ec6a92205e6cb39d24a99f3b12d6c0fa2dbff8064bc6a2e23c0c250cc7f", + "source_bytes": 1425, + "public_sha256": "70d07ec6a92205e6cb39d24a99f3b12d6c0fa2dbff8064bc6a2e23c0c250cc7f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "40e252c71c4328d2975ee6a369d218f281b71621", + "source_lfs_sha256": null, + "source_sha256": "8a1eb415bdfbf4d46faf63558d4a147686c432aa50c9a5dbf712a70c4b078fdb", + "source_bytes": 471630, + "public_sha256": "8a1eb415bdfbf4d46faf63558d4a147686c432aa50c9a5dbf712a70c4b078fdb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "89ed484d82c7a26e7e52960fc4e31fe2a5cbefb8", + "source_lfs_sha256": null, + "source_sha256": "62b3642937badec991b1b3c1664b74c2ff590207c645d3e78872666ed9664a35", + "source_bytes": 419, + "public_sha256": "62b3642937badec991b1b3c1664b74c2ff590207c645d3e78872666ed9664a35", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "deb7b442fc9e89b5a9512f723c65f140ec7a2ea7", + "source_lfs_sha256": null, + "source_sha256": "258635de8cdc9a3c2415a2f96e76aafdf84d15b9a73fd142b030b42bc2a54036", + "source_bytes": 595, + "public_sha256": "258635de8cdc9a3c2415a2f96e76aafdf84d15b9a73fd142b030b42bc2a54036", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "6fe6f73b8c70742928eb286c0e5f37a85cf64ae7", + "source_lfs_sha256": null, + "source_sha256": "b97cc6962489b966e32e0147d588331b5173604b0cdc116b756832a5b38db5de", + "source_bytes": 1529, + "public_sha256": "b97cc6962489b966e32e0147d588331b5173604b0cdc116b756832a5b38db5de", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "d9d3f9a68f43e1db80810c77dbbc03bc7d48b1f2", + "source_lfs_sha256": null, + "source_sha256": "39e48b0835f6ef536728eb538d817e2ac555c95e8cb9e52095d1b142a8e2aba2", + "source_bytes": 2424, + "public_sha256": "39e48b0835f6ef536728eb538d817e2ac555c95e8cb9e52095d1b142a8e2aba2", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "d9d3f9a68f43e1db80810c77dbbc03bc7d48b1f2", + "source_lfs_sha256": null, + "source_sha256": "39e48b0835f6ef536728eb538d817e2ac555c95e8cb9e52095d1b142a8e2aba2", + "source_bytes": 2424, + "public_sha256": "39e48b0835f6ef536728eb538d817e2ac555c95e8cb9e52095d1b142a8e2aba2", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "5af1af29dae793010304c3c8792a153c0f3a84ab", + "source_lfs_sha256": null, + "source_sha256": "fa35f3295d5af9df7caddc34d96507c4bf13182546b7c6632711d5c789a3b1ae", + "source_bytes": 1818, + "public_sha256": "fa35f3295d5af9df7caddc34d96507c4bf13182546b7c6632711d5c789a3b1ae", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "c1bf6f309428116acd41c3f5e0e8c69a9d220975", + "source_lfs_sha256": null, + "source_sha256": "ee1eeeda632b5d6cfe382b60d4c56c58e43468f702c717eba3cf10022bb59509", + "source_bytes": 2898, + "public_sha256": "ee1eeeda632b5d6cfe382b60d4c56c58e43468f702c717eba3cf10022bb59509", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "c1bf6f309428116acd41c3f5e0e8c69a9d220975", + "source_lfs_sha256": null, + "source_sha256": "ee1eeeda632b5d6cfe382b60d4c56c58e43468f702c717eba3cf10022bb59509", + "source_bytes": 2898, + "public_sha256": "ee1eeeda632b5d6cfe382b60d4c56c58e43468f702c717eba3cf10022bb59509", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "0ff418b9a06080c88d284f3df26650aaab780103", + "source_lfs_sha256": null, + "source_sha256": "44a3144d4a8b6d775a588476e4468914f2fde55b7a27a4d779d0507354d9f5d4", + "source_bytes": 1689898, + "public_sha256": "44a3144d4a8b6d775a588476e4468914f2fde55b7a27a4d779d0507354d9f5d4", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "b839fb5e863c18d94f9947a11e0660896a443d88", + "source_lfs_sha256": null, + "source_sha256": "f4e383a72e46d6d8f8c9d273857b3af61032b1413bc89d44ce44d6e35cbdf915", + "source_bytes": 220123, + "public_sha256": "f4e383a72e46d6d8f8c9d273857b3af61032b1413bc89d44ce44d6e35cbdf915", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-mix-crossover-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "2a1b065a88038c03974befba06f8a6b07cac0529", + "source_lfs_sha256": null, + "source_sha256": "6bbec8b9c46621e0d201ccc02cf8ed53c63fb4ed38af918972ff89fa228d9e0a", + "source_bytes": 1505538, + "public_sha256": "6bbec8b9c46621e0d201ccc02cf8ed53c63fb4ed38af918972ff89fa228d9e0a", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "ba9b283c78cb55117df739b110f6bf49124c5aff", + "source_lfs_sha256": null, + "source_sha256": "68b0b1a4f234253837ab26be4861559f70bd8ce4a0c3e9ab50e27dc2e433269f", + "source_bytes": 1526, + "public_sha256": "68b0b1a4f234253837ab26be4861559f70bd8ce4a0c3e9ab50e27dc2e433269f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "0c0c7f38ca8f9788804599ff0293a5ad869edc30", + "source_lfs_sha256": null, + "source_sha256": "49c3aab97123ef0d7cdbcb54a14e260cc9976b7c2136374c47810a72ec7a8dc9", + "source_bytes": 1410, + "public_sha256": "49c3aab97123ef0d7cdbcb54a14e260cc9976b7c2136374c47810a72ec7a8dc9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "fb93374d9b40c3dd1d3f4f521b934b5e2627ef05", + "source_lfs_sha256": null, + "source_sha256": "1c32cb4ad7359f682df92549aeedd917a9bbf7e83c76b05c255c1f2204eeaf65", + "source_bytes": 1756015, + "public_sha256": "1c32cb4ad7359f682df92549aeedd917a9bbf7e83c76b05c255c1f2204eeaf65", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "929a9f7c8836d9187add984ec2cf55199c0d9942", + "source_lfs_sha256": null, + "source_sha256": "d7b4ebefd4f557ac10eb664d67f9cce6ce725151fd888a4faf2097399549b631", + "source_bytes": 684, + "public_sha256": "d7b4ebefd4f557ac10eb664d67f9cce6ce725151fd888a4faf2097399549b631", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "d35c3861cb8a9ac7f86ec72ab402c283d6cabb77", + "source_lfs_sha256": null, + "source_sha256": "b8a6449fb24a81f7ab6b733ad95f9694fc2207e3b9a2cc3b158055e935e5e057", + "source_bytes": 908, + "public_sha256": "b8a6449fb24a81f7ab6b733ad95f9694fc2207e3b9a2cc3b158055e935e5e057", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "a0e3aa956ab89252ec188e41589d0ffaa9d2e573", + "source_lfs_sha256": null, + "source_sha256": "c0a396461ddc9a8b2ac3fd02cf98197478032c448da7c26464158488a65f165a", + "source_bytes": 1513, + "public_sha256": "c0a396461ddc9a8b2ac3fd02cf98197478032c448da7c26464158488a65f165a", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/promoted_128k.jsonl", + "source_git_blob": "c9db79fd01f383f78cf8d573023740c3da80a0d5", + "source_lfs_sha256": null, + "source_sha256": "6adf6d2f19461298f78b30f52c598f8252881562ec4b967a697ef53dc9d2b112", + "source_bytes": 3884649, + "public_sha256": "6adf6d2f19461298f78b30f52c598f8252881562ec4b967a697ef53dc9d2b112", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/promoted_16k.jsonl", + "source_git_blob": "872435ef88339438ea434a16e7f8f5663e906d0c", + "source_lfs_sha256": null, + "source_sha256": "7db446f01fbc33e405478b9f119990557c52ae4c1c89ffc8845c27f83e2624ec", + "source_bytes": 503975, + "public_sha256": "7db446f01fbc33e405478b9f119990557c52ae4c1c89ffc8845c27f83e2624ec", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/promoted_32k.jsonl", + "source_git_blob": "87610e9c06555200d0ba0ad994eb29d487f95582", + "source_lfs_sha256": null, + "source_sha256": "a0a7a681adc011f49e9735c27ab349178caa24034edcbbb8aca6d9065865efd0", + "source_bytes": 916528, + "public_sha256": "a0a7a681adc011f49e9735c27ab349178caa24034edcbbb8aca6d9065865efd0", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/promoted_64k.jsonl", + "source_git_blob": "962aadf278dd6410563a6b9b9230470c994e1473", + "source_lfs_sha256": null, + "source_sha256": "cee693639fb4ddccd7c24b934a32c36738f55263a06f8fb31dffa0193c252d30", + "source_bytes": 1706670, + "public_sha256": "cee693639fb4ddccd7c24b934a32c36738f55263a06f8fb31dffa0193c252d30", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "68e6c588ed78f5922c28c2b035e2a7a5d2597e4e", + "source_lfs_sha256": null, + "source_sha256": "6aafa7c293e119e94082b17bc375455be1c17a69b42527ebfd02bae287af25d5", + "source_bytes": 2708, + "public_sha256": "6aafa7c293e119e94082b17bc375455be1c17a69b42527ebfd02bae287af25d5", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "f451d6bd008ff297a8869da45b27adbbfc89f01d", + "source_lfs_sha256": null, + "source_sha256": "9915298f1e54aff6559f6fb29a9c6bc2d917c84eee66099d6413342d688afb9f", + "source_bytes": 2454, + "public_sha256": "9915298f1e54aff6559f6fb29a9c6bc2d917c84eee66099d6413342d688afb9f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "592d56c2d9677848b00843ddb5a20b9379943a70", + "source_lfs_sha256": null, + "source_sha256": "06d760bb9274f5e9ebd0b0b61782a82e2ce3f06da93926734422590830db96c1", + "source_bytes": 1802, + "public_sha256": "06d760bb9274f5e9ebd0b0b61782a82e2ce3f06da93926734422590830db96c1", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "ceab8992e77fa35f3d254a366b4c740294f72486", + "source_lfs_sha256": null, + "source_sha256": "2fef0efefe4e584027ba276f9cdb755c9c661656783d5328a33b378c105f049a", + "source_bytes": 6283, + "public_sha256": "2fef0efefe4e584027ba276f9cdb755c9c661656783d5328a33b378c105f049a", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "5fd62579b18cf1e28fddc6a765ceb3d0266d75d1", + "source_lfs_sha256": null, + "source_sha256": "2a46a60ef3bcc781590bc1e29f475fdef44955d72b508dccab5c27b44f761f11", + "source_bytes": 5911, + "public_sha256": "2a46a60ef3bcc781590bc1e29f475fdef44955d72b508dccab5c27b44f761f11", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "3c500112278876166a7bdb447e74c45722d49b54", + "source_lfs_sha256": null, + "source_sha256": "7be565f53103c995a9a30163d51414f8ceaa6129e9c8708ad6a4edb3379f2141", + "source_bytes": 7011822, + "public_sha256": "7be565f53103c995a9a30163d51414f8ceaa6129e9c8708ad6a4edb3379f2141", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "c13e2c94ab55205aaaf2cad6fccc6da46a8f2dfb", + "source_lfs_sha256": null, + "source_sha256": "600f6372d795c10bf4984f9c21b2adee23c71fc7ffcbae06266cb38764503455", + "source_bytes": 1529882, + "public_sha256": "600f6372d795c10bf4984f9c21b2adee23c71fc7ffcbae06266cb38764503455", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-finance-nvidia-market-segment-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "afe06dc4652dfd1c96d25af87e5ed1fb3a4e6b9e", + "source_lfs_sha256": null, + "source_sha256": "fad0c0812333b8447b8378eb7b09436bf35789c49d827d472838fb5f5946e277", + "source_bytes": 5601327, + "public_sha256": "fad0c0812333b8447b8378eb7b09436bf35789c49d827d472838fb5f5946e277", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "495f09d7cbdf311649898ce77eec0ebb5d9c56c5", + "source_lfs_sha256": null, + "source_sha256": "702660c53ee160bf4e3e5eba9ecd360aa46581ef86170f7188c7ab5924faaf06", + "source_bytes": 1528, + "public_sha256": "702660c53ee160bf4e3e5eba9ecd360aa46581ef86170f7188c7ab5924faaf06", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "fb7a8d9b5421feea14b16ccbc4a9a540002ee17b", + "source_lfs_sha256": null, + "source_sha256": "62f6dac922badfa4eb7f4c64cbe4a04b6e6d0a58a33b4ff0388f2a3c004c4107", + "source_bytes": 1412, + "public_sha256": "62f6dac922badfa4eb7f4c64cbe4a04b6e6d0a58a33b4ff0388f2a3c004c4107", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "f2b6c4600fd932056476c1f32bf2fd5d1b54fd1c", + "source_lfs_sha256": null, + "source_sha256": "b622fde74fbb884b42b94176a3edcb314adacc34f2181abcf3b3e95f48186792", + "source_bytes": 429687, + "public_sha256": "b622fde74fbb884b42b94176a3edcb314adacc34f2181abcf3b3e95f48186792", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "4969ac397c90d3298299ef8f902d1c8206c535ec", + "source_lfs_sha256": null, + "source_sha256": "974bab68f92e53fb11ee94b884fc273411687798d0e9c13693a73c87a6daeaeb", + "source_bytes": 418, + "public_sha256": "974bab68f92e53fb11ee94b884fc273411687798d0e9c13693a73c87a6daeaeb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "c05a1dec564b322312a0a63cc28028d49056c5dc", + "source_lfs_sha256": null, + "source_sha256": "c25ce79e21cb575f653eeca880692d1bc1c1cf323925813d92bde55726f18c11", + "source_bytes": 594, + "public_sha256": "c25ce79e21cb575f653eeca880692d1bc1c1cf323925813d92bde55726f18c11", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "93a0550336dbcb385888879d381969e1eb563870", + "source_lfs_sha256": null, + "source_sha256": "fdc38544298d40f24212cbc9622508d0f5951146aa6d03861dd30d25780c7dbf", + "source_bytes": 1512, + "public_sha256": "fdc38544298d40f24212cbc9622508d0f5951146aa6d03861dd30d25780c7dbf", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "389e74686978fbe824084a443747e63c4ce11766", + "source_lfs_sha256": null, + "source_sha256": "134b9d8a408f424b0376a4b9dd6c2974e1e2e4ef560d8616a35454e0920e883d", + "source_bytes": 2622, + "public_sha256": "134b9d8a408f424b0376a4b9dd6c2974e1e2e4ef560d8616a35454e0920e883d", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "eaeb82edb668fd92fd4442c2be3ea2f9c5a1f76b", + "source_lfs_sha256": null, + "source_sha256": "f38c610c6c53b63be17b73dfc060fb954e8fa22db37f295494e58147c1b26ef7", + "source_bytes": 2380, + "public_sha256": "f38c610c6c53b63be17b73dfc060fb954e8fa22db37f295494e58147c1b26ef7", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "b2f7de3094dc27e8e26305ebea3ff9984da35869", + "source_lfs_sha256": null, + "source_sha256": "726964512b7528d0b21d810cf2b9fdf6f1093b16f086b6d6da44be8dd0c4c8a7", + "source_bytes": 1801, + "public_sha256": "726964512b7528d0b21d810cf2b9fdf6f1093b16f086b6d6da44be8dd0c4c8a7", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "0aba30f62c2406dbd35df2bc93a5605e974f954c", + "source_lfs_sha256": null, + "source_sha256": "dfcf48f559922a99ab9846dd09c1bc774bc427aaeb0654da0d5c84325e1cada9", + "source_bytes": 3065, + "public_sha256": "dfcf48f559922a99ab9846dd09c1bc774bc427aaeb0654da0d5c84325e1cada9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "6304b680dfed707b75196e121dddaf38a5d64639", + "source_lfs_sha256": null, + "source_sha256": "779fa0359115d18431a3bdedfb327eb22faac03c4a5ffbe48cc3ca7e0fd6f0c1", + "source_bytes": 2801, + "public_sha256": "779fa0359115d18431a3bdedfb327eb22faac03c4a5ffbe48cc3ca7e0fd6f0c1", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "7793fcf6077790672eb25170ed69afc273fa3a9b", + "source_lfs_sha256": null, + "source_sha256": "92fe60212139142bb0e36985a015480174c88aa9cf58a43662382a1281b33b75", + "source_bytes": 2439373, + "public_sha256": "92fe60212139142bb0e36985a015480174c88aa9cf58a43662382a1281b33b75", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "7723ddd8228046ef00bfb334c28f4af9349bc7d2", + "source_lfs_sha256": null, + "source_sha256": "f5d16b8238db54b1587f70a76d55188e7b7f67dce108c1bad502460b090ff019", + "source_bytes": 78005, + "public_sha256": "f5d16b8238db54b1587f70a76d55188e7b7f67dce108c1bad502460b090ff019", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-acme-issuance-32k-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "8df01e6eac3a8481fc4a4a1184792e32c262af86", + "source_lfs_sha256": null, + "source_sha256": "93cca4386104fc4f47236bd1cf30e424319fc6385dbae1e35e06cd2a2fc96cdb", + "source_bytes": 2371111, + "public_sha256": "93cca4386104fc4f47236bd1cf30e424319fc6385dbae1e35e06cd2a2fc96cdb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "ae9e2718977c05f2094db2b8f2eef7e7e1b3d2aa", + "source_lfs_sha256": null, + "source_sha256": "7965154d01a18238ebc4ffaefa555c84d8f4ad62504db294ed9902d1ff815401", + "source_bytes": 1523, + "public_sha256": "7965154d01a18238ebc4ffaefa555c84d8f4ad62504db294ed9902d1ff815401", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "ea5183be430ee73b6c7741db6b1319d38d960a39", + "source_lfs_sha256": null, + "source_sha256": "8854980822fa21d935c79a58b96e3fd5c0da62c2db7e59d72c3e1eeabd57bf11", + "source_bytes": 1407, + "public_sha256": "8854980822fa21d935c79a58b96e3fd5c0da62c2db7e59d72c3e1eeabd57bf11", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "0aa4cc9dfb4c74ff331386ee402e05ae022281b8", + "source_lfs_sha256": null, + "source_sha256": "0ba298c5e915137f7be74dd3a1acaae954a23e7d22ab568b1680a3fa03eb61be", + "source_bytes": 799146, + "public_sha256": "0ba298c5e915137f7be74dd3a1acaae954a23e7d22ab568b1680a3fa03eb61be", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "af20490d600c2d00b6d90f4a276ab2eed1244ad9", + "source_lfs_sha256": null, + "source_sha256": "8565680039434baae7a40377816842f35fd565b7872941c698ad9dc72384a8e9", + "source_bytes": 419, + "public_sha256": "8565680039434baae7a40377816842f35fd565b7872941c698ad9dc72384a8e9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "439ef8b3492e19876a2f8b812eede894fe004f82", + "source_lfs_sha256": null, + "source_sha256": "bb7c16ecbe87d6336547eda0f0c4318f7c7fa880d481553a5f3b95a81dcc2090", + "source_bytes": 595, + "public_sha256": "bb7c16ecbe87d6336547eda0f0c4318f7c7fa880d481553a5f3b95a81dcc2090", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "73506e94911d308c1bcf70addc110df8f6ec2a07", + "source_lfs_sha256": null, + "source_sha256": "b7c76b79fb4abda52d248fa33edb794d3a0acadd720e9e510a407d509b8d33a7", + "source_bytes": 1505, + "public_sha256": "b7c76b79fb4abda52d248fa33edb794d3a0acadd720e9e510a407d509b8d33a7", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "b11004918348cab081dc2de37758b12359f9a730", + "source_lfs_sha256": null, + "source_sha256": "58356ee4e544143494ca4254ab259b39dc80b007fa53ccc8aa5b78320826a566", + "source_bytes": 2605, + "public_sha256": "58356ee4e544143494ca4254ab259b39dc80b007fa53ccc8aa5b78320826a566", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "b4a69e98ef61d56481ddcadca8d3757ffb5e2fcd", + "source_lfs_sha256": null, + "source_sha256": "edc236b383062a857117c047ce60d83fe1085d371ecc1d3f27fd2be870aaf42b", + "source_bytes": 2363, + "public_sha256": "edc236b383062a857117c047ce60d83fe1085d371ecc1d3f27fd2be870aaf42b", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "6af92979bd032511e57f35fa87b5ba2a1c1fcbc3", + "source_lfs_sha256": null, + "source_sha256": "d1b544400e6639a4f40510616de1724cd634689e5915bb0012466543788588af", + "source_bytes": 1794, + "public_sha256": "d1b544400e6639a4f40510616de1724cd634689e5915bb0012466543788588af", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "f8643a7338d09992569bd9734a2a40b0175090d0", + "source_lfs_sha256": null, + "source_sha256": "1e5271781791b451ad2ae801bc56b41275bb6a1c4e8bad0a1f5318b65301530d", + "source_bytes": 3023, + "public_sha256": "1e5271781791b451ad2ae801bc56b41275bb6a1c4e8bad0a1f5318b65301530d", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "02579e663482a6b3e646f6adafbc277ce7480a4e", + "source_lfs_sha256": null, + "source_sha256": "2275372f15712eb711b00f08904ebe42c1a65ac6b8e9cdb3f854f0bf70416bc2", + "source_bytes": 2759, + "public_sha256": "2275372f15712eb711b00f08904ebe42c1a65ac6b8e9cdb3f854f0bf70416bc2", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "8dda48fb896f8edcbc41804afa1210f8d603d73b", + "source_lfs_sha256": null, + "source_sha256": "46609c8b1c622aa486e96e86feeb620304e35897ba1c1800afc575bd400150eb", + "source_bytes": 4019539, + "public_sha256": "46609c8b1c622aa486e96e86feeb620304e35897ba1c1800afc575bd400150eb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "74f8b430d52f61ce05127389887d60d9855b1363", + "source_lfs_sha256": null, + "source_sha256": "919b55209adae3d6f44f1c1ebc849c643bb99ab266fa37752848e53e3214fecc", + "source_bytes": 88906, + "public_sha256": "919b55209adae3d6f44f1c1ebc849c643bb99ab266fa37752848e53e3214fecc", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-dnssec-64k-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "6cbf4310789d0309407eb388566a4130e2c32b8b", + "source_lfs_sha256": null, + "source_sha256": "916a2b96d37accf5eae0785343498bed90fa2eebc4a60a3d5038ea593f85379a", + "source_bytes": 3942652, + "public_sha256": "916a2b96d37accf5eae0785343498bed90fa2eebc4a60a3d5038ea593f85379a", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "7674f48b7a84273f0b6c202cd45dc7f1ee77f2b9", + "source_lfs_sha256": null, + "source_sha256": "7b6b5dcfdb28230435c16bf654abee1becc8526f1b0b9e1c866a2b7e3da104a1", + "source_bytes": 1550, + "public_sha256": "7b6b5dcfdb28230435c16bf654abee1becc8526f1b0b9e1c866a2b7e3da104a1", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "4b6293fa362de43deee99012db09ed6352aa588f", + "source_lfs_sha256": null, + "source_sha256": "8be3e9082e9b4d03f6372efb07c0c7e1e53b1656fe8aecedc9f9bb80453084e1", + "source_bytes": 1434, + "public_sha256": "8be3e9082e9b4d03f6372efb07c0c7e1e53b1656fe8aecedc9f9bb80453084e1", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "0613cf1b29031a5f3b46e1b5b80fc2f3313193ef", + "source_lfs_sha256": null, + "source_sha256": "b00affe8e4aa3e73ec7fde82e3952c9b7e3a702910250324b453b7c1d8304c47", + "source_bytes": 1578962, + "public_sha256": "b00affe8e4aa3e73ec7fde82e3952c9b7e3a702910250324b453b7c1d8304c47", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "18714c64d73dd7e5cfb9e056aa3dcab812da7445", + "source_lfs_sha256": null, + "source_sha256": "549a9ca8da0850ca9b3092901ad1ce8f9d8c1aa88c6b1b8c1396c4fba54d60ce", + "source_bytes": 423, + "public_sha256": "549a9ca8da0850ca9b3092901ad1ce8f9d8c1aa88c6b1b8c1396c4fba54d60ce", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "0187bc5e56ec3d35e00c4bbb4c10ab1c2fe06103", + "source_lfs_sha256": null, + "source_sha256": "a59519496a0e982025a3f1851fddc562f83d218cdc6aed920a60886a015b9149", + "source_bytes": 599, + "public_sha256": "a59519496a0e982025a3f1851fddc562f83d218cdc6aed920a60886a015b9149", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "128359f5fecf63100ec47eeaa6d34fb29a339da7", + "source_lfs_sha256": null, + "source_sha256": "2213f505de4dadf0f32853df9b82d5bca7ce91a07174ed35fad381f6b638eef9", + "source_bytes": 1525, + "public_sha256": "2213f505de4dadf0f32853df9b82d5bca7ce91a07174ed35fad381f6b638eef9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "6571d4fe451a9e0fd1ea7f09e0e713b7ae90c0e6", + "source_lfs_sha256": null, + "source_sha256": "8bcf9310868e0c4cc714a6d4f72ddec9476a79c5fd617a7ae7325ab3642983d5", + "source_bytes": 2645, + "public_sha256": "8bcf9310868e0c4cc714a6d4f72ddec9476a79c5fd617a7ae7325ab3642983d5", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "5c2f1f7eb64b6599633b755ac80ff047679c2d70", + "source_lfs_sha256": null, + "source_sha256": "590eea269a7be12e6499eb12f2ef5e7ca7255d895e14b82f8db879ffb2d491ad", + "source_bytes": 2403, + "public_sha256": "590eea269a7be12e6499eb12f2ef5e7ca7255d895e14b82f8db879ffb2d491ad", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "a0bfb7c3a00a048ec46a23dc39281c768329b217", + "source_lfs_sha256": null, + "source_sha256": "68ae36270707f3e78f2045d56633b20981d42393d7f35a829568b88ccf26ed06", + "source_bytes": 1813, + "public_sha256": "68ae36270707f3e78f2045d56633b20981d42393d7f35a829568b88ccf26ed06", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "b1537f742d1e000a304f889affa13b0013012c02", + "source_lfs_sha256": null, + "source_sha256": "d622ac51eb96e79e6816c6339c6cf15c96e49f9bfd7fb102b51450fbf957a78c", + "source_bytes": 3082, + "public_sha256": "d622ac51eb96e79e6816c6339c6cf15c96e49f9bfd7fb102b51450fbf957a78c", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "762c00c024c10d1e098b8b32399a114deb1ddae4", + "source_lfs_sha256": null, + "source_sha256": "f2b1ac1b5fb076194d74d1aaf23d761e81330d62c9c06e3c1dda321ad11a8037", + "source_bytes": 2818, + "public_sha256": "f2b1ac1b5fb076194d74d1aaf23d761e81330d62c9c06e3c1dda321ad11a8037", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "14711d7475f1fa444d89498a1303765e4c2fdbe1", + "source_lfs_sha256": null, + "source_sha256": "87a9efd64ad3eb07ee94bab67fa7a19229b1a4d2c101d109976d1abda23aed7a", + "source_bytes": 9172590, + "public_sha256": "87a9efd64ad3eb07ee94bab67fa7a19229b1a4d2c101d109976d1abda23aed7a", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "095000eac1cd16ee89d9a654d4b850fd0778fbf6", + "source_lfs_sha256": null, + "source_sha256": "4821146351961b49f48644c2e64cde762d492cfaea9b0ef7b08eeeff5d0fc098", + "source_bytes": 113112, + "public_sha256": "4821146351961b49f48644c2e64cde762d492cfaea9b0ef7b08eeeff5d0fc098", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http-semantics-succession-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "acb08b8597803236de5f8346e9c8442893b2e99c", + "source_lfs_sha256": null, + "source_sha256": "30562c5e9b47b43a96b31ea8c494b6a92121468c5d3bc364fdf149cb980a0748", + "source_bytes": 9076547, + "public_sha256": "30562c5e9b47b43a96b31ea8c494b6a92121468c5d3bc364fdf149cb980a0748", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "7fa76091a59840d721c7885889b65f8b213176e5", + "source_lfs_sha256": null, + "source_sha256": "cf0bde02e4ef72a80afc6f4eef04a66d7fd12d960f0f87b283dfeddcf6f73512", + "source_bytes": 1532, + "public_sha256": "cf0bde02e4ef72a80afc6f4eef04a66d7fd12d960f0f87b283dfeddcf6f73512", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "64744c0e71dc9ea54c4e48c35fdc02c2884562bc", + "source_lfs_sha256": null, + "source_sha256": "8b2728e2440e4d821de539aa3ce54a6fdc8c30a1e10714366306db562d0e3bff", + "source_bytes": 1416, + "public_sha256": "8b2728e2440e4d821de539aa3ce54a6fdc8c30a1e10714366306db562d0e3bff", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "9b6a70290686c15d4d00f9fd2e496560e9b9a184", + "source_lfs_sha256": null, + "source_sha256": "06b330066621117a122fb749365638a3bb972719b9bba00ce66381e4c42f6ea3", + "source_bytes": 782197, + "public_sha256": "06b330066621117a122fb749365638a3bb972719b9bba00ce66381e4c42f6ea3", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "1f976d62f1eefecc4ce2976f6e8d9fcec1122632", + "source_lfs_sha256": null, + "source_sha256": "92c1ed4f418ff079ba80cd039516a637c5266e66e8d5015cde0b2cac9e5ee1e3", + "source_bytes": 419, + "public_sha256": "92c1ed4f418ff079ba80cd039516a637c5266e66e8d5015cde0b2cac9e5ee1e3", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "e8894feb41884001f9313f88f7eb2fbb77045c8a", + "source_lfs_sha256": null, + "source_sha256": "f26a78fb8b45339053aaa84b9c2243ddc272efcc46e57e630340ee9c2d2bf79e", + "source_bytes": 595, + "public_sha256": "f26a78fb8b45339053aaa84b9c2243ddc272efcc46e57e630340ee9c2d2bf79e", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "ba86125dca5ce6655b1a9610bac6ef609374124d", + "source_lfs_sha256": null, + "source_sha256": "53909980f0d2c09d5305d2e48f7664932954ec8c330f0daba875f457602cb980", + "source_bytes": 1504, + "public_sha256": "53909980f0d2c09d5305d2e48f7664932954ec8c330f0daba875f457602cb980", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "9be3aaf40e79174814da6c757c74690fa57fbdbc", + "source_lfs_sha256": null, + "source_sha256": "0ee7a6f6357e5b59913569caf3120989fdacce55355d90ef12fbd34e21ee060f", + "source_bytes": 2598, + "public_sha256": "0ee7a6f6357e5b59913569caf3120989fdacce55355d90ef12fbd34e21ee060f", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "dbe27336fd46387bc2816bff9b8c0cc088bfa011", + "source_lfs_sha256": null, + "source_sha256": "39009869ab3575828d0ddf9fdd6ca8413c430899535de4c6049f77a08c9032e1", + "source_bytes": 2356, + "public_sha256": "39009869ab3575828d0ddf9fdd6ca8413c430899535de4c6049f77a08c9032e1", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "c8a4952a617805cae0a30b4bc47d78edb7009f97", + "source_lfs_sha256": null, + "source_sha256": "36febbe842933d413885c05525c7ddbafc6e217f1e87623c8c811a1129f28680", + "source_bytes": 1793, + "public_sha256": "36febbe842933d413885c05525c7ddbafc6e217f1e87623c8c811a1129f28680", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "1465ed33108bbbf5d3dbd72c9fe413ce453d7166", + "source_lfs_sha256": null, + "source_sha256": "0ed355b0f93b4b00d7b7581bbbeb5af31bc28c76641cc6436279fb7bbf4ad9b5", + "source_bytes": 3017, + "public_sha256": "0ed355b0f93b4b00d7b7581bbbeb5af31bc28c76641cc6436279fb7bbf4ad9b5", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "66167b5e1f0969ef59c3166f58fcf405f854034e", + "source_lfs_sha256": null, + "source_sha256": "b2ae3f92acfa5c2569e54f027eb56d47fd9b555fc1f5a166401823477c3d2569", + "source_bytes": 2753, + "public_sha256": "b2ae3f92acfa5c2569e54f027eb56d47fd9b555fc1f5a166401823477c3d2569", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "c6df5dc05cbacc19d3ec071810e2e24306931595", + "source_lfs_sha256": null, + "source_sha256": "7fe36036faba5476741e9f0f95158101ce68c52a901aca0653f60f86f8713e13", + "source_bytes": 3744445, + "public_sha256": "7fe36036faba5476741e9f0f95158101ce68c52a901aca0653f60f86f8713e13", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "a7cc19024afd3c48da8f8e8c5fb44fc1ae051b18", + "source_lfs_sha256": null, + "source_sha256": "b249a8aff840b98837b6fe1ea3386412be3bf2c1c687fd6aded0ddcd459e4831", + "source_bytes": 69967, + "public_sha256": "b249a8aff840b98837b6fe1ea3386412be3bf2c1c687fd6aded0ddcd459e4831", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-http2-succession-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "d3960c8a9d5ddaf3f933757781d07a699c1da6d5", + "source_lfs_sha256": null, + "source_sha256": "3adec88d1d9cdb2636cc171d71b1a5c45eaa9b9ace8c270759eb870b868afdc2", + "source_bytes": 3684705, + "public_sha256": "3adec88d1d9cdb2636cc171d71b1a5c45eaa9b9ace8c270759eb870b868afdc2", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "cfad485818895262d51d25e5cdba2a44e0ef5bb6", + "source_lfs_sha256": null, + "source_sha256": "0852185fdaec41e525f3a949137967839c7a5dcec14d06ecbc1e0f7ae623d6ee", + "source_bytes": 1529, + "public_sha256": "0852185fdaec41e525f3a949137967839c7a5dcec14d06ecbc1e0f7ae623d6ee", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "ceec2958bd9f8342d223abdee88e8d28e14c14d4", + "source_lfs_sha256": null, + "source_sha256": "f94ab94c9a26ac954c6ab1efd78fc6a8b970fa66d680aff2febaf20e05652bc6", + "source_bytes": 1413, + "public_sha256": "f94ab94c9a26ac954c6ab1efd78fc6a8b970fa66d680aff2febaf20e05652bc6", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "a6072afd79386d3552d81e40f84907b489c4ffcc", + "source_lfs_sha256": null, + "source_sha256": "4ab2535e8a9169ca4cd64c1c4664742676a6b8bd2884f6705fd2a4d210d72ecb", + "source_bytes": 796757, + "public_sha256": "4ab2535e8a9169ca4cd64c1c4664742676a6b8bd2884f6705fd2a4d210d72ecb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "fefba4e7cb47da885f89fcdc88b9dbd6aef45ca6", + "source_lfs_sha256": null, + "source_sha256": "8dbf7d984a91b9ee4ece120754d2c732af3a82dfccc59d7e8afacd32e7b05395", + "source_bytes": 419, + "public_sha256": "8dbf7d984a91b9ee4ece120754d2c732af3a82dfccc59d7e8afacd32e7b05395", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "6636dabca1d7142e8a22b46df0dea04f593854e4", + "source_lfs_sha256": null, + "source_sha256": "fae017e7fb25e6af900f06ff73e4e739c81ed76536303f3caa3bf3c50c347f1a", + "source_bytes": 595, + "public_sha256": "fae017e7fb25e6af900f06ff73e4e739c81ed76536303f3caa3bf3c50c347f1a", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "c8b1369a21bc5f320c1c11c4a8c2d1bb5b1cb414", + "source_lfs_sha256": null, + "source_sha256": "db3e786070f57d5bb39ce31dc9657a60315f43eb3120393f13f2ce64843e5f03", + "source_bytes": 1508, + "public_sha256": "db3e786070f57d5bb39ce31dc9657a60315f43eb3120393f13f2ce64843e5f03", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "668165f3fc70fecc515d6ae2abed13e7e9aacbf2", + "source_lfs_sha256": null, + "source_sha256": "2faeb8ed1b0f7175b1110f8cfb63e0830806a8b0ed021e096acc719d2cc469e6", + "source_bytes": 2610, + "public_sha256": "2faeb8ed1b0f7175b1110f8cfb63e0830806a8b0ed021e096acc719d2cc469e6", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "c2d7d0c25576c2de44f914fb76b2dd4363c0a4e9", + "source_lfs_sha256": null, + "source_sha256": "a84ef81a6a48314ab3eaed9e41d312e9c56617fbe7a350cb5fdc77daaf6d7470", + "source_bytes": 2368, + "public_sha256": "a84ef81a6a48314ab3eaed9e41d312e9c56617fbe7a350cb5fdc77daaf6d7470", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "cc5efed797051f53783eea3234bb5d104e7e1ad2", + "source_lfs_sha256": null, + "source_sha256": "e0eaf12cf4bdd79a11c17143da5f1c8379dd106ba8c62ca19c9c645d1b1cc8ff", + "source_bytes": 1797, + "public_sha256": "e0eaf12cf4bdd79a11c17143da5f1c8379dd106ba8c62ca19c9c645d1b1cc8ff", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "b2d34b865022c3b4aa95e87dab992da85e659a80", + "source_lfs_sha256": null, + "source_sha256": "7762086636f527c262b02261587d6f587521cc1ce53d9876a359fadf0982bfd5", + "source_bytes": 3041, + "public_sha256": "7762086636f527c262b02261587d6f587521cc1ce53d9876a359fadf0982bfd5", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "708a240e7c9413dd79f6b66dd6bb1a1154de127c", + "source_lfs_sha256": null, + "source_sha256": "35423917d14de9acb1bc59f695d710d456bf9333b564110b467ea8fa9af87473", + "source_bytes": 2777, + "public_sha256": "35423917d14de9acb1bc59f695d710d456bf9333b564110b467ea8fa9af87473", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "a6db1448cd83199acb3a729048210f31a5b9df10", + "source_lfs_sha256": null, + "source_sha256": "6f1f56fa426336ce97a059d0193e6625fb365fea9c97225b1f7ea6e78805bc52", + "source_bytes": 5127504, + "public_sha256": "6f1f56fa426336ce97a059d0193e6625fb365fea9c97225b1f7ea6e78805bc52", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "0915c3b2badd26266ee6d1e2fdf5c93104e18c2f", + "source_lfs_sha256": null, + "source_sha256": "990ad8fc3b3f98abc254ea358be4b67d39bcf92d5cf86fd995ddcb8f0c3891fa", + "source_bytes": 76551, + "public_sha256": "990ad8fc3b3f98abc254ea358be4b67d39bcf92d5cf86fd995ddcb8f0c3891fa", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-pkix-path-succession-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "d620d07d956097d9d66e026147e92b2a1d9ee217", + "source_lfs_sha256": null, + "source_sha256": "845df800de6b2e80557b0fca8113d40cf0f570fbc2e98d3ab6da5ca754381112", + "source_bytes": 5063056, + "public_sha256": "845df800de6b2e80557b0fca8113d40cf0f570fbc2e98d3ab6da5ca754381112", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "37c49ff49e9a676137c1fcd826053ed468187834", + "source_lfs_sha256": null, + "source_sha256": "490321df856d98a584ef9e2018dc00f753cf378a9aaaa5559e6ec25c2d09e380", + "source_bytes": 1530, + "public_sha256": "490321df856d98a584ef9e2018dc00f753cf378a9aaaa5559e6ec25c2d09e380", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "96246ed1fff2f67f38736bdb4e248115ec630f76", + "source_lfs_sha256": null, + "source_sha256": "a98b3eb6666ab04abfb4b2e2c436148da2cd472cb6c55c6f3b6b458349a0c9a9", + "source_bytes": 1414, + "public_sha256": "a98b3eb6666ab04abfb4b2e2c436148da2cd472cb6c55c6f3b6b458349a0c9a9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "79889a0598ce5e11975027531c8f790c295ee42a", + "source_lfs_sha256": null, + "source_sha256": "3a1c842dc60cbec8ead820301682a0e402a1c667068679f159a9cc1ccb1a3862", + "source_bytes": 436269, + "public_sha256": "3a1c842dc60cbec8ead820301682a0e402a1c667068679f159a9cc1ccb1a3862", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "80c2d5a9559e9642cd52b6320cee891fd733d6b8", + "source_lfs_sha256": null, + "source_sha256": "306b0ae1b97e4a85dab84cd8969428fd460b89729ff861b65bbd102df096e225", + "source_bytes": 418, + "public_sha256": "306b0ae1b97e4a85dab84cd8969428fd460b89729ff861b65bbd102df096e225", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "b7b9db05b12f658f737aca0624f4c43240dec9ce", + "source_lfs_sha256": null, + "source_sha256": "8d9640467f5797d49787f39a913dda63b5370b20319a2a101ae86944abefde63", + "source_bytes": 594, + "public_sha256": "8d9640467f5797d49787f39a913dda63b5370b20319a2a101ae86944abefde63", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "269f3115b1cff942a01f2a84865e6ba577439063", + "source_lfs_sha256": null, + "source_sha256": "9289738efac9019f4cac4f89cc4d9bd845a386e285ff0282e11cece5c0249c64", + "source_bytes": 1515, + "public_sha256": "9289738efac9019f4cac4f89cc4d9bd845a386e285ff0282e11cece5c0249c64", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "b9bb5eb5af3269189ddede606ee3a70cae8d6770", + "source_lfs_sha256": null, + "source_sha256": "7e816992efd891599344090960dfa2d44d2eebee4cd96bed4d79a1a9b10e76b0", + "source_bytes": 2631, + "public_sha256": "7e816992efd891599344090960dfa2d44d2eebee4cd96bed4d79a1a9b10e76b0", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "7279f7955ee36032b0c871f329cf5dc2d9cec845", + "source_lfs_sha256": null, + "source_sha256": "377d20dc15a787caf57fd7d701e8276837f481b3ccd5474e7378d6f084e4fa21", + "source_bytes": 2389, + "public_sha256": "377d20dc15a787caf57fd7d701e8276837f481b3ccd5474e7378d6f084e4fa21", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "a1e7871c51a175c1eaa7ff83af4d4421bb51588a", + "source_lfs_sha256": null, + "source_sha256": "9a9d342a3e53f2bdfbd9a0589eff26d39be79ee15f524e0cd2bf3ae150496d63", + "source_bytes": 1804, + "public_sha256": "9a9d342a3e53f2bdfbd9a0589eff26d39be79ee15f524e0cd2bf3ae150496d63", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "5df0a6c4397b0094484775131f3bc6444f790b43", + "source_lfs_sha256": null, + "source_sha256": "8672cc3c33d1cfb7ba39622036d8acda32e0055c5049166747d3b3c2dbe2637c", + "source_bytes": 3083, + "public_sha256": "8672cc3c33d1cfb7ba39622036d8acda32e0055c5049166747d3b3c2dbe2637c", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "3774df3f6f3a32f4b501ac5fa1e8e2d8b1c2411c", + "source_lfs_sha256": null, + "source_sha256": "218e653d212436d962aa04fbdd6bb8f436c4a44f4c4d0a4bbd8de06a1bbe08ca", + "source_bytes": 2819, + "public_sha256": "218e653d212436d962aa04fbdd6bb8f436c4a44f4c4d0a4bbd8de06a1bbe08ca", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "ccf9fd897bf48bbc975a94def1b5b74dc82e24ee", + "source_lfs_sha256": null, + "source_sha256": "60bb53a9b087aed86814999051851fe968395b8bed7e96d49560009c09bc9575", + "source_bytes": 2308035, + "public_sha256": "60bb53a9b087aed86814999051851fe968395b8bed7e96d49560009c09bc9575", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "98aaeb941335fca3750f2a2d2c3d950993fbcd74", + "source_lfs_sha256": null, + "source_sha256": "c6fb79aa5ac2da53894198733ba558d0cb4791496f1e683abc00883a665ab4b9", + "source_bytes": 88035, + "public_sha256": "c6fb79aa5ac2da53894198733ba558d0cb4791496f1e683abc00883a665ab4b9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-ssh-architecture-32k-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "2c13b291fff0db4fcfb5de0c2dc3d74539cf600f", + "source_lfs_sha256": null, + "source_sha256": "752d007fb01378ac629c74c9735f3205d79795a4f448c35c414030e646cef88c", + "source_bytes": 2230781, + "public_sha256": "752d007fb01378ac629c74c9735f3205d79795a4f448c35c414030e646cef88c", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/candidate_report.json", + "source_git_blob": "5b92d5f42b18f1c8981671e2c2dd994e7bb06190", + "source_lfs_sha256": null, + "source_sha256": "002ac8396e1aedea44534f8f119bfdcd5ff27b260a32b0e52225051e5fdd1237", + "source_bytes": 1536, + "public_sha256": "002ac8396e1aedea44534f8f119bfdcd5ff27b260a32b0e52225051e5fdd1237", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/candidate_report.jsonl", + "source_git_blob": "62c0055be2af870a68b83eff8fef4b701ee20524", + "source_lfs_sha256": null, + "source_sha256": "2e91c7bb9beea789f56d4ca913c9fb5411f26701ed3f021b4b223ca45f995e23", + "source_bytes": 1420, + "public_sha256": "2e91c7bb9beea789f56d4ca913c9fb5411f26701ed3f021b4b223ca45f995e23", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/eval.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/eval_audits.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/eval_candidates.jsonl", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/llamafactory/B5.json", + "source_git_blob": "4bbe609aae2e4b575f5f37460002bc538ad5e85b", + "source_lfs_sha256": null, + "source_sha256": "76dd5bd93a3fd3bdd4a6a24e4aea863713b2fdd4089f210b781fc1f720df415e", + "source_bytes": 2549917, + "public_sha256": "76dd5bd93a3fd3bdd4a6a24e4aea863713b2fdd4089f210b781fc1f720df415e", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/llamafactory/B5.meta.json", + "source_git_blob": "3e4671b86ae080615084a1e4965a3bd65bbed214", + "source_lfs_sha256": null, + "source_sha256": "8fe57d7b71b52d5da34117a54addec6d1c3bb2781bc1f11b81ccf4547bce7656", + "source_bytes": 509, + "public_sha256": "8fe57d7b71b52d5da34117a54addec6d1c3bb2781bc1f11b81ccf4547bce7656", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/llamafactory/dataset_info.json", + "source_git_blob": "2e26af10d47fc01cfbf3332e493fca65afc62ee3", + "source_lfs_sha256": null, + "source_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "source_bytes": 307, + "public_sha256": "3cb3dc8c5d6a77d4b469bfd6276df3eab73cdcfb420ab8a00219efee6944bffb", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/llamafactory/export_summary.json", + "source_git_blob": "31beb9bd7f8572d8767c55f99d87376e828010a6", + "source_lfs_sha256": null, + "source_sha256": "006825eae9f9cb21322b0afb9ed5258243710c5ac95a7c04a5adb63ec24e6a96", + "source_bytes": 701, + "public_sha256": "006825eae9f9cb21322b0afb9ed5258243710c5ac95a7c04a5adb63ec24e6a96", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/llamafactory/training_export_manifest.json", + "source_git_blob": "0bb9fd32aab8cce982079eea7aad535934f19252", + "source_lfs_sha256": null, + "source_sha256": "015ab24c92de2c56de2a723f9e9102b16a38023fc44f4fcc6611cc53f9ed7a31", + "source_bytes": 1520, + "public_sha256": "015ab24c92de2c56de2a723f9e9102b16a38023fc44f4fcc6611cc53f9ed7a31", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/promoted_128k.jsonl", + "source_git_blob": "58f7019c877437cfa909af3612d634ef96cb11d4", + "source_lfs_sha256": null, + "source_sha256": "387553873b410b0e26c5b1c1ce76d7fa8f48c49b7caf8eac5280d26021b457be", + "source_bytes": 7029318, + "public_sha256": "387553873b410b0e26c5b1c1ce76d7fa8f48c49b7caf8eac5280d26021b457be", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/promoted_64k.jsonl", + "source_git_blob": "5974022ace68ef8a42a28044bcb103a2f94be4f8", + "source_lfs_sha256": null, + "source_sha256": "de14c13b2fae192f6a0609563232cef48f6202a61f3e87de308d2bef66946054", + "source_bytes": 5374863, + "public_sha256": "de14c13b2fae192f6a0609563232cef48f6202a61f3e87de308d2bef66946054", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/quality_report.json", + "source_git_blob": "8badb251aa8633fb4c4d1a568528f8cd175427bd", + "source_lfs_sha256": null, + "source_sha256": "08b37d29cce4bd2c09c7864cbe56e4f14dbec8beb599be1dd365b5f4641cb98a", + "source_bytes": 2656, + "public_sha256": "08b37d29cce4bd2c09c7864cbe56e4f14dbec8beb599be1dd365b5f4641cb98a", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/quality_report.jsonl", + "source_git_blob": "ead672f01a61f4032721e75d7d44ecc204acba3e", + "source_lfs_sha256": null, + "source_sha256": "6127ae7c3d9a2a3fa57b2caaa7d4b0701811d64f9661a59dc6790075979008e0", + "source_bytes": 2410, + "public_sha256": "6127ae7c3d9a2a3fa57b2caaa7d4b0701811d64f9661a59dc6790075979008e0", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/release_gate_receipt.json", + "source_git_blob": "c9ca39bea21caae420c3f1d9c18b20b923396255", + "source_lfs_sha256": null, + "source_sha256": "efd1f7e280eafd3dd03ddf06d64425b138f9717066581b68998c421e8979682c", + "source_bytes": 1808, + "public_sha256": "efd1f7e280eafd3dd03ddf06d64425b138f9717066581b68998c421e8979682c", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/release_selection.json", + "source_git_blob": "71ac53186dd32b97bbc0c06b4fc12cf21df699d8", + "source_lfs_sha256": null, + "source_sha256": "f199f235e474ab17916606d821e6e1ccdcb5d42eecaddf9e3d6d6a7d409642e0", + "source_bytes": 4138, + "public_sha256": "f199f235e474ab17916606d821e6e1ccdcb5d42eecaddf9e3d6d6a7d409642e0", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/release_selection.jsonl", + "source_git_blob": "f9d277cd17302bfedcf6038dee71cd14bbf3c714", + "source_lfs_sha256": null, + "source_sha256": "72f6849a6a3288cee10dd7b5d9b9a3617faeef2f17ca552389ed237032751adc", + "source_bytes": 3838, + "public_sha256": "72f6849a6a3288cee10dd7b5d9b9a3617faeef2f17ca552389ed237032751adc", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/train.jsonl", + "source_git_blob": "59b8b060e947f8c6c02805db3073abfd95bfa258", + "source_lfs_sha256": "01ef1c0f38aa021a47017902146a908f5cd62d729a8f0a7640581ed4538cccf4", + "source_sha256": "01ef1c0f38aa021a47017902146a908f5cd62d729a8f0a7640581ed4538cccf4", + "source_bytes": 12404181, + "public_sha256": "01ef1c0f38aa021a47017902146a908f5cd62d729a8f0a7640581ed4538cccf4", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/train_audits.jsonl", + "source_git_blob": "b489aa4337e1fdf59977d7498b71ff488671246e", + "source_lfs_sha256": null, + "source_sha256": "143967b59099a110c40c4007afe5c751813d2d630f9078b10a88024cf6294af9", + "source_bytes": 180283, + "public_sha256": "143967b59099a110c40c4007afe5c751813d2d630f9078b10a88024cf6294af9", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/train_candidates.jsonl", + "source_git_blob": "d19d2f4e7dd727b6b0224940cbd59840faa8ddaa", + "source_lfs_sha256": "6c202213590e1bfae6d21c55fc76f28bd6a36118742d59516f1dd535eb3b4866", + "source_sha256": "6c202213590e1bfae6d21c55fc76f28bd6a36118742d59516f1dd535eb3b4866", + "source_bytes": 12250510, + "public_sha256": "6c202213590e1bfae6d21c55fc76f28bd6a36118742d59516f1dd535eb3b4866", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/P64_TRAINING_MANIFEST.json", + "source_git_blob": "9a5afa029dd2869adbf0d48b0acbc2aecc8d1348", + "source_lfs_sha256": null, + "source_sha256": "8e2c0aeb220f2117699af7bfec3cdb355567434dddc26b61b026b1db830ac459", + "source_bytes": 70530, + "public_sha256": "2e6b1ee059f4d50f36e3bded63d2712231167baf96766132b1d7d39b1e0592da", + "operation": "add_sanitized_file", + "audit_status": "passed", + "context_decision": "302 approved operational path prefixes replaced; JSON keys, numeric lexemes and nonpath scientific content unchanged", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/TASKBANK_P64.yaml", + "source_git_blob": "db4d07fb025025acf98c987e502fa4653c6a09b3", + "source_lfs_sha256": null, + "source_sha256": "fee14c63bc6396991ac6b145ca7837c563c592070b832443a2a9490d05380460", + "source_bytes": 387, + "public_sha256": "fee14c63bc6396991ac6b145ca7837c563c592070b832443a2a9490d05380460", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/codeforge_eval.jsonl", + "source_git_blob": "1b240e5e7a461b5a081baf963d2d96faa0293c23", + "source_lfs_sha256": "b6705df17f1c099084dcb2d0a56a3e29376ac80fb4a83eefa4a7c9a81f9d7bee", + "source_sha256": "b6705df17f1c099084dcb2d0a56a3e29376ac80fb4a83eefa4a7c9a81f9d7bee", + "source_bytes": 56658476, + "public_sha256": "b6705df17f1c099084dcb2d0a56a3e29376ac80fb4a83eefa4a7c9a81f9d7bee", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/codeforge_train.jsonl", + "source_git_blob": "f09b57c89b02d5b5fa2c728d62fde62cc79a3da9", + "source_lfs_sha256": "e78852013aaa49ea54e65e0d475d5b8e0c61dc1e096ddcfcd4a28ccb3cd7cbae", + "source_sha256": "e78852013aaa49ea54e65e0d475d5b8e0c61dc1e096ddcfcd4a28ccb3cd7cbae", + "source_bytes": 242734380, + "public_sha256": "e78852013aaa49ea54e65e0d475d5b8e0c61dc1e096ddcfcd4a28ccb3cd7cbae", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/dataset_info.json", + "source_git_blob": "bd4bd4c68610d116b4f0b195d969947a932d6500", + "source_lfs_sha256": null, + "source_sha256": "c41d28e54498091aaf1a8d3f8931a27c298923e84800ed53afb0e5b8315ed301", + "source_bytes": 850, + "public_sha256": "c41d28e54498091aaf1a8d3f8931a27c298923e84800ed53afb0e5b8315ed301", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/finance_eval.jsonl", + "source_git_blob": "cb7eb11d2f5102b40539b44707eda2b114ee55f0", + "source_lfs_sha256": "3ee5d59d48d6f87c032b1afb8ded4f9b255aa6be12cf51d79eb54d32b6875b89", + "source_sha256": "3ee5d59d48d6f87c032b1afb8ded4f9b255aa6be12cf51d79eb54d32b6875b89", + "source_bytes": 240577104, + "public_sha256": "3ee5d59d48d6f87c032b1afb8ded4f9b255aa6be12cf51d79eb54d32b6875b89", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/finance_train.jsonl", + "source_git_blob": "02151bdfc7ada218eba897c684176a83b21c587b", + "source_lfs_sha256": "18cce1a697eeacd468b1a1ac278c42734a91711894ec65c31cadd79bb170da57", + "source_sha256": "18cce1a697eeacd468b1a1ac278c42734a91711894ec65c31cadd79bb170da57", + "source_bytes": 653089550, + "public_sha256": "18cce1a697eeacd468b1a1ac278c42734a91711894ec65c31cadd79bb170da57", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_owner_authorized_download_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/sample_index.jsonl", + "source_git_blob": "9c994746808d0384120024c66ce531f116fa66e4", + "source_lfs_sha256": null, + "source_sha256": "1e767e5f1fe489887db917615aa071155e33aa3859909ce9a5563a551f9debec", + "source_bytes": 796305, + "public_sha256": "1e767e5f1fe489887db917615aa071155e33aa3859909ce9a5563a551f9debec", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "All parsed task records and strings reviewed: preserve public code/specification material, public references, institution attribution and fictional example-domain contacts; exact protocol/test fixtures are not operational secrets", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/alphabet.stderr.txt", + "source_git_blob": "f9a22f13d3c2bd30b5099838db584d07c7e0249c", + "source_lfs_sha256": null, + "source_sha256": "cc895459700ed89292333afb615f5aac9313a8505ae0e2ca94e0cd71a571e247", + "source_bytes": 203, + "public_sha256": "cc895459700ed89292333afb615f5aac9313a8505ae0e2ca94e0cd71a571e247", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/alphabet.stdout.txt", + "source_git_blob": "29a36bc653444b719cbf483b8d439a2be2c87069", + "source_lfs_sha256": null, + "source_sha256": "dc55a2ded6b3731f9cbaee7d0465e536b100a9057562317d36fc5753658f190d", + "source_bytes": 170, + "public_sha256": "dc55a2ded6b3731f9cbaee7d0465e536b100a9057562317d36fc5753658f190d", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/amazon.stderr.txt", + "source_git_blob": "ca8dcb38124ec3b5ec07846b0418144c79f814ee", + "source_lfs_sha256": null, + "source_sha256": "e868aff568484b220a06803e7017afed97554f2a345dc997c63be395991da472", + "source_bytes": 203, + "public_sha256": "e868aff568484b220a06803e7017afed97554f2a345dc997c63be395991da472", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/amazon.stdout.txt", + "source_git_blob": "5b72ef660d09edd488a11f12d049ea7deacda833", + "source_lfs_sha256": null, + "source_sha256": "7abd1d1dd131d28bab9fa188e1471a4f59014dcedf49d359427acf6d9a653ccc", + "source_bytes": 170, + "public_sha256": "7abd1d1dd131d28bab9fa188e1471a4f59014dcedf49d359427acf6d9a653ccc", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/amd.stderr.txt", + "source_git_blob": "542316cc980f4c7143eff86aa8b426473d1e931c", + "source_lfs_sha256": null, + "source_sha256": "fec4f1d46156e957cf8ade672c0006f28e00e27bea99bc83c2196bd1936e1dd6", + "source_bytes": 203, + "public_sha256": "fec4f1d46156e957cf8ade672c0006f28e00e27bea99bc83c2196bd1936e1dd6", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/amd.stdout.txt", + "source_git_blob": "6bdc75499c6afbeb41bdc9606c790090fdaed179", + "source_lfs_sha256": null, + "source_sha256": "44ebd13f58de954a56759ed40ca11921c7fee3b3316583977db394cf2b773289", + "source_bytes": 190, + "public_sha256": "44ebd13f58de954a56759ed40ca11921c7fee3b3316583977db394cf2b773289", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/codeforge.stderr.txt", + "source_git_blob": "18acb7570dd74514418d969bf17c92601ed59fae", + "source_lfs_sha256": null, + "source_sha256": "05b56bba2468728d89450067dc8b63092b87b3e6af4681755532261399bb25c2", + "source_bytes": 203, + "public_sha256": "05b56bba2468728d89450067dc8b63092b87b3e6af4681755532261399bb25c2", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/codeforge.stdout.txt", + "source_git_blob": "a5ce7a6e767092eda23b8b729e8797e4acda3667", + "source_lfs_sha256": null, + "source_sha256": "f03387e6a1fea8fb09e5c540567217baaa1ce10b0dc117ec8fff6d3cc55fb9c4", + "source_bytes": 555, + "public_sha256": "f03387e6a1fea8fb09e5c540567217baaa1ce10b0dc117ec8fff6d3cc55fb9c4", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/intel.stderr.txt", + "source_git_blob": "2f29bfde17acd4977ee4ac771d34c48446d77790", + "source_lfs_sha256": null, + "source_sha256": "8af952cb45c1a6eb4ce0214b990c7ef4f9311d176cd77172c08a070b24096f00", + "source_bytes": 203, + "public_sha256": "8af952cb45c1a6eb4ce0214b990c7ef4f9311d176cd77172c08a070b24096f00", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/intel.stdout.txt", + "source_git_blob": "4dc9a8edaebc362a8151b7414ab68069aa2a5ca1", + "source_lfs_sha256": null, + "source_sha256": "b3866d34e5636279075a1312d51107534510febd86ba24d07e2cd75dec59eb0c", + "source_bytes": 170, + "public_sha256": "b3866d34e5636279075a1312d51107534510febd86ba24d07e2cd75dec59eb0c", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/meta.stderr.txt", + "source_git_blob": "c28207414a5843ef4099222b26c5ea232f57b549", + "source_lfs_sha256": null, + "source_sha256": "6a502380c840f1de381a131251056c29f742b5aa0d74e0badca7b30c30aa4b92", + "source_bytes": 203, + "public_sha256": "6a502380c840f1de381a131251056c29f742b5aa0d74e0badca7b30c30aa4b92", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/meta.stdout.txt", + "source_git_blob": "c4440e0bc4b29fdebc647a72ba491ce01837c239", + "source_lfs_sha256": null, + "source_sha256": "bba45f92c70adb32534f91cf969bc511fe3b9aba6930032da74b23f19d178db7", + "source_bytes": 170, + "public_sha256": "bba45f92c70adb32534f91cf969bc511fe3b9aba6930032da74b23f19d178db7", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/micron.stderr.txt", + "source_git_blob": "7c8d976b3a5b0668058b94d837bffcc32db7929a", + "source_lfs_sha256": null, + "source_sha256": "e3407338f1df5a0488d64695e62de0481b1d2665a4e17a60cc86720944d75dca", + "source_bytes": 203, + "public_sha256": "e3407338f1df5a0488d64695e62de0481b1d2665a4e17a60cc86720944d75dca", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/micron.stdout.txt", + "source_git_blob": "fe112b1d62b7aa62e8ca5dfe42d3655748c1d8c9", + "source_lfs_sha256": null, + "source_sha256": "0fc495b43da97e9c9bd85d3f6b5c84607f7fa78ff78f2a4fbf9d68d7457e13c5", + "source_bytes": 170, + "public_sha256": "0fc495b43da97e9c9bd85d3f6b5c84607f7fa78ff78f2a4fbf9d68d7457e13c5", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/nvidia.stderr.txt", + "source_git_blob": "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391", + "source_lfs_sha256": null, + "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "source_bytes": 0, + "public_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/nvidia.stdout.txt", + "source_git_blob": "df02404ae0cfb9cb187c09d89d9c91b8ab58f2b2", + "source_lfs_sha256": null, + "source_sha256": "f4c3abf78447d05b5ef88054735a7eb088586ed7a5f55a768f999d49c27c5079", + "source_bytes": 205, + "public_sha256": "f4c3abf78447d05b5ef88054735a7eb088586ed7a5f55a768f999d49c27c5079", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/siriusxm.stderr.txt", + "source_git_blob": "5f6203a3ea82835862ab8341b42c5c0d75d0a842", + "source_lfs_sha256": null, + "source_sha256": "792e32a48e3bfbee7349458dccd310e9700bb986cb5747c78c0df1db2973998a", + "source_bytes": 203, + "public_sha256": "792e32a48e3bfbee7349458dccd310e9700bb986cb5747c78c0df1db2973998a", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "local-probe-train-ready/p64-primary-training-v2/source_validation/siriusxm.stdout.txt", + "source_git_blob": "3eeb139e03f123cd8b36ea7376d7cdeb31e52760", + "source_lfs_sha256": null, + "source_sha256": "b7cdbf7eaa3ab3474e573cba9967f3e36c627002da470c292b0656b0c1fe0e65", + "source_bytes": 170, + "public_sha256": "b7cdbf7eaa3ab3474e573cba9967f3e36c627002da470c292b0656b0c1fe0e65", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + }, + { + "path": "release_inventory.json", + "source_git_blob": "c177cbff404eee331ff76d17586738116c7b39cb", + "source_lfs_sha256": null, + "source_sha256": "9bfbda84431a4cd3cad0ad511e2be5b8b05336bdcb3acaa0ffe3e85869533145", + "source_bytes": 3940, + "public_sha256": "9bfbda84431a4cd3cad0ad511e2be5b8b05336bdcb3acaa0ffe3e85869533145", + "operation": "copy_exact_source_object", + "audit_status": "passed", + "context_decision": "no_confidentiality_candidates", + "audit_mode": "full_immutable_git_blob_hash_verified" + } + ] +} diff --git a/README.md b/README.md new file mode 100644 index 0000000000000000000000000000000000000000..9aa32af385091d59412a00021dbb66b98106dff9 --- /dev/null +++ b/README.md @@ -0,0 +1,63 @@ +--- +pretty_name: LongWorld Real Workflow P6 Probe +language: + - en +task_categories: + - question-answering + - text-generation +size_categories: + - n<1K +license: other +tags: + - long-context + - synthetic-data + - github + - arxiv +--- + +# LongWorld research data + +This repository preserves the workflow products and local-probe exports from the original research-data snapshot. The P6 probe described below is one retained product; additional exports remain under `local-probe-train-ready/` with their original admission labels. + +[ARTIFACTS.json](ARTIFACTS.json) records source and public file hashes. Machine-specific paths in the P64 training manifest use `/WORKSPACE/` aliases; numeric values, scientific identifiers, keys, and result labels are unchanged. + +## P6 workflow probe + +This is a public, local-engineering release for long-context SFT experiments. It contains 542 promoted rows from 12 causally executable worlds. The release passed the `p6-source-dependent-probe-12-v1` quality profile and is not a production-approved dataset. + +## What is in this release + +- 428 train rows and 114 eval rows; split boundaries are world-level. +- 164 16K, 222 32K, and 156 exact-tokenizer-verified 64K rows. +- Four company, four research-lab, and four code-forge worlds. +- 61 unique base tasks, 24 executable proofs, 23 answer programs, and 7 verified real-source relation graphs. +- 16 real-source exact-64K rows, 8 real-source base tasks, and 3 real-source families. +- Full, counterfactual, and ordered-artifact views. Counterfactual answers are obtained by executable replay. + +The real-source worlds use signed exports of public GitHub workflow records and official arXiv metadata/source archives. The source text is parsed into state, causal relations, proof requirements, and answers. Synthetic workflow events supply controlled counterfactuals and delayed effects; they are explicitly classified and are not represented as authentic source records. + +## Quality controls + +The release gate requires strict replay, source-bundle binding, source-to-state-to-answer dependence, factual/counterfactual twins, remove-one failure, single-essential-document insufficiency, body-corruption failure, contiguous-window and sparse/dense retrieval checks, semantic-growth checks, exact 64K token verification, zero exact row duplicates, and world/domain quotas. The passing receipt is included at `05_promoted/release_gate_pass.json`. + +Training exports use unique ShareGPT rows. `B5w.datasets.yaml` applies sampler weight 2 to the single `B5w.json` file; no second copied weight shard is included. + +## Unseen evaluation status + +The local-engineering split artifacts report topology/operator and source-document-family axes as ready. World/entity is only `world_atomic_only`, so production publication of that axis is blocked. Domain-composition is blocked because this 12-world probe has fewer than two composition groups. These limitations must be resolved before a production unseen claim. + +## Provenance and trust scope + +This package is signed with local probe role keys and is marked `local_engineering`. It does not have an independently controlled KMS approval or a production trust-root pin. The OpenReview exporter and parser are implemented and tested, but live OpenReview acquisition was blocked by an upstream HTTP 403 during this run; no OpenReview response is present in these rows. arXiv and GitHub provenance is included only when the exact signed replay bundles verify. + +## Intended use and limitations + +Use this release for controlled SFT/CPT pipeline probes, ablations, and evaluator development. Do not describe it as a production-approved corpus, an OpenReview corpus, or a general long-context benchmark. The dataset is small, English-heavy, and intentionally contains only three domains. It does not yet establish unseen domain-composition generalization or external benchmark gains. + +Source documents retain their original provenance and applicable licenses. The aggregate package uses `license: other` because source-specific terms differ; users must review source metadata and applicable licenses before redistribution or model release. Public availability does not change the local-engineering status or establish production approval. + +## Related resources + +- [128k baseline data](https://huggingface.co/datasets/Xnhyacinth/longworld-128k-sft-baselines): ACC, LongTrace and LongMIT inputs. +- [Research checkpoint exports](https://huggingface.co/Xnhyacinth/LongWorld-Research-Checkpoints): three self-contained final step-680 variants—Instruct ACC, Base ACC and Base LongTrace. + diff --git a/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/promoted_64k.jsonl b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/promoted_64k.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..5974022ace68ef8a42a28044bcb103a2f94be4f8 --- /dev/null +++ b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/promoted_64k.jsonl @@ -0,0 +1,3 @@ +{"actual_context_tokens": 65493, "answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "answer_program_id": "ietf.tls13_handshake_succession.v1", "artifact_classification": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef:0:41965:80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49", "source_char_end": 41965, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:5077", "source_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "evidence_role": "causal_supporting", "provenance_id": "source-span-sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd:335784:337410:0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5", "source_char_end": 337410, "source_char_start": 335784, "source_origin": "real_public", "source_record_id": "ietf:rfc:8446", "source_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "evidence_role": "causal_supporting", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:0:35597:ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6", "source_char_end": 35597, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:5246", "source_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "counterfactual_child_text_sha256": "f9892786fca5daa1aca587f899ae9a4d74912e13b89740492729f5c3fe986064", "counterfactual_operation_local_byte_end": 447, "counterfactual_operation_local_byte_start": 361, "counterfactual_operation_local_char_end": 447, "counterfactual_operation_local_char_start": 361, "counterfactual_operation_source_char_end": 447, "counterfactual_operation_source_char_start": 361, "counterfactual_parent_provenance_id": "source-span-sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd:0:38698:61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e", "counterfactual_parent_sidecar_sha256": "e77fe08f470f227b433656f0fac14e677653bdb1f53faa939175c5c883608e76", "counterfactual_parent_source_binding_sha256": "fc34a690fcb8ffffb64df92f1df724ea75d43ce88c10daf432fb085fee09ddbe", "counterfactual_parent_source_origin": "real_public", "counterfactual_parent_text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e", "evidence_role": "causal_gold", "provenance_id": "counterfactual-projection-sha256:f9892786fca5daa1aca587f899ae9a4d74912e13b89740492729f5c3fe986064", "source_char_end": 38698, "source_char_start": 0, "source_origin": "synthetic_counterfactual", "source_record_id": "ietf:rfc:8446", "source_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "hybrid_causal"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "evidence_role": "causal_supporting", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:35597:72013:3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69", "source_char_end": 72013, "source_char_start": 35597, "source_origin": "real_public", "source_record_id": "ietf:rfc:5246", "source_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6:0:342:d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b", "source_char_end": 342, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "source_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:72013:109924:596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef", "source_char_end": 109924, "source_char_start": 72013, "source_origin": "real_public", "source_record_id": "ietf:rfc:5246", "source_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db:0:21461:bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a", "source_char_end": 21461, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:6961", "source_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a:0:16341:4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473", "source_char_end": 16341, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:5705", "source_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0:0:55067:a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813", "source_char_end": 55067, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:6066", "source_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}], "attestation": {"digest": "ca07540a41db0914cc2b959b5eb42e45194a1800eaa1138555c1fbefc1e7d111", "environment": "probe", "key_id": "probe-promotion-c9a5dab33e62d6c983d4696be812c27f", "purpose": "sft_row", "role": "promotion", "scheme": "hmac-sha256-v2"}, "authentic_source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}], "authentic_source_relation_id": "e0dcb1ad1463850449f0", "base_task_id": "776021e7d8e7f94c336a", "cf_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "composition_method": "counterfactual_twin", "content_gate_eligible": true, "context": "Question:\nResolve the effective TLS 1.3 handshake protocol succession at 2026-01-31T00:00:00Z for this scenario (canonical JSON): {\"protocol\":\"tls\",\"publication\":\"rfc8446\",\"succession\":\"obsoletes_and_updates\",\"version\":\"1.3\"}. Return exactly one JSON object with these keys in this order: [\"current_protocol\",\"obsoletes_session_tickets\",\"obsoletes_tls12\",\"obsoletes_ocsp_stapling\",\"updates_exporters\",\"updates_extensions\"]. Use this exact per-field output codebook (canonical JSON): {\"current_protocol\":{\"code\":\"TLS_1_3_RFC8446\",\"meaning\":\"RFC 8446 is TLS protocol version 1.3\"},\"obsoletes_ocsp_stapling\":{\"code\":\"RFC6961\",\"meaning\":\"RFC 8446 obsoletes multiple certificate status requests\"},\"obsoletes_session_tickets\":{\"code\":\"RFC5077\",\"meaning\":\"RFC 8446 obsoletes TLS session resumption tickets\"},\"obsoletes_tls12\":{\"code\":\"RFC5246\",\"meaning\":\"RFC 8446 obsoletes TLS protocol version 1.2\"},\"updates_exporters\":{\"code\":\"RFC5705\",\"meaning\":\"RFC 8446 updates TLS keying material exporters\"},\"updates_extensions\":{\"code\":\"RFC6066\",\"meaning\":\"RFC 8446 updates TLS extension definitions\"}}. Resolve each field independently from the supplied RFC graph; use the string \"UNKNOWN\" only for a field whose required evidence or relation is absent.\n\nContext (internal records):\nNetwork Working Group J. Salowey\nRequest for Comments: 5077 H. Zhou\nObsoletes: 4507 Cisco Systems\nCategory: Standards Track P. Eronen\n Nokia\n H. Tschofenig\n Nokia Siemens Networks\n January 2008\n\n\n Transport Layer Security (TLS) Session Resumption without\n Server-Side State\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document describes a mechanism that enables the Transport Layer\n Security (TLS) server to resume sessions and avoid keeping per-client\n session state. The TLS server encapsulates the session state into a\n ticket and forwards it to the client. The client can subsequently\n resume a session using the obtained ticket. This document obsoletes\n RFC 4507.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 1]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3. Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . 4\n 3.2. SessionTicket TLS Extension . . . . . . . . . . . . . . . 7\n 3.3. NewSessionTicket Handshake Message . . . . . . . . . . . . 8\n 3.4. Interaction with TLS Session ID . . . . . . . . . . . . . 9\n 4. Recommended Ticket Construction . . . . . . . . . . . . . . . 10\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . 12\n 5.1. Invalidating Sessions . . . . . . . . . . . . . . . . . . 12\n 5.2. Stolen Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.3. Forged Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.4. Denial of Service Attacks . . . . . . . . . . . . . . . . 12\n 5.5. Ticket Protection Key Management . . . . . . . . . . . . . 13\n 5.6. Ticket Lifetime . . . . . . . . . . . . . . . . . . . . . 13\n 5.7. Alternate Ticket Formats and Distribution Schemes . . . . 13\n 5.8. Identity Privacy, Anonymity, and Unlinkability . . . . . . 14\n 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 14\n 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 15\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 15\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 15\n Appendix A. Discussion of Changes to RFC 4507 . . . . . . . . . . 17\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 2]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n1. Introduction\n\n This document defines a way to resume a Transport Layer Security\n (TLS) session without requiring session-specific state at the TLS\n server. This mechanism may be used with any TLS ciphersuite. This\n document applies to both TLS 1.0 defined in [RFC2246], and TLS 1.1\n defined in [RFC4346]. The mechanism makes use of TLS extensions\n defined in [RFC4366] and defines a new TLS message type.\n\n This mechanism is useful in the following situations:\n\n 1. servers that handle a large number of transactions from different\n users\n\n 2. servers that desire to cache sessions for a long time\n\n 3. ability to load balance requests across servers\n\n 4. embedded servers with little memory\n\n This document obsoletes RFC 4507 [RFC4507] to correct an error in the\n encoding that caused the specification to differ from deployed\n implementations. At the time of this writing, there are no known\n implementations that follow the encoding specified in RFC 4507. This\n update to RFC 4507 aligns the document with currently deployed\n implementations. More details of the change are given in Appendix A.\n\n2. Terminology\n\n Within this document, the term 'ticket' refers to a cryptographically\n protected data structure that is created and consumed by the server\n to rebuild session-specific state.\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Protocol\n\n This specification describes a mechanism to distribute encrypted\n session-state information to the client in the form of a ticket and a\n mechanism to present the ticket back to the server. The ticket is\n created by a TLS server and sent to a TLS client. The TLS client\n presents the ticket to the TLS server to resume a session.\n Implementations of this specification are expected to support both\n mechanisms. Other specifications can take advantage of the session\n tickets, perhaps specifying alternative means for distribution or\n selection. For example, a separate specification may describe an\n\n\n\nSalowey, et al. Standards Track [Page 3]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n alternate way to distribute a ticket and use the TLS extension in\n this document to resume the session. This behavior is beyond the\n scope of the document and would need to be described in a separate\n specification.\n\n3.1. Overview\n\n The client indicates that it supports this mechanism by including a\n SessionTicket TLS extension in the ClientHello message. The\n extension will be empty if the client does not already possess a\n ticket for the server. The server sends an empty SessionTicket\n extension to indicate that it will send a new session ticket using\n the NewSessionTicket handshake message. The extension is described\n in Section 3.2.\n\n If the server wants to use this mechanism, it stores its session\n state (such as ciphersuite and master secret) to a ticket that is\n encrypted and integrity-protected by a key known only to the server.\n The ticket is distributed to the client using the NewSessionTicket\n TLS handshake message described in Section 3.3. This message is sent\n during the TLS handshake before the ChangeCipherSpec message, after\n the server has successfully verified the client's Finished message.\n\n Client Server\n\n ClientHello\n (empty SessionTicket extension)-------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1: Message Flow for Full Handshake Issuing New Session Ticket\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 4]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n The client caches this ticket along with the master secret and other\n parameters associated with the current session. When the client\n wishes to resume the session, it includes the ticket in the\n SessionTicket extension within the ClientHello message. Appendix A\n provides a detailed description of the encoding of the extension and\n changes from RFC 4507. The server then decrypts the received ticket,\n verifies the ticket's validity, retrieves the session state from the\n contents of the ticket, and uses this state to resume the session.\n The interaction with the TLS Session ID is described in Section 3.4.\n If the server successfully verifies the client's ticket, then it may\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello.\n\n Client Server\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2: Message Flow for Abbreviated Handshake Using New Session\n Ticket\n\n A recommended ticket format is given in Section 4.\n\n If the server cannot or does not want to honor the ticket, then it\n can initiate a full handshake with the client.\n\n In the case that the server does not wish to issue a new ticket at\n this time, it just completes the handshake without including a\n SessionTicket extension or NewSessionTicket handshake message. This\n is shown below (this flow is identical to Figure 1 in RFC 4346,\n except for the SessionTicket extension in the first message):\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 5]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 3: Message Flow for Server Completing Full Handshake Without\n Issuing New Session Ticket\n\n It is also permissible to have an exchange similar to Figure 3 using\n the abbreviated handshake defined in Figure 2 of RFC 4346, where the\n client uses the SessionTicket extension to resume the session, but\n the server does not wish to issue a new ticket, and therefore does\n not send a SessionTicket extension.\n\n If the server rejects the ticket, it may still wish to issue a new\n ticket after performing the full handshake as shown below (this flow\n is identical to Figure 1, except the SessionTicket extension in the\n ClientHello is not empty):\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 6]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 4: Message Flow for Server Rejecting Ticket, Performing Full\n Handshake, and Issuing New Session Ticket\n\n3.2. SessionTicket TLS Extension\n\n The SessionTicket TLS extension is based on [RFC4366]. The format of\n the ticket is an opaque structure used to carry session-specific\n state information. This extension may be sent in the ClientHello and\n ServerHello.\n\n If the client possesses a ticket that it wants to use to resume a\n session, then it includes the ticket in the SessionTicket extension\n in the ClientHello. If the client does not have a ticket and is\n prepared to receive one in the NewSessionTicket handshake message,\n then it MUST include a zero-length ticket in the SessionTicket\n extension. If the client is not prepared to receive a ticket in the\n NewSessionTicket handshake message, then it MUST NOT include a\n SessionTicket extension unless it is sending a non-empty ticket it\n received through some other means from the server.\n\n The server uses a zero-length SessionTicket extension to indicate to\n the client that it will send a new session ticket using the\n NewSessionTicket handshake message described in Section 3.3. The\n server MUST send this extension in the ServerHello if it wishes to\n issue a new ticket to the client using the NewSessionTicket handshake\n message. The server MUST NOT send this extension if it does not\n receive one in the ClientHello.\n\n\n\n\nSalowey, et al. Standards Track [Page 7]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n If the server fails to verify the ticket, then it falls back to\n performing a full handshake. If the ticket is accepted by the server\n but the handshake fails, the client SHOULD delete the ticket.\n\n The SessionTicket extension has been assigned the number 35. The\n extension_data field of SessionTicket extension contains the ticket.\n\n3.3. NewSessionTicket Handshake Message\n\n This message is sent by the server during the TLS handshake before\n the ChangeCipherSpec message. This message MUST be sent if the\n server included a SessionTicket extension in the ServerHello. This\n message MUST NOT be sent if the server did not include a\n SessionTicket extension in the ServerHello. This message is included\n in the hash used to create and verify the Finished message. In the\n case of a full handshake, the server MUST verify the client's\n Finished message before sending the ticket. The client MUST NOT\n treat the ticket as valid until it has verified the server's Finished\n message. If the server determines that it does not want to include a\n ticket after it has included the SessionTicket extension in the\n ServerHello, then it sends a zero-length ticket in the\n NewSessionTicket handshake message.\n\n If the server successfully verifies the client's ticket, then it MAY\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello in the abbreviated handshake. The client\n should start using the new ticket as soon as possible after it\n verifies the server's Finished message for new connections. Note\n that since the updated ticket is issued before the handshake\n completes, it is possible that the client may not put the new ticket\n into use before it initiates new connections. The server MUST NOT\n assume that the client actually received the updated ticket until it\n successfully verifies the client's Finished message.\n\n The NewSessionTicket handshake message has been assigned the number 4\n and its definition is given at the end of this section. The\n ticket_lifetime_hint field contains a hint from the server about how\n long the ticket should be stored. The value indicates the lifetime\n in seconds as a 32-bit unsigned integer in network byte order\n relative to when the ticket is received. A value of zero is reserved\n to indicate that the lifetime of the ticket is unspecified. A client\n SHOULD delete the ticket and associated state when the time expires.\n It MAY delete the ticket earlier based on local policy. A server MAY\n treat a ticket as valid for a shorter or longer period of time than\n what is stated in the ticket_lifetime_hint.\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 8]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n HandshakeType msg_type;\n uint24 length;\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case session_ticket: NewSessionTicket; /* NEW */\n } body;\n } Handshake;\n\n\n struct {\n uint32 ticket_lifetime_hint;\n opaque ticket<0..2^16-1>;\n } NewSessionTicket;\n\n3.4. Interaction with TLS Session ID\n\n If a server is planning on issuing a session ticket to a client that\n does not present one, it SHOULD include an empty Session ID in the\n ServerHello. If the server rejects the ticket and falls back to the\n full handshake then it may include a non-empty Session ID to indicate\n its support for stateful session resumption. If the client receives\n a session ticket from the server, then it discards any Session ID\n that was sent in the ServerHello.\n\n When presenting a ticket, the client MAY generate and include a\n Session ID in the TLS ClientHello. If the server accepts the ticket\n and the Session ID is not empty, then it MUST respond with the same\n Session ID present in the ClientHello. This allows the client to\n easily differentiate when the server is resuming a session from when\n it is falling back to a full handshake. Since the client generates a\n Session ID, the server MUST NOT rely upon the Session ID having a\n particular value when validating the ticket. If a ticket is\n presented by the client, the server MUST NOT attempt to use the\n Session ID in the ClientHello for stateful session resumption.\n Alternatively, the client MAY include an empty Session ID in the\n ClientHello. In this case, the client ignores the Session ID sent in\n the ServerHello and determines if the server is resuming a session by\n the subsequent handshake messages.\n\n\n\nSalowey, et al. Standards Track [Page 9]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n4. Recommended Ticket Construction\n\n This section describes a recommended format and protection for the\n ticket. Note that the ticket is opaque to the client, so the\n structure is not subject to interoperability concerns, and\n implementations may diverge from this format. If implementations do\n diverge from this format, they must take security concerns seriously.\n Clients MUST NOT examine the ticket under the assumption that it\n complies with this document.\n\n The server uses two different keys: one 128-bit key for Advanced\n Encryption Standard (AES) [AES] in Cipher Block Chaining (CBC) mode\n [CBC] encryption and one 256-bit key for HMAC-SHA-256 [RFC4634].\n\n The ticket is structured as follows:\n\n struct {\n opaque key_name[16];\n opaque iv[16];\n opaque encrypted_state<0..2^16-1>;\n opaque mac[32];\n } ticket;\n\n Here, key_name serves to identify a particular set of keys used to\n protect the ticket. It enables the server to easily recognize\n tickets it has issued. The key_name should be randomly generated to\n avoid collisions between servers. One possibility is to generate new\n random keys and key_name every time the server is started.\n\n The actual state information in encrypted_state is encrypted using\n 128-bit AES in CBC mode with the given IV. The Message\n Authentication Code (MAC) is calculated using HMAC-SHA-256 over\n key_name (16 octets) and IV (16 octets), followed by the length of\n the encrypted_state field (2 octets) and its contents (variable\n length).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 10]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n ProtocolVersion protocol_version;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n opaque master_secret[48];\n ClientIdentity client_identity;\n uint32 timestamp;\n } StatePlaintext;\n\n enum {\n anonymous(0),\n certificate_based(1),\n psk(2)\n } ClientAuthenticationType;\n\n struct {\n ClientAuthenticationType client_authentication_type;\n select (ClientAuthenticationType) {\n case anonymous: struct {};\n case certificate_based:\n ASN.1Cert certificate_list<0..2^24-1>;\n case psk:\n opaque psk_identity<0..2^16-1>; /* from [RFC4279] */\n };\n } ClientIdentity;\n\n The structure StatePlaintext stores the TLS session state including\n the master_secret. The timestamp within this structure allows the\n TLS server to expire tickets. To cover the authentication and key\n exchange protocols provided by TLS, the ClientIdentity structure\n contains the authentication type of the client used in the initial\n exchange (see ClientAuthenticationType). To offer the TLS server\n with the same capabilities for authentication and authorization, a\n certificate list is included in case of public-key-based\n authentication. The TLS server is therefore able to inspect a number\n of different attributes within these certificates. A specific\n implementation might choose to store a subset of this information or\n additional information. Other authentication mechanisms, such as\n Kerberos [RFC2712], would require different client identity data.\n Other TLS extensions may require the inclusion of additional data in\n the StatePlaintext structure.\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 11]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5. Security Considerations\n\n This section addresses security issues related to the usage of a\n ticket. Tickets must be authenticated and encrypted to prevent\n modification or eavesdropping by an attacker. Several attacks\n described below will be possible if this is not carefully done.\n\n Implementations should take care to ensure that the processing of\n tickets does not increase the chance of denial of service as\n described below.\n\n5.1. Invalidating Sessions\n\n The TLS specification requires that TLS sessions be invalidated when\n errors occur. [CSSC] discusses the security implications of this in\n detail. In the analysis within this paper, failure to invalidate\n sessions does not pose a security risk. This is because the TLS\n handshake uses a non-reversible function to derive keys for a session\n so information about one session does not provide an advantage to\n attack the master secret or a different session. If a session\n invalidation scheme is used, the implementation should verify the\n integrity of the ticket before using the contents to invalidate a\n session to ensure that an attacker cannot invalidate a chosen\n session.\n\n5.2. Stolen Tickets\n\n An eavesdropper or man-in-the-middle may obtain the ticket and\n attempt to use it to establish a session with the server; however,\n since the ticket is encrypted and the attacker does not know the\n secret key, a stolen ticket does not help an attacker resume a\n session. A TLS server MUST use strong encryption and integrity\n protection for the ticket to prevent an attacker from using a brute\n force mechanism to obtain the ticket's contents.\n\n5.3. Forged Tickets\n\n A malicious user could forge or alter a ticket in order to resume a\n session, to extend its lifetime, to impersonate another user, or to\n gain additional privileges. This attack is not possible if the\n ticket is protected using a strong integrity protection algorithm\n such as a keyed HMAC-SHA-256.\n\n5.4. Denial of Service Attacks\n\n The key_name field defined in the recommended ticket format helps the\n server efficiently reject tickets that it did not issue. However, an\n adversary could store or generate a large number of tickets to send\n\n\n\nSalowey, et al. Standards Track [Page 12]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n to the TLS server for verification. To minimize the possibility of a\n denial of service, the verification of the ticket should be\n lightweight (e.g., using efficient symmetric key cryptographic\n algorithms).\n\n5.5. Ticket Protection Key Management\n\n A full description of the management of the keys used to protect the\n ticket is beyond the scope of this document. A list of RECOMMENDED\n practices is given below.\n\n o The keys should be generated securely following the randomness\n recommendations in [RFC4086].\n\n o The keys and cryptographic protection algorithms should be at\n least 128 bits in strength. Some ciphersuites and applications\n may require cryptographic protection greater than 128 bits in\n strength.\n\n o The keys should not be used for any purpose other than generating\n and verifying tickets.\n\n o The keys should be changed regularly.\n\n o The keys should be changed if the ticket format or cryptographic\n protection algorithms change.\n\n5.6. Ticket Lifetime\n\n The TLS server controls the lifetime of the ticket. Servers\n determine the acceptable lifetime based on the operational and\n security requirements of the environments in which they are deployed.\n The ticket lifetime may be longer than the 24-hour lifetime\n recommended in [RFC4346]. TLS clients may be given a hint of the\n lifetime of the ticket. Since the lifetime of a ticket may be\n unspecified, a client has its own local policy that determines when\n it discards tickets.\n\n5.7. Alternate Ticket Formats and Distribution Schemes\n\n If the ticket format or distribution scheme defined in this document\n is not used, then great care must be taken in analyzing the security\n of the solution. In particular, if confidential information, such as\n a secret key, is transferred to the client, it MUST be done using\n secure communication so as to prevent attackers from obtaining or\n modifying the key. Also, the ticket MUST have its integrity and\n confidentiality protected with strong cryptographic techniques to\n prevent a breach in the security of the system.\n\n\n\nSalowey, et al. Standards Track [Page 13]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5.8. Identity Privacy, Anonymity, and Unlinkability\n\n This document mandates that the content of the ticket is\n confidentiality protected in order to avoid leakage of its content,\n such as user-relevant information. As such, it prevents disclosure\n of potentially sensitive information carried within the ticket.\n\n The initial handshake exchange, which was used to obtain the ticket,\n might not provide identity confidentiality of the client based on the\n properties of TLS. Another relevant security threat is the ability\n for an on-path adversary to observe multiple TLS handshakes where the\n same ticket is used, therefore concluding they belong to the same\n communication endpoints. Application designers that use the ticket\n mechanism described in this document should consider that\n unlinkability [ANON] is not necessarily provided.\n\n While a full discussion of these topics is beyond the scope of this\n document, it should be noted that it is possible to issue a ticket\n using a TLS renegotiation handshake that occurs after a secure tunnel\n has been established by a previous handshake. This may help address\n some privacy and unlinkability issues in some environments.\n\n6. Acknowledgements\n\n The authors would like to thank the following people for their help\n with preparing and reviewing this document: Eric Rescorla, Mohamad\n Badra, Tim Dierks, Nelson Bolyard, Nancy Cam-Winget, David McGrew,\n Rob Dugal, Russ Housley, Amir Herzberg, Bernard Aboba, and members of\n the TLS working group.\n\n [CSSC] describes a solution that is very similar to the one described\n in this document and gives a detailed analysis of the security\n considerations involved. [RFC2712] describes a mechanism for using\n Kerberos [RFC4120] in TLS ciphersuites, which helped inspire the use\n of tickets to avoid server state. [RFC4851] makes use of a similar\n mechanism to avoid maintaining server state for the cryptographic\n tunnel. [SC97] also investigates the concept of stateless sessions.\n\n The authors would also like to thank Jan Nordqvist, who found the\n encoding error in RFC 4507, corrected by this document. In addition\n Nagendra Modadugu, Wan-Teh Chang, and Michael D'Errico provided\n useful feedback during the review of this document.\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 14]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n7. IANA Considerations\n\n IANA has assigned a TLS extension number of 35 to the SessionTicket\n TLS extension from the TLS registry of ExtensionType values defined\n in [RFC4366].\n\n IANA has assigned a TLS HandshakeType number 4 to the\n NewSessionTicket handshake type from the TLS registry of\n HandshakeType values defined in [RFC4346].\n\n This document does not require any actions or assignments from IANA.\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2246] Dierks, T. and C. Allen, \"The TLS Protocol Version 1.0\",\n RFC 2246, January 1999.\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346, April 2006.\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, April 2006.\n\n [RFC4507] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 4507, May 2006.\n\n8.2. Informative References\n\n [AES] National Institute of Standards and Technology, \"Advanced\n Encryption Standard (AES)\", Federal Information Processing\n Standards (FIPS) Publication 197, November 2001.\n\n [ANON] Pfitzmann, A. and M. Hansen, \"Anonymity, Unlinkability,\n Unobservability, Pseudonymity, and Identity Management - A\n Consolidated Proposal for Terminology\", http://\n dud.inf.tu-dresden.de/literatur/\n Anon_Terminology_v0.26-1.pdf Version 0.26, December 2005.\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 15]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n [CBC] National Institute of Standards and Technology,\n \"Recommendation for Block Cipher Modes of Operation -\n Methods and Techniques\", NIST Special Publication 800-38A,\n December 2001.\n\n [CSSC] Shacham, H., Boneh, D., and E. Rescorla, \"Client-side\n caching for TLS\", Transactions on Information and System\n Security (TISSEC) , Volume 7, Issue 4, November 2004.\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [RFC4086] Eastlake, D., Schiller, J., and S. Crocker, \"Randomness\n Requirements for Security\", BCP 106, RFC 4086, June 2005.\n\n [RFC4120] Neuman, C., Yu, T., Hartman, S., and K. Raeburn, \"The\n Kerberos Network Authentication Service (V5)\", RFC 4120,\n July 2005.\n\n [RFC4279] Eronen, P. and H. Tschofenig, \"Pre-Shared Key Ciphersuites\n for Transport Layer Security (TLS)\", RFC 4279,\n December 2005.\n\n [RFC4634] Eastlake, D. and T. Hansen, \"US Secure Hash Algorithms\n (SHA and HMAC-SHA)\", RFC 4634, July 2006.\n\n [RFC4851] Cam-Winget, N., McGrew, D., Salowey, J., and H. Zhou, \"The\n Flexible Authentication via Secure Tunneling Extensible\n Authentication Protocol Method (EAP-FAST)\", RFC 4851,\n May 2007.\n\n [SC97] Aura, T. and P. Nikander, \"Stateless Connections\",\n Proceedings of the First International Conference on\n Information and Communication Security (ICICS '97) , 1997.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 16]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAppendix A. Discussion of Changes to RFC 4507\n\n RFC 4507 [RFC4507] defines a mechanism to resume a TLS session\n without maintaining server side state by specifying an encrypted\n ticket that is maintained on the client. The client presents this\n ticket to the server in a SessionTicket hello extension. The\n encoding in RFC 4507 used the XDR style encoding specified in TLS\n [RFC4346].\n\n An error in the encoding caused the specification to differ from\n deployed implementations. At the time of this writing there are no\n known implementations that follow the encoding specified in RFC 4507.\n This update to RFC 4507 aligns the document with these currently\n deployed implementations.\n\n Erroneous encoding in RFC 4507 resulted in two length fields; one for\n the extension contents and one for the ticket itself. Hence, for a\n ticket that is 256 bytes long and begins with the hex value FF FF,\n the encoding of the extension would be as follows according to RFC\n 4507:\n\n 00 23 Ticket Extension type 35\n 01 02 Length of extension contents\n 01 00 Length of ticket\n FF FF .. .. Actual ticket\n\n The update proposed in this document reflects what implementations\n actually encode, namely it removes the redundant length field. So,\n for a ticket that is 256 bytes long and begins with the hex value FF\n FF, the encoding of the extension would be as follows according to\n this update:\n\n 00 23 Extension type 35\n 01 00 Length of extension contents (ticket)\n FF FF .. .. Actual ticket\n\n A server implemented according to RFC 4507 receiving a ticket\n extension from a client conforming to this document would interpret\n the first two bytes of the ticket as the length of this ticket. This\n will result in either an inconsistent length field or in the\n processing of a ticket missing the first two bytes. In the first\n case, the server should reject the request based on a malformed\n length. In the second case, the server should reject the ticket\n based on a malformed ticket, incorrect key version, or failed\n decryption. A server implementation based on this update receiving\n an RFC 4507 extension would interpret the first length field as the\n\n\n\n\n\nSalowey, et al. Standards Track [Page 17]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n length of the ticket and include the second two length bytes as the\n first bytes in the ticket, resulting in the ticket being rejected\n based on a malformed ticket, incorrect key version, or failed\n decryption.\n\n Note that the encoding of an empty SessionTicket extension was\n ambiguous in RFC 4507. An RFC 4507 implementation may have encoded\n it as:\n\n 00 23 Extension type 35\n 00 02 Length of extension contents\n 00 00 Length of ticket\n\n or it may have encoded it the same way as this update:\n\n 00 23 Extension type 35\n 00 00 Length of extension contents\n\n A server wishing to support RFC 4507 clients should respond to an\n empty SessionTicket extension encoded the same way as it received it.\n\n A server implementation can construct tickets such that it can detect\n an RFC 4507 implementation, if one existed, by including a cookie at\n the beginning of the tickets that can be differentiated from a valid\n length. For example, if an implementation constructed tickets to\n start with the hex values FF FF, then it could determine where the\n ticket begins and determine the length correctly from the type of\n length fields present.\n\n This document makes a few additional changes to RFC 4507 listed\n below.\n\n o Clarifying that the server can allow session resumption using a\n ticket without issuing a new ticket in Section 3.1.\n\n o Clarifying that the lifetime is relative to when the ticket is\n received in section 3.3.\n\n o Clarifying that the NewSessionTicket handshake message is included\n in the hash generated for the Finished messages in Section 3.3.\n\n o Clarifying the interaction with TLS Session ID in Section 3.4.\n\n o Recommending the use of SHA-256 for the integrity protection of\n the ticket in Section 4.\n\n o Clarifying that additional data can be included in the\n StatePlaintext structure in Section 4.\n\n\n\nSalowey, et al. Standards Track [Page 18]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAuthors' Addresses\n\n Joseph Salowey\n Cisco Systems\n 2901 3rd Ave\n Seattle, WA 98121\n US\n\n EMail: [redacted-email]\n\n\n Hao Zhou\n Cisco Systems\n 4125 Highlander Parkway\n Richfield, OH 44286\n US\n\n EMail: [redacted-email]\n\n\n Pasi Eronen\n Nokia Research Center\n P.O. Box 407\n FIN-00045 Nokia Group\n Finland\n\n EMail: [redacted-email]\n\n\n Hannes Tschofenig\n Nokia Siemens Networks\n Otto-Hahn-Ring 6\n Munich, Bayern 81739\n Germany\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 19]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nFull Copyright Statement\n\n Copyright (C) The IETF Trust (2008).\n\n This document is subject to the rights, licenses and restrictions\n contained in BCP 78, and except as set forth therein, the authors\n retain all their rights.\n\n This document and the information contained herein are provided on an\n \"AS IS\" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\n OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\n THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\n OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\n THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\n WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\n\nIntellectual Property\n\n The IETF takes no position regarding the validity or scope of any\n Intellectual Property Rights or other rights that might be claimed to\n pertain to the implementation or use of the technology described in\n this document or the extent to which any license under such rights\n might or might not be available; nor does it represent that it has\n made any independent effort to identify any such rights. Information\n on the procedures with respect to rights in RFC documents can be\n found in BCP 78 and BCP 79.\n\n Copies of IPR disclosures made to the IETF Secretariat and any\n assurances of licenses to be made available, or the result of an\n attempt made to obtain a general license or permission for the use of\n such proprietary rights by implementers or users of this\n specification can be obtained from the IETF on-line IPR repository at\n http://www.ietf.org/ipr.\n\n The IETF invites any interested party to bring to its attention any\n copyrights, patents or patent applications, or other proprietary\n rights that may cover technology that may be required to implement\n this standard. Please address the information to the IETF at\n ietf-ipr@ietf.org.\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 20]\n\n===== DOCUMENT =====\n\n\nRFC 8446 TLS August 2018\n\n\n Sam Scott\n Royal Holloway, University of London\n [redacted-email]\n\n Thomas Shrimpton\n University of Florida\n [redacted-email]\n\n Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n Brian Smith\n Independent\n [redacted-email]\n\n Brian Sniffen\n Akamai Technologies\n [redacted-email]\n\n Nick Sullivan\n Cloudflare Inc.\n [redacted-email]\n\n Bjoern Tackmann\n University of California, San Diego\n [redacted-email]\n\n Tim Taubert\n Mozilla\n [redacted-email]\n\n Martin Thomson\n Mozilla\n [redacted-email]\n\n Hannes Tschofenig\n Arm Limited\n [redacted-email]\n\n Sean Turner\n sn3rd\n [redacted-email]\n\n Steven Valdez\n Google\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 159]\n\f\nRFC 8446 TLS August 2018\n\n\n Filippo Valsorda\n Cloudflare Inc.\n [redacted-email]\n\n Thyla van der Merwe\n Royal Holloway, University of London\n [redacted-email]\n\n Victor Vasiliev\n Google\n [redacted-email]\n\n Hoeteck Wee\n Ecole Normale Superieure, Paris\n [redacted-email]\n\n Tom Weinstein\n\n David Wong\n NCC Group\n [redacted-email]\n\n Christopher A. Wood\n Apple Inc.\n [redacted-email]\n\n Tim Wright\n Vodafone\n [redacted-email]\n\n Peter Wu\n Independent\n [redacted-email]\n\n Kazu Yamamoto\n Internet Initiative Japan Inc.\n [redacted-email]\n\nAuthor's Address\n\n Eric Rescorla\n Mozilla\n\n Email: [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 160]\n\n===== DOCUMENT =====\n\nNetwork Working Group T. Dierks\nRequest for Comments: 5246 Independent\nObsoletes: 3268, 4346, 4366 E. Rescorla\nUpdates: 4492 RTFM, Inc.\nCategory: Standards Track August 2008\n\n\n The Transport Layer Security (TLS) Protocol\n Version 1.2\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document specifies Version 1.2 of the Transport Layer Security\n (TLS) protocol. The TLS protocol provides communications security\n over the Internet. The protocol allows client/server applications to\n communicate in a way that is designed to prevent eavesdropping,\n tampering, or message forgery.\n\nTable of Contents\n\n 1. Introduction ....................................................4\n 1.1. Requirements Terminology ...................................5\n 1.2. Major Differences from TLS 1.1 .............................5\n 2. Goals ...........................................................6\n 3. Goals of This Document ..........................................7\n 4. Presentation Language ...........................................7\n 4.1. Basic Block Size ...........................................7\n 4.2. Miscellaneous ..............................................8\n 4.3. Vectors ....................................................8\n 4.4. Numbers ....................................................9\n 4.5. Enumerateds ................................................9\n 4.6. Constructed Types .........................................10\n 4.6.1. Variants ...........................................10\n 4.7. Cryptographic Attributes ..................................12\n 4.8. Constants .................................................14\n 5. HMAC and the Pseudorandom Function .............................14\n 6. The TLS Record Protocol ........................................15\n 6.1. Connection States .........................................16\n 6.2. Record Layer ..............................................19\n 6.2.1. Fragmentation ......................................19\n\n\n\nDierks & Rescorla Standards Track [Page 1]\n\f\nRFC 5246 TLS August 2008\n\n\n 6.2.2. Record Compression and Decompression ...............20\n 6.2.3. Record Payload Protection ..........................21\n 6.2.3.1. Null or Standard Stream Cipher ............22\n 6.2.3.2. CBC Block Cipher ..........................22\n 6.2.3.3. AEAD Ciphers ..............................24\n 6.3. Key Calculation ...........................................25\n 7. The TLS Handshaking Protocols ..................................26\n 7.1. Change Cipher Spec Protocol ...............................27\n 7.2. Alert Protocol ............................................28\n 7.2.1. Closure Alerts .....................................29\n 7.2.2. Error Alerts .......................................30\n 7.3. Handshake Protocol Overview ...............................33\n 7.4. Handshake Protocol ........................................37\n 7.4.1. Hello Messages .....................................38\n 7.4.1.1. Hello Request .............................38\n 7.4.1.2. Client Hello ..............................39\n 7.4.1.3. Server Hello ..............................42\n 7.4.1.4. Hello Extensions ..........................44\n 7.4.1.4.1. Signature Algorithms ...........45\n 7.4.2. Server Certificate .................................47\n 7.4.3. Server Key Exchange Message ........................50\n 7.4.4. Certificate Request ................................53\n 7.4.5. Server Hello Done ..................................55\n 7.4.6. Client Certificate .................................55\n 7.4.7. Client Key Exchange Message ........................57\n 7.4.7.1. RSA-Encrypted Premaster Secret Message ....58\n 7.4.7.2. Client Diffie-Hellman Public Value ........61\n 7.4.8. Certificate Verify .................................62\n 7.4.9. Finished ...........................................63\n 8. Cryptographic Computations .....................................64\n 8.1. Computing the Master Secret ...............................64\n 8.1.1. RSA ................................................65\n 8.1.2. Diffie-Hellman .....................................65\n 9. Mandatory Cipher Suites ........................................65\n 10. Application Data Protocol .....................................65\n 11. Security Considerations .......................................65\n 12. IANA Considerations ...........................................65\n Appendix A. Protocol Data Structures and Constant Values ..........68\n A.1. Record Layer ..............................................68\n A.2. Change Cipher Specs Message ...............................69\n A.3. Alert Messages ............................................69\n A.4. Handshake Protocol ........................................70\n A.4.1. Hello Messages .....................................71\n A.4.2. Server Authentication and Key Exchange Messages ....72\n A.4.3. Client Authentication and Key Exchange Messages ....74\n A.4.4. Handshake Finalization Message .....................74\n A.5. The Cipher Suite ..........................................75\n A.6. The Security Parameters ...................................77\n\n\n\nDierks & Rescorla Standards Track [Page 2]\n\f\nRFC 5246 TLS August 2008\n\n\n A.7. Changes to RFC 4492 .......................................78\n Appendix B. Glossary ..............................................78\n Appendix C. Cipher Suite Definitions ..............................83\n Appendix D. Implementation Notes ..................................85\n D.1. Random Number Generation and Seeding ......................85\n D.2. Certificates and Authentication ...........................85\n D.3. Cipher Suites .............................................85\n D.4. Implementation Pitfalls ...................................85\n Appendix E. Backward Compatibility ................................87\n E.1. Compatibility with TLS 1.0/1.1 and SSL 3.0 ................87\n E.2. Compatibility with SSL 2.0 ................................88\n E.3. Avoiding Man-in-the-Middle Version Rollback ...............90\n Appendix F. Security Analysis .....................................91\n F.1. Handshake Protocol ........................................91\n F.1.1. Authentication and Key Exchange ....................91\n F.1.1.1. Anonymous Key Exchange ....................91\n F.1.1.2. RSA Key Exchange and Authentication .......92\n F.1.1.3. Diffie-Hellman Key Exchange with\n Authentication ............................92\n F.1.2. Version Rollback Attacks ...........................93\n F.1.3. Detecting Attacks Against the Handshake Protocol ...94\n F.1.4. Resuming Sessions ..................................94\n F.2. Protecting Application Data ...............................94\n F.3. Explicit IVs ..............................................95\n F.4. Security of Composite Cipher Modes ........................95\n F.5. Denial of Service .........................................96\n F.6. Final Notes ...............................................96\n Normative References ..............................................97\n Informative References ............................................98\n Working Group Information ........................................101\n Contributors .....................................................101\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 3]\n\f\nRFC 5246 TLS August 2008\n\n\n1. Introduction\n\n The primary goal of the TLS protocol is to provide privacy and data\n integrity between two communicating applications. The protocol is\n composed of two layers: the TLS Record Protocol and the TLS Handshake\n Protocol. At the lowest level, layered on top of some reliable\n transport protocol (e.g., TCP [TCP]), is the TLS Record Protocol.\n The TLS Record Protocol provides connection security that has two\n basic properties:\n\n - The connection is private. Symmetric cryptography is used for\n data encryption (e.g., AES [AES], RC4 [SCH], etc.). The keys for\n this symmetric encryption are generated uniquely for each\n connection and are based on a secret negotiated by another\n protocol (such as the TLS Handshake Protocol). The Record\n Protocol can also be used without encryption.\n\n - The connection is reliable. Message transport includes a message\n integrity check using a keyed MAC. Secure hash functions (e.g.,\n SHA-1, etc.) are used for MAC computations. The Record Protocol\n can operate without a MAC, but is generally only used in this mode\n while another protocol is using the Record Protocol as a transport\n for negotiating security parameters.\n\n The TLS Record Protocol is used for encapsulation of various higher-\n level protocols. One such encapsulated protocol, the TLS Handshake\n Protocol, allows the server and client to authenticate each other and\n to negotiate an encryption algorithm and cryptographic keys before\n the application protocol transmits or receives its first byte of\n data. The TLS Handshake Protocol provides connection security that\n has three basic properties:\n\n - The peer's identity can be authenticated using asymmetric, or\n public key, cryptography (e.g., RSA [RSA], DSA [DSS], etc.). This\n authentication can be made optional, but is generally required for\n at least one of the peers.\n\n - The negotiation of a shared secret is secure: the negotiated\n secret is unavailable to eavesdroppers, and for any authenticated\n connection the secret cannot be obtained, even by an attacker who\n can place himself in the middle of the connection.\n\n - The negotiation is reliable: no attacker can modify the\n negotiation communication without being detected by the parties to\n the communication.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 4]\n\f\nRFC 5246 TLS August 2008\n\n\n One advantage of TLS is that it is application protocol independent.\n Higher-level protocols can layer on top of the TLS protocol\n transparently. The TLS standard, however, does not specify how\n protocols add security with TLS; the decisions on how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n1.1. Requirements Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [REQ].\n\n1.2. Major Differences from TLS 1.1\n\n This document is a revision of the TLS 1.1 [TLS1.1] protocol which\n contains improved flexibility, particularly for negotiation of\n cryptographic algorithms. The major changes are:\n\n - The MD5/SHA-1 combination in the pseudorandom function (PRF) has\n been replaced with cipher-suite-specified PRFs. All cipher suites\n in this document use P_SHA256.\n\n - The MD5/SHA-1 combination in the digitally-signed element has been\n replaced with a single hash. Signed elements now include a field\n that explicitly specifies the hash algorithm used.\n\n - Substantial cleanup to the client's and server's ability to\n specify which hash and signature algorithms they will accept.\n Note that this also relaxes some of the constraints on signature\n and hash algorithms from previous versions of TLS.\n\n - Addition of support for authenticated encryption with additional\n data modes.\n\n - TLS Extensions definition and AES Cipher Suites were merged in\n from external [TLSEXT] and [TLSAES].\n\n - Tighter checking of EncryptedPreMasterSecret version numbers.\n\n - Tightened up a number of requirements.\n\n - Verify_data length now depends on the cipher suite (default is\n still 12).\n\n - Cleaned up description of Bleichenbacher/Klima attack defenses.\n\n\n\n\nDierks & Rescorla Standards Track [Page 5]\n\f\nRFC 5246 TLS August 2008\n\n\n - Alerts MUST now be sent in many cases.\n\n - After a certificate_request, if no certificates are available,\n clients now MUST send an empty certificate list.\n\n - TLS_RSA_WITH_AES_128_CBC_SHA is now the mandatory to implement\n cipher suite.\n\n - Added HMAC-SHA256 cipher suites.\n\n - Removed IDEA and DES cipher suites. They are now deprecated and\n will be documented in a separate document.\n\n - Support for the SSLv2 backward-compatible hello is now a MAY, not\n a SHOULD, with sending it a SHOULD NOT. Support will probably\n become a SHOULD NOT in the future.\n\n - Added limited \"fall-through\" to the presentation language to allow\n multiple case arms to have the same encoding.\n\n - Added an Implementation Pitfalls sections\n\n - The usual clarifications and editorial work.\n\n2. Goals\n\n The goals of the TLS protocol, in order of priority, are as follows:\n\n 1. Cryptographic security: TLS should be used to establish a secure\n connection between two parties.\n\n 2. Interoperability: Independent programmers should be able to\n develop applications utilizing TLS that can successfully exchange\n cryptographic parameters without knowledge of one another's code.\n\n 3. Extensibility: TLS seeks to provide a framework into which new\n public key and bulk encryption methods can be incorporated as\n necessary. This will also accomplish two sub-goals: preventing\n the need to create a new protocol (and risking the introduction of\n possible new weaknesses) and avoiding the need to implement an\n entire new security library.\n\n 4. Relative efficiency: Cryptographic operations tend to be highly\n CPU intensive, particularly public key operations. For this\n reason, the TLS protocol has incorporated an optional session\n caching scheme to reduce the number of connections that need to be\n established from scratch. Additionally, care has been taken to\n reduce network activity.\n\n\n\nDierks & Rescorla Standards Track [Page 6]\n\f\nRFC 5246 TLS August 2008\n\n\n3. Goals of This Document\n\n This document and the TLS protocol itself are based on the SSL 3.0\n Protocol Specification as published by Netscape. The differences\n between this protocol and SSL 3.0 are not dramatic, but they are\n significant enough that the various versions of TLS and SSL 3.0 do\n not interoperate (although each protocol incorporates a mechanism by\n which an implementation can back down to prior versions). This\n document is intended primarily for readers who will be implementing\n the protocol and for those doing cryptographic analysis of it. The\n specification has been written with this in mind, and it is intended\n to reflect the needs of those two groups. For that reason, many of\n the algorithm-dependent data structures and rules are included in the\n body of the text (as opposed to in an appendix), providing easier\n access to them.\n\n This document is not intended to supply any details of service\n definition or of interface definition, although it does cover select\n areas of policy as they are required for the maintenance of solid\n security.\n\n4. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used. The syntax draws from\n several sources in its structure. Although it resembles the\n programming language \"C\" in its syntax and XDR [XDR] in both its\n syntax and intent, it would be risky to draw too many parallels. The\n purpose of this presentation language is to document TLS only; it has\n no general application beyond that particular goal.\n\n4.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 7]\n\f\nRFC 5246 TLS August 2008\n\n\n4.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" double\n brackets.\n\n Single-byte entities containing uninterpreted data are of type\n opaque.\n\n4.3. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* 3 consecutive 3 byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n In the following example, mandatory is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 4.4). On the\n other hand, longer can represent up to 800 bytes of data, or 400\n uint16 elements, and it may be empty. Its encoding will include a\n\n\n\nDierks & Rescorla Standards Track [Page 8]\n\f\nRFC 5246 TLS August 2008\n\n\n two-byte actual length field prepended to the vector. The length of\n an encoded vector must be an even multiple of the length of a single\n element (for example, a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is 2 bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n4.4. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are formed from fixed-length series of bytes\n concatenated as described in Section 4.1 and are also unsigned. The\n following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are stored in\n network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n Note that in some cases (e.g., DH parameters) it is necessary to\n represent integers as opaque vectors. In such cases, they are\n represented as unsigned integers (i.e., leading zero octets are not\n required even if the most significant bit is set).\n\n4.5. Enumerateds\n\n An additional sparse data type is available called enum. A field of\n type enum can only assume the values declared in the definition.\n Each definition is a different type. Only enumerateds of the same\n type may be assigned or compared. Every element of an enumerated\n must be assigned a value, as demonstrated in the following example.\n Since the elements of the enumerated are not ordered, they can be\n assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n\n\n\nDierks & Rescorla Standards Track [Page 9]\n\f\nRFC 5246 TLS August 2008\n\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n For enumerateds that are never converted to external representation,\n the numerical information may be omitted.\n\n enum { low, medium, high } Amount;\n\n4.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax for definition is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } [[T]];\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n Structure definitions may be embedded.\n\n4.6.1. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. There\n must be a case arm for every element of the enumeration declared in\n the select. Case arms have limited fall-through: if two case arms\n follow in immediate succession with no fields in between, then they\n\n\n\nDierks & Rescorla Standards Track [Page 10]\n\f\nRFC 5246 TLS August 2008\n\n\n both contain the same fields. Thus, in the example below, \"orange\"\n and \"banana\" both contain V2. Note that this is a new piece of\n syntax in TLS 1.2.\n\n The body of the variant structure may be given a label for reference.\n The mechanism by which the variant is selected at runtime is not\n prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1;\n case e2: Te2;\n case e3: case e4: Te3;\n ....\n case en: Ten;\n } [[fv]];\n } [[Tv]];\n\n For example:\n\n enum { apple, orange, banana } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n select (VariantTag) { /* value of selector is implicit */\n case apple:\n V1; /* VariantBody, tag = apple */\n case orange:\n case banana:\n V2; /* VariantBody, tag = orange or banana */\n } variant_body; /* optional label on variant */\n } VariantRecord;\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 11]\n\f\nRFC 5246 TLS August 2008\n\n\n4.7. Cryptographic Attributes\n\n The five cryptographic operations -- digital signing, stream cipher\n encryption, block cipher encryption, authenticated encryption with\n additional data (AEAD) encryption, and public key encryption -- are\n designated digitally-signed, stream-ciphered, block-ciphered, aead-\n ciphered, and public-key-encrypted, respectively. A field's\n cryptographic processing is specified by prepending an appropriate\n key word designation before the field's type specification.\n Cryptographic keys are implied by the current session state (see\n Section 6.1).\n\n A digitally-signed element is encoded as a struct DigitallySigned:\n\n struct {\n SignatureAndHashAlgorithm algorithm;\n opaque signature<0..2^16-1>;\n } DigitallySigned;\n\n The algorithm field specifies the algorithm used (see Section\n 7.4.1.4.1 for the definition of this field). Note that the\n introduction of the algorithm field is a change from previous\n versions. The signature is a digital signature using those\n algorithms over the contents of the element. The contents themselves\n do not appear on the wire but are simply calculated. The length of\n the signature is specified by the signing algorithm and key.\n\n In RSA signing, the opaque vector contains the signature generated\n using the RSASSA-PKCS1-v1_5 signature scheme defined in [PKCS1]. As\n discussed in [PKCS1], the DigestInfo MUST be DER-encoded [X680]\n [X690]. For hash algorithms without parameters (which includes\n SHA-1), the DigestInfo.AlgorithmIdentifier.parameters field MUST be\n NULL, but implementations MUST accept both without parameters and\n with NULL parameters. Note that earlier versions of TLS used a\n different RSA signature scheme that did not include a DigestInfo\n encoding.\n\n In DSA, the 20 bytes of the SHA-1 hash are run directly through the\n Digital Signing Algorithm with no additional hashing. This produces\n two values, r and s. The DSA signature is an opaque vector, as\n above, the contents of which are the DER encoding of:\n\n Dss-Sig-Value ::= SEQUENCE {\n r INTEGER,\n s INTEGER\n }\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 12]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: In current terminology, DSA refers to the Digital Signature\n Algorithm and DSS refers to the NIST standard. In the original SSL\n and TLS specs, \"DSS\" was used universally. This document uses \"DSA\"\n to refer to the algorithm, \"DSS\" to refer to the standard, and it\n uses \"DSS\" in the code point definitions for historical continuity.\n\n In stream cipher encryption, the plaintext is exclusive-ORed with an\n identical amount of output generated from a cryptographically secure\n keyed pseudorandom number generator.\n\n In block cipher encryption, every block of plaintext encrypts to a\n block of ciphertext. All block cipher encryption is done in CBC\n (Cipher Block Chaining) mode, and all items that are block-ciphered\n will be an exact multiple of the cipher block length.\n\n In AEAD encryption, the plaintext is simultaneously encrypted and\n integrity protected. The input may be of any length, and aead-\n ciphered output is generally larger than the input in order to\n accommodate the integrity check value.\n\n In public key encryption, a public key algorithm is used to encrypt\n data in such a way that it can be decrypted only with the matching\n private key. A public-key-encrypted element is encoded as an opaque\n vector <0..2^16-1>, where the length is specified by the encryption\n algorithm and key.\n\n RSA encryption is done using the RSAES-PKCS1-v1_5 encryption scheme\n defined in [PKCS1].\n\n In the following example\n\n stream-ciphered struct {\n uint8 field1;\n uint8 field2;\n digitally-signed opaque {\n uint8 field3<0..255>;\n uint8 field4;\n };\n } UserType;\n\n The contents of the inner struct (field3 and field4) are used as\n input for the signature/hash algorithm, and then the entire structure\n is encrypted with a stream cipher. The length of this structure, in\n bytes, would be equal to two bytes for field1 and field2, plus two\n bytes for the signature and hash algorithm, plus two bytes for the\n length of the signature, plus the length of the output of the signing\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 13]\n\f\nRFC 5246 TLS August 2008\n\n\n algorithm. The length of the signature is known because the\n algorithm and key used for the signing are known prior to encoding or\n decoding this structure.\n\n4.8. Constants\n\n Typed constants can be defined for purposes of specification by\n declaring a symbol of the desired type and assigning values to it.\n\n Under-specified types (opaque, variable-length vectors, and\n structures that contain opaque) cannot be assigned values. No fields\n of a multi-element structure or vector may be elided.\n\n For example:\n\n struct {\n uint8 f1;\n uint8 f2;\n } Example1;\n\n Example1 ex1 = {1, 4}; /* assigns f1 = 1, f2 = 4 */\n\n5. HMAC and the Pseudorandom Function\n\n The TLS record layer uses a keyed Message Authentication Code (MAC)\n to protect message integrity. The cipher suites defined in this\n document use a construction known as HMAC, described in [HMAC], which\n is based on a hash function. Other cipher suites MAY define their\n own MAC constructions, if needed.\n\n In addition, a construction is required to do expansion of secrets\n into blocks of data for the purposes of key generation or validation.\n This pseudorandom function (PRF) takes as input a secret, a seed, and\n an identifying label and produces an output of arbitrary length.\n\n In this section, we define one PRF, based on HMAC. This PRF with the\n SHA-256 hash function is used for all cipher suites defined in this\n document and in TLS documents published prior to this document when\n TLS 1.2 is negotiated. New cipher suites MUST explicitly specify a\n PRF and, in general, SHOULD use the TLS PRF with SHA-256 or a\n stronger standard hash function.\n\n First, we define a data expansion function, P_hash(secret, data),\n that uses a single hash function to expand a secret and seed into an\n arbitrary quantity of output:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 14]\n\f\nRFC 5246 TLS August 2008\n\n\n P_hash(secret, seed) = HMAC_hash(secret, A(1) + seed) +\n HMAC_hash(secret, A(2) + seed) +\n HMAC_hash(secret, A(3) + seed) + ...\n\n where + indicates concatenation.\n\n A() is defined as:\n\n A(0) = seed\n A(i) = HMAC_hash(secret, A(i-1))\n\n P_hash can be iterated as many times as necessary to produce the\n required quantity of data. For example, if P_SHA256 is being used to\n create 80 bytes of data, it will have to be iterated three times\n (through A(3)), creating 96 bytes of output data; the last 16 bytes\n of the final iteration will then be discarded, leaving 80 bytes of\n output data.\n\n TLS's PRF is created by applying P_hash to the secret as:\n\n PRF(secret, label, seed) = P_(secret, label + seed)\n\n The label is an ASCII string. It should be included in the exact\n form it is given without a length byte or trailing null character.\n For example, the label \"slithy toves\" would be processed by hashing\n the following bytes:\n\n 73 6C 69 74 68 79 20 74 6F 76 65 73\n\n6. The TLS Record Protocol\n\n The TLS Record Protocol is a layered protocol. At each layer,\n messages may include fields for length, description, and content.\n The Record Protocol takes messages to be transmitted, fragments the\n data into manageable blocks, optionally compresses the data, applies\n a MAC, encrypts, and transmits the result. Received data is\n decrypted, verified, decompressed, reassembled, and then delivered to\n higher-level clients.\n\n Four protocols that use the record protocol are described in this\n document: the handshake protocol, the alert protocol, the change\n cipher spec protocol, and the application data protocol. In order to\n allow extension of the TLS protocol, additional record content types\n can be supported by the record protocol. New record content type\n values are assigned by IANA in the TLS Content Type Registry as\n described in Section 12.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 15]\n\f\nRFC 5246 TLS August 2008\n\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST send an\n unexpected_message alert.\n\n Any protocol designed for use over TLS must be carefully designed to\n deal with all possible attacks against it. As a practical matter,\n this means that the protocol designer must be aware of what security\n properties TLS does and does not provide and cannot safely rely on\n the latter.\n\n Note in particular that type and length of a record are not protected\n by encryption. If this information is itself sensitive, application\n designers may wish to take steps (padding, cover traffic) to minimize\n information leakage.\n\n6.1. Connection States\n\n\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 8446 Mozilla\nObsoletes: 5077, 5246, 6961 August 2018\nUpdates: 5705, 6066\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS) Protocol Version 1.3\n\nAbstract\n\n TLS allows client/server applications to communicate\n over the Internet in a way that is designed to prevent eavesdropping,\n tampering, and message forgery.\n\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077,\n 5246, and 6961. This document also specifies new requirements for\n TLS 1.2 implementations.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 7841.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n https://www.rfc-editor.org/info/rfc8446.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 8446 TLS August 2018\n\n\nCopyright Notice\n\n Copyright (c) 2018 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (https://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 8446 TLS August 2018\n\n\nTable of Contents\n\n 1. Introduction ....................................................6\n 1.1. Conventions and Terminology ................................7\n 1.2. Major Differences from TLS 1.2 .............................8\n 1.3. Updates Affecting TLS 1.2 ..................................9\n 2. Protocol Overview ..............................................10\n 2.1. Incorrect DHE Share .......................................14\n 2.2. Resumption and Pre-Shared Key (PSK) .......................15\n 2.3. 0-RTT Data ................................................17\n 3. Presentation Language ..........................................19\n 3.1. Basic Block Size ..........................................19\n 3.2. Miscellaneous .............................................20\n 3.3. Numbers ...................................................20\n 3.4. Vectors ...................................................20\n 3.5. Enumerateds ...............................................21\n 3.6. Constructed Types .........................................22\n 3.7. Constants .................................................23\n 3.8. Variants ..................................................23\n 4. Handshake Protocol .............................................24\n 4.1. Key Exchange Messages .....................................25\n 4.1.1. Cryptographic Negotiation ..........................26\n 4.1.2. Client Hello .......................................27\n 4.1.3. Server Hello .......................................31\n 4.1.4. Hello Retry Request ................................33\n 4.2. Extensions ................................................35\n 4.2.1. Supported Versions .................................39\n 4.2.2. Cookie .............................................40\n 4.2.3. Signature Algorithms ...............................41\n 4.2.4. Certificate Authorities ............................45\n 4.2.5. OID Filters ........................................45\n 4.2.6. Post-Handshake Client Authentication ...............47\n 4.2.7. Supported Groups ...................................47\n 4.2.8. Key Share ..........................................48\n 4.2.9. Pre-Shared Key Exchange Modes ......................51\n 4.2.10. Early Data Indication .............................52\n 4.2.11. Pre-Shared Key Extension ..........................55\n 4.3. Server Parameters .........................................59\n 4.3.1. Encrypted Extensions ...............................60\n 4.3.2. Certificate Request ................................60\n 4.4. Authentication Messages ...................................61\n 4.4.1. The Transcript Hash ................................63\n 4.4.2. Certificate ........................................64\n 4.4.3. Certificate Verify .................................69\n 4.4.4. Finished ...........................................71\n 4.5. End of Early Data .........................................72\n\n\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 8446 TLS August 2018\n\n\n 4.6. Post-Handshake Messages ...................................73\n 4.6.1. New Session Ticket Message .........................73\n 4.6.2. Post-Handshake Authentication ......................75\n 4.6.3. Key and Initialization Vector Update ...............76\n 5. Record Protocol ................................................77\n 5.1. Record Layer ..............................................78\n 5.2. Record Payload Protection .................................80\n 5.3. Per-Record Nonce ..........................................82\n 5.4. Record Padding ............................................83\n 5.5. Limits on Key Usage .......................................84\n 6. Alert Protocol .................................................85\n 6.1. Closure Alerts ............................................87\n 6.2. Error Alerts ..............................................88\n 7. Cryptographic Computations .....................................90\n 7.1. Key Schedule ..............................................91\n 7.2. Updating Traffic Secrets ..................................94\n 7.3. Traffic Key Calculation ...................................95\n 7.4. (EC)DHE Shared Secret Calculation .........................95\n 7.4.1. Finite Field Diffie-Hellman ........................95\n 7.4.2. Elliptic Curve Diffie-Hellman ......................96\n 7.5. Exporters .................................................97\n 8. 0-RTT and Anti-Replay ..........................................98\n 8.1. Single-Use Tickets ........................................99\n 8.2. Client Hello Recording ....................................99\n 8.3. Freshness Checks .........................................101\n 9. Compliance Requirements .......................................102\n 9.1. Mandatory-to-Implement Cipher Suites .....................102\n 9.2. Mandatory-to-Implement Extensions ........................103\n 9.3. Protocol Invariants ......................................104\n 10. Security Considerations ......................................106\n 11. IANA Considerations ..........................................106\n 12. References ...................................................109\n 12.1. Normative References ....................................109\n 12.2. Informative References ..................................112\n Appendix A. State Machine ........................................120\n A.1. Client ....................................................120\n A.2. Server ....................................................121\n Appendix B. Protocol Data Structures and Constant Values .........122\n B.1. Record Layer ..............................................122\n B.2. Alert Messages ............................................123\n B.3. Handshake Protocol ........................................124\n B.3.1. Key Exchange Messages .................................125\n B.3.2. Server Parameters Messages ............................131\n B.3.3. Authentication Messages ...............................132\n B.3.4. Ticket Establishment ..................................132\n B.3.5. Updating Keys .........................................133\n B.4. Cipher Suites .............................................133\n\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 8446 TLS August 2018\n\n\n Appendix C. Implementation Notes .................................134\n C.1. Random Number Generation and Seeding ......................134\n C.2. Certificates and Authentication ...........................135\n C.3. Implementation Pitfalls ...................................135\n C.4. Client Tracking Prevention ................................137\n C.5. Unauthenticated Operation .................................137\n Appendix D. Backward Compatibility ...............................138\n D.1. Negotiating with an Older Server ..........................139\n D.2. Negotiating with an Older Client ..........................139\n D.3. 0-RTT Backward Compatibility ..............................140\n D.4. Middlebox Compatibility Mode ..............................140\n D.5. Security Restrictions Related to Backward Compatibility ...141\n Appendix E. Overview of Security Properties ......................142\n E.1. Handshake .................................................142\n E.1.1. Key Derivation and HKDF ...............................145\n E.1.2. Client Authentication .................................146\n E.1.3. 0-RTT .................................................146\n E.1.4. Exporter Independence .................................146\n E.1.5. Post-Compromise Security ..............................146\n E.1.6. External References ...................................147\n E.2. Record Layer ..............................................147\n E.2.1. External References ...................................148\n E.3. Traffic Analysis ..........................................148\n E.4. Side-Channel Attacks ......................................149\n E.5. Replay Attacks on 0-RTT ...................................150\n E.5.1. Replay and Exporters ..................................151\n E.6. PSK Identity Exposure .....................................152\n E.7. Sharing PSKs ..............................................152\n E.8. Attacks on Static RSA .....................................152\n Contributors .....................................................153\n Author's Address .................................................160\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 8446 TLS August 2018\n\n\n1. Introduction\n\n The primary goal of TLS is to provide a secure channel between two\n communicating peers; the only requirement from the underlying\n transport is a reliable, in-order data stream. Specifically, the\n secure channel should provide the following properties:\n\n - Authentication: The server side of the channel is always\n authenticated; the client side is optionally authenticated.\n Authentication can happen via asymmetric cryptography (e.g., RSA\n [RSA], the Elliptic Curve Digital Signature Algorithm (ECDSA)\n [ECDSA], or the Edwards-Curve Digital Signature Algorithm (EdDSA)\n [RFC8032]) or a symmetric pre-shared key (PSK).\n\n - Confidentiality: Data sent over the channel after establishment is\n only visible to the endpoints. TLS does not hide the length of\n the data it transmits, though endpoints are able to pad TLS\n records in order to obscure lengths and improve protection against\n traffic analysis techniques.\n\n - Integrity: Data sent over the channel after establishment cannot\n be modified by attackers without detection.\n\n These properties should be true even in the face of an attacker who\n has complete control of the network, as described in [RFC3552]. See\n Appendix E for a more complete statement of the relevant security\n properties.\n\n TLS consists of two primary components:\n\n - A handshake protocol (Section 4) that authenticates the\n communicating parties, negotiates cryptographic modes and\n parameters, and establishes shared keying material. The handshake\n protocol is designed to resist tampering; an active attacker\n should not be able to force the peers to negotiate different\n parameters than they would if the connection were not under\n attack.\n\n - A record protocol (Section 5) that uses the parameters established\n by the handshake protocol to protect traffic between the\n communicating peers. The record protocol divides traffic up into\n a series of records, each of which is independently protected\n using the traffic keys.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 8446 TLS August 2018\n\n\n TLS is application protocol independent; higher-level protocols can\n layer on top of TLS transparently. The TLS standard, however, does\n not specify how protocols add security with TLS; how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n This document defines TLS version 1.3. While TLS 1.3 is not directly\n compatible with previous versions, all versions of TLS incorporate a\n versioning mechanism which allows clients and servers to\n interoperably negotiate a common version if one is supported by both\n peers.\n\n This document supersedes and obsoletes previous versions of TLS,\n including version 1.2 [RFC5246]. It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2. Because TLS 1.3 changes the way keys are\n derived, it updates [RFC5705] as described in Section 7.5. It also\n changes how Online Certificate Status Protocol (OCSP) messages are\n carried and therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in Section 4.4.2.1.\n\n1.1. Conventions and Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all\n capitals, as shown here.\n\n The following terms are used:\n\n client: The endpoint initiating the TLS connection.\n\n connection: A transport-layer connection between two endpoints.\n\n endpoint: Either the client or server of the connection.\n\n handshake: An initial negotiation between client and server that\n establishes the parameters of their subsequent interactions\n within TLS.\n\n peer: An endpoint. When discussing a particular endpoint, \"peer\"\n refers to the endpoint that is not the primary subject of\n discussion.\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\f\nRFC 8446 TLS August 2018\n\n\n receiver: An endpoint that is receiving records.\n\n sender: An endpoint that is transmitting records.\n\n server: The endpoint that did not initiate the TLS connection.\n\n1.2. Major Differences from TLS 1.2\n\n The following is a list of the major functional differences between\n TLS 1.2 and TLS 1.3. It is not intended to be exhaustive, and there\n are many minor differences.\n\n - The list of supported symmetric encryption algorithms has been\n pruned of all algorithms that are considered legacy. Those that\n remain are all Authenticated Encryption with Associated Data\n (AEAD) algorithms. The cipher suite concept has been changed to\n separate the authentication and key exchange mechanisms from the\n record protection algorithm (including secret key length) and a\n hash to be used with both the key derivation function and\n handshake message authentication code (MAC).\n\n - A zero round-trip time (0-RTT) mode was added, saving a round trip\n at connection setup for some application data, at the cost of\n certain security properties.\n\n - Static RSA and Diffie-Hellman cipher suites have been removed; all\n public-key based key exchange mechanisms now provide forward\n secrecy.\n\n - All handshake messages after the ServerHello are now encrypted.\n The newly introduced EncryptedExtensions message allows various\n extensions previously sent in the clear in the ServerHello to also\n enjoy confidentiality protection.\n\n - The key derivation functions have been redesigned. The new design\n allows easier analysis by cryptographers due to their improved key\n separation properties. The HMAC-based Extract-and-Expand Key\n Derivation Function (HKDF) is used as an underlying primitive.\n\n - The handshake state machine has been significantly restructured to\n be more consistent and to remove superfluous messages such as\n ChangeCipherSpec (except when needed for middlebox compatibility).\n\n - Elliptic curve algorithms are now in the base spec, and new\n signature algorithms, such as EdDSA, are included. TLS 1.3\n removed point format negotiation in favor of a single point format\n for each curve.\n\n\n\n\nRescorla Standards Track [Page 8]\n\f\nRFC 8446 TLS August 2018\n\n\n - Other cryptographic improvements were made, including changing the\n RSA padding to use the RSA Probabilistic Signature Scheme\n (RSASSA-PSS), and the removal of compression, the Digital\n Signature Algorithm (DSA), and custom Ephemeral Diffie-Hellman\n (DHE) groups.\n\n - The TLS 1.2 version negotiation mechanism has been deprecated in\n favor of a version list in an extension. This increases\n compatibility with existing servers that incorrectly implemented\n version negotiation.\n\n - Session resumption with and without server-side state as well as\n the PSK-based cipher suites of earlier TLS versions have been\n replaced by a single new PSK exchange.\n\n - References have been updated to point to the updated versions of\n RFCs, as appropriate (e.g., RFC 5280 rather than RFC 3280).\n\n1.3. Updates Affecting TLS 1.2\n\n This document defines several changes that optionally affect\n implementations of TLS 1.2, including those which do not also support\n TLS 1.3:\n\n - A version downgrade protection mechanism is described in\n Section 4.1.3.\n\n - RSASSA-PSS signature schemes are defined in Section 4.2.3.\n\n - The \"supported_versions\" ClientHello extension can be used to\n negotiate the version of TLS to use, in preference to the\n legacy_version field of the ClientHello.\n\n - The \"signature_algorithms_cert\" extension allows a client to\n indicate which signature algorithms it can validate in X.509\n certificates.\n\n Additionally, this document clarifies some compliance requirements\n for earlier versions of TLS; see Section 9.3.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 9]\n\f\nRFC 8446 TLS August 2018\n\n\n2. Protocol Overview\n\n The cryptographic parameters used by the secure channel are produced\n by the TLS handshake protocol. This sub-protocol of TLS is used by\n the client and server when first communicating with each other. The\n handshake protocol allows peers to negotiate a protocol version,\n select cryptographic algorithms, optionally authenticate each other,\n and establish shared secret keying material. Once the handshake is\n complete, the peers use the established keys to protect the\n application-layer traffic.\n\n A failure of the handshake or other protocol error triggers the\n termination of the connection, optionally preceded by an alert\n message (Section 6).\n\n TLS supports three basic key exchange modes:\n\n - (EC)DHE (Diffie-Hellman over either finite fields or elliptic\n curves)\n\n - PSK-only\n\n - PSK with (EC)DHE\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 10]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 1 below shows the basic full TLS handshake:\n\n Client Server\n\nKey ^ ClientHello\nExch | + key_share*\n | + signature_algorithms*\n | + psk_key_exchange_modes*\n v + pre_shared_key* -------->\n ServerHello ^ Key\n + key_share* | Exch\n + pre_shared_key* v\n {EncryptedExtensions} ^ Server\n {CertificateRequest*} v Params\n {Certificate*} ^\n {CertificateVerify*} | Auth\n {Finished} v\n <-------- [Application Data*]\n ^ {Certificate*}\nAuth | {CertificateVerify*}\n v {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 1: Message Flow for Full TLS Handshake\n\n The handshake can be thought of as having three phases (indicated in\n the diagram above):\n\n - Key Exchange: Establish shared keying material and select the\n cryptographic parameters. Everything after this phase is\n encrypted.\n\n - Server Parameters: Establish other handshake parameters\n (whether the client is authenticated, application-layer protocol\n support, etc.).\n\n\n\n\nRescorla Standards Track [Page 11]\n\f\nRFC 8446 TLS August 2018\n\n\n - Authentication: Authenticate the server (and, optionally, the\n client) and provide key confirmation and handshake integrity.\n\n In the Key Exchange phase, the client sends the ClientHello\n (Section 4.1.2) message, which contains a random nonce\n (ClientHello.random); its offered protocol versions; a list of\n symmetric cipher/HKDF hash pairs; either a set of Diffie-Hellman key\n shares (in the \"key_share\" (Section 4.2.8) extension), a set of\n pre-shared key labels (in the \"pre_shared_key\" (Section 4.2.11)\n extension), or both; and potentially additional extensions.\n Additional fields and/or messages may also be present for middlebox\n compatibility.\n\n The server processes the ClientHello and determines the appropriate\n cryptographic parameters for the connection. It then responds with\n its own ServerHello (Section 4.1.3), which indicates the negotiated\n connection parameters. The combination of the ClientHello and the\n ServerHello determines the shared keys. If (EC)DHE key establishment\n is in use, then the ServerHello contains a \"key_share\" extension with\n the server's ephemeral Diffie-Hellman share; the server's share MUST\n be in the same group as one of the client's shares. If PSK key\n establishment is in use, then the ServerHello contains a\n \"pre_shared_key\" extension indicating which of the client's offered\n PSKs was selected. Note that implementations can use (EC)DHE and PSK\n together, in which case both extensions will be supplied.\n\n The server then sends two messages to establish the Server\n Parameters:\n\n EncryptedExtensions: responses to ClientHello extensions that are\n not required to determine the cryptographic parameters, other than\n those that are specific to individual certificates.\n [Section 4.3.1]\n\n CertificateRequest: if certificate-based client authentication is\n desired, the desired parameters for that certificate. This\n message is omitted if client authentication is not desired.\n [Section 4.3.2]\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 12]\n\f\nRFC 8446 TLS August 2018\n\n\n Finally, the client and server exchange Authentication messages. TLS\n uses the same set of messages every time that certificate-based\n authentication is needed. (PSK-based authentication happens as a\n side effect of key exchange.) Specifically:\n\n Certificate: The certificate of the endpoint and any per-certificate\n extensions. This message is omitted by the server if not\n authenticating with a certificate and by the client if the server\n did not send CertificateRequest (thus indicating that the client\n should not authenticate with a certificate). Note that if raw\n public keys [RFC7250] or the cached information extension\n [RFC7924] are in use, then this message will not contain a\n certificate but rather some other value corresponding to the\n server's long-term key. [Section 4.4.2]\n\n CertificateVerify: A signature over the entire handshake using the\n private key corresponding to the public key in the Certificate\n message. This message is omitted if the endpoint is not\n authenticating via a certificate. [Section 4.4.3]\n\n Finished: A MAC (Message Authentication Code) over the entire\n handshake. This message provides key confirmation, binds the\n endpoint's identity to the exchanged keys, and in PSK mode also\n authenticates the handshake. [Section 4.4.4]\n\n Upon receiving the server's messages, the client responds with its\n Authentication messages, namely Certificate and CertificateVerify (if\n requested), and Finished.\n\n At this point, the handshake is complete, and the client and server\n derive the keying material required by the record layer to exchange\n application-layer data protected through authenticated encryption.\n Application Data MUST NOT be sent prior to sending the Finished\n message, except as specified in Section 2.3. Note that while the\n server may send Application Data prior to receiving the client's\n Authentication messages, any data sent at that point is, of course,\n being sent to an unauthenticated peer.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 13]\n\f\nRFC 8446 TLS August 2018\n\n\n2.1. Incorrect DHE Share\n\n If the client has not provided a sufficient \"key_share\" extension\n (e.g., it includes only DHE or ECDHE groups unacceptable to or\n unsupported by the server), the server corrects the mismatch with a\n HelloRetryRequest and the client needs to restart the handshake with\n an appropriate \"key_share\" extension, as shown in Figure 2. If no\n common cryptographic parameters can be negotiated, the server MUST\n abort the handshake with an appropriate alert.\n\n Client Server\n\n ClientHello\n + key_share -------->\n HelloRetryRequest\n <-------- + key_share\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 2: Message Flow for a Full Handshake with\n Mismatched Parameters\n\n Note: The handshake transcript incorporates the initial\n ClientHello/HelloRetryRequest exchange; it is not reset with the\n new ClientHello.\n\n TLS also allows several optimized variants of the basic handshake, as\n described in the following sections.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 14]\n\f\nRFC 8446 TLS August 2018\n\n\n2.2. Resumption and Pre-Shared Key (PSK)\n\n Although TLS PSKs can be established out of band, PSKs can also be\n established in a previous connection and then used to establish a new\n connection (\"session resumption\" or \"resuming\" with a PSK). Once a\n handshake has completed, the server can send the client a PSK\n identity that corresponds to a unique key derived from the initial\n handshake (see Section 4.6.1). The client can then use that PSK\n identity in future handshakes to negotiate the use of the associated\n PSK. If the server accepts the PSK, then the security context of the\n new connection is cryptographically tied to the original connection\n and the key derived from the initial handshake is used to bootstrap\n the cryptographic state instead of a full handshake. In TLS 1.2 and\n below, this functionality was provided by \"session IDs\" and \"session\n tickets\" [RFC5077]. Both mechanisms are obsoleted in TLS 1.3.\n\n PSKs can be used with (EC)DHE key exchange in order to provide\n forward secrecy in combination with shared keys, or can be used\n alone, at the cost of losing forward secrecy for the application\n data.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 15]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 3 shows a pair of handshakes in which the first handshake\n establishes a PSK and the second handshake uses it:\n\n Client Server\n\n Initial Handshake:\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n <-------- [NewSessionTicket]\n [Application Data] <-------> [Application Data]\n\n\n Subsequent Handshake:\n ClientHello\n + key_share*\n + pre_shared_key -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n {Finished}\n <-------- [Application Data*]\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 3: Message Flow for Resumption and PSK\n\n As the server is authenticating via a PSK, it does not send a\n Certificate or a CertificateVerify message. When a client offers\n resumption via a PSK, it SHOULD also supply a \"key_share\" extension\n to the server to allow the server to decline resumption and fall back\n to a full handshake, if needed. The server responds with a\n \"pre_shared_key\" extension to negotiate the use of PSK key\n establishment and can (as shown here) respond with a \"key_share\"\n extension to do (EC)DHE key establishment, thus providing forward\n secrecy.\n\n\n\n\nRescorla Standards Track [Page 16]\n\f\nRFC 8446 TLS August 2018\n\n\n When PSKs are provisioned out of band, the PSK identity and the KDF\n hash algorithm to be used with the PSK MUST also be provisioned.\n\n Note: When using an out-of-band provisioned pre-shared secret, a\n critical consideration is using sufficient entropy during the key\n generation, as discussed in [RFC4086]. Deriving a shared secret\n from a password or other low-entropy sources is not secure. A\n low-entropy secret, or password, is subject to dictionary attacks\n based on the PSK binder. The specified PSK authentication is not\n a strong password-based authenticated key exchange even when used\n with Diffie-Hellman key establishment. Specifically, it does not\n prevent an attacker that can observe the handshake from performing\n a brute-force attack on the password/pre-shared key.\n\n2.3. 0-RTT Data\n\n When clients and servers share a PSK (either obtained externally or\n via a previous handshake), TLS 1.3 allows clients to send data on the\n first flight (\"early data\"). The client uses the PSK to authenticate\n the server and to encrypt the early data.\n\n As shown in Figure 4, the 0-RTT data is just added to the 1-RTT\n handshake in the first flight. The rest of the handshake uses the\n same messages as for a 1-RTT handshake with PSK resumption.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 17]\n\f\nRFC 8446 TLS August 2018\n\n\n Client Server\n\n ClientHello\n + early_data\n + key_share*\n + psk_key_exchange_modes\n + pre_shared_key\n (Application Data*) -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n + early_data*\n {Finished}\n <-------- [Application Data*]\n (EndOfEarlyData)\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n () Indicates messages protected using keys\n derived from a client_early_traffic_secret.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 4: Message Flow for a 0-RTT Handshake\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 18]\n\f\nRFC 8446 TLS August 2018\n\n\n IMPORTANT NOTE: The security properties for 0-RTT data are weaker\n than those for other kinds of TLS data. Specifically:\n\n 1. This data is not forward secret, as it is encrypted solely under\n keys derived using the offered PSK.\n\n 2. There are no guarantees of non-replay between connections.\n Protection against replay for ordinary TLS 1.3 1-RTT data is\n provided via the server's Random value, but 0-RTT data does not\n depend on the ServerHello and therefore has weaker guarantees.\n This is especially relevant if the data is authenticated either\n with TLS client authentication or inside the application\n protocol. The same warnings apply to any use of the\n early_exporter_master_secret.\n\n\n\n===== DOCUMENT =====\n\n A TLS connection state is the operating environment of the TLS Record\n Protocol. It specifies a compression algorithm, an encryption\n algorithm, and a MAC algorithm. In addition, the parameters for\n these algorithms are known: the MAC key and the bulk encryption keys\n for the connection in both the read and the write directions.\n Logically, there are always four connection states outstanding: the\n current read and write states, and the pending read and write states.\n All records are processed under the current read and write states.\n The security parameters for the pending states can be set by the TLS\n Handshake Protocol, and the ChangeCipherSpec can selectively make\n either of the pending states current, in which case the appropriate\n current state is disposed of and replaced with the pending state; the\n pending state is then reinitialized to an empty state. It is illegal\n to make a state that has not been initialized with security\n parameters a current state. The initial current state always\n specifies that no encryption, compression, or MAC will be used.\n\n The security parameters for a TLS Connection read and write state are\n set by providing the following values:\n\n connection end\n Whether this entity is considered the \"client\" or the \"server\" in\n this connection.\n\n PRF algorithm\n An algorithm used to generate keys from the master secret (see\n Sections 5 and 6.3).\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 16]\n\f\nRFC 5246 TLS August 2008\n\n\n bulk encryption algorithm\n An algorithm to be used for bulk encryption. This specification\n includes the key size of this algorithm, whether it is a block,\n stream, or AEAD cipher, the block size of the cipher (if\n appropriate), and the lengths of explicit and implicit\n initialization vectors (or nonces).\n\n MAC algorithm\n An algorithm to be used for message authentication. This\n specification includes the size of the value returned by the MAC\n algorithm.\n\n compression algorithm\n An algorithm to be used for data compression. This specification\n must include all information the algorithm requires to do\n compression.\n\n master secret\n A 48-byte secret shared between the two peers in the connection.\n\n client random\n A 32-byte value provided by the client.\n\n server random\n A 32-byte value provided by the server.\n\n These parameters are defined in the presentation language as:\n\n enum { server, client } ConnectionEnd;\n\n enum { tls_prf_sha256 } PRFAlgorithm;\n\n enum { null, rc4, 3des, aes }\n BulkCipherAlgorithm;\n\n enum { stream, block, aead } CipherType;\n\n enum { null, hmac_md5, hmac_sha1, hmac_sha256,\n hmac_sha384, hmac_sha512} MACAlgorithm;\n\n enum { null(0), (255) } CompressionMethod;\n\n /* The algorithms specified in CompressionMethod, PRFAlgorithm,\n BulkCipherAlgorithm, and MACAlgorithm may be added to. */\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 17]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ConnectionEnd entity;\n PRFAlgorithm prf_algorithm;\n BulkCipherAlgorithm bulk_cipher_algorithm;\n CipherType cipher_type;\n uint8 enc_key_length;\n uint8 block_length;\n uint8 fixed_iv_length;\n uint8 record_iv_length;\n MACAlgorithm mac_algorithm;\n uint8 mac_length;\n uint8 mac_key_length;\n CompressionMethod compression_algorithm;\n opaque master_secret[48];\n opaque client_random[32];\n opaque server_random[32];\n } SecurityParameters;\n\n The record layer will use the security parameters to generate the\n following six items (some of which are not required by all ciphers,\n and are thus empty):\n\n client write MAC key\n server write MAC key\n client write encryption key\n server write encryption key\n client write IV\n server write IV\n\n The client write parameters are used by the server when receiving and\n processing records and vice versa. The algorithm used for generating\n these items from the security parameters is described in Section 6.3.\n\n Once the security parameters have been set and the keys have been\n generated, the connection states can be instantiated by making them\n the current states. These current states MUST be updated for each\n record processed. Each connection state includes the following\n elements:\n\n compression state\n The current state of the compression algorithm.\n\n cipher state\n The current state of the encryption algorithm. This will consist\n of the scheduled key for that connection. For stream ciphers,\n this will also contain whatever state information is necessary to\n allow the stream to continue to encrypt or decrypt data.\n\n\n\n\nDierks & Rescorla Standards Track [Page 18]\n\f\nRFC 5246 TLS August 2008\n\n\n MAC key\n The MAC key for this connection, as generated above.\n\n sequence number\n Each connection state contains a sequence number, which is\n maintained separately for read and write states. The sequence\n number MUST be set to zero whenever a connection state is made the\n active state. Sequence numbers are of type uint64 and may not\n exceed 2^64-1. Sequence numbers do not wrap. If a TLS\n implementation would need to wrap a sequence number, it must\n renegotiate instead. A sequence number is incremented after each\n record: specifically, the first record transmitted under a\n particular connection state MUST use sequence number 0.\n\n6.2. Record Layer\n\n The TLS record layer receives uninterpreted data from higher layers\n in non-empty blocks of arbitrary size.\n\n6.2.1. Fragmentation\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Client\n message boundaries are not preserved in the record layer (i.e.,\n multiple client messages of the same ContentType MAY be coalesced\n into a single TLSPlaintext record, or a single message MAY be\n fragmented across several records).\n\n struct {\n uint8 major;\n uint8 minor;\n } ProtocolVersion;\n\n enum {\n change_cipher_spec(20), alert(21), handshake(22),\n application_data(23), (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type\n The higher-level protocol used to process the enclosed fragment.\n\n\n\n\nDierks & Rescorla Standards Track [Page 19]\n\f\nRFC 5246 TLS August 2008\n\n\n version\n The version of the protocol being employed. This document\n describes TLS Version 1.2, which uses the version { 3, 3 }. The\n version value 3.3 is historical, deriving from the use of {3, 1}\n for TLS 1.0. (See Appendix A.1.) Note that a client that\n supports multiple versions of TLS may not know what version will\n be employed before it receives the ServerHello. See Appendix E\n for discussion about what record layer version number should be\n employed for ClientHello.\n\n length\n The length (in bytes) of the following TLSPlaintext.fragment. The\n length MUST NOT exceed 2^14.\n\n fragment\n The application data. This data is transparent and treated as an\n independent block to be dealt with by the higher-level protocol\n specified by the type field.\n\n Implementations MUST NOT send zero-length fragments of Handshake,\n Alert, or ChangeCipherSpec content types. Zero-length fragments of\n Application data MAY be sent as they are potentially useful as a\n traffic analysis countermeasure.\n\n Note: Data of different TLS record layer content types MAY be\n interleaved. Application data is generally of lower precedence for\n transmission than other content types. However, records MUST be\n delivered to the network in the same order as they are protected by\n the record layer. Recipients MUST receive and process interleaved\n application layer traffic during handshakes subsequent to the first\n one on a connection.\n\n6.2.2. Record Compression and Decompression\n\n All records are compressed using the compression algorithm defined in\n the current session state. There is always an active compression\n algorithm; however, initially it is defined as\n CompressionMethod.null. The compression algorithm translates a\n TLSPlaintext structure into a TLSCompressed structure. Compression\n functions are initialized with default state information whenever a\n connection state is made active. [RFC3749] describes compression\n algorithms for TLS.\n\n Compression must be lossless and may not increase the content length\n by more than 1024 bytes. If the decompression function encounters a\n TLSCompressed.fragment that would decompress to a length in excess of\n 2^14 bytes, it MUST report a fatal decompression failure error.\n\n\n\n\nDierks & Rescorla Standards Track [Page 20]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ContentType type; /* same as TLSPlaintext.type */\n ProtocolVersion version;/* same as TLSPlaintext.version */\n uint16 length;\n opaque fragment[TLSCompressed.length];\n } TLSCompressed;\n\n length\n The length (in bytes) of the following TLSCompressed.fragment.\n The length MUST NOT exceed 2^14 + 1024.\n\n fragment\n The compressed form of TLSPlaintext.fragment.\n\n Note: A CompressionMethod.null operation is an identity operation;\n no fields are altered.\n\n Implementation note: Decompression functions are responsible for\n ensuring that messages cannot cause internal buffer overflows.\n\n6.2.3. Record Payload Protection\n\n The encryption and MAC functions translate a TLSCompressed\n structure into a TLSCiphertext. The decryption functions reverse\n the process. The MAC of the record also includes a sequence\n number so that missing, extra, or repeated messages are\n detectable.\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n select (SecurityParameters.cipher_type) {\n case stream: GenericStreamCipher;\n case block: GenericBlockCipher;\n case aead: GenericAEADCipher;\n } fragment;\n } TLSCiphertext;\n\n type\n The type field is identical to TLSCompressed.type.\n\n version\n The version field is identical to TLSCompressed.version.\n\n length\n The length (in bytes) of the following TLSCiphertext.fragment.\n The length MUST NOT exceed 2^14 + 2048.\n\n\n\nDierks & Rescorla Standards Track [Page 21]\n\f\nRFC 5246 TLS August 2008\n\n\n fragment\n The encrypted form of TLSCompressed.fragment, with the MAC.\n\n6.2.3.1. Null or Standard Stream Cipher\n\n Stream ciphers (including BulkCipherAlgorithm.null; see Appendix A.6)\n convert TLSCompressed.fragment structures to and from stream\n TLSCiphertext.fragment structures.\n\n stream-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n } GenericStreamCipher;\n\n The MAC is generated as:\n\n MAC(MAC_write_key, seq_num +\n TLSCompressed.type +\n TLSCompressed.version +\n TLSCompressed.length +\n TLSCompressed.fragment);\n\n where \"+\" denotes concatenation.\n\n seq_num\n The sequence number for this record.\n\n MAC\n The MAC algorithm specified by SecurityParameters.mac_algorithm.\n\n Note that the MAC is computed before encryption. The stream cipher\n encrypts the entire block, including the MAC. For stream ciphers\n that do not use a synchronization vector (such as RC4), the stream\n cipher state from the end of one record is simply used on the\n subsequent packet. If the cipher suite is TLS_NULL_WITH_NULL_NULL,\n encryption consists of the identity operation (i.e., the data is not\n encrypted, and the MAC size is zero, implying that no MAC is used).\n For both null and stream ciphers, TLSCiphertext.length is\n TLSCompressed.length plus SecurityParameters.mac_length.\n\n6.2.3.2. CBC Block Cipher\n\n For block ciphers (such as 3DES or AES), the encryption and MAC\n functions convert TLSCompressed.fragment structures to and from block\n TLSCiphertext.fragment structures.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 22]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n opaque IV[SecurityParameters.record_iv_length];\n block-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n uint8 padding[GenericBlockCipher.padding_length];\n uint8 padding_length;\n };\n } GenericBlockCipher;\n\n The MAC is generated as described in Section 6.2.3.1.\n\n IV\n The Initialization Vector (IV) SHOULD be chosen at random, and\n MUST be unpredictable. Note that in versions of TLS prior to 1.1,\n there was no IV field, and the last ciphertext block of the\n previous record (the \"CBC residue\") was used as the IV. This was\n changed to prevent the attacks described in [CBCATT]. For block\n ciphers, the IV length is of length\n SecurityParameters.record_iv_length, which is equal to the\n SecurityParameters.block_size.\n\n padding\n Padding that is added to force the length of the plaintext to be\n an integral multiple of the block cipher's block length. The\n padding MAY be any length up to 255 bytes, as long as it results\n in the TLSCiphertext.length being an integral multiple of the\n block length. Lengths longer than necessary might be desirable to\n frustrate attacks on a protocol that are based on analysis of the\n lengths of exchanged messages. Each uint8 in the padding data\n vector MUST be filled with the padding length value. The receiver\n MUST check this padding and MUST use the bad_record_mac alert to\n indicate padding errors.\n\n padding_length\n The padding length MUST be such that the total size of the\n GenericBlockCipher structure is a multiple of the cipher's block\n length. Legal values range from zero to 255, inclusive. This\n length specifies the length of the padding field exclusive of the\n padding_length field itself.\n\n The encrypted data length (TLSCiphertext.length) is one more than the\n sum of SecurityParameters.block_length, TLSCompressed.length,\n SecurityParameters.mac_length, and padding_length.\n\n Example: If the block length is 8 bytes, the content length\n (TLSCompressed.length) is 61 bytes, and the MAC length is 20 bytes,\n then the length before padding is 82 bytes (this does not include the\n\n\n\nDierks & Rescorla Standards Track [Page 23]\n\f\nRFC 5246 TLS August 2008\n\n\n IV. Thus, the padding length modulo 8 must be equal to 6 in order to\n make the total length an even multiple of 8 bytes (the block length).\n The padding length can be 6, 14, 22, and so on, through 254. If the\n padding length were the minimum necessary, 6, the padding would be 6\n bytes, each containing the value 6. Thus, the last 8 octets of the\n GenericBlockCipher before block encryption would be xx 06 06 06 06 06\n 06 06, where xx is the last octet of the MAC.\n\n Note: With block ciphers in CBC mode (Cipher Block Chaining), it is\n critical that the entire plaintext of the record be known before any\n ciphertext is transmitted. Otherwise, it is possible for the\n attacker to mount the attack described in [CBCATT].\n\n Implementation note: Canvel et al. [CBCTIME] have demonstrated a\n timing attack on CBC padding based on the time required to compute\n the MAC. In order to defend against this attack, implementations\n MUST ensure that record processing time is essentially the same\n whether or not the padding is correct. In general, the best way to\n do this is to compute the MAC even if the padding is incorrect, and\n only then reject the packet. For instance, if the pad appears to be\n incorrect, the implementation might assume a zero-length pad and then\n compute the MAC. This leaves a small timing channel, since MAC\n performance depends to some extent on the size of the data fragment,\n but it is not believed to be large enough to be exploitable, due to\n the large block size of existing MACs and the small size of the\n timing signal.\n\n6.2.3.3. AEAD Ciphers\n\n For AEAD [AEAD] ciphers (such as [CCM] or [GCM]), the AEAD function\n converts TLSCompressed.fragment structures to and from AEAD\n TLSCiphertext.fragment structures.\n\n struct {\n opaque nonce_explicit[SecurityParameters.record_iv_length];\n aead-ciphered struct {\n opaque content[TLSCompressed.length];\n };\n } GenericAEADCipher;\n\n AEAD ciphers take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [AEAD]. The key is either the\n client_write_key or the server_write_key. No MAC key is used.\n\n Each AEAD cipher suite MUST specify how the nonce supplied to the\n AEAD operation is constructed, and what is the length of the\n GenericAEADCipher.nonce_explicit part. In many cases, it is\n\n\n\nDierks & Rescorla Standards Track [Page 24]\n\f\nRFC 5246 TLS August 2008\n\n\n appropriate to use the partially implicit nonce technique described\n in Section 3.2.1 of [AEAD]; with record_iv_length being the length of\n the explicit part. In this case, the implicit part SHOULD be derived\n from key_block as client_write_iv and server_write_iv (as described\n in Section 6.3), and the explicit part is included in\n GenericAEAEDCipher.nonce_explicit.\n\n The plaintext is the TLSCompressed.fragment.\n\n The additional authenticated data, which we denote as\n additional_data, is defined as follows:\n\n additional_data = seq_num + TLSCompressed.type +\n TLSCompressed.version + TLSCompressed.length;\n\n where \"+\" denotes concatenation.\n\n The aead_output consists of the ciphertext output by the AEAD\n encryption operation. The length will generally be larger than\n TLSCompressed.length, but by an amount that varies with the AEAD\n cipher. Since the ciphers might incorporate padding, the amount of\n overhead could vary with different TLSCompressed.length values. Each\n AEAD cipher MUST NOT produce an expansion of greater than 1024 bytes.\n Symbolically,\n\n AEADEncrypted = AEAD-Encrypt(write_key, nonce, plaintext,\n additional_data)\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, the \"additional_data\", and the AEADEncrypted value. The\n output is either the plaintext or an error indicating that the\n decryption failed. There is no separate integrity check. That is:\n\n TLSCompressed.fragment = AEAD-Decrypt(write_key, nonce,\n AEADEncrypted,\n additional_data)\n\n If the decryption fails, a fatal bad_record_mac alert MUST be\n generated.\n\n6.3. Key Calculation\n\n The Record Protocol requires an algorithm to generate keys required\n by the current connection state (see Appendix A.6) from the security\n parameters provided by the handshake protocol.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 25]\n\f\nRFC 5246 TLS August 2008\n\n\n The master secret is expanded into a sequence of secure bytes, which\n is then split to a client write MAC key, a server write MAC key, a\n client write encryption key, and a server write encryption key. Each\n of these is generated from the byte sequence in that order. Unused\n values are empty. Some AEAD ciphers may additionally require a\n client write IV and a server write IV (see Section 6.2.3.3).\n\n When keys and MAC keys are generated, the master secret is used as an\n entropy source.\n\n To generate the key material, compute\n\n key_block = PRF(SecurityParameters.master_secret,\n \"key expansion\",\n SecurityParameters.server_random +\n SecurityParameters.client_random);\n\n until enough output has been generated. Then, the key_block is\n partitioned as follows:\n\n client_write_MAC_key[SecurityParameters.mac_key_length]\n server_write_MAC_key[SecurityParameters.mac_key_length]\n client_write_key[SecurityParameters.enc_key_length]\n server_write_key[SecurityParameters.enc_key_length]\n client_write_IV[SecurityParameters.fixed_iv_length]\n server_write_IV[SecurityParameters.fixed_iv_length]\n\n Currently, the client_write_IV and server_write_IV are only generated\n for implicit nonce techniques as described in Section 3.2.1 of\n [AEAD].\n\n Implementation note: The currently defined cipher suite which\n requires the most material is AES_256_CBC_SHA256. It requires 2 x 32\n byte keys and 2 x 32 byte MAC keys, for a total 128 bytes of key\n material.\n\n7. The TLS Handshaking Protocols\n\n TLS has three subprotocols that are used to allow peers to agree upon\n security parameters for the record layer, to authenticate themselves,\n to instantiate negotiated security parameters, and to report error\n conditions to each other.\n\n The Handshake Protocol is responsible for negotiating a session,\n which consists of the following items:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 26]\n\f\nRFC 5246 TLS August 2008\n\n\n session identifier\n An arbitrary byte sequence chosen by the server to identify an\n active or resumable session state.\n\n peer certificate\n X509v3 [PKIX] certificate of the peer. This element of the state\n may be null.\n\n compression method\n The algorithm used to compress data prior to encryption.\n\n cipher spec\n Specifies the pseudorandom function (PRF) used to generate keying\n material, the bulk data encryption algorithm (such as null, AES,\n etc.) and the MAC algorithm (such as HMAC-SHA1). It also defines\n cryptographic attributes such as the mac_length. (See Appendix\n A.6 for formal definition.)\n\n master secret\n 48-byte secret shared between the client and server.\n\n is resumable\n A flag indicating whether the session can be used to initiate new\n connections.\n\n These items are then used to create security parameters for use by\n the record layer when protecting application data. Many connections\n can be instantiated using the same session through the resumption\n feature of the TLS Handshake Protocol.\n\n7.1. Change Cipher Spec Protocol\n\n The change cipher spec protocol exists to signal transitions in\n ciphering strategies. The protocol consists of a single message,\n which is encrypted and compressed under the current (not the pending)\n connection state. The message consists of a single byte of value 1.\n\n struct {\n enum { change_cipher_spec(1), (255) } type;\n } ChangeCipherSpec;\n\n The ChangeCipherSpec message is sent by both the client and the\n server to notify the receiving party that subsequent records will be\n protected under the newly negotiated CipherSpec and keys. Reception\n of this message causes the receiver to instruct the record layer to\n immediately copy the read pending state into the read current state.\n Immediately after sending this message, the sender MUST instruct the\n record layer to make the write pending state the write active state.\n\n\n\nDierks & Rescorla Standards Track [Page 27]\n\f\nRFC 5246 TLS August 2008\n\n\n (See Section 6.1.) The ChangeCipherSpec message is sent during the\n handshake after the security parameters have been agreed upon, but\n before the verifying Finished message is sent.\n\n Note: If a rehandshake occurs while data is flowing on a connection,\n the communicating parties may continue to send data using the old\n CipherSpec. However, once the ChangeCipherSpec has been sent, the\n new CipherSpec MUST be used. The first side to send the\n ChangeCipherSpec does not know that the other side has finished\n computing the new keying material (e.g., if it has to perform a\n time-consuming public key operation). Thus, a small window of time,\n during which the recipient must buffer the data, MAY exist. In\n practice, with modern machines this interval is likely to be fairly\n short.\n\n7.2. Alert Protocol\n\n One of the content types supported by the TLS record layer is the\n alert type. Alert messages convey the severity of the message\n (warning or fatal) and a description of the alert. Alert messages\n with a level of fatal result in the immediate termination of the\n connection. In this case, other connections corresponding to the\n session may continue, but the session identifier MUST be invalidated,\n preventing the failed session from being used to establish new\n connections. Like other messages, alert messages are encrypted and\n compressed, as specified by the current connection state.\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n\n\n\nDierks & Rescorla Standards Track [Page 28]\n\f\nRFC 5246 TLS August 2008\n\n\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n7.2.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack. Either party may\n initiate the exchange of closing messages.\n\n close_notify\n This message notifies the recipient that the sender will not send\n any more messages on this connection. Note that as of TLS 1.1,\n failure to properly close a connection no longer requires that a\n session not be resumed. This is a change from TLS 1.0 to conform\n with widespread implementation practice.\n\n Either party may initiate a close by sending a close_notify alert.\n Any data received after a closure alert is ignored.\n\n Unless some other fatal alert has been transmitted, each party is\n required to send a close_notify alert before closing the write side\n of the connection. The other party MUST respond with a close_notify\n alert of its own and close down the connection immediately,\n discarding any pending writes. It is not required for the initiator\n of the close to wait for the responding close_notify alert before\n closing the read side of the connection.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation must receive the responding\n close_notify alert before indicating to the application layer that\n the TLS connection has ended. If the application protocol will not\n transfer any additional data, but will only close the underlying\n transport connection, then the implementation MAY choose to close the\n transport without waiting for the responding close_notify. No part\n\n\n\n\nDierks & Rescorla Standards Track [Page 29]\n\f\nRFC 5246 TLS August 2008\n\n\n of this standard should be taken to dictate the manner in which a\n usage profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing a connection reliably delivers\n pending data before destroying the transport.\n\n7.2.2. Error Alerts\n\n Error handling in the TLS Handshake protocol is very simple. When an\n error is detected, the detecting party sends a message to the other\n party. Upon transmission or receipt of a fatal alert message, both\n parties immediately close the connection. Servers and clients MUST\n forget any session-identifiers, keys, and secrets associated with a\n failed connection. Thus, any connection terminated with a fatal\n alert MUST NOT be resumed.\n\n Whenever an implementation encounters a condition which is defined as\n a fatal alert, it MUST send the appropriate alert prior to closing\n the connection. For all errors where an alert level is not\n explicitly specified, the sending party MAY determine at its\n discretion whether to treat this as a fatal error or not. If the\n implementation chooses to send an alert but intends to close the\n connection immediately afterwards, it MUST send that alert at the\n fatal alert level.\n\n If an alert with a level of warning is sent and received, generally\n the connection can continue normally. If the receiving party decides\n not to proceed with the connection (e.g., after having received a\n no_renegotiation alert that it is not willing to accept), it SHOULD\n send a fatal alert to terminate the connection. Given this, the\n sending party cannot, in general, know how the receiving party will\n behave. Therefore, warning alerts are not very useful when the\n sending party wants to continue the connection, and thus are\n sometimes omitted. For example, if a peer decides to accept an\n expired certificate (perhaps after confirming this with the user) and\n wants to continue the connection, it would not generally send a\n certificate_expired alert.\n\n The following error alerts are defined:\n\n unexpected_message\n An inappropriate message was received. This alert is always fatal\n and should never be observed in communication between proper\n implementations.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 30]\n\f\nRFC 5246 TLS August 2008\n\n\n bad_record_mac\n This alert is returned if a record is received with an incorrect\n MAC. This alert also MUST be returned if an alert is sent because\n a TLSCiphertext decrypted in an invalid way: either it wasn't an\n even multiple of the block length, or its padding values, when\n checked, weren't correct. This message is always fatal and should\n never be observed in communication between proper implementations\n (except when messages were corrupted in the network).\n\n decryption_failed_RESERVED\n This alert was used in some earlier versions of TLS, and may have\n permitted certain attacks against the CBC mode [CBCATT]. It MUST\n NOT be sent by compliant implementations.\n\n record_overflow\n A TLSCiphertext record was received that had a length more than\n 2^14+2048 bytes, or a record decrypted to a TLSCompressed record\n with more than 2^14+1024 bytes. This message is always fatal and\n should never be observed in communication between proper\n implementations (except when messages were corrupted in the\n network).\n\n decompression_failure\n The decompression function received improper input (e.g., data\n that would expand to excessive length). This message is always\n fatal and should never be observed in communication between proper\n implementations.\n\n handshake_failure\n Reception of a handshake_failure alert message indicates that the\n sender was unable to negotiate an acceptable set of security\n parameters given the options available. This is a fatal error.\n\n no_certificate_RESERVED\n This alert was used in SSLv3 but not any version of TLS. It MUST\n NOT be sent by compliant implementations.\n\n bad_certificate\n A certificate was corrupt, contained signatures that did not\n verify correctly, etc.\n\n unsupported_certificate\n A certificate was of an unsupported type.\n\n certificate_revoked\n A certificate was revoked by its signer.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 31]\n\f\nRFC 5246 TLS August 2008\n\n\n certificate_expired\n A certificate has expired or is not currently valid.\n\n certificate_unknown\n Some other (unspecified) issue arose in processing the\n certificate, rendering it unacceptable.\n\n illegal_parameter\n A field in the handshake was out of range or inconsistent with\n other fields. This message is always fatal.\n\n unknown_ca\n A valid certificate chain or partial chain was received, but the\n certificate was not accepted because the CA certificate could not\n be located or couldn't be matched with a known, trusted CA. This\n message is always fatal.\n\n access_denied\n A valid certificate was received, but when access control was\n applied, the sender decided not to proceed with negotiation. This\n message is always fatal.\n\n decode_error\n A message could not be decoded because some field was out of the\n specified range or the length of the message was incorrect. This\n message is always fatal and should never be observed in\n communication between proper implementations (except when messages\n were corrupted in the network).\n\n decrypt_error\n A handshake cryptographic operation failed, including being unable\n to correctly verify a signature or validate a Finished message.\n This message is always fatal.\n\n export_restriction_RESERVED\n This alert was used in some earlier versions of TLS. It MUST NOT\n be sent by compliant implementations.\n\n protocol_version\n The protocol version the client has attempted to negotiate is\n recognized but not supported. (For example, old protocol versions\n might be avoided for security reasons.) This message is always\n fatal.\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 32]\n\f\nRFC 5246 TLS August 2008\n\n\n insufficient_security\n Returned instead of handshake_failure when a negotiation has\n failed specifically because the server requires ciphers more\n secure than those supported by the client. This message is always\n fatal.\n\n internal_error\n An internal error unrelated to the peer or the correctness of the\n protocol (such as a memory allocation failure) makes it impossible\n to continue. This message is always fatal.\n\n\n\n===== DOCUMENT =====\n\nNetwork Working Group E. Rescorla\nInternet-Draft RTFM, Inc.\nObsoletes: 5077, 5246, 6961 (if March 20, 2018\n approved)\nUpdates: 4492, 5705, 6066 (if approved)\nIntended status: Standards Track\nExpires: September 21, 2018\n\n\n\n===== DOCUMENT =====\n\n user_canceled\n This handshake is being canceled for some reason unrelated to a\n protocol failure. If the user cancels an operation after the\n handshake is complete, just closing the connection by sending a\n close_notify is more appropriate. This alert should be followed\n by a close_notify. This message is generally a warning.\n\n no_renegotiation\n Sent by the client in response to a hello request or by the server\n in response to a client hello after initial handshaking. Either\n of these would normally lead to renegotiation; when that is not\n appropriate, the recipient should respond with this alert. At\n that point, the original requester can decide whether to proceed\n with the connection. One case where this would be appropriate is\n where a server has spawned a process to satisfy a request; the\n process might receive security parameters (key length,\n authentication, etc.) at startup, and it might be difficult to\n communicate changes to these parameters after that point. This\n message is always a warning.\n\n unsupported_extension\n sent by clients that receive an extended server hello containing\n an extension that they did not put in the corresponding client\n hello. This message is always fatal.\n\n New Alert values are assigned by IANA as described in Section 12.\n\n7.3. Handshake Protocol Overview\n\n The cryptographic parameters of the session state are produced by the\n TLS Handshake Protocol, which operates on top of the TLS record\n layer. When a TLS client and server first start communicating, they\n agree on a protocol version, select cryptographic algorithms,\n optionally authenticate each other, and use public-key encryption\n techniques to generate shared secrets.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 33]\n\f\nRFC 5246 TLS August 2008\n\n\n The TLS Handshake Protocol involves the following steps:\n\n - Exchange hello messages to agree on algorithms, exchange random\n values, and check for session resumption.\n\n - Exchange the necessary cryptographic parameters to allow the\n client and server to agree on a premaster secret.\n\n - Exchange certificates and cryptographic information to allow the\n client and server to authenticate themselves.\n\n - Generate a master secret from the premaster secret and exchanged\n random values.\n\n - Provide security parameters to the record layer.\n\n - Allow the client and server to verify that their peer has\n calculated the same security parameters and that the handshake\n occurred without tampering by an attacker.\n\n Note that higher layers should not be overly reliant on whether TLS\n always negotiates the strongest possible connection between two\n peers. There are a number of ways in which a man-in-the-middle\n attacker can attempt to make two entities drop down to the least\n secure method they support. The protocol has been designed to\n minimize this risk, but there are still attacks available: for\n example, an attacker could block access to the port a secure service\n runs on, or attempt to get the peers to negotiate an unauthenticated\n connection. The fundamental rule is that higher levels must be\n cognizant of what their security requirements are and never transmit\n information over a channel less secure than what they require. The\n TLS protocol is secure in that any cipher suite offers its promised\n level of security: if you negotiate 3DES with a 1024-bit RSA key\n exchange with a host whose certificate you have verified, you can\n expect to be that secure.\n\n These goals are achieved by the handshake protocol, which can be\n summarized as follows: The client sends a ClientHello message to\n which the server must respond with a ServerHello message, or else a\n fatal error will occur and the connection will fail. The ClientHello\n and ServerHello are used to establish security enhancement\n capabilities between client and server. The ClientHello and\n ServerHello establish the following attributes: Protocol Version,\n Session ID, Cipher Suite, and Compression Method. Additionally, two\n random values are generated and exchanged: ClientHello.random and\n ServerHello.random.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 34]\n\f\nRFC 5246 TLS August 2008\n\n\n The actual key exchange uses up to four messages: the server\n Certificate, the ServerKeyExchange, the client Certificate, and the\n ClientKeyExchange. New key exchange methods can be created by\n specifying a format for these messages and by defining the use of the\n messages to allow the client and server to agree upon a shared\n secret. This secret MUST be quite long; currently defined key\n exchange methods exchange secrets that range from 46 bytes upwards.\n\n Following the hello messages, the server will send its certificate in\n a Certificate message if it is to be authenticated. Additionally, a\n ServerKeyExchange message may be sent, if it is required (e.g., if\n the server has no certificate, or if its certificate is for signing\n only). If the server is authenticated, it may request a certificate\n from the client, if that is appropriate to the cipher suite selected.\n Next, the server will send the ServerHelloDone message, indicating\n that the hello-message phase of the handshake is complete. The\n server will then wait for a client response. If the server has sent\n a CertificateRequest message, the client MUST send the Certificate\n message. The ClientKeyExchange message is now sent, and the content\n of that message will depend on the public key algorithm selected\n between the ClientHello and the ServerHello. If the client has sent\n a certificate with signing ability, a digitally-signed\n CertificateVerify message is sent to explicitly verify possession of\n the private key in the certificate.\n\n At this point, a ChangeCipherSpec message is sent by the client, and\n the client copies the pending Cipher Spec into the current Cipher\n Spec. The client then immediately sends the Finished message under\n the new algorithms, keys, and secrets. In response, the server will\n send its own ChangeCipherSpec message, transfer the pending to the\n current Cipher Spec, and send its Finished message under the new\n Cipher Spec. At this point, the handshake is complete, and the\n client and server may begin to exchange application layer data. (See\n flow chart below.) Application data MUST NOT be sent prior to the\n completion of the first handshake (before a cipher suite other than\n TLS_NULL_WITH_NULL_NULL is established).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 35]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1. Message flow for a full handshake\n\n * Indicates optional or situation-dependent messages that are not\n always sent.\n\n Note: To help avoid pipeline stalls, ChangeCipherSpec is an\n independent TLS protocol content type, and is not actually a TLS\n handshake message.\n\n When the client and server decide to resume a previous session or\n duplicate an existing session (instead of negotiating new security\n parameters), the message flow is as follows:\n\n The client sends a ClientHello using the Session ID of the session to\n be resumed. The server then checks its session cache for a match.\n If a match is found, and the server is willing to re-establish the\n connection under the specified session state, it will send a\n ServerHello with the same Session ID value. At this point, both\n client and server MUST send ChangeCipherSpec messages and proceed\n directly to Finished messages. Once the re-establishment is\n complete, the client and server MAY begin to exchange application\n layer data. (See flow chart below.) If a Session ID match is not\n found, the server generates a new session ID, and the TLS client and\n server perform a full handshake.\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 36]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2. Message flow for an abbreviated handshake\n\n The contents and significance of each message will be presented in\n detail in the following sections.\n\n7.4. Handshake Protocol\n\n The TLS Handshake Protocol is one of the defined higher-level clients\n of the TLS Record Protocol. This protocol is used to negotiate the\n secure attributes of a session. Handshake messages are supplied to\n the TLS record layer, where they are encapsulated within one or more\n TLSPlaintext structures, which are processed and transmitted as\n specified by the current active session state.\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n } body;\n } Handshake;\n\n\n\nDierks & Rescorla Standards Track [Page 37]\n\f\nRFC 5246 TLS August 2008\n\n\n The handshake protocol messages are presented below in the order they\n MUST be sent; sending handshake messages in an unexpected order\n results in a fatal error. Unneeded handshake messages can be\n omitted, however. Note one exception to the ordering: the\n Certificate message is used twice in the handshake (from server to\n client, then from client to server), but described only in its first\n position. The one message that is not bound by these ordering rules\n is the HelloRequest message, which can be sent at any time, but which\n SHOULD be ignored by the client if it arrives in the middle of a\n handshake.\n\n New handshake message types are assigned by IANA as described in\n Section 12.\n\n7.4.1. Hello Messages\n\n The hello phase messages are used to exchange security enhancement\n capabilities between the client and server. When a new session\n begins, the record layer's connection state encryption, hash, and\n compression algorithms are initialized to null. The current\n connection state is used for renegotiation messages.\n\n7.4.1.1. Hello Request\n\n When this message will be sent:\n\n The HelloRequest message MAY be sent by the server at any time.\n\n Meaning of this message:\n\n HelloRequest is a simple notification that the client should begin\n the negotiation process anew. In response, the client should send\n a ClientHello message when convenient. This message is not\n intended to establish which side is the client or server but\n merely to initiate a new negotiation. Servers SHOULD NOT send a\n HelloRequest immediately upon the client's initial connection. It\n is the client's job to send a ClientHello at that time.\n\n This message will be ignored by the client if the client is\n currently negotiating a session. This message MAY be ignored by\n the client if it does not wish to renegotiate a session, or the\n client may, if it wishes, respond with a no_renegotiation alert.\n Since handshake messages are intended to have transmission\n precedence over application data, it is expected that the\n negotiation will begin before no more than a few records are\n received from the client. If the server sends a HelloRequest but\n does not receive a ClientHello in response, it may close the\n connection with a fatal alert.\n\n\n\nDierks & Rescorla Standards Track [Page 38]\n\f\nRFC 5246 TLS August 2008\n\n\n After sending a HelloRequest, servers SHOULD NOT repeat the\n request until the subsequent handshake negotiation is complete.\n\n Structure of this message:\n\n struct { } HelloRequest;\n\n This message MUST NOT be included in the message hashes that are\n maintained throughout the handshake and used in the Finished messages\n and the certificate verify message.\n\n7.4.1.2. Client Hello\n\n When this message will be sent:\n\n When a client first connects to a server, it is required to send\n the ClientHello as its first message. The client can also send a\n ClientHello in response to a HelloRequest or on its own initiative\n in order to renegotiate the security parameters in an existing\n connection.\n\n Structure of this message:\n\n The ClientHello message includes a random structure, which is used\n later in the protocol.\n\n struct {\n uint32 gmt_unix_time;\n opaque random_bytes[28];\n } Random;\n\n gmt_unix_time\n The current time and date in standard UNIX 32-bit format\n (seconds since the midnight starting Jan 1, 1970, UTC, ignoring\n leap seconds) according to the sender's internal clock. Clocks\n are not required to be set correctly by the basic TLS protocol;\n higher-level or application protocols may define additional\n requirements. Note that, for historical reasons, the data\n element is named using GMT, the predecessor of the current\n worldwide time base, UTC.\n\n random_bytes\n 28 bytes generated by a secure random number generator.\n\n The ClientHello message includes a variable-length session\n identifier. If not empty, the value identifies a session between the\n same client and server whose security parameters the client wishes to\n reuse. The session identifier MAY be from an earlier connection,\n\n\n\nDierks & Rescorla Standards Track [Page 39]\n\f\nRFC 5246 TLS August 2008\n\n\n this connection, or from another currently active connection. The\n second option is useful if the client only wishes to update the\n random structures and derived values of a connection, and the third\n option makes it possible to establish several independent secure\n connections without repeating the full handshake protocol. These\n independent connections may occur sequentially or simultaneously; a\n SessionID becomes valid when the handshake negotiating it completes\n with the exchange of Finished messages and persists until it is\n removed due to aging or because a fatal error was encountered on a\n connection associated with the session. The actual contents of the\n SessionID are defined by the server.\n\n opaque SessionID<0..32>;\n\n Warning: Because the SessionID is transmitted without encryption or\n immediate MAC protection, servers MUST NOT place confidential\n information in session identifiers or let the contents of fake\n session identifiers cause any breach of security. (Note that the\n content of the handshake as a whole, including the SessionID, is\n protected by the Finished messages exchanged at the end of the\n handshake.)\n\n The cipher suite list, passed from the client to the server in the\n ClientHello message, contains the combinations of cryptographic\n algorithms supported by the client in order of the client's\n preference (favorite choice first). Each cipher suite defines a key\n exchange algorithm, a bulk encryption algorithm (including secret key\n length), a MAC algorithm, and a PRF. The server will select a cipher\n suite or, if no acceptable choices are presented, return a handshake\n failure alert and close the connection. If the list contains cipher\n suites the server does not recognize, support, or wish to use, the\n server MUST ignore those cipher suites, and process the remaining\n ones as usual.\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n The ClientHello includes a list of compression algorithms supported\n by the client, ordered according to the client's preference.\n\n enum { null(0), (255) } CompressionMethod;\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 40]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ProtocolVersion client_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suites<2..2^16-2>;\n CompressionMethod compression_methods<1..2^8-1>;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ClientHello;\n\n TLS allows extensions to follow the compression_methods field in an\n extensions block. The presence of extensions can be detected by\n determining whether there are bytes following the compression_methods\n at the end of the ClientHello. Note that this method of detecting\n optional data differs from the normal TLS method of having a\n variable-length field, but it is used for compatibility with TLS\n before extensions were defined.\n\n client_version\n The version of the TLS protocol by which the client wishes to\n communicate during this session. This SHOULD be the latest\n (highest valued) version supported by the client. For this\n version of the specification, the version will be 3.3 (see\n Appendix E for details about backward compatibility).\n\n random\n A client-generated random structure.\n\n session_id\n The ID of a session the client wishes to use for this connection.\n This field is empty if no session_id is available, or if the\n client wishes to generate new security parameters.\n\n cipher_suites\n This is a list of the cryptographic options supported by the\n client, with the client's first preference first. If the\n session_id field is not empty (implying a session resumption\n request), this vector MUST include at least the cipher_suite from\n that session. Values are defined in Appendix A.5.\n\n compression_methods\n This is a list of the compression methods supported by the client,\n sorted by client preference. If the session_id field is not empty\n (implying a session resumption request), it MUST include the\n\n\n\nDierks & Rescorla Standards Track [Page 41]\n\f\nRFC 5246 TLS August 2008\n\n\n compression_method from that session. This vector MUST contain,\n and all implementations MUST support, CompressionMethod.null.\n Thus, a client and server will always be able to agree on a\n compression method.\n\n extensions\n Clients MAY request extended functionality from servers by sending\n data in the extensions field. The actual \"Extension\" format is\n defined in Section 7.4.1.4.\n\n In the event that a client requests additional functionality using\n extensions, and this functionality is not supplied by the server, the\n client MAY abort the handshake. A server MUST accept ClientHello\n messages both with and without the extensions field, and (as for all\n other messages) it MUST check that the amount of data in the message\n precisely matches one of these formats; if not, then it MUST send a\n fatal \"decode_error\" alert.\n\n After sending the ClientHello message, the client waits for a\n ServerHello message. Any handshake message returned by the server,\n except for a HelloRequest, is treated as a fatal error.\n\n7.4.1.3. Server Hello\n\n When this message will be sent:\n\n The server will send this message in response to a ClientHello\n message when it was able to find an acceptable set of algorithms.\n If it cannot find such a match, it will respond with a handshake\n failure alert.\n\n Structure of this message:\n\n struct {\n ProtocolVersion server_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ServerHello;\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 42]\n\f\nRFC 5246 TLS August 2008\n\n\n The presence of extensions can be detected by determining whether\n there are bytes following the compression_method field at the end of\n the ServerHello.\n\n server_version\n This field will contain the lower of that suggested by the client\n in the client hello and the highest supported by the server. For\n this version of the specification, the version is 3.3. (See\n Appendix E for details about backward compatibility.)\n\n random\n This structure is generated by the server and MUST be\n independently generated from the ClientHello.random.\n\n session_id\n This is the identity of the session corresponding to this\n connection. If the ClientHello.session_id was non-empty, the\n server will look in its session cache for a match. If a match is\n found and the server is willing to establish the new connection\n using the specified session state, the server will respond with\n the same value as was supplied by the client. This indicates a\n resumed session and dictates that the parties must proceed\n directly to the Finished messages. Otherwise, this field will\n contain a different value identifying the new session. The server\n may return an empty session_id to indicate that the session will\n not be cached and therefore cannot be resumed. If a session is\n resumed, it must be resumed using the same cipher suite it was\n originally negotiated with. Note that there is no requirement\n that the server resume any session even if it had formerly\n provided a session_id. Clients MUST be prepared to do a full\n negotiation -- including negotiating new cipher suites -- during\n any handshake.\n\n cipher_suite\n The single cipher suite selected by the server from the list in\n ClientHello.cipher_suites. For resumed sessions, this field is\n the value from the state of the session being resumed.\n\n compression_method\n The single compression algorithm selected by the server from the\n list in ClientHello.compression_methods. For resumed sessions,\n this field is the value from the resumed session state.\n\n extensions\n A list of extensions. Note that only extensions offered by the\n client can appear in the server's list.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 43]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.1.4. Hello Extensions\n\n The extension format is:\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n signature_algorithms(13), (65535)\n } ExtensionType;\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The initial set of extensions is defined in a companion document\n [TLSEXT]. The list of extension types is maintained by IANA as\n described in Section 12.\n\n An extension type MUST NOT appear in the ServerHello unless the same\n extension type appeared in the corresponding ClientHello. If a\n client receives an extension type in ServerHello that it did not\n request in the associated ClientHello, it MUST abort the handshake\n with an unsupported_extension fatal alert.\n\n Nonetheless, \"server-oriented\" extensions may be provided in the\n future within this framework. Such an extension (say, of type x)\n would require the client to first send an extension of type x in a\n ClientHello with empty extension_data to indicate that it supports\n the extension type. In this case, the client is offering the\n capability to understand the extension type, and the server is taking\n the client up on its offer.\n\n When multiple extensions of different types are present in the\n ClientHello or ServerHello messages, the extensions MAY appear in any\n order. There MUST NOT be more than one extension of the same type.\n\n Finally, note that extensions can be sent both when starting a new\n session and when requesting session resumption. Indeed, a client\n that requests session resumption does not in general know whether the\n server will accept this request, and therefore it SHOULD send the\n same extensions as it would send if it were not attempting\n resumption.\n\n\n\nDierks & Rescorla Standards Track [Page 44]\n\f\nRFC 5246 TLS August 2008\n\n\n In general, the specification of each extension type needs to\n describe the effect of the extension both during full handshake and\n session resumption. Most current TLS extensions are relevant only\n when a session is initiated: when an older session is resumed, the\n server does not process these extensions in Client Hello, and does\n not include them in Server Hello. However, some extensions may\n specify different behavior during session resumption.\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions, and some are simply refusals to support particular\n features. In general, error alerts should be used for the former,\n and a field in the server extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem.\n\n Often the fact that the extension fields are included in the\n inputs to the Finished message hashes will be sufficient, but\n extreme care is needed when the extension changes the meaning of\n messages sent in the handshake phase. Designers and implementors\n should be aware of the fact that until the handshake has been\n authenticated, active attackers can modify messages and insert,\n remove, or replace extensions.\n\n - It would be technically possible to use extensions to change major\n aspects of the design of TLS; for example the design of cipher\n suite negotiation. This is not recommended; it would be more\n appropriate to define a new version of TLS -- particularly since\n the TLS handshake algorithms have specific protection against\n version rollback attacks based on the version number, and the\n possibility of version rollback should be a significant\n consideration in any major design change.\n\n7.4.1.4.1. Signature Algorithms\n\n The client uses the \"signature_algorithms\" extension to indicate to\n the server which signature/hash algorithm pairs may be used in\n digital signatures. The \"extension_data\" field of this extension\n contains a \"supported_signature_algorithms\" value.\n\n\n\nDierks & Rescorla Standards Track [Page 45]\n\f\nRFC 5246 TLS August 2008\n\n\n enum {\n none(0), md5(1), sha1(2), sha224(3), sha256(4), sha384(5),\n sha512(6), (255)\n } HashAlgorithm;\n\n enum { anonymous(0), rsa(1), dsa(2), ecdsa(3), (255) }\n SignatureAlgorithm;\n\n struct {\n HashAlgorithm hash;\n SignatureAlgorithm signature;\n } SignatureAndHashAlgorithm;\n\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2..2^16-2>;\n\n Each SignatureAndHashAlgorithm value lists a single hash/signature\n pair that the client is willing to verify. The values are indicated\n in descending order of preference.\n\n Note: Because not all signature algorithms and hash algorithms may be\n accepted by an implementation (e.g., DSA with SHA-1, but not\n SHA-256), algorithms here are listed in pairs.\n\n hash\n This field indicates the hash algorithm which may be used. The\n values indicate support for unhashed data, MD5 [MD5], SHA-1,\n SHA-224, SHA-256, SHA-384, and SHA-512 [SHS], respectively. The\n \"none\" value is provided for future extensibility, in case of a\n signature algorithm which does not require hashing before signing.\n\n signature\n This field indicates the signature algorithm that may be used.\n The values indicate anonymous signatures, RSASSA-PKCS1-v1_5\n [PKCS1] and DSA [DSS], and ECDSA [ECDSA], respectively. The\n \"anonymous\" value is meaningless in this context but used in\n Section 7.4.3. It MUST NOT appear in this extension.\n\n The semantics of this extension are somewhat complicated because the\n cipher suite indicates permissible signature algorithms but not hash\n algorithms. Sections 7.4.2 and 7.4.3 describe the appropriate rules.\n\n If the client supports only the default hash and signature algorithms\n (listed in this section), it MAY omit the signature_algorithms\n extension. If the client does not support the default algorithms, or\n supports other hash and signature algorithms (and it is willing to\n use them for verifying messages sent by the server, i.e., server\n certificates and server key exchange), it MUST send the\n\n\n\nDierks & Rescorla Standards Track [Page 46]\n\f\nRFC 5246 TLS August 2008\n\n\n signature_algorithms extension, listing the algorithms it is willing\n to accept.\n\n If the client does not send the signature_algorithms extension, the\n server MUST do the following:\n\n - If the negotiated key exchange algorithm is one of (RSA, DHE_RSA,\n DH_RSA, RSA_PSK, ECDH_RSA, ECDHE_RSA), behave as if client had\n sent the value {sha1,rsa}.\n\n - If the negotiated key exchange algorithm is one of (DHE_DSS,\n DH_DSS), behave as if the client had sent the value {sha1,dsa}.\n\n - If the negotiated key exchange algorithm is one of (ECDH_ECDSA,\n ECDHE_ECDSA), behave as if the client had sent value {sha1,ecdsa}.\n\n Note: this is a change from TLS 1.1 where there are no explicit\n rules, but as a practical matter one can assume that the peer\n supports MD5 and SHA-1.\n\n Note: this extension is not meaningful for TLS versions prior to 1.2.\n Clients MUST NOT offer it if they are offering prior versions.\n However, even if clients do offer it, the rules specified in [TLSEXT]\n require servers to ignore extensions they do not understand.\n\n Servers MUST NOT send this extension. TLS servers MUST support\n receiving this extension.\n\n When performing session resumption, this extension is not included in\n Server Hello, and the server ignores the extension in Client Hello\n (if present).\n\n7.4.2. Server Certificate\n\n When this message will be sent:\n\n The server MUST send a Certificate message whenever the agreed-\n upon key exchange method uses certificates for authentication\n (this includes all key exchange methods defined in this document\n except DH_anon). This message will always immediately follow the\n ServerHello message.\n\n Meaning of this message:\n\n This message conveys the server's certificate chain to the client.\n\n The certificate MUST be appropriate for the negotiated cipher\n suite's key exchange algorithm and any negotiated extensions.\n\n\n\nDierks & Rescorla Standards Track [Page 47]\n\f\nRFC 5246 TLS August 2008\n\n\n Structure of this message:\n\n opaque ASN.1Cert<1..2^24-1>;\n\n struct {\n ASN.1Cert certificate_list<0..2^24-1>;\n } Certificate;\n\n certificate_list\n This is a sequence (chain) of certificates. The sender's\n certificate MUST come first in the list. Each following\n certificate MUST directly certify the one preceding it. Because\n certificate validation requires that root keys be distributed\n independently, the self-signed certificate that specifies the root\n certificate authority MAY be omitted from the chain, under the\n assumption that the remote end must already possess it in order to\n validate it in any case.\n\n The same message type and structure will be used for the client's\n response to a certificate request message. Note that a client MAY\n send no certificates if it does not have an appropriate certificate\n to send in response to the server's authentication request.\n\n Note: PKCS #7 [PKCS7] is not used as the format for the certificate\n vector because PKCS #6 [PKCS6] extended certificates are not used.\n Also, PKCS #7 defines a SET rather than a SEQUENCE, making the task\n of parsing the list more difficult.\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3, unless explicitly negotiated\n otherwise (e.g., [TLSPGP]).\n\n - The end entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected key exchange\n algorithm.\n\n Key Exchange Alg. Certificate Key Type\n\n RSA RSA public key; the certificate MUST allow the\n RSA_PSK key to be used for encryption (the\n keyEncipherment bit MUST be set if the key\n usage extension is present).\n Note: RSA_PSK is defined in [TLSPSK].\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 48]\n\f\nRFC 5246 TLS August 2008\n\n\n DHE_RSA RSA public key; the certificate MUST allow the\n ECDHE_RSA key to be used for signing (the\n digitalSignature bit MUST be set if the key\n usage extension is present) with the signature\n scheme and hash algorithm that will be employed\n in the server key exchange message.\n Note: ECDHE_RSA is defined in [TLSECC].\n\n DHE_DSS DSA public key; the certificate MUST allow the\n key to be used for signing with the hash\n algorithm that will be employed in the server\n key exchange message.\n\n DH_DSS Diffie-Hellman public key; the keyAgreement bit\n DH_RSA MUST be set if the key usage extension is\n present.\n\n ECDH_ECDSA ECDH-capable public key; the public key MUST\n ECDH_RSA use a curve and point format supported by the\n client, as described in [TLSECC].\n\n ECDHE_ECDSA ECDSA-capable public key; the certificate MUST\n allow the key to be used for signing with the\n hash algorithm that will be employed in the\n server key exchange message. The public key\n MUST use a curve and point format supported by\n the client, as described in [TLSECC].\n\n - The \"server_name\" and \"trusted_ca_keys\" extensions [TLSEXT] are\n used to guide certificate selection.\n\n\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) Y. Pettersen\nRequest for Comments: 6961 June 2013\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\n\nAbstract\n\n This document defines the Transport Layer Security (TLS) Certificate\n Status Version 2 Extension to allow clients to specify and support\n several certificate status methods. (The use of the Certificate\n Status extension is commonly referred to as \"OCSP stapling\".) Also\n defined is a new method based on the Online Certificate Status\n Protocol (OCSP) that servers can use to provide status information\n about not only the server's own certificate but also the status of\n intermediate certificates in the chain.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6961.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 1]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nCopyright Notice\n\n Copyright (c) 2013 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n1. Introduction\n\n The Transport Layer Security (TLS) Extension [RFC6066] framework\n defines, among other extensions, the Certificate Status extension\n (also referred to as \"OCSP stapling\") that clients can use to request\n the server's copy of the current status of its certificate. The\n benefits of this extension include a reduced number of roundtrips and\n network delays for the client to verify the status of the server's\n certificate and a reduced load on the certificate issuer's status\n response servers, thus solving a problem that can become significant\n when the issued certificate is presented by a frequently visited\n server.\n\n There are two problems with the existing Certificate Status\n extension. First, it does not provide functionality to request the\n status information about intermediate Certification Authority (CA)\n certificates, which means the client has to request status\n information through other methods, such as Certificate Revocation\n Lists (CRLs), introducing further delays. Second, the current format\n of the extension and requirements in the TLS protocol prevent a\n client from offering the server multiple status methods.\n\n\n\nPettersen Standards Track [Page 2]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n Many CAs are now issuing intermediate CA certificates that not only\n specify the publication point for their CRLs in a CRL Distribution\n Point [RFC5280] but also specify a URL for their OCSP [RFC6960]\n server in Authority Information Access [RFC5280]. Given that\n client-cached CRLs are frequently out of date, clients would benefit\n from using OCSP to access up-to-date status information about\n intermediate CA certificates. The benefit to the issuing CA is less\n clear, as providing the bandwidth for the OCSP responder can be\n costly, especially for CAs with many high-traffic subscriber sites,\n and this cost is a concern for many CAs. There are cases where OCSP\n requests for a single high-traffic site caused significant network\n problems for the issuing CA.\n\n Clients will benefit from the TLS server providing certificate status\n information regardless of type, not just for the server certificate\n but also for the intermediate CA certificates. Combining the status\n checks into one extension will reduce the roundtrips needed to\n complete the handshake by the client to just those needed for\n negotiating the TLS connection. Also, for the Certification\n Authorities, the load on their servers will depend on the number of\n certificates they have issued, not on the number of visitors to those\n sites. Additionally, using this extension significantly reduces\n privacy concerns around the clients informing the certificate issuer\n about which sites they are visiting.\n\n For such a new system to be introduced seamlessly, clients need to be\n able to indicate support for the existing OCSP Certificate Status\n method and a new multiple-OCSP mode.\n\n Unfortunately, the definition of the Certificate Status extension\n only allows a single Certificate Status extension to be defined in a\n single extension record in the handshake, and the TLS protocol\n [RFC5246] only allows a single record in the extension list for any\n given extension. This means that it is not possible for clients to\n indicate support for new methods while still supporting older\n methods, which would cause problems for interoperability between\n newer clients and older servers. This will not just be an issue for\n the multiple status request mode proposed above but also for any\n other future status methods that might be introduced. This will be\n true not just for the current PKIX infrastructure [RFC5280] but also\n for alternative PKI structures.\n\n The solution to this problem is to define a new extension,\n \"status_request_v2\", with an extended format that allows the client\n to indicate support for multiple status request methods. This is\n implemented using a list of CertificateStatusRequestItemV2 records in\n the extension record. As the server will select the single status\n\n\n\n\nPettersen Standards Track [Page 3]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n method based on the selected cipher suite and the certificate\n presented, no significant changes are needed in the server's\n extension format.\n\n1.1. Requirements Language\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [RFC2119].\n\n1.2. Presentation Language\n\n This document defines protocol structures using the same conventions\n and presentation language as defined in Section 4 of [RFC5246].\n\n2. Multiple Certificate Status Extension\n\n2.1. New Extension\n\n The extension defined by this document is indicated by\n \"status_request_v2\" in the ExtensionType enum (originally defined by\n [RFC6066]), which uses the following value:\n\n enum {\n status_request_v2(17), (65535)\n } ExtensionType;\n\n2.2. Multiple Certificate Status Request Record\n\n Clients that support a certificate status protocol like OCSP may send\n the \"status_request_v2\" extension to the server in order to use the\n TLS handshake to transfer such data instead of downloading it through\n separate connections. When using this extension, the\n \"extension_data\" field (defined in Section 7.4.1.4 of [RFC5246]) of\n the extension SHALL contain a CertificateStatusRequestListV2 where:\n\n struct {\n CertificateStatusType status_type;\n uint16 request_length; /* Length of request field in bytes */\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n case ocsp_multi: OCSPStatusRequest;\n } request;\n } CertificateStatusRequestItemV2;\n\n enum { ocsp(1), ocsp_multi(2), (255) } CertificateStatusType;\n\n\n\n\n\nPettersen Standards Track [Page 4]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n struct {\n CertificateStatusRequestItemV2\n certificate_status_req_list<1..2^16-1>;\n } CertificateStatusRequestListV2;\n\n In the OCSPStatusRequest (originally defined by [RFC6066]), the\n \"ResponderID\" provides a list of OCSP responders that the client\n trusts. A zero-length \"responder_id_list\" sequence has the special\n meaning that the responders are implicitly known to the server, e.g.,\n by prior arrangement, or are identified by the certificates used by\n the server. \"Extensions\" is a DER encoding [X.690] of the OCSP\n request extensions, and if the server supports the forwarding of OCSP\n request extensions, this value MUST be forwarded without\n modification.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC6960]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a DER-encoded zero-length ASN.1 SEQUENCE,\n which is not valid for the \"Extensions\" type).\n\n Servers that support a client's selection of responders using the\n ResponderID field could implement this selection by matching the\n responder ID values from the client's list with the ResponderIDs of\n known OCSP responders, either by using a binary compare of the values\n or a hash calculation and compare method.\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement). This\n has been clarified in [RFC6960].\n\n The items in the list of CertificateStatusRequestItemV2 entries are\n ordered according to the client's preference (favorite choice first).\n\n A server that receives a client hello containing the\n \"status_request_v2\" extension MAY return a suitable certificate\n status response message to the client along with the server's\n\n\n\nPettersen Standards Track [Page 5]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n certificate message. If OCSP is requested, it SHOULD use the\n information contained in the extension when selecting an OCSP\n responder and SHOULD include request_extensions in the OCSP request.\n\n The server returns a certificate status response along with its\n certificate by sending a \"CertificateStatus\" message (originally\n defined by [RFC6066]) immediately after the \"Certificate\" message\n (Section 7.4.2 of [RFC5246]) (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message in response to a \"status_request_v2\"\n request, then the server MUST have included an extension of type\n \"status_request_v2\" with empty \"extension_data\" in the extended\n server hello.\n\n The \"CertificateStatus\" message is conveyed using the handshake\n message type \"certificate_status\" (defined in [RFC6066]) as follows\n (updated from the definition in [RFC6066]):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n case ocsp_multi: OCSPResponseList;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<0..2^24-1>;\n\n struct {\n OCSPResponse ocsp_response_list<1..2^24-1>;\n } OCSPResponseList;\n\n An \"OCSPResponse\" element (originally defined by [RFC6066]) contains\n a complete, DER-encoded OCSP response (using the ASN.1 [X.680] type\n OCSPResponse defined in [RFC6960]). Only one OCSP response, with a\n length of at least one byte, may be sent for status_type \"ocsp\".\n\n An \"ocsp_response_list\" contains a list of \"OCSPResponse\" elements,\n as specified above, each containing the OCSP response for the\n matching corresponding certificate in the server's Certificate TLS\n handshake message. That is, the first entry is the OCSP response for\n the first certificate in the Certificate list, the second entry is\n the response for the second certificate, and so on. The list MAY\n contain fewer OCSP responses than there were certificates in the\n Certificate handshake message, but there MUST NOT be more responses\n than there were certificates in the list. Individual elements of the\n list MAY have a length of 0 (zero) bytes if the server does not have\n the OCSP response for that particular certificate stored, in which\n\n\n\nPettersen Standards Track [Page 6]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n case the client MUST act as if a response was not received for that\n particular certificate. If the client receives a\n \"ocsp_response_list\" that does not contain a response for one or more\n of the certificates in the completed certificate chain, the client\n SHOULD attempt to validate the certificate using an alternative\n retrieval method, such as downloading the relevant CRL; OCSP SHOULD\n in this situation only be used for the end-entity certificate, not\n intermediate CA certificates, for reasons stated above.\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if it has received a \"status_request_v2\" extension in\n the client hello message and has sent a \"status_request_v2\" extension\n in the server hello message. Additionally, note that a server MUST\n NOT send the \"CertificateStatus\" message unless it received either a\n \"status_request\" or \"status_request_v2\" extension in the client hello\n message and sent a corresponding \"status_request\" or\n \"status_request_v2\" extension in the server hello message.\n\n Clients requesting an OCSP response and receiving one or more OCSP\n responses in a \"CertificateStatus\" message MUST check the OCSP\n response(s) and abort the handshake if the response is a \"revoked\"\n status or other unacceptable responses (as determined by client\n policy) with a bad_certificate_status_response(113) alert. This\n alert is always fatal.\n\n If the OCSP response received from the server does not result in a\n definite \"good\" or \"revoked\" status, it is inconclusive. A TLS\n client in such a case MAY check the validity of the server\n certificate through other means, e.g., by directly querying the\n certificate issuer. If such processing still results in an\n inconclusive response, then the application using the TLS connection\n will have to decide whether to close the connection or not. Note\n that this problem cannot be decided by the generic TLS client code\n without information from the application. If the application doesn't\n provide any such information, then the client MUST abort the\n connection, since the server certificate has not been sufficiently\n validated.\n\n An example of where the application might wish to continue is with\n EAP-TLS (Extensible Authentication Protocol - TLS), where the\n application can use another mechanism to check the status of a\n certificate once it obtains network access. In this case, the\n application could have the client continue with the handshake, but it\n MUST NOT disclose a username and password until it has fully\n validated the server certificate.\n\n\n\n\n\n\nPettersen Standards Track [Page 7]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n3. IANA Considerations\n\n Section 2.1 defines the new TLS extension status_request_v2 (17)\n enum, which has been added to the \"ExtensionType Values\" list in the\n IANA \"Transport Layer Security (TLS) Extensions\" registry.\n\n Section 2.2 describes a TLS CertificateStatusType registry that is\n now maintained by IANA. The \"TLS Certificate Status Types\" registry\n has been created under the \"Transport Layer Security (TLS)\n Extensions\" registry. CertificateStatusType values are to be\n assigned via IETF Review as defined in [RFC5226]. The initial\n registry corresponds to the definition of \"CertificateStatusType\" in\n Section 2.2.\n\n Value Description Reference\n -----------------------------------------\n 0 Reserved [RFC6961]\n 1 ocsp [RFC6066] [RFC6961]\n 2 ocsp_multi [RFC6961]\n 3-255 Unassigned\n\n4. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security considerations for the particular extension\n specified in this document are given below. In general, implementers\n should continue to monitor the state of the art and address any\n weaknesses identified.\n\n4.1. Security Considerations for status_request_v2\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC6960] for further details.\n\n This extension allows the client to send arbitrary data to the\n server. The server implementers need to handle such data carefully\n to avoid introducing security vulnerabilities.\n\n The security considerations of [RFC6960] apply to OCSP requests and\n responses.\n\n\n\n\nPettersen Standards Track [Page 8]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n5. Acknowledgements\n\n This document is based on [RFC6066], authored by Donald Eastlake 3rd.\n\n6. References\n\n6.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246, August 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC6066] Eastlake, D., \"Transport Layer Security (TLS) Extensions:\n Extension Definitions\", RFC 6066, January 2011.\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, June 2013.\n\n [X.680] ITU-T Recommendation X.680 (2008) | ISO/IEC 8824-1:2008,\n \"Abstract Syntax Notation One (ASN.1): Specification of\n basic notation\", November 2008.\n\n [X.690] ITU-T Recommendation X.690 (2008) | ISO/IEC 8825-1:2008,\n \"ASN.1 encoding rules: Specification of Basic Encoding\n Rules (BER), Canonical Encoding Rules (CER) and\n Distinguished Encoding Rules (DER)\", November 2008.\n\n6.2. Informative References\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and C.\n Adams, \"X.509 Internet Public Key Infrastructure Online\n Certificate Status Protocol - OCSP\", RFC 2560, June 1999.\n\n\n\n\n\n\nPettersen Standards Track [Page 9]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nAuthor's Address\n\n Yngve N. Pettersen\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 10]\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 5705 RTFM, Inc.\nCategory: Standards Track March 2010\nISSN: 2070-1721\n\n\n Keying Material Exporters for Transport Layer Security (TLS)\n\nAbstract\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n to perform key establishment but then use some of the keying material\n for their own purposes. This document describes a general mechanism\n for allowing that.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/5705.\n\nCopyright Notice\n\n Copyright (c) 2010 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2\n 2. Conventions Used In This Document . . . . . . . . . . . . . . . 3\n 3. Binding to Application Contexts . . . . . . . . . . . . . . . . 3\n 4. Exporter Definition . . . . . . . . . . . . . . . . . . . . . . 4\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . . 5\n 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 6\n 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . 6\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 7\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 7\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 7\n\n1. Introduction\n\n Note: The mechanism described in this document was previously known\n as \"TLS Extractors\" but was changed to avoid a name conflict\n with the use of the term \"Extractor\" in the cryptographic\n community.\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n [RFC5246] or Datagram TLS (DTLS) [RFC4347] to perform key\n establishment but then use some of the keying material for their own\n purposes. A typical example is DTLS-SRTP [DTLS-SRTP], a key\n management scheme for the Secure Real-time Transport Protocol (SRTP)\n that uses DTLS to perform a key exchange and negotiate the SRTP\n [RFC3711] protection suite and then uses the DTLS master_secret to\n generate the SRTP keys.\n\n These applications imply a need to be able to export keying material\n (later called Exported Keying Material or EKM) from TLS/DTLS to an\n application or protocol residing at an upper layer, and to securely\n agree on the upper-layer context where the keying material will be\n used. The mechanism for exporting the keying material has the\n following requirements:\n\n o Both client and server need to be able to export the same EKM\n value.\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n o EKM values should be indistinguishable from random data to\n attackers who don't know the master_secret.\n\n o It should be possible to export multiple EKM values from the same\n TLS/DTLS association.\n\n o Knowing one EKM value should not reveal any useful information\n about the master_secret or about other EKM values.\n\n The mechanism described in this document is intended to fulfill these\n requirements. This mechanism is compatible with all versions of TLS.\n\n2. Conventions Used In This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Binding to Application Contexts\n\n In addition to using an exporter to obtain keying material, an\n application using the keying material has to securely establish the\n upper-layer context where the keying material will be used. The\n details of this context depend on the application, but it could\n include things such as algorithms and parameters that will be used\n with the keys, identifier(s) for the endpoint(s) who will use the\n keys, identifier(s) for the session(s) where the keys will be used,\n and the lifetime(s) for the context and/or keys. At a minimum, there\n should be some mechanism for signaling that an exporter will be used.\n\n This specification does not mandate a single mechanism for agreeing\n on such context; instead, there are several possibilities that can be\n used (and can complement each other). For example:\n\n o Information about the upper-layer context can be included in the\n optional data after the exporter label (see Section 4).\n\n o Information about the upper-layer context can be exchanged in TLS\n extensions included in the ClientHello and ServerHello messages.\n This approach is used in [DTLS-SRTP]. The handshake messages are\n protected by the Finished messages, so once the handshake\n completes, the peers will have the same view of the information.\n Extensions also allow a limited form of negotiation: for example,\n the TLS client could propose several alternatives for some context\n parameters, and the TLS server could select one of them.\n\n o The upper-layer protocol can include its own handshake, which can\n be protected using the keys exported by TLS.\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n No matter how the context is agreed, it is required that it has one\n part that indicates which application will use the exported keys.\n This part is the disambiguating label string (see Section 4).\n\n It is important to note that just embedding TLS messages in the\n upper-layer protocol may not automatically secure all the important\n context information, since the upper-layer messages are not covered\n by TLS Finished messages.\n\n4. Exporter Definition\n\n The output of the exporter is intended to be used in a single scope,\n which is associated with the TLS session, the label, and the context\n value.\n\n The exporter takes three input values:\n\n o a disambiguating label string,\n\n o a per-association context value provided by the application using\n the exporter, and\n\n o a length value.\n\n If no context is provided, it then computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random\n )[length]\n\n If context is provided, it computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random +\n context_value_length + context_value\n )[length]\n\n Where PRF is the TLS Pseudorandom Function in use for the session.\n The output is a pseudorandom bit string of length bytes generated\n from the master_secret. (This construction allows for\n interoperability with older exporter-type constructions which do not\n use context values, e.g., [RFC5281]).\n\n Labels here have the same definition as in TLS, i.e., an ASCII string\n with no terminating NULL. Label values beginning with \"EXPERIMENTAL\"\n MAY be used for private use without registration. All other label\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n values MUST be registered via Specification Required as described by\n RFC 5226 [RFC5226]. Note that exporter labels have the potential to\n collide with existing PRF labels. In order to prevent this, labels\n SHOULD begin with \"EXPORTER\". This is not a MUST because there are\n existing uses that have labels which do not begin with this prefix.\n\n The context value allows the application using the exporter to mix\n its own data with the TLS PRF for the exporter output. One example\n of where this might be useful is an authentication setting where the\n client credentials are valid for more than one identity; the context\n value could then be used to mix the expected identity into the keying\n material, thus preventing substitution attacks. The context value\n length is encoded as an unsigned, 16-bit quantity (uint16; see\n [RFC5246], Section 4.4) representing the length of the context value.\n The context MAY be zero length. Because the context value is mixed\n with the master_secret via the PRF, it is safe to mix confidential\n information into the exporter, provided that the master_secret will\n not be known to the attacker.\n\n5. Security Considerations\n\n The prime security requirement for exporter outputs is that they be\n independent. More formally, after a particular TLS session, if an\n adversary is allowed to choose multiple (label, context value) pairs\n and is given the output of the PRF for those values, the attacker is\n still unable to distinguish between the output of the PRF for a\n (label, context value) pair (different from the ones that it\n submitted) and a random value of the same length. In particular,\n there may be settings, such as the one described in Section 4, where\n the attacker can control the context value; such an attacker MUST NOT\n be able to predict the output of the exporter. Similarly, an\n attacker who does not know the master secret should not be able to\n distinguish valid exporter outputs from random values. The current\n set of TLS PRFs is believed to meet this objective, provided the\n master secret is randomly generated.\n\n Because an exporter produces the same value if applied twice with the\n same label to the same master_secret, it is critical that two EKM\n values generated with the same label not be used for two different\n purposes -- hence, the requirement for IANA registration. However,\n because exporters depend on the TLS PRF, it is not a threat to the\n use of an EKM value generated from one label to reveal an EKM value\n generated from another label.\n\n With certain TLS cipher suites, the TLS master secret is not\n necessarily unique to a single TLS session. In particular, with RSA\n key exchange, a malicious party acting as TLS server in one session\n and as TLS client in another session can cause those two sessions to\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n have the same TLS master secret (though the sessions must be\n established simultaneously to get adequate control of the Random\n values). Applications using the EKM need to consider this in how\n they use the EKM; in some cases, requiring the use of other cipher\n suites (such as those using a Diffie-Hellman key exchange) may be\n advisable.\n\n Designing a secure mechanism that uses exporters is not necessarily\n straightforward. This document only provides the exporter mechanism,\n but the problem of agreeing on the surrounding context and the\n meaning of the information passed to and from the exporter remains.\n Any new uses of the exporter mechanism should be subject to careful\n review.\n\n6. IANA Considerations\n\n IANA has created a TLS Exporter Label registry for this purpose. The\n initial contents of the registry are given below:\n\n Value Reference Note\n ----------------------------- --------- ----\n client finished [RFC5246] (1)\n server finished [RFC5246] (1)\n master secret [RFC5246] (1)\n key expansion [RFC5246] (1)\n client EAP encryption [RFC5216]\n ttls keying material [RFC5281]\n ttls challenge [RFC5281]\n\n Note: (1) These entries are reserved and MUST NOT be used for the\n purpose described in RFC 5705, in order to avoid confusion with\n similar, but distinct, use in RFC 5246.\n\n Future values are allocated via the RFC 5226 Specification Required\n policy. The label is a string consisting of printable ASCII\n characters. IANA MUST also verify that one label is not a prefix of\n any other label. For example, labels \"key\" or \"master secretary\" are\n forbidden.\n\n7. Acknowledgments\n\n Thanks to Pasi Eronen for valuable comments and for the contents of\n the IANA section and Section 3. Thanks to David McGrew for helpful\n discussion of the security considerations and to Vijay Gurbani and\n Alfred Hoenes for editorial comments.\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246,\n August 2008.\n\n8.2. Informative References\n\n [DTLS-SRTP] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for Secure\n Real-time Transport Protocol (SRTP)\", Work in Progress,\n February 2009.\n\n [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and\n K. Norrman, \"The Secure Real-time Transport Protocol\n (SRTP)\", RFC 3711, March 2004.\n\n [RFC4347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security\", RFC 4347, April 2006.\n\n [RFC5216] Simon, D., Aboba, B., and R. Hurst, \"The EAP-TLS\n Authentication Protocol\", RFC 5216, March 2008.\n\n [RFC5281] Funk, P. and S. Blake-Wilson, \"Extensible Authentication\n Protocol Tunneled Transport Layer Security Authenticated\n Protocol Version 0 (EAP-TTLSv0)\", RFC 5281, August 2008.\n\nAuthor's Address\n\n Eric Rescorla\n RTFM, Inc.\n 2064 Edgewood Drive\n Palo Alto, CA 94303\n USA\n\n EMail: [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) D. Eastlake 3rd\nRequest for Comments: 6066 Huawei\nObsoletes: 4366 January 2011\nCategory: Standards Track\nISSN: 2070-1721\n\n\n Transport Layer Security (TLS) Extensions: Extension Definitions\n\nAbstract\n\n This document provides specifications for existing TLS extensions.\n It is a companion document for RFC 5246, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\". The extensions specified are\n server_name, max_fragment_length, client_certificate_url,\n trusted_ca_keys, truncated_hmac, and status_request.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6066.\n\nCopyright Notice\n\n Copyright (c) 2011 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n\n\n\n\n\nEastlake Standards Track [Page 1]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction ....................................................3\n 1.1. Specific Extensions Covered ................................3\n 1.2. Conventions Used in This Document ..........................5\n 2. Extensions to the Handshake Protocol ............................5\n 3. Server Name Indication ..........................................6\n 4. Maximum Fragment Length Negotiation .............................8\n 5. Client Certificate URLs .........................................9\n 6. Trusted CA Indication ..........................................12\n 7. Truncated HMAC .................................................13\n 8. Certificate Status Request .....................................14\n 9. Error Alerts ...................................................16\n 10. IANA Considerations ...........................................17\n 10.1. pkipath MIME Type Registration ...........................17\n 10.2. Reference for TLS Alerts, TLS HandshakeTypes, and\n ExtensionTypes ...........................................19\n 11. Security Considerations .......................................19\n 11.1. Security Considerations for server_name ..................19\n 11.2. Security Considerations for max_fragment_length ..........20\n 11.3. Security Considerations for client_certificate_url .......20\n 11.4. Security Considerations for trusted_ca_keys ..............21\n 11.5. Security Considerations for truncated_hmac ...............21\n 11.6. Security Considerations for status_request ...............22\n 12. Normative References ..........................................22\n 13. Informative References ........................................23\n Appendix A. Changes from RFC 4366 .................................24\n Appendix B. Acknowledgements ......................................25\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 2]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1. Introduction\n\n The Transport Layer Security (TLS) Protocol Version 1.2 is specified\n in [RFC5246]. That specification includes the framework for\n extensions to TLS, considerations in designing such extensions (see\n Section 7.4.1.4 of [RFC5246]), and IANA Considerations for the\n allocation of new extension code points; however, it does not specify\n any particular extensions other than Signature Algorithms (see\n Section 7.4.1.4.1 of [RFC5246]).\n\n This document provides the specifications for existing TLS\n extensions. It is, for the most part, the adaptation and editing of\n material from RFC 4366, which covered TLS extensions for TLS 1.0 (RFC\n 2246) and TLS 1.1 (RFC 4346).\n\n1.1. Specific Extensions Covered\n\n The extensions described here focus on extending the functionality\n provided by the TLS protocol message formats. Other issues, such as\n the addition of new cipher suites, are deferred.\n\n The extension types defined in this document are:\n\n enum {\n server_name(0), max_fragment_length(1),\n client_certificate_url(2), trusted_ca_keys(3),\n truncated_hmac(4), status_request(5), (65535)\n } ExtensionType;\n\n Specifically, the extensions described in this document:\n\n - Allow TLS clients to provide to the TLS server the name of the\n server they are contacting. This functionality is desirable in\n order to facilitate secure connections to servers that host\n multiple 'virtual' servers at a single underlying network address.\n\n - Allow TLS clients and servers to negotiate the maximum fragment\n length to be sent. This functionality is desirable as a result of\n memory constraints among some clients, and bandwidth constraints\n among some access networks.\n\n - Allow TLS clients and servers to negotiate the use of client\n certificate URLs. This functionality is desirable in order to\n conserve memory on constrained clients.\n\n\n\n\n\n\n\nEastlake Standards Track [Page 3]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - Allow TLS clients to indicate to TLS servers which certification\n authority (CA) root keys they possess. This functionality is\n desirable in order to prevent multiple handshake failures\n involving TLS clients that are only able to store a small number\n of CA root keys due to memory limitations.\n\n - Allow TLS clients and servers to negotiate the use of truncated\n Message Authentication Codes (MACs). This functionality is\n desirable in order to conserve bandwidth in constrained access\n networks.\n\n - Allow TLS clients and servers to negotiate that the server sends\n the client certificate status information (e.g., an Online\n Certificate Status Protocol (OCSP) [RFC2560] response) during a\n TLS handshake. This functionality is desirable in order to avoid\n sending a Certificate Revocation List (CRL) over a constrained\n access network and therefore saving bandwidth.\n\n TLS clients and servers may use the extensions described in this\n document. The extensions are designed to be backwards compatible,\n meaning that TLS clients that support the extensions can talk to TLS\n servers that do not support the extensions, and vice versa.\n\n Note that any messages associated with these extensions that are sent\n during the TLS handshake MUST be included in the hash calculations\n involved in \"Finished\" messages.\n\n Note also that all the extensions defined in this document are\n relevant only when a session is initiated. A client that requests\n session resumption does not in general know whether the server will\n accept this request, and therefore it SHOULD send the same extensions\n as it would send if it were not attempting resumption. When a client\n includes one or more of the defined extension types in an extended\n client hello while requesting session resumption:\n\n - The server name indication extension MAY be used by the server\n when deciding whether or not to resume a session as described in\n Section 3.\n\n - If the resumption request is denied, the use of the extensions is\n negotiated as normal.\n\n - If, on the other hand, the older session is resumed, then the\n server MUST ignore the extensions and send a server hello\n containing none of the extension types. In this case, the\n functionality of these extensions negotiated during the original\n session initiation is applied to the resumed session.\n\n\n\n\nEastlake Standards Track [Page 4]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1.2. Conventions Used in This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n [RFC2119].\n\n2. Extensions to the Handshake Protocol\n\n This document specifies the use of two new handshake messages,\n \"CertificateURL\" and \"CertificateStatus\". These messages are\n described in Sections 5 and 8, respectively. The new handshake\n message structure therefore becomes:\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), certificate_url(21), certificate_status(22),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case certificate_url: CertificateURL;\n case certificate_status: CertificateStatus;\n } body;\n } Handshake;\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 5]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n3. Server Name Indication\n\n TLS does not provide a mechanism for a client to tell a server the\n name of the server it is contacting. It may be desirable for clients\n to provide this information to facilitate secure connections to\n servers that host multiple 'virtual' servers at a single underlying\n network address.\n\n In order to provide any of the server names, clients MAY include an\n extension of type \"server_name\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"ServerNameList\" where:\n\n struct {\n NameType name_type;\n select (name_type) {\n case host_name: HostName;\n } name;\n } ServerName;\n\n enum {\n host_name(0), (255)\n } NameType;\n\n opaque HostName<1..2^16-1>;\n\n struct {\n ServerName server_name_list<1..2^16-1>\n } ServerNameList;\n\n The ServerNameList MUST NOT contain more than one name of the same\n name_type. If the server understood the ClientHello extension but\n does not recognize the server name, the server SHOULD take one of two\n actions: either abort the handshake by sending a fatal-level\n unrecognized_name(112) alert or continue the handshake. It is NOT\n RECOMMENDED to send a warning-level unrecognized_name(112) alert,\n because the client's behavior in response to warning-level alerts is\n unpredictable. If there is a mismatch between the server name used\n by the client application and the server name of the credential\n chosen by the server, this mismatch will become apparent when the\n client application performs the server endpoint identification, at\n which point the client application will have to decide whether to\n proceed with the communication. TLS implementations are encouraged\n to make information available to application callers about warning-\n level alerts that were received or sent during a TLS handshake. Such\n information can be useful for diagnostic purposes.\n\n\n\n\n\nEastlake Standards Track [Page 6]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note: Earlier versions of this specification permitted multiple\n names of the same name_type. In practice, current client\n implementations only send one name, and the client cannot\n necessarily find out which name the server selected. Multiple\n names of the same name_type are therefore now prohibited.\n\n Currently, the only server names supported are DNS hostnames;\n however, this does not imply any dependency of TLS on DNS, and other\n name types may be added in the future (by an RFC that updates this\n document). The data structure associated with the host_name NameType\n is a variable-length vector that begins with a 16-bit length. For\n backward compatibility, all future data structures associated with\n new NameTypes MUST begin with a 16-bit length field. TLS MAY treat\n provided server names as opaque data and pass the names and types to\n the application.\n\n \"HostName\" contains the fully qualified DNS hostname of the server,\n as understood by the client. The hostname is represented as a byte\n string using ASCII encoding without a trailing dot. This allows the\n support of internationalized domain names through the use of A-labels\n defined in [RFC5890]. DNS hostnames are case-insensitive. The\n algorithm to compare hostnames is described in [RFC5890], Section\n 2.3.2.4.\n\n Literal IPv4 and IPv6 addresses are not permitted in \"HostName\".\n\n It is RECOMMENDED that clients include an extension of type\n \"server_name\" in the client hello whenever they locate a server by a\n supported name type.\n\n A server that receives a client hello containing the \"server_name\"\n extension MAY use the information contained in the extension to guide\n its selection of an appropriate certificate to return to the client,\n and/or other aspects of security policy. In this event, the server\n SHALL include an extension of type \"server_name\" in the (extended)\n server hello. The \"extension_data\" field of this extension SHALL be\n empty.\n\n When the server is deciding whether or not to accept a request to\n resume a session, the contents of a server_name extension MAY be used\n in the lookup of the session in the session cache. The client SHOULD\n include the same server_name extension in the session resumption\n request as it did in the full handshake that established the session.\n A server that implements this extension MUST NOT accept the request\n to resume the session if the server_name extension contains a\n different name. Instead, it proceeds with a full handshake to\n establish a new session. When resuming a session, the server MUST\n NOT include a server_name extension in the server hello.\n\n\n\nEastlake Standards Track [Page 7]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n If an application negotiates a server name using an application\n protocol and then upgrades to TLS, and if a server_name extension is\n sent, then the extension SHOULD contain the same name that was\n negotiated in the application protocol. If the server_name is\n established in the TLS session handshake, the client SHOULD NOT\n attempt to request a different server name at the application layer.\n\n4. Maximum Fragment Length Negotiation\n\n Without this extension, TLS specifies a fixed maximum plaintext\n fragment length of 2^14 bytes. It may be desirable for constrained\n clients to negotiate a smaller maximum fragment length due to memory\n limitations or bandwidth limitations.\n\n In order to negotiate smaller maximum fragment lengths, clients MAY\n include an extension of type \"max_fragment_length\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain:\n\n enum{\n 2^9(1), 2^10(2), 2^11(3), 2^12(4), (255)\n } MaxFragmentLength;\n\n whose value is the desired maximum fragment length. The allowed\n values for this field are: 2^9, 2^10, 2^11, and 2^12.\n\n Servers that receive an extended client hello containing a\n \"max_fragment_length\" extension MAY accept the requested maximum\n fragment length by including an extension of type\n \"max_fragment_length\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL contain a\n \"MaxFragmentLength\" whose value is the same as the requested maximum\n fragment length.\n\n If a server receives a maximum fragment length negotiation request\n for a value other than the allowed values, it MUST abort the\n handshake with an \"illegal_parameter\" alert. Similarly, if a client\n receives a maximum fragment length negotiation response that differs\n from the length it requested, it MUST also abort the handshake with\n an \"illegal_parameter\" alert.\n\n Once a maximum fragment length other than 2^14 has been successfully\n negotiated, the client and server MUST immediately begin fragmenting\n messages (including handshake messages) to ensure that no fragment\n larger than the negotiated length is sent. Note that TLS already\n requires clients and servers to support fragmentation of handshake\n messages.\n\n\n\n\nEastlake Standards Track [Page 8]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n The negotiated length applies for the duration of the session\n including session resumptions.\n\n The negotiated length limits the input that the record layer may\n process without fragmentation (that is, the maximum value of\n TLSPlaintext.length; see [RFC5246], Section 6.2.1). Note that the\n output of the record layer may be larger. For example, if the\n negotiated length is 2^9=512, then, when using currently defined\n cipher suites (those defined in [RFC5246] and [RFC2712]) and null\n compression, the record-layer output can be at most 805 bytes: 5\n bytes of headers, 512 bytes of application data, 256 bytes of\n padding, and 32 bytes of MAC. This means that in this event a TLS\n record-layer peer receiving a TLS record-layer message larger than\n 805 bytes MUST discard the message and send a \"record_overflow\"\n alert, without decrypting the message. When this extension is used\n with Datagram Transport Layer Security (DTLS), implementations SHOULD\n NOT generate record_overflow alerts unless the packet passes message\n authentication.\n\n5. Client Certificate URLs\n\n Without this extension, TLS specifies that when client authentication\n is performed, client certificates are sent by clients to servers\n during the TLS handshake. It may be desirable for constrained\n clients to send certificate URLs in place of certificates, so that\n they do not need to store their certificates and can therefore save\n memory.\n\n In order to negotiate sending certificate URLs to a server, clients\n MAY include an extension of type \"client_certificate_url\" in the\n (extended) client hello. The \"extension_data\" field of this\n extension SHALL be empty.\n\n (Note that it is necessary to negotiate the use of client certificate\n URLs in order to avoid \"breaking\" existing TLS servers.)\n\n Servers that receive an extended client hello containing a\n \"client_certificate_url\" extension MAY indicate that they are willing\n to accept certificate URLs by including an extension of type\n \"client_certificate_url\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n After negotiation of the use of client certificate URLs has been\n successfully completed (by exchanging hellos including\n \"client_certificate_url\" extensions), clients MAY send a\n \"CertificateURL\" message in place of a \"Certificate\" message as\n follows (see also Section 2):\n\n\n\n\nEastlake Standards Track [Page 9]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n enum {\n individual_certs(0), pkipath(1), (255)\n } CertChainType;\n\n struct {\n CertChainType type;\n URLAndHash url_and_hash_list<1..2^16-1>;\n } CertificateURL;\n\n struct {\n opaque url<1..2^16-1>;\n unint8 padding;\n opaque SHA1Hash[20];\n } URLAndHash;\n\n Here, \"url_and_hash_list\" contains a sequence of URLs and hashes.\n Each \"url\" MUST be an absolute URI reference according to [RFC3986]\n that can be immediately used to fetch the certificate(s).\n\n When X.509 certificates are used, there are two possibilities:\n\n - If CertificateURL.type is \"individual_certs\", each URL refers to a\n single DER-encoded X.509v3 certificate, with the URL for the\n client's certificate first.\n\n - If CertificateURL.type is \"pkipath\", the list contains a single\n URL referring to a DER-encoded certificate chain, using the type\n PkiPath described in Section 10.1.\n\n When any other certificate format is used, the specification that\n describes use of that format in TLS should define the encoding format\n of certificates or certificate chains, and any constraint on their\n ordering.\n\n The \"padding\" byte MUST be 0x01. It is present to make the structure\n backwards compatible.\n\n The hash corresponding to each URL is the SHA-1 hash of the\n certificate or certificate chain (in the case of X.509 certificates,\n the DER-encoded certificate or the DER-encoded PkiPath).\n\n Note that when a list of URLs for X.509 certificates is used, the\n ordering of URLs is the same as that used in the TLS Certificate\n message (see [RFC5246], Section 7.4.2), but opposite to the order in\n which certificates are encoded in PkiPath. In either case, the self-\n signed root certificate MAY be omitted from the chain, under the\n assumption that the server must already possess it in order to\n validate it.\n\n\n\nEastlake Standards Track [Page 10]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Servers receiving \"CertificateURL\" SHALL attempt to retrieve the\n client's certificate chain from the URLs and then process the\n certificate chain as usual. A cached copy of the content of any URL\n in the chain MAY be used, provided that the SHA-1 hash matches the\n hash of the cached copy.\n\n Servers that support this extension MUST support the 'http' URI\n scheme for certificate URLs and MAY support other schemes. Use of\n other schemes than 'http', 'https', or 'ftp' may create unexpected\n problems.\n\n If the protocol used is HTTP, then the HTTP server can be configured\n to use the Cache-Control and Expires directives described in\n [RFC2616] to specify whether and for how long certificates or\n certificate chains should be cached.\n\n The TLS server MUST NOT follow HTTP redirects when retrieving the\n certificates or certificate chain. The URLs used in this extension\n MUST NOT be chosen to depend on such redirects.\n\n If the protocol used to retrieve certificates or certificate chains\n returns a MIME-formatted response (as HTTP does), then the following\n MIME Content-Types SHALL be used: when a single X.509v3 certificate\n is returned, the Content-Type is \"application/pkix-cert\" [RFC2585],\n and when a chain of X.509v3 certificates is returned, the Content-\n Type is \"application/pkix-pkipath\" (Section 10.1).\n\n The server MUST check that the SHA-1 hash of the contents of the\n object retrieved from that URL (after decoding any MIME Content-\n Transfer-Encoding) matches the given hash. If any retrieved object\n does not have the correct SHA-1 hash, the server MUST abort the\n handshake with a bad_certificate_hash_value(114) alert. This alert\n is always fatal.\n\n Clients may choose to send either \"Certificate\" or \"CertificateURL\"\n after successfully negotiating the option to send certificate URLs.\n The option to send a certificate is included to provide flexibility\n to clients possessing multiple certificates.\n\n If a server is unable to obtain certificates in a given\n CertificateURL, it MUST send a fatal certificate_unobtainable(111)\n alert if it requires the certificates to complete the handshake. If\n the server does not require the certificates, then the server\n continues the handshake. The server MAY send a warning-level alert\n in this case. Clients receiving such an alert SHOULD log the alert\n and continue with the handshake if possible.\n\n\n\n\n\nEastlake Standards Track [Page 11]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n6. Trusted CA Indication\n\n Constrained clients that, due to memory limitations, possess only a\n small number of CA root keys may wish to indicate to servers which\n root keys they possess, in order to avoid repeated handshake\n failures.\n\n In order to indicate which CA root keys they possess, clients MAY\n include an extension of type \"trusted_ca_keys\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain \"TrustedAuthorities\" where:\n\n struct {\n TrustedAuthority trusted_authorities_list<0..2^16-1>;\n } TrustedAuthorities;\n\n struct {\n IdentifierType identifier_type;\n select (identifier_type) {\n case pre_agreed: struct {};\n case key_sha1_hash: SHA1Hash;\n case x509_name: DistinguishedName;\n case cert_sha1_hash: SHA1Hash;\n } identifier;\n } TrustedAuthority;\n\n enum {\n pre_agreed(0), key_sha1_hash(1), x509_name(2),\n cert_sha1_hash(3), (255)\n } IdentifierType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n Here, \"TrustedAuthorities\" provides a list of CA root key identifiers\n that the client possesses. Each CA root key is identified via\n either:\n\n - \"pre_agreed\": no CA root key identity supplied.\n\n - \"key_sha1_hash\": contains the SHA-1 hash of the CA root key. For\n Digital Signature Algorithm (DSA) and Elliptic Curve Digital\n Signature Algorithm (ECDSA) keys, this is the hash of the\n \"subjectPublicKey\" value. For RSA keys, the hash is of the big-\n endian byte string representation of the modulus without any\n initial zero-valued bytes. (This copies the key hash formats\n deployed in other environments.)\n\n\n\n\n\nEastlake Standards Track [Page 12]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - \"x509_name\": contains the DER-encoded X.509 DistinguishedName of\n the CA.\n\n - \"cert_sha1_hash\": contains the SHA-1 hash of a DER-encoded\n Certificate containing the CA root key.\n\n Note that clients may include none, some, or all of the CA root keys\n they possess in this extension.\n\n Note also that it is possible that a key hash or a Distinguished Name\n alone may not uniquely identify a certificate issuer (for example, if\n a particular CA has multiple key pairs). However, here we assume\n this is the case following the use of Distinguished Names to identify\n certificate issuers in TLS.\n\n The option to include no CA root keys is included to allow the client\n to indicate possession of some pre-defined set of CA root keys.\n\n Servers that receive a client hello containing the \"trusted_ca_keys\"\n extension MAY use the information contained in the extension to guide\n their selection of an appropriate certificate chain to return to the\n client. In this event, the server SHALL include an extension of type\n \"trusted_ca_keys\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n7. Truncated HMAC\n\n Currently defined TLS cipher suites use the MAC construction HMAC\n [RFC2104] to authenticate record-layer communications. In TLS, the\n entire output of the hash function is used as the MAC tag. However,\n it may be desirable in constrained environments to save bandwidth by\n truncating the output of the hash function to 80 bits when forming\n MAC tags.\n\n In order to negotiate the use of 80-bit truncated HMAC, clients MAY\n include an extension of type \"truncated_hmac\" in the extended client\n hello. The \"extension_data\" field of this extension SHALL be empty.\n\n Servers that receive an extended hello containing a \"truncated_hmac\"\n extension MAY agree to use a truncated HMAC by including an extension\n of type \"truncated_hmac\", with empty \"extension_data\", in the\n extended server hello.\n\n Note that if new cipher suites are added that do not use HMAC, and\n the session negotiates one of these cipher suites, this extension\n will have no effect. It is strongly recommended that any new cipher\n suites using other MACs consider the MAC size an integral part of the\n\n\n\n\nEastlake Standards Track [Page 13]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n cipher suite definition, taking into account both security and\n bandwidth considerations.\n\n If HMAC truncation has been successfully negotiated during a TLS\n handshake, and the negotiated cipher suite uses HMAC, both the client\n and the server pass this fact to the TLS record layer along with the\n other negotiated security parameters. Subsequently during the\n session, clients and servers MUST use truncated HMACs, calculated as\n specified in [RFC2104]. That is, SecurityParameters.mac_length is 10\n bytes, and only the first 10 bytes of the HMAC output are transmitted\n and checked. Note that this extension does not affect the\n calculation of the pseudo-random function (PRF) as part of\n handshaking or key derivation.\n\n The negotiated HMAC truncation size applies for the duration of the\n session including session resumptions.\n\n8. Certificate Status Request\n\n Constrained clients may wish to use a certificate-status protocol\n such as OCSP [RFC2560] to check the validity of server certificates,\n in order to avoid transmission of CRLs and therefore save bandwidth\n on constrained networks. This extension allows for such information\n to be sent in the TLS handshake, saving roundtrips and resources.\n\n In order to indicate their desire to receive certificate status\n information, clients MAY include an extension of type\n \"status_request\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"CertificateStatusRequest\" where:\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n } request;\n } CertificateStatusRequest;\n\n enum { ocsp(1), (255) } CertificateStatusType;\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n\n\n\nEastlake Standards Track [Page 14]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n In the OCSPStatusRequest, the \"ResponderIDs\" provides a list of OCSP\n responders that the client trusts. A zero-length \"responder_id_list\"\n sequence has the special meaning that the responders are implicitly\n known to the server, e.g., by prior arrangement. \"Extensions\" is a\n DER encoding of OCSP request extensions.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC2560]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a zero-length ASN.1 SEQUENCE, which is not\n valid for the \"Extensions\" type).\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement).\n\n Servers that receive a client hello containing the \"status_request\"\n extension MAY return a suitable certificate status response to the\n client along with their certificate. If OCSP is requested, they\n SHOULD use the information contained in the extension when selecting\n an OCSP responder and SHOULD include request_extensions in the OCSP\n request.\n\n Servers return a certificate response along with their certificate by\n sending a \"CertificateStatus\" message immediately after the\n \"Certificate\" message (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message, then the server MUST have included an\n extension of type \"status_request\" with empty \"extension_data\" in the\n extended server hello. The \"CertificateStatus\" message is conveyed\n using the handshake message type \"certificate_status\" as follows (see\n also Section 2):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<1..2^24-1>;\n\n An \"ocsp_response\" contains a complete, DER-encoded OCSP response\n (using the ASN.1 type OCSPResponse defined in [RFC2560]). Only one\n OCSP response may be sent.\n\n\n\n\nEastlake Standards Track [Page 15]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if has received a \"status_request\" extension in the\n client hello message and has sent a \"status_request\" extension in the\n server hello message.\n\n Note in addition that a server MUST NOT send the \"CertificateStatus\"\n message unless it received a \"status_request\" extension in the client\n hello message and sent a \"status_request\" extension in the server\n hello message.\n\n Clients requesting an OCSP response and receiving an OCSP response in\n a \"CertificateStatus\" message MUST check the OCSP response and abort\n the handshake if the response is not satisfactory with\n bad_certificate_status_response(113) alert. This alert is always\n fatal.\n\n9. Error Alerts\n\n Four new error alerts are defined for use with the TLS extensions\n defined in this document. To avoid \"breaking\" existing clients and\n servers, these alerts MUST NOT be sent unless the sending party has\n received an extended hello message from the party they are\n communicating with. These error alerts are conveyed using the\n following syntax. The new alerts are the last four, as indicated by\n the comments on the same line as the error alert number.\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n /* 41 is not defined, for historical reasons */\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction(60),\n protocol_version(70),\n insufficient_security(71),\n\n\n\nEastlake Standards Track [Page 16]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n certificate_unobtainable(111), /* new */\n unrecognized_name(112), /* new */\n bad_certificate_status_response(113), /* new */\n bad_certificate_hash_value(114), /* new */\n (255)\n } AlertDescription;\n\n \"certificate_unobtainable\" is described in Section 5.\n \"unrecognized_name\" is described in Section 3.\n \"bad_certificate_status_response\" is described in Section 8.\n \"bad_certificate_hash_value\" is described in Section 5.\n\n10. IANA Considerations\n\n IANA Considerations for TLS extensions and the creation of a registry\n are covered in Section 12 of [RFC5246] except for the registration of\n MIME type application/pkix-pkipath, which appears below.\n\n The IANA TLS extensions and MIME type application/pkix-pkipath\n registry entries that reference RFC 4366 have been updated to\n reference this document.\n\n10.1. pkipath MIME Type Registration\n\n MIME media type name: application\n MIME subtype name: pkix-pkipath\n Required parameters: none\n\n Optional parameters: version (default value is \"1\")\n\n Encoding considerations:\n Binary; this MIME type is a DER encoding of the ASN.1 type\n PkiPath, defined as follows:\n PkiPath ::= SEQUENCE OF Certificate\n PkiPath is used to represent a certification path. Within the\n sequence, the order of certificates is such that the subject of\n the first certificate is the issuer of the second certificate,\n etc.\n This is identical to the definition published in [X509-4th-TC1];\n note that it is different from that in [X509-4th].\n\n All Certificates MUST conform to [RFC5280]. (This should be\n interpreted as a requirement to encode only PKIX-conformant\n certificates using this type. It does not necessarily require\n\n\n\nEastlake Standards Track [Page 17]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n that all certificates that are not strictly PKIX-conformant must\n be rejected by relying parties, although the security consequences\n of accepting any such certificates should be considered\n carefully.)\n\n DER (as opposed to BER) encoding MUST be used. If this type is\n sent over a 7-bit transport, base64 encoding SHOULD be used.\n\n Security considerations:\n The security considerations of [X509-4th] and [RFC5280] (or any\n updates to them) apply, as well as those of any protocol that uses\n this type (e.g., TLS).\n\n Note that this type only specifies a certificate chain that can be\n assessed for validity according to the relying party's existing\n configuration of trusted CAs; it is not intended to be used to\n specify any change to that configuration.\n\n Interoperability considerations:\n No specific interoperability problems are known with this type,\n but for recommendations relating to X.509 certificates in general,\n see [RFC5280].\n\n Published specification: This document and [RFC5280].\n\n Applications that use this media type:\n TLS. It may also be used by other protocols or for general\n interchange of PKIX certificate chains.\n\n Additional information:\n Magic number(s): DER-encoded ASN.1 can be easily recognized.\n Further parsing is required to distinguish it from other ASN.1\n types.\n File extension(s): .pkipath\n Macintosh File Type Code(s): not specified\n\n Person & email address to contact for further information:\n Magnus Nystrom <[redacted-email]>\n\n Intended usage: COMMON\n\n Change controller: IESG <[redacted-email]>\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 18]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n10.2. Reference for TLS Alerts, TLS HandshakeTypes, and ExtensionTypes\n\n The following values in the TLS Alert Registry have been updated to\n reference this document:\n\n 111 certificate_unobtainable\n 112 unrecognized_name\n 113 bad_certificate_status_response\n 114 bad_certificate_hash_value\n\n The following values in the TLS HandshakeType Registry have been\n updated to reference this document:\n\n 21 certificate_url\n 22 certificate_status\n\n The following ExtensionType values have been updated to reference\n this document:\n\n 0 server_name\n 1 max_fragment_length\n 2 client_certificate_url\n 3 trusted_ca_keys\n 4 truncated_hmac\n 5 status_request\n\n11. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security Considerations for particular extensions\n specified in this document are given below.\n\n In general, implementers should continue to monitor the state of the\n art and address any weaknesses identified.\n\n11.1. Security Considerations for server_name\n\n If a single server hosts several domains, then clearly it is\n necessary for the owners of each domain to ensure that this satisfies\n their security needs. Apart from this, server_name does not appear\n to introduce significant security issues.\n\n Since it is possible for a client to present a different server_name\n in the application protocol, application server implementations that\n rely upon these names being the same MUST check to make sure the\n client did not present a different name in the application protocol.\n\n\n\n\n\nEastlake Standards Track [Page 19]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Implementations MUST ensure that a buffer overflow does not occur,\n whatever the values of the length fields in server_name.\n\n11.2. Security Considerations for max_fragment_length\n\n The maximum fragment length takes effect immediately, including for\n handshake messages. However, that does not introduce any security\n complications that are not already present in TLS, since TLS requires\n implementations to be able to handle fragmented handshake messages.\n\n Note that, as described in Section 4, once a non-null cipher suite\n has been activated, the effective maximum fragment length depends on\n the cipher suite and compression method, as well as on the negotiated\n max_fragment_length. This must be taken into account when sizing\n buffers and checking for buffer overflow.\n\n11.3. Security Considerations for client_certificate_url\n\n Support for client_certificate_url involves the server's acting as a\n client in another URI-scheme-dependent protocol. The server\n therefore becomes subject to many of the same security concerns that\n clients of the URI scheme are subject to, with the added concern that\n the client can attempt to prompt the server to connect to some\n (possibly weird-looking) URL.\n\n In general, this issue means that an attacker might use the server to\n indirectly attack another host that is vulnerable to some security\n flaw. It also introduces the possibility of denial-of-service\n attacks in which an attacker makes many connections to the server,\n each of which results in the server's attempting a connection to the\n target of the attack.\n\n Note that the server may be behind a firewall or otherwise able to\n access hosts that would not be directly accessible from the public\n Internet. This could exacerbate the potential security and denial-\n of-service problems described above, as well as allow the existence\n of internal hosts to be confirmed when they would otherwise be\n hidden.\n\n The detailed security concerns involved will depend on the URI\n schemes supported by the server. In the case of HTTP, the concerns\n are similar to those that apply to a publicly accessible HTTP proxy\n server. In the case of HTTPS, loops and deadlocks may be created,\n and this should be addressed. In the case of FTP, attacks arise that\n are similar to FTP bounce attacks.\n\n\n\n\n\n\nEastlake Standards Track [Page 20]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n As a result of this issue, it is RECOMMENDED that the\n client_certificate_url extension should have to be specifically\n enabled by a server administrator, rather than be enabled by default.\n It is also RECOMMENDED that URI schemes be enabled by the\n administrator individually, and only a minimal set of schemes be\n enabled. Unusual protocols that offer limited security or whose\n security is not well understood SHOULD be avoided.\n\n As discussed in [RFC3986], URLs that specify ports other than the\n default may cause problems, as may very long URLs (which are more\n likely to be useful in exploiting buffer overflow bugs).\n\n This extension continues to use SHA-1 (as in RFC 4366) and does not\n provide algorithm agility. The property required of SHA-1 in this\n case is second pre-image resistance, not collision resistance.\n Furthermore, even if second pre-image attacks against SHA-1 are found\n in the future, an attack against client_certificate_url would require\n a second pre-image that is accepted as a valid certificate by the\n server and contains the same public key.\n\n Also note that HTTP caching proxies are common on the Internet, and\n some proxies do not check for the latest version of an object\n correctly. If a request using HTTP (or another caching protocol)\n goes through a misconfigured or otherwise broken proxy, the proxy may\n return an out-of-date response.\n\n11.4. Security Considerations for trusted_ca_keys\n\n Potentially, the CA root keys a client possesses could be regarded as\n confidential information. As a result, the CA root key indication\n extension should be used with care.\n\n The use of the SHA-1 certificate hash alternative ensures that each\n certificate is specified unambiguously. This context does not\n require a cryptographic hash function, so the use of SHA-1 is\n considered acceptable, and no algorithm agility is provided.\n\n11.5. Security Considerations for truncated_hmac\n\n It is possible that truncated MACs are weaker than \"un-truncated\"\n MACs. However, no significant weaknesses are currently known or\n expected to exist for HMAC with MD5 or SHA-1, truncated to 80 bits.\n\n Note that the output length of a MAC need not be as long as the\n length of a symmetric cipher key, since forging of MAC values cannot\n be done off-line: in TLS, a single failed MAC guess will cause the\n immediate termination of the TLS session.\n\n\n\n\nEastlake Standards Track [Page 21]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Since the MAC algorithm only takes effect after all handshake\n messages that affect extension parameters have been authenticated by\n the hashes in the Finished messages, it is not possible for an active\n attacker to force negotiation of the truncated HMAC extension where\n it would not otherwise be used (to the extent that the handshake\n authentication is secure). Therefore, in the event that any security\n problems were found with truncated HMAC in the future, if either the\n client or the server for a given session were updated to take the\n problem into account, it would be able to veto use of this extension.\n\n11.6. Security Considerations for status_request\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC2560] for further details.\n\n12. Normative References\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC:\n Keyed-Hashing for Message Authentication\", RFC 2104,\n February 1997.\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and\n C. Adams, \"X.509 Internet Public Key Infrastructure\n Online Certificate Status Protocol - OCSP\", RFC 2560,\n June 1999.\n\n [RFC2585] Housley, R. and P. Hoffman, \"Internet X.509 Public Key\n Infrastructure Operational Protocols: FTP and HTTP\",\n RFC 2585, May 1999.\n\n [RFC2616] Fielding, R., Gettys, J., Mogul, J., Frystyk, H.,\n Masinter, L., Leach, P., and T. Berners-Lee,\n \"Hypertext Transfer Protocol -- HTTP/1.1\", RFC 2616,\n June 1999.\n\n [RFC3986] Berners-Lee, T., Fielding, R., and L. Masinter,\n \"Uniform Resource Identifier (URI): Generic Syntax\",\n STD 66, RFC 3986, January 2005.\n\n\n\nEastlake Standards Track [Page 22]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246, August\n 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation\n List (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC5890] Klensin, J., \"Internationalized Domain Names for\n Applications (IDNA): Definitions and Document\n Framework\", RFC 5890, August 2010.\n\n13. Informative References\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [X509-4th] ITU-T Recommendation X.509 (2000) | ISO/IEC\n 9594-8:2001, \"Information Systems - Open Systems\n Interconnection - The Directory: Public key and\n attribute certificate frameworks\".\n\n [X509-4th-TC1] ITU-T Recommendation X.509(2000) Corrigendum 1(2001) |\n ISO/IEC 9594-8:2001/Cor.1:2002, Technical Corrigendum\n 1 to ISO/IEC 9594:8:2001.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 23]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix A. Changes from RFC 4366\n\n The significant changes between RFC 4366 and this document are\n described below.\n\n RFC 4366 described both general extension mechanisms (for the TLS\n handshake and client and server hellos) as well as specific\n extensions. RFC 4366 was associated with RFC 4346, TLS 1.1. The\n client and server hello extension mechanisms have been moved into RFC\n 5246, TLS 1.2, so this document, which is associated with RFC 5246,\n includes only the handshake extension mechanisms and the specific\n extensions from RFC 4366. RFC 5246 also specifies the unknown\n extension error and new extension specification considerations, so\n that material has been removed from this document.\n\n The Server Name extension now specifies only ASCII representation,\n eliminating UTF-8. It is provided that the ServerNameList can\n contain more than only one name of any particular name_type. If a\n server name is provided but not recognized, the server should either\n continue the handshake without an error or send a fatal error.\n Sending a warning-level message is not recommended because client\n behavior will be unpredictable. Provision was added for the user\n using the server_name extension in deciding whether or not to resume\n a session. Furthermore, this extension should be the same in a\n session resumption request as it was in the full handshake that\n established the session. Such a resumption request must not be\n accepted if the server_name extension is different, but instead a\n full handshake must be done to possibly establish a new session.\n\n The Client Certificate URLs extension has been changed to make the\n presence of a hash mandatory.\n\n For the case of DTLS, the requirement to report an overflow of the\n negotiated maximum fragment length is made conditional on passing\n authentication.\n\n TLS servers are now prohibited from following HTTP redirects when\n retrieving certificates.\n\n The material was also re-organized in minor ways. For example,\n information as to which errors are fatal is moved from the \"Error\n Alerts\" section to the individual extension specifications.\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 24]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix B. Acknowledgements\n\n This document is based on material from RFC 4366 for which the\n authors were S. Blake-Wilson, M. Nystrom, D. Hopwood, J. Mikkelsen,\n and T. Wright. Other contributors include Joseph Salowey, Alexey\n Melnikov, Peter Saint-Andre, and Adrian Farrel.\n\nAuthor's Address\n\n Donald Eastlake 3rd\n Huawei\n 155 Beaver Street\n Milford, MA 01757 USA\n\n Phone: +1-508-333-2270\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 25]\n\nAnswer using only the documents and the scenario in the question. Return exactly the requested JSON object; use \"UNKNOWN\" only per unresolved field.", "context_source_relation_count": 22, "counterfactual_twin": {"byte_end": 447, "byte_start": 361, "char_end": 447, "char_start": 361, "evidence_id": "current_protocol", "parent_source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd", "parent_text_sha256": "d596886622bceee8500b3f3c9d869d947b4e70a2f05ea65fdd75666b46debd81", "parent_value": " ", "provenance_operation": "exclude_exact_source_span", "record_id": "ietf:rfc:8446", "source_manifest_sha256": "287ca4031b223ad818d21781c575905d5ed30eb7e4e0e38fef196cf1ca1865b4", "source_origin": "synthetic_counterfactual", "text_sha256": "98ef2decaa389708a963efbd12018cad07c13f824ea89269ccae34e25ced38b7", "value": "This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol."}, "data_stage": "train_ready", "dependency_class": "long_range_retrieval", "diagnostic_only": true, "difficulty": {"context_tokens": 65493, "evidence_span_tokens": 65114, "max_evidence_distance": 65121, "proof_depth": 2, "state_updates": 10}, "document_context": "Network Working Group J. Salowey\nRequest for Comments: 5077 H. Zhou\nObsoletes: 4507 Cisco Systems\nCategory: Standards Track P. Eronen\n Nokia\n H. Tschofenig\n Nokia Siemens Networks\n January 2008\n\n\n Transport Layer Security (TLS) Session Resumption without\n Server-Side State\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document describes a mechanism that enables the Transport Layer\n Security (TLS) server to resume sessions and avoid keeping per-client\n session state. The TLS server encapsulates the session state into a\n ticket and forwards it to the client. The client can subsequently\n resume a session using the obtained ticket. This document obsoletes\n RFC 4507.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 1]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3. Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . 4\n 3.2. SessionTicket TLS Extension . . . . . . . . . . . . . . . 7\n 3.3. NewSessionTicket Handshake Message . . . . . . . . . . . . 8\n 3.4. Interaction with TLS Session ID . . . . . . . . . . . . . 9\n 4. Recommended Ticket Construction . . . . . . . . . . . . . . . 10\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . 12\n 5.1. Invalidating Sessions . . . . . . . . . . . . . . . . . . 12\n 5.2. Stolen Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.3. Forged Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.4. Denial of Service Attacks . . . . . . . . . . . . . . . . 12\n 5.5. Ticket Protection Key Management . . . . . . . . . . . . . 13\n 5.6. Ticket Lifetime . . . . . . . . . . . . . . . . . . . . . 13\n 5.7. Alternate Ticket Formats and Distribution Schemes . . . . 13\n 5.8. Identity Privacy, Anonymity, and Unlinkability . . . . . . 14\n 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 14\n 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 15\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 15\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 15\n Appendix A. Discussion of Changes to RFC 4507 . . . . . . . . . . 17\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 2]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n1. Introduction\n\n This document defines a way to resume a Transport Layer Security\n (TLS) session without requiring session-specific state at the TLS\n server. This mechanism may be used with any TLS ciphersuite. This\n document applies to both TLS 1.0 defined in [RFC2246], and TLS 1.1\n defined in [RFC4346]. The mechanism makes use of TLS extensions\n defined in [RFC4366] and defines a new TLS message type.\n\n This mechanism is useful in the following situations:\n\n 1. servers that handle a large number of transactions from different\n users\n\n 2. servers that desire to cache sessions for a long time\n\n 3. ability to load balance requests across servers\n\n 4. embedded servers with little memory\n\n This document obsoletes RFC 4507 [RFC4507] to correct an error in the\n encoding that caused the specification to differ from deployed\n implementations. At the time of this writing, there are no known\n implementations that follow the encoding specified in RFC 4507. This\n update to RFC 4507 aligns the document with currently deployed\n implementations. More details of the change are given in Appendix A.\n\n2. Terminology\n\n Within this document, the term 'ticket' refers to a cryptographically\n protected data structure that is created and consumed by the server\n to rebuild session-specific state.\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Protocol\n\n This specification describes a mechanism to distribute encrypted\n session-state information to the client in the form of a ticket and a\n mechanism to present the ticket back to the server. The ticket is\n created by a TLS server and sent to a TLS client. The TLS client\n presents the ticket to the TLS server to resume a session.\n Implementations of this specification are expected to support both\n mechanisms. Other specifications can take advantage of the session\n tickets, perhaps specifying alternative means for distribution or\n selection. For example, a separate specification may describe an\n\n\n\nSalowey, et al. Standards Track [Page 3]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n alternate way to distribute a ticket and use the TLS extension in\n this document to resume the session. This behavior is beyond the\n scope of the document and would need to be described in a separate\n specification.\n\n3.1. Overview\n\n The client indicates that it supports this mechanism by including a\n SessionTicket TLS extension in the ClientHello message. The\n extension will be empty if the client does not already possess a\n ticket for the server. The server sends an empty SessionTicket\n extension to indicate that it will send a new session ticket using\n the NewSessionTicket handshake message. The extension is described\n in Section 3.2.\n\n If the server wants to use this mechanism, it stores its session\n state (such as ciphersuite and master secret) to a ticket that is\n encrypted and integrity-protected by a key known only to the server.\n The ticket is distributed to the client using the NewSessionTicket\n TLS handshake message described in Section 3.3. This message is sent\n during the TLS handshake before the ChangeCipherSpec message, after\n the server has successfully verified the client's Finished message.\n\n Client Server\n\n ClientHello\n (empty SessionTicket extension)-------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1: Message Flow for Full Handshake Issuing New Session Ticket\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 4]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n The client caches this ticket along with the master secret and other\n parameters associated with the current session. When the client\n wishes to resume the session, it includes the ticket in the\n SessionTicket extension within the ClientHello message. Appendix A\n provides a detailed description of the encoding of the extension and\n changes from RFC 4507. The server then decrypts the received ticket,\n verifies the ticket's validity, retrieves the session state from the\n contents of the ticket, and uses this state to resume the session.\n The interaction with the TLS Session ID is described in Section 3.4.\n If the server successfully verifies the client's ticket, then it may\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello.\n\n Client Server\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2: Message Flow for Abbreviated Handshake Using New Session\n Ticket\n\n A recommended ticket format is given in Section 4.\n\n If the server cannot or does not want to honor the ticket, then it\n can initiate a full handshake with the client.\n\n In the case that the server does not wish to issue a new ticket at\n this time, it just completes the handshake without including a\n SessionTicket extension or NewSessionTicket handshake message. This\n is shown below (this flow is identical to Figure 1 in RFC 4346,\n except for the SessionTicket extension in the first message):\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 5]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 3: Message Flow for Server Completing Full Handshake Without\n Issuing New Session Ticket\n\n It is also permissible to have an exchange similar to Figure 3 using\n the abbreviated handshake defined in Figure 2 of RFC 4346, where the\n client uses the SessionTicket extension to resume the session, but\n the server does not wish to issue a new ticket, and therefore does\n not send a SessionTicket extension.\n\n If the server rejects the ticket, it may still wish to issue a new\n ticket after performing the full handshake as shown below (this flow\n is identical to Figure 1, except the SessionTicket extension in the\n ClientHello is not empty):\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 6]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 4: Message Flow for Server Rejecting Ticket, Performing Full\n Handshake, and Issuing New Session Ticket\n\n3.2. SessionTicket TLS Extension\n\n The SessionTicket TLS extension is based on [RFC4366]. The format of\n the ticket is an opaque structure used to carry session-specific\n state information. This extension may be sent in the ClientHello and\n ServerHello.\n\n If the client possesses a ticket that it wants to use to resume a\n session, then it includes the ticket in the SessionTicket extension\n in the ClientHello. If the client does not have a ticket and is\n prepared to receive one in the NewSessionTicket handshake message,\n then it MUST include a zero-length ticket in the SessionTicket\n extension. If the client is not prepared to receive a ticket in the\n NewSessionTicket handshake message, then it MUST NOT include a\n SessionTicket extension unless it is sending a non-empty ticket it\n received through some other means from the server.\n\n The server uses a zero-length SessionTicket extension to indicate to\n the client that it will send a new session ticket using the\n NewSessionTicket handshake message described in Section 3.3. The\n server MUST send this extension in the ServerHello if it wishes to\n issue a new ticket to the client using the NewSessionTicket handshake\n message. The server MUST NOT send this extension if it does not\n receive one in the ClientHello.\n\n\n\n\nSalowey, et al. Standards Track [Page 7]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n If the server fails to verify the ticket, then it falls back to\n performing a full handshake. If the ticket is accepted by the server\n but the handshake fails, the client SHOULD delete the ticket.\n\n The SessionTicket extension has been assigned the number 35. The\n extension_data field of SessionTicket extension contains the ticket.\n\n3.3. NewSessionTicket Handshake Message\n\n This message is sent by the server during the TLS handshake before\n the ChangeCipherSpec message. This message MUST be sent if the\n server included a SessionTicket extension in the ServerHello. This\n message MUST NOT be sent if the server did not include a\n SessionTicket extension in the ServerHello. This message is included\n in the hash used to create and verify the Finished message. In the\n case of a full handshake, the server MUST verify the client's\n Finished message before sending the ticket. The client MUST NOT\n treat the ticket as valid until it has verified the server's Finished\n message. If the server determines that it does not want to include a\n ticket after it has included the SessionTicket extension in the\n ServerHello, then it sends a zero-length ticket in the\n NewSessionTicket handshake message.\n\n If the server successfully verifies the client's ticket, then it MAY\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello in the abbreviated handshake. The client\n should start using the new ticket as soon as possible after it\n verifies the server's Finished message for new connections. Note\n that since the updated ticket is issued before the handshake\n completes, it is possible that the client may not put the new ticket\n into use before it initiates new connections. The server MUST NOT\n assume that the client actually received the updated ticket until it\n successfully verifies the client's Finished message.\n\n The NewSessionTicket handshake message has been assigned the number 4\n and its definition is given at the end of this section. The\n ticket_lifetime_hint field contains a hint from the server about how\n long the ticket should be stored. The value indicates the lifetime\n in seconds as a 32-bit unsigned integer in network byte order\n relative to when the ticket is received. A value of zero is reserved\n to indicate that the lifetime of the ticket is unspecified. A client\n SHOULD delete the ticket and associated state when the time expires.\n It MAY delete the ticket earlier based on local policy. A server MAY\n treat a ticket as valid for a shorter or longer period of time than\n what is stated in the ticket_lifetime_hint.\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 8]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n HandshakeType msg_type;\n uint24 length;\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case session_ticket: NewSessionTicket; /* NEW */\n } body;\n } Handshake;\n\n\n struct {\n uint32 ticket_lifetime_hint;\n opaque ticket<0..2^16-1>;\n } NewSessionTicket;\n\n3.4. Interaction with TLS Session ID\n\n If a server is planning on issuing a session ticket to a client that\n does not present one, it SHOULD include an empty Session ID in the\n ServerHello. If the server rejects the ticket and falls back to the\n full handshake then it may include a non-empty Session ID to indicate\n its support for stateful session resumption. If the client receives\n a session ticket from the server, then it discards any Session ID\n that was sent in the ServerHello.\n\n When presenting a ticket, the client MAY generate and include a\n Session ID in the TLS ClientHello. If the server accepts the ticket\n and the Session ID is not empty, then it MUST respond with the same\n Session ID present in the ClientHello. This allows the client to\n easily differentiate when the server is resuming a session from when\n it is falling back to a full handshake. Since the client generates a\n Session ID, the server MUST NOT rely upon the Session ID having a\n particular value when validating the ticket. If a ticket is\n presented by the client, the server MUST NOT attempt to use the\n Session ID in the ClientHello for stateful session resumption.\n Alternatively, the client MAY include an empty Session ID in the\n ClientHello. In this case, the client ignores the Session ID sent in\n the ServerHello and determines if the server is resuming a session by\n the subsequent handshake messages.\n\n\n\nSalowey, et al. Standards Track [Page 9]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n4. Recommended Ticket Construction\n\n This section describes a recommended format and protection for the\n ticket. Note that the ticket is opaque to the client, so the\n structure is not subject to interoperability concerns, and\n implementations may diverge from this format. If implementations do\n diverge from this format, they must take security concerns seriously.\n Clients MUST NOT examine the ticket under the assumption that it\n complies with this document.\n\n The server uses two different keys: one 128-bit key for Advanced\n Encryption Standard (AES) [AES] in Cipher Block Chaining (CBC) mode\n [CBC] encryption and one 256-bit key for HMAC-SHA-256 [RFC4634].\n\n The ticket is structured as follows:\n\n struct {\n opaque key_name[16];\n opaque iv[16];\n opaque encrypted_state<0..2^16-1>;\n opaque mac[32];\n } ticket;\n\n Here, key_name serves to identify a particular set of keys used to\n protect the ticket. It enables the server to easily recognize\n tickets it has issued. The key_name should be randomly generated to\n avoid collisions between servers. One possibility is to generate new\n random keys and key_name every time the server is started.\n\n The actual state information in encrypted_state is encrypted using\n 128-bit AES in CBC mode with the given IV. The Message\n Authentication Code (MAC) is calculated using HMAC-SHA-256 over\n key_name (16 octets) and IV (16 octets), followed by the length of\n the encrypted_state field (2 octets) and its contents (variable\n length).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 10]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n ProtocolVersion protocol_version;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n opaque master_secret[48];\n ClientIdentity client_identity;\n uint32 timestamp;\n } StatePlaintext;\n\n enum {\n anonymous(0),\n certificate_based(1),\n psk(2)\n } ClientAuthenticationType;\n\n struct {\n ClientAuthenticationType client_authentication_type;\n select (ClientAuthenticationType) {\n case anonymous: struct {};\n case certificate_based:\n ASN.1Cert certificate_list<0..2^24-1>;\n case psk:\n opaque psk_identity<0..2^16-1>; /* from [RFC4279] */\n };\n } ClientIdentity;\n\n The structure StatePlaintext stores the TLS session state including\n the master_secret. The timestamp within this structure allows the\n TLS server to expire tickets. To cover the authentication and key\n exchange protocols provided by TLS, the ClientIdentity structure\n contains the authentication type of the client used in the initial\n exchange (see ClientAuthenticationType). To offer the TLS server\n with the same capabilities for authentication and authorization, a\n certificate list is included in case of public-key-based\n authentication. The TLS server is therefore able to inspect a number\n of different attributes within these certificates. A specific\n implementation might choose to store a subset of this information or\n additional information. Other authentication mechanisms, such as\n Kerberos [RFC2712], would require different client identity data.\n Other TLS extensions may require the inclusion of additional data in\n the StatePlaintext structure.\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 11]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5. Security Considerations\n\n This section addresses security issues related to the usage of a\n ticket. Tickets must be authenticated and encrypted to prevent\n modification or eavesdropping by an attacker. Several attacks\n described below will be possible if this is not carefully done.\n\n Implementations should take care to ensure that the processing of\n tickets does not increase the chance of denial of service as\n described below.\n\n5.1. Invalidating Sessions\n\n The TLS specification requires that TLS sessions be invalidated when\n errors occur. [CSSC] discusses the security implications of this in\n detail. In the analysis within this paper, failure to invalidate\n sessions does not pose a security risk. This is because the TLS\n handshake uses a non-reversible function to derive keys for a session\n so information about one session does not provide an advantage to\n attack the master secret or a different session. If a session\n invalidation scheme is used, the implementation should verify the\n integrity of the ticket before using the contents to invalidate a\n session to ensure that an attacker cannot invalidate a chosen\n session.\n\n5.2. Stolen Tickets\n\n An eavesdropper or man-in-the-middle may obtain the ticket and\n attempt to use it to establish a session with the server; however,\n since the ticket is encrypted and the attacker does not know the\n secret key, a stolen ticket does not help an attacker resume a\n session. A TLS server MUST use strong encryption and integrity\n protection for the ticket to prevent an attacker from using a brute\n force mechanism to obtain the ticket's contents.\n\n5.3. Forged Tickets\n\n A malicious user could forge or alter a ticket in order to resume a\n session, to extend its lifetime, to impersonate another user, or to\n gain additional privileges. This attack is not possible if the\n ticket is protected using a strong integrity protection algorithm\n such as a keyed HMAC-SHA-256.\n\n5.4. Denial of Service Attacks\n\n The key_name field defined in the recommended ticket format helps the\n server efficiently reject tickets that it did not issue. However, an\n adversary could store or generate a large number of tickets to send\n\n\n\nSalowey, et al. Standards Track [Page 12]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n to the TLS server for verification. To minimize the possibility of a\n denial of service, the verification of the ticket should be\n lightweight (e.g., using efficient symmetric key cryptographic\n algorithms).\n\n5.5. Ticket Protection Key Management\n\n A full description of the management of the keys used to protect the\n ticket is beyond the scope of this document. A list of RECOMMENDED\n practices is given below.\n\n o The keys should be generated securely following the randomness\n recommendations in [RFC4086].\n\n o The keys and cryptographic protection algorithms should be at\n least 128 bits in strength. Some ciphersuites and applications\n may require cryptographic protection greater than 128 bits in\n strength.\n\n o The keys should not be used for any purpose other than generating\n and verifying tickets.\n\n o The keys should be changed regularly.\n\n o The keys should be changed if the ticket format or cryptographic\n protection algorithms change.\n\n5.6. Ticket Lifetime\n\n The TLS server controls the lifetime of the ticket. Servers\n determine the acceptable lifetime based on the operational and\n security requirements of the environments in which they are deployed.\n The ticket lifetime may be longer than the 24-hour lifetime\n recommended in [RFC4346]. TLS clients may be given a hint of the\n lifetime of the ticket. Since the lifetime of a ticket may be\n unspecified, a client has its own local policy that determines when\n it discards tickets.\n\n5.7. Alternate Ticket Formats and Distribution Schemes\n\n If the ticket format or distribution scheme defined in this document\n is not used, then great care must be taken in analyzing the security\n of the solution. In particular, if confidential information, such as\n a secret key, is transferred to the client, it MUST be done using\n secure communication so as to prevent attackers from obtaining or\n modifying the key. Also, the ticket MUST have its integrity and\n confidentiality protected with strong cryptographic techniques to\n prevent a breach in the security of the system.\n\n\n\nSalowey, et al. Standards Track [Page 13]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5.8. Identity Privacy, Anonymity, and Unlinkability\n\n This document mandates that the content of the ticket is\n confidentiality protected in order to avoid leakage of its content,\n such as user-relevant information. As such, it prevents disclosure\n of potentially sensitive information carried within the ticket.\n\n The initial handshake exchange, which was used to obtain the ticket,\n might not provide identity confidentiality of the client based on the\n properties of TLS. Another relevant security threat is the ability\n for an on-path adversary to observe multiple TLS handshakes where the\n same ticket is used, therefore concluding they belong to the same\n communication endpoints. Application designers that use the ticket\n mechanism described in this document should consider that\n unlinkability [ANON] is not necessarily provided.\n\n While a full discussion of these topics is beyond the scope of this\n document, it should be noted that it is possible to issue a ticket\n using a TLS renegotiation handshake that occurs after a secure tunnel\n has been established by a previous handshake. This may help address\n some privacy and unlinkability issues in some environments.\n\n6. Acknowledgements\n\n The authors would like to thank the following people for their help\n with preparing and reviewing this document: Eric Rescorla, Mohamad\n Badra, Tim Dierks, Nelson Bolyard, Nancy Cam-Winget, David McGrew,\n Rob Dugal, Russ Housley, Amir Herzberg, Bernard Aboba, and members of\n the TLS working group.\n\n [CSSC] describes a solution that is very similar to the one described\n in this document and gives a detailed analysis of the security\n considerations involved. [RFC2712] describes a mechanism for using\n Kerberos [RFC4120] in TLS ciphersuites, which helped inspire the use\n of tickets to avoid server state. [RFC4851] makes use of a similar\n mechanism to avoid maintaining server state for the cryptographic\n tunnel. [SC97] also investigates the concept of stateless sessions.\n\n The authors would also like to thank Jan Nordqvist, who found the\n encoding error in RFC 4507, corrected by this document. In addition\n Nagendra Modadugu, Wan-Teh Chang, and Michael D'Errico provided\n useful feedback during the review of this document.\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 14]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n7. IANA Considerations\n\n IANA has assigned a TLS extension number of 35 to the SessionTicket\n TLS extension from the TLS registry of ExtensionType values defined\n in [RFC4366].\n\n IANA has assigned a TLS HandshakeType number 4 to the\n NewSessionTicket handshake type from the TLS registry of\n HandshakeType values defined in [RFC4346].\n\n This document does not require any actions or assignments from IANA.\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2246] Dierks, T. and C. Allen, \"The TLS Protocol Version 1.0\",\n RFC 2246, January 1999.\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346, April 2006.\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, April 2006.\n\n [RFC4507] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 4507, May 2006.\n\n8.2. Informative References\n\n [AES] National Institute of Standards and Technology, \"Advanced\n Encryption Standard (AES)\", Federal Information Processing\n Standards (FIPS) Publication 197, November 2001.\n\n [ANON] Pfitzmann, A. and M. Hansen, \"Anonymity, Unlinkability,\n Unobservability, Pseudonymity, and Identity Management - A\n Consolidated Proposal for Terminology\", http://\n dud.inf.tu-dresden.de/literatur/\n Anon_Terminology_v0.26-1.pdf Version 0.26, December 2005.\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 15]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n [CBC] National Institute of Standards and Technology,\n \"Recommendation for Block Cipher Modes of Operation -\n Methods and Techniques\", NIST Special Publication 800-38A,\n December 2001.\n\n [CSSC] Shacham, H., Boneh, D., and E. Rescorla, \"Client-side\n caching for TLS\", Transactions on Information and System\n Security (TISSEC) , Volume 7, Issue 4, November 2004.\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [RFC4086] Eastlake, D., Schiller, J., and S. Crocker, \"Randomness\n Requirements for Security\", BCP 106, RFC 4086, June 2005.\n\n [RFC4120] Neuman, C., Yu, T., Hartman, S., and K. Raeburn, \"The\n Kerberos Network Authentication Service (V5)\", RFC 4120,\n July 2005.\n\n [RFC4279] Eronen, P. and H. Tschofenig, \"Pre-Shared Key Ciphersuites\n for Transport Layer Security (TLS)\", RFC 4279,\n December 2005.\n\n [RFC4634] Eastlake, D. and T. Hansen, \"US Secure Hash Algorithms\n (SHA and HMAC-SHA)\", RFC 4634, July 2006.\n\n [RFC4851] Cam-Winget, N., McGrew, D., Salowey, J., and H. Zhou, \"The\n Flexible Authentication via Secure Tunneling Extensible\n Authentication Protocol Method (EAP-FAST)\", RFC 4851,\n May 2007.\n\n [SC97] Aura, T. and P. Nikander, \"Stateless Connections\",\n Proceedings of the First International Conference on\n Information and Communication Security (ICICS '97) , 1997.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 16]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAppendix A. Discussion of Changes to RFC 4507\n\n RFC 4507 [RFC4507] defines a mechanism to resume a TLS session\n without maintaining server side state by specifying an encrypted\n ticket that is maintained on the client. The client presents this\n ticket to the server in a SessionTicket hello extension. The\n encoding in RFC 4507 used the XDR style encoding specified in TLS\n [RFC4346].\n\n An error in the encoding caused the specification to differ from\n deployed implementations. At the time of this writing there are no\n known implementations that follow the encoding specified in RFC 4507.\n This update to RFC 4507 aligns the document with these currently\n deployed implementations.\n\n Erroneous encoding in RFC 4507 resulted in two length fields; one for\n the extension contents and one for the ticket itself. Hence, for a\n ticket that is 256 bytes long and begins with the hex value FF FF,\n the encoding of the extension would be as follows according to RFC\n 4507:\n\n 00 23 Ticket Extension type 35\n 01 02 Length of extension contents\n 01 00 Length of ticket\n FF FF .. .. Actual ticket\n\n The update proposed in this document reflects what implementations\n actually encode, namely it removes the redundant length field. So,\n for a ticket that is 256 bytes long and begins with the hex value FF\n FF, the encoding of the extension would be as follows according to\n this update:\n\n 00 23 Extension type 35\n 01 00 Length of extension contents (ticket)\n FF FF .. .. Actual ticket\n\n A server implemented according to RFC 4507 receiving a ticket\n extension from a client conforming to this document would interpret\n the first two bytes of the ticket as the length of this ticket. This\n will result in either an inconsistent length field or in the\n processing of a ticket missing the first two bytes. In the first\n case, the server should reject the request based on a malformed\n length. In the second case, the server should reject the ticket\n based on a malformed ticket, incorrect key version, or failed\n decryption. A server implementation based on this update receiving\n an RFC 4507 extension would interpret the first length field as the\n\n\n\n\n\nSalowey, et al. Standards Track [Page 17]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n length of the ticket and include the second two length bytes as the\n first bytes in the ticket, resulting in the ticket being rejected\n based on a malformed ticket, incorrect key version, or failed\n decryption.\n\n Note that the encoding of an empty SessionTicket extension was\n ambiguous in RFC 4507. An RFC 4507 implementation may have encoded\n it as:\n\n 00 23 Extension type 35\n 00 02 Length of extension contents\n 00 00 Length of ticket\n\n or it may have encoded it the same way as this update:\n\n 00 23 Extension type 35\n 00 00 Length of extension contents\n\n A server wishing to support RFC 4507 clients should respond to an\n empty SessionTicket extension encoded the same way as it received it.\n\n A server implementation can construct tickets such that it can detect\n an RFC 4507 implementation, if one existed, by including a cookie at\n the beginning of the tickets that can be differentiated from a valid\n length. For example, if an implementation constructed tickets to\n start with the hex values FF FF, then it could determine where the\n ticket begins and determine the length correctly from the type of\n length fields present.\n\n This document makes a few additional changes to RFC 4507 listed\n below.\n\n o Clarifying that the server can allow session resumption using a\n ticket without issuing a new ticket in Section 3.1.\n\n o Clarifying that the lifetime is relative to when the ticket is\n received in section 3.3.\n\n o Clarifying that the NewSessionTicket handshake message is included\n in the hash generated for the Finished messages in Section 3.3.\n\n o Clarifying the interaction with TLS Session ID in Section 3.4.\n\n o Recommending the use of SHA-256 for the integrity protection of\n the ticket in Section 4.\n\n o Clarifying that additional data can be included in the\n StatePlaintext structure in Section 4.\n\n\n\nSalowey, et al. Standards Track [Page 18]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAuthors' Addresses\n\n Joseph Salowey\n Cisco Systems\n 2901 3rd Ave\n Seattle, WA 98121\n US\n\n EMail: [redacted-email]\n\n\n Hao Zhou\n Cisco Systems\n 4125 Highlander Parkway\n Richfield, OH 44286\n US\n\n EMail: [redacted-email]\n\n\n Pasi Eronen\n Nokia Research Center\n P.O. Box 407\n FIN-00045 Nokia Group\n Finland\n\n EMail: [redacted-email]\n\n\n Hannes Tschofenig\n Nokia Siemens Networks\n Otto-Hahn-Ring 6\n Munich, Bayern 81739\n Germany\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 19]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nFull Copyright Statement\n\n Copyright (C) The IETF Trust (2008).\n\n This document is subject to the rights, licenses and restrictions\n contained in BCP 78, and except as set forth therein, the authors\n retain all their rights.\n\n This document and the information contained herein are provided on an\n \"AS IS\" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\n OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\n THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\n OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\n THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\n WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\n\nIntellectual Property\n\n The IETF takes no position regarding the validity or scope of any\n Intellectual Property Rights or other rights that might be claimed to\n pertain to the implementation or use of the technology described in\n this document or the extent to which any license under such rights\n might or might not be available; nor does it represent that it has\n made any independent effort to identify any such rights. Information\n on the procedures with respect to rights in RFC documents can be\n found in BCP 78 and BCP 79.\n\n Copies of IPR disclosures made to the IETF Secretariat and any\n assurances of licenses to be made available, or the result of an\n attempt made to obtain a general license or permission for the use of\n such proprietary rights by implementers or users of this\n specification can be obtained from the IETF on-line IPR repository at\n http://www.ietf.org/ipr.\n\n The IETF invites any interested party to bring to its attention any\n copyrights, patents or patent applications, or other proprietary\n rights that may cover technology that may be required to implement\n this standard. Please address the information to the IETF at\n ietf-ipr@ietf.org.\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 20]\n\n===== DOCUMENT =====\n\n\nRFC 8446 TLS August 2018\n\n\n Sam Scott\n Royal Holloway, University of London\n [redacted-email]\n\n Thomas Shrimpton\n University of Florida\n [redacted-email]\n\n Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n Brian Smith\n Independent\n [redacted-email]\n\n Brian Sniffen\n Akamai Technologies\n [redacted-email]\n\n Nick Sullivan\n Cloudflare Inc.\n [redacted-email]\n\n Bjoern Tackmann\n University of California, San Diego\n [redacted-email]\n\n Tim Taubert\n Mozilla\n [redacted-email]\n\n Martin Thomson\n Mozilla\n [redacted-email]\n\n Hannes Tschofenig\n Arm Limited\n [redacted-email]\n\n Sean Turner\n sn3rd\n [redacted-email]\n\n Steven Valdez\n Google\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 159]\n\f\nRFC 8446 TLS August 2018\n\n\n Filippo Valsorda\n Cloudflare Inc.\n [redacted-email]\n\n Thyla van der Merwe\n Royal Holloway, University of London\n [redacted-email]\n\n Victor Vasiliev\n Google\n [redacted-email]\n\n Hoeteck Wee\n Ecole Normale Superieure, Paris\n [redacted-email]\n\n Tom Weinstein\n\n David Wong\n NCC Group\n [redacted-email]\n\n Christopher A. Wood\n Apple Inc.\n [redacted-email]\n\n Tim Wright\n Vodafone\n [redacted-email]\n\n Peter Wu\n Independent\n [redacted-email]\n\n Kazu Yamamoto\n Internet Initiative Japan Inc.\n [redacted-email]\n\nAuthor's Address\n\n Eric Rescorla\n Mozilla\n\n Email: [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 160]\n\n===== DOCUMENT =====\n\nNetwork Working Group T. Dierks\nRequest for Comments: 5246 Independent\nObsoletes: 3268, 4346, 4366 E. Rescorla\nUpdates: 4492 RTFM, Inc.\nCategory: Standards Track August 2008\n\n\n The Transport Layer Security (TLS) Protocol\n Version 1.2\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document specifies Version 1.2 of the Transport Layer Security\n (TLS) protocol. The TLS protocol provides communications security\n over the Internet. The protocol allows client/server applications to\n communicate in a way that is designed to prevent eavesdropping,\n tampering, or message forgery.\n\nTable of Contents\n\n 1. Introduction ....................................................4\n 1.1. Requirements Terminology ...................................5\n 1.2. Major Differences from TLS 1.1 .............................5\n 2. Goals ...........................................................6\n 3. Goals of This Document ..........................................7\n 4. Presentation Language ...........................................7\n 4.1. Basic Block Size ...........................................7\n 4.2. Miscellaneous ..............................................8\n 4.3. Vectors ....................................................8\n 4.4. Numbers ....................................................9\n 4.5. Enumerateds ................................................9\n 4.6. Constructed Types .........................................10\n 4.6.1. Variants ...........................................10\n 4.7. Cryptographic Attributes ..................................12\n 4.8. Constants .................................................14\n 5. HMAC and the Pseudorandom Function .............................14\n 6. The TLS Record Protocol ........................................15\n 6.1. Connection States .........................................16\n 6.2. Record Layer ..............................................19\n 6.2.1. Fragmentation ......................................19\n\n\n\nDierks & Rescorla Standards Track [Page 1]\n\f\nRFC 5246 TLS August 2008\n\n\n 6.2.2. Record Compression and Decompression ...............20\n 6.2.3. Record Payload Protection ..........................21\n 6.2.3.1. Null or Standard Stream Cipher ............22\n 6.2.3.2. CBC Block Cipher ..........................22\n 6.2.3.3. AEAD Ciphers ..............................24\n 6.3. Key Calculation ...........................................25\n 7. The TLS Handshaking Protocols ..................................26\n 7.1. Change Cipher Spec Protocol ...............................27\n 7.2. Alert Protocol ............................................28\n 7.2.1. Closure Alerts .....................................29\n 7.2.2. Error Alerts .......................................30\n 7.3. Handshake Protocol Overview ...............................33\n 7.4. Handshake Protocol ........................................37\n 7.4.1. Hello Messages .....................................38\n 7.4.1.1. Hello Request .............................38\n 7.4.1.2. Client Hello ..............................39\n 7.4.1.3. Server Hello ..............................42\n 7.4.1.4. Hello Extensions ..........................44\n 7.4.1.4.1. Signature Algorithms ...........45\n 7.4.2. Server Certificate .................................47\n 7.4.3. Server Key Exchange Message ........................50\n 7.4.4. Certificate Request ................................53\n 7.4.5. Server Hello Done ..................................55\n 7.4.6. Client Certificate .................................55\n 7.4.7. Client Key Exchange Message ........................57\n 7.4.7.1. RSA-Encrypted Premaster Secret Message ....58\n 7.4.7.2. Client Diffie-Hellman Public Value ........61\n 7.4.8. Certificate Verify .................................62\n 7.4.9. Finished ...........................................63\n 8. Cryptographic Computations .....................................64\n 8.1. Computing the Master Secret ...............................64\n 8.1.1. RSA ................................................65\n 8.1.2. Diffie-Hellman .....................................65\n 9. Mandatory Cipher Suites ........................................65\n 10. Application Data Protocol .....................................65\n 11. Security Considerations .......................................65\n 12. IANA Considerations ...........................................65\n Appendix A. Protocol Data Structures and Constant Values ..........68\n A.1. Record Layer ..............................................68\n A.2. Change Cipher Specs Message ...............................69\n A.3. Alert Messages ............................................69\n A.4. Handshake Protocol ........................................70\n A.4.1. Hello Messages .....................................71\n A.4.2. Server Authentication and Key Exchange Messages ....72\n A.4.3. Client Authentication and Key Exchange Messages ....74\n A.4.4. Handshake Finalization Message .....................74\n A.5. The Cipher Suite ..........................................75\n A.6. The Security Parameters ...................................77\n\n\n\nDierks & Rescorla Standards Track [Page 2]\n\f\nRFC 5246 TLS August 2008\n\n\n A.7. Changes to RFC 4492 .......................................78\n Appendix B. Glossary ..............................................78\n Appendix C. Cipher Suite Definitions ..............................83\n Appendix D. Implementation Notes ..................................85\n D.1. Random Number Generation and Seeding ......................85\n D.2. Certificates and Authentication ...........................85\n D.3. Cipher Suites .............................................85\n D.4. Implementation Pitfalls ...................................85\n Appendix E. Backward Compatibility ................................87\n E.1. Compatibility with TLS 1.0/1.1 and SSL 3.0 ................87\n E.2. Compatibility with SSL 2.0 ................................88\n E.3. Avoiding Man-in-the-Middle Version Rollback ...............90\n Appendix F. Security Analysis .....................................91\n F.1. Handshake Protocol ........................................91\n F.1.1. Authentication and Key Exchange ....................91\n F.1.1.1. Anonymous Key Exchange ....................91\n F.1.1.2. RSA Key Exchange and Authentication .......92\n F.1.1.3. Diffie-Hellman Key Exchange with\n Authentication ............................92\n F.1.2. Version Rollback Attacks ...........................93\n F.1.3. Detecting Attacks Against the Handshake Protocol ...94\n F.1.4. Resuming Sessions ..................................94\n F.2. Protecting Application Data ...............................94\n F.3. Explicit IVs ..............................................95\n F.4. Security of Composite Cipher Modes ........................95\n F.5. Denial of Service .........................................96\n F.6. Final Notes ...............................................96\n Normative References ..............................................97\n Informative References ............................................98\n Working Group Information ........................................101\n Contributors .....................................................101\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 3]\n\f\nRFC 5246 TLS August 2008\n\n\n1. Introduction\n\n The primary goal of the TLS protocol is to provide privacy and data\n integrity between two communicating applications. The protocol is\n composed of two layers: the TLS Record Protocol and the TLS Handshake\n Protocol. At the lowest level, layered on top of some reliable\n transport protocol (e.g., TCP [TCP]), is the TLS Record Protocol.\n The TLS Record Protocol provides connection security that has two\n basic properties:\n\n - The connection is private. Symmetric cryptography is used for\n data encryption (e.g., AES [AES], RC4 [SCH], etc.). The keys for\n this symmetric encryption are generated uniquely for each\n connection and are based on a secret negotiated by another\n protocol (such as the TLS Handshake Protocol). The Record\n Protocol can also be used without encryption.\n\n - The connection is reliable. Message transport includes a message\n integrity check using a keyed MAC. Secure hash functions (e.g.,\n SHA-1, etc.) are used for MAC computations. The Record Protocol\n can operate without a MAC, but is generally only used in this mode\n while another protocol is using the Record Protocol as a transport\n for negotiating security parameters.\n\n The TLS Record Protocol is used for encapsulation of various higher-\n level protocols. One such encapsulated protocol, the TLS Handshake\n Protocol, allows the server and client to authenticate each other and\n to negotiate an encryption algorithm and cryptographic keys before\n the application protocol transmits or receives its first byte of\n data. The TLS Handshake Protocol provides connection security that\n has three basic properties:\n\n - The peer's identity can be authenticated using asymmetric, or\n public key, cryptography (e.g., RSA [RSA], DSA [DSS], etc.). This\n authentication can be made optional, but is generally required for\n at least one of the peers.\n\n - The negotiation of a shared secret is secure: the negotiated\n secret is unavailable to eavesdroppers, and for any authenticated\n connection the secret cannot be obtained, even by an attacker who\n can place himself in the middle of the connection.\n\n - The negotiation is reliable: no attacker can modify the\n negotiation communication without being detected by the parties to\n the communication.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 4]\n\f\nRFC 5246 TLS August 2008\n\n\n One advantage of TLS is that it is application protocol independent.\n Higher-level protocols can layer on top of the TLS protocol\n transparently. The TLS standard, however, does not specify how\n protocols add security with TLS; the decisions on how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n1.1. Requirements Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [REQ].\n\n1.2. Major Differences from TLS 1.1\n\n This document is a revision of the TLS 1.1 [TLS1.1] protocol which\n contains improved flexibility, particularly for negotiation of\n cryptographic algorithms. The major changes are:\n\n - The MD5/SHA-1 combination in the pseudorandom function (PRF) has\n been replaced with cipher-suite-specified PRFs. All cipher suites\n in this document use P_SHA256.\n\n - The MD5/SHA-1 combination in the digitally-signed element has been\n replaced with a single hash. Signed elements now include a field\n that explicitly specifies the hash algorithm used.\n\n - Substantial cleanup to the client's and server's ability to\n specify which hash and signature algorithms they will accept.\n Note that this also relaxes some of the constraints on signature\n and hash algorithms from previous versions of TLS.\n\n - Addition of support for authenticated encryption with additional\n data modes.\n\n - TLS Extensions definition and AES Cipher Suites were merged in\n from external [TLSEXT] and [TLSAES].\n\n - Tighter checking of EncryptedPreMasterSecret version numbers.\n\n - Tightened up a number of requirements.\n\n - Verify_data length now depends on the cipher suite (default is\n still 12).\n\n - Cleaned up description of Bleichenbacher/Klima attack defenses.\n\n\n\n\nDierks & Rescorla Standards Track [Page 5]\n\f\nRFC 5246 TLS August 2008\n\n\n - Alerts MUST now be sent in many cases.\n\n - After a certificate_request, if no certificates are available,\n clients now MUST send an empty certificate list.\n\n - TLS_RSA_WITH_AES_128_CBC_SHA is now the mandatory to implement\n cipher suite.\n\n - Added HMAC-SHA256 cipher suites.\n\n - Removed IDEA and DES cipher suites. They are now deprecated and\n will be documented in a separate document.\n\n - Support for the SSLv2 backward-compatible hello is now a MAY, not\n a SHOULD, with sending it a SHOULD NOT. Support will probably\n become a SHOULD NOT in the future.\n\n - Added limited \"fall-through\" to the presentation language to allow\n multiple case arms to have the same encoding.\n\n - Added an Implementation Pitfalls sections\n\n - The usual clarifications and editorial work.\n\n2. Goals\n\n The goals of the TLS protocol, in order of priority, are as follows:\n\n 1. Cryptographic security: TLS should be used to establish a secure\n connection between two parties.\n\n 2. Interoperability: Independent programmers should be able to\n develop applications utilizing TLS that can successfully exchange\n cryptographic parameters without knowledge of one another's code.\n\n 3. Extensibility: TLS seeks to provide a framework into which new\n public key and bulk encryption methods can be incorporated as\n necessary. This will also accomplish two sub-goals: preventing\n the need to create a new protocol (and risking the introduction of\n possible new weaknesses) and avoiding the need to implement an\n entire new security library.\n\n 4. Relative efficiency: Cryptographic operations tend to be highly\n CPU intensive, particularly public key operations. For this\n reason, the TLS protocol has incorporated an optional session\n caching scheme to reduce the number of connections that need to be\n established from scratch. Additionally, care has been taken to\n reduce network activity.\n\n\n\nDierks & Rescorla Standards Track [Page 6]\n\f\nRFC 5246 TLS August 2008\n\n\n3. Goals of This Document\n\n This document and the TLS protocol itself are based on the SSL 3.0\n Protocol Specification as published by Netscape. The differences\n between this protocol and SSL 3.0 are not dramatic, but they are\n significant enough that the various versions of TLS and SSL 3.0 do\n not interoperate (although each protocol incorporates a mechanism by\n which an implementation can back down to prior versions). This\n document is intended primarily for readers who will be implementing\n the protocol and for those doing cryptographic analysis of it. The\n specification has been written with this in mind, and it is intended\n to reflect the needs of those two groups. For that reason, many of\n the algorithm-dependent data structures and rules are included in the\n body of the text (as opposed to in an appendix), providing easier\n access to them.\n\n This document is not intended to supply any details of service\n definition or of interface definition, although it does cover select\n areas of policy as they are required for the maintenance of solid\n security.\n\n4. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used. The syntax draws from\n several sources in its structure. Although it resembles the\n programming language \"C\" in its syntax and XDR [XDR] in both its\n syntax and intent, it would be risky to draw too many parallels. The\n purpose of this presentation language is to document TLS only; it has\n no general application beyond that particular goal.\n\n4.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 7]\n\f\nRFC 5246 TLS August 2008\n\n\n4.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" double\n brackets.\n\n Single-byte entities containing uninterpreted data are of type\n opaque.\n\n4.3. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* 3 consecutive 3 byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n In the following example, mandatory is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 4.4). On the\n other hand, longer can represent up to 800 bytes of data, or 400\n uint16 elements, and it may be empty. Its encoding will include a\n\n\n\nDierks & Rescorla Standards Track [Page 8]\n\f\nRFC 5246 TLS August 2008\n\n\n two-byte actual length field prepended to the vector. The length of\n an encoded vector must be an even multiple of the length of a single\n element (for example, a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is 2 bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n4.4. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are formed from fixed-length series of bytes\n concatenated as described in Section 4.1 and are also unsigned. The\n following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are stored in\n network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n Note that in some cases (e.g., DH parameters) it is necessary to\n represent integers as opaque vectors. In such cases, they are\n represented as unsigned integers (i.e., leading zero octets are not\n required even if the most significant bit is set).\n\n4.5. Enumerateds\n\n An additional sparse data type is available called enum. A field of\n type enum can only assume the values declared in the definition.\n Each definition is a different type. Only enumerateds of the same\n type may be assigned or compared. Every element of an enumerated\n must be assigned a value, as demonstrated in the following example.\n Since the elements of the enumerated are not ordered, they can be\n assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n\n\n\nDierks & Rescorla Standards Track [Page 9]\n\f\nRFC 5246 TLS August 2008\n\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n For enumerateds that are never converted to external representation,\n the numerical information may be omitted.\n\n enum { low, medium, high } Amount;\n\n4.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax for definition is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } [[T]];\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n Structure definitions may be embedded.\n\n4.6.1. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. There\n must be a case arm for every element of the enumeration declared in\n the select. Case arms have limited fall-through: if two case arms\n follow in immediate succession with no fields in between, then they\n\n\n\nDierks & Rescorla Standards Track [Page 10]\n\f\nRFC 5246 TLS August 2008\n\n\n both contain the same fields. Thus, in the example below, \"orange\"\n and \"banana\" both contain V2. Note that this is a new piece of\n syntax in TLS 1.2.\n\n The body of the variant structure may be given a label for reference.\n The mechanism by which the variant is selected at runtime is not\n prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1;\n case e2: Te2;\n case e3: case e4: Te3;\n ....\n case en: Ten;\n } [[fv]];\n } [[Tv]];\n\n For example:\n\n enum { apple, orange, banana } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n select (VariantTag) { /* value of selector is implicit */\n case apple:\n V1; /* VariantBody, tag = apple */\n case orange:\n case banana:\n V2; /* VariantBody, tag = orange or banana */\n } variant_body; /* optional label on variant */\n } VariantRecord;\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 11]\n\f\nRFC 5246 TLS August 2008\n\n\n4.7. Cryptographic Attributes\n\n The five cryptographic operations -- digital signing, stream cipher\n encryption, block cipher encryption, authenticated encryption with\n additional data (AEAD) encryption, and public key encryption -- are\n designated digitally-signed, stream-ciphered, block-ciphered, aead-\n ciphered, and public-key-encrypted, respectively. A field's\n cryptographic processing is specified by prepending an appropriate\n key word designation before the field's type specification.\n Cryptographic keys are implied by the current session state (see\n Section 6.1).\n\n A digitally-signed element is encoded as a struct DigitallySigned:\n\n struct {\n SignatureAndHashAlgorithm algorithm;\n opaque signature<0..2^16-1>;\n } DigitallySigned;\n\n The algorithm field specifies the algorithm used (see Section\n 7.4.1.4.1 for the definition of this field). Note that the\n introduction of the algorithm field is a change from previous\n versions. The signature is a digital signature using those\n algorithms over the contents of the element. The contents themselves\n do not appear on the wire but are simply calculated. The length of\n the signature is specified by the signing algorithm and key.\n\n In RSA signing, the opaque vector contains the signature generated\n using the RSASSA-PKCS1-v1_5 signature scheme defined in [PKCS1]. As\n discussed in [PKCS1], the DigestInfo MUST be DER-encoded [X680]\n [X690]. For hash algorithms without parameters (which includes\n SHA-1), the DigestInfo.AlgorithmIdentifier.parameters field MUST be\n NULL, but implementations MUST accept both without parameters and\n with NULL parameters. Note that earlier versions of TLS used a\n different RSA signature scheme that did not include a DigestInfo\n encoding.\n\n In DSA, the 20 bytes of the SHA-1 hash are run directly through the\n Digital Signing Algorithm with no additional hashing. This produces\n two values, r and s. The DSA signature is an opaque vector, as\n above, the contents of which are the DER encoding of:\n\n Dss-Sig-Value ::= SEQUENCE {\n r INTEGER,\n s INTEGER\n }\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 12]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: In current terminology, DSA refers to the Digital Signature\n Algorithm and DSS refers to the NIST standard. In the original SSL\n and TLS specs, \"DSS\" was used universally. This document uses \"DSA\"\n to refer to the algorithm, \"DSS\" to refer to the standard, and it\n uses \"DSS\" in the code point definitions for historical continuity.\n\n In stream cipher encryption, the plaintext is exclusive-ORed with an\n identical amount of output generated from a cryptographically secure\n keyed pseudorandom number generator.\n\n In block cipher encryption, every block of plaintext encrypts to a\n block of ciphertext. All block cipher encryption is done in CBC\n (Cipher Block Chaining) mode, and all items that are block-ciphered\n will be an exact multiple of the cipher block length.\n\n In AEAD encryption, the plaintext is simultaneously encrypted and\n integrity protected. The input may be of any length, and aead-\n ciphered output is generally larger than the input in order to\n accommodate the integrity check value.\n\n In public key encryption, a public key algorithm is used to encrypt\n data in such a way that it can be decrypted only with the matching\n private key. A public-key-encrypted element is encoded as an opaque\n vector <0..2^16-1>, where the length is specified by the encryption\n algorithm and key.\n\n RSA encryption is done using the RSAES-PKCS1-v1_5 encryption scheme\n defined in [PKCS1].\n\n In the following example\n\n stream-ciphered struct {\n uint8 field1;\n uint8 field2;\n digitally-signed opaque {\n uint8 field3<0..255>;\n uint8 field4;\n };\n } UserType;\n\n The contents of the inner struct (field3 and field4) are used as\n input for the signature/hash algorithm, and then the entire structure\n is encrypted with a stream cipher. The length of this structure, in\n bytes, would be equal to two bytes for field1 and field2, plus two\n bytes for the signature and hash algorithm, plus two bytes for the\n length of the signature, plus the length of the output of the signing\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 13]\n\f\nRFC 5246 TLS August 2008\n\n\n algorithm. The length of the signature is known because the\n algorithm and key used for the signing are known prior to encoding or\n decoding this structure.\n\n4.8. Constants\n\n Typed constants can be defined for purposes of specification by\n declaring a symbol of the desired type and assigning values to it.\n\n Under-specified types (opaque, variable-length vectors, and\n structures that contain opaque) cannot be assigned values. No fields\n of a multi-element structure or vector may be elided.\n\n For example:\n\n struct {\n uint8 f1;\n uint8 f2;\n } Example1;\n\n Example1 ex1 = {1, 4}; /* assigns f1 = 1, f2 = 4 */\n\n5. HMAC and the Pseudorandom Function\n\n The TLS record layer uses a keyed Message Authentication Code (MAC)\n to protect message integrity. The cipher suites defined in this\n document use a construction known as HMAC, described in [HMAC], which\n is based on a hash function. Other cipher suites MAY define their\n own MAC constructions, if needed.\n\n In addition, a construction is required to do expansion of secrets\n into blocks of data for the purposes of key generation or validation.\n This pseudorandom function (PRF) takes as input a secret, a seed, and\n an identifying label and produces an output of arbitrary length.\n\n In this section, we define one PRF, based on HMAC. This PRF with the\n SHA-256 hash function is used for all cipher suites defined in this\n document and in TLS documents published prior to this document when\n TLS 1.2 is negotiated. New cipher suites MUST explicitly specify a\n PRF and, in general, SHOULD use the TLS PRF with SHA-256 or a\n stronger standard hash function.\n\n First, we define a data expansion function, P_hash(secret, data),\n that uses a single hash function to expand a secret and seed into an\n arbitrary quantity of output:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 14]\n\f\nRFC 5246 TLS August 2008\n\n\n P_hash(secret, seed) = HMAC_hash(secret, A(1) + seed) +\n HMAC_hash(secret, A(2) + seed) +\n HMAC_hash(secret, A(3) + seed) + ...\n\n where + indicates concatenation.\n\n A() is defined as:\n\n A(0) = seed\n A(i) = HMAC_hash(secret, A(i-1))\n\n P_hash can be iterated as many times as necessary to produce the\n required quantity of data. For example, if P_SHA256 is being used to\n create 80 bytes of data, it will have to be iterated three times\n (through A(3)), creating 96 bytes of output data; the last 16 bytes\n of the final iteration will then be discarded, leaving 80 bytes of\n output data.\n\n TLS's PRF is created by applying P_hash to the secret as:\n\n PRF(secret, label, seed) = P_(secret, label + seed)\n\n The label is an ASCII string. It should be included in the exact\n form it is given without a length byte or trailing null character.\n For example, the label \"slithy toves\" would be processed by hashing\n the following bytes:\n\n 73 6C 69 74 68 79 20 74 6F 76 65 73\n\n6. The TLS Record Protocol\n\n The TLS Record Protocol is a layered protocol. At each layer,\n messages may include fields for length, description, and content.\n The Record Protocol takes messages to be transmitted, fragments the\n data into manageable blocks, optionally compresses the data, applies\n a MAC, encrypts, and transmits the result. Received data is\n decrypted, verified, decompressed, reassembled, and then delivered to\n higher-level clients.\n\n Four protocols that use the record protocol are described in this\n document: the handshake protocol, the alert protocol, the change\n cipher spec protocol, and the application data protocol. In order to\n allow extension of the TLS protocol, additional record content types\n can be supported by the record protocol. New record content type\n values are assigned by IANA in the TLS Content Type Registry as\n described in Section 12.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 15]\n\f\nRFC 5246 TLS August 2008\n\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST send an\n unexpected_message alert.\n\n Any protocol designed for use over TLS must be carefully designed to\n deal with all possible attacks against it. As a practical matter,\n this means that the protocol designer must be aware of what security\n properties TLS does and does not provide and cannot safely rely on\n the latter.\n\n Note in particular that type and length of a record are not protected\n by encryption. If this information is itself sensitive, application\n designers may wish to take steps (padding, cover traffic) to minimize\n information leakage.\n\n6.1. Connection States\n\n\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 8446 Mozilla\nObsoletes: 5077, 5246, 6961 August 2018\nUpdates: 5705, 6066\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS) Protocol Version 1.3\n\nAbstract\n\n TLS allows client/server applications to communicate\n over the Internet in a way that is designed to prevent eavesdropping,\n tampering, and message forgery.\n\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077,\n 5246, and 6961. This document also specifies new requirements for\n TLS 1.2 implementations.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 7841.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n https://www.rfc-editor.org/info/rfc8446.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 8446 TLS August 2018\n\n\nCopyright Notice\n\n Copyright (c) 2018 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (https://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 8446 TLS August 2018\n\n\nTable of Contents\n\n 1. Introduction ....................................................6\n 1.1. Conventions and Terminology ................................7\n 1.2. Major Differences from TLS 1.2 .............................8\n 1.3. Updates Affecting TLS 1.2 ..................................9\n 2. Protocol Overview ..............................................10\n 2.1. Incorrect DHE Share .......................................14\n 2.2. Resumption and Pre-Shared Key (PSK) .......................15\n 2.3. 0-RTT Data ................................................17\n 3. Presentation Language ..........................................19\n 3.1. Basic Block Size ..........................................19\n 3.2. Miscellaneous .............................................20\n 3.3. Numbers ...................................................20\n 3.4. Vectors ...................................................20\n 3.5. Enumerateds ...............................................21\n 3.6. Constructed Types .........................................22\n 3.7. Constants .................................................23\n 3.8. Variants ..................................................23\n 4. Handshake Protocol .............................................24\n 4.1. Key Exchange Messages .....................................25\n 4.1.1. Cryptographic Negotiation ..........................26\n 4.1.2. Client Hello .......................................27\n 4.1.3. Server Hello .......................................31\n 4.1.4. Hello Retry Request ................................33\n 4.2. Extensions ................................................35\n 4.2.1. Supported Versions .................................39\n 4.2.2. Cookie .............................................40\n 4.2.3. Signature Algorithms ...............................41\n 4.2.4. Certificate Authorities ............................45\n 4.2.5. OID Filters ........................................45\n 4.2.6. Post-Handshake Client Authentication ...............47\n 4.2.7. Supported Groups ...................................47\n 4.2.8. Key Share ..........................................48\n 4.2.9. Pre-Shared Key Exchange Modes ......................51\n 4.2.10. Early Data Indication .............................52\n 4.2.11. Pre-Shared Key Extension ..........................55\n 4.3. Server Parameters .........................................59\n 4.3.1. Encrypted Extensions ...............................60\n 4.3.2. Certificate Request ................................60\n 4.4. Authentication Messages ...................................61\n 4.4.1. The Transcript Hash ................................63\n 4.4.2. Certificate ........................................64\n 4.4.3. Certificate Verify .................................69\n 4.4.4. Finished ...........................................71\n 4.5. End of Early Data .........................................72\n\n\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 8446 TLS August 2018\n\n\n 4.6. Post-Handshake Messages ...................................73\n 4.6.1. New Session Ticket Message .........................73\n 4.6.2. Post-Handshake Authentication ......................75\n 4.6.3. Key and Initialization Vector Update ...............76\n 5. Record Protocol ................................................77\n 5.1. Record Layer ..............................................78\n 5.2. Record Payload Protection .................................80\n 5.3. Per-Record Nonce ..........................................82\n 5.4. Record Padding ............................................83\n 5.5. Limits on Key Usage .......................................84\n 6. Alert Protocol .................................................85\n 6.1. Closure Alerts ............................................87\n 6.2. Error Alerts ..............................................88\n 7. Cryptographic Computations .....................................90\n 7.1. Key Schedule ..............................................91\n 7.2. Updating Traffic Secrets ..................................94\n 7.3. Traffic Key Calculation ...................................95\n 7.4. (EC)DHE Shared Secret Calculation .........................95\n 7.4.1. Finite Field Diffie-Hellman ........................95\n 7.4.2. Elliptic Curve Diffie-Hellman ......................96\n 7.5. Exporters .................................................97\n 8. 0-RTT and Anti-Replay ..........................................98\n 8.1. Single-Use Tickets ........................................99\n 8.2. Client Hello Recording ....................................99\n 8.3. Freshness Checks .........................................101\n 9. Compliance Requirements .......................................102\n 9.1. Mandatory-to-Implement Cipher Suites .....................102\n 9.2. Mandatory-to-Implement Extensions ........................103\n 9.3. Protocol Invariants ......................................104\n 10. Security Considerations ......................................106\n 11. IANA Considerations ..........................................106\n 12. References ...................................................109\n 12.1. Normative References ....................................109\n 12.2. Informative References ..................................112\n Appendix A. State Machine ........................................120\n A.1. Client ....................................................120\n A.2. Server ....................................................121\n Appendix B. Protocol Data Structures and Constant Values .........122\n B.1. Record Layer ..............................................122\n B.2. Alert Messages ............................................123\n B.3. Handshake Protocol ........................................124\n B.3.1. Key Exchange Messages .................................125\n B.3.2. Server Parameters Messages ............................131\n B.3.3. Authentication Messages ...............................132\n B.3.4. Ticket Establishment ..................................132\n B.3.5. Updating Keys .........................................133\n B.4. Cipher Suites .............................................133\n\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 8446 TLS August 2018\n\n\n Appendix C. Implementation Notes .................................134\n C.1. Random Number Generation and Seeding ......................134\n C.2. Certificates and Authentication ...........................135\n C.3. Implementation Pitfalls ...................................135\n C.4. Client Tracking Prevention ................................137\n C.5. Unauthenticated Operation .................................137\n Appendix D. Backward Compatibility ...............................138\n D.1. Negotiating with an Older Server ..........................139\n D.2. Negotiating with an Older Client ..........................139\n D.3. 0-RTT Backward Compatibility ..............................140\n D.4. Middlebox Compatibility Mode ..............................140\n D.5. Security Restrictions Related to Backward Compatibility ...141\n Appendix E. Overview of Security Properties ......................142\n E.1. Handshake .................................................142\n E.1.1. Key Derivation and HKDF ...............................145\n E.1.2. Client Authentication .................................146\n E.1.3. 0-RTT .................................................146\n E.1.4. Exporter Independence .................................146\n E.1.5. Post-Compromise Security ..............................146\n E.1.6. External References ...................................147\n E.2. Record Layer ..............................................147\n E.2.1. External References ...................................148\n E.3. Traffic Analysis ..........................................148\n E.4. Side-Channel Attacks ......................................149\n E.5. Replay Attacks on 0-RTT ...................................150\n E.5.1. Replay and Exporters ..................................151\n E.6. PSK Identity Exposure .....................................152\n E.7. Sharing PSKs ..............................................152\n E.8. Attacks on Static RSA .....................................152\n Contributors .....................................................153\n Author's Address .................................................160\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 8446 TLS August 2018\n\n\n1. Introduction\n\n The primary goal of TLS is to provide a secure channel between two\n communicating peers; the only requirement from the underlying\n transport is a reliable, in-order data stream. Specifically, the\n secure channel should provide the following properties:\n\n - Authentication: The server side of the channel is always\n authenticated; the client side is optionally authenticated.\n Authentication can happen via asymmetric cryptography (e.g., RSA\n [RSA], the Elliptic Curve Digital Signature Algorithm (ECDSA)\n [ECDSA], or the Edwards-Curve Digital Signature Algorithm (EdDSA)\n [RFC8032]) or a symmetric pre-shared key (PSK).\n\n - Confidentiality: Data sent over the channel after establishment is\n only visible to the endpoints. TLS does not hide the length of\n the data it transmits, though endpoints are able to pad TLS\n records in order to obscure lengths and improve protection against\n traffic analysis techniques.\n\n - Integrity: Data sent over the channel after establishment cannot\n be modified by attackers without detection.\n\n These properties should be true even in the face of an attacker who\n has complete control of the network, as described in [RFC3552]. See\n Appendix E for a more complete statement of the relevant security\n properties.\n\n TLS consists of two primary components:\n\n - A handshake protocol (Section 4) that authenticates the\n communicating parties, negotiates cryptographic modes and\n parameters, and establishes shared keying material. The handshake\n protocol is designed to resist tampering; an active attacker\n should not be able to force the peers to negotiate different\n parameters than they would if the connection were not under\n attack.\n\n - A record protocol (Section 5) that uses the parameters established\n by the handshake protocol to protect traffic between the\n communicating peers. The record protocol divides traffic up into\n a series of records, each of which is independently protected\n using the traffic keys.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 8446 TLS August 2018\n\n\n TLS is application protocol independent; higher-level protocols can\n layer on top of TLS transparently. The TLS standard, however, does\n not specify how protocols add security with TLS; how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n This document defines TLS version 1.3. While TLS 1.3 is not directly\n compatible with previous versions, all versions of TLS incorporate a\n versioning mechanism which allows clients and servers to\n interoperably negotiate a common version if one is supported by both\n peers.\n\n This document supersedes and obsoletes previous versions of TLS,\n including version 1.2 [RFC5246]. It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2. Because TLS 1.3 changes the way keys are\n derived, it updates [RFC5705] as described in Section 7.5. It also\n changes how Online Certificate Status Protocol (OCSP) messages are\n carried and therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in Section 4.4.2.1.\n\n1.1. Conventions and Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all\n capitals, as shown here.\n\n The following terms are used:\n\n client: The endpoint initiating the TLS connection.\n\n connection: A transport-layer connection between two endpoints.\n\n endpoint: Either the client or server of the connection.\n\n handshake: An initial negotiation between client and server that\n establishes the parameters of their subsequent interactions\n within TLS.\n\n peer: An endpoint. When discussing a particular endpoint, \"peer\"\n refers to the endpoint that is not the primary subject of\n discussion.\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\f\nRFC 8446 TLS August 2018\n\n\n receiver: An endpoint that is receiving records.\n\n sender: An endpoint that is transmitting records.\n\n server: The endpoint that did not initiate the TLS connection.\n\n1.2. Major Differences from TLS 1.2\n\n The following is a list of the major functional differences between\n TLS 1.2 and TLS 1.3. It is not intended to be exhaustive, and there\n are many minor differences.\n\n - The list of supported symmetric encryption algorithms has been\n pruned of all algorithms that are considered legacy. Those that\n remain are all Authenticated Encryption with Associated Data\n (AEAD) algorithms. The cipher suite concept has been changed to\n separate the authentication and key exchange mechanisms from the\n record protection algorithm (including secret key length) and a\n hash to be used with both the key derivation function and\n handshake message authentication code (MAC).\n\n - A zero round-trip time (0-RTT) mode was added, saving a round trip\n at connection setup for some application data, at the cost of\n certain security properties.\n\n - Static RSA and Diffie-Hellman cipher suites have been removed; all\n public-key based key exchange mechanisms now provide forward\n secrecy.\n\n - All handshake messages after the ServerHello are now encrypted.\n The newly introduced EncryptedExtensions message allows various\n extensions previously sent in the clear in the ServerHello to also\n enjoy confidentiality protection.\n\n - The key derivation functions have been redesigned. The new design\n allows easier analysis by cryptographers due to their improved key\n separation properties. The HMAC-based Extract-and-Expand Key\n Derivation Function (HKDF) is used as an underlying primitive.\n\n - The handshake state machine has been significantly restructured to\n be more consistent and to remove superfluous messages such as\n ChangeCipherSpec (except when needed for middlebox compatibility).\n\n - Elliptic curve algorithms are now in the base spec, and new\n signature algorithms, such as EdDSA, are included. TLS 1.3\n removed point format negotiation in favor of a single point format\n for each curve.\n\n\n\n\nRescorla Standards Track [Page 8]\n\f\nRFC 8446 TLS August 2018\n\n\n - Other cryptographic improvements were made, including changing the\n RSA padding to use the RSA Probabilistic Signature Scheme\n (RSASSA-PSS), and the removal of compression, the Digital\n Signature Algorithm (DSA), and custom Ephemeral Diffie-Hellman\n (DHE) groups.\n\n - The TLS 1.2 version negotiation mechanism has been deprecated in\n favor of a version list in an extension. This increases\n compatibility with existing servers that incorrectly implemented\n version negotiation.\n\n - Session resumption with and without server-side state as well as\n the PSK-based cipher suites of earlier TLS versions have been\n replaced by a single new PSK exchange.\n\n - References have been updated to point to the updated versions of\n RFCs, as appropriate (e.g., RFC 5280 rather than RFC 3280).\n\n1.3. Updates Affecting TLS 1.2\n\n This document defines several changes that optionally affect\n implementations of TLS 1.2, including those which do not also support\n TLS 1.3:\n\n - A version downgrade protection mechanism is described in\n Section 4.1.3.\n\n - RSASSA-PSS signature schemes are defined in Section 4.2.3.\n\n - The \"supported_versions\" ClientHello extension can be used to\n negotiate the version of TLS to use, in preference to the\n legacy_version field of the ClientHello.\n\n - The \"signature_algorithms_cert\" extension allows a client to\n indicate which signature algorithms it can validate in X.509\n certificates.\n\n Additionally, this document clarifies some compliance requirements\n for earlier versions of TLS; see Section 9.3.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 9]\n\f\nRFC 8446 TLS August 2018\n\n\n2. Protocol Overview\n\n The cryptographic parameters used by the secure channel are produced\n by the TLS handshake protocol. This sub-protocol of TLS is used by\n the client and server when first communicating with each other. The\n handshake protocol allows peers to negotiate a protocol version,\n select cryptographic algorithms, optionally authenticate each other,\n and establish shared secret keying material. Once the handshake is\n complete, the peers use the established keys to protect the\n application-layer traffic.\n\n A failure of the handshake or other protocol error triggers the\n termination of the connection, optionally preceded by an alert\n message (Section 6).\n\n TLS supports three basic key exchange modes:\n\n - (EC)DHE (Diffie-Hellman over either finite fields or elliptic\n curves)\n\n - PSK-only\n\n - PSK with (EC)DHE\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 10]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 1 below shows the basic full TLS handshake:\n\n Client Server\n\nKey ^ ClientHello\nExch | + key_share*\n | + signature_algorithms*\n | + psk_key_exchange_modes*\n v + pre_shared_key* -------->\n ServerHello ^ Key\n + key_share* | Exch\n + pre_shared_key* v\n {EncryptedExtensions} ^ Server\n {CertificateRequest*} v Params\n {Certificate*} ^\n {CertificateVerify*} | Auth\n {Finished} v\n <-------- [Application Data*]\n ^ {Certificate*}\nAuth | {CertificateVerify*}\n v {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 1: Message Flow for Full TLS Handshake\n\n The handshake can be thought of as having three phases (indicated in\n the diagram above):\n\n - Key Exchange: Establish shared keying material and select the\n cryptographic parameters. Everything after this phase is\n encrypted.\n\n - Server Parameters: Establish other handshake parameters\n (whether the client is authenticated, application-layer protocol\n support, etc.).\n\n\n\n\nRescorla Standards Track [Page 11]\n\f\nRFC 8446 TLS August 2018\n\n\n - Authentication: Authenticate the server (and, optionally, the\n client) and provide key confirmation and handshake integrity.\n\n In the Key Exchange phase, the client sends the ClientHello\n (Section 4.1.2) message, which contains a random nonce\n (ClientHello.random); its offered protocol versions; a list of\n symmetric cipher/HKDF hash pairs; either a set of Diffie-Hellman key\n shares (in the \"key_share\" (Section 4.2.8) extension), a set of\n pre-shared key labels (in the \"pre_shared_key\" (Section 4.2.11)\n extension), or both; and potentially additional extensions.\n Additional fields and/or messages may also be present for middlebox\n compatibility.\n\n The server processes the ClientHello and determines the appropriate\n cryptographic parameters for the connection. It then responds with\n its own ServerHello (Section 4.1.3), which indicates the negotiated\n connection parameters. The combination of the ClientHello and the\n ServerHello determines the shared keys. If (EC)DHE key establishment\n is in use, then the ServerHello contains a \"key_share\" extension with\n the server's ephemeral Diffie-Hellman share; the server's share MUST\n be in the same group as one of the client's shares. If PSK key\n establishment is in use, then the ServerHello contains a\n \"pre_shared_key\" extension indicating which of the client's offered\n PSKs was selected. Note that implementations can use (EC)DHE and PSK\n together, in which case both extensions will be supplied.\n\n The server then sends two messages to establish the Server\n Parameters:\n\n EncryptedExtensions: responses to ClientHello extensions that are\n not required to determine the cryptographic parameters, other than\n those that are specific to individual certificates.\n [Section 4.3.1]\n\n CertificateRequest: if certificate-based client authentication is\n desired, the desired parameters for that certificate. This\n message is omitted if client authentication is not desired.\n [Section 4.3.2]\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 12]\n\f\nRFC 8446 TLS August 2018\n\n\n Finally, the client and server exchange Authentication messages. TLS\n uses the same set of messages every time that certificate-based\n authentication is needed. (PSK-based authentication happens as a\n side effect of key exchange.) Specifically:\n\n Certificate: The certificate of the endpoint and any per-certificate\n extensions. This message is omitted by the server if not\n authenticating with a certificate and by the client if the server\n did not send CertificateRequest (thus indicating that the client\n should not authenticate with a certificate). Note that if raw\n public keys [RFC7250] or the cached information extension\n [RFC7924] are in use, then this message will not contain a\n certificate but rather some other value corresponding to the\n server's long-term key. [Section 4.4.2]\n\n CertificateVerify: A signature over the entire handshake using the\n private key corresponding to the public key in the Certificate\n message. This message is omitted if the endpoint is not\n authenticating via a certificate. [Section 4.4.3]\n\n Finished: A MAC (Message Authentication Code) over the entire\n handshake. This message provides key confirmation, binds the\n endpoint's identity to the exchanged keys, and in PSK mode also\n authenticates the handshake. [Section 4.4.4]\n\n Upon receiving the server's messages, the client responds with its\n Authentication messages, namely Certificate and CertificateVerify (if\n requested), and Finished.\n\n At this point, the handshake is complete, and the client and server\n derive the keying material required by the record layer to exchange\n application-layer data protected through authenticated encryption.\n Application Data MUST NOT be sent prior to sending the Finished\n message, except as specified in Section 2.3. Note that while the\n server may send Application Data prior to receiving the client's\n Authentication messages, any data sent at that point is, of course,\n being sent to an unauthenticated peer.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 13]\n\f\nRFC 8446 TLS August 2018\n\n\n2.1. Incorrect DHE Share\n\n If the client has not provided a sufficient \"key_share\" extension\n (e.g., it includes only DHE or ECDHE groups unacceptable to or\n unsupported by the server), the server corrects the mismatch with a\n HelloRetryRequest and the client needs to restart the handshake with\n an appropriate \"key_share\" extension, as shown in Figure 2. If no\n common cryptographic parameters can be negotiated, the server MUST\n abort the handshake with an appropriate alert.\n\n Client Server\n\n ClientHello\n + key_share -------->\n HelloRetryRequest\n <-------- + key_share\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 2: Message Flow for a Full Handshake with\n Mismatched Parameters\n\n Note: The handshake transcript incorporates the initial\n ClientHello/HelloRetryRequest exchange; it is not reset with the\n new ClientHello.\n\n TLS also allows several optimized variants of the basic handshake, as\n described in the following sections.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 14]\n\f\nRFC 8446 TLS August 2018\n\n\n2.2. Resumption and Pre-Shared Key (PSK)\n\n Although TLS PSKs can be established out of band, PSKs can also be\n established in a previous connection and then used to establish a new\n connection (\"session resumption\" or \"resuming\" with a PSK). Once a\n handshake has completed, the server can send the client a PSK\n identity that corresponds to a unique key derived from the initial\n handshake (see Section 4.6.1). The client can then use that PSK\n identity in future handshakes to negotiate the use of the associated\n PSK. If the server accepts the PSK, then the security context of the\n new connection is cryptographically tied to the original connection\n and the key derived from the initial handshake is used to bootstrap\n the cryptographic state instead of a full handshake. In TLS 1.2 and\n below, this functionality was provided by \"session IDs\" and \"session\n tickets\" [RFC5077]. Both mechanisms are obsoleted in TLS 1.3.\n\n PSKs can be used with (EC)DHE key exchange in order to provide\n forward secrecy in combination with shared keys, or can be used\n alone, at the cost of losing forward secrecy for the application\n data.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 15]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 3 shows a pair of handshakes in which the first handshake\n establishes a PSK and the second handshake uses it:\n\n Client Server\n\n Initial Handshake:\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n <-------- [NewSessionTicket]\n [Application Data] <-------> [Application Data]\n\n\n Subsequent Handshake:\n ClientHello\n + key_share*\n + pre_shared_key -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n {Finished}\n <-------- [Application Data*]\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 3: Message Flow for Resumption and PSK\n\n As the server is authenticating via a PSK, it does not send a\n Certificate or a CertificateVerify message. When a client offers\n resumption via a PSK, it SHOULD also supply a \"key_share\" extension\n to the server to allow the server to decline resumption and fall back\n to a full handshake, if needed. The server responds with a\n \"pre_shared_key\" extension to negotiate the use of PSK key\n establishment and can (as shown here) respond with a \"key_share\"\n extension to do (EC)DHE key establishment, thus providing forward\n secrecy.\n\n\n\n\nRescorla Standards Track [Page 16]\n\f\nRFC 8446 TLS August 2018\n\n\n When PSKs are provisioned out of band, the PSK identity and the KDF\n hash algorithm to be used with the PSK MUST also be provisioned.\n\n Note: When using an out-of-band provisioned pre-shared secret, a\n critical consideration is using sufficient entropy during the key\n generation, as discussed in [RFC4086]. Deriving a shared secret\n from a password or other low-entropy sources is not secure. A\n low-entropy secret, or password, is subject to dictionary attacks\n based on the PSK binder. The specified PSK authentication is not\n a strong password-based authenticated key exchange even when used\n with Diffie-Hellman key establishment. Specifically, it does not\n prevent an attacker that can observe the handshake from performing\n a brute-force attack on the password/pre-shared key.\n\n2.3. 0-RTT Data\n\n When clients and servers share a PSK (either obtained externally or\n via a previous handshake), TLS 1.3 allows clients to send data on the\n first flight (\"early data\"). The client uses the PSK to authenticate\n the server and to encrypt the early data.\n\n As shown in Figure 4, the 0-RTT data is just added to the 1-RTT\n handshake in the first flight. The rest of the handshake uses the\n same messages as for a 1-RTT handshake with PSK resumption.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 17]\n\f\nRFC 8446 TLS August 2018\n\n\n Client Server\n\n ClientHello\n + early_data\n + key_share*\n + psk_key_exchange_modes\n + pre_shared_key\n (Application Data*) -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n + early_data*\n {Finished}\n <-------- [Application Data*]\n (EndOfEarlyData)\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n () Indicates messages protected using keys\n derived from a client_early_traffic_secret.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 4: Message Flow for a 0-RTT Handshake\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 18]\n\f\nRFC 8446 TLS August 2018\n\n\n IMPORTANT NOTE: The security properties for 0-RTT data are weaker\n than those for other kinds of TLS data. Specifically:\n\n 1. This data is not forward secret, as it is encrypted solely under\n keys derived using the offered PSK.\n\n 2. There are no guarantees of non-replay between connections.\n Protection against replay for ordinary TLS 1.3 1-RTT data is\n provided via the server's Random value, but 0-RTT data does not\n depend on the ServerHello and therefore has weaker guarantees.\n This is especially relevant if the data is authenticated either\n with TLS client authentication or inside the application\n protocol. The same warnings apply to any use of the\n early_exporter_master_secret.\n\n\n\n===== DOCUMENT =====\n\n A TLS connection state is the operating environment of the TLS Record\n Protocol. It specifies a compression algorithm, an encryption\n algorithm, and a MAC algorithm. In addition, the parameters for\n these algorithms are known: the MAC key and the bulk encryption keys\n for the connection in both the read and the write directions.\n Logically, there are always four connection states outstanding: the\n current read and write states, and the pending read and write states.\n All records are processed under the current read and write states.\n The security parameters for the pending states can be set by the TLS\n Handshake Protocol, and the ChangeCipherSpec can selectively make\n either of the pending states current, in which case the appropriate\n current state is disposed of and replaced with the pending state; the\n pending state is then reinitialized to an empty state. It is illegal\n to make a state that has not been initialized with security\n parameters a current state. The initial current state always\n specifies that no encryption, compression, or MAC will be used.\n\n The security parameters for a TLS Connection read and write state are\n set by providing the following values:\n\n connection end\n Whether this entity is considered the \"client\" or the \"server\" in\n this connection.\n\n PRF algorithm\n An algorithm used to generate keys from the master secret (see\n Sections 5 and 6.3).\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 16]\n\f\nRFC 5246 TLS August 2008\n\n\n bulk encryption algorithm\n An algorithm to be used for bulk encryption. This specification\n includes the key size of this algorithm, whether it is a block,\n stream, or AEAD cipher, the block size of the cipher (if\n appropriate), and the lengths of explicit and implicit\n initialization vectors (or nonces).\n\n MAC algorithm\n An algorithm to be used for message authentication. This\n specification includes the size of the value returned by the MAC\n algorithm.\n\n compression algorithm\n An algorithm to be used for data compression. This specification\n must include all information the algorithm requires to do\n compression.\n\n master secret\n A 48-byte secret shared between the two peers in the connection.\n\n client random\n A 32-byte value provided by the client.\n\n server random\n A 32-byte value provided by the server.\n\n These parameters are defined in the presentation language as:\n\n enum { server, client } ConnectionEnd;\n\n enum { tls_prf_sha256 } PRFAlgorithm;\n\n enum { null, rc4, 3des, aes }\n BulkCipherAlgorithm;\n\n enum { stream, block, aead } CipherType;\n\n enum { null, hmac_md5, hmac_sha1, hmac_sha256,\n hmac_sha384, hmac_sha512} MACAlgorithm;\n\n enum { null(0), (255) } CompressionMethod;\n\n /* The algorithms specified in CompressionMethod, PRFAlgorithm,\n BulkCipherAlgorithm, and MACAlgorithm may be added to. */\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 17]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ConnectionEnd entity;\n PRFAlgorithm prf_algorithm;\n BulkCipherAlgorithm bulk_cipher_algorithm;\n CipherType cipher_type;\n uint8 enc_key_length;\n uint8 block_length;\n uint8 fixed_iv_length;\n uint8 record_iv_length;\n MACAlgorithm mac_algorithm;\n uint8 mac_length;\n uint8 mac_key_length;\n CompressionMethod compression_algorithm;\n opaque master_secret[48];\n opaque client_random[32];\n opaque server_random[32];\n } SecurityParameters;\n\n The record layer will use the security parameters to generate the\n following six items (some of which are not required by all ciphers,\n and are thus empty):\n\n client write MAC key\n server write MAC key\n client write encryption key\n server write encryption key\n client write IV\n server write IV\n\n The client write parameters are used by the server when receiving and\n processing records and vice versa. The algorithm used for generating\n these items from the security parameters is described in Section 6.3.\n\n Once the security parameters have been set and the keys have been\n generated, the connection states can be instantiated by making them\n the current states. These current states MUST be updated for each\n record processed. Each connection state includes the following\n elements:\n\n compression state\n The current state of the compression algorithm.\n\n cipher state\n The current state of the encryption algorithm. This will consist\n of the scheduled key for that connection. For stream ciphers,\n this will also contain whatever state information is necessary to\n allow the stream to continue to encrypt or decrypt data.\n\n\n\n\nDierks & Rescorla Standards Track [Page 18]\n\f\nRFC 5246 TLS August 2008\n\n\n MAC key\n The MAC key for this connection, as generated above.\n\n sequence number\n Each connection state contains a sequence number, which is\n maintained separately for read and write states. The sequence\n number MUST be set to zero whenever a connection state is made the\n active state. Sequence numbers are of type uint64 and may not\n exceed 2^64-1. Sequence numbers do not wrap. If a TLS\n implementation would need to wrap a sequence number, it must\n renegotiate instead. A sequence number is incremented after each\n record: specifically, the first record transmitted under a\n particular connection state MUST use sequence number 0.\n\n6.2. Record Layer\n\n The TLS record layer receives uninterpreted data from higher layers\n in non-empty blocks of arbitrary size.\n\n6.2.1. Fragmentation\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Client\n message boundaries are not preserved in the record layer (i.e.,\n multiple client messages of the same ContentType MAY be coalesced\n into a single TLSPlaintext record, or a single message MAY be\n fragmented across several records).\n\n struct {\n uint8 major;\n uint8 minor;\n } ProtocolVersion;\n\n enum {\n change_cipher_spec(20), alert(21), handshake(22),\n application_data(23), (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type\n The higher-level protocol used to process the enclosed fragment.\n\n\n\n\nDierks & Rescorla Standards Track [Page 19]\n\f\nRFC 5246 TLS August 2008\n\n\n version\n The version of the protocol being employed. This document\n describes TLS Version 1.2, which uses the version { 3, 3 }. The\n version value 3.3 is historical, deriving from the use of {3, 1}\n for TLS 1.0. (See Appendix A.1.) Note that a client that\n supports multiple versions of TLS may not know what version will\n be employed before it receives the ServerHello. See Appendix E\n for discussion about what record layer version number should be\n employed for ClientHello.\n\n length\n The length (in bytes) of the following TLSPlaintext.fragment. The\n length MUST NOT exceed 2^14.\n\n fragment\n The application data. This data is transparent and treated as an\n independent block to be dealt with by the higher-level protocol\n specified by the type field.\n\n Implementations MUST NOT send zero-length fragments of Handshake,\n Alert, or ChangeCipherSpec content types. Zero-length fragments of\n Application data MAY be sent as they are potentially useful as a\n traffic analysis countermeasure.\n\n Note: Data of different TLS record layer content types MAY be\n interleaved. Application data is generally of lower precedence for\n transmission than other content types. However, records MUST be\n delivered to the network in the same order as they are protected by\n the record layer. Recipients MUST receive and process interleaved\n application layer traffic during handshakes subsequent to the first\n one on a connection.\n\n6.2.2. Record Compression and Decompression\n\n All records are compressed using the compression algorithm defined in\n the current session state. There is always an active compression\n algorithm; however, initially it is defined as\n CompressionMethod.null. The compression algorithm translates a\n TLSPlaintext structure into a TLSCompressed structure. Compression\n functions are initialized with default state information whenever a\n connection state is made active. [RFC3749] describes compression\n algorithms for TLS.\n\n Compression must be lossless and may not increase the content length\n by more than 1024 bytes. If the decompression function encounters a\n TLSCompressed.fragment that would decompress to a length in excess of\n 2^14 bytes, it MUST report a fatal decompression failure error.\n\n\n\n\nDierks & Rescorla Standards Track [Page 20]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ContentType type; /* same as TLSPlaintext.type */\n ProtocolVersion version;/* same as TLSPlaintext.version */\n uint16 length;\n opaque fragment[TLSCompressed.length];\n } TLSCompressed;\n\n length\n The length (in bytes) of the following TLSCompressed.fragment.\n The length MUST NOT exceed 2^14 + 1024.\n\n fragment\n The compressed form of TLSPlaintext.fragment.\n\n Note: A CompressionMethod.null operation is an identity operation;\n no fields are altered.\n\n Implementation note: Decompression functions are responsible for\n ensuring that messages cannot cause internal buffer overflows.\n\n6.2.3. Record Payload Protection\n\n The encryption and MAC functions translate a TLSCompressed\n structure into a TLSCiphertext. The decryption functions reverse\n the process. The MAC of the record also includes a sequence\n number so that missing, extra, or repeated messages are\n detectable.\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n select (SecurityParameters.cipher_type) {\n case stream: GenericStreamCipher;\n case block: GenericBlockCipher;\n case aead: GenericAEADCipher;\n } fragment;\n } TLSCiphertext;\n\n type\n The type field is identical to TLSCompressed.type.\n\n version\n The version field is identical to TLSCompressed.version.\n\n length\n The length (in bytes) of the following TLSCiphertext.fragment.\n The length MUST NOT exceed 2^14 + 2048.\n\n\n\nDierks & Rescorla Standards Track [Page 21]\n\f\nRFC 5246 TLS August 2008\n\n\n fragment\n The encrypted form of TLSCompressed.fragment, with the MAC.\n\n6.2.3.1. Null or Standard Stream Cipher\n\n Stream ciphers (including BulkCipherAlgorithm.null; see Appendix A.6)\n convert TLSCompressed.fragment structures to and from stream\n TLSCiphertext.fragment structures.\n\n stream-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n } GenericStreamCipher;\n\n The MAC is generated as:\n\n MAC(MAC_write_key, seq_num +\n TLSCompressed.type +\n TLSCompressed.version +\n TLSCompressed.length +\n TLSCompressed.fragment);\n\n where \"+\" denotes concatenation.\n\n seq_num\n The sequence number for this record.\n\n MAC\n The MAC algorithm specified by SecurityParameters.mac_algorithm.\n\n Note that the MAC is computed before encryption. The stream cipher\n encrypts the entire block, including the MAC. For stream ciphers\n that do not use a synchronization vector (such as RC4), the stream\n cipher state from the end of one record is simply used on the\n subsequent packet. If the cipher suite is TLS_NULL_WITH_NULL_NULL,\n encryption consists of the identity operation (i.e., the data is not\n encrypted, and the MAC size is zero, implying that no MAC is used).\n For both null and stream ciphers, TLSCiphertext.length is\n TLSCompressed.length plus SecurityParameters.mac_length.\n\n6.2.3.2. CBC Block Cipher\n\n For block ciphers (such as 3DES or AES), the encryption and MAC\n functions convert TLSCompressed.fragment structures to and from block\n TLSCiphertext.fragment structures.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 22]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n opaque IV[SecurityParameters.record_iv_length];\n block-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n uint8 padding[GenericBlockCipher.padding_length];\n uint8 padding_length;\n };\n } GenericBlockCipher;\n\n The MAC is generated as described in Section 6.2.3.1.\n\n IV\n The Initialization Vector (IV) SHOULD be chosen at random, and\n MUST be unpredictable. Note that in versions of TLS prior to 1.1,\n there was no IV field, and the last ciphertext block of the\n previous record (the \"CBC residue\") was used as the IV. This was\n changed to prevent the attacks described in [CBCATT]. For block\n ciphers, the IV length is of length\n SecurityParameters.record_iv_length, which is equal to the\n SecurityParameters.block_size.\n\n padding\n Padding that is added to force the length of the plaintext to be\n an integral multiple of the block cipher's block length. The\n padding MAY be any length up to 255 bytes, as long as it results\n in the TLSCiphertext.length being an integral multiple of the\n block length. Lengths longer than necessary might be desirable to\n frustrate attacks on a protocol that are based on analysis of the\n lengths of exchanged messages. Each uint8 in the padding data\n vector MUST be filled with the padding length value. The receiver\n MUST check this padding and MUST use the bad_record_mac alert to\n indicate padding errors.\n\n padding_length\n The padding length MUST be such that the total size of the\n GenericBlockCipher structure is a multiple of the cipher's block\n length. Legal values range from zero to 255, inclusive. This\n length specifies the length of the padding field exclusive of the\n padding_length field itself.\n\n The encrypted data length (TLSCiphertext.length) is one more than the\n sum of SecurityParameters.block_length, TLSCompressed.length,\n SecurityParameters.mac_length, and padding_length.\n\n Example: If the block length is 8 bytes, the content length\n (TLSCompressed.length) is 61 bytes, and the MAC length is 20 bytes,\n then the length before padding is 82 bytes (this does not include the\n\n\n\nDierks & Rescorla Standards Track [Page 23]\n\f\nRFC 5246 TLS August 2008\n\n\n IV. Thus, the padding length modulo 8 must be equal to 6 in order to\n make the total length an even multiple of 8 bytes (the block length).\n The padding length can be 6, 14, 22, and so on, through 254. If the\n padding length were the minimum necessary, 6, the padding would be 6\n bytes, each containing the value 6. Thus, the last 8 octets of the\n GenericBlockCipher before block encryption would be xx 06 06 06 06 06\n 06 06, where xx is the last octet of the MAC.\n\n Note: With block ciphers in CBC mode (Cipher Block Chaining), it is\n critical that the entire plaintext of the record be known before any\n ciphertext is transmitted. Otherwise, it is possible for the\n attacker to mount the attack described in [CBCATT].\n\n Implementation note: Canvel et al. [CBCTIME] have demonstrated a\n timing attack on CBC padding based on the time required to compute\n the MAC. In order to defend against this attack, implementations\n MUST ensure that record processing time is essentially the same\n whether or not the padding is correct. In general, the best way to\n do this is to compute the MAC even if the padding is incorrect, and\n only then reject the packet. For instance, if the pad appears to be\n incorrect, the implementation might assume a zero-length pad and then\n compute the MAC. This leaves a small timing channel, since MAC\n performance depends to some extent on the size of the data fragment,\n but it is not believed to be large enough to be exploitable, due to\n the large block size of existing MACs and the small size of the\n timing signal.\n\n6.2.3.3. AEAD Ciphers\n\n For AEAD [AEAD] ciphers (such as [CCM] or [GCM]), the AEAD function\n converts TLSCompressed.fragment structures to and from AEAD\n TLSCiphertext.fragment structures.\n\n struct {\n opaque nonce_explicit[SecurityParameters.record_iv_length];\n aead-ciphered struct {\n opaque content[TLSCompressed.length];\n };\n } GenericAEADCipher;\n\n AEAD ciphers take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [AEAD]. The key is either the\n client_write_key or the server_write_key. No MAC key is used.\n\n Each AEAD cipher suite MUST specify how the nonce supplied to the\n AEAD operation is constructed, and what is the length of the\n GenericAEADCipher.nonce_explicit part. In many cases, it is\n\n\n\nDierks & Rescorla Standards Track [Page 24]\n\f\nRFC 5246 TLS August 2008\n\n\n appropriate to use the partially implicit nonce technique described\n in Section 3.2.1 of [AEAD]; with record_iv_length being the length of\n the explicit part. In this case, the implicit part SHOULD be derived\n from key_block as client_write_iv and server_write_iv (as described\n in Section 6.3), and the explicit part is included in\n GenericAEAEDCipher.nonce_explicit.\n\n The plaintext is the TLSCompressed.fragment.\n\n The additional authenticated data, which we denote as\n additional_data, is defined as follows:\n\n additional_data = seq_num + TLSCompressed.type +\n TLSCompressed.version + TLSCompressed.length;\n\n where \"+\" denotes concatenation.\n\n The aead_output consists of the ciphertext output by the AEAD\n encryption operation. The length will generally be larger than\n TLSCompressed.length, but by an amount that varies with the AEAD\n cipher. Since the ciphers might incorporate padding, the amount of\n overhead could vary with different TLSCompressed.length values. Each\n AEAD cipher MUST NOT produce an expansion of greater than 1024 bytes.\n Symbolically,\n\n AEADEncrypted = AEAD-Encrypt(write_key, nonce, plaintext,\n additional_data)\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, the \"additional_data\", and the AEADEncrypted value. The\n output is either the plaintext or an error indicating that the\n decryption failed. There is no separate integrity check. That is:\n\n TLSCompressed.fragment = AEAD-Decrypt(write_key, nonce,\n AEADEncrypted,\n additional_data)\n\n If the decryption fails, a fatal bad_record_mac alert MUST be\n generated.\n\n6.3. Key Calculation\n\n The Record Protocol requires an algorithm to generate keys required\n by the current connection state (see Appendix A.6) from the security\n parameters provided by the handshake protocol.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 25]\n\f\nRFC 5246 TLS August 2008\n\n\n The master secret is expanded into a sequence of secure bytes, which\n is then split to a client write MAC key, a server write MAC key, a\n client write encryption key, and a server write encryption key. Each\n of these is generated from the byte sequence in that order. Unused\n values are empty. Some AEAD ciphers may additionally require a\n client write IV and a server write IV (see Section 6.2.3.3).\n\n When keys and MAC keys are generated, the master secret is used as an\n entropy source.\n\n To generate the key material, compute\n\n key_block = PRF(SecurityParameters.master_secret,\n \"key expansion\",\n SecurityParameters.server_random +\n SecurityParameters.client_random);\n\n until enough output has been generated. Then, the key_block is\n partitioned as follows:\n\n client_write_MAC_key[SecurityParameters.mac_key_length]\n server_write_MAC_key[SecurityParameters.mac_key_length]\n client_write_key[SecurityParameters.enc_key_length]\n server_write_key[SecurityParameters.enc_key_length]\n client_write_IV[SecurityParameters.fixed_iv_length]\n server_write_IV[SecurityParameters.fixed_iv_length]\n\n Currently, the client_write_IV and server_write_IV are only generated\n for implicit nonce techniques as described in Section 3.2.1 of\n [AEAD].\n\n Implementation note: The currently defined cipher suite which\n requires the most material is AES_256_CBC_SHA256. It requires 2 x 32\n byte keys and 2 x 32 byte MAC keys, for a total 128 bytes of key\n material.\n\n7. The TLS Handshaking Protocols\n\n TLS has three subprotocols that are used to allow peers to agree upon\n security parameters for the record layer, to authenticate themselves,\n to instantiate negotiated security parameters, and to report error\n conditions to each other.\n\n The Handshake Protocol is responsible for negotiating a session,\n which consists of the following items:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 26]\n\f\nRFC 5246 TLS August 2008\n\n\n session identifier\n An arbitrary byte sequence chosen by the server to identify an\n active or resumable session state.\n\n peer certificate\n X509v3 [PKIX] certificate of the peer. This element of the state\n may be null.\n\n compression method\n The algorithm used to compress data prior to encryption.\n\n cipher spec\n Specifies the pseudorandom function (PRF) used to generate keying\n material, the bulk data encryption algorithm (such as null, AES,\n etc.) and the MAC algorithm (such as HMAC-SHA1). It also defines\n cryptographic attributes such as the mac_length. (See Appendix\n A.6 for formal definition.)\n\n master secret\n 48-byte secret shared between the client and server.\n\n is resumable\n A flag indicating whether the session can be used to initiate new\n connections.\n\n These items are then used to create security parameters for use by\n the record layer when protecting application data. Many connections\n can be instantiated using the same session through the resumption\n feature of the TLS Handshake Protocol.\n\n7.1. Change Cipher Spec Protocol\n\n The change cipher spec protocol exists to signal transitions in\n ciphering strategies. The protocol consists of a single message,\n which is encrypted and compressed under the current (not the pending)\n connection state. The message consists of a single byte of value 1.\n\n struct {\n enum { change_cipher_spec(1), (255) } type;\n } ChangeCipherSpec;\n\n The ChangeCipherSpec message is sent by both the client and the\n server to notify the receiving party that subsequent records will be\n protected under the newly negotiated CipherSpec and keys. Reception\n of this message causes the receiver to instruct the record layer to\n immediately copy the read pending state into the read current state.\n Immediately after sending this message, the sender MUST instruct the\n record layer to make the write pending state the write active state.\n\n\n\nDierks & Rescorla Standards Track [Page 27]\n\f\nRFC 5246 TLS August 2008\n\n\n (See Section 6.1.) The ChangeCipherSpec message is sent during the\n handshake after the security parameters have been agreed upon, but\n before the verifying Finished message is sent.\n\n Note: If a rehandshake occurs while data is flowing on a connection,\n the communicating parties may continue to send data using the old\n CipherSpec. However, once the ChangeCipherSpec has been sent, the\n new CipherSpec MUST be used. The first side to send the\n ChangeCipherSpec does not know that the other side has finished\n computing the new keying material (e.g., if it has to perform a\n time-consuming public key operation). Thus, a small window of time,\n during which the recipient must buffer the data, MAY exist. In\n practice, with modern machines this interval is likely to be fairly\n short.\n\n7.2. Alert Protocol\n\n One of the content types supported by the TLS record layer is the\n alert type. Alert messages convey the severity of the message\n (warning or fatal) and a description of the alert. Alert messages\n with a level of fatal result in the immediate termination of the\n connection. In this case, other connections corresponding to the\n session may continue, but the session identifier MUST be invalidated,\n preventing the failed session from being used to establish new\n connections. Like other messages, alert messages are encrypted and\n compressed, as specified by the current connection state.\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n\n\n\nDierks & Rescorla Standards Track [Page 28]\n\f\nRFC 5246 TLS August 2008\n\n\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n7.2.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack. Either party may\n initiate the exchange of closing messages.\n\n close_notify\n This message notifies the recipient that the sender will not send\n any more messages on this connection. Note that as of TLS 1.1,\n failure to properly close a connection no longer requires that a\n session not be resumed. This is a change from TLS 1.0 to conform\n with widespread implementation practice.\n\n Either party may initiate a close by sending a close_notify alert.\n Any data received after a closure alert is ignored.\n\n Unless some other fatal alert has been transmitted, each party is\n required to send a close_notify alert before closing the write side\n of the connection. The other party MUST respond with a close_notify\n alert of its own and close down the connection immediately,\n discarding any pending writes. It is not required for the initiator\n of the close to wait for the responding close_notify alert before\n closing the read side of the connection.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation must receive the responding\n close_notify alert before indicating to the application layer that\n the TLS connection has ended. If the application protocol will not\n transfer any additional data, but will only close the underlying\n transport connection, then the implementation MAY choose to close the\n transport without waiting for the responding close_notify. No part\n\n\n\n\nDierks & Rescorla Standards Track [Page 29]\n\f\nRFC 5246 TLS August 2008\n\n\n of this standard should be taken to dictate the manner in which a\n usage profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing a connection reliably delivers\n pending data before destroying the transport.\n\n7.2.2. Error Alerts\n\n Error handling in the TLS Handshake protocol is very simple. When an\n error is detected, the detecting party sends a message to the other\n party. Upon transmission or receipt of a fatal alert message, both\n parties immediately close the connection. Servers and clients MUST\n forget any session-identifiers, keys, and secrets associated with a\n failed connection. Thus, any connection terminated with a fatal\n alert MUST NOT be resumed.\n\n Whenever an implementation encounters a condition which is defined as\n a fatal alert, it MUST send the appropriate alert prior to closing\n the connection. For all errors where an alert level is not\n explicitly specified, the sending party MAY determine at its\n discretion whether to treat this as a fatal error or not. If the\n implementation chooses to send an alert but intends to close the\n connection immediately afterwards, it MUST send that alert at the\n fatal alert level.\n\n If an alert with a level of warning is sent and received, generally\n the connection can continue normally. If the receiving party decides\n not to proceed with the connection (e.g., after having received a\n no_renegotiation alert that it is not willing to accept), it SHOULD\n send a fatal alert to terminate the connection. Given this, the\n sending party cannot, in general, know how the receiving party will\n behave. Therefore, warning alerts are not very useful when the\n sending party wants to continue the connection, and thus are\n sometimes omitted. For example, if a peer decides to accept an\n expired certificate (perhaps after confirming this with the user) and\n wants to continue the connection, it would not generally send a\n certificate_expired alert.\n\n The following error alerts are defined:\n\n unexpected_message\n An inappropriate message was received. This alert is always fatal\n and should never be observed in communication between proper\n implementations.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 30]\n\f\nRFC 5246 TLS August 2008\n\n\n bad_record_mac\n This alert is returned if a record is received with an incorrect\n MAC. This alert also MUST be returned if an alert is sent because\n a TLSCiphertext decrypted in an invalid way: either it wasn't an\n even multiple of the block length, or its padding values, when\n checked, weren't correct. This message is always fatal and should\n never be observed in communication between proper implementations\n (except when messages were corrupted in the network).\n\n decryption_failed_RESERVED\n This alert was used in some earlier versions of TLS, and may have\n permitted certain attacks against the CBC mode [CBCATT]. It MUST\n NOT be sent by compliant implementations.\n\n record_overflow\n A TLSCiphertext record was received that had a length more than\n 2^14+2048 bytes, or a record decrypted to a TLSCompressed record\n with more than 2^14+1024 bytes. This message is always fatal and\n should never be observed in communication between proper\n implementations (except when messages were corrupted in the\n network).\n\n decompression_failure\n The decompression function received improper input (e.g., data\n that would expand to excessive length). This message is always\n fatal and should never be observed in communication between proper\n implementations.\n\n handshake_failure\n Reception of a handshake_failure alert message indicates that the\n sender was unable to negotiate an acceptable set of security\n parameters given the options available. This is a fatal error.\n\n no_certificate_RESERVED\n This alert was used in SSLv3 but not any version of TLS. It MUST\n NOT be sent by compliant implementations.\n\n bad_certificate\n A certificate was corrupt, contained signatures that did not\n verify correctly, etc.\n\n unsupported_certificate\n A certificate was of an unsupported type.\n\n certificate_revoked\n A certificate was revoked by its signer.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 31]\n\f\nRFC 5246 TLS August 2008\n\n\n certificate_expired\n A certificate has expired or is not currently valid.\n\n certificate_unknown\n Some other (unspecified) issue arose in processing the\n certificate, rendering it unacceptable.\n\n illegal_parameter\n A field in the handshake was out of range or inconsistent with\n other fields. This message is always fatal.\n\n unknown_ca\n A valid certificate chain or partial chain was received, but the\n certificate was not accepted because the CA certificate could not\n be located or couldn't be matched with a known, trusted CA. This\n message is always fatal.\n\n access_denied\n A valid certificate was received, but when access control was\n applied, the sender decided not to proceed with negotiation. This\n message is always fatal.\n\n decode_error\n A message could not be decoded because some field was out of the\n specified range or the length of the message was incorrect. This\n message is always fatal and should never be observed in\n communication between proper implementations (except when messages\n were corrupted in the network).\n\n decrypt_error\n A handshake cryptographic operation failed, including being unable\n to correctly verify a signature or validate a Finished message.\n This message is always fatal.\n\n export_restriction_RESERVED\n This alert was used in some earlier versions of TLS. It MUST NOT\n be sent by compliant implementations.\n\n protocol_version\n The protocol version the client has attempted to negotiate is\n recognized but not supported. (For example, old protocol versions\n might be avoided for security reasons.) This message is always\n fatal.\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 32]\n\f\nRFC 5246 TLS August 2008\n\n\n insufficient_security\n Returned instead of handshake_failure when a negotiation has\n failed specifically because the server requires ciphers more\n secure than those supported by the client. This message is always\n fatal.\n\n internal_error\n An internal error unrelated to the peer or the correctness of the\n protocol (such as a memory allocation failure) makes it impossible\n to continue. This message is always fatal.\n\n\n\n===== DOCUMENT =====\n\nNetwork Working Group E. Rescorla\nInternet-Draft RTFM, Inc.\nObsoletes: 5077, 5246, 6961 (if March 20, 2018\n approved)\nUpdates: 4492, 5705, 6066 (if approved)\nIntended status: Standards Track\nExpires: September 21, 2018\n\n\n\n===== DOCUMENT =====\n\n user_canceled\n This handshake is being canceled for some reason unrelated to a\n protocol failure. If the user cancels an operation after the\n handshake is complete, just closing the connection by sending a\n close_notify is more appropriate. This alert should be followed\n by a close_notify. This message is generally a warning.\n\n no_renegotiation\n Sent by the client in response to a hello request or by the server\n in response to a client hello after initial handshaking. Either\n of these would normally lead to renegotiation; when that is not\n appropriate, the recipient should respond with this alert. At\n that point, the original requester can decide whether to proceed\n with the connection. One case where this would be appropriate is\n where a server has spawned a process to satisfy a request; the\n process might receive security parameters (key length,\n authentication, etc.) at startup, and it might be difficult to\n communicate changes to these parameters after that point. This\n message is always a warning.\n\n unsupported_extension\n sent by clients that receive an extended server hello containing\n an extension that they did not put in the corresponding client\n hello. This message is always fatal.\n\n New Alert values are assigned by IANA as described in Section 12.\n\n7.3. Handshake Protocol Overview\n\n The cryptographic parameters of the session state are produced by the\n TLS Handshake Protocol, which operates on top of the TLS record\n layer. When a TLS client and server first start communicating, they\n agree on a protocol version, select cryptographic algorithms,\n optionally authenticate each other, and use public-key encryption\n techniques to generate shared secrets.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 33]\n\f\nRFC 5246 TLS August 2008\n\n\n The TLS Handshake Protocol involves the following steps:\n\n - Exchange hello messages to agree on algorithms, exchange random\n values, and check for session resumption.\n\n - Exchange the necessary cryptographic parameters to allow the\n client and server to agree on a premaster secret.\n\n - Exchange certificates and cryptographic information to allow the\n client and server to authenticate themselves.\n\n - Generate a master secret from the premaster secret and exchanged\n random values.\n\n - Provide security parameters to the record layer.\n\n - Allow the client and server to verify that their peer has\n calculated the same security parameters and that the handshake\n occurred without tampering by an attacker.\n\n Note that higher layers should not be overly reliant on whether TLS\n always negotiates the strongest possible connection between two\n peers. There are a number of ways in which a man-in-the-middle\n attacker can attempt to make two entities drop down to the least\n secure method they support. The protocol has been designed to\n minimize this risk, but there are still attacks available: for\n example, an attacker could block access to the port a secure service\n runs on, or attempt to get the peers to negotiate an unauthenticated\n connection. The fundamental rule is that higher levels must be\n cognizant of what their security requirements are and never transmit\n information over a channel less secure than what they require. The\n TLS protocol is secure in that any cipher suite offers its promised\n level of security: if you negotiate 3DES with a 1024-bit RSA key\n exchange with a host whose certificate you have verified, you can\n expect to be that secure.\n\n These goals are achieved by the handshake protocol, which can be\n summarized as follows: The client sends a ClientHello message to\n which the server must respond with a ServerHello message, or else a\n fatal error will occur and the connection will fail. The ClientHello\n and ServerHello are used to establish security enhancement\n capabilities between client and server. The ClientHello and\n ServerHello establish the following attributes: Protocol Version,\n Session ID, Cipher Suite, and Compression Method. Additionally, two\n random values are generated and exchanged: ClientHello.random and\n ServerHello.random.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 34]\n\f\nRFC 5246 TLS August 2008\n\n\n The actual key exchange uses up to four messages: the server\n Certificate, the ServerKeyExchange, the client Certificate, and the\n ClientKeyExchange. New key exchange methods can be created by\n specifying a format for these messages and by defining the use of the\n messages to allow the client and server to agree upon a shared\n secret. This secret MUST be quite long; currently defined key\n exchange methods exchange secrets that range from 46 bytes upwards.\n\n Following the hello messages, the server will send its certificate in\n a Certificate message if it is to be authenticated. Additionally, a\n ServerKeyExchange message may be sent, if it is required (e.g., if\n the server has no certificate, or if its certificate is for signing\n only). If the server is authenticated, it may request a certificate\n from the client, if that is appropriate to the cipher suite selected.\n Next, the server will send the ServerHelloDone message, indicating\n that the hello-message phase of the handshake is complete. The\n server will then wait for a client response. If the server has sent\n a CertificateRequest message, the client MUST send the Certificate\n message. The ClientKeyExchange message is now sent, and the content\n of that message will depend on the public key algorithm selected\n between the ClientHello and the ServerHello. If the client has sent\n a certificate with signing ability, a digitally-signed\n CertificateVerify message is sent to explicitly verify possession of\n the private key in the certificate.\n\n At this point, a ChangeCipherSpec message is sent by the client, and\n the client copies the pending Cipher Spec into the current Cipher\n Spec. The client then immediately sends the Finished message under\n the new algorithms, keys, and secrets. In response, the server will\n send its own ChangeCipherSpec message, transfer the pending to the\n current Cipher Spec, and send its Finished message under the new\n Cipher Spec. At this point, the handshake is complete, and the\n client and server may begin to exchange application layer data. (See\n flow chart below.) Application data MUST NOT be sent prior to the\n completion of the first handshake (before a cipher suite other than\n TLS_NULL_WITH_NULL_NULL is established).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 35]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1. Message flow for a full handshake\n\n * Indicates optional or situation-dependent messages that are not\n always sent.\n\n Note: To help avoid pipeline stalls, ChangeCipherSpec is an\n independent TLS protocol content type, and is not actually a TLS\n handshake message.\n\n When the client and server decide to resume a previous session or\n duplicate an existing session (instead of negotiating new security\n parameters), the message flow is as follows:\n\n The client sends a ClientHello using the Session ID of the session to\n be resumed. The server then checks its session cache for a match.\n If a match is found, and the server is willing to re-establish the\n connection under the specified session state, it will send a\n ServerHello with the same Session ID value. At this point, both\n client and server MUST send ChangeCipherSpec messages and proceed\n directly to Finished messages. Once the re-establishment is\n complete, the client and server MAY begin to exchange application\n layer data. (See flow chart below.) If a Session ID match is not\n found, the server generates a new session ID, and the TLS client and\n server perform a full handshake.\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 36]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2. Message flow for an abbreviated handshake\n\n The contents and significance of each message will be presented in\n detail in the following sections.\n\n7.4. Handshake Protocol\n\n The TLS Handshake Protocol is one of the defined higher-level clients\n of the TLS Record Protocol. This protocol is used to negotiate the\n secure attributes of a session. Handshake messages are supplied to\n the TLS record layer, where they are encapsulated within one or more\n TLSPlaintext structures, which are processed and transmitted as\n specified by the current active session state.\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n } body;\n } Handshake;\n\n\n\nDierks & Rescorla Standards Track [Page 37]\n\f\nRFC 5246 TLS August 2008\n\n\n The handshake protocol messages are presented below in the order they\n MUST be sent; sending handshake messages in an unexpected order\n results in a fatal error. Unneeded handshake messages can be\n omitted, however. Note one exception to the ordering: the\n Certificate message is used twice in the handshake (from server to\n client, then from client to server), but described only in its first\n position. The one message that is not bound by these ordering rules\n is the HelloRequest message, which can be sent at any time, but which\n SHOULD be ignored by the client if it arrives in the middle of a\n handshake.\n\n New handshake message types are assigned by IANA as described in\n Section 12.\n\n7.4.1. Hello Messages\n\n The hello phase messages are used to exchange security enhancement\n capabilities between the client and server. When a new session\n begins, the record layer's connection state encryption, hash, and\n compression algorithms are initialized to null. The current\n connection state is used for renegotiation messages.\n\n7.4.1.1. Hello Request\n\n When this message will be sent:\n\n The HelloRequest message MAY be sent by the server at any time.\n\n Meaning of this message:\n\n HelloRequest is a simple notification that the client should begin\n the negotiation process anew. In response, the client should send\n a ClientHello message when convenient. This message is not\n intended to establish which side is the client or server but\n merely to initiate a new negotiation. Servers SHOULD NOT send a\n HelloRequest immediately upon the client's initial connection. It\n is the client's job to send a ClientHello at that time.\n\n This message will be ignored by the client if the client is\n currently negotiating a session. This message MAY be ignored by\n the client if it does not wish to renegotiate a session, or the\n client may, if it wishes, respond with a no_renegotiation alert.\n Since handshake messages are intended to have transmission\n precedence over application data, it is expected that the\n negotiation will begin before no more than a few records are\n received from the client. If the server sends a HelloRequest but\n does not receive a ClientHello in response, it may close the\n connection with a fatal alert.\n\n\n\nDierks & Rescorla Standards Track [Page 38]\n\f\nRFC 5246 TLS August 2008\n\n\n After sending a HelloRequest, servers SHOULD NOT repeat the\n request until the subsequent handshake negotiation is complete.\n\n Structure of this message:\n\n struct { } HelloRequest;\n\n This message MUST NOT be included in the message hashes that are\n maintained throughout the handshake and used in the Finished messages\n and the certificate verify message.\n\n7.4.1.2. Client Hello\n\n When this message will be sent:\n\n When a client first connects to a server, it is required to send\n the ClientHello as its first message. The client can also send a\n ClientHello in response to a HelloRequest or on its own initiative\n in order to renegotiate the security parameters in an existing\n connection.\n\n Structure of this message:\n\n The ClientHello message includes a random structure, which is used\n later in the protocol.\n\n struct {\n uint32 gmt_unix_time;\n opaque random_bytes[28];\n } Random;\n\n gmt_unix_time\n The current time and date in standard UNIX 32-bit format\n (seconds since the midnight starting Jan 1, 1970, UTC, ignoring\n leap seconds) according to the sender's internal clock. Clocks\n are not required to be set correctly by the basic TLS protocol;\n higher-level or application protocols may define additional\n requirements. Note that, for historical reasons, the data\n element is named using GMT, the predecessor of the current\n worldwide time base, UTC.\n\n random_bytes\n 28 bytes generated by a secure random number generator.\n\n The ClientHello message includes a variable-length session\n identifier. If not empty, the value identifies a session between the\n same client and server whose security parameters the client wishes to\n reuse. The session identifier MAY be from an earlier connection,\n\n\n\nDierks & Rescorla Standards Track [Page 39]\n\f\nRFC 5246 TLS August 2008\n\n\n this connection, or from another currently active connection. The\n second option is useful if the client only wishes to update the\n random structures and derived values of a connection, and the third\n option makes it possible to establish several independent secure\n connections without repeating the full handshake protocol. These\n independent connections may occur sequentially or simultaneously; a\n SessionID becomes valid when the handshake negotiating it completes\n with the exchange of Finished messages and persists until it is\n removed due to aging or because a fatal error was encountered on a\n connection associated with the session. The actual contents of the\n SessionID are defined by the server.\n\n opaque SessionID<0..32>;\n\n Warning: Because the SessionID is transmitted without encryption or\n immediate MAC protection, servers MUST NOT place confidential\n information in session identifiers or let the contents of fake\n session identifiers cause any breach of security. (Note that the\n content of the handshake as a whole, including the SessionID, is\n protected by the Finished messages exchanged at the end of the\n handshake.)\n\n The cipher suite list, passed from the client to the server in the\n ClientHello message, contains the combinations of cryptographic\n algorithms supported by the client in order of the client's\n preference (favorite choice first). Each cipher suite defines a key\n exchange algorithm, a bulk encryption algorithm (including secret key\n length), a MAC algorithm, and a PRF. The server will select a cipher\n suite or, if no acceptable choices are presented, return a handshake\n failure alert and close the connection. If the list contains cipher\n suites the server does not recognize, support, or wish to use, the\n server MUST ignore those cipher suites, and process the remaining\n ones as usual.\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n The ClientHello includes a list of compression algorithms supported\n by the client, ordered according to the client's preference.\n\n enum { null(0), (255) } CompressionMethod;\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 40]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ProtocolVersion client_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suites<2..2^16-2>;\n CompressionMethod compression_methods<1..2^8-1>;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ClientHello;\n\n TLS allows extensions to follow the compression_methods field in an\n extensions block. The presence of extensions can be detected by\n determining whether there are bytes following the compression_methods\n at the end of the ClientHello. Note that this method of detecting\n optional data differs from the normal TLS method of having a\n variable-length field, but it is used for compatibility with TLS\n before extensions were defined.\n\n client_version\n The version of the TLS protocol by which the client wishes to\n communicate during this session. This SHOULD be the latest\n (highest valued) version supported by the client. For this\n version of the specification, the version will be 3.3 (see\n Appendix E for details about backward compatibility).\n\n random\n A client-generated random structure.\n\n session_id\n The ID of a session the client wishes to use for this connection.\n This field is empty if no session_id is available, or if the\n client wishes to generate new security parameters.\n\n cipher_suites\n This is a list of the cryptographic options supported by the\n client, with the client's first preference first. If the\n session_id field is not empty (implying a session resumption\n request), this vector MUST include at least the cipher_suite from\n that session. Values are defined in Appendix A.5.\n\n compression_methods\n This is a list of the compression methods supported by the client,\n sorted by client preference. If the session_id field is not empty\n (implying a session resumption request), it MUST include the\n\n\n\nDierks & Rescorla Standards Track [Page 41]\n\f\nRFC 5246 TLS August 2008\n\n\n compression_method from that session. This vector MUST contain,\n and all implementations MUST support, CompressionMethod.null.\n Thus, a client and server will always be able to agree on a\n compression method.\n\n extensions\n Clients MAY request extended functionality from servers by sending\n data in the extensions field. The actual \"Extension\" format is\n defined in Section 7.4.1.4.\n\n In the event that a client requests additional functionality using\n extensions, and this functionality is not supplied by the server, the\n client MAY abort the handshake. A server MUST accept ClientHello\n messages both with and without the extensions field, and (as for all\n other messages) it MUST check that the amount of data in the message\n precisely matches one of these formats; if not, then it MUST send a\n fatal \"decode_error\" alert.\n\n After sending the ClientHello message, the client waits for a\n ServerHello message. Any handshake message returned by the server,\n except for a HelloRequest, is treated as a fatal error.\n\n7.4.1.3. Server Hello\n\n When this message will be sent:\n\n The server will send this message in response to a ClientHello\n message when it was able to find an acceptable set of algorithms.\n If it cannot find such a match, it will respond with a handshake\n failure alert.\n\n Structure of this message:\n\n struct {\n ProtocolVersion server_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ServerHello;\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 42]\n\f\nRFC 5246 TLS August 2008\n\n\n The presence of extensions can be detected by determining whether\n there are bytes following the compression_method field at the end of\n the ServerHello.\n\n server_version\n This field will contain the lower of that suggested by the client\n in the client hello and the highest supported by the server. For\n this version of the specification, the version is 3.3. (See\n Appendix E for details about backward compatibility.)\n\n random\n This structure is generated by the server and MUST be\n independently generated from the ClientHello.random.\n\n session_id\n This is the identity of the session corresponding to this\n connection. If the ClientHello.session_id was non-empty, the\n server will look in its session cache for a match. If a match is\n found and the server is willing to establish the new connection\n using the specified session state, the server will respond with\n the same value as was supplied by the client. This indicates a\n resumed session and dictates that the parties must proceed\n directly to the Finished messages. Otherwise, this field will\n contain a different value identifying the new session. The server\n may return an empty session_id to indicate that the session will\n not be cached and therefore cannot be resumed. If a session is\n resumed, it must be resumed using the same cipher suite it was\n originally negotiated with. Note that there is no requirement\n that the server resume any session even if it had formerly\n provided a session_id. Clients MUST be prepared to do a full\n negotiation -- including negotiating new cipher suites -- during\n any handshake.\n\n cipher_suite\n The single cipher suite selected by the server from the list in\n ClientHello.cipher_suites. For resumed sessions, this field is\n the value from the state of the session being resumed.\n\n compression_method\n The single compression algorithm selected by the server from the\n list in ClientHello.compression_methods. For resumed sessions,\n this field is the value from the resumed session state.\n\n extensions\n A list of extensions. Note that only extensions offered by the\n client can appear in the server's list.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 43]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.1.4. Hello Extensions\n\n The extension format is:\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n signature_algorithms(13), (65535)\n } ExtensionType;\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The initial set of extensions is defined in a companion document\n [TLSEXT]. The list of extension types is maintained by IANA as\n described in Section 12.\n\n An extension type MUST NOT appear in the ServerHello unless the same\n extension type appeared in the corresponding ClientHello. If a\n client receives an extension type in ServerHello that it did not\n request in the associated ClientHello, it MUST abort the handshake\n with an unsupported_extension fatal alert.\n\n Nonetheless, \"server-oriented\" extensions may be provided in the\n future within this framework. Such an extension (say, of type x)\n would require the client to first send an extension of type x in a\n ClientHello with empty extension_data to indicate that it supports\n the extension type. In this case, the client is offering the\n capability to understand the extension type, and the server is taking\n the client up on its offer.\n\n When multiple extensions of different types are present in the\n ClientHello or ServerHello messages, the extensions MAY appear in any\n order. There MUST NOT be more than one extension of the same type.\n\n Finally, note that extensions can be sent both when starting a new\n session and when requesting session resumption. Indeed, a client\n that requests session resumption does not in general know whether the\n server will accept this request, and therefore it SHOULD send the\n same extensions as it would send if it were not attempting\n resumption.\n\n\n\nDierks & Rescorla Standards Track [Page 44]\n\f\nRFC 5246 TLS August 2008\n\n\n In general, the specification of each extension type needs to\n describe the effect of the extension both during full handshake and\n session resumption. Most current TLS extensions are relevant only\n when a session is initiated: when an older session is resumed, the\n server does not process these extensions in Client Hello, and does\n not include them in Server Hello. However, some extensions may\n specify different behavior during session resumption.\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions, and some are simply refusals to support particular\n features. In general, error alerts should be used for the former,\n and a field in the server extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem.\n\n Often the fact that the extension fields are included in the\n inputs to the Finished message hashes will be sufficient, but\n extreme care is needed when the extension changes the meaning of\n messages sent in the handshake phase. Designers and implementors\n should be aware of the fact that until the handshake has been\n authenticated, active attackers can modify messages and insert,\n remove, or replace extensions.\n\n - It would be technically possible to use extensions to change major\n aspects of the design of TLS; for example the design of cipher\n suite negotiation. This is not recommended; it would be more\n appropriate to define a new version of TLS -- particularly since\n the TLS handshake algorithms have specific protection against\n version rollback attacks based on the version number, and the\n possibility of version rollback should be a significant\n consideration in any major design change.\n\n7.4.1.4.1. Signature Algorithms\n\n The client uses the \"signature_algorithms\" extension to indicate to\n the server which signature/hash algorithm pairs may be used in\n digital signatures. The \"extension_data\" field of this extension\n contains a \"supported_signature_algorithms\" value.\n\n\n\nDierks & Rescorla Standards Track [Page 45]\n\f\nRFC 5246 TLS August 2008\n\n\n enum {\n none(0), md5(1), sha1(2), sha224(3), sha256(4), sha384(5),\n sha512(6), (255)\n } HashAlgorithm;\n\n enum { anonymous(0), rsa(1), dsa(2), ecdsa(3), (255) }\n SignatureAlgorithm;\n\n struct {\n HashAlgorithm hash;\n SignatureAlgorithm signature;\n } SignatureAndHashAlgorithm;\n\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2..2^16-2>;\n\n Each SignatureAndHashAlgorithm value lists a single hash/signature\n pair that the client is willing to verify. The values are indicated\n in descending order of preference.\n\n Note: Because not all signature algorithms and hash algorithms may be\n accepted by an implementation (e.g., DSA with SHA-1, but not\n SHA-256), algorithms here are listed in pairs.\n\n hash\n This field indicates the hash algorithm which may be used. The\n values indicate support for unhashed data, MD5 [MD5], SHA-1,\n SHA-224, SHA-256, SHA-384, and SHA-512 [SHS], respectively. The\n \"none\" value is provided for future extensibility, in case of a\n signature algorithm which does not require hashing before signing.\n\n signature\n This field indicates the signature algorithm that may be used.\n The values indicate anonymous signatures, RSASSA-PKCS1-v1_5\n [PKCS1] and DSA [DSS], and ECDSA [ECDSA], respectively. The\n \"anonymous\" value is meaningless in this context but used in\n Section 7.4.3. It MUST NOT appear in this extension.\n\n The semantics of this extension are somewhat complicated because the\n cipher suite indicates permissible signature algorithms but not hash\n algorithms. Sections 7.4.2 and 7.4.3 describe the appropriate rules.\n\n If the client supports only the default hash and signature algorithms\n (listed in this section), it MAY omit the signature_algorithms\n extension. If the client does not support the default algorithms, or\n supports other hash and signature algorithms (and it is willing to\n use them for verifying messages sent by the server, i.e., server\n certificates and server key exchange), it MUST send the\n\n\n\nDierks & Rescorla Standards Track [Page 46]\n\f\nRFC 5246 TLS August 2008\n\n\n signature_algorithms extension, listing the algorithms it is willing\n to accept.\n\n If the client does not send the signature_algorithms extension, the\n server MUST do the following:\n\n - If the negotiated key exchange algorithm is one of (RSA, DHE_RSA,\n DH_RSA, RSA_PSK, ECDH_RSA, ECDHE_RSA), behave as if client had\n sent the value {sha1,rsa}.\n\n - If the negotiated key exchange algorithm is one of (DHE_DSS,\n DH_DSS), behave as if the client had sent the value {sha1,dsa}.\n\n - If the negotiated key exchange algorithm is one of (ECDH_ECDSA,\n ECDHE_ECDSA), behave as if the client had sent value {sha1,ecdsa}.\n\n Note: this is a change from TLS 1.1 where there are no explicit\n rules, but as a practical matter one can assume that the peer\n supports MD5 and SHA-1.\n\n Note: this extension is not meaningful for TLS versions prior to 1.2.\n Clients MUST NOT offer it if they are offering prior versions.\n However, even if clients do offer it, the rules specified in [TLSEXT]\n require servers to ignore extensions they do not understand.\n\n Servers MUST NOT send this extension. TLS servers MUST support\n receiving this extension.\n\n When performing session resumption, this extension is not included in\n Server Hello, and the server ignores the extension in Client Hello\n (if present).\n\n7.4.2. Server Certificate\n\n When this message will be sent:\n\n The server MUST send a Certificate message whenever the agreed-\n upon key exchange method uses certificates for authentication\n (this includes all key exchange methods defined in this document\n except DH_anon). This message will always immediately follow the\n ServerHello message.\n\n Meaning of this message:\n\n This message conveys the server's certificate chain to the client.\n\n The certificate MUST be appropriate for the negotiated cipher\n suite's key exchange algorithm and any negotiated extensions.\n\n\n\nDierks & Rescorla Standards Track [Page 47]\n\f\nRFC 5246 TLS August 2008\n\n\n Structure of this message:\n\n opaque ASN.1Cert<1..2^24-1>;\n\n struct {\n ASN.1Cert certificate_list<0..2^24-1>;\n } Certificate;\n\n certificate_list\n This is a sequence (chain) of certificates. The sender's\n certificate MUST come first in the list. Each following\n certificate MUST directly certify the one preceding it. Because\n certificate validation requires that root keys be distributed\n independently, the self-signed certificate that specifies the root\n certificate authority MAY be omitted from the chain, under the\n assumption that the remote end must already possess it in order to\n validate it in any case.\n\n The same message type and structure will be used for the client's\n response to a certificate request message. Note that a client MAY\n send no certificates if it does not have an appropriate certificate\n to send in response to the server's authentication request.\n\n Note: PKCS #7 [PKCS7] is not used as the format for the certificate\n vector because PKCS #6 [PKCS6] extended certificates are not used.\n Also, PKCS #7 defines a SET rather than a SEQUENCE, making the task\n of parsing the list more difficult.\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3, unless explicitly negotiated\n otherwise (e.g., [TLSPGP]).\n\n - The end entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected key exchange\n algorithm.\n\n Key Exchange Alg. Certificate Key Type\n\n RSA RSA public key; the certificate MUST allow the\n RSA_PSK key to be used for encryption (the\n keyEncipherment bit MUST be set if the key\n usage extension is present).\n Note: RSA_PSK is defined in [TLSPSK].\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 48]\n\f\nRFC 5246 TLS August 2008\n\n\n DHE_RSA RSA public key; the certificate MUST allow the\n ECDHE_RSA key to be used for signing (the\n digitalSignature bit MUST be set if the key\n usage extension is present) with the signature\n scheme and hash algorithm that will be employed\n in the server key exchange message.\n Note: ECDHE_RSA is defined in [TLSECC].\n\n DHE_DSS DSA public key; the certificate MUST allow the\n key to be used for signing with the hash\n algorithm that will be employed in the server\n key exchange message.\n\n DH_DSS Diffie-Hellman public key; the keyAgreement bit\n DH_RSA MUST be set if the key usage extension is\n present.\n\n ECDH_ECDSA ECDH-capable public key; the public key MUST\n ECDH_RSA use a curve and point format supported by the\n client, as described in [TLSECC].\n\n ECDHE_ECDSA ECDSA-capable public key; the certificate MUST\n allow the key to be used for signing with the\n hash algorithm that will be employed in the\n server key exchange message. The public key\n MUST use a curve and point format supported by\n the client, as described in [TLSECC].\n\n - The \"server_name\" and \"trusted_ca_keys\" extensions [TLSEXT] are\n used to guide certificate selection.\n\n\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) Y. Pettersen\nRequest for Comments: 6961 June 2013\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\n\nAbstract\n\n This document defines the Transport Layer Security (TLS) Certificate\n Status Version 2 Extension to allow clients to specify and support\n several certificate status methods. (The use of the Certificate\n Status extension is commonly referred to as \"OCSP stapling\".) Also\n defined is a new method based on the Online Certificate Status\n Protocol (OCSP) that servers can use to provide status information\n about not only the server's own certificate but also the status of\n intermediate certificates in the chain.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6961.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 1]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nCopyright Notice\n\n Copyright (c) 2013 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n1. Introduction\n\n The Transport Layer Security (TLS) Extension [RFC6066] framework\n defines, among other extensions, the Certificate Status extension\n (also referred to as \"OCSP stapling\") that clients can use to request\n the server's copy of the current status of its certificate. The\n benefits of this extension include a reduced number of roundtrips and\n network delays for the client to verify the status of the server's\n certificate and a reduced load on the certificate issuer's status\n response servers, thus solving a problem that can become significant\n when the issued certificate is presented by a frequently visited\n server.\n\n There are two problems with the existing Certificate Status\n extension. First, it does not provide functionality to request the\n status information about intermediate Certification Authority (CA)\n certificates, which means the client has to request status\n information through other methods, such as Certificate Revocation\n Lists (CRLs), introducing further delays. Second, the current format\n of the extension and requirements in the TLS protocol prevent a\n client from offering the server multiple status methods.\n\n\n\nPettersen Standards Track [Page 2]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n Many CAs are now issuing intermediate CA certificates that not only\n specify the publication point for their CRLs in a CRL Distribution\n Point [RFC5280] but also specify a URL for their OCSP [RFC6960]\n server in Authority Information Access [RFC5280]. Given that\n client-cached CRLs are frequently out of date, clients would benefit\n from using OCSP to access up-to-date status information about\n intermediate CA certificates. The benefit to the issuing CA is less\n clear, as providing the bandwidth for the OCSP responder can be\n costly, especially for CAs with many high-traffic subscriber sites,\n and this cost is a concern for many CAs. There are cases where OCSP\n requests for a single high-traffic site caused significant network\n problems for the issuing CA.\n\n Clients will benefit from the TLS server providing certificate status\n information regardless of type, not just for the server certificate\n but also for the intermediate CA certificates. Combining the status\n checks into one extension will reduce the roundtrips needed to\n complete the handshake by the client to just those needed for\n negotiating the TLS connection. Also, for the Certification\n Authorities, the load on their servers will depend on the number of\n certificates they have issued, not on the number of visitors to those\n sites. Additionally, using this extension significantly reduces\n privacy concerns around the clients informing the certificate issuer\n about which sites they are visiting.\n\n For such a new system to be introduced seamlessly, clients need to be\n able to indicate support for the existing OCSP Certificate Status\n method and a new multiple-OCSP mode.\n\n Unfortunately, the definition of the Certificate Status extension\n only allows a single Certificate Status extension to be defined in a\n single extension record in the handshake, and the TLS protocol\n [RFC5246] only allows a single record in the extension list for any\n given extension. This means that it is not possible for clients to\n indicate support for new methods while still supporting older\n methods, which would cause problems for interoperability between\n newer clients and older servers. This will not just be an issue for\n the multiple status request mode proposed above but also for any\n other future status methods that might be introduced. This will be\n true not just for the current PKIX infrastructure [RFC5280] but also\n for alternative PKI structures.\n\n The solution to this problem is to define a new extension,\n \"status_request_v2\", with an extended format that allows the client\n to indicate support for multiple status request methods. This is\n implemented using a list of CertificateStatusRequestItemV2 records in\n the extension record. As the server will select the single status\n\n\n\n\nPettersen Standards Track [Page 3]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n method based on the selected cipher suite and the certificate\n presented, no significant changes are needed in the server's\n extension format.\n\n1.1. Requirements Language\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [RFC2119].\n\n1.2. Presentation Language\n\n This document defines protocol structures using the same conventions\n and presentation language as defined in Section 4 of [RFC5246].\n\n2. Multiple Certificate Status Extension\n\n2.1. New Extension\n\n The extension defined by this document is indicated by\n \"status_request_v2\" in the ExtensionType enum (originally defined by\n [RFC6066]), which uses the following value:\n\n enum {\n status_request_v2(17), (65535)\n } ExtensionType;\n\n2.2. Multiple Certificate Status Request Record\n\n Clients that support a certificate status protocol like OCSP may send\n the \"status_request_v2\" extension to the server in order to use the\n TLS handshake to transfer such data instead of downloading it through\n separate connections. When using this extension, the\n \"extension_data\" field (defined in Section 7.4.1.4 of [RFC5246]) of\n the extension SHALL contain a CertificateStatusRequestListV2 where:\n\n struct {\n CertificateStatusType status_type;\n uint16 request_length; /* Length of request field in bytes */\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n case ocsp_multi: OCSPStatusRequest;\n } request;\n } CertificateStatusRequestItemV2;\n\n enum { ocsp(1), ocsp_multi(2), (255) } CertificateStatusType;\n\n\n\n\n\nPettersen Standards Track [Page 4]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n struct {\n CertificateStatusRequestItemV2\n certificate_status_req_list<1..2^16-1>;\n } CertificateStatusRequestListV2;\n\n In the OCSPStatusRequest (originally defined by [RFC6066]), the\n \"ResponderID\" provides a list of OCSP responders that the client\n trusts. A zero-length \"responder_id_list\" sequence has the special\n meaning that the responders are implicitly known to the server, e.g.,\n by prior arrangement, or are identified by the certificates used by\n the server. \"Extensions\" is a DER encoding [X.690] of the OCSP\n request extensions, and if the server supports the forwarding of OCSP\n request extensions, this value MUST be forwarded without\n modification.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC6960]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a DER-encoded zero-length ASN.1 SEQUENCE,\n which is not valid for the \"Extensions\" type).\n\n Servers that support a client's selection of responders using the\n ResponderID field could implement this selection by matching the\n responder ID values from the client's list with the ResponderIDs of\n known OCSP responders, either by using a binary compare of the values\n or a hash calculation and compare method.\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement). This\n has been clarified in [RFC6960].\n\n The items in the list of CertificateStatusRequestItemV2 entries are\n ordered according to the client's preference (favorite choice first).\n\n A server that receives a client hello containing the\n \"status_request_v2\" extension MAY return a suitable certificate\n status response message to the client along with the server's\n\n\n\nPettersen Standards Track [Page 5]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n certificate message. If OCSP is requested, it SHOULD use the\n information contained in the extension when selecting an OCSP\n responder and SHOULD include request_extensions in the OCSP request.\n\n The server returns a certificate status response along with its\n certificate by sending a \"CertificateStatus\" message (originally\n defined by [RFC6066]) immediately after the \"Certificate\" message\n (Section 7.4.2 of [RFC5246]) (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message in response to a \"status_request_v2\"\n request, then the server MUST have included an extension of type\n \"status_request_v2\" with empty \"extension_data\" in the extended\n server hello.\n\n The \"CertificateStatus\" message is conveyed using the handshake\n message type \"certificate_status\" (defined in [RFC6066]) as follows\n (updated from the definition in [RFC6066]):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n case ocsp_multi: OCSPResponseList;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<0..2^24-1>;\n\n struct {\n OCSPResponse ocsp_response_list<1..2^24-1>;\n } OCSPResponseList;\n\n An \"OCSPResponse\" element (originally defined by [RFC6066]) contains\n a complete, DER-encoded OCSP response (using the ASN.1 [X.680] type\n OCSPResponse defined in [RFC6960]). Only one OCSP response, with a\n length of at least one byte, may be sent for status_type \"ocsp\".\n\n An \"ocsp_response_list\" contains a list of \"OCSPResponse\" elements,\n as specified above, each containing the OCSP response for the\n matching corresponding certificate in the server's Certificate TLS\n handshake message. That is, the first entry is the OCSP response for\n the first certificate in the Certificate list, the second entry is\n the response for the second certificate, and so on. The list MAY\n contain fewer OCSP responses than there were certificates in the\n Certificate handshake message, but there MUST NOT be more responses\n than there were certificates in the list. Individual elements of the\n list MAY have a length of 0 (zero) bytes if the server does not have\n the OCSP response for that particular certificate stored, in which\n\n\n\nPettersen Standards Track [Page 6]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n case the client MUST act as if a response was not received for that\n particular certificate. If the client receives a\n \"ocsp_response_list\" that does not contain a response for one or more\n of the certificates in the completed certificate chain, the client\n SHOULD attempt to validate the certificate using an alternative\n retrieval method, such as downloading the relevant CRL; OCSP SHOULD\n in this situation only be used for the end-entity certificate, not\n intermediate CA certificates, for reasons stated above.\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if it has received a \"status_request_v2\" extension in\n the client hello message and has sent a \"status_request_v2\" extension\n in the server hello message. Additionally, note that a server MUST\n NOT send the \"CertificateStatus\" message unless it received either a\n \"status_request\" or \"status_request_v2\" extension in the client hello\n message and sent a corresponding \"status_request\" or\n \"status_request_v2\" extension in the server hello message.\n\n Clients requesting an OCSP response and receiving one or more OCSP\n responses in a \"CertificateStatus\" message MUST check the OCSP\n response(s) and abort the handshake if the response is a \"revoked\"\n status or other unacceptable responses (as determined by client\n policy) with a bad_certificate_status_response(113) alert. This\n alert is always fatal.\n\n If the OCSP response received from the server does not result in a\n definite \"good\" or \"revoked\" status, it is inconclusive. A TLS\n client in such a case MAY check the validity of the server\n certificate through other means, e.g., by directly querying the\n certificate issuer. If such processing still results in an\n inconclusive response, then the application using the TLS connection\n will have to decide whether to close the connection or not. Note\n that this problem cannot be decided by the generic TLS client code\n without information from the application. If the application doesn't\n provide any such information, then the client MUST abort the\n connection, since the server certificate has not been sufficiently\n validated.\n\n An example of where the application might wish to continue is with\n EAP-TLS (Extensible Authentication Protocol - TLS), where the\n application can use another mechanism to check the status of a\n certificate once it obtains network access. In this case, the\n application could have the client continue with the handshake, but it\n MUST NOT disclose a username and password until it has fully\n validated the server certificate.\n\n\n\n\n\n\nPettersen Standards Track [Page 7]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n3. IANA Considerations\n\n Section 2.1 defines the new TLS extension status_request_v2 (17)\n enum, which has been added to the \"ExtensionType Values\" list in the\n IANA \"Transport Layer Security (TLS) Extensions\" registry.\n\n Section 2.2 describes a TLS CertificateStatusType registry that is\n now maintained by IANA. The \"TLS Certificate Status Types\" registry\n has been created under the \"Transport Layer Security (TLS)\n Extensions\" registry. CertificateStatusType values are to be\n assigned via IETF Review as defined in [RFC5226]. The initial\n registry corresponds to the definition of \"CertificateStatusType\" in\n Section 2.2.\n\n Value Description Reference\n -----------------------------------------\n 0 Reserved [RFC6961]\n 1 ocsp [RFC6066] [RFC6961]\n 2 ocsp_multi [RFC6961]\n 3-255 Unassigned\n\n4. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security considerations for the particular extension\n specified in this document are given below. In general, implementers\n should continue to monitor the state of the art and address any\n weaknesses identified.\n\n4.1. Security Considerations for status_request_v2\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC6960] for further details.\n\n This extension allows the client to send arbitrary data to the\n server. The server implementers need to handle such data carefully\n to avoid introducing security vulnerabilities.\n\n The security considerations of [RFC6960] apply to OCSP requests and\n responses.\n\n\n\n\nPettersen Standards Track [Page 8]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n5. Acknowledgements\n\n This document is based on [RFC6066], authored by Donald Eastlake 3rd.\n\n6. References\n\n6.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246, August 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC6066] Eastlake, D., \"Transport Layer Security (TLS) Extensions:\n Extension Definitions\", RFC 6066, January 2011.\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, June 2013.\n\n [X.680] ITU-T Recommendation X.680 (2008) | ISO/IEC 8824-1:2008,\n \"Abstract Syntax Notation One (ASN.1): Specification of\n basic notation\", November 2008.\n\n [X.690] ITU-T Recommendation X.690 (2008) | ISO/IEC 8825-1:2008,\n \"ASN.1 encoding rules: Specification of Basic Encoding\n Rules (BER), Canonical Encoding Rules (CER) and\n Distinguished Encoding Rules (DER)\", November 2008.\n\n6.2. Informative References\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and C.\n Adams, \"X.509 Internet Public Key Infrastructure Online\n Certificate Status Protocol - OCSP\", RFC 2560, June 1999.\n\n\n\n\n\n\nPettersen Standards Track [Page 9]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nAuthor's Address\n\n Yngve N. Pettersen\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 10]\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 5705 RTFM, Inc.\nCategory: Standards Track March 2010\nISSN: 2070-1721\n\n\n Keying Material Exporters for Transport Layer Security (TLS)\n\nAbstract\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n to perform key establishment but then use some of the keying material\n for their own purposes. This document describes a general mechanism\n for allowing that.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/5705.\n\nCopyright Notice\n\n Copyright (c) 2010 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2\n 2. Conventions Used In This Document . . . . . . . . . . . . . . . 3\n 3. Binding to Application Contexts . . . . . . . . . . . . . . . . 3\n 4. Exporter Definition . . . . . . . . . . . . . . . . . . . . . . 4\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . . 5\n 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 6\n 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . 6\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 7\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 7\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 7\n\n1. Introduction\n\n Note: The mechanism described in this document was previously known\n as \"TLS Extractors\" but was changed to avoid a name conflict\n with the use of the term \"Extractor\" in the cryptographic\n community.\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n [RFC5246] or Datagram TLS (DTLS) [RFC4347] to perform key\n establishment but then use some of the keying material for their own\n purposes. A typical example is DTLS-SRTP [DTLS-SRTP], a key\n management scheme for the Secure Real-time Transport Protocol (SRTP)\n that uses DTLS to perform a key exchange and negotiate the SRTP\n [RFC3711] protection suite and then uses the DTLS master_secret to\n generate the SRTP keys.\n\n These applications imply a need to be able to export keying material\n (later called Exported Keying Material or EKM) from TLS/DTLS to an\n application or protocol residing at an upper layer, and to securely\n agree on the upper-layer context where the keying material will be\n used. The mechanism for exporting the keying material has the\n following requirements:\n\n o Both client and server need to be able to export the same EKM\n value.\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n o EKM values should be indistinguishable from random data to\n attackers who don't know the master_secret.\n\n o It should be possible to export multiple EKM values from the same\n TLS/DTLS association.\n\n o Knowing one EKM value should not reveal any useful information\n about the master_secret or about other EKM values.\n\n The mechanism described in this document is intended to fulfill these\n requirements. This mechanism is compatible with all versions of TLS.\n\n2. Conventions Used In This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Binding to Application Contexts\n\n In addition to using an exporter to obtain keying material, an\n application using the keying material has to securely establish the\n upper-layer context where the keying material will be used. The\n details of this context depend on the application, but it could\n include things such as algorithms and parameters that will be used\n with the keys, identifier(s) for the endpoint(s) who will use the\n keys, identifier(s) for the session(s) where the keys will be used,\n and the lifetime(s) for the context and/or keys. At a minimum, there\n should be some mechanism for signaling that an exporter will be used.\n\n This specification does not mandate a single mechanism for agreeing\n on such context; instead, there are several possibilities that can be\n used (and can complement each other). For example:\n\n o Information about the upper-layer context can be included in the\n optional data after the exporter label (see Section 4).\n\n o Information about the upper-layer context can be exchanged in TLS\n extensions included in the ClientHello and ServerHello messages.\n This approach is used in [DTLS-SRTP]. The handshake messages are\n protected by the Finished messages, so once the handshake\n completes, the peers will have the same view of the information.\n Extensions also allow a limited form of negotiation: for example,\n the TLS client could propose several alternatives for some context\n parameters, and the TLS server could select one of them.\n\n o The upper-layer protocol can include its own handshake, which can\n be protected using the keys exported by TLS.\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n No matter how the context is agreed, it is required that it has one\n part that indicates which application will use the exported keys.\n This part is the disambiguating label string (see Section 4).\n\n It is important to note that just embedding TLS messages in the\n upper-layer protocol may not automatically secure all the important\n context information, since the upper-layer messages are not covered\n by TLS Finished messages.\n\n4. Exporter Definition\n\n The output of the exporter is intended to be used in a single scope,\n which is associated with the TLS session, the label, and the context\n value.\n\n The exporter takes three input values:\n\n o a disambiguating label string,\n\n o a per-association context value provided by the application using\n the exporter, and\n\n o a length value.\n\n If no context is provided, it then computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random\n )[length]\n\n If context is provided, it computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random +\n context_value_length + context_value\n )[length]\n\n Where PRF is the TLS Pseudorandom Function in use for the session.\n The output is a pseudorandom bit string of length bytes generated\n from the master_secret. (This construction allows for\n interoperability with older exporter-type constructions which do not\n use context values, e.g., [RFC5281]).\n\n Labels here have the same definition as in TLS, i.e., an ASCII string\n with no terminating NULL. Label values beginning with \"EXPERIMENTAL\"\n MAY be used for private use without registration. All other label\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n values MUST be registered via Specification Required as described by\n RFC 5226 [RFC5226]. Note that exporter labels have the potential to\n collide with existing PRF labels. In order to prevent this, labels\n SHOULD begin with \"EXPORTER\". This is not a MUST because there are\n existing uses that have labels which do not begin with this prefix.\n\n The context value allows the application using the exporter to mix\n its own data with the TLS PRF for the exporter output. One example\n of where this might be useful is an authentication setting where the\n client credentials are valid for more than one identity; the context\n value could then be used to mix the expected identity into the keying\n material, thus preventing substitution attacks. The context value\n length is encoded as an unsigned, 16-bit quantity (uint16; see\n [RFC5246], Section 4.4) representing the length of the context value.\n The context MAY be zero length. Because the context value is mixed\n with the master_secret via the PRF, it is safe to mix confidential\n information into the exporter, provided that the master_secret will\n not be known to the attacker.\n\n5. Security Considerations\n\n The prime security requirement for exporter outputs is that they be\n independent. More formally, after a particular TLS session, if an\n adversary is allowed to choose multiple (label, context value) pairs\n and is given the output of the PRF for those values, the attacker is\n still unable to distinguish between the output of the PRF for a\n (label, context value) pair (different from the ones that it\n submitted) and a random value of the same length. In particular,\n there may be settings, such as the one described in Section 4, where\n the attacker can control the context value; such an attacker MUST NOT\n be able to predict the output of the exporter. Similarly, an\n attacker who does not know the master secret should not be able to\n distinguish valid exporter outputs from random values. The current\n set of TLS PRFs is believed to meet this objective, provided the\n master secret is randomly generated.\n\n Because an exporter produces the same value if applied twice with the\n same label to the same master_secret, it is critical that two EKM\n values generated with the same label not be used for two different\n purposes -- hence, the requirement for IANA registration. However,\n because exporters depend on the TLS PRF, it is not a threat to the\n use of an EKM value generated from one label to reveal an EKM value\n generated from another label.\n\n With certain TLS cipher suites, the TLS master secret is not\n necessarily unique to a single TLS session. In particular, with RSA\n key exchange, a malicious party acting as TLS server in one session\n and as TLS client in another session can cause those two sessions to\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n have the same TLS master secret (though the sessions must be\n established simultaneously to get adequate control of the Random\n values). Applications using the EKM need to consider this in how\n they use the EKM; in some cases, requiring the use of other cipher\n suites (such as those using a Diffie-Hellman key exchange) may be\n advisable.\n\n Designing a secure mechanism that uses exporters is not necessarily\n straightforward. This document only provides the exporter mechanism,\n but the problem of agreeing on the surrounding context and the\n meaning of the information passed to and from the exporter remains.\n Any new uses of the exporter mechanism should be subject to careful\n review.\n\n6. IANA Considerations\n\n IANA has created a TLS Exporter Label registry for this purpose. The\n initial contents of the registry are given below:\n\n Value Reference Note\n ----------------------------- --------- ----\n client finished [RFC5246] (1)\n server finished [RFC5246] (1)\n master secret [RFC5246] (1)\n key expansion [RFC5246] (1)\n client EAP encryption [RFC5216]\n ttls keying material [RFC5281]\n ttls challenge [RFC5281]\n\n Note: (1) These entries are reserved and MUST NOT be used for the\n purpose described in RFC 5705, in order to avoid confusion with\n similar, but distinct, use in RFC 5246.\n\n Future values are allocated via the RFC 5226 Specification Required\n policy. The label is a string consisting of printable ASCII\n characters. IANA MUST also verify that one label is not a prefix of\n any other label. For example, labels \"key\" or \"master secretary\" are\n forbidden.\n\n7. Acknowledgments\n\n Thanks to Pasi Eronen for valuable comments and for the contents of\n the IANA section and Section 3. Thanks to David McGrew for helpful\n discussion of the security considerations and to Vijay Gurbani and\n Alfred Hoenes for editorial comments.\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246,\n August 2008.\n\n8.2. Informative References\n\n [DTLS-SRTP] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for Secure\n Real-time Transport Protocol (SRTP)\", Work in Progress,\n February 2009.\n\n [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and\n K. Norrman, \"The Secure Real-time Transport Protocol\n (SRTP)\", RFC 3711, March 2004.\n\n [RFC4347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security\", RFC 4347, April 2006.\n\n [RFC5216] Simon, D., Aboba, B., and R. Hurst, \"The EAP-TLS\n Authentication Protocol\", RFC 5216, March 2008.\n\n [RFC5281] Funk, P. and S. Blake-Wilson, \"Extensible Authentication\n Protocol Tunneled Transport Layer Security Authenticated\n Protocol Version 0 (EAP-TTLSv0)\", RFC 5281, August 2008.\n\nAuthor's Address\n\n Eric Rescorla\n RTFM, Inc.\n 2064 Edgewood Drive\n Palo Alto, CA 94303\n USA\n\n EMail: [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) D. Eastlake 3rd\nRequest for Comments: 6066 Huawei\nObsoletes: 4366 January 2011\nCategory: Standards Track\nISSN: 2070-1721\n\n\n Transport Layer Security (TLS) Extensions: Extension Definitions\n\nAbstract\n\n This document provides specifications for existing TLS extensions.\n It is a companion document for RFC 5246, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\". The extensions specified are\n server_name, max_fragment_length, client_certificate_url,\n trusted_ca_keys, truncated_hmac, and status_request.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6066.\n\nCopyright Notice\n\n Copyright (c) 2011 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n\n\n\n\n\nEastlake Standards Track [Page 1]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction ....................................................3\n 1.1. Specific Extensions Covered ................................3\n 1.2. Conventions Used in This Document ..........................5\n 2. Extensions to the Handshake Protocol ............................5\n 3. Server Name Indication ..........................................6\n 4. Maximum Fragment Length Negotiation .............................8\n 5. Client Certificate URLs .........................................9\n 6. Trusted CA Indication ..........................................12\n 7. Truncated HMAC .................................................13\n 8. Certificate Status Request .....................................14\n 9. Error Alerts ...................................................16\n 10. IANA Considerations ...........................................17\n 10.1. pkipath MIME Type Registration ...........................17\n 10.2. Reference for TLS Alerts, TLS HandshakeTypes, and\n ExtensionTypes ...........................................19\n 11. Security Considerations .......................................19\n 11.1. Security Considerations for server_name ..................19\n 11.2. Security Considerations for max_fragment_length ..........20\n 11.3. Security Considerations for client_certificate_url .......20\n 11.4. Security Considerations for trusted_ca_keys ..............21\n 11.5. Security Considerations for truncated_hmac ...............21\n 11.6. Security Considerations for status_request ...............22\n 12. Normative References ..........................................22\n 13. Informative References ........................................23\n Appendix A. Changes from RFC 4366 .................................24\n Appendix B. Acknowledgements ......................................25\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 2]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1. Introduction\n\n The Transport Layer Security (TLS) Protocol Version 1.2 is specified\n in [RFC5246]. That specification includes the framework for\n extensions to TLS, considerations in designing such extensions (see\n Section 7.4.1.4 of [RFC5246]), and IANA Considerations for the\n allocation of new extension code points; however, it does not specify\n any particular extensions other than Signature Algorithms (see\n Section 7.4.1.4.1 of [RFC5246]).\n\n This document provides the specifications for existing TLS\n extensions. It is, for the most part, the adaptation and editing of\n material from RFC 4366, which covered TLS extensions for TLS 1.0 (RFC\n 2246) and TLS 1.1 (RFC 4346).\n\n1.1. Specific Extensions Covered\n\n The extensions described here focus on extending the functionality\n provided by the TLS protocol message formats. Other issues, such as\n the addition of new cipher suites, are deferred.\n\n The extension types defined in this document are:\n\n enum {\n server_name(0), max_fragment_length(1),\n client_certificate_url(2), trusted_ca_keys(3),\n truncated_hmac(4), status_request(5), (65535)\n } ExtensionType;\n\n Specifically, the extensions described in this document:\n\n - Allow TLS clients to provide to the TLS server the name of the\n server they are contacting. This functionality is desirable in\n order to facilitate secure connections to servers that host\n multiple 'virtual' servers at a single underlying network address.\n\n - Allow TLS clients and servers to negotiate the maximum fragment\n length to be sent. This functionality is desirable as a result of\n memory constraints among some clients, and bandwidth constraints\n among some access networks.\n\n - Allow TLS clients and servers to negotiate the use of client\n certificate URLs. This functionality is desirable in order to\n conserve memory on constrained clients.\n\n\n\n\n\n\n\nEastlake Standards Track [Page 3]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - Allow TLS clients to indicate to TLS servers which certification\n authority (CA) root keys they possess. This functionality is\n desirable in order to prevent multiple handshake failures\n involving TLS clients that are only able to store a small number\n of CA root keys due to memory limitations.\n\n - Allow TLS clients and servers to negotiate the use of truncated\n Message Authentication Codes (MACs). This functionality is\n desirable in order to conserve bandwidth in constrained access\n networks.\n\n - Allow TLS clients and servers to negotiate that the server sends\n the client certificate status information (e.g., an Online\n Certificate Status Protocol (OCSP) [RFC2560] response) during a\n TLS handshake. This functionality is desirable in order to avoid\n sending a Certificate Revocation List (CRL) over a constrained\n access network and therefore saving bandwidth.\n\n TLS clients and servers may use the extensions described in this\n document. The extensions are designed to be backwards compatible,\n meaning that TLS clients that support the extensions can talk to TLS\n servers that do not support the extensions, and vice versa.\n\n Note that any messages associated with these extensions that are sent\n during the TLS handshake MUST be included in the hash calculations\n involved in \"Finished\" messages.\n\n Note also that all the extensions defined in this document are\n relevant only when a session is initiated. A client that requests\n session resumption does not in general know whether the server will\n accept this request, and therefore it SHOULD send the same extensions\n as it would send if it were not attempting resumption. When a client\n includes one or more of the defined extension types in an extended\n client hello while requesting session resumption:\n\n - The server name indication extension MAY be used by the server\n when deciding whether or not to resume a session as described in\n Section 3.\n\n - If the resumption request is denied, the use of the extensions is\n negotiated as normal.\n\n - If, on the other hand, the older session is resumed, then the\n server MUST ignore the extensions and send a server hello\n containing none of the extension types. In this case, the\n functionality of these extensions negotiated during the original\n session initiation is applied to the resumed session.\n\n\n\n\nEastlake Standards Track [Page 4]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1.2. Conventions Used in This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n [RFC2119].\n\n2. Extensions to the Handshake Protocol\n\n This document specifies the use of two new handshake messages,\n \"CertificateURL\" and \"CertificateStatus\". These messages are\n described in Sections 5 and 8, respectively. The new handshake\n message structure therefore becomes:\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), certificate_url(21), certificate_status(22),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case certificate_url: CertificateURL;\n case certificate_status: CertificateStatus;\n } body;\n } Handshake;\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 5]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n3. Server Name Indication\n\n TLS does not provide a mechanism for a client to tell a server the\n name of the server it is contacting. It may be desirable for clients\n to provide this information to facilitate secure connections to\n servers that host multiple 'virtual' servers at a single underlying\n network address.\n\n In order to provide any of the server names, clients MAY include an\n extension of type \"server_name\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"ServerNameList\" where:\n\n struct {\n NameType name_type;\n select (name_type) {\n case host_name: HostName;\n } name;\n } ServerName;\n\n enum {\n host_name(0), (255)\n } NameType;\n\n opaque HostName<1..2^16-1>;\n\n struct {\n ServerName server_name_list<1..2^16-1>\n } ServerNameList;\n\n The ServerNameList MUST NOT contain more than one name of the same\n name_type. If the server understood the ClientHello extension but\n does not recognize the server name, the server SHOULD take one of two\n actions: either abort the handshake by sending a fatal-level\n unrecognized_name(112) alert or continue the handshake. It is NOT\n RECOMMENDED to send a warning-level unrecognized_name(112) alert,\n because the client's behavior in response to warning-level alerts is\n unpredictable. If there is a mismatch between the server name used\n by the client application and the server name of the credential\n chosen by the server, this mismatch will become apparent when the\n client application performs the server endpoint identification, at\n which point the client application will have to decide whether to\n proceed with the communication. TLS implementations are encouraged\n to make information available to application callers about warning-\n level alerts that were received or sent during a TLS handshake. Such\n information can be useful for diagnostic purposes.\n\n\n\n\n\nEastlake Standards Track [Page 6]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note: Earlier versions of this specification permitted multiple\n names of the same name_type. In practice, current client\n implementations only send one name, and the client cannot\n necessarily find out which name the server selected. Multiple\n names of the same name_type are therefore now prohibited.\n\n Currently, the only server names supported are DNS hostnames;\n however, this does not imply any dependency of TLS on DNS, and other\n name types may be added in the future (by an RFC that updates this\n document). The data structure associated with the host_name NameType\n is a variable-length vector that begins with a 16-bit length. For\n backward compatibility, all future data structures associated with\n new NameTypes MUST begin with a 16-bit length field. TLS MAY treat\n provided server names as opaque data and pass the names and types to\n the application.\n\n \"HostName\" contains the fully qualified DNS hostname of the server,\n as understood by the client. The hostname is represented as a byte\n string using ASCII encoding without a trailing dot. This allows the\n support of internationalized domain names through the use of A-labels\n defined in [RFC5890]. DNS hostnames are case-insensitive. The\n algorithm to compare hostnames is described in [RFC5890], Section\n 2.3.2.4.\n\n Literal IPv4 and IPv6 addresses are not permitted in \"HostName\".\n\n It is RECOMMENDED that clients include an extension of type\n \"server_name\" in the client hello whenever they locate a server by a\n supported name type.\n\n A server that receives a client hello containing the \"server_name\"\n extension MAY use the information contained in the extension to guide\n its selection of an appropriate certificate to return to the client,\n and/or other aspects of security policy. In this event, the server\n SHALL include an extension of type \"server_name\" in the (extended)\n server hello. The \"extension_data\" field of this extension SHALL be\n empty.\n\n When the server is deciding whether or not to accept a request to\n resume a session, the contents of a server_name extension MAY be used\n in the lookup of the session in the session cache. The client SHOULD\n include the same server_name extension in the session resumption\n request as it did in the full handshake that established the session.\n A server that implements this extension MUST NOT accept the request\n to resume the session if the server_name extension contains a\n different name. Instead, it proceeds with a full handshake to\n establish a new session. When resuming a session, the server MUST\n NOT include a server_name extension in the server hello.\n\n\n\nEastlake Standards Track [Page 7]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n If an application negotiates a server name using an application\n protocol and then upgrades to TLS, and if a server_name extension is\n sent, then the extension SHOULD contain the same name that was\n negotiated in the application protocol. If the server_name is\n established in the TLS session handshake, the client SHOULD NOT\n attempt to request a different server name at the application layer.\n\n4. Maximum Fragment Length Negotiation\n\n Without this extension, TLS specifies a fixed maximum plaintext\n fragment length of 2^14 bytes. It may be desirable for constrained\n clients to negotiate a smaller maximum fragment length due to memory\n limitations or bandwidth limitations.\n\n In order to negotiate smaller maximum fragment lengths, clients MAY\n include an extension of type \"max_fragment_length\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain:\n\n enum{\n 2^9(1), 2^10(2), 2^11(3), 2^12(4), (255)\n } MaxFragmentLength;\n\n whose value is the desired maximum fragment length. The allowed\n values for this field are: 2^9, 2^10, 2^11, and 2^12.\n\n Servers that receive an extended client hello containing a\n \"max_fragment_length\" extension MAY accept the requested maximum\n fragment length by including an extension of type\n \"max_fragment_length\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL contain a\n \"MaxFragmentLength\" whose value is the same as the requested maximum\n fragment length.\n\n If a server receives a maximum fragment length negotiation request\n for a value other than the allowed values, it MUST abort the\n handshake with an \"illegal_parameter\" alert. Similarly, if a client\n receives a maximum fragment length negotiation response that differs\n from the length it requested, it MUST also abort the handshake with\n an \"illegal_parameter\" alert.\n\n Once a maximum fragment length other than 2^14 has been successfully\n negotiated, the client and server MUST immediately begin fragmenting\n messages (including handshake messages) to ensure that no fragment\n larger than the negotiated length is sent. Note that TLS already\n requires clients and servers to support fragmentation of handshake\n messages.\n\n\n\n\nEastlake Standards Track [Page 8]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n The negotiated length applies for the duration of the session\n including session resumptions.\n\n The negotiated length limits the input that the record layer may\n process without fragmentation (that is, the maximum value of\n TLSPlaintext.length; see [RFC5246], Section 6.2.1). Note that the\n output of the record layer may be larger. For example, if the\n negotiated length is 2^9=512, then, when using currently defined\n cipher suites (those defined in [RFC5246] and [RFC2712]) and null\n compression, the record-layer output can be at most 805 bytes: 5\n bytes of headers, 512 bytes of application data, 256 bytes of\n padding, and 32 bytes of MAC. This means that in this event a TLS\n record-layer peer receiving a TLS record-layer message larger than\n 805 bytes MUST discard the message and send a \"record_overflow\"\n alert, without decrypting the message. When this extension is used\n with Datagram Transport Layer Security (DTLS), implementations SHOULD\n NOT generate record_overflow alerts unless the packet passes message\n authentication.\n\n5. Client Certificate URLs\n\n Without this extension, TLS specifies that when client authentication\n is performed, client certificates are sent by clients to servers\n during the TLS handshake. It may be desirable for constrained\n clients to send certificate URLs in place of certificates, so that\n they do not need to store their certificates and can therefore save\n memory.\n\n In order to negotiate sending certificate URLs to a server, clients\n MAY include an extension of type \"client_certificate_url\" in the\n (extended) client hello. The \"extension_data\" field of this\n extension SHALL be empty.\n\n (Note that it is necessary to negotiate the use of client certificate\n URLs in order to avoid \"breaking\" existing TLS servers.)\n\n Servers that receive an extended client hello containing a\n \"client_certificate_url\" extension MAY indicate that they are willing\n to accept certificate URLs by including an extension of type\n \"client_certificate_url\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n After negotiation of the use of client certificate URLs has been\n successfully completed (by exchanging hellos including\n \"client_certificate_url\" extensions), clients MAY send a\n \"CertificateURL\" message in place of a \"Certificate\" message as\n follows (see also Section 2):\n\n\n\n\nEastlake Standards Track [Page 9]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n enum {\n individual_certs(0), pkipath(1), (255)\n } CertChainType;\n\n struct {\n CertChainType type;\n URLAndHash url_and_hash_list<1..2^16-1>;\n } CertificateURL;\n\n struct {\n opaque url<1..2^16-1>;\n unint8 padding;\n opaque SHA1Hash[20];\n } URLAndHash;\n\n Here, \"url_and_hash_list\" contains a sequence of URLs and hashes.\n Each \"url\" MUST be an absolute URI reference according to [RFC3986]\n that can be immediately used to fetch the certificate(s).\n\n When X.509 certificates are used, there are two possibilities:\n\n - If CertificateURL.type is \"individual_certs\", each URL refers to a\n single DER-encoded X.509v3 certificate, with the URL for the\n client's certificate first.\n\n - If CertificateURL.type is \"pkipath\", the list contains a single\n URL referring to a DER-encoded certificate chain, using the type\n PkiPath described in Section 10.1.\n\n When any other certificate format is used, the specification that\n describes use of that format in TLS should define the encoding format\n of certificates or certificate chains, and any constraint on their\n ordering.\n\n The \"padding\" byte MUST be 0x01. It is present to make the structure\n backwards compatible.\n\n The hash corresponding to each URL is the SHA-1 hash of the\n certificate or certificate chain (in the case of X.509 certificates,\n the DER-encoded certificate or the DER-encoded PkiPath).\n\n Note that when a list of URLs for X.509 certificates is used, the\n ordering of URLs is the same as that used in the TLS Certificate\n message (see [RFC5246], Section 7.4.2), but opposite to the order in\n which certificates are encoded in PkiPath. In either case, the self-\n signed root certificate MAY be omitted from the chain, under the\n assumption that the server must already possess it in order to\n validate it.\n\n\n\nEastlake Standards Track [Page 10]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Servers receiving \"CertificateURL\" SHALL attempt to retrieve the\n client's certificate chain from the URLs and then process the\n certificate chain as usual. A cached copy of the content of any URL\n in the chain MAY be used, provided that the SHA-1 hash matches the\n hash of the cached copy.\n\n Servers that support this extension MUST support the 'http' URI\n scheme for certificate URLs and MAY support other schemes. Use of\n other schemes than 'http', 'https', or 'ftp' may create unexpected\n problems.\n\n If the protocol used is HTTP, then the HTTP server can be configured\n to use the Cache-Control and Expires directives described in\n [RFC2616] to specify whether and for how long certificates or\n certificate chains should be cached.\n\n The TLS server MUST NOT follow HTTP redirects when retrieving the\n certificates or certificate chain. The URLs used in this extension\n MUST NOT be chosen to depend on such redirects.\n\n If the protocol used to retrieve certificates or certificate chains\n returns a MIME-formatted response (as HTTP does), then the following\n MIME Content-Types SHALL be used: when a single X.509v3 certificate\n is returned, the Content-Type is \"application/pkix-cert\" [RFC2585],\n and when a chain of X.509v3 certificates is returned, the Content-\n Type is \"application/pkix-pkipath\" (Section 10.1).\n\n The server MUST check that the SHA-1 hash of the contents of the\n object retrieved from that URL (after decoding any MIME Content-\n Transfer-Encoding) matches the given hash. If any retrieved object\n does not have the correct SHA-1 hash, the server MUST abort the\n handshake with a bad_certificate_hash_value(114) alert. This alert\n is always fatal.\n\n Clients may choose to send either \"Certificate\" or \"CertificateURL\"\n after successfully negotiating the option to send certificate URLs.\n The option to send a certificate is included to provide flexibility\n to clients possessing multiple certificates.\n\n If a server is unable to obtain certificates in a given\n CertificateURL, it MUST send a fatal certificate_unobtainable(111)\n alert if it requires the certificates to complete the handshake. If\n the server does not require the certificates, then the server\n continues the handshake. The server MAY send a warning-level alert\n in this case. Clients receiving such an alert SHOULD log the alert\n and continue with the handshake if possible.\n\n\n\n\n\nEastlake Standards Track [Page 11]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n6. Trusted CA Indication\n\n Constrained clients that, due to memory limitations, possess only a\n small number of CA root keys may wish to indicate to servers which\n root keys they possess, in order to avoid repeated handshake\n failures.\n\n In order to indicate which CA root keys they possess, clients MAY\n include an extension of type \"trusted_ca_keys\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain \"TrustedAuthorities\" where:\n\n struct {\n TrustedAuthority trusted_authorities_list<0..2^16-1>;\n } TrustedAuthorities;\n\n struct {\n IdentifierType identifier_type;\n select (identifier_type) {\n case pre_agreed: struct {};\n case key_sha1_hash: SHA1Hash;\n case x509_name: DistinguishedName;\n case cert_sha1_hash: SHA1Hash;\n } identifier;\n } TrustedAuthority;\n\n enum {\n pre_agreed(0), key_sha1_hash(1), x509_name(2),\n cert_sha1_hash(3), (255)\n } IdentifierType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n Here, \"TrustedAuthorities\" provides a list of CA root key identifiers\n that the client possesses. Each CA root key is identified via\n either:\n\n - \"pre_agreed\": no CA root key identity supplied.\n\n - \"key_sha1_hash\": contains the SHA-1 hash of the CA root key. For\n Digital Signature Algorithm (DSA) and Elliptic Curve Digital\n Signature Algorithm (ECDSA) keys, this is the hash of the\n \"subjectPublicKey\" value. For RSA keys, the hash is of the big-\n endian byte string representation of the modulus without any\n initial zero-valued bytes. (This copies the key hash formats\n deployed in other environments.)\n\n\n\n\n\nEastlake Standards Track [Page 12]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - \"x509_name\": contains the DER-encoded X.509 DistinguishedName of\n the CA.\n\n - \"cert_sha1_hash\": contains the SHA-1 hash of a DER-encoded\n Certificate containing the CA root key.\n\n Note that clients may include none, some, or all of the CA root keys\n they possess in this extension.\n\n Note also that it is possible that a key hash or a Distinguished Name\n alone may not uniquely identify a certificate issuer (for example, if\n a particular CA has multiple key pairs). However, here we assume\n this is the case following the use of Distinguished Names to identify\n certificate issuers in TLS.\n\n The option to include no CA root keys is included to allow the client\n to indicate possession of some pre-defined set of CA root keys.\n\n Servers that receive a client hello containing the \"trusted_ca_keys\"\n extension MAY use the information contained in the extension to guide\n their selection of an appropriate certificate chain to return to the\n client. In this event, the server SHALL include an extension of type\n \"trusted_ca_keys\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n7. Truncated HMAC\n\n Currently defined TLS cipher suites use the MAC construction HMAC\n [RFC2104] to authenticate record-layer communications. In TLS, the\n entire output of the hash function is used as the MAC tag. However,\n it may be desirable in constrained environments to save bandwidth by\n truncating the output of the hash function to 80 bits when forming\n MAC tags.\n\n In order to negotiate the use of 80-bit truncated HMAC, clients MAY\n include an extension of type \"truncated_hmac\" in the extended client\n hello. The \"extension_data\" field of this extension SHALL be empty.\n\n Servers that receive an extended hello containing a \"truncated_hmac\"\n extension MAY agree to use a truncated HMAC by including an extension\n of type \"truncated_hmac\", with empty \"extension_data\", in the\n extended server hello.\n\n Note that if new cipher suites are added that do not use HMAC, and\n the session negotiates one of these cipher suites, this extension\n will have no effect. It is strongly recommended that any new cipher\n suites using other MACs consider the MAC size an integral part of the\n\n\n\n\nEastlake Standards Track [Page 13]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n cipher suite definition, taking into account both security and\n bandwidth considerations.\n\n If HMAC truncation has been successfully negotiated during a TLS\n handshake, and the negotiated cipher suite uses HMAC, both the client\n and the server pass this fact to the TLS record layer along with the\n other negotiated security parameters. Subsequently during the\n session, clients and servers MUST use truncated HMACs, calculated as\n specified in [RFC2104]. That is, SecurityParameters.mac_length is 10\n bytes, and only the first 10 bytes of the HMAC output are transmitted\n and checked. Note that this extension does not affect the\n calculation of the pseudo-random function (PRF) as part of\n handshaking or key derivation.\n\n The negotiated HMAC truncation size applies for the duration of the\n session including session resumptions.\n\n8. Certificate Status Request\n\n Constrained clients may wish to use a certificate-status protocol\n such as OCSP [RFC2560] to check the validity of server certificates,\n in order to avoid transmission of CRLs and therefore save bandwidth\n on constrained networks. This extension allows for such information\n to be sent in the TLS handshake, saving roundtrips and resources.\n\n In order to indicate their desire to receive certificate status\n information, clients MAY include an extension of type\n \"status_request\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"CertificateStatusRequest\" where:\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n } request;\n } CertificateStatusRequest;\n\n enum { ocsp(1), (255) } CertificateStatusType;\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n\n\n\nEastlake Standards Track [Page 14]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n In the OCSPStatusRequest, the \"ResponderIDs\" provides a list of OCSP\n responders that the client trusts. A zero-length \"responder_id_list\"\n sequence has the special meaning that the responders are implicitly\n known to the server, e.g., by prior arrangement. \"Extensions\" is a\n DER encoding of OCSP request extensions.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC2560]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a zero-length ASN.1 SEQUENCE, which is not\n valid for the \"Extensions\" type).\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement).\n\n Servers that receive a client hello containing the \"status_request\"\n extension MAY return a suitable certificate status response to the\n client along with their certificate. If OCSP is requested, they\n SHOULD use the information contained in the extension when selecting\n an OCSP responder and SHOULD include request_extensions in the OCSP\n request.\n\n Servers return a certificate response along with their certificate by\n sending a \"CertificateStatus\" message immediately after the\n \"Certificate\" message (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message, then the server MUST have included an\n extension of type \"status_request\" with empty \"extension_data\" in the\n extended server hello. The \"CertificateStatus\" message is conveyed\n using the handshake message type \"certificate_status\" as follows (see\n also Section 2):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<1..2^24-1>;\n\n An \"ocsp_response\" contains a complete, DER-encoded OCSP response\n (using the ASN.1 type OCSPResponse defined in [RFC2560]). Only one\n OCSP response may be sent.\n\n\n\n\nEastlake Standards Track [Page 15]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if has received a \"status_request\" extension in the\n client hello message and has sent a \"status_request\" extension in the\n server hello message.\n\n Note in addition that a server MUST NOT send the \"CertificateStatus\"\n message unless it received a \"status_request\" extension in the client\n hello message and sent a \"status_request\" extension in the server\n hello message.\n\n Clients requesting an OCSP response and receiving an OCSP response in\n a \"CertificateStatus\" message MUST check the OCSP response and abort\n the handshake if the response is not satisfactory with\n bad_certificate_status_response(113) alert. This alert is always\n fatal.\n\n9. Error Alerts\n\n Four new error alerts are defined for use with the TLS extensions\n defined in this document. To avoid \"breaking\" existing clients and\n servers, these alerts MUST NOT be sent unless the sending party has\n received an extended hello message from the party they are\n communicating with. These error alerts are conveyed using the\n following syntax. The new alerts are the last four, as indicated by\n the comments on the same line as the error alert number.\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n /* 41 is not defined, for historical reasons */\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction(60),\n protocol_version(70),\n insufficient_security(71),\n\n\n\nEastlake Standards Track [Page 16]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n certificate_unobtainable(111), /* new */\n unrecognized_name(112), /* new */\n bad_certificate_status_response(113), /* new */\n bad_certificate_hash_value(114), /* new */\n (255)\n } AlertDescription;\n\n \"certificate_unobtainable\" is described in Section 5.\n \"unrecognized_name\" is described in Section 3.\n \"bad_certificate_status_response\" is described in Section 8.\n \"bad_certificate_hash_value\" is described in Section 5.\n\n10. IANA Considerations\n\n IANA Considerations for TLS extensions and the creation of a registry\n are covered in Section 12 of [RFC5246] except for the registration of\n MIME type application/pkix-pkipath, which appears below.\n\n The IANA TLS extensions and MIME type application/pkix-pkipath\n registry entries that reference RFC 4366 have been updated to\n reference this document.\n\n10.1. pkipath MIME Type Registration\n\n MIME media type name: application\n MIME subtype name: pkix-pkipath\n Required parameters: none\n\n Optional parameters: version (default value is \"1\")\n\n Encoding considerations:\n Binary; this MIME type is a DER encoding of the ASN.1 type\n PkiPath, defined as follows:\n PkiPath ::= SEQUENCE OF Certificate\n PkiPath is used to represent a certification path. Within the\n sequence, the order of certificates is such that the subject of\n the first certificate is the issuer of the second certificate,\n etc.\n This is identical to the definition published in [X509-4th-TC1];\n note that it is different from that in [X509-4th].\n\n All Certificates MUST conform to [RFC5280]. (This should be\n interpreted as a requirement to encode only PKIX-conformant\n certificates using this type. It does not necessarily require\n\n\n\nEastlake Standards Track [Page 17]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n that all certificates that are not strictly PKIX-conformant must\n be rejected by relying parties, although the security consequences\n of accepting any such certificates should be considered\n carefully.)\n\n DER (as opposed to BER) encoding MUST be used. If this type is\n sent over a 7-bit transport, base64 encoding SHOULD be used.\n\n Security considerations:\n The security considerations of [X509-4th] and [RFC5280] (or any\n updates to them) apply, as well as those of any protocol that uses\n this type (e.g., TLS).\n\n Note that this type only specifies a certificate chain that can be\n assessed for validity according to the relying party's existing\n configuration of trusted CAs; it is not intended to be used to\n specify any change to that configuration.\n\n Interoperability considerations:\n No specific interoperability problems are known with this type,\n but for recommendations relating to X.509 certificates in general,\n see [RFC5280].\n\n Published specification: This document and [RFC5280].\n\n Applications that use this media type:\n TLS. It may also be used by other protocols or for general\n interchange of PKIX certificate chains.\n\n Additional information:\n Magic number(s): DER-encoded ASN.1 can be easily recognized.\n Further parsing is required to distinguish it from other ASN.1\n types.\n File extension(s): .pkipath\n Macintosh File Type Code(s): not specified\n\n Person & email address to contact for further information:\n Magnus Nystrom <[redacted-email]>\n\n Intended usage: COMMON\n\n Change controller: IESG <[redacted-email]>\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 18]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n10.2. Reference for TLS Alerts, TLS HandshakeTypes, and ExtensionTypes\n\n The following values in the TLS Alert Registry have been updated to\n reference this document:\n\n 111 certificate_unobtainable\n 112 unrecognized_name\n 113 bad_certificate_status_response\n 114 bad_certificate_hash_value\n\n The following values in the TLS HandshakeType Registry have been\n updated to reference this document:\n\n 21 certificate_url\n 22 certificate_status\n\n The following ExtensionType values have been updated to reference\n this document:\n\n 0 server_name\n 1 max_fragment_length\n 2 client_certificate_url\n 3 trusted_ca_keys\n 4 truncated_hmac\n 5 status_request\n\n11. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security Considerations for particular extensions\n specified in this document are given below.\n\n In general, implementers should continue to monitor the state of the\n art and address any weaknesses identified.\n\n11.1. Security Considerations for server_name\n\n If a single server hosts several domains, then clearly it is\n necessary for the owners of each domain to ensure that this satisfies\n their security needs. Apart from this, server_name does not appear\n to introduce significant security issues.\n\n Since it is possible for a client to present a different server_name\n in the application protocol, application server implementations that\n rely upon these names being the same MUST check to make sure the\n client did not present a different name in the application protocol.\n\n\n\n\n\nEastlake Standards Track [Page 19]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Implementations MUST ensure that a buffer overflow does not occur,\n whatever the values of the length fields in server_name.\n\n11.2. Security Considerations for max_fragment_length\n\n The maximum fragment length takes effect immediately, including for\n handshake messages. However, that does not introduce any security\n complications that are not already present in TLS, since TLS requires\n implementations to be able to handle fragmented handshake messages.\n\n Note that, as described in Section 4, once a non-null cipher suite\n has been activated, the effective maximum fragment length depends on\n the cipher suite and compression method, as well as on the negotiated\n max_fragment_length. This must be taken into account when sizing\n buffers and checking for buffer overflow.\n\n11.3. Security Considerations for client_certificate_url\n\n Support for client_certificate_url involves the server's acting as a\n client in another URI-scheme-dependent protocol. The server\n therefore becomes subject to many of the same security concerns that\n clients of the URI scheme are subject to, with the added concern that\n the client can attempt to prompt the server to connect to some\n (possibly weird-looking) URL.\n\n In general, this issue means that an attacker might use the server to\n indirectly attack another host that is vulnerable to some security\n flaw. It also introduces the possibility of denial-of-service\n attacks in which an attacker makes many connections to the server,\n each of which results in the server's attempting a connection to the\n target of the attack.\n\n Note that the server may be behind a firewall or otherwise able to\n access hosts that would not be directly accessible from the public\n Internet. This could exacerbate the potential security and denial-\n of-service problems described above, as well as allow the existence\n of internal hosts to be confirmed when they would otherwise be\n hidden.\n\n The detailed security concerns involved will depend on the URI\n schemes supported by the server. In the case of HTTP, the concerns\n are similar to those that apply to a publicly accessible HTTP proxy\n server. In the case of HTTPS, loops and deadlocks may be created,\n and this should be addressed. In the case of FTP, attacks arise that\n are similar to FTP bounce attacks.\n\n\n\n\n\n\nEastlake Standards Track [Page 20]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n As a result of this issue, it is RECOMMENDED that the\n client_certificate_url extension should have to be specifically\n enabled by a server administrator, rather than be enabled by default.\n It is also RECOMMENDED that URI schemes be enabled by the\n administrator individually, and only a minimal set of schemes be\n enabled. Unusual protocols that offer limited security or whose\n security is not well understood SHOULD be avoided.\n\n As discussed in [RFC3986], URLs that specify ports other than the\n default may cause problems, as may very long URLs (which are more\n likely to be useful in exploiting buffer overflow bugs).\n\n This extension continues to use SHA-1 (as in RFC 4366) and does not\n provide algorithm agility. The property required of SHA-1 in this\n case is second pre-image resistance, not collision resistance.\n Furthermore, even if second pre-image attacks against SHA-1 are found\n in the future, an attack against client_certificate_url would require\n a second pre-image that is accepted as a valid certificate by the\n server and contains the same public key.\n\n Also note that HTTP caching proxies are common on the Internet, and\n some proxies do not check for the latest version of an object\n correctly. If a request using HTTP (or another caching protocol)\n goes through a misconfigured or otherwise broken proxy, the proxy may\n return an out-of-date response.\n\n11.4. Security Considerations for trusted_ca_keys\n\n Potentially, the CA root keys a client possesses could be regarded as\n confidential information. As a result, the CA root key indication\n extension should be used with care.\n\n The use of the SHA-1 certificate hash alternative ensures that each\n certificate is specified unambiguously. This context does not\n require a cryptographic hash function, so the use of SHA-1 is\n considered acceptable, and no algorithm agility is provided.\n\n11.5. Security Considerations for truncated_hmac\n\n It is possible that truncated MACs are weaker than \"un-truncated\"\n MACs. However, no significant weaknesses are currently known or\n expected to exist for HMAC with MD5 or SHA-1, truncated to 80 bits.\n\n Note that the output length of a MAC need not be as long as the\n length of a symmetric cipher key, since forging of MAC values cannot\n be done off-line: in TLS, a single failed MAC guess will cause the\n immediate termination of the TLS session.\n\n\n\n\nEastlake Standards Track [Page 21]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Since the MAC algorithm only takes effect after all handshake\n messages that affect extension parameters have been authenticated by\n the hashes in the Finished messages, it is not possible for an active\n attacker to force negotiation of the truncated HMAC extension where\n it would not otherwise be used (to the extent that the handshake\n authentication is secure). Therefore, in the event that any security\n problems were found with truncated HMAC in the future, if either the\n client or the server for a given session were updated to take the\n problem into account, it would be able to veto use of this extension.\n\n11.6. Security Considerations for status_request\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC2560] for further details.\n\n12. Normative References\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC:\n Keyed-Hashing for Message Authentication\", RFC 2104,\n February 1997.\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and\n C. Adams, \"X.509 Internet Public Key Infrastructure\n Online Certificate Status Protocol - OCSP\", RFC 2560,\n June 1999.\n\n [RFC2585] Housley, R. and P. Hoffman, \"Internet X.509 Public Key\n Infrastructure Operational Protocols: FTP and HTTP\",\n RFC 2585, May 1999.\n\n [RFC2616] Fielding, R., Gettys, J., Mogul, J., Frystyk, H.,\n Masinter, L., Leach, P., and T. Berners-Lee,\n \"Hypertext Transfer Protocol -- HTTP/1.1\", RFC 2616,\n June 1999.\n\n [RFC3986] Berners-Lee, T., Fielding, R., and L. Masinter,\n \"Uniform Resource Identifier (URI): Generic Syntax\",\n STD 66, RFC 3986, January 2005.\n\n\n\nEastlake Standards Track [Page 22]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246, August\n 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation\n List (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC5890] Klensin, J., \"Internationalized Domain Names for\n Applications (IDNA): Definitions and Document\n Framework\", RFC 5890, August 2010.\n\n13. Informative References\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [X509-4th] ITU-T Recommendation X.509 (2000) | ISO/IEC\n 9594-8:2001, \"Information Systems - Open Systems\n Interconnection - The Directory: Public key and\n attribute certificate frameworks\".\n\n [X509-4th-TC1] ITU-T Recommendation X.509(2000) Corrigendum 1(2001) |\n ISO/IEC 9594-8:2001/Cor.1:2002, Technical Corrigendum\n 1 to ISO/IEC 9594:8:2001.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 23]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix A. Changes from RFC 4366\n\n The significant changes between RFC 4366 and this document are\n described below.\n\n RFC 4366 described both general extension mechanisms (for the TLS\n handshake and client and server hellos) as well as specific\n extensions. RFC 4366 was associated with RFC 4346, TLS 1.1. The\n client and server hello extension mechanisms have been moved into RFC\n 5246, TLS 1.2, so this document, which is associated with RFC 5246,\n includes only the handshake extension mechanisms and the specific\n extensions from RFC 4366. RFC 5246 also specifies the unknown\n extension error and new extension specification considerations, so\n that material has been removed from this document.\n\n The Server Name extension now specifies only ASCII representation,\n eliminating UTF-8. It is provided that the ServerNameList can\n contain more than only one name of any particular name_type. If a\n server name is provided but not recognized, the server should either\n continue the handshake without an error or send a fatal error.\n Sending a warning-level message is not recommended because client\n behavior will be unpredictable. Provision was added for the user\n using the server_name extension in deciding whether or not to resume\n a session. Furthermore, this extension should be the same in a\n session resumption request as it was in the full handshake that\n established the session. Such a resumption request must not be\n accepted if the server_name extension is different, but instead a\n full handshake must be done to possibly establish a new session.\n\n The Client Certificate URLs extension has been changed to make the\n presence of a hash mandatory.\n\n For the case of DTLS, the requirement to report an overflow of the\n negotiated maximum fragment length is made conditional on passing\n authentication.\n\n TLS servers are now prohibited from following HTTP redirects when\n retrieving certificates.\n\n The material was also re-organized in minor ways. For example,\n information as to which errors are fatal is moved from the \"Error\n Alerts\" section to the individual extension specifications.\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 24]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix B. Acknowledgements\n\n This document is based on material from RFC 4366 for which the\n authors were S. Blake-Wilson, M. Nystrom, D. Hopwood, J. Mikkelsen,\n and T. Wright. Other contributors include Joseph Salowey, Alexey\n Melnikov, Peter Saint-Andre, and Adrian Farrel.\n\nAuthor's Address\n\n Donald Eastlake 3rd\n Huawei\n 155 Beaver Street\n Milford, MA 01757 USA\n\n Phone: +1-508-333-2270\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 25]", "domain": "standards", "dossier_id": "a00388a585cd2f4d89b3", "essential_artifact_ids": ["ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"], "event_count": 5, "evidence_distance": 65121, "evidence_span_tokens": 65114, "executable_proof_id": "0c64de0f18afd43b8e63", "generation_integration": "task_replay_promotion_v1", "graph": {"hop_count": 2, "n_essential_artifacts": 7, "n_essential_events": 6, "proof_depth": 2}, "holdout": {"group_id": "af2707ad6cb50ab3", "strategy": "world"}, "hop_count": 2, "hybrid_causal_edges": [], "ietf_requirement_task": {"answer": {"current_protocol": "TLS_1_3_RFC8446", "obsoletes_ocsp_stapling": "RFC6961", "obsoletes_session_tickets": "RFC5077", "obsoletes_tls12": "RFC5246", "updates_exporters": "RFC5705", "updates_extensions": "RFC6066"}, "answer_program_id": "ietf.tls13_handshake_succession.v1", "cutoff": "2026-01-31T00:00:00Z", "essential_evidence_ids": ["current_protocol", "obsoletes_session_tickets", "obsoletes_tls12", "obsoletes_ocsp_stapling", "updates_exporters", "updates_extensions"], "essential_relation_ids": ["ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "ietf:updates:ietf:rfc:8446:ietf:rfc:6066"], "evidence_items": [{"char_end": 447, "char_start": 361, "evidence_id": "current_protocol", "evidence_quote": "This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol.", "quote_sha256": "213f297f8fd7940d76115d7115fd430d489bced81930ca9d9319c2eaddff65ea", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15703, "char_start": 15575, "evidence_id": "obsoletes_session_tickets", "evidence_quote": "It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2.", "quote_sha256": "957220687c3a075f9479d4799d2a2a6bc3065f4b5f968376a62cd724206781ff", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15573, "char_start": 15473, "evidence_id": "obsoletes_tls12", "evidence_quote": "This document supersedes and obsoletes previous versions of TLS,\n including version 1.2 [RFC5246].", "quote_sha256": "0ffd331074a612c946a7f65eb52848b1224f4497f7dad642bd1c34ac255fe340", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15989, "char_start": 15934, "evidence_id": "obsoletes_ocsp_stapling", "evidence_quote": "obsoletes [RFC6961] as\n described in Section 4.4.2.1.", "quote_sha256": "836a51c84442de8cd356808fe29a5bc51913ffc3248fa087461cea3fbe9f26cd", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15807, "char_start": 15705, "evidence_id": "updates_exporters", "evidence_quote": "Because TLS 1.3 changes the way keys are\n derived, it updates [RFC5705] as described in Section 7.5.", "quote_sha256": "e5c566bced7efdd9df1c7483ab71fa5ce896e674e1a35613667bd90d52341b0d", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15989, "char_start": 15902, "evidence_id": "updates_extensions", "evidence_quote": "therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in Section 4.4.2.1.", "quote_sha256": "46d4d3c9d3c46874270771eceb979e3b19a7f09528933d4e9071941d096f5b9f", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}], "query_type": "protocol_succession_resolution", "question": "Resolve the effective TLS 1.3 handshake protocol succession at 2026-01-31T00:00:00Z for this scenario (canonical JSON): {\"protocol\":\"tls\",\"publication\":\"rfc8446\",\"succession\":\"obsoletes_and_updates\",\"version\":\"1.3\"}. Return exactly one JSON object with these keys in this order: [\"current_protocol\",\"obsoletes_session_tickets\",\"obsoletes_tls12\",\"obsoletes_ocsp_stapling\",\"updates_exporters\",\"updates_extensions\"]. Use this exact per-field output codebook (canonical JSON): {\"current_protocol\":{\"code\":\"TLS_1_3_RFC8446\",\"meaning\":\"RFC 8446 is TLS protocol version 1.3\"},\"obsoletes_ocsp_stapling\":{\"code\":\"RFC6961\",\"meaning\":\"RFC 8446 obsoletes multiple certificate status requests\"},\"obsoletes_session_tickets\":{\"code\":\"RFC5077\",\"meaning\":\"RFC 8446 obsoletes TLS session resumption tickets\"},\"obsoletes_tls12\":{\"code\":\"RFC5246\",\"meaning\":\"RFC 8446 obsoletes TLS protocol version 1.2\"},\"updates_exporters\":{\"code\":\"RFC5705\",\"meaning\":\"RFC 8446 updates TLS keying material exporters\"},\"updates_extensions\":{\"code\":\"RFC6066\",\"meaning\":\"RFC 8446 updates TLS extension definitions\"}}. Resolve each field independently from the supplied RFC graph; use the string \"UNKNOWN\" only for a field whose required evidence or relation is absent.", "scenario": {"protocol": "tls", "publication": "rfc8446", "succession": "obsoletes_and_updates", "version": "1.3"}, "schema_version": "longworld.ietf-tls13-handshake-succession-task.v1", "source_manifest": {"authorization": {"allowed_actions": ["fetch_datatracker_document", "fetch_datatracker_relation", "fetch_draft_revision", "fetch_rfc"], "basis": "IETF open-records research export for a closed published_as plus Updates/Obsoletes graph. Distinct from HTTP/3/QUIC packing, OAuth, and RFC 9421 revision stacking. RFC 5077 and RFC 6961 are included because RFC 8446 authentically obsoletes them.", "record_id": "p57-ietf-tls13-handshake-graph-v1", "reviewed_at": "2026-09-07T04:00:00Z", "scope": "public IETF Datatracker, Internet-Draft archive, and RFC Editor records for the TLS 1.3 handshake requirement graph over RFC 8446 and its authentic Updates/Obsoletes targets"}, "data_stage": "source_inventory", "fetch_inventory_sha256": "c86bc411fe4e31382a729a63b451fff99251ebacc2b1323213b4fa8cba891c36", "fetch_receipt": {"allowed_actions": ["fetch_datatracker_document", "fetch_datatracker_relation", "fetch_draft_revision", "fetch_rfc"], "completed_at": "2026-09-07T04:14:41.225467Z", "max_retries": 2, "policy_url": "https://www.ietf.org/about/open-records/", "request_file": "ietf_fetch_request.json", "request_sha256": "577344d9c626e814c0e57a20755290a4e9bdaa66c1e7b545370366600ef34c5f", "requests_per_second": 1.0, "retrievals": [{"content_type": "application/json", "final_url": "https://datatracker.ietf.org/api/v1/doc/document/draft-ietf-tls-tls13/", "kind": "datatracker_document", "observed_at": "2026-09-07T04:14:30.932265Z", "redirect_chain": [], "requested_url": "https://datatracker.ietf.org/api/v1/doc/document/draft-ietf-tls-tls13/", "retrieval_file": "datatracker-draft-ietf-tls-tls13.json", "sha256": "a632cb66584c8f5d364d909b76eed121c1a9d00e341908463e019efad0faee0f", "status": 200}, {"content_type": "application/json", "final_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=draft-ietf-tls-tls13&limit=100", "kind": "datatracker_relation", "observed_at": "2026-09-07T04:14:32.295913Z", "redirect_chain": [], "requested_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=draft-ietf-tls-tls13&limit=100", "retrieval_file": "datatracker-draft-ietf-tls-tls13-relations.json", "sha256": "3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "kind": "draft_revision", "observed_at": "2026-09-07T04:14:32.641074Z", "redirect_chain": [], "requested_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "retrieval_file": "draft-ietf-tls-tls13-28.txt", "sha256": "9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:33.763234Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "retrieval_file": "rfc5077.txt", "sha256": "e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:34.735753Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "retrieval_file": "rfc5246.txt", "sha256": "2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:35.724001Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "retrieval_file": "rfc5705.txt", "sha256": "29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:36.718981Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "retrieval_file": "rfc6066.txt", "sha256": "6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:37.734605Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "retrieval_file": "rfc6961.txt", "sha256": "c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:38.737647Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "retrieval_file": "rfc8446.txt", "sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd", "status": 200}, {"content_type": "application/json", "final_url": "https://datatracker.ietf.org/api/v1/doc/document/rfc8446/", "kind": "datatracker_document", "observed_at": "2026-09-07T04:14:39.917810Z", "redirect_chain": [], "requested_url": "https://datatracker.ietf.org/api/v1/doc/document/rfc8446/", "retrieval_file": "datatracker-rfc8446.json", "sha256": "89b01010d1ff76dc260539d5a453dde0b116d857db43c9e09d4a3dc062c9ce31", "status": 200}, {"content_type": "application/json", "final_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=rfc8446&limit=100", "kind": "datatracker_relation", "observed_at": "2026-09-07T04:14:41.148823Z", "redirect_chain": [], "requested_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=rfc8446&limit=100", "retrieval_file": "datatracker-rfc8446-relations.json", "sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7", "status": 200}], "started_at": "2026-09-07T04:14:30.455103Z", "user_agent_sha256": "1c8887e16ad14ac078ad6e7a52c6cdf9f988eb709fa5671439746106b1265652"}, "generated_at": "2026-09-07T04:16:16.549092Z", "generation_integration": "disabled", "hybrid_train_ready": false, "n": 7, "n_relations": 11, "n_supporting_records": 4, "production_eligible": false, "records": [{"draft_references": [], "excluded_rfc_relation_targets": [{"kind": "obsoletes", "reason": "not_requested", "rfc_number": 4507}], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5077 H. Zhou", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "5077"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 4507 Cisco Systems", "fact_id": "obsoletes:4507", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "4507"}], "kind": "rfc", "occurred_at": "2008-01-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 4, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef", "record_id": "ietf:rfc:5077", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "rfc_number": 5077, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef", "source_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "temporal_semantics": "source_event_time", "text": "Network Working Group J. Salowey\nRequest for Comments: 5077 H. Zhou\nObsoletes: 4507 Cisco Systems\nCategory: Standards Track P. Eronen\n Nokia\n H. Tschofenig\n Nokia Siemens Networks\n January 2008\n\n\n Transport Layer Security (TLS) Session Resumption without\n Server-Side State\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document describes a mechanism that enables the Transport Layer\n Security (TLS) server to resume sessions and avoid keeping per-client\n session state. The TLS server encapsulates the session state into a\n ticket and forwards it to the client. The client can subsequently\n resume a session using the obtained ticket. This document obsoletes\n RFC 4507.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 1]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3. Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . 4\n 3.2. SessionTicket TLS Extension . . . . . . . . . . . . . . . 7\n 3.3. NewSessionTicket Handshake Message . . . . . . . . . . . . 8\n 3.4. Interaction with TLS Session ID . . . . . . . . . . . . . 9\n 4. Recommended Ticket Construction . . . . . . . . . . . . . . . 10\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . 12\n 5.1. Invalidating Sessions . . . . . . . . . . . . . . . . . . 12\n 5.2. Stolen Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.3. Forged Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.4. Denial of Service Attacks . . . . . . . . . . . . . . . . 12\n 5.5. Ticket Protection Key Management . . . . . . . . . . . . . 13\n 5.6. Ticket Lifetime . . . . . . . . . . . . . . . . . . . . . 13\n 5.7. Alternate Ticket Formats and Distribution Schemes . . . . 13\n 5.8. Identity Privacy, Anonymity, and Unlinkability . . . . . . 14\n 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 14\n 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 15\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 15\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 15\n Appendix A. Discussion of Changes to RFC 4507 . . . . . . . . . . 17\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 2]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n1. Introduction\n\n This document defines a way to resume a Transport Layer Security\n (TLS) session without requiring session-specific state at the TLS\n server. This mechanism may be used with any TLS ciphersuite. This\n document applies to both TLS 1.0 defined in [RFC2246], and TLS 1.1\n defined in [RFC4346]. The mechanism makes use of TLS extensions\n defined in [RFC4366] and defines a new TLS message type.\n\n This mechanism is useful in the following situations:\n\n 1. servers that handle a large number of transactions from different\n users\n\n 2. servers that desire to cache sessions for a long time\n\n 3. ability to load balance requests across servers\n\n 4. embedded servers with little memory\n\n This document obsoletes RFC 4507 [RFC4507] to correct an error in the\n encoding that caused the specification to differ from deployed\n implementations. At the time of this writing, there are no known\n implementations that follow the encoding specified in RFC 4507. This\n update to RFC 4507 aligns the document with currently deployed\n implementations. More details of the change are given in Appendix A.\n\n2. Terminology\n\n Within this document, the term 'ticket' refers to a cryptographically\n protected data structure that is created and consumed by the server\n to rebuild session-specific state.\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Protocol\n\n This specification describes a mechanism to distribute encrypted\n session-state information to the client in the form of a ticket and a\n mechanism to present the ticket back to the server. The ticket is\n created by a TLS server and sent to a TLS client. The TLS client\n presents the ticket to the TLS server to resume a session.\n Implementations of this specification are expected to support both\n mechanisms. Other specifications can take advantage of the session\n tickets, perhaps specifying alternative means for distribution or\n selection. For example, a separate specification may describe an\n\n\n\nSalowey, et al. Standards Track [Page 3]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n alternate way to distribute a ticket and use the TLS extension in\n this document to resume the session. This behavior is beyond the\n scope of the document and would need to be described in a separate\n specification.\n\n3.1. Overview\n\n The client indicates that it supports this mechanism by including a\n SessionTicket TLS extension in the ClientHello message. The\n extension will be empty if the client does not already possess a\n ticket for the server. The server sends an empty SessionTicket\n extension to indicate that it will send a new session ticket using\n the NewSessionTicket handshake message. The extension is described\n in Section 3.2.\n\n If the server wants to use this mechanism, it stores its session\n state (such as ciphersuite and master secret) to a ticket that is\n encrypted and integrity-protected by a key known only to the server.\n The ticket is distributed to the client using the NewSessionTicket\n TLS handshake message described in Section 3.3. This message is sent\n during the TLS handshake before the ChangeCipherSpec message, after\n the server has successfully verified the client's Finished message.\n\n Client Server\n\n ClientHello\n (empty SessionTicket extension)-------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1: Message Flow for Full Handshake Issuing New Session Ticket\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 4]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n The client caches this ticket along with the master secret and other\n parameters associated with the current session. When the client\n wishes to resume the session, it includes the ticket in the\n SessionTicket extension within the ClientHello message. Appendix A\n provides a detailed description of the encoding of the extension and\n changes from RFC 4507. The server then decrypts the received ticket,\n verifies the ticket's validity, retrieves the session state from the\n contents of the ticket, and uses this state to resume the session.\n The interaction with the TLS Session ID is described in Section 3.4.\n If the server successfully verifies the client's ticket, then it may\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello.\n\n Client Server\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2: Message Flow for Abbreviated Handshake Using New Session\n Ticket\n\n A recommended ticket format is given in Section 4.\n\n If the server cannot or does not want to honor the ticket, then it\n can initiate a full handshake with the client.\n\n In the case that the server does not wish to issue a new ticket at\n this time, it just completes the handshake without including a\n SessionTicket extension or NewSessionTicket handshake message. This\n is shown below (this flow is identical to Figure 1 in RFC 4346,\n except for the SessionTicket extension in the first message):\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 5]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 3: Message Flow for Server Completing Full Handshake Without\n Issuing New Session Ticket\n\n It is also permissible to have an exchange similar to Figure 3 using\n the abbreviated handshake defined in Figure 2 of RFC 4346, where the\n client uses the SessionTicket extension to resume the session, but\n the server does not wish to issue a new ticket, and therefore does\n not send a SessionTicket extension.\n\n If the server rejects the ticket, it may still wish to issue a new\n ticket after performing the full handshake as shown below (this flow\n is identical to Figure 1, except the SessionTicket extension in the\n ClientHello is not empty):\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 6]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 4: Message Flow for Server Rejecting Ticket, Performing Full\n Handshake, and Issuing New Session Ticket\n\n3.2. SessionTicket TLS Extension\n\n The SessionTicket TLS extension is based on [RFC4366]. The format of\n the ticket is an opaque structure used to carry session-specific\n state information. This extension may be sent in the ClientHello and\n ServerHello.\n\n If the client possesses a ticket that it wants to use to resume a\n session, then it includes the ticket in the SessionTicket extension\n in the ClientHello. If the client does not have a ticket and is\n prepared to receive one in the NewSessionTicket handshake message,\n then it MUST include a zero-length ticket in the SessionTicket\n extension. If the client is not prepared to receive a ticket in the\n NewSessionTicket handshake message, then it MUST NOT include a\n SessionTicket extension unless it is sending a non-empty ticket it\n received through some other means from the server.\n\n The server uses a zero-length SessionTicket extension to indicate to\n the client that it will send a new session ticket using the\n NewSessionTicket handshake message described in Section 3.3. The\n server MUST send this extension in the ServerHello if it wishes to\n issue a new ticket to the client using the NewSessionTicket handshake\n message. The server MUST NOT send this extension if it does not\n receive one in the ClientHello.\n\n\n\n\nSalowey, et al. Standards Track [Page 7]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n If the server fails to verify the ticket, then it falls back to\n performing a full handshake. If the ticket is accepted by the server\n but the handshake fails, the client SHOULD delete the ticket.\n\n The SessionTicket extension has been assigned the number 35. The\n extension_data field of SessionTicket extension contains the ticket.\n\n3.3. NewSessionTicket Handshake Message\n\n This message is sent by the server during the TLS handshake before\n the ChangeCipherSpec message. This message MUST be sent if the\n server included a SessionTicket extension in the ServerHello. This\n message MUST NOT be sent if the server did not include a\n SessionTicket extension in the ServerHello. This message is included\n in the hash used to create and verify the Finished message. In the\n case of a full handshake, the server MUST verify the client's\n Finished message before sending the ticket. The client MUST NOT\n treat the ticket as valid until it has verified the server's Finished\n message. If the server determines that it does not want to include a\n ticket after it has included the SessionTicket extension in the\n ServerHello, then it sends a zero-length ticket in the\n NewSessionTicket handshake message.\n\n If the server successfully verifies the client's ticket, then it MAY\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello in the abbreviated handshake. The client\n should start using the new ticket as soon as possible after it\n verifies the server's Finished message for new connections. Note\n that since the updated ticket is issued before the handshake\n completes, it is possible that the client may not put the new ticket\n into use before it initiates new connections. The server MUST NOT\n assume that the client actually received the updated ticket until it\n successfully verifies the client's Finished message.\n\n The NewSessionTicket handshake message has been assigned the number 4\n and its definition is given at the end of this section. The\n ticket_lifetime_hint field contains a hint from the server about how\n long the ticket should be stored. The value indicates the lifetime\n in seconds as a 32-bit unsigned integer in network byte order\n relative to when the ticket is received. A value of zero is reserved\n to indicate that the lifetime of the ticket is unspecified. A client\n SHOULD delete the ticket and associated state when the time expires.\n It MAY delete the ticket earlier based on local policy. A server MAY\n treat a ticket as valid for a shorter or longer period of time than\n what is stated in the ticket_lifetime_hint.\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 8]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n HandshakeType msg_type;\n uint24 length;\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case session_ticket: NewSessionTicket; /* NEW */\n } body;\n } Handshake;\n\n\n struct {\n uint32 ticket_lifetime_hint;\n opaque ticket<0..2^16-1>;\n } NewSessionTicket;\n\n3.4. Interaction with TLS Session ID\n\n If a server is planning on issuing a session ticket to a client that\n does not present one, it SHOULD include an empty Session ID in the\n ServerHello. If the server rejects the ticket and falls back to the\n full handshake then it may include a non-empty Session ID to indicate\n its support for stateful session resumption. If the client receives\n a session ticket from the server, then it discards any Session ID\n that was sent in the ServerHello.\n\n When presenting a ticket, the client MAY generate and include a\n Session ID in the TLS ClientHello. If the server accepts the ticket\n and the Session ID is not empty, then it MUST respond with the same\n Session ID present in the ClientHello. This allows the client to\n easily differentiate when the server is resuming a session from when\n it is falling back to a full handshake. Since the client generates a\n Session ID, the server MUST NOT rely upon the Session ID having a\n particular value when validating the ticket. If a ticket is\n presented by the client, the server MUST NOT attempt to use the\n Session ID in the ClientHello for stateful session resumption.\n Alternatively, the client MAY include an empty Session ID in the\n ClientHello. In this case, the client ignores the Session ID sent in\n the ServerHello and determines if the server is resuming a session by\n the subsequent handshake messages.\n\n\n\nSalowey, et al. Standards Track [Page 9]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n4. Recommended Ticket Construction\n\n This section describes a recommended format and protection for the\n ticket. Note that the ticket is opaque to the client, so the\n structure is not subject to interoperability concerns, and\n implementations may diverge from this format. If implementations do\n diverge from this format, they must take security concerns seriously.\n Clients MUST NOT examine the ticket under the assumption that it\n complies with this document.\n\n The server uses two different keys: one 128-bit key for Advanced\n Encryption Standard (AES) [AES] in Cipher Block Chaining (CBC) mode\n [CBC] encryption and one 256-bit key for HMAC-SHA-256 [RFC4634].\n\n The ticket is structured as follows:\n\n struct {\n opaque key_name[16];\n opaque iv[16];\n opaque encrypted_state<0..2^16-1>;\n opaque mac[32];\n } ticket;\n\n Here, key_name serves to identify a particular set of keys used to\n protect the ticket. It enables the server to easily recognize\n tickets it has issued. The key_name should be randomly generated to\n avoid collisions between servers. One possibility is to generate new\n random keys and key_name every time the server is started.\n\n The actual state information in encrypted_state is encrypted using\n 128-bit AES in CBC mode with the given IV. The Message\n Authentication Code (MAC) is calculated using HMAC-SHA-256 over\n key_name (16 octets) and IV (16 octets), followed by the length of\n the encrypted_state field (2 octets) and its contents (variable\n length).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 10]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n ProtocolVersion protocol_version;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n opaque master_secret[48];\n ClientIdentity client_identity;\n uint32 timestamp;\n } StatePlaintext;\n\n enum {\n anonymous(0),\n certificate_based(1),\n psk(2)\n } ClientAuthenticationType;\n\n struct {\n ClientAuthenticationType client_authentication_type;\n select (ClientAuthenticationType) {\n case anonymous: struct {};\n case certificate_based:\n ASN.1Cert certificate_list<0..2^24-1>;\n case psk:\n opaque psk_identity<0..2^16-1>; /* from [RFC4279] */\n };\n } ClientIdentity;\n\n The structure StatePlaintext stores the TLS session state including\n the master_secret. The timestamp within this structure allows the\n TLS server to expire tickets. To cover the authentication and key\n exchange protocols provided by TLS, the ClientIdentity structure\n contains the authentication type of the client used in the initial\n exchange (see ClientAuthenticationType). To offer the TLS server\n with the same capabilities for authentication and authorization, a\n certificate list is included in case of public-key-based\n authentication. The TLS server is therefore able to inspect a number\n of different attributes within these certificates. A specific\n implementation might choose to store a subset of this information or\n additional information. Other authentication mechanisms, such as\n Kerberos [RFC2712], would require different client identity data.\n Other TLS extensions may require the inclusion of additional data in\n the StatePlaintext structure.\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 11]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5. Security Considerations\n\n This section addresses security issues related to the usage of a\n ticket. Tickets must be authenticated and encrypted to prevent\n modification or eavesdropping by an attacker. Several attacks\n described below will be possible if this is not carefully done.\n\n Implementations should take care to ensure that the processing of\n tickets does not increase the chance of denial of service as\n described below.\n\n5.1. Invalidating Sessions\n\n The TLS specification requires that TLS sessions be invalidated when\n errors occur. [CSSC] discusses the security implications of this in\n detail. In the analysis within this paper, failure to invalidate\n sessions does not pose a security risk. This is because the TLS\n handshake uses a non-reversible function to derive keys for a session\n so information about one session does not provide an advantage to\n attack the master secret or a different session. If a session\n invalidation scheme is used, the implementation should verify the\n integrity of the ticket before using the contents to invalidate a\n session to ensure that an attacker cannot invalidate a chosen\n session.\n\n5.2. Stolen Tickets\n\n An eavesdropper or man-in-the-middle may obtain the ticket and\n attempt to use it to establish a session with the server; however,\n since the ticket is encrypted and the attacker does not know the\n secret key, a stolen ticket does not help an attacker resume a\n session. A TLS server MUST use strong encryption and integrity\n protection for the ticket to prevent an attacker from using a brute\n force mechanism to obtain the ticket's contents.\n\n5.3. Forged Tickets\n\n A malicious user could forge or alter a ticket in order to resume a\n session, to extend its lifetime, to impersonate another user, or to\n gain additional privileges. This attack is not possible if the\n ticket is protected using a strong integrity protection algorithm\n such as a keyed HMAC-SHA-256.\n\n5.4. Denial of Service Attacks\n\n The key_name field defined in the recommended ticket format helps the\n server efficiently reject tickets that it did not issue. However, an\n adversary could store or generate a large number of tickets to send\n\n\n\nSalowey, et al. Standards Track [Page 12]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n to the TLS server for verification. To minimize the possibility of a\n denial of service, the verification of the ticket should be\n lightweight (e.g., using efficient symmetric key cryptographic\n algorithms).\n\n5.5. Ticket Protection Key Management\n\n A full description of the management of the keys used to protect the\n ticket is beyond the scope of this document. A list of RECOMMENDED\n practices is given below.\n\n o The keys should be generated securely following the randomness\n recommendations in [RFC4086].\n\n o The keys and cryptographic protection algorithms should be at\n least 128 bits in strength. Some ciphersuites and applications\n may require cryptographic protection greater than 128 bits in\n strength.\n\n o The keys should not be used for any purpose other than generating\n and verifying tickets.\n\n o The keys should be changed regularly.\n\n o The keys should be changed if the ticket format or cryptographic\n protection algorithms change.\n\n5.6. Ticket Lifetime\n\n The TLS server controls the lifetime of the ticket. Servers\n determine the acceptable lifetime based on the operational and\n security requirements of the environments in which they are deployed.\n The ticket lifetime may be longer than the 24-hour lifetime\n recommended in [RFC4346]. TLS clients may be given a hint of the\n lifetime of the ticket. Since the lifetime of a ticket may be\n unspecified, a client has its own local policy that determines when\n it discards tickets.\n\n5.7. Alternate Ticket Formats and Distribution Schemes\n\n If the ticket format or distribution scheme defined in this document\n is not used, then great care must be taken in analyzing the security\n of the solution. In particular, if confidential information, such as\n a secret key, is transferred to the client, it MUST be done using\n secure communication so as to prevent attackers from obtaining or\n modifying the key. Also, the ticket MUST have its integrity and\n confidentiality protected with strong cryptographic techniques to\n prevent a breach in the security of the system.\n\n\n\nSalowey, et al. Standards Track [Page 13]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5.8. Identity Privacy, Anonymity, and Unlinkability\n\n This document mandates that the content of the ticket is\n confidentiality protected in order to avoid leakage of its content,\n such as user-relevant information. As such, it prevents disclosure\n of potentially sensitive information carried within the ticket.\n\n The initial handshake exchange, which was used to obtain the ticket,\n might not provide identity confidentiality of the client based on the\n properties of TLS. Another relevant security threat is the ability\n for an on-path adversary to observe multiple TLS handshakes where the\n same ticket is used, therefore concluding they belong to the same\n communication endpoints. Application designers that use the ticket\n mechanism described in this document should consider that\n unlinkability [ANON] is not necessarily provided.\n\n While a full discussion of these topics is beyond the scope of this\n document, it should be noted that it is possible to issue a ticket\n using a TLS renegotiation handshake that occurs after a secure tunnel\n has been established by a previous handshake. This may help address\n some privacy and unlinkability issues in some environments.\n\n6. Acknowledgements\n\n The authors would like to thank the following people for their help\n with preparing and reviewing this document: Eric Rescorla, Mohamad\n Badra, Tim Dierks, Nelson Bolyard, Nancy Cam-Winget, David McGrew,\n Rob Dugal, Russ Housley, Amir Herzberg, Bernard Aboba, and members of\n the TLS working group.\n\n [CSSC] describes a solution that is very similar to the one described\n in this document and gives a detailed analysis of the security\n considerations involved. [RFC2712] describes a mechanism for using\n Kerberos [RFC4120] in TLS ciphersuites, which helped inspire the use\n of tickets to avoid server state. [RFC4851] makes use of a similar\n mechanism to avoid maintaining server state for the cryptographic\n tunnel. [SC97] also investigates the concept of stateless sessions.\n\n The authors would also like to thank Jan Nordqvist, who found the\n encoding error in RFC 4507, corrected by this document. In addition\n Nagendra Modadugu, Wan-Teh Chang, and Michael D'Errico provided\n useful feedback during the review of this document.\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 14]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n7. IANA Considerations\n\n IANA has assigned a TLS extension number of 35 to the SessionTicket\n TLS extension from the TLS registry of ExtensionType values defined\n in [RFC4366].\n\n IANA has assigned a TLS HandshakeType number 4 to the\n NewSessionTicket handshake type from the TLS registry of\n HandshakeType values defined in [RFC4346].\n\n This document does not require any actions or assignments from IANA.\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2246] Dierks, T. and C. Allen, \"The TLS Protocol Version 1.0\",\n RFC 2246, January 1999.\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346, April 2006.\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, April 2006.\n\n [RFC4507] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 4507, May 2006.\n\n8.2. Informative References\n\n [AES] National Institute of Standards and Technology, \"Advanced\n Encryption Standard (AES)\", Federal Information Processing\n Standards (FIPS) Publication 197, November 2001.\n\n [ANON] Pfitzmann, A. and M. Hansen, \"Anonymity, Unlinkability,\n Unobservability, Pseudonymity, and Identity Management - A\n Consolidated Proposal for Terminology\", http://\n dud.inf.tu-dresden.de/literatur/\n Anon_Terminology_v0.26-1.pdf Version 0.26, December 2005.\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 15]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n [CBC] National Institute of Standards and Technology,\n \"Recommendation for Block Cipher Modes of Operation -\n Methods and Techniques\", NIST Special Publication 800-38A,\n December 2001.\n\n [CSSC] Shacham, H., Boneh, D., and E. Rescorla, \"Client-side\n caching for TLS\", Transactions on Information and System\n Security (TISSEC) , Volume 7, Issue 4, November 2004.\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [RFC4086] Eastlake, D., Schiller, J., and S. Crocker, \"Randomness\n Requirements for Security\", BCP 106, RFC 4086, June 2005.\n\n [RFC4120] Neuman, C., Yu, T., Hartman, S., and K. Raeburn, \"The\n Kerberos Network Authentication Service (V5)\", RFC 4120,\n July 2005.\n\n [RFC4279] Eronen, P. and H. Tschofenig, \"Pre-Shared Key Ciphersuites\n for Transport Layer Security (TLS)\", RFC 4279,\n December 2005.\n\n [RFC4634] Eastlake, D. and T. Hansen, \"US Secure Hash Algorithms\n (SHA and HMAC-SHA)\", RFC 4634, July 2006.\n\n [RFC4851] Cam-Winget, N., McGrew, D., Salowey, J., and H. Zhou, \"The\n Flexible Authentication via Secure Tunneling Extensible\n Authentication Protocol Method (EAP-FAST)\", RFC 4851,\n May 2007.\n\n [SC97] Aura, T. and P. Nikander, \"Stateless Connections\",\n Proceedings of the First International Conference on\n Information and Communication Security (ICICS '97) , 1997.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 16]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAppendix A. Discussion of Changes to RFC 4507\n\n RFC 4507 [RFC4507] defines a mechanism to resume a TLS session\n without maintaining server side state by specifying an encrypted\n ticket that is maintained on the client. The client presents this\n ticket to the server in a SessionTicket hello extension. The\n encoding in RFC 4507 used the XDR style encoding specified in TLS\n [RFC4346].\n\n An error in the encoding caused the specification to differ from\n deployed implementations. At the time of this writing there are no\n known implementations that follow the encoding specified in RFC 4507.\n This update to RFC 4507 aligns the document with these currently\n deployed implementations.\n\n Erroneous encoding in RFC 4507 resulted in two length fields; one for\n the extension contents and one for the ticket itself. Hence, for a\n ticket that is 256 bytes long and begins with the hex value FF FF,\n the encoding of the extension would be as follows according to RFC\n 4507:\n\n 00 23 Ticket Extension type 35\n 01 02 Length of extension contents\n 01 00 Length of ticket\n FF FF .. .. Actual ticket\n\n The update proposed in this document reflects what implementations\n actually encode, namely it removes the redundant length field. So,\n for a ticket that is 256 bytes long and begins with the hex value FF\n FF, the encoding of the extension would be as follows according to\n this update:\n\n 00 23 Extension type 35\n 01 00 Length of extension contents (ticket)\n FF FF .. .. Actual ticket\n\n A server implemented according to RFC 4507 receiving a ticket\n extension from a client conforming to this document would interpret\n the first two bytes of the ticket as the length of this ticket. This\n will result in either an inconsistent length field or in the\n processing of a ticket missing the first two bytes. In the first\n case, the server should reject the request based on a malformed\n length. In the second case, the server should reject the ticket\n based on a malformed ticket, incorrect key version, or failed\n decryption. A server implementation based on this update receiving\n an RFC 4507 extension would interpret the first length field as the\n\n\n\n\n\nSalowey, et al. Standards Track [Page 17]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n length of the ticket and include the second two length bytes as the\n first bytes in the ticket, resulting in the ticket being rejected\n based on a malformed ticket, incorrect key version, or failed\n decryption.\n\n Note that the encoding of an empty SessionTicket extension was\n ambiguous in RFC 4507. An RFC 4507 implementation may have encoded\n it as:\n\n 00 23 Extension type 35\n 00 02 Length of extension contents\n 00 00 Length of ticket\n\n or it may have encoded it the same way as this update:\n\n 00 23 Extension type 35\n 00 00 Length of extension contents\n\n A server wishing to support RFC 4507 clients should respond to an\n empty SessionTicket extension encoded the same way as it received it.\n\n A server implementation can construct tickets such that it can detect\n an RFC 4507 implementation, if one existed, by including a cookie at\n the beginning of the tickets that can be differentiated from a valid\n length. For example, if an implementation constructed tickets to\n start with the hex values FF FF, then it could determine where the\n ticket begins and determine the length correctly from the type of\n length fields present.\n\n This document makes a few additional changes to RFC 4507 listed\n below.\n\n o Clarifying that the server can allow session resumption using a\n ticket without issuing a new ticket in Section 3.1.\n\n o Clarifying that the lifetime is relative to when the ticket is\n received in section 3.3.\n\n o Clarifying that the NewSessionTicket handshake message is included\n in the hash generated for the Finished messages in Section 3.3.\n\n o Clarifying the interaction with TLS Session ID in Section 3.4.\n\n o Recommending the use of SHA-256 for the integrity protection of\n the ticket in Section 4.\n\n o Clarifying that additional data can be included in the\n StatePlaintext structure in Section 4.\n\n\n\nSalowey, et al. Standards Track [Page 18]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAuthors' Addresses\n\n Joseph Salowey\n Cisco Systems\n 2901 3rd Ave\n Seattle, WA 98121\n US\n\n EMail: [redacted-email]\n\n\n Hao Zhou\n Cisco Systems\n 4125 Highlander Parkway\n Richfield, OH 44286\n US\n\n EMail: [redacted-email]\n\n\n Pasi Eronen\n Nokia Research Center\n P.O. Box 407\n FIN-00045 Nokia Group\n Finland\n\n EMail: [redacted-email]\n\n\n Hannes Tschofenig\n Nokia Siemens Networks\n Otto-Hahn-Ring 6\n Munich, Bayern 81739\n Germany\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 19]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nFull Copyright Statement\n\n Copyright (C) The IETF Trust (2008).\n\n This document is subject to the rights, licenses and restrictions\n contained in BCP 78, and except as set forth therein, the authors\n retain all their rights.\n\n This document and the information contained herein are provided on an\n \"AS IS\" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\n OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\n THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\n OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\n THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\n WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\n\nIntellectual Property\n\n The IETF takes no position regarding the validity or scope of any\n Intellectual Property Rights or other rights that might be claimed to\n pertain to the implementation or use of the technology described in\n this document or the extent to which any license under such rights\n might or might not be available; nor does it represent that it has\n made any independent effort to identify any such rights. Information\n on the procedures with respect to rights in RFC documents can be\n found in BCP 78 and BCP 79.\n\n Copies of IPR disclosures made to the IETF Secretariat and any\n assurances of licenses to be made available, or the result of an\n attempt made to obtain a general license or permission for the use of\n such proprietary rights by implementers or users of this\n specification can be obtained from the IETF on-line IPR repository at\n http://www.ietf.org/ipr.\n\n The IETF invites any interested party to bring to its attention any\n copyrights, patents or patent applications, or other proprietary\n rights that may cover technology that may be required to implement\n this standard. Please address the information to the IETF at\n ietf-ipr@ietf.org.\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 20]", "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"draft_references": [], "excluded_rfc_relation_targets": [{"kind": "obsoletes", "reason": "not_requested", "rfc_number": 3268}, {"kind": "obsoletes", "reason": "not_requested", "rfc_number": 4346}, {"kind": "obsoletes", "reason": "not_requested", "rfc_number": 4366}, {"kind": "updates", "reason": "not_requested", "rfc_number": 4492}], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5246 Independent", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "5246"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 3268, 4346, 4366 E. Rescorla", "fact_id": "obsoletes:3268", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "3268"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 3268, 4346, 4366 E. Rescorla", "fact_id": "obsoletes:4346", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "4346"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 3268, 4346, 4366 E. Rescorla", "fact_id": "obsoletes:4366", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "4366"}, {"char_end": 291, "char_start": 219, "evidence_quote": "Updates: 4492 RTFM, Inc.", "fact_id": "updates:4492", "field": "updates", "raw_char_end": 297, "raw_char_start": 225, "value": "4492"}], "kind": "rfc", "occurred_at": "2008-08-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 22, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb", "record_id": "ietf:rfc:5246", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "rfc_number": 5246, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb", "source_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "temporal_semantics": "source_event_time", "text": "Network Working Group T. Dierks\nRequest for Comments: 5246 Independent\nObsoletes: 3268, 4346, 4366 E. Rescorla\nUpdates: 4492 RTFM, Inc.\nCategory: Standards Track August 2008\n\n\n The Transport Layer Security (TLS) Protocol\n Version 1.2\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document specifies Version 1.2 of the Transport Layer Security\n (TLS) protocol. The TLS protocol provides communications security\n over the Internet. The protocol allows client/server applications to\n communicate in a way that is designed to prevent eavesdropping,\n tampering, or message forgery.\n\nTable of Contents\n\n 1. Introduction ....................................................4\n 1.1. Requirements Terminology ...................................5\n 1.2. Major Differences from TLS 1.1 .............................5\n 2. Goals ...........................................................6\n 3. Goals of This Document ..........................................7\n 4. Presentation Language ...........................................7\n 4.1. Basic Block Size ...........................................7\n 4.2. Miscellaneous ..............................................8\n 4.3. Vectors ....................................................8\n 4.4. Numbers ....................................................9\n 4.5. Enumerateds ................................................9\n 4.6. Constructed Types .........................................10\n 4.6.1. Variants ...........................................10\n 4.7. Cryptographic Attributes ..................................12\n 4.8. Constants .................................................14\n 5. HMAC and the Pseudorandom Function .............................14\n 6. The TLS Record Protocol ........................................15\n 6.1. Connection States .........................................16\n 6.2. Record Layer ..............................................19\n 6.2.1. Fragmentation ......................................19\n\n\n\nDierks & Rescorla Standards Track [Page 1]\n\f\nRFC 5246 TLS August 2008\n\n\n 6.2.2. Record Compression and Decompression ...............20\n 6.2.3. Record Payload Protection ..........................21\n 6.2.3.1. Null or Standard Stream Cipher ............22\n 6.2.3.2. CBC Block Cipher ..........................22\n 6.2.3.3. AEAD Ciphers ..............................24\n 6.3. Key Calculation ...........................................25\n 7. The TLS Handshaking Protocols ..................................26\n 7.1. Change Cipher Spec Protocol ...............................27\n 7.2. Alert Protocol ............................................28\n 7.2.1. Closure Alerts .....................................29\n 7.2.2. Error Alerts .......................................30\n 7.3. Handshake Protocol Overview ...............................33\n 7.4. Handshake Protocol ........................................37\n 7.4.1. Hello Messages .....................................38\n 7.4.1.1. Hello Request .............................38\n 7.4.1.2. Client Hello ..............................39\n 7.4.1.3. Server Hello ..............................42\n 7.4.1.4. Hello Extensions ..........................44\n 7.4.1.4.1. Signature Algorithms ...........45\n 7.4.2. Server Certificate .................................47\n 7.4.3. Server Key Exchange Message ........................50\n 7.4.4. Certificate Request ................................53\n 7.4.5. Server Hello Done ..................................55\n 7.4.6. Client Certificate .................................55\n 7.4.7. Client Key Exchange Message ........................57\n 7.4.7.1. RSA-Encrypted Premaster Secret Message ....58\n 7.4.7.2. Client Diffie-Hellman Public Value ........61\n 7.4.8. Certificate Verify .................................62\n 7.4.9. Finished ...........................................63\n 8. Cryptographic Computations .....................................64\n 8.1. Computing the Master Secret ...............................64\n 8.1.1. RSA ................................................65\n 8.1.2. Diffie-Hellman .....................................65\n 9. Mandatory Cipher Suites ........................................65\n 10. Application Data Protocol .....................................65\n 11. Security Considerations .......................................65\n 12. IANA Considerations ...........................................65\n Appendix A. Protocol Data Structures and Constant Values ..........68\n A.1. Record Layer ..............................................68\n A.2. Change Cipher Specs Message ...............................69\n A.3. Alert Messages ............................................69\n A.4. Handshake Protocol ........................................70\n A.4.1. Hello Messages .....................................71\n A.4.2. Server Authentication and Key Exchange Messages ....72\n A.4.3. Client Authentication and Key Exchange Messages ....74\n A.4.4. Handshake Finalization Message .....................74\n A.5. The Cipher Suite ..........................................75\n A.6. The Security Parameters ...................................77\n\n\n\nDierks & Rescorla Standards Track [Page 2]\n\f\nRFC 5246 TLS August 2008\n\n\n A.7. Changes to RFC 4492 .......................................78\n Appendix B. Glossary ..............................................78\n Appendix C. Cipher Suite Definitions ..............................83\n Appendix D. Implementation Notes ..................................85\n D.1. Random Number Generation and Seeding ......................85\n D.2. Certificates and Authentication ...........................85\n D.3. Cipher Suites .............................................85\n D.4. Implementation Pitfalls ...................................85\n Appendix E. Backward Compatibility ................................87\n E.1. Compatibility with TLS 1.0/1.1 and SSL 3.0 ................87\n E.2. Compatibility with SSL 2.0 ................................88\n E.3. Avoiding Man-in-the-Middle Version Rollback ...............90\n Appendix F. Security Analysis .....................................91\n F.1. Handshake Protocol ........................................91\n F.1.1. Authentication and Key Exchange ....................91\n F.1.1.1. Anonymous Key Exchange ....................91\n F.1.1.2. RSA Key Exchange and Authentication .......92\n F.1.1.3. Diffie-Hellman Key Exchange with\n Authentication ............................92\n F.1.2. Version Rollback Attacks ...........................93\n F.1.3. Detecting Attacks Against the Handshake Protocol ...94\n F.1.4. Resuming Sessions ..................................94\n F.2. Protecting Application Data ...............................94\n F.3. Explicit IVs ..............................................95\n F.4. Security of Composite Cipher Modes ........................95\n F.5. Denial of Service .........................................96\n F.6. Final Notes ...............................................96\n Normative References ..............................................97\n Informative References ............................................98\n Working Group Information ........................................101\n Contributors .....................................................101\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 3]\n\f\nRFC 5246 TLS August 2008\n\n\n1. Introduction\n\n The primary goal of the TLS protocol is to provide privacy and data\n integrity between two communicating applications. The protocol is\n composed of two layers: the TLS Record Protocol and the TLS Handshake\n Protocol. At the lowest level, layered on top of some reliable\n transport protocol (e.g., TCP [TCP]), is the TLS Record Protocol.\n The TLS Record Protocol provides connection security that has two\n basic properties:\n\n - The connection is private. Symmetric cryptography is used for\n data encryption (e.g., AES [AES], RC4 [SCH], etc.). The keys for\n this symmetric encryption are generated uniquely for each\n connection and are based on a secret negotiated by another\n protocol (such as the TLS Handshake Protocol). The Record\n Protocol can also be used without encryption.\n\n - The connection is reliable. Message transport includes a message\n integrity check using a keyed MAC. Secure hash functions (e.g.,\n SHA-1, etc.) are used for MAC computations. The Record Protocol\n can operate without a MAC, but is generally only used in this mode\n while another protocol is using the Record Protocol as a transport\n for negotiating security parameters.\n\n The TLS Record Protocol is used for encapsulation of various higher-\n level protocols. One such encapsulated protocol, the TLS Handshake\n Protocol, allows the server and client to authenticate each other and\n to negotiate an encryption algorithm and cryptographic keys before\n the application protocol transmits or receives its first byte of\n data. The TLS Handshake Protocol provides connection security that\n has three basic properties:\n\n - The peer's identity can be authenticated using asymmetric, or\n public key, cryptography (e.g., RSA [RSA], DSA [DSS], etc.). This\n authentication can be made optional, but is generally required for\n at least one of the peers.\n\n - The negotiation of a shared secret is secure: the negotiated\n secret is unavailable to eavesdroppers, and for any authenticated\n connection the secret cannot be obtained, even by an attacker who\n can place himself in the middle of the connection.\n\n - The negotiation is reliable: no attacker can modify the\n negotiation communication without being detected by the parties to\n the communication.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 4]\n\f\nRFC 5246 TLS August 2008\n\n\n One advantage of TLS is that it is application protocol independent.\n Higher-level protocols can layer on top of the TLS protocol\n transparently. The TLS standard, however, does not specify how\n protocols add security with TLS; the decisions on how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n1.1. Requirements Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [REQ].\n\n1.2. Major Differences from TLS 1.1\n\n This document is a revision of the TLS 1.1 [TLS1.1] protocol which\n contains improved flexibility, particularly for negotiation of\n cryptographic algorithms. The major changes are:\n\n - The MD5/SHA-1 combination in the pseudorandom function (PRF) has\n been replaced with cipher-suite-specified PRFs. All cipher suites\n in this document use P_SHA256.\n\n - The MD5/SHA-1 combination in the digitally-signed element has been\n replaced with a single hash. Signed elements now include a field\n that explicitly specifies the hash algorithm used.\n\n - Substantial cleanup to the client's and server's ability to\n specify which hash and signature algorithms they will accept.\n Note that this also relaxes some of the constraints on signature\n and hash algorithms from previous versions of TLS.\n\n - Addition of support for authenticated encryption with additional\n data modes.\n\n - TLS Extensions definition and AES Cipher Suites were merged in\n from external [TLSEXT] and [TLSAES].\n\n - Tighter checking of EncryptedPreMasterSecret version numbers.\n\n - Tightened up a number of requirements.\n\n - Verify_data length now depends on the cipher suite (default is\n still 12).\n\n - Cleaned up description of Bleichenbacher/Klima attack defenses.\n\n\n\n\nDierks & Rescorla Standards Track [Page 5]\n\f\nRFC 5246 TLS August 2008\n\n\n - Alerts MUST now be sent in many cases.\n\n - After a certificate_request, if no certificates are available,\n clients now MUST send an empty certificate list.\n\n - TLS_RSA_WITH_AES_128_CBC_SHA is now the mandatory to implement\n cipher suite.\n\n - Added HMAC-SHA256 cipher suites.\n\n - Removed IDEA and DES cipher suites. They are now deprecated and\n will be documented in a separate document.\n\n - Support for the SSLv2 backward-compatible hello is now a MAY, not\n a SHOULD, with sending it a SHOULD NOT. Support will probably\n become a SHOULD NOT in the future.\n\n - Added limited \"fall-through\" to the presentation language to allow\n multiple case arms to have the same encoding.\n\n - Added an Implementation Pitfalls sections\n\n - The usual clarifications and editorial work.\n\n2. Goals\n\n The goals of the TLS protocol, in order of priority, are as follows:\n\n 1. Cryptographic security: TLS should be used to establish a secure\n connection between two parties.\n\n 2. Interoperability: Independent programmers should be able to\n develop applications utilizing TLS that can successfully exchange\n cryptographic parameters without knowledge of one another's code.\n\n 3. Extensibility: TLS seeks to provide a framework into which new\n public key and bulk encryption methods can be incorporated as\n necessary. This will also accomplish two sub-goals: preventing\n the need to create a new protocol (and risking the introduction of\n possible new weaknesses) and avoiding the need to implement an\n entire new security library.\n\n 4. Relative efficiency: Cryptographic operations tend to be highly\n CPU intensive, particularly public key operations. For this\n reason, the TLS protocol has incorporated an optional session\n caching scheme to reduce the number of connections that need to be\n established from scratch. Additionally, care has been taken to\n reduce network activity.\n\n\n\nDierks & Rescorla Standards Track [Page 6]\n\f\nRFC 5246 TLS August 2008\n\n\n3. Goals of This Document\n\n This document and the TLS protocol itself are based on the SSL 3.0\n Protocol Specification as published by Netscape. The differences\n between this protocol and SSL 3.0 are not dramatic, but they are\n significant enough that the various versions of TLS and SSL 3.0 do\n not interoperate (although each protocol incorporates a mechanism by\n which an implementation can back down to prior versions). This\n document is intended primarily for readers who will be implementing\n the protocol and for those doing cryptographic analysis of it. The\n specification has been written with this in mind, and it is intended\n to reflect the needs of those two groups. For that reason, many of\n the algorithm-dependent data structures and rules are included in the\n body of the text (as opposed to in an appendix), providing easier\n access to them.\n\n This document is not intended to supply any details of service\n definition or of interface definition, although it does cover select\n areas of policy as they are required for the maintenance of solid\n security.\n\n4. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used. The syntax draws from\n several sources in its structure. Although it resembles the\n programming language \"C\" in its syntax and XDR [XDR] in both its\n syntax and intent, it would be risky to draw too many parallels. The\n purpose of this presentation language is to document TLS only; it has\n no general application beyond that particular goal.\n\n4.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 7]\n\f\nRFC 5246 TLS August 2008\n\n\n4.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" double\n brackets.\n\n Single-byte entities containing uninterpreted data are of type\n opaque.\n\n4.3. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* 3 consecutive 3 byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n In the following example, mandatory is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 4.4). On the\n other hand, longer can represent up to 800 bytes of data, or 400\n uint16 elements, and it may be empty. Its encoding will include a\n\n\n\nDierks & Rescorla Standards Track [Page 8]\n\f\nRFC 5246 TLS August 2008\n\n\n two-byte actual length field prepended to the vector. The length of\n an encoded vector must be an even multiple of the length of a single\n element (for example, a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is 2 bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n4.4. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are formed from fixed-length series of bytes\n concatenated as described in Section 4.1 and are also unsigned. The\n following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are stored in\n network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n Note that in some cases (e.g., DH parameters) it is necessary to\n represent integers as opaque vectors. In such cases, they are\n represented as unsigned integers (i.e., leading zero octets are not\n required even if the most significant bit is set).\n\n4.5. Enumerateds\n\n An additional sparse data type is available called enum. A field of\n type enum can only assume the values declared in the definition.\n Each definition is a different type. Only enumerateds of the same\n type may be assigned or compared. Every element of an enumerated\n must be assigned a value, as demonstrated in the following example.\n Since the elements of the enumerated are not ordered, they can be\n assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n\n\n\nDierks & Rescorla Standards Track [Page 9]\n\f\nRFC 5246 TLS August 2008\n\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n For enumerateds that are never converted to external representation,\n the numerical information may be omitted.\n\n enum { low, medium, high } Amount;\n\n4.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax for definition is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } [[T]];\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n Structure definitions may be embedded.\n\n4.6.1. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. There\n must be a case arm for every element of the enumeration declared in\n the select. Case arms have limited fall-through: if two case arms\n follow in immediate succession with no fields in between, then they\n\n\n\nDierks & Rescorla Standards Track [Page 10]\n\f\nRFC 5246 TLS August 2008\n\n\n both contain the same fields. Thus, in the example below, \"orange\"\n and \"banana\" both contain V2. Note that this is a new piece of\n syntax in TLS 1.2.\n\n The body of the variant structure may be given a label for reference.\n The mechanism by which the variant is selected at runtime is not\n prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1;\n case e2: Te2;\n case e3: case e4: Te3;\n ....\n case en: Ten;\n } [[fv]];\n } [[Tv]];\n\n For example:\n\n enum { apple, orange, banana } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n select (VariantTag) { /* value of selector is implicit */\n case apple:\n V1; /* VariantBody, tag = apple */\n case orange:\n case banana:\n V2; /* VariantBody, tag = orange or banana */\n } variant_body; /* optional label on variant */\n } VariantRecord;\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 11]\n\f\nRFC 5246 TLS August 2008\n\n\n4.7. Cryptographic Attributes\n\n The five cryptographic operations -- digital signing, stream cipher\n encryption, block cipher encryption, authenticated encryption with\n additional data (AEAD) encryption, and public key encryption -- are\n designated digitally-signed, stream-ciphered, block-ciphered, aead-\n ciphered, and public-key-encrypted, respectively. A field's\n cryptographic processing is specified by prepending an appropriate\n key word designation before the field's type specification.\n Cryptographic keys are implied by the current session state (see\n Section 6.1).\n\n A digitally-signed element is encoded as a struct DigitallySigned:\n\n struct {\n SignatureAndHashAlgorithm algorithm;\n opaque signature<0..2^16-1>;\n } DigitallySigned;\n\n The algorithm field specifies the algorithm used (see Section\n 7.4.1.4.1 for the definition of this field). Note that the\n introduction of the algorithm field is a change from previous\n versions. The signature is a digital signature using those\n algorithms over the contents of the element. The contents themselves\n do not appear on the wire but are simply calculated. The length of\n the signature is specified by the signing algorithm and key.\n\n In RSA signing, the opaque vector contains the signature generated\n using the RSASSA-PKCS1-v1_5 signature scheme defined in [PKCS1]. As\n discussed in [PKCS1], the DigestInfo MUST be DER-encoded [X680]\n [X690]. For hash algorithms without parameters (which includes\n SHA-1), the DigestInfo.AlgorithmIdentifier.parameters field MUST be\n NULL, but implementations MUST accept both without parameters and\n with NULL parameters. Note that earlier versions of TLS used a\n different RSA signature scheme that did not include a DigestInfo\n encoding.\n\n In DSA, the 20 bytes of the SHA-1 hash are run directly through the\n Digital Signing Algorithm with no additional hashing. This produces\n two values, r and s. The DSA signature is an opaque vector, as\n above, the contents of which are the DER encoding of:\n\n Dss-Sig-Value ::= SEQUENCE {\n r INTEGER,\n s INTEGER\n }\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 12]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: In current terminology, DSA refers to the Digital Signature\n Algorithm and DSS refers to the NIST standard. In the original SSL\n and TLS specs, \"DSS\" was used universally. This document uses \"DSA\"\n to refer to the algorithm, \"DSS\" to refer to the standard, and it\n uses \"DSS\" in the code point definitions for historical continuity.\n\n In stream cipher encryption, the plaintext is exclusive-ORed with an\n identical amount of output generated from a cryptographically secure\n keyed pseudorandom number generator.\n\n In block cipher encryption, every block of plaintext encrypts to a\n block of ciphertext. All block cipher encryption is done in CBC\n (Cipher Block Chaining) mode, and all items that are block-ciphered\n will be an exact multiple of the cipher block length.\n\n In AEAD encryption, the plaintext is simultaneously encrypted and\n integrity protected. The input may be of any length, and aead-\n ciphered output is generally larger than the input in order to\n accommodate the integrity check value.\n\n In public key encryption, a public key algorithm is used to encrypt\n data in such a way that it can be decrypted only with the matching\n private key. A public-key-encrypted element is encoded as an opaque\n vector <0..2^16-1>, where the length is specified by the encryption\n algorithm and key.\n\n RSA encryption is done using the RSAES-PKCS1-v1_5 encryption scheme\n defined in [PKCS1].\n\n In the following example\n\n stream-ciphered struct {\n uint8 field1;\n uint8 field2;\n digitally-signed opaque {\n uint8 field3<0..255>;\n uint8 field4;\n };\n } UserType;\n\n The contents of the inner struct (field3 and field4) are used as\n input for the signature/hash algorithm, and then the entire structure\n is encrypted with a stream cipher. The length of this structure, in\n bytes, would be equal to two bytes for field1 and field2, plus two\n bytes for the signature and hash algorithm, plus two bytes for the\n length of the signature, plus the length of the output of the signing\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 13]\n\f\nRFC 5246 TLS August 2008\n\n\n algorithm. The length of the signature is known because the\n algorithm and key used for the signing are known prior to encoding or\n decoding this structure.\n\n4.8. Constants\n\n Typed constants can be defined for purposes of specification by\n declaring a symbol of the desired type and assigning values to it.\n\n Under-specified types (opaque, variable-length vectors, and\n structures that contain opaque) cannot be assigned values. No fields\n of a multi-element structure or vector may be elided.\n\n For example:\n\n struct {\n uint8 f1;\n uint8 f2;\n } Example1;\n\n Example1 ex1 = {1, 4}; /* assigns f1 = 1, f2 = 4 */\n\n5. HMAC and the Pseudorandom Function\n\n The TLS record layer uses a keyed Message Authentication Code (MAC)\n to protect message integrity. The cipher suites defined in this\n document use a construction known as HMAC, described in [HMAC], which\n is based on a hash function. Other cipher suites MAY define their\n own MAC constructions, if needed.\n\n In addition, a construction is required to do expansion of secrets\n into blocks of data for the purposes of key generation or validation.\n This pseudorandom function (PRF) takes as input a secret, a seed, and\n an identifying label and produces an output of arbitrary length.\n\n In this section, we define one PRF, based on HMAC. This PRF with the\n SHA-256 hash function is used for all cipher suites defined in this\n document and in TLS documents published prior to this document when\n TLS 1.2 is negotiated. New cipher suites MUST explicitly specify a\n PRF and, in general, SHOULD use the TLS PRF with SHA-256 or a\n stronger standard hash function.\n\n First, we define a data expansion function, P_hash(secret, data),\n that uses a single hash function to expand a secret and seed into an\n arbitrary quantity of output:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 14]\n\f\nRFC 5246 TLS August 2008\n\n\n P_hash(secret, seed) = HMAC_hash(secret, A(1) + seed) +\n HMAC_hash(secret, A(2) + seed) +\n HMAC_hash(secret, A(3) + seed) + ...\n\n where + indicates concatenation.\n\n A() is defined as:\n\n A(0) = seed\n A(i) = HMAC_hash(secret, A(i-1))\n\n P_hash can be iterated as many times as necessary to produce the\n required quantity of data. For example, if P_SHA256 is being used to\n create 80 bytes of data, it will have to be iterated three times\n (through A(3)), creating 96 bytes of output data; the last 16 bytes\n of the final iteration will then be discarded, leaving 80 bytes of\n output data.\n\n TLS's PRF is created by applying P_hash to the secret as:\n\n PRF(secret, label, seed) = P_(secret, label + seed)\n\n The label is an ASCII string. It should be included in the exact\n form it is given without a length byte or trailing null character.\n For example, the label \"slithy toves\" would be processed by hashing\n the following bytes:\n\n 73 6C 69 74 68 79 20 74 6F 76 65 73\n\n6. The TLS Record Protocol\n\n The TLS Record Protocol is a layered protocol. At each layer,\n messages may include fields for length, description, and content.\n The Record Protocol takes messages to be transmitted, fragments the\n data into manageable blocks, optionally compresses the data, applies\n a MAC, encrypts, and transmits the result. Received data is\n decrypted, verified, decompressed, reassembled, and then delivered to\n higher-level clients.\n\n Four protocols that use the record protocol are described in this\n document: the handshake protocol, the alert protocol, the change\n cipher spec protocol, and the application data protocol. In order to\n allow extension of the TLS protocol, additional record content types\n can be supported by the record protocol. New record content type\n values are assigned by IANA in the TLS Content Type Registry as\n described in Section 12.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 15]\n\f\nRFC 5246 TLS August 2008\n\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST send an\n unexpected_message alert.\n\n Any protocol designed for use over TLS must be carefully designed to\n deal with all possible attacks against it. As a practical matter,\n this means that the protocol designer must be aware of what security\n properties TLS does and does not provide and cannot safely rely on\n the latter.\n\n Note in particular that type and length of a record are not protected\n by encryption. If this information is itself sensitive, application\n designers may wish to take steps (padding, cover traffic) to minimize\n information leakage.\n\n6.1. Connection States\n\n A TLS connection state is the operating environment of the TLS Record\n Protocol. It specifies a compression algorithm, an encryption\n algorithm, and a MAC algorithm. In addition, the parameters for\n these algorithms are known: the MAC key and the bulk encryption keys\n for the connection in both the read and the write directions.\n Logically, there are always four connection states outstanding: the\n current read and write states, and the pending read and write states.\n All records are processed under the current read and write states.\n The security parameters for the pending states can be set by the TLS\n Handshake Protocol, and the ChangeCipherSpec can selectively make\n either of the pending states current, in which case the appropriate\n current state is disposed of and replaced with the pending state; the\n pending state is then reinitialized to an empty state. It is illegal\n to make a state that has not been initialized with security\n parameters a current state. The initial current state always\n specifies that no encryption, compression, or MAC will be used.\n\n The security parameters for a TLS Connection read and write state are\n set by providing the following values:\n\n connection end\n Whether this entity is considered the \"client\" or the \"server\" in\n this connection.\n\n PRF algorithm\n An algorithm used to generate keys from the master secret (see\n Sections 5 and 6.3).\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 16]\n\f\nRFC 5246 TLS August 2008\n\n\n bulk encryption algorithm\n An algorithm to be used for bulk encryption. This specification\n includes the key size of this algorithm, whether it is a block,\n stream, or AEAD cipher, the block size of the cipher (if\n appropriate), and the lengths of explicit and implicit\n initialization vectors (or nonces).\n\n MAC algorithm\n An algorithm to be used for message authentication. This\n specification includes the size of the value returned by the MAC\n algorithm.\n\n compression algorithm\n An algorithm to be used for data compression. This specification\n must include all information the algorithm requires to do\n compression.\n\n master secret\n A 48-byte secret shared between the two peers in the connection.\n\n client random\n A 32-byte value provided by the client.\n\n server random\n A 32-byte value provided by the server.\n\n These parameters are defined in the presentation language as:\n\n enum { server, client } ConnectionEnd;\n\n enum { tls_prf_sha256 } PRFAlgorithm;\n\n enum { null, rc4, 3des, aes }\n BulkCipherAlgorithm;\n\n enum { stream, block, aead } CipherType;\n\n enum { null, hmac_md5, hmac_sha1, hmac_sha256,\n hmac_sha384, hmac_sha512} MACAlgorithm;\n\n enum { null(0), (255) } CompressionMethod;\n\n /* The algorithms specified in CompressionMethod, PRFAlgorithm,\n BulkCipherAlgorithm, and MACAlgorithm may be added to. */\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 17]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ConnectionEnd entity;\n PRFAlgorithm prf_algorithm;\n BulkCipherAlgorithm bulk_cipher_algorithm;\n CipherType cipher_type;\n uint8 enc_key_length;\n uint8 block_length;\n uint8 fixed_iv_length;\n uint8 record_iv_length;\n MACAlgorithm mac_algorithm;\n uint8 mac_length;\n uint8 mac_key_length;\n CompressionMethod compression_algorithm;\n opaque master_secret[48];\n opaque client_random[32];\n opaque server_random[32];\n } SecurityParameters;\n\n The record layer will use the security parameters to generate the\n following six items (some of which are not required by all ciphers,\n and are thus empty):\n\n client write MAC key\n server write MAC key\n client write encryption key\n server write encryption key\n client write IV\n server write IV\n\n The client write parameters are used by the server when receiving and\n processing records and vice versa. The algorithm used for generating\n these items from the security parameters is described in Section 6.3.\n\n Once the security parameters have been set and the keys have been\n generated, the connection states can be instantiated by making them\n the current states. These current states MUST be updated for each\n record processed. Each connection state includes the following\n elements:\n\n compression state\n The current state of the compression algorithm.\n\n cipher state\n The current state of the encryption algorithm. This will consist\n of the scheduled key for that connection. For stream ciphers,\n this will also contain whatever state information is necessary to\n allow the stream to continue to encrypt or decrypt data.\n\n\n\n\nDierks & Rescorla Standards Track [Page 18]\n\f\nRFC 5246 TLS August 2008\n\n\n MAC key\n The MAC key for this connection, as generated above.\n\n sequence number\n Each connection state contains a sequence number, which is\n maintained separately for read and write states. The sequence\n number MUST be set to zero whenever a connection state is made the\n active state. Sequence numbers are of type uint64 and may not\n exceed 2^64-1. Sequence numbers do not wrap. If a TLS\n implementation would need to wrap a sequence number, it must\n renegotiate instead. A sequence number is incremented after each\n record: specifically, the first record transmitted under a\n particular connection state MUST use sequence number 0.\n\n6.2. Record Layer\n\n The TLS record layer receives uninterpreted data from higher layers\n in non-empty blocks of arbitrary size.\n\n6.2.1. Fragmentation\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Client\n message boundaries are not preserved in the record layer (i.e.,\n multiple client messages of the same ContentType MAY be coalesced\n into a single TLSPlaintext record, or a single message MAY be\n fragmented across several records).\n\n struct {\n uint8 major;\n uint8 minor;\n } ProtocolVersion;\n\n enum {\n change_cipher_spec(20), alert(21), handshake(22),\n application_data(23), (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type\n The higher-level protocol used to process the enclosed fragment.\n\n\n\n\nDierks & Rescorla Standards Track [Page 19]\n\f\nRFC 5246 TLS August 2008\n\n\n version\n The version of the protocol being employed. This document\n describes TLS Version 1.2, which uses the version { 3, 3 }. The\n version value 3.3 is historical, deriving from the use of {3, 1}\n for TLS 1.0. (See Appendix A.1.) Note that a client that\n supports multiple versions of TLS may not know what version will\n be employed before it receives the ServerHello. See Appendix E\n for discussion about what record layer version number should be\n employed for ClientHello.\n\n length\n The length (in bytes) of the following TLSPlaintext.fragment. The\n length MUST NOT exceed 2^14.\n\n fragment\n The application data. This data is transparent and treated as an\n independent block to be dealt with by the higher-level protocol\n specified by the type field.\n\n Implementations MUST NOT send zero-length fragments of Handshake,\n Alert, or ChangeCipherSpec content types. Zero-length fragments of\n Application data MAY be sent as they are potentially useful as a\n traffic analysis countermeasure.\n\n Note: Data of different TLS record layer content types MAY be\n interleaved. Application data is generally of lower precedence for\n transmission than other content types. However, records MUST be\n delivered to the network in the same order as they are protected by\n the record layer. Recipients MUST receive and process interleaved\n application layer traffic during handshakes subsequent to the first\n one on a connection.\n\n6.2.2. Record Compression and Decompression\n\n All records are compressed using the compression algorithm defined in\n the current session state. There is always an active compression\n algorithm; however, initially it is defined as\n CompressionMethod.null. The compression algorithm translates a\n TLSPlaintext structure into a TLSCompressed structure. Compression\n functions are initialized with default state information whenever a\n connection state is made active. [RFC3749] describes compression\n algorithms for TLS.\n\n Compression must be lossless and may not increase the content length\n by more than 1024 bytes. If the decompression function encounters a\n TLSCompressed.fragment that would decompress to a length in excess of\n 2^14 bytes, it MUST report a fatal decompression failure error.\n\n\n\n\nDierks & Rescorla Standards Track [Page 20]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ContentType type; /* same as TLSPlaintext.type */\n ProtocolVersion version;/* same as TLSPlaintext.version */\n uint16 length;\n opaque fragment[TLSCompressed.length];\n } TLSCompressed;\n\n length\n The length (in bytes) of the following TLSCompressed.fragment.\n The length MUST NOT exceed 2^14 + 1024.\n\n fragment\n The compressed form of TLSPlaintext.fragment.\n\n Note: A CompressionMethod.null operation is an identity operation;\n no fields are altered.\n\n Implementation note: Decompression functions are responsible for\n ensuring that messages cannot cause internal buffer overflows.\n\n6.2.3. Record Payload Protection\n\n The encryption and MAC functions translate a TLSCompressed\n structure into a TLSCiphertext. The decryption functions reverse\n the process. The MAC of the record also includes a sequence\n number so that missing, extra, or repeated messages are\n detectable.\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n select (SecurityParameters.cipher_type) {\n case stream: GenericStreamCipher;\n case block: GenericBlockCipher;\n case aead: GenericAEADCipher;\n } fragment;\n } TLSCiphertext;\n\n type\n The type field is identical to TLSCompressed.type.\n\n version\n The version field is identical to TLSCompressed.version.\n\n length\n The length (in bytes) of the following TLSCiphertext.fragment.\n The length MUST NOT exceed 2^14 + 2048.\n\n\n\nDierks & Rescorla Standards Track [Page 21]\n\f\nRFC 5246 TLS August 2008\n\n\n fragment\n The encrypted form of TLSCompressed.fragment, with the MAC.\n\n6.2.3.1. Null or Standard Stream Cipher\n\n Stream ciphers (including BulkCipherAlgorithm.null; see Appendix A.6)\n convert TLSCompressed.fragment structures to and from stream\n TLSCiphertext.fragment structures.\n\n stream-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n } GenericStreamCipher;\n\n The MAC is generated as:\n\n MAC(MAC_write_key, seq_num +\n TLSCompressed.type +\n TLSCompressed.version +\n TLSCompressed.length +\n TLSCompressed.fragment);\n\n where \"+\" denotes concatenation.\n\n seq_num\n The sequence number for this record.\n\n MAC\n The MAC algorithm specified by SecurityParameters.mac_algorithm.\n\n Note that the MAC is computed before encryption. The stream cipher\n encrypts the entire block, including the MAC. For stream ciphers\n that do not use a synchronization vector (such as RC4), the stream\n cipher state from the end of one record is simply used on the\n subsequent packet. If the cipher suite is TLS_NULL_WITH_NULL_NULL,\n encryption consists of the identity operation (i.e., the data is not\n encrypted, and the MAC size is zero, implying that no MAC is used).\n For both null and stream ciphers, TLSCiphertext.length is\n TLSCompressed.length plus SecurityParameters.mac_length.\n\n6.2.3.2. CBC Block Cipher\n\n For block ciphers (such as 3DES or AES), the encryption and MAC\n functions convert TLSCompressed.fragment structures to and from block\n TLSCiphertext.fragment structures.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 22]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n opaque IV[SecurityParameters.record_iv_length];\n block-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n uint8 padding[GenericBlockCipher.padding_length];\n uint8 padding_length;\n };\n } GenericBlockCipher;\n\n The MAC is generated as described in Section 6.2.3.1.\n\n IV\n The Initialization Vector (IV) SHOULD be chosen at random, and\n MUST be unpredictable. Note that in versions of TLS prior to 1.1,\n there was no IV field, and the last ciphertext block of the\n previous record (the \"CBC residue\") was used as the IV. This was\n changed to prevent the attacks described in [CBCATT]. For block\n ciphers, the IV length is of length\n SecurityParameters.record_iv_length, which is equal to the\n SecurityParameters.block_size.\n\n padding\n Padding that is added to force the length of the plaintext to be\n an integral multiple of the block cipher's block length. The\n padding MAY be any length up to 255 bytes, as long as it results\n in the TLSCiphertext.length being an integral multiple of the\n block length. Lengths longer than necessary might be desirable to\n frustrate attacks on a protocol that are based on analysis of the\n lengths of exchanged messages. Each uint8 in the padding data\n vector MUST be filled with the padding length value. The receiver\n MUST check this padding and MUST use the bad_record_mac alert to\n indicate padding errors.\n\n padding_length\n The padding length MUST be such that the total size of the\n GenericBlockCipher structure is a multiple of the cipher's block\n length. Legal values range from zero to 255, inclusive. This\n length specifies the length of the padding field exclusive of the\n padding_length field itself.\n\n The encrypted data length (TLSCiphertext.length) is one more than the\n sum of SecurityParameters.block_length, TLSCompressed.length,\n SecurityParameters.mac_length, and padding_length.\n\n Example: If the block length is 8 bytes, the content length\n (TLSCompressed.length) is 61 bytes, and the MAC length is 20 bytes,\n then the length before padding is 82 bytes (this does not include the\n\n\n\nDierks & Rescorla Standards Track [Page 23]\n\f\nRFC 5246 TLS August 2008\n\n\n IV. Thus, the padding length modulo 8 must be equal to 6 in order to\n make the total length an even multiple of 8 bytes (the block length).\n The padding length can be 6, 14, 22, and so on, through 254. If the\n padding length were the minimum necessary, 6, the padding would be 6\n bytes, each containing the value 6. Thus, the last 8 octets of the\n GenericBlockCipher before block encryption would be xx 06 06 06 06 06\n 06 06, where xx is the last octet of the MAC.\n\n Note: With block ciphers in CBC mode (Cipher Block Chaining), it is\n critical that the entire plaintext of the record be known before any\n ciphertext is transmitted. Otherwise, it is possible for the\n attacker to mount the attack described in [CBCATT].\n\n Implementation note: Canvel et al. [CBCTIME] have demonstrated a\n timing attack on CBC padding based on the time required to compute\n the MAC. In order to defend against this attack, implementations\n MUST ensure that record processing time is essentially the same\n whether or not the padding is correct. In general, the best way to\n do this is to compute the MAC even if the padding is incorrect, and\n only then reject the packet. For instance, if the pad appears to be\n incorrect, the implementation might assume a zero-length pad and then\n compute the MAC. This leaves a small timing channel, since MAC\n performance depends to some extent on the size of the data fragment,\n but it is not believed to be large enough to be exploitable, due to\n the large block size of existing MACs and the small size of the\n timing signal.\n\n6.2.3.3. AEAD Ciphers\n\n For AEAD [AEAD] ciphers (such as [CCM] or [GCM]), the AEAD function\n converts TLSCompressed.fragment structures to and from AEAD\n TLSCiphertext.fragment structures.\n\n struct {\n opaque nonce_explicit[SecurityParameters.record_iv_length];\n aead-ciphered struct {\n opaque content[TLSCompressed.length];\n };\n } GenericAEADCipher;\n\n AEAD ciphers take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [AEAD]. The key is either the\n client_write_key or the server_write_key. No MAC key is used.\n\n Each AEAD cipher suite MUST specify how the nonce supplied to the\n AEAD operation is constructed, and what is the length of the\n GenericAEADCipher.nonce_explicit part. In many cases, it is\n\n\n\nDierks & Rescorla Standards Track [Page 24]\n\f\nRFC 5246 TLS August 2008\n\n\n appropriate to use the partially implicit nonce technique described\n in Section 3.2.1 of [AEAD]; with record_iv_length being the length of\n the explicit part. In this case, the implicit part SHOULD be derived\n from key_block as client_write_iv and server_write_iv (as described\n in Section 6.3), and the explicit part is included in\n GenericAEAEDCipher.nonce_explicit.\n\n The plaintext is the TLSCompressed.fragment.\n\n The additional authenticated data, which we denote as\n additional_data, is defined as follows:\n\n additional_data = seq_num + TLSCompressed.type +\n TLSCompressed.version + TLSCompressed.length;\n\n where \"+\" denotes concatenation.\n\n The aead_output consists of the ciphertext output by the AEAD\n encryption operation. The length will generally be larger than\n TLSCompressed.length, but by an amount that varies with the AEAD\n cipher. Since the ciphers might incorporate padding, the amount of\n overhead could vary with different TLSCompressed.length values. Each\n AEAD cipher MUST NOT produce an expansion of greater than 1024 bytes.\n Symbolically,\n\n AEADEncrypted = AEAD-Encrypt(write_key, nonce, plaintext,\n additional_data)\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, the \"additional_data\", and the AEADEncrypted value. The\n output is either the plaintext or an error indicating that the\n decryption failed. There is no separate integrity check. That is:\n\n TLSCompressed.fragment = AEAD-Decrypt(write_key, nonce,\n AEADEncrypted,\n additional_data)\n\n If the decryption fails, a fatal bad_record_mac alert MUST be\n generated.\n\n6.3. Key Calculation\n\n The Record Protocol requires an algorithm to generate keys required\n by the current connection state (see Appendix A.6) from the security\n parameters provided by the handshake protocol.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 25]\n\f\nRFC 5246 TLS August 2008\n\n\n The master secret is expanded into a sequence of secure bytes, which\n is then split to a client write MAC key, a server write MAC key, a\n client write encryption key, and a server write encryption key. Each\n of these is generated from the byte sequence in that order. Unused\n values are empty. Some AEAD ciphers may additionally require a\n client write IV and a server write IV (see Section 6.2.3.3).\n\n When keys and MAC keys are generated, the master secret is used as an\n entropy source.\n\n To generate the key material, compute\n\n key_block = PRF(SecurityParameters.master_secret,\n \"key expansion\",\n SecurityParameters.server_random +\n SecurityParameters.client_random);\n\n until enough output has been generated. Then, the key_block is\n partitioned as follows:\n\n client_write_MAC_key[SecurityParameters.mac_key_length]\n server_write_MAC_key[SecurityParameters.mac_key_length]\n client_write_key[SecurityParameters.enc_key_length]\n server_write_key[SecurityParameters.enc_key_length]\n client_write_IV[SecurityParameters.fixed_iv_length]\n server_write_IV[SecurityParameters.fixed_iv_length]\n\n Currently, the client_write_IV and server_write_IV are only generated\n for implicit nonce techniques as described in Section 3.2.1 of\n [AEAD].\n\n Implementation note: The currently defined cipher suite which\n requires the most material is AES_256_CBC_SHA256. It requires 2 x 32\n byte keys and 2 x 32 byte MAC keys, for a total 128 bytes of key\n material.\n\n7. The TLS Handshaking Protocols\n\n TLS has three subprotocols that are used to allow peers to agree upon\n security parameters for the record layer, to authenticate themselves,\n to instantiate negotiated security parameters, and to report error\n conditions to each other.\n\n The Handshake Protocol is responsible for negotiating a session,\n which consists of the following items:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 26]\n\f\nRFC 5246 TLS August 2008\n\n\n session identifier\n An arbitrary byte sequence chosen by the server to identify an\n active or resumable session state.\n\n peer certificate\n X509v3 [PKIX] certificate of the peer. This element of the state\n may be null.\n\n compression method\n The algorithm used to compress data prior to encryption.\n\n cipher spec\n Specifies the pseudorandom function (PRF) used to generate keying\n material, the bulk data encryption algorithm (such as null, AES,\n etc.) and the MAC algorithm (such as HMAC-SHA1). It also defines\n cryptographic attributes such as the mac_length. (See Appendix\n A.6 for formal definition.)\n\n master secret\n 48-byte secret shared between the client and server.\n\n is resumable\n A flag indicating whether the session can be used to initiate new\n connections.\n\n These items are then used to create security parameters for use by\n the record layer when protecting application data. Many connections\n can be instantiated using the same session through the resumption\n feature of the TLS Handshake Protocol.\n\n7.1. Change Cipher Spec Protocol\n\n The change cipher spec protocol exists to signal transitions in\n ciphering strategies. The protocol consists of a single message,\n which is encrypted and compressed under the current (not the pending)\n connection state. The message consists of a single byte of value 1.\n\n struct {\n enum { change_cipher_spec(1), (255) } type;\n } ChangeCipherSpec;\n\n The ChangeCipherSpec message is sent by both the client and the\n server to notify the receiving party that subsequent records will be\n protected under the newly negotiated CipherSpec and keys. Reception\n of this message causes the receiver to instruct the record layer to\n immediately copy the read pending state into the read current state.\n Immediately after sending this message, the sender MUST instruct the\n record layer to make the write pending state the write active state.\n\n\n\nDierks & Rescorla Standards Track [Page 27]\n\f\nRFC 5246 TLS August 2008\n\n\n (See Section 6.1.) The ChangeCipherSpec message is sent during the\n handshake after the security parameters have been agreed upon, but\n before the verifying Finished message is sent.\n\n Note: If a rehandshake occurs while data is flowing on a connection,\n the communicating parties may continue to send data using the old\n CipherSpec. However, once the ChangeCipherSpec has been sent, the\n new CipherSpec MUST be used. The first side to send the\n ChangeCipherSpec does not know that the other side has finished\n computing the new keying material (e.g., if it has to perform a\n time-consuming public key operation). Thus, a small window of time,\n during which the recipient must buffer the data, MAY exist. In\n practice, with modern machines this interval is likely to be fairly\n short.\n\n7.2. Alert Protocol\n\n One of the content types supported by the TLS record layer is the\n alert type. Alert messages convey the severity of the message\n (warning or fatal) and a description of the alert. Alert messages\n with a level of fatal result in the immediate termination of the\n connection. In this case, other connections corresponding to the\n session may continue, but the session identifier MUST be invalidated,\n preventing the failed session from being used to establish new\n connections. Like other messages, alert messages are encrypted and\n compressed, as specified by the current connection state.\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n\n\n\nDierks & Rescorla Standards Track [Page 28]\n\f\nRFC 5246 TLS August 2008\n\n\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n7.2.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack. Either party may\n initiate the exchange of closing messages.\n\n close_notify\n This message notifies the recipient that the sender will not send\n any more messages on this connection. Note that as of TLS 1.1,\n failure to properly close a connection no longer requires that a\n session not be resumed. This is a change from TLS 1.0 to conform\n with widespread implementation practice.\n\n Either party may initiate a close by sending a close_notify alert.\n Any data received after a closure alert is ignored.\n\n Unless some other fatal alert has been transmitted, each party is\n required to send a close_notify alert before closing the write side\n of the connection. The other party MUST respond with a close_notify\n alert of its own and close down the connection immediately,\n discarding any pending writes. It is not required for the initiator\n of the close to wait for the responding close_notify alert before\n closing the read side of the connection.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation must receive the responding\n close_notify alert before indicating to the application layer that\n the TLS connection has ended. If the application protocol will not\n transfer any additional data, but will only close the underlying\n transport connection, then the implementation MAY choose to close the\n transport without waiting for the responding close_notify. No part\n\n\n\n\nDierks & Rescorla Standards Track [Page 29]\n\f\nRFC 5246 TLS August 2008\n\n\n of this standard should be taken to dictate the manner in which a\n usage profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing a connection reliably delivers\n pending data before destroying the transport.\n\n7.2.2. Error Alerts\n\n Error handling in the TLS Handshake protocol is very simple. When an\n error is detected, the detecting party sends a message to the other\n party. Upon transmission or receipt of a fatal alert message, both\n parties immediately close the connection. Servers and clients MUST\n forget any session-identifiers, keys, and secrets associated with a\n failed connection. Thus, any connection terminated with a fatal\n alert MUST NOT be resumed.\n\n Whenever an implementation encounters a condition which is defined as\n a fatal alert, it MUST send the appropriate alert prior to closing\n the connection. For all errors where an alert level is not\n explicitly specified, the sending party MAY determine at its\n discretion whether to treat this as a fatal error or not. If the\n implementation chooses to send an alert but intends to close the\n connection immediately afterwards, it MUST send that alert at the\n fatal alert level.\n\n If an alert with a level of warning is sent and received, generally\n the connection can continue normally. If the receiving party decides\n not to proceed with the connection (e.g., after having received a\n no_renegotiation alert that it is not willing to accept), it SHOULD\n send a fatal alert to terminate the connection. Given this, the\n sending party cannot, in general, know how the receiving party will\n behave. Therefore, warning alerts are not very useful when the\n sending party wants to continue the connection, and thus are\n sometimes omitted. For example, if a peer decides to accept an\n expired certificate (perhaps after confirming this with the user) and\n wants to continue the connection, it would not generally send a\n certificate_expired alert.\n\n The following error alerts are defined:\n\n unexpected_message\n An inappropriate message was received. This alert is always fatal\n and should never be observed in communication between proper\n implementations.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 30]\n\f\nRFC 5246 TLS August 2008\n\n\n bad_record_mac\n This alert is returned if a record is received with an incorrect\n MAC. This alert also MUST be returned if an alert is sent because\n a TLSCiphertext decrypted in an invalid way: either it wasn't an\n even multiple of the block length, or its padding values, when\n checked, weren't correct. This message is always fatal and should\n never be observed in communication between proper implementations\n (except when messages were corrupted in the network).\n\n decryption_failed_RESERVED\n This alert was used in some earlier versions of TLS, and may have\n permitted certain attacks against the CBC mode [CBCATT]. It MUST\n NOT be sent by compliant implementations.\n\n record_overflow\n A TLSCiphertext record was received that had a length more than\n 2^14+2048 bytes, or a record decrypted to a TLSCompressed record\n with more than 2^14+1024 bytes. This message is always fatal and\n should never be observed in communication between proper\n implementations (except when messages were corrupted in the\n network).\n\n decompression_failure\n The decompression function received improper input (e.g., data\n that would expand to excessive length). This message is always\n fatal and should never be observed in communication between proper\n implementations.\n\n handshake_failure\n Reception of a handshake_failure alert message indicates that the\n sender was unable to negotiate an acceptable set of security\n parameters given the options available. This is a fatal error.\n\n no_certificate_RESERVED\n This alert was used in SSLv3 but not any version of TLS. It MUST\n NOT be sent by compliant implementations.\n\n bad_certificate\n A certificate was corrupt, contained signatures that did not\n verify correctly, etc.\n\n unsupported_certificate\n A certificate was of an unsupported type.\n\n certificate_revoked\n A certificate was revoked by its signer.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 31]\n\f\nRFC 5246 TLS August 2008\n\n\n certificate_expired\n A certificate has expired or is not currently valid.\n\n certificate_unknown\n Some other (unspecified) issue arose in processing the\n certificate, rendering it unacceptable.\n\n illegal_parameter\n A field in the handshake was out of range or inconsistent with\n other fields. This message is always fatal.\n\n unknown_ca\n A valid certificate chain or partial chain was received, but the\n certificate was not accepted because the CA certificate could not\n be located or couldn't be matched with a known, trusted CA. This\n message is always fatal.\n\n access_denied\n A valid certificate was received, but when access control was\n applied, the sender decided not to proceed with negotiation. This\n message is always fatal.\n\n decode_error\n A message could not be decoded because some field was out of the\n specified range or the length of the message was incorrect. This\n message is always fatal and should never be observed in\n communication between proper implementations (except when messages\n were corrupted in the network).\n\n decrypt_error\n A handshake cryptographic operation failed, including being unable\n to correctly verify a signature or validate a Finished message.\n This message is always fatal.\n\n export_restriction_RESERVED\n This alert was used in some earlier versions of TLS. It MUST NOT\n be sent by compliant implementations.\n\n protocol_version\n The protocol version the client has attempted to negotiate is\n recognized but not supported. (For example, old protocol versions\n might be avoided for security reasons.) This message is always\n fatal.\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 32]\n\f\nRFC 5246 TLS August 2008\n\n\n insufficient_security\n Returned instead of handshake_failure when a negotiation has\n failed specifically because the server requires ciphers more\n secure than those supported by the client. This message is always\n fatal.\n\n internal_error\n An internal error unrelated to the peer or the correctness of the\n protocol (such as a memory allocation failure) makes it impossible\n to continue. This message is always fatal.\n\n user_canceled\n This handshake is being canceled for some reason unrelated to a\n protocol failure. If the user cancels an operation after the\n handshake is complete, just closing the connection by sending a\n close_notify is more appropriate. This alert should be followed\n by a close_notify. This message is generally a warning.\n\n no_renegotiation\n Sent by the client in response to a hello request or by the server\n in response to a client hello after initial handshaking. Either\n of these would normally lead to renegotiation; when that is not\n appropriate, the recipient should respond with this alert. At\n that point, the original requester can decide whether to proceed\n with the connection. One case where this would be appropriate is\n where a server has spawned a process to satisfy a request; the\n process might receive security parameters (key length,\n authentication, etc.) at startup, and it might be difficult to\n communicate changes to these parameters after that point. This\n message is always a warning.\n\n unsupported_extension\n sent by clients that receive an extended server hello containing\n an extension that they did not put in the corresponding client\n hello. This message is always fatal.\n\n New Alert values are assigned by IANA as described in Section 12.\n\n7.3. Handshake Protocol Overview\n\n The cryptographic parameters of the session state are produced by the\n TLS Handshake Protocol, which operates on top of the TLS record\n layer. When a TLS client and server first start communicating, they\n agree on a protocol version, select cryptographic algorithms,\n optionally authenticate each other, and use public-key encryption\n techniques to generate shared secrets.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 33]\n\f\nRFC 5246 TLS August 2008\n\n\n The TLS Handshake Protocol involves the following steps:\n\n - Exchange hello messages to agree on algorithms, exchange random\n values, and check for session resumption.\n\n - Exchange the necessary cryptographic parameters to allow the\n client and server to agree on a premaster secret.\n\n - Exchange certificates and cryptographic information to allow the\n client and server to authenticate themselves.\n\n - Generate a master secret from the premaster secret and exchanged\n random values.\n\n - Provide security parameters to the record layer.\n\n - Allow the client and server to verify that their peer has\n calculated the same security parameters and that the handshake\n occurred without tampering by an attacker.\n\n Note that higher layers should not be overly reliant on whether TLS\n always negotiates the strongest possible connection between two\n peers. There are a number of ways in which a man-in-the-middle\n attacker can attempt to make two entities drop down to the least\n secure method they support. The protocol has been designed to\n minimize this risk, but there are still attacks available: for\n example, an attacker could block access to the port a secure service\n runs on, or attempt to get the peers to negotiate an unauthenticated\n connection. The fundamental rule is that higher levels must be\n cognizant of what their security requirements are and never transmit\n information over a channel less secure than what they require. The\n TLS protocol is secure in that any cipher suite offers its promised\n level of security: if you negotiate 3DES with a 1024-bit RSA key\n exchange with a host whose certificate you have verified, you can\n expect to be that secure.\n\n These goals are achieved by the handshake protocol, which can be\n summarized as follows: The client sends a ClientHello message to\n which the server must respond with a ServerHello message, or else a\n fatal error will occur and the connection will fail. The ClientHello\n and ServerHello are used to establish security enhancement\n capabilities between client and server. The ClientHello and\n ServerHello establish the following attributes: Protocol Version,\n Session ID, Cipher Suite, and Compression Method. Additionally, two\n random values are generated and exchanged: ClientHello.random and\n ServerHello.random.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 34]\n\f\nRFC 5246 TLS August 2008\n\n\n The actual key exchange uses up to four messages: the server\n Certificate, the ServerKeyExchange, the client Certificate, and the\n ClientKeyExchange. New key exchange methods can be created by\n specifying a format for these messages and by defining the use of the\n messages to allow the client and server to agree upon a shared\n secret. This secret MUST be quite long; currently defined key\n exchange methods exchange secrets that range from 46 bytes upwards.\n\n Following the hello messages, the server will send its certificate in\n a Certificate message if it is to be authenticated. Additionally, a\n ServerKeyExchange message may be sent, if it is required (e.g., if\n the server has no certificate, or if its certificate is for signing\n only). If the server is authenticated, it may request a certificate\n from the client, if that is appropriate to the cipher suite selected.\n Next, the server will send the ServerHelloDone message, indicating\n that the hello-message phase of the handshake is complete. The\n server will then wait for a client response. If the server has sent\n a CertificateRequest message, the client MUST send the Certificate\n message. The ClientKeyExchange message is now sent, and the content\n of that message will depend on the public key algorithm selected\n between the ClientHello and the ServerHello. If the client has sent\n a certificate with signing ability, a digitally-signed\n CertificateVerify message is sent to explicitly verify possession of\n the private key in the certificate.\n\n At this point, a ChangeCipherSpec message is sent by the client, and\n the client copies the pending Cipher Spec into the current Cipher\n Spec. The client then immediately sends the Finished message under\n the new algorithms, keys, and secrets. In response, the server will\n send its own ChangeCipherSpec message, transfer the pending to the\n current Cipher Spec, and send its Finished message under the new\n Cipher Spec. At this point, the handshake is complete, and the\n client and server may begin to exchange application layer data. (See\n flow chart below.) Application data MUST NOT be sent prior to the\n completion of the first handshake (before a cipher suite other than\n TLS_NULL_WITH_NULL_NULL is established).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 35]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1. Message flow for a full handshake\n\n * Indicates optional or situation-dependent messages that are not\n always sent.\n\n Note: To help avoid pipeline stalls, ChangeCipherSpec is an\n independent TLS protocol content type, and is not actually a TLS\n handshake message.\n\n When the client and server decide to resume a previous session or\n duplicate an existing session (instead of negotiating new security\n parameters), the message flow is as follows:\n\n The client sends a ClientHello using the Session ID of the session to\n be resumed. The server then checks its session cache for a match.\n If a match is found, and the server is willing to re-establish the\n connection under the specified session state, it will send a\n ServerHello with the same Session ID value. At this point, both\n client and server MUST send ChangeCipherSpec messages and proceed\n directly to Finished messages. Once the re-establishment is\n complete, the client and server MAY begin to exchange application\n layer data. (See flow chart below.) If a Session ID match is not\n found, the server generates a new session ID, and the TLS client and\n server perform a full handshake.\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 36]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2. Message flow for an abbreviated handshake\n\n The contents and significance of each message will be presented in\n detail in the following sections.\n\n7.4. Handshake Protocol\n\n The TLS Handshake Protocol is one of the defined higher-level clients\n of the TLS Record Protocol. This protocol is used to negotiate the\n secure attributes of a session. Handshake messages are supplied to\n the TLS record layer, where they are encapsulated within one or more\n TLSPlaintext structures, which are processed and transmitted as\n specified by the current active session state.\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n } body;\n } Handshake;\n\n\n\nDierks & Rescorla Standards Track [Page 37]\n\f\nRFC 5246 TLS August 2008\n\n\n The handshake protocol messages are presented below in the order they\n MUST be sent; sending handshake messages in an unexpected order\n results in a fatal error. Unneeded handshake messages can be\n omitted, however. Note one exception to the ordering: the\n Certificate message is used twice in the handshake (from server to\n client, then from client to server), but described only in its first\n position. The one message that is not bound by these ordering rules\n is the HelloRequest message, which can be sent at any time, but which\n SHOULD be ignored by the client if it arrives in the middle of a\n handshake.\n\n New handshake message types are assigned by IANA as described in\n Section 12.\n\n7.4.1. Hello Messages\n\n The hello phase messages are used to exchange security enhancement\n capabilities between the client and server. When a new session\n begins, the record layer's connection state encryption, hash, and\n compression algorithms are initialized to null. The current\n connection state is used for renegotiation messages.\n\n7.4.1.1. Hello Request\n\n When this message will be sent:\n\n The HelloRequest message MAY be sent by the server at any time.\n\n Meaning of this message:\n\n HelloRequest is a simple notification that the client should begin\n the negotiation process anew. In response, the client should send\n a ClientHello message when convenient. This message is not\n intended to establish which side is the client or server but\n merely to initiate a new negotiation. Servers SHOULD NOT send a\n HelloRequest immediately upon the client's initial connection. It\n is the client's job to send a ClientHello at that time.\n\n This message will be ignored by the client if the client is\n currently negotiating a session. This message MAY be ignored by\n the client if it does not wish to renegotiate a session, or the\n client may, if it wishes, respond with a no_renegotiation alert.\n Since handshake messages are intended to have transmission\n precedence over application data, it is expected that the\n negotiation will begin before no more than a few records are\n received from the client. If the server sends a HelloRequest but\n does not receive a ClientHello in response, it may close the\n connection with a fatal alert.\n\n\n\nDierks & Rescorla Standards Track [Page 38]\n\f\nRFC 5246 TLS August 2008\n\n\n After sending a HelloRequest, servers SHOULD NOT repeat the\n request until the subsequent handshake negotiation is complete.\n\n Structure of this message:\n\n struct { } HelloRequest;\n\n This message MUST NOT be included in the message hashes that are\n maintained throughout the handshake and used in the Finished messages\n and the certificate verify message.\n\n7.4.1.2. Client Hello\n\n When this message will be sent:\n\n When a client first connects to a server, it is required to send\n the ClientHello as its first message. The client can also send a\n ClientHello in response to a HelloRequest or on its own initiative\n in order to renegotiate the security parameters in an existing\n connection.\n\n Structure of this message:\n\n The ClientHello message includes a random structure, which is used\n later in the protocol.\n\n struct {\n uint32 gmt_unix_time;\n opaque random_bytes[28];\n } Random;\n\n gmt_unix_time\n The current time and date in standard UNIX 32-bit format\n (seconds since the midnight starting Jan 1, 1970, UTC, ignoring\n leap seconds) according to the sender's internal clock. Clocks\n are not required to be set correctly by the basic TLS protocol;\n higher-level or application protocols may define additional\n requirements. Note that, for historical reasons, the data\n element is named using GMT, the predecessor of the current\n worldwide time base, UTC.\n\n random_bytes\n 28 bytes generated by a secure random number generator.\n\n The ClientHello message includes a variable-length session\n identifier. If not empty, the value identifies a session between the\n same client and server whose security parameters the client wishes to\n reuse. The session identifier MAY be from an earlier connection,\n\n\n\nDierks & Rescorla Standards Track [Page 39]\n\f\nRFC 5246 TLS August 2008\n\n\n this connection, or from another currently active connection. The\n second option is useful if the client only wishes to update the\n random structures and derived values of a connection, and the third\n option makes it possible to establish several independent secure\n connections without repeating the full handshake protocol. These\n independent connections may occur sequentially or simultaneously; a\n SessionID becomes valid when the handshake negotiating it completes\n with the exchange of Finished messages and persists until it is\n removed due to aging or because a fatal error was encountered on a\n connection associated with the session. The actual contents of the\n SessionID are defined by the server.\n\n opaque SessionID<0..32>;\n\n Warning: Because the SessionID is transmitted without encryption or\n immediate MAC protection, servers MUST NOT place confidential\n information in session identifiers or let the contents of fake\n session identifiers cause any breach of security. (Note that the\n content of the handshake as a whole, including the SessionID, is\n protected by the Finished messages exchanged at the end of the\n handshake.)\n\n The cipher suite list, passed from the client to the server in the\n ClientHello message, contains the combinations of cryptographic\n algorithms supported by the client in order of the client's\n preference (favorite choice first). Each cipher suite defines a key\n exchange algorithm, a bulk encryption algorithm (including secret key\n length), a MAC algorithm, and a PRF. The server will select a cipher\n suite or, if no acceptable choices are presented, return a handshake\n failure alert and close the connection. If the list contains cipher\n suites the server does not recognize, support, or wish to use, the\n server MUST ignore those cipher suites, and process the remaining\n ones as usual.\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n The ClientHello includes a list of compression algorithms supported\n by the client, ordered according to the client's preference.\n\n enum { null(0), (255) } CompressionMethod;\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 40]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ProtocolVersion client_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suites<2..2^16-2>;\n CompressionMethod compression_methods<1..2^8-1>;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ClientHello;\n\n TLS allows extensions to follow the compression_methods field in an\n extensions block. The presence of extensions can be detected by\n determining whether there are bytes following the compression_methods\n at the end of the ClientHello. Note that this method of detecting\n optional data differs from the normal TLS method of having a\n variable-length field, but it is used for compatibility with TLS\n before extensions were defined.\n\n client_version\n The version of the TLS protocol by which the client wishes to\n communicate during this session. This SHOULD be the latest\n (highest valued) version supported by the client. For this\n version of the specification, the version will be 3.3 (see\n Appendix E for details about backward compatibility).\n\n random\n A client-generated random structure.\n\n session_id\n The ID of a session the client wishes to use for this connection.\n This field is empty if no session_id is available, or if the\n client wishes to generate new security parameters.\n\n cipher_suites\n This is a list of the cryptographic options supported by the\n client, with the client's first preference first. If the\n session_id field is not empty (implying a session resumption\n request), this vector MUST include at least the cipher_suite from\n that session. Values are defined in Appendix A.5.\n\n compression_methods\n This is a list of the compression methods supported by the client,\n sorted by client preference. If the session_id field is not empty\n (implying a session resumption request), it MUST include the\n\n\n\nDierks & Rescorla Standards Track [Page 41]\n\f\nRFC 5246 TLS August 2008\n\n\n compression_method from that session. This vector MUST contain,\n and all implementations MUST support, CompressionMethod.null.\n Thus, a client and server will always be able to agree on a\n compression method.\n\n extensions\n Clients MAY request extended functionality from servers by sending\n data in the extensions field. The actual \"Extension\" format is\n defined in Section 7.4.1.4.\n\n In the event that a client requests additional functionality using\n extensions, and this functionality is not supplied by the server, the\n client MAY abort the handshake. A server MUST accept ClientHello\n messages both with and without the extensions field, and (as for all\n other messages) it MUST check that the amount of data in the message\n precisely matches one of these formats; if not, then it MUST send a\n fatal \"decode_error\" alert.\n\n After sending the ClientHello message, the client waits for a\n ServerHello message. Any handshake message returned by the server,\n except for a HelloRequest, is treated as a fatal error.\n\n7.4.1.3. Server Hello\n\n When this message will be sent:\n\n The server will send this message in response to a ClientHello\n message when it was able to find an acceptable set of algorithms.\n If it cannot find such a match, it will respond with a handshake\n failure alert.\n\n Structure of this message:\n\n struct {\n ProtocolVersion server_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ServerHello;\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 42]\n\f\nRFC 5246 TLS August 2008\n\n\n The presence of extensions can be detected by determining whether\n there are bytes following the compression_method field at the end of\n the ServerHello.\n\n server_version\n This field will contain the lower of that suggested by the client\n in the client hello and the highest supported by the server. For\n this version of the specification, the version is 3.3. (See\n Appendix E for details about backward compatibility.)\n\n random\n This structure is generated by the server and MUST be\n independently generated from the ClientHello.random.\n\n session_id\n This is the identity of the session corresponding to this\n connection. If the ClientHello.session_id was non-empty, the\n server will look in its session cache for a match. If a match is\n found and the server is willing to establish the new connection\n using the specified session state, the server will respond with\n the same value as was supplied by the client. This indicates a\n resumed session and dictates that the parties must proceed\n directly to the Finished messages. Otherwise, this field will\n contain a different value identifying the new session. The server\n may return an empty session_id to indicate that the session will\n not be cached and therefore cannot be resumed. If a session is\n resumed, it must be resumed using the same cipher suite it was\n originally negotiated with. Note that there is no requirement\n that the server resume any session even if it had formerly\n provided a session_id. Clients MUST be prepared to do a full\n negotiation -- including negotiating new cipher suites -- during\n any handshake.\n\n cipher_suite\n The single cipher suite selected by the server from the list in\n ClientHello.cipher_suites. For resumed sessions, this field is\n the value from the state of the session being resumed.\n\n compression_method\n The single compression algorithm selected by the server from the\n list in ClientHello.compression_methods. For resumed sessions,\n this field is the value from the resumed session state.\n\n extensions\n A list of extensions. Note that only extensions offered by the\n client can appear in the server's list.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 43]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.1.4. Hello Extensions\n\n The extension format is:\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n signature_algorithms(13), (65535)\n } ExtensionType;\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The initial set of extensions is defined in a companion document\n [TLSEXT]. The list of extension types is maintained by IANA as\n described in Section 12.\n\n An extension type MUST NOT appear in the ServerHello unless the same\n extension type appeared in the corresponding ClientHello. If a\n client receives an extension type in ServerHello that it did not\n request in the associated ClientHello, it MUST abort the handshake\n with an unsupported_extension fatal alert.\n\n Nonetheless, \"server-oriented\" extensions may be provided in the\n future within this framework. Such an extension (say, of type x)\n would require the client to first send an extension of type x in a\n ClientHello with empty extension_data to indicate that it supports\n the extension type. In this case, the client is offering the\n capability to understand the extension type, and the server is taking\n the client up on its offer.\n\n When multiple extensions of different types are present in the\n ClientHello or ServerHello messages, the extensions MAY appear in any\n order. There MUST NOT be more than one extension of the same type.\n\n Finally, note that extensions can be sent both when starting a new\n session and when requesting session resumption. Indeed, a client\n that requests session resumption does not in general know whether the\n server will accept this request, and therefore it SHOULD send the\n same extensions as it would send if it were not attempting\n resumption.\n\n\n\nDierks & Rescorla Standards Track [Page 44]\n\f\nRFC 5246 TLS August 2008\n\n\n In general, the specification of each extension type needs to\n describe the effect of the extension both during full handshake and\n session resumption. Most current TLS extensions are relevant only\n when a session is initiated: when an older session is resumed, the\n server does not process these extensions in Client Hello, and does\n not include them in Server Hello. However, some extensions may\n specify different behavior during session resumption.\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions, and some are simply refusals to support particular\n features. In general, error alerts should be used for the former,\n and a field in the server extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem.\n\n Often the fact that the extension fields are included in the\n inputs to the Finished message hashes will be sufficient, but\n extreme care is needed when the extension changes the meaning of\n messages sent in the handshake phase. Designers and implementors\n should be aware of the fact that until the handshake has been\n authenticated, active attackers can modify messages and insert,\n remove, or replace extensions.\n\n - It would be technically possible to use extensions to change major\n aspects of the design of TLS; for example the design of cipher\n suite negotiation. This is not recommended; it would be more\n appropriate to define a new version of TLS -- particularly since\n the TLS handshake algorithms have specific protection against\n version rollback attacks based on the version number, and the\n possibility of version rollback should be a significant\n consideration in any major design change.\n\n7.4.1.4.1. Signature Algorithms\n\n The client uses the \"signature_algorithms\" extension to indicate to\n the server which signature/hash algorithm pairs may be used in\n digital signatures. The \"extension_data\" field of this extension\n contains a \"supported_signature_algorithms\" value.\n\n\n\nDierks & Rescorla Standards Track [Page 45]\n\f\nRFC 5246 TLS August 2008\n\n\n enum {\n none(0), md5(1), sha1(2), sha224(3), sha256(4), sha384(5),\n sha512(6), (255)\n } HashAlgorithm;\n\n enum { anonymous(0), rsa(1), dsa(2), ecdsa(3), (255) }\n SignatureAlgorithm;\n\n struct {\n HashAlgorithm hash;\n SignatureAlgorithm signature;\n } SignatureAndHashAlgorithm;\n\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2..2^16-2>;\n\n Each SignatureAndHashAlgorithm value lists a single hash/signature\n pair that the client is willing to verify. The values are indicated\n in descending order of preference.\n\n Note: Because not all signature algorithms and hash algorithms may be\n accepted by an implementation (e.g., DSA with SHA-1, but not\n SHA-256), algorithms here are listed in pairs.\n\n hash\n This field indicates the hash algorithm which may be used. The\n values indicate support for unhashed data, MD5 [MD5], SHA-1,\n SHA-224, SHA-256, SHA-384, and SHA-512 [SHS], respectively. The\n \"none\" value is provided for future extensibility, in case of a\n signature algorithm which does not require hashing before signing.\n\n signature\n This field indicates the signature algorithm that may be used.\n The values indicate anonymous signatures, RSASSA-PKCS1-v1_5\n [PKCS1] and DSA [DSS], and ECDSA [ECDSA], respectively. The\n \"anonymous\" value is meaningless in this context but used in\n Section 7.4.3. It MUST NOT appear in this extension.\n\n The semantics of this extension are somewhat complicated because the\n cipher suite indicates permissible signature algorithms but not hash\n algorithms. Sections 7.4.2 and 7.4.3 describe the appropriate rules.\n\n If the client supports only the default hash and signature algorithms\n (listed in this section), it MAY omit the signature_algorithms\n extension. If the client does not support the default algorithms, or\n supports other hash and signature algorithms (and it is willing to\n use them for verifying messages sent by the server, i.e., server\n certificates and server key exchange), it MUST send the\n\n\n\nDierks & Rescorla Standards Track [Page 46]\n\f\nRFC 5246 TLS August 2008\n\n\n signature_algorithms extension, listing the algorithms it is willing\n to accept.\n\n If the client does not send the signature_algorithms extension, the\n server MUST do the following:\n\n - If the negotiated key exchange algorithm is one of (RSA, DHE_RSA,\n DH_RSA, RSA_PSK, ECDH_RSA, ECDHE_RSA), behave as if client had\n sent the value {sha1,rsa}.\n\n - If the negotiated key exchange algorithm is one of (DHE_DSS,\n DH_DSS), behave as if the client had sent the value {sha1,dsa}.\n\n - If the negotiated key exchange algorithm is one of (ECDH_ECDSA,\n ECDHE_ECDSA), behave as if the client had sent value {sha1,ecdsa}.\n\n Note: this is a change from TLS 1.1 where there are no explicit\n rules, but as a practical matter one can assume that the peer\n supports MD5 and SHA-1.\n\n Note: this extension is not meaningful for TLS versions prior to 1.2.\n Clients MUST NOT offer it if they are offering prior versions.\n However, even if clients do offer it, the rules specified in [TLSEXT]\n require servers to ignore extensions they do not understand.\n\n Servers MUST NOT send this extension. TLS servers MUST support\n receiving this extension.\n\n When performing session resumption, this extension is not included in\n Server Hello, and the server ignores the extension in Client Hello\n (if present).\n\n7.4.2. Server Certificate\n\n When this message will be sent:\n\n The server MUST send a Certificate message whenever the agreed-\n upon key exchange method uses certificates for authentication\n (this includes all key exchange methods defined in this document\n except DH_anon). This message will always immediately follow the\n ServerHello message.\n\n Meaning of this message:\n\n This message conveys the server's certificate chain to the client.\n\n The certificate MUST be appropriate for the negotiated cipher\n suite's key exchange algorithm and any negotiated extensions.\n\n\n\nDierks & Rescorla Standards Track [Page 47]\n\f\nRFC 5246 TLS August 2008\n\n\n Structure of this message:\n\n opaque ASN.1Cert<1..2^24-1>;\n\n struct {\n ASN.1Cert certificate_list<0..2^24-1>;\n } Certificate;\n\n certificate_list\n This is a sequence (chain) of certificates. The sender's\n certificate MUST come first in the list. Each following\n certificate MUST directly certify the one preceding it. Because\n certificate validation requires that root keys be distributed\n independently, the self-signed certificate that specifies the root\n certificate authority MAY be omitted from the chain, under the\n assumption that the remote end must already possess it in order to\n validate it in any case.\n\n The same message type and structure will be used for the client's\n response to a certificate request message. Note that a client MAY\n send no certificates if it does not have an appropriate certificate\n to send in response to the server's authentication request.\n\n Note: PKCS #7 [PKCS7] is not used as the format for the certificate\n vector because PKCS #6 [PKCS6] extended certificates are not used.\n Also, PKCS #7 defines a SET rather than a SEQUENCE, making the task\n of parsing the list more difficult.\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3, unless explicitly negotiated\n otherwise (e.g., [TLSPGP]).\n\n - The end entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected key exchange\n algorithm.\n\n Key Exchange Alg. Certificate Key Type\n\n RSA RSA public key; the certificate MUST allow the\n RSA_PSK key to be used for encryption (the\n keyEncipherment bit MUST be set if the key\n usage extension is present).\n Note: RSA_PSK is defined in [TLSPSK].\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 48]\n\f\nRFC 5246 TLS August 2008\n\n\n DHE_RSA RSA public key; the certificate MUST allow the\n ECDHE_RSA key to be used for signing (the\n digitalSignature bit MUST be set if the key\n usage extension is present) with the signature\n scheme and hash algorithm that will be employed\n in the server key exchange message.\n Note: ECDHE_RSA is defined in [TLSECC].\n\n DHE_DSS DSA public key; the certificate MUST allow the\n key to be used for signing with the hash\n algorithm that will be employed in the server\n key exchange message.\n\n DH_DSS Diffie-Hellman public key; the keyAgreement bit\n DH_RSA MUST be set if the key usage extension is\n present.\n\n ECDH_ECDSA ECDH-capable public key; the public key MUST\n ECDH_RSA use a curve and point format supported by the\n client, as described in [TLSECC].\n\n ECDHE_ECDSA ECDSA-capable public key; the certificate MUST\n allow the key to be used for signing with the\n hash algorithm that will be employed in the\n server key exchange message. The public key\n MUST use a curve and point format supported by\n the client, as described in [TLSECC].\n\n - The \"server_name\" and \"trusted_ca_keys\" extensions [TLSEXT] are\n used to guide certificate selection.\n\n If the client provided a \"signature_algorithms\" extension, then all\n certificates provided by the server MUST be signed by a\n hash/signature algorithm pair that appears in that extension. Note\n that this implies that a certificate containing a key for one\n signature algorithm MAY be signed using a different signature\n algorithm (for instance, an RSA key signed with a DSA key). This is\n a departure from TLS 1.1, which required that the algorithms be the\n same. Note that this also implies that the DH_DSS, DH_RSA,\n ECDH_ECDSA, and ECDH_RSA key exchange algorithms do not restrict the\n algorithm used to sign the certificate. Fixed DH certificates MAY be\n signed with any hash/signature algorithm pair appearing in the\n extension. The names DH_DSS, DH_RSA, ECDH_ECDSA, and ECDH_RSA are\n historical.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 49]\n\f\nRFC 5246 TLS August 2008\n\n\n If the server has multiple certificates, it chooses one of them based\n on the above-mentioned criteria (in addition to other criteria, such\n as transport layer endpoint, local configuration and preferences,\n etc.). If the server has a single certificate, it SHOULD attempt to\n validate that it meets these criteria.\n\n Note that there are certificates that use algorithms and/or algorithm\n combinations that cannot be currently used with TLS. For example, a\n certificate with RSASSA-PSS signature key (id-RSASSA-PSS OID in\n SubjectPublicKeyInfo) cannot be used because TLS defines no\n corresponding signature algorithm.\n\n As cipher suites that specify new key exchange methods are specified\n for the TLS protocol, they will imply the certificate format and the\n required encoded keying information.\n\n7.4.3. Server Key Exchange Message\n\n When this message will be sent:\n\n This message will be sent immediately after the server Certificate\n message (or the ServerHello message, if this is an anonymous\n negotiation).\n\n The ServerKeyExchange message is sent by the server only when the\n server Certificate message (if sent) does not contain enough data\n to allow the client to exchange a premaster secret. This is true\n for the following key exchange methods:\n\n DHE_DSS\n DHE_RSA\n DH_anon\n\n It is not legal to send the ServerKeyExchange message for the\n following key exchange methods:\n\n RSA\n DH_DSS\n DH_RSA\n\n Other key exchange algorithms, such as those defined in [TLSECC],\n MUST specify whether the ServerKeyExchange message is sent or not;\n and if the message is sent, its contents.\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 50]\n\f\nRFC 5246 TLS August 2008\n\n\n Meaning of this message:\n\n This message conveys cryptographic information to allow the client\n to communicate the premaster secret: a Diffie-Hellman public key\n with which the client can complete a key exchange (with the result\n being the premaster secret) or a public key for some other\n algorithm.\n\n Structure of this message:\n\n enum { dhe_dss, dhe_rsa, dh_anon, rsa, dh_dss, dh_rsa\n /* may be extended, e.g., for ECDH -- see [TLSECC] */\n } KeyExchangeAlgorithm;\n\n struct {\n opaque dh_p<1..2^16-1>;\n opaque dh_g<1..2^16-1>;\n opaque dh_Ys<1..2^16-1>;\n } ServerDHParams; /* Ephemeral DH parameters */\n\n dh_p\n The prime modulus used for the Diffie-Hellman operation.\n\n dh_g\n The generator used for the Diffie-Hellman operation.\n\n dh_Ys\n The server's Diffie-Hellman public value (g^X mod p).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 51]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n select (KeyExchangeAlgorithm) {\n case dh_anon:\n ServerDHParams params;\n case dhe_dss:\n case dhe_rsa:\n ServerDHParams params;\n digitally-signed struct {\n opaque client_random[32];\n opaque server_random[32];\n ServerDHParams params;\n } signed_params;\n case rsa:\n case dh_dss:\n case dh_rsa:\n struct {} ;\n /* message is omitted for rsa, dh_dss, and dh_rsa */\n /* may be extended, e.g., for ECDH -- see [TLSECC] */\n };\n } ServerKeyExchange;\n\n params\n The server's key exchange parameters.\n\n signed_params\n For non-anonymous key exchanges, a signature over the server's\n key exchange parameters.\n\n If the client has offered the \"signature_algorithms\" extension, the\n signature algorithm and hash algorithm MUST be a pair listed in that\n extension. Note that there is a possibility for inconsistencies\n here. For instance, the client might offer DHE_DSS key exchange but\n omit any DSA pairs from its \"signature_algorithms\" extension. In\n order to negotiate correctly, the server MUST check any candidate\n cipher suites against the \"signature_algorithms\" extension before\n selecting them. This is somewhat inelegant but is a compromise\n designed to minimize changes to the original cipher suite design.\n\n In addition, the hash and signature algorithms MUST be compatible\n with the key in the server's end-entity certificate. RSA keys MAY be\n used with any permitted hash algorithm, subject to restrictions in\n the certificate, if any.\n\n Because DSA signatures do not contain any secure indication of hash\n algorithm, there is a risk of hash substitution if multiple hashes\n may be used with any key. Currently, DSA [DSS] may only be used with\n SHA-1. Future revisions of DSS [DSS-3] are expected to allow the use\n of other digest algorithms with DSA, as well as guidance as to which\n\n\n\nDierks & Rescorla Standards Track [Page 52]\n\f\nRFC 5246 TLS August 2008\n\n\n digest algorithms should be used with each key size. In addition,\n future revisions of [PKIX] may specify mechanisms for certificates to\n indicate which digest algorithms are to be used with DSA.\n\n As additional cipher suites are defined for TLS that include new key\n exchange algorithms, the server key exchange message will be sent if\n and only if the certificate type associated with the key exchange\n algorithm does not provide enough information for the client to\n exchange a premaster secret.\n\n7.4.4. Certificate Request\n\n When this message will be sent:\n\n A non-anonymous server can optionally request a certificate from\n the client, if appropriate for the selected cipher suite. This\n message, if sent, will immediately follow the ServerKeyExchange\n message (if it is sent; otherwise, this message follows the\n server's Certificate message).\n\n Structure of this message:\n\n enum {\n rsa_sign(1), dss_sign(2), rsa_fixed_dh(3), dss_fixed_dh(4),\n rsa_ephemeral_dh_RESERVED(5), dss_ephemeral_dh_RESERVED(6),\n fortezza_dms_RESERVED(20), (255)\n } ClientCertificateType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n ClientCertificateType certificate_types<1..2^8-1>;\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2^16-1>;\n DistinguishedName certificate_authorities<0..2^16-1>;\n } CertificateRequest;\n\n certificate_types\n A list of the types of certificate types that the client may\n offer.\n\n rsa_sign a certificate containing an RSA key\n dss_sign a certificate containing a DSA key\n rsa_fixed_dh a certificate containing a static DH key.\n dss_fixed_dh a certificate containing a static DH key\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 53]\n\f\nRFC 5246 TLS August 2008\n\n\n supported_signature_algorithms\n A list of the hash/signature algorithm pairs that the server is\n able to verify, listed in descending order of preference.\n\n certificate_authorities\n A list of the distinguished names [X501] of acceptable\n certificate_authorities, represented in DER-encoded format. These\n distinguished names may specify a desired distinguished name for a\n root CA or for a subordinate CA; thus, this message can be used to\n describe known roots as well as a desired authorization space. If\n the certificate_authorities list is empty, then the client MAY\n send any certificate of the appropriate ClientCertificateType,\n unless there is some external arrangement to the contrary.\n\n The interaction of the certificate_types and\n supported_signature_algorithms fields is somewhat complicated.\n certificate_types has been present in TLS since SSLv3, but was\n somewhat underspecified. Much of its functionality is superseded by\n supported_signature_algorithms. The following rules apply:\n\n - Any certificates provided by the client MUST be signed using a\n hash/signature algorithm pair found in\n supported_signature_algorithms.\n\n - The end-entity certificate provided by the client MUST contain a\n key that is compatible with certificate_types. If the key is a\n signature key, it MUST be usable with some hash/signature\n algorithm pair in supported_signature_algorithms.\n\n - For historical reasons, the names of some client certificate types\n include the algorithm used to sign the certificate. For example,\n in earlier versions of TLS, rsa_fixed_dh meant a certificate\n signed with RSA and containing a static DH key. In TLS 1.2, this\n functionality has been obsoleted by the\n supported_signature_algorithms, and the certificate type no longer\n restricts the algorithm used to sign the certificate. For\n example, if the server sends dss_fixed_dh certificate type and\n {{sha1, dsa}, {sha1, rsa}} signature types, the client MAY reply\n with a certificate containing a static DH key, signed with RSA-\n SHA1.\n\n New ClientCertificateType values are assigned by IANA as described in\n Section 12.\n\n Note: Values listed as RESERVED may not be used. They were used in\n SSLv3.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 54]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: It is a fatal handshake_failure alert for an anonymous server\n to request client authentication.\n\n7.4.5. Server Hello Done\n\n When this message will be sent:\n\n The ServerHelloDone message is sent by the server to indicate the\n end of the ServerHello and associated messages. After sending\n this message, the server will wait for a client response.\n\n Meaning of this message:\n\n This message means that the server is done sending messages to\n support the key exchange, and the client can proceed with its\n phase of the key exchange.\n\n Upon receipt of the ServerHelloDone message, the client SHOULD\n verify that the server provided a valid certificate, if required,\n and check that the server hello parameters are acceptable.\n\n Structure of this message:\n\n struct { } ServerHelloDone;\n\n7.4.6. Client Certificate\n\n When this message will be sent:\n\n This is the first message the client can send after receiving a\n ServerHelloDone message. This message is only sent if the server\n requests a certificate. If no suitable certificate is available,\n the client MUST send a certificate message containing no\n certificates. That is, the certificate_list structure has a\n length of zero. If the client does not send any certificates, the\n server MAY at its discretion either continue the handshake without\n client authentication, or respond with a fatal handshake_failure\n alert. Also, if some aspect of the certificate chain was\n unacceptable (e.g., it was not signed by a known, trusted CA), the\n server MAY at its discretion either continue the handshake\n (considering the client unauthenticated) or send a fatal alert.\n\n Client certificates are sent using the Certificate structure\n defined in Section 7.4.2.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 55]\n\f\nRFC 5246 TLS August 2008\n\n\n Meaning of this message:\n\n This message conveys the client's certificate chain to the server;\n the server will use it when verifying the CertificateVerify\n message (when the client authentication is based on signing) or\n calculating the premaster secret (for non-ephemeral Diffie-\n Hellman). The certificate MUST be appropriate for the negotiated\n cipher suite's key exchange algorithm, and any negotiated\n extensions.\n\n In particular:\n\n - The certificate type MUST be X.509v3, unless explicitly negotiated\n otherwise (e.g., [TLSPGP]).\n\n - The end-entity certificate's public key (and associated\n restrictions) has to be compatible with the certificate types\n listed in CertificateRequest:\n\n Client Cert. Type Certificate Key Type\n\n rsa_sign RSA public key; the certificate MUST allow the\n key to be used for signing with the signature\n scheme and hash algorithm that will be\n employed in the certificate verify message.\n\n dss_sign DSA public key; the certificate MUST allow the\n key to be used for signing with the hash\n algorithm that will be employed in the\n certificate verify message.\n\n ecdsa_sign ECDSA-capable public key; the certificate MUST\n allow the key to be used for signing with the\n hash algorithm that will be employed in the\n certificate verify message; the public key\n MUST use a curve and point format supported by\n the server.\n\n rsa_fixed_dh Diffie-Hellman public key; MUST use the same\n dss_fixed_dh parameters as server's key.\n\n rsa_fixed_ecdh ECDH-capable public key; MUST use the\n ecdsa_fixed_ecdh same curve as the server's key, and MUST use a\n point format supported by the server.\n\n - If the certificate_authorities list in the certificate request\n message was non-empty, one of the certificates in the certificate\n chain SHOULD be issued by one of the listed CAs.\n\n\n\nDierks & Rescorla Standards Track [Page 56]\n\f\nRFC 5246 TLS August 2008\n\n\n - The certificates MUST be signed using an acceptable hash/\n signature algorithm pair, as described in Section 7.4.4. Note\n that this relaxes the constraints on certificate-signing\n algorithms found in prior versions of TLS.\n\n Note that, as with the server certificate, there are certificates\n that use algorithms/algorithm combinations that cannot be currently\n used with TLS.\n\n7.4.7. Client Key Exchange Message\n\n When this message will be sent:\n\n This message is always sent by the client. It MUST immediately\n follow the client certificate message, if it is sent. Otherwise,\n it MUST be the first message sent by the client after it receives\n the ServerHelloDone message.\n\n Meaning of this message:\n\n With this message, the premaster secret is set, either by direct\n transmission of the RSA-encrypted secret or by the transmission of\n Diffie-Hellman parameters that will allow each side to agree upon\n the same premaster secret.\n\n When the client is using an ephemeral Diffie-Hellman exponent,\n then this message contains the client's Diffie-Hellman public\n value. If the client is sending a certificate containing a static\n DH exponent (i.e., it is doing fixed_dh client authentication),\n then this message MUST be sent but MUST be empty.\n\n Structure of this message:\n\n The choice of messages depends on which key exchange method has\n been selected. See Section 7.4.3 for the KeyExchangeAlgorithm\n definition.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 57]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n select (KeyExchangeAlgorithm) {\n case rsa:\n EncryptedPreMasterSecret;\n case dhe_dss:\n case dhe_rsa:\n case dh_dss:\n case dh_rsa:\n case dh_anon:\n ClientDiffieHellmanPublic;\n } exchange_keys;\n } ClientKeyExchange;\n\n7.4.7.1. RSA-Encrypted Premaster Secret Message\n\n Meaning of this message:\n\n If RSA is being used for key agreement and authentication, the\n client generates a 48-byte premaster secret, encrypts it using the\n public key from the server's certificate, and sends the result in\n an encrypted premaster secret message. This structure is a\n variant of the ClientKeyExchange message and is not a message in\n itself.\n\n Structure of this message:\n\n struct {\n ProtocolVersion client_version;\n opaque random[46];\n } PreMasterSecret;\n\n client_version\n The latest (newest) version supported by the client. This is\n used to detect version rollback attacks.\n\n random\n 46 securely-generated random bytes.\n\n struct {\n public-key-encrypted PreMasterSecret pre_master_secret;\n } EncryptedPreMasterSecret;\n\n pre_master_secret\n This random value is generated by the client and is used to\n generate the master secret, as specified in Section 8.1.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 58]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: The version number in the PreMasterSecret is the version\n offered by the client in the ClientHello.client_version, not the\n version negotiated for the connection. This feature is designed to\n prevent rollback attacks. Unfortunately, some old implementations\n use the negotiated version instead, and therefore checking the\n version number may lead to failure to interoperate with such\n incorrect client implementations.\n\n Client implementations MUST always send the correct version number in\n PreMasterSecret. If ClientHello.client_version is TLS 1.1 or higher,\n server implementations MUST check the version number as described in\n the note below. If the version number is TLS 1.0 or earlier, server\n implementations SHOULD check the version number, but MAY have a\n configuration option to disable the check. Note that if the check\n fails, the PreMasterSecret SHOULD be randomized as described below.\n\n Note: Attacks discovered by Bleichenbacher [BLEI] and Klima et al.\n [KPR03] can be used to attack a TLS server that reveals whether a\n particular message, when decrypted, is properly PKCS#1 formatted,\n contains a valid PreMasterSecret structure, or has the correct\n version number.\n\n As described by Klima [KPR03], these vulnerabilities can be avoided\n by treating incorrectly formatted message blocks and/or mismatched\n version numbers in a manner indistinguishable from correctly\n formatted RSA blocks. In other words:\n\n 1. Generate a string R of 46 random bytes\n\n 2. Decrypt the message to recover the plaintext M\n\n 3. If the PKCS#1 padding is not correct, or the length of message\n M is not exactly 48 bytes:\n pre_master_secret = ClientHello.client_version || R\n else If ClientHello.client_version <= TLS 1.0, and version\n number check is explicitly disabled:\n pre_master_secret = M\n else:\n pre_master_secret = ClientHello.client_version || M[2..47]\n\n Note that explicitly constructing the pre_master_secret with the\n ClientHello.client_version produces an invalid master_secret if the\n client has sent the wrong version in the original pre_master_secret.\n\n An alternative approach is to treat a version number mismatch as a\n PKCS-1 formatting error and randomize the premaster secret\n completely:\n\n\n\n\nDierks & Rescorla Standards Track [Page 59]\n\f\nRFC 5246 TLS August 2008\n\n\n 1. Generate a string R of 48 random bytes\n\n 2. Decrypt the message to recover the plaintext M\n\n 3. If the PKCS#1 padding is not correct, or the length of message\n M is not exactly 48 bytes:\n pre_master_secret = R\n else If ClientHello.client_version <= TLS 1.0, and version\n number check is explicitly disabled:\n premaster secret = M\n else If M[0..1] != ClientHello.client_version:\n premaster secret = R\n else:\n premaster secret = M\n\n Although no practical attacks against this construction are known,\n Klima et al. [KPR03] describe some theoretical attacks, and therefore\n the first construction described is RECOMMENDED.\n\n In any case, a TLS server MUST NOT generate an alert if processing an\n RSA-encrypted premaster secret message fails, or the version number\n is not as expected. Instead, it MUST continue the handshake with a\n randomly generated premaster secret. It may be useful to log the\n real cause of failure for troubleshooting purposes; however, care\n must be taken to avoid leaking the information to an attacker\n (through, e.g., timing, log files, or other channels.)\n\n The RSAES-OAEP encryption scheme defined in [PKCS1] is more secure\n against the Bleichenbacher attack. However, for maximal\n compatibility with earlier versions of TLS, this specification uses\n the RSAES-PKCS1-v1_5 scheme. No variants of the Bleichenbacher\n attack are known to exist provided that the above recommendations are\n followed.\n\n Implementation note: Public-key-encrypted data is represented as an\n opaque vector <0..2^16-1> (see Section 4.7). Thus, the RSA-encrypted\n PreMasterSecret in a ClientKeyExchange is preceded by two length\n bytes. These bytes are redundant in the case of RSA because the\n EncryptedPreMasterSecret is the only data in the ClientKeyExchange\n and its length can therefore be unambiguously determined. The SSLv3\n specification was not clear about the encoding of public-key-\n encrypted data, and therefore many SSLv3 implementations do not\n include the length bytes -- they encode the RSA-encrypted data\n directly in the ClientKeyExchange message.\n\n This specification requires correct encoding of the\n EncryptedPreMasterSecret complete with length bytes. The resulting\n PDU is incompatible with many SSLv3 implementations. Implementors\n\n\n\nDierks & Rescorla Standards Track [Page 60]\n\f\nRFC 5246 TLS August 2008\n\n\n upgrading from SSLv3 MUST modify their implementations to generate\n and accept the correct encoding. Implementors who wish to be\n compatible with both SSLv3 and TLS should make their implementation's\n behavior dependent on the protocol version.\n\n Implementation note: It is now known that remote timing-based attacks\n on TLS are possible, at least when the client and server are on the\n same LAN. Accordingly, implementations that use static RSA keys MUST\n use RSA blinding or some other anti-timing technique, as described in\n [TIMING].\n\n7.4.7.2. Client Diffie-Hellman Public Value\n\n Meaning of this message:\n\n This structure conveys the client's Diffie-Hellman public value\n (Yc) if it was not already included in the client's certificate.\n The encoding used for Yc is determined by the enumerated\n PublicValueEncoding. This structure is a variant of the client\n key exchange message, and not a message in itself.\n\n Structure of this message:\n\n enum { implicit, explicit } PublicValueEncoding;\n\n implicit\n If the client has sent a certificate which contains a suitable\n Diffie-Hellman key (for fixed_dh client authentication), then\n Yc is implicit and does not need to be sent again. In this\n case, the client key exchange message will be sent, but it MUST\n be empty.\n\n explicit\n Yc needs to be sent.\n\n struct {\n select (PublicValueEncoding) {\n case implicit: struct { };\n case explicit: opaque dh_Yc<1..2^16-1>;\n } dh_public;\n } ClientDiffieHellmanPublic;\n\n dh_Yc\n The client's Diffie-Hellman public value (Yc).\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 61]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.8. Certificate Verify\n\n When this message will be sent:\n\n This message is used to provide explicit verification of a client\n certificate. This message is only sent following a client\n certificate that has signing capability (i.e., all certificates\n except those containing fixed Diffie-Hellman parameters). When\n sent, it MUST immediately follow the client key exchange message.\n\n Structure of this message:\n\n struct {\n digitally-signed struct {\n opaque handshake_messages[handshake_messages_length];\n }\n } CertificateVerify;\n\n Here handshake_messages refers to all handshake messages sent or\n received, starting at client hello and up to, but not including,\n this message, including the type and length fields of the\n handshake messages. This is the concatenation of all the\n Handshake structures (as defined in Section 7.4) exchanged thus\n far. Note that this requires both sides to either buffer the\n messages or compute running hashes for all potential hash\n algorithms up to the time of the CertificateVerify computation.\n Servers can minimize this computation cost by offering a\n restricted set of digest algorithms in the CertificateRequest\n message.\n\n The hash and signature algorithms used in the signature MUST be\n one of those present in the supported_signature_algorithms field\n of the CertificateRequest message. In addition, the hash and\n signature algorithms MUST be compatible with the key in the\n client's end-entity certificate. RSA keys MAY be used with any\n permitted hash algorithm, subject to restrictions in the\n certificate, if any.\n\n Because DSA signatures do not contain any secure indication of\n hash algorithm, there is a risk of hash substitution if multiple\n hashes may be used with any key. Currently, DSA [DSS] may only be\n used with SHA-1. Future revisions of DSS [DSS-3] are expected to\n allow the use of other digest algorithms with DSA, as well as\n guidance as to which digest algorithms should be used with each\n key size. In addition, future revisions of [PKIX] may specify\n mechanisms for certificates to indicate which digest algorithms\n are to be used with DSA.\n\n\n\n\nDierks & Rescorla Standards Track [Page 62]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.9. Finished\n\n When this message will be sent:\n\n A Finished message is always sent immediately after a change\n cipher spec message to verify that the key exchange and\n authentication processes were successful. It is essential that a\n change cipher spec message be received between the other handshake\n messages and the Finished message.\n\n Meaning of this message:\n\n The Finished message is the first one protected with the just\n negotiated algorithms, keys, and secrets. Recipients of Finished\n messages MUST verify that the contents are correct. Once a side\n has sent its Finished message and received and validated the\n Finished message from its peer, it may begin to send and receive\n application data over the connection.\n\n Structure of this message:\n\n struct {\n opaque verify_data[verify_data_length];\n } Finished;\n\n verify_data\n PRF(master_secret, finished_label, Hash(handshake_messages))\n [0..verify_data_length-1];\n\n finished_label\n For Finished messages sent by the client, the string\n \"client finished\". For Finished messages sent by the server,\n the string \"server finished\".\n\n Hash denotes a Hash of the handshake messages. For the PRF\n defined in Section 5, the Hash MUST be the Hash used as the basis\n for the PRF. Any cipher suite which defines a different PRF MUST\n also define the Hash to use in the Finished computation.\n\n In previous versions of TLS, the verify_data was always 12 octets\n long. In the current version of TLS, it depends on the cipher\n suite. Any cipher suite which does not explicitly specify\n verify_data_length has a verify_data_length equal to 12. This\n includes all existing cipher suites. Note that this\n representation has the same encoding as with previous versions.\n Future cipher suites MAY specify other lengths but such length\n MUST be at least 12 bytes.\n\n\n\n\nDierks & Rescorla Standards Track [Page 63]\n\f\nRFC 5246 TLS August 2008\n\n\n handshake_messages\n All of the data from all messages in this handshake (not\n including any HelloRequest messages) up to, but not including,\n this message. This is only data visible at the handshake layer\n and does not include record layer headers. This is the\n concatenation of all the Handshake structures as defined in\n Section 7.4, exchanged thus far.\n\n It is a fatal error if a Finished message is not preceded by a\n ChangeCipherSpec message at the appropriate point in the handshake.\n\n The value handshake_messages includes all handshake messages starting\n at ClientHello up to, but not including, this Finished message. This\n may be different from handshake_messages in Section 7.4.8 because it\n would include the CertificateVerify message (if sent). Also, the\n handshake_messages for the Finished message sent by the client will\n be different from that for the Finished message sent by the server,\n because the one that is sent second will include the prior one.\n\n Note: ChangeCipherSpec messages, alerts, and any other record types\n are not handshake messages and are not included in the hash\n computations. Also, HelloRequest messages are omitted from handshake\n hashes.\n\n8. Cryptographic Computations\n\n In order to begin connection protection, the TLS Record Protocol\n requires specification of a suite of algorithms, a master secret, and\n the client and server random values. The authentication, encryption,\n and MAC algorithms are determined by the cipher_suite selected by the\n server and revealed in the ServerHello message. The compression\n algorithm is negotiated in the hello messages, and the random values\n are exchanged in the hello messages. All that remains is to\n calculate the master secret.\n\n8.1. Computing the Master Secret\n\n For all key exchange methods, the same algorithm is used to convert\n the pre_master_secret into the master_secret. The pre_master_secret\n should be deleted from memory once the master_secret has been\n computed.\n\n master_secret = PRF(pre_master_secret, \"master secret\",\n ClientHello.random + ServerHello.random)\n [0..47];\n\n The master secret is always exactly 48 bytes in length. The length\n of the premaster secret will vary depending on key exchange method.\n\n\n\nDierks & Rescorla Standards Track [Page 64]\n\f\nRFC 5246 TLS August 2008\n\n\n8.1.1. RSA\n\n When RSA is used for server authentication and key exchange, a 48-\n byte pre_master_secret is generated by the client, encrypted under\n the server's public key, and sent to the server. The server uses its\n private key to decrypt the pre_master_secret. Both parties then\n convert the pre_master_secret into the master_secret, as specified\n above.\n\n8.1.2. Diffie-Hellman\n\n A conventional Diffie-Hellman computation is performed. The\n negotiated key (Z) is used as the pre_master_secret, and is converted\n into the master_secret, as specified above. Leading bytes of Z that\n contain all zero bits are stripped before it is used as the\n pre_master_secret.\n\n Note: Diffie-Hellman parameters are specified by the server and may\n be either ephemeral or contained within the server's certificate.\n\n9. Mandatory Cipher Suites\n\n In the absence of an application profile standard specifying\n otherwise, a TLS-compliant application MUST implement the cipher\n suite TLS_RSA_WITH_AES_128_CBC_SHA (see Appendix A.5 for the\n definition).\n\n10. Application Data Protocol\n\n Application data messages are carried by the record layer and are\n fragmented, compressed, and encrypted based on the current connection\n state. The messages are treated as transparent data to the record\n layer.\n\n11. Security Considerations\n\n Security issues are discussed throughout this memo, especially in\n Appendices D, E, and F.\n\n12. IANA Considerations\n\n This document uses several registries that were originally created in\n [TLS1.1]. IANA has updated these to reference this document. The\n registries and their allocation policies (unchanged from [TLS1.1])\n are listed below.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 65]\n\f\nRFC 5246 TLS August 2008\n\n\n - TLS ClientCertificateType Identifiers Registry: Future values in\n the range 0-63 (decimal) inclusive are assigned via Standards\n Action [RFC2434]. Values in the range 64-223 (decimal) inclusive\n are assigned via Specification Required [RFC2434]. Values from\n 224-255 (decimal) inclusive are reserved for Private Use\n [RFC2434].\n\n - TLS Cipher Suite Registry: Future values with the first byte in\n the range 0-191 (decimal) inclusive are assigned via Standards\n Action [RFC2434]. Values with the first byte in the range 192-254\n (decimal) are assigned via Specification Required [RFC2434].\n Values with the first byte 255 (decimal) are reserved for Private\n Use [RFC2434].\n\n - This document defines several new HMAC-SHA256-based cipher suites,\n whose values (in Appendix A.5) have been allocated from the TLS\n Cipher Suite registry.\n\n - TLS ContentType Registry: Future values are allocated via\n Standards Action [RFC2434].\n\n - TLS Alert Registry: Future values are allocated via Standards\n Action [RFC2434].\n\n - TLS HandshakeType Registry: Future values are allocated via\n Standards Action [RFC2434].\n\n This document also uses a registry originally created in [RFC4366].\n IANA has updated it to reference this document. The registry and its\n allocation policy (unchanged from [RFC4366]) is listed below:\n\n - TLS ExtensionType Registry: Future values are allocated via IETF\n Consensus [RFC2434]. IANA has updated this registry to include\n the signature_algorithms extension and its corresponding value\n (see Section 7.4.1.4).\n\n In addition, this document defines two new registries to be\n maintained by IANA:\n\n - TLS SignatureAlgorithm Registry: The registry has been initially\n populated with the values described in Section 7.4.1.4.1. Future\n values in the range 0-63 (decimal) inclusive are assigned via\n Standards Action [RFC2434]. Values in the range 64-223 (decimal)\n inclusive are assigned via Specification Required [RFC2434].\n Values from 224-255 (decimal) inclusive are reserved for Private\n Use [RFC2434].\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 66]\n\f\nRFC 5246 TLS August 2008\n\n\n - TLS HashAlgorithm Registry: The registry has been initially\n populated with the values described in Section 7.4.1.4.1. Future\n values in the range 0-63 (decimal) inclusive are assigned via\n Standards Action [RFC2434]. Values in the range 64-223 (decimal)\n inclusive are assigned via Specification Required [RFC2434].\n Values from 224-255 (decimal) inclusive are reserved for Private\n Use [RFC2434].\n\n This document also uses the TLS Compression Method Identifiers\n Registry, defined in [RFC3749]. IANA has allocated value 0 for\n the \"null\" compression method.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 67]\n\f\nRFC 5246 TLS August 2008\n\n\nAppendix A. Protocol Data Structures and Constant Values\n\n This section describes protocol types and constants.\n\nA.1. Record Layer\n\n struct {\n uint8 major;\n uint8 minor;\n } ProtocolVersion;\n\n ProtocolVersion version = { 3, 3 }; /* TLS v1.2*/\n\n enum {\n change_cipher_spec(20), alert(21), handshake(22),\n application_data(23), (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSCompressed.length];\n } TLSCompressed;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n select (SecurityParameters.cipher_type) {\n case stream: GenericStreamCipher;\n case block: GenericBlockCipher;\n case aead: GenericAEADCipher;\n } fragment;\n } TLSCiphertext;\n\n stream-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n } GenericStreamCipher;\n\n\n\n\nDierks & Rescorla Standards Track [Page 68]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n opaque IV[SecurityParameters.record_iv_length];\n block-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n uint8 padding[GenericBlockCipher.padding_length];\n uint8 padding_length;\n };\n } GenericBlockCipher;\n\n struct {\n opaque nonce_explicit[SecurityParameters.record_iv_length];\n aead-ciphered struct {\n opaque content[TLSCompressed.length];\n };\n } GenericAEADCipher;\n\nA.2. Change Cipher Specs Message\n\n struct {\n enum { change_cipher_spec(1), (255) } type;\n } ChangeCipherSpec;\n\nA.3. Alert Messages\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction_RESERVED(60),\n protocol_version(70),\n\n\n\nDierks & Rescorla Standards Track [Page 69]\n\f\nRFC 5246 TLS August 2008\n\n\n insufficient_security(71),\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110), /* new */\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\nA.4. Handshake Protocol\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20)\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type;\n uint24 length;\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n } body;\n } Handshake;\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 70]\n\f\nRFC 5246 TLS August 2008\n\n\nA.4.1. Hello Messages\n\n struct { } HelloRequest;\n\n struct {\n uint32 gmt_unix_time;\n opaque random_bytes[28];\n } Random;\n\n opaque SessionID<0..32>;\n\n uint8 CipherSuite[2];\n\n enum { null(0), (255) } CompressionMethod;\n\n struct {\n ProtocolVersion client_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suites<2..2^16-2>;\n CompressionMethod compression_methods<1..2^8-1>;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ClientHello;\n\n struct {\n ProtocolVersion server_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ServerHello;\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n\n\n\nDierks & Rescorla Standards Track [Page 71]\n\f\nRFC 5246 TLS August 2008\n\n\n enum {\n signature_algorithms(13), (65535)\n } ExtensionType;\n\n enum{\n none(0), md5(1), sha1(2), sha224(3), sha256(4), sha384(5),\n sha512(6), (255)\n } HashAlgorithm;\n enum {\n anonymous(0), rsa(1), dsa(2), ecdsa(3), (255)\n } SignatureAlgorithm;\n\n struct {\n HashAlgorithm hash;\n SignatureAlgorithm signature;\n } SignatureAndHashAlgorithm;\n\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2..2^16-1>;\n\nA.4.2. Server Authentication and Key Exchange Messages\n\n opaque ASN.1Cert<2^24-1>;\n\n struct {\n ASN.1Cert certificate_list<0..2^24-1>;\n } Certificate;\n\n enum { dhe_dss, dhe_rsa, dh_anon, rsa,dh_dss, dh_rsa\n /* may be extended, e.g., for ECDH -- see [TLSECC] */\n } KeyExchangeAlgorithm;\n\n struct {\n opaque dh_p<1..2^16-1>;\n opaque dh_g<1..2^16-1>;\n opaque dh_Ys<1..2^16-1>;\n } ServerDHParams; /* Ephemeral DH parameters */\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 72]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n select (KeyExchangeAlgorithm) {\n case dh_anon:\n ServerDHParams params;\n case dhe_dss:\n case dhe_rsa:\n ServerDHParams params;\n digitally-signed struct {\n opaque client_random[32];\n opaque server_random[32];\n ServerDHParams params;\n } signed_params;\n case rsa:\n case dh_dss:\n case dh_rsa:\n struct {} ;\n /* message is omitted for rsa, dh_dss, and dh_rsa */\n /* may be extended, e.g., for ECDH -- see [TLSECC] */\n } ServerKeyExchange;\n\n enum {\n rsa_sign(1), dss_sign(2), rsa_fixed_dh(3), dss_fixed_dh(4),\n rsa_ephemeral_dh_RESERVED(5), dss_ephemeral_dh_RESERVED(6),\n fortezza_dms_RESERVED(20),\n (255)\n } ClientCertificateType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n ClientCertificateType certificate_types<1..2^8-1>;\n DistinguishedName certificate_authorities<0..2^16-1>;\n } CertificateRequest;\n\n struct { } ServerHelloDone;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 73]\n\f\nRFC 5246 TLS August 2008\n\n\nA.4.3. Client Authentication and Key Exchange Messages\n\n struct {\n select (KeyExchangeAlgorithm) {\n case rsa:\n EncryptedPreMasterSecret;\n case dhe_dss:\n case dhe_rsa:\n case dh_dss:\n case dh_rsa:\n case dh_anon:\n ClientDiffieHellmanPublic;\n } exchange_keys;\n } ClientKeyExchange;\n\n struct {\n ProtocolVersion client_version;\n opaque random[46];\n } PreMasterSecret;\n\n struct {\n public-key-encrypted PreMasterSecret pre_master_secret;\n } EncryptedPreMasterSecret;\n\n enum { implicit, explicit } PublicValueEncoding;\n\n struct {\n select (PublicValueEncoding) {\n case implicit: struct {};\n case explicit: opaque DH_Yc<1..2^16-1>;\n } dh_public;\n } ClientDiffieHellmanPublic;\n\n struct {\n digitally-signed struct {\n opaque handshake_messages[handshake_messages_length];\n }\n } CertificateVerify;\n\nA.4.4. Handshake Finalization Message\n\n struct {\n opaque verify_data[verify_data_length];\n } Finished;\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 74]\n\f\nRFC 5246 TLS August 2008\n\n\nA.5. The Cipher Suite\n\n The following values define the cipher suite codes used in the\n ClientHello and ServerHello messages.\n\n A cipher suite defines a cipher specification supported in TLS\n Version 1.2.\n\n TLS_NULL_WITH_NULL_NULL is specified and is the initial state of a\n TLS connection during the first handshake on that channel, but MUST\n NOT be negotiated, as it provides no more protection than an\n unsecured connection.\n\n CipherSuite TLS_NULL_WITH_NULL_NULL = { 0x00,0x00 };\n\n The following CipherSuite definitions require that the server provide\n an RSA certificate that can be used for key exchange. The server may\n request any signature-capable certificate in the certificate request\n message.\n\n CipherSuite TLS_RSA_WITH_NULL_MD5 = { 0x00,0x01 };\n CipherSuite TLS_RSA_WITH_NULL_SHA = { 0x00,0x02 };\n CipherSuite TLS_RSA_WITH_NULL_SHA256 = { 0x00,0x3B };\n CipherSuite TLS_RSA_WITH_RC4_128_MD5 = { 0x00,0x04 };\n CipherSuite TLS_RSA_WITH_RC4_128_SHA = { 0x00,0x05 };\n CipherSuite TLS_RSA_WITH_3DES_EDE_CBC_SHA = { 0x00,0x0A };\n CipherSuite TLS_RSA_WITH_AES_128_CBC_SHA = { 0x00,0x2F };\n CipherSuite TLS_RSA_WITH_AES_256_CBC_SHA = { 0x00,0x35 };\n CipherSuite TLS_RSA_WITH_AES_128_CBC_SHA256 = { 0x00,0x3C };\n CipherSuite TLS_RSA_WITH_AES_256_CBC_SHA256 = { 0x00,0x3D };\n\n The following cipher suite definitions are used for server-\n authenticated (and optionally client-authenticated) Diffie-Hellman.\n DH denotes cipher suites in which the server's certificate contains\n the Diffie-Hellman parameters signed by the certificate authority\n (CA). DHE denotes ephemeral Diffie-Hellman, where the Diffie-Hellman\n parameters are signed by a signature-capable certificate, which has\n been signed by the CA. The signing algorithm used by the server is\n specified after the DHE component of the CipherSuite name. The\n server can request any signature-capable certificate from the client\n for client authentication, or it may request a Diffie-Hellman\n certificate. Any Diffie-Hellman certificate provided by the client\n must use the parameters (group and generator) described by the\n server.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 75]\n\f\nRFC 5246 TLS August 2008\n\n\n CipherSuite TLS_DH_DSS_WITH_3DES_EDE_CBC_SHA = { 0x00,0x0D };\n CipherSuite TLS_DH_RSA_WITH_3DES_EDE_CBC_SHA = { 0x00,0x10 };\n CipherSuite TLS_DHE_DSS_WITH_3DES_EDE_CBC_SHA = { 0x00,0x13 };\n CipherSuite TLS_DHE_RSA_WITH_3DES_EDE_CBC_SHA = { 0x00,0x16 };\n CipherSuite TLS_DH_DSS_WITH_AES_128_CBC_SHA = { 0x00,0x30 };\n CipherSuite TLS_DH_RSA_WITH_AES_128_CBC_SHA = { 0x00,0x31 };\n CipherSuite TLS_DHE_DSS_WITH_AES_128_CBC_SHA = { 0x00,0x32 };\n CipherSuite TLS_DHE_RSA_WITH_AES_128_CBC_SHA = { 0x00,0x33 };\n CipherSuite TLS_DH_DSS_WITH_AES_256_CBC_SHA = { 0x00,0x36 };\n CipherSuite TLS_DH_RSA_WITH_AES_256_CBC_SHA = { 0x00,0x37 };\n CipherSuite TLS_DHE_DSS_WITH_AES_256_CBC_SHA = { 0x00,0x38 };\n CipherSuite TLS_DHE_RSA_WITH_AES_256_CBC_SHA = { 0x00,0x39 };\n CipherSuite TLS_DH_DSS_WITH_AES_128_CBC_SHA256 = { 0x00,0x3E };\n CipherSuite TLS_DH_RSA_WITH_AES_128_CBC_SHA256 = { 0x00,0x3F };\n CipherSuite TLS_DHE_DSS_WITH_AES_128_CBC_SHA256 = { 0x00,0x40 };\n CipherSuite TLS_DHE_RSA_WITH_AES_128_CBC_SHA256 = { 0x00,0x67 };\n CipherSuite TLS_DH_DSS_WITH_AES_256_CBC_SHA256 = { 0x00,0x68 };\n CipherSuite TLS_DH_RSA_WITH_AES_256_CBC_SHA256 = { 0x00,0x69 };\n CipherSuite TLS_DHE_DSS_WITH_AES_256_CBC_SHA256 = { 0x00,0x6A };\n CipherSuite TLS_DHE_RSA_WITH_AES_256_CBC_SHA256 = { 0x00,0x6B };\n\n The following cipher suites are used for completely anonymous\n Diffie-Hellman communications in which neither party is\n authenticated. Note that this mode is vulnerable to man-in-the-\n middle attacks. Using this mode therefore is of limited use: These\n cipher suites MUST NOT be used by TLS 1.2 implementations unless the\n application layer has specifically requested to allow anonymous key\n exchange. (Anonymous key exchange may sometimes be acceptable, for\n example, to support opportunistic encryption when no set-up for\n authentication is in place, or when TLS is used as part of more\n complex security protocols that have other means to ensure\n authentication.)\n\n CipherSuite TLS_DH_anon_WITH_RC4_128_MD5 = { 0x00,0x18 };\n CipherSuite TLS_DH_anon_WITH_3DES_EDE_CBC_SHA = { 0x00,0x1B };\n CipherSuite TLS_DH_anon_WITH_AES_128_CBC_SHA = { 0x00,0x34 };\n CipherSuite TLS_DH_anon_WITH_AES_256_CBC_SHA = { 0x00,0x3A };\n CipherSuite TLS_DH_anon_WITH_AES_128_CBC_SHA256 = { 0x00,0x6C };\n CipherSuite TLS_DH_anon_WITH_AES_256_CBC_SHA256 = { 0x00,0x6D };\n\n Note that using non-anonymous key exchange without actually verifying\n the key exchange is essentially equivalent to anonymous key exchange,\n and the same precautions apply. While non-anonymous key exchange\n will generally involve a higher computational and communicational\n cost than anonymous key exchange, it may be in the interest of\n interoperability not to disable non-anonymous key exchange when the\n application layer is allowing anonymous key exchange.\n\n\n\n\nDierks & Rescorla Standards Track [Page 76]\n\f\nRFC 5246 TLS August 2008\n\n\n New cipher suite values have been assigned by IANA as described in\n Section 12.\n\n Note: The cipher suite values { 0x00, 0x1C } and { 0x00, 0x1D } are\n reserved to avoid collision with Fortezza-based cipher suites in\n SSL 3.\n\nA.6. The Security Parameters\n\n These security parameters are determined by the TLS Handshake\n Protocol and provided as parameters to the TLS record layer in order\n to initialize a connection state. SecurityParameters includes:\n\n enum { null(0), (255) } CompressionMethod;\n\n enum { server, client } ConnectionEnd;\n\n enum { tls_prf_sha256 } PRFAlgorithm;\n\n enum { null, rc4, 3des, aes } BulkCipherAlgorithm;\n\n enum { stream, block, aead } CipherType;\n\n enum { null, hmac_md5, hmac_sha1, hmac_sha256, hmac_sha384,\n hmac_sha512} MACAlgorithm;\n\n /* Other values may be added to the algorithms specified in\n CompressionMethod, PRFAlgorithm, BulkCipherAlgorithm, and\n MACAlgorithm. */\n\n struct {\n ConnectionEnd entity;\n PRFAlgorithm prf_algorithm;\n BulkCipherAlgorithm bulk_cipher_algorithm;\n CipherType cipher_type;\n uint8 enc_key_length;\n uint8 block_length;\n uint8 fixed_iv_length;\n uint8 record_iv_length;\n MACAlgorithm mac_algorithm;\n uint8 mac_length;\n uint8 mac_key_length;\n CompressionMethod compression_algorithm;\n opaque master_secret[48];\n opaque client_random[32];\n opaque server_random[32];\n } SecurityParameters;\n\n\n\n\nDierks & Rescorla Standards Track [Page 77]\n\f\nRFC 5246 TLS August 2008\n\n\nA.7. Changes to RFC 4492\n\n RFC 4492 [TLSECC] adds Elliptic Curve cipher suites to TLS. This\n document changes some of the structures used in that document. This\n section details the required changes for implementors of both RFC\n 4492 and TLS 1.2. Implementors of TLS 1.2 who are not implementing\n RFC 4492 do not need to read this section.\n\n This document adds a \"signature_algorithm\" field to the digitally-\n signed element in order to identify the signature and digest\n algorithms used to create a signature. This change applies to\n digital signatures formed using ECDSA as well, thus allowing ECDSA\n signatures to be used with digest algorithms other than SHA-1,\n provided such use is compatible with the certificate and any\n restrictions imposed by future revisions of [PKIX].\n\n As described in Sections 7.4.2 and 7.4.6, the restrictions on the\n signature algorithms used to sign certificates are no longer tied to\n the cipher suite (when used by the server) or the\n ClientCertificateType (when used by the client). Thus, the\n restrictions on the algorithm used to sign certificates specified in\n Sections 2 and 3 of RFC 4492 are also relaxed. As in this document,\n the restrictions on the keys in the end-entity certificate remain.\n\nAppendix B. Glossary\n\n Advanced Encryption Standard (AES)\n AES [AES] is a widely used symmetric encryption algorithm. AES is\n a block cipher with a 128-, 192-, or 256-bit keys and a 16-byte\n block size. TLS currently only supports the 128- and 256-bit key\n sizes.\n\n application protocol\n An application protocol is a protocol that normally layers\n directly on top of the transport layer (e.g., TCP/IP). Examples\n include HTTP, TELNET, FTP, and SMTP.\n\n asymmetric cipher\n See public key cryptography.\n\n authenticated encryption with additional data (AEAD)\n A symmetric encryption algorithm that simultaneously provides\n confidentiality and message integrity.\n\n authentication\n Authentication is the ability of one entity to determine the\n identity of another entity.\n\n\n\n\nDierks & Rescorla Standards Track [Page 78]\n\f\nRFC 5246 TLS August 2008\n\n\n block cipher\n A block cipher is an algorithm that operates on plaintext in\n groups of bits, called blocks. 64 bits was, and 128 bits is, a\n common block size.\n\n bulk cipher\n A symmetric encryption algorithm used to encrypt large quantities\n of data.\n\n cipher block chaining (CBC)\n CBC is a mode in which every plaintext block encrypted with a\n block cipher is first exclusive-ORed with the previous ciphertext\n block (or, in the case of the first block, with the initialization\n vector). For decryption, every block is first decrypted, then\n exclusive-ORed with the previous ciphertext block (or IV).\n\n certificate\n As part of the X.509 protocol (a.k.a. ISO Authentication\n framework), certificates are assigned by a trusted Certificate\n Authority and provide a strong binding between a party's identity\n or some other attributes and its public key.\n\n client\n The application entity that initiates a TLS connection to a\n server. This may or may not imply that the client initiated the\n underlying transport connection. The primary operational\n difference between the server and client is that the server is\n generally authenticated, while the client is only optionally\n authenticated.\n\n client write key\n The key used to encrypt data written by the client.\n\n client write MAC key\n The secret data used to authenticate data written by the client.\n\n connection\n A connection is a transport (in the OSI layering model definition)\n that provides a suitable type of service. For TLS, such\n connections are peer-to-peer relationships. The connections are\n transient. Every connection is associated with one session.\n\n Data Encryption Standard\n DES [DES] still is a very widely used symmetric encryption\n algorithm although it is considered as rather weak now. DES is a\n block cipher with a 56-bit key and an 8-byte block size. Note\n that in TLS, for key generation purposes, DES is treated as having\n an 8-byte key length (64 bits), but it still only provides 56 bits\n\n\n\nDierks & Rescorla Standards Track [Page 79]\n\f\nRFC 5246 TLS August 2008\n\n\n of protection. (The low bit of each key byte is presumed to be\n set to produce odd parity in that key byte.) DES can also be\n operated in a mode [3DES] where three independent keys and three\n encryptions are used for each block of data; this uses 168 bits of\n key (24 bytes in the TLS key generation method) and provides the\n equivalent of 112 bits of security.\n\n Digital Signature Standard (DSS)\n A standard for digital signing, including the Digital Signing\n Algorithm, approved by the National Institute of Standards and\n Technology, defined in NIST FIPS PUB 186-2, \"Digital Signature\n Standard\", published January 2000 by the U.S. Department of\n Commerce [DSS]. A significant update [DSS-3] has been drafted and\n was published in March 2006.\n\n digital signatures\n Digital signatures utilize public key cryptography and one-way\n hash functions to produce a signature of the data that can be\n authenticated, and is difficult to forge or repudiate.\n\n handshake An initial negotiation between client and server that\n establishes the parameters of their transactions.\n\n Initialization Vector (IV)\n When a block cipher is used in CBC mode, the initialization vector\n is exclusive-ORed with the first plaintext block prior to\n encryption.\n\n Message Authentication Code (MAC)\n A Message Authentication Code is a one-way hash computed from a\n message and some secret data. It is difficult to forge without\n knowing the secret data. Its purpose is to detect if the message\n has been altered.\n\n master secret\n Secure secret data used for generating encryption keys, MAC\n secrets, and IVs.\n\n MD5\n MD5 [MD5] is a hashing function that converts an arbitrarily long\n data stream into a hash of fixed size (16 bytes). Due to\n significant progress in cryptanalysis, at the time of publication\n of this document, MD5 no longer can be considered a 'secure'\n hashing function.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 80]\n\f\nRFC 5246 TLS August 2008\n\n\n public key cryptography\n A class of cryptographic techniques employing two-key ciphers.\n Messages encrypted with the public key can only be decrypted with\n the associated private key. Conversely, messages signed with the\n private key can be verified with the public key.\n\n one-way hash function\n A one-way transformation that converts an arbitrary amount of data\n into a fixed-length hash. It is computationally hard to reverse\n the transformation or to find collisions. MD5 and SHA are\n examples of one-way hash functions.\n\n RC4\n A stream cipher invented by Ron Rivest. A compatible cipher is\n described in [SCH].\n\n RSA\n A very widely used public key algorithm that can be used for\n either encryption or digital signing. [RSA]\n\n server\n The server is the application entity that responds to requests for\n connections from clients. See also \"client\".\n\n session\n A TLS session is an association between a client and a server.\n Sessions are created by the handshake protocol. Sessions define a\n set of cryptographic security parameters that can be shared among\n multiple connections. Sessions are used to avoid the expensive\n negotiation of new security parameters for each connection.\n\n session identifier\n A session identifier is a value generated by a server that\n identifies a particular session.\n\n server write key\n The key used to encrypt data written by the server.\n\n server write MAC key\n The secret data used to authenticate data written by the server.\n\n SHA\n The Secure Hash Algorithm [SHS] is defined in FIPS PUB 180-2. It\n produces a 20-byte output. Note that all references to SHA\n (without a numerical suffix) actually use the modified SHA-1\n algorithm.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 81]\n\f\nRFC 5246 TLS August 2008\n\n\n SHA-256\n The 256-bit Secure Hash Algorithm is defined in FIPS PUB 180-2.\n It produces a 32-byte output.\n\n SSL\n Netscape's Secure Socket Layer protocol [SSL3]. TLS is based on\n SSL Version 3.0.\n\n stream cipher\n An encryption algorithm that converts a key into a\n cryptographically strong keystream, which is then exclusive-ORed\n with the plaintext.\n\n symmetric cipher\n See bulk cipher.\n\n Transport Layer Security (TLS)\n This protocol; also, the Transport Layer Security working group of\n the Internet Engineering Task Force (IETF). See \"Working Group\n Information\" at the end of this document (see page 99).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 82]\n\f\nRFC 5246 TLS August 2008\n\n\nAppendix C. Cipher Suite Definitions\n\nCipher Suite Key Cipher Mac\n Exchange\n\nTLS_NULL_WITH_NULL_NULL NULL NULL NULL\nTLS_RSA_WITH_NULL_MD5 RSA NULL MD5\nTLS_RSA_WITH_NULL_SHA RSA NULL SHA\nTLS_RSA_WITH_NULL_SHA256 RSA NULL SHA256\nTLS_RSA_WITH_RC4_128_MD5 RSA RC4_128 MD5\nTLS_RSA_WITH_RC4_128_SHA RSA RC4_128 SHA\nTLS_RSA_WITH_3DES_EDE_CBC_SHA RSA 3DES_EDE_CBC SHA\nTLS_RSA_WITH_AES_128_CBC_SHA RSA AES_128_CBC SHA\nTLS_RSA_WITH_AES_256_CBC_SHA RSA AES_256_CBC SHA\nTLS_RSA_WITH_AES_128_CBC_SHA256 RSA AES_128_CBC SHA256\nTLS_RSA_WITH_AES_256_CBC_SHA256 RSA AES_256_CBC SHA256\nTLS_DH_DSS_WITH_3DES_EDE_CBC_SHA DH_DSS 3DES_EDE_CBC SHA\nTLS_DH_RSA_WITH_3DES_EDE_CBC_SHA DH_RSA 3DES_EDE_CBC SHA\nTLS_DHE_DSS_WITH_3DES_EDE_CBC_SHA DHE_DSS 3DES_EDE_CBC SHA\nTLS_DHE_RSA_WITH_3DES_EDE_CBC_SHA DHE_RSA 3DES_EDE_CBC SHA\nTLS_DH_anon_WITH_RC4_128_MD5 DH_anon RC4_128 MD5\nTLS_DH_anon_WITH_3DES_EDE_CBC_SHA DH_anon 3DES_EDE_CBC SHA\nTLS_DH_DSS_WITH_AES_128_CBC_SHA DH_DSS AES_128_CBC SHA\nTLS_DH_RSA_WITH_AES_128_CBC_SHA DH_RSA AES_128_CBC SHA\nTLS_DHE_DSS_WITH_AES_128_CBC_SHA DHE_DSS AES_128_CBC SHA\nTLS_DHE_RSA_WITH_AES_128_CBC_SHA DHE_RSA AES_128_CBC SHA\nTLS_DH_anon_WITH_AES_128_CBC_SHA DH_anon AES_128_CBC SHA\nTLS_DH_DSS_WITH_AES_256_CBC_SHA DH_DSS AES_256_CBC SHA\nTLS_DH_RSA_WITH_AES_256_CBC_SHA DH_RSA AES_256_CBC SHA\nTLS_DHE_DSS_WITH_AES_256_CBC_SHA DHE_DSS AES_256_CBC SHA\nTLS_DHE_RSA_WITH_AES_256_CBC_SHA DHE_RSA AES_256_CBC SHA\nTLS_DH_anon_WITH_AES_256_CBC_SHA DH_anon AES_256_CBC SHA\nTLS_DH_DSS_WITH_AES_128_CBC_SHA256 DH_DSS AES_128_CBC SHA256\nTLS_DH_RSA_WITH_AES_128_CBC_SHA256 DH_RSA AES_128_CBC SHA256\nTLS_DHE_DSS_WITH_AES_128_CBC_SHA256 DHE_DSS AES_128_CBC SHA256\nTLS_DHE_RSA_WITH_AES_128_CBC_SHA256 DHE_RSA AES_128_CBC SHA256\nTLS_DH_anon_WITH_AES_128_CBC_SHA256 DH_anon AES_128_CBC SHA256\nTLS_DH_DSS_WITH_AES_256_CBC_SHA256 DH_DSS AES_256_CBC SHA256\nTLS_DH_RSA_WITH_AES_256_CBC_SHA256 DH_RSA AES_256_CBC SHA256\nTLS_DHE_DSS_WITH_AES_256_CBC_SHA256 DHE_DSS AES_256_CBC SHA256\nTLS_DHE_RSA_WITH_AES_256_CBC_SHA256 DHE_RSA AES_256_CBC SHA256\nTLS_DH_anon_WITH_AES_256_CBC_SHA256 DH_anon AES_256_CBC SHA256\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 83]\n\f\nRFC 5246 TLS August 2008\n\n\n Key IV Block\nCipher Type Material Size Size\n------------ ------ -------- ---- -----\nNULL Stream 0 0 N/A\nRC4_128 Stream 16 0 N/A\n3DES_EDE_CBC Block 24 8 8\nAES_128_CBC Block 16 16 16\nAES_256_CBC Block 32 16 16\n\n\nMAC Algorithm mac_length mac_key_length\n-------- ----------- ---------- --------------\nNULL N/A 0 0\nMD5 HMAC-MD5 16 16\nSHA HMAC-SHA1 20 20\nSHA256 HMAC-SHA256 32 32\n\n Type\n Indicates whether this is a stream cipher or a block cipher\n running in CBC mode.\n\n Key Material\n The number of bytes from the key_block that are used for\n generating the write keys.\n\n IV Size\n The amount of data needed to be generated for the initialization\n vector. Zero for stream ciphers; equal to the block size for\n block ciphers (this is equal to\n SecurityParameters.record_iv_length).\n\n Block Size\n The amount of data a block cipher enciphers in one chunk; a block\n cipher running in CBC mode can only encrypt an even multiple of\n its block size.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 84]\n\f\nRFC 5246 TLS August 2008\n\n\nAppendix D. Implementation Notes\n\n The TLS protocol cannot prevent many common security mistakes. This\n section provides several recommendations to assist implementors.\n\nD.1. Random Number Generation and Seeding\n\n TLS requires a cryptographically secure pseudorandom number generator\n (PRNG). Care must be taken in designing and seeding PRNGs. PRNGs\n based on secure hash operations, most notably SHA-1, are acceptable,\n but cannot provide more security than the size of the random number\n generator state.\n\n To estimate the amount of seed material being produced, add the\n number of bits of unpredictable information in each seed byte. For\n example, keystroke timing values taken from a PC compatible's 18.2 Hz\n timer provide 1 or 2 secure bits each, even though the total size of\n the counter value is 16 bits or more. Seeding a 128-bit PRNG would\n thus require approximately 100 such timer values.\n\n [RANDOM] provides guidance on the generation of random values.\n\nD.2. Certificates and Authentication\n\n Implementations are responsible for verifying the integrity of\n certificates and should generally support certificate revocation\n messages. Certificates should always be verified to ensure proper\n signing by a trusted Certificate Authority (CA). The selection and\n addition of trusted CAs should be done very carefully. Users should\n be able to view information about the certificate and root CA.\n\nD.3. Cipher Suites\n\n TLS supports a range of key sizes and security levels, including some\n that provide no or minimal security. A proper implementation will\n probably not support many cipher suites. For instance, anonymous\n Diffie-Hellman is strongly discouraged because it cannot prevent man-\n in-the-middle attacks. Applications should also enforce minimum and\n maximum key sizes. For example, certificate chains containing 512-\n bit RSA keys or signatures are not appropriate for high-security\n applications.\n\nD.4. Implementation Pitfalls\n\n Implementation experience has shown that certain parts of earlier TLS\n specifications are not easy to understand, and have been a source of\n interoperability and security problems. Many of these areas have\n\n\n\n\nDierks & Rescorla Standards Track [Page 85]\n\f\nRFC 5246 TLS August 2008\n\n\n been clarified in this document, but this appendix contains a short\n list of the most important things that require special attention from\n implementors.\n\n TLS protocol issues:\n\n - Do you correctly handle handshake messages that are fragmented to\n multiple TLS records (see Section 6.2.1)? Including corner cases\n like a ClientHello that is split to several small fragments? Do\n you fragment handshake messages that exceed the maximum fragment\n size? In particular, the certificate and certificate request\n handshake messages can be large enough to require fragmentation.\n\n - Do you ignore the TLS record layer version number in all TLS\n records before ServerHello (see Appendix E.1)?\n\n - Do you handle TLS extensions in ClientHello correctly, including\n omitting the extensions field completely?\n\n - Do you support renegotiation, both client and server initiated?\n While renegotiation is an optional feature, supporting it is\n highly recommended.\n\n - When the server has requested a client certificate, but no\n suitable certificate is available, do you correctly send an empty\n Certificate message, instead of omitting the whole message (see\n Section 7.4.6)?\n\n Cryptographic details:\n\n - In the RSA-encrypted Premaster Secret, do you correctly send and\n verify the version number? When an error is encountered, do you\n continue the handshake to avoid the Bleichenbacher attack (see\n Section 7.4.7.1)?\n\n - What countermeasures do you use to prevent timing attacks against\n RSA decryption and signing operations (see Section 7.4.7.1)?\n\n - When verifying RSA signatures, do you accept both NULL and missing\n parameters (see Section 4.7)? Do you verify that the RSA padding\n doesn't have additional data after the hash value? [FI06]\n\n - When using Diffie-Hellman key exchange, do you correctly strip\n leading zero bytes from the negotiated key (see Section 8.1.2)?\n\n - Does your TLS client check that the Diffie-Hellman parameters sent\n by the server are acceptable (see Section F.1.1.3)?\n\n\n\n\nDierks & Rescorla Standards Track [Page 86]\n\f\nRFC 5246 TLS August 2008\n\n\n - How do you generate unpredictable IVs for CBC mode ciphers (see\n Section 6.2.3.2)?\n\n - Do you accept long CBC mode padding (up to 255 bytes; see Section\n 6.2.3.2)?\n\n - How do you address CBC mode timing attacks (Section 6.2.3.2)?\n\n - Do you use a strong and, most importantly, properly seeded random\n number generator (see Appendix D.1) for generating the premaster\n secret (for RSA key exchange), Diffie-Hellman private values, the\n DSA \"k\" parameter, and other security-critical values?\n\nAppendix E. Backward Compatibility\n\nE.1. Compatibility with TLS 1.0/1.1 and SSL 3.0\n\n Since there are various versions of TLS (1.0, 1.1, 1.2, and any\n future versions) and SSL (2.0 and 3.0), means are needed to negotiate\n the specific protocol version to use. The TLS protocol provides a\n built-in mechanism for version negotiation so as not to bother other\n protocol components with the complexities of version selection.\n\n TLS versions 1.0, 1.1, and 1.2, and SSL 3.0 are very similar, and use\n compatible ClientHello messages; thus, supporting all of them is\n relatively easy. Similarly, servers can easily handle clients trying\n to use future versions of TLS as long as the ClientHello format\n remains compatible, and the client supports the highest protocol\n version available in the server.\n\n A TLS 1.2 client who wishes to negotiate with such older servers will\n send a normal TLS 1.2 ClientHello, containing { 3, 3 } (TLS 1.2) in\n ClientHello.client_version. If the server does not support this\n version, it will respond with a ServerHello containing an older\n version number. If the client agrees to use this version, the\n negotiation will proceed as appropriate for the negotiated protocol.\n\n If the version chosen by the server is not supported by the client\n (or not acceptable), the client MUST send a \"protocol_version\" alert\n message and close the connection.\n\n If a TLS server receives a ClientHello containing a version number\n greater than the highest version supported by the server, it MUST\n reply according to the highest version supported by the server.\n\n A TLS server can also receive a ClientHello containing a version\n number smaller than the highest supported version. If the server\n wishes to negotiate with old clients, it will proceed as appropriate\n\n\n\nDierks & Rescorla Standards Track [Page 87]\n\f\nRFC 5246 TLS August 2008\n\n\n for the highest version supported by the server that is not greater\n than ClientHello.client_version. For example, if the server supports\n TLS 1.0, 1.1, and 1.2, and client_version is TLS 1.0, the server will\n proceed with a TLS 1.0 ServerHello. If server supports (or is\n willing to use) only versions greater than client_version, it MUST\n send a \"protocol_version\" alert message and close the connection.\n\n Whenever a client already knows the highest protocol version known to\n a server (for example, when resuming a session), it SHOULD initiate\n the connection in that native protocol.\n\n Note: some server implementations are known to implement version\n negotiation incorrectly. For example, there are buggy TLS 1.0\n servers that simply close the connection when the client offers a\n version newer than TLS 1.0. Also, it is known that some servers will\n refuse the connection if any TLS extensions are included in\n ClientHello. Interoperability with such buggy servers is a complex\n topic beyond the scope of this document, and may require multiple\n connection attempts by the client.\n\n Earlier versions of the TLS specification were not fully clear on\n what the record layer version number (TLSPlaintext.version) should\n contain when sending ClientHello (i.e., before it is known which\n version of the protocol will be employed). Thus, TLS servers\n compliant with this specification MUST accept any value {03,XX} as\n the record layer version number for ClientHello.\n\n TLS clients that wish to negotiate with older servers MAY send any\n value {03,XX} as the record layer version number. Typical values\n would be {03,00}, the lowest version number supported by the client,\n and the value of ClientHello.client_version. No single value will\n guarantee interoperability with all old servers, but this is a\n complex topic beyond the scope of this document.\n\nE.2. Compatibility with SSL 2.0\n\n TLS 1.2 clients that wish to support SSL 2.0 servers MUST send\n version 2.0 CLIENT-HELLO messages defined in [SSL2]. The message\n MUST contain the same version number as would be used for ordinary\n ClientHello, and MUST encode the supported TLS cipher suites in the\n CIPHER-SPECS-DATA field as described below.\n\n Warning: The ability to send version 2.0 CLIENT-HELLO messages will\n be phased out with all due haste, since the newer ClientHello format\n provides better mechanisms for moving to newer versions and\n negotiating extensions. TLS 1.2 clients SHOULD NOT support SSL 2.0.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 88]\n\f\nRFC 5246 TLS August 2008\n\n\n However, even TLS servers that do not support SSL 2.0 MAY accept\n version 2.0 CLIENT-HELLO messages. The message is presented below in\n sufficient detail for TLS server implementors; the true definition is\n still assumed to be [SSL2].\n\n For negotiation purposes, 2.0 CLIENT-HELLO is interpreted the same\n way as a ClientHello with a \"null\" compression method and no\n extensions. Note that this message MUST be sent directly on the\n wire, not wrapped as a TLS record. For the purposes of calculating\n Finished and CertificateVerify, the msg_length field is not\n considered to be a part of the handshake message.\n\n uint8 V2CipherSpec[3];\n struct {\n uint16 msg_length;\n uint8 msg_type;\n Version version;\n uint16 cipher_spec_length;\n uint16 session_id_length;\n uint16 challenge_length;\n V2CipherSpec cipher_specs[V2ClientHello.cipher_spec_length];\n opaque session_id[V2ClientHello.session_id_length];\n opaque challenge[V2ClientHello.challenge_length;\n } V2ClientHello;\n\n msg_length\n The highest bit MUST be 1; the remaining bits contain the length\n of the following data in bytes.\n\n msg_type\n This field, in conjunction with the version field, identifies a\n version 2 ClientHello message. The value MUST be 1.\n\n version\n Equal to ClientHello.client_version.\n\n cipher_spec_length\n This field is the total length of the field cipher_specs. It\n cannot be zero and MUST be a multiple of the V2CipherSpec length\n (3).\n\n session_id_length\n This field MUST have a value of zero for a client that claims to\n support TLS 1.2.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 89]\n\f\nRFC 5246 TLS August 2008\n\n\n challenge_length\n The length in bytes of the client's challenge to the server to\n authenticate itself. Historically, permissible values are between\n 16 and 32 bytes inclusive. When using the SSLv2 backward-\n compatible handshake the client SHOULD use a 32-byte challenge.\n\n cipher_specs\n This is a list of all CipherSpecs the client is willing and able\n to use. In addition to the 2.0 cipher specs defined in [SSL2],\n this includes the TLS cipher suites normally sent in\n ClientHello.cipher_suites, with each cipher suite prefixed by a\n zero byte. For example, the TLS cipher suite {0x00,0x0A} would be\n sent as {0x00,0x00,0x0A}.\n\n session_id\n This field MUST be empty.\n\n challenge\n Corresponds to ClientHello.random. If the challenge length is\n less than 32, the TLS server will pad the data with leading (note:\n not trailing) zero bytes to make it 32 bytes long.\n\n Note: Requests to resume a TLS session MUST use a TLS client hello.\n\nE.3. Avoiding Man-in-the-Middle Version Rollback\n\n When TLS clients fall back to Version 2.0 compatibility mode, they\n MUST use special PKCS#1 block formatting. This is done so that TLS\n servers will reject Version 2.0 sessions with TLS-capable clients.\n\n When a client negotiates SSL 2.0 but also supports TLS, it MUST set\n the right-hand (least-significant) 8 random bytes of the PKCS padding\n (not including the terminal null of the padding) for the RSA\n encryption of the ENCRYPTED-KEY-DATA field of the CLIENT-MASTER-KEY\n to 0x03 (the other padding bytes are random).\n\n When a TLS-capable server negotiates SSL 2.0 it SHOULD, after\n decrypting the ENCRYPTED-KEY-DATA field, check that these 8 padding\n bytes are 0x03. If they are not, the server SHOULD generate a random\n value for SECRET-KEY-DATA, and continue the handshake (which will\n eventually fail since the keys will not match). Note that reporting\n the error situation to the client could make the server vulnerable to\n attacks described in [BLEI].\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 90]\n\f\nRFC 5246 TLS August 2008\n\n\nAppendix F. Security Analysis\n\n The TLS protocol is designed to establish a secure connection between\n a client and a server communicating over an insecure channel. This\n document makes several traditional assumptions, including that\n attackers have substantial computational resources and cannot obtain\n secret information from sources outside the protocol. Attackers are\n assumed to have the ability to capture, modify, delete, replay, and\n otherwise tamper with messages sent over the communication channel.\n This appendix outlines how TLS has been designed to resist a variety\n of attacks.\n\nF.1. Handshake Protocol\n\n The handshake protocol is responsible for selecting a cipher spec and\n generating a master secret, which together comprise the primary\n cryptographic parameters associated with a secure session. The\n handshake protocol can also optionally authenticate parties who have\n certificates signed by a trusted certificate authority.\n\nF.1.1. Authentication and Key Exchange\n\n TLS supports three authentication modes: authentication of both\n parties, server authentication with an unauthenticated client, and\n total anonymity. Whenever the server is authenticated, the channel\n is secure against man-in-the-middle attacks, but completely anonymous\n sessions are inherently vulnerable to such attacks. Anonymous\n servers cannot authenticate clients. If the server is authenticated,\n its certificate message must provide a valid certificate chain\n leading to an acceptable certificate authority. Similarly,\n authenticated clients must supply an acceptable certificate to the\n server. Each party is responsible for verifying that the other's\n certificate is valid and has not expired or been revoked.\n\n The general goal of the key exchange process is to create a\n pre_master_secret known to the communicating parties and not to\n attackers. The pre_master_secret will be used to generate the\n master_secret (see Section 8.1). The master_secret is required to\n generate the Finished messages, encryption keys, and MAC keys (see\n Sections 7.4.9 and 6.3). By sending a correct Finished message,\n parties thus prove that they know the correct pre_master_secret.\n\nF.1.1.1. Anonymous Key Exchange\n\n Completely anonymous sessions can be established using Diffie-Hellman\n for key exchange. The server's public parameters are contained in\n the server key exchange message, and the client's are sent in the\n\n\n\n\nDierks & Rescorla Standards Track [Page 91]\n\f\nRFC 5246 TLS August 2008\n\n\n client key exchange message. Eavesdroppers who do not know the\n private values should not be able to find the Diffie-Hellman result\n (i.e., the pre_master_secret).\n\n Warning: Completely anonymous connections only provide protection\n against passive eavesdropping. Unless an independent tamper-proof\n channel is used to verify that the Finished messages were not\n replaced by an attacker, server authentication is required in\n environments where active man-in-the-middle attacks are a concern.\n\nF.1.1.2. RSA Key Exchange and Authentication\n\n With RSA, key exchange and server authentication are combined. The\n public key is contained in the server's certificate. Note that\n compromise of the server's static RSA key results in a loss of\n confidentiality for all sessions protected under that static key.\n TLS users desiring Perfect Forward Secrecy should use DHE cipher\n suites. The damage done by exposure of a private key can be limited\n by changing one's private key (and certificate) frequently.\n\n After verifying the server's certificate, the client encrypts a\n pre_master_secret with the server's public key. By successfully\n decoding the pre_master_secret and producing a correct Finished\n message, the server demonstrates that it knows the private key\n corresponding to the server certificate.\n\n When RSA is used for key exchange, clients are authenticated using\n the certificate verify message (see Section 7.4.8). The client signs\n a value derived from all preceding handshake messages. These\n handshake messages include the server certificate, which binds the\n signature to the server, and ServerHello.random, which binds the\n signature to the current handshake process.\n\nF.1.1.3. Diffie-Hellman Key Exchange with Authentication\n\n When Diffie-Hellman key exchange is used, the server can either\n supply a certificate containing fixed Diffie-Hellman parameters or\n use the server key exchange message to send a set of temporary\n Diffie-Hellman parameters signed with a DSA or RSA certificate.\n Temporary parameters are hashed with the hello.random values before\n signing to ensure that attackers do not replay old parameters. In\n either case, the client can verify the certificate or signature to\n ensure that the parameters belong to the server.\n\n If the client has a certificate containing fixed Diffie-Hellman\n parameters, its certificate contains the information required to\n complete the key exchange. Note that in this case the client and\n server will generate the same Diffie-Hellman result (i.e.,\n\n\n\nDierks & Rescorla Standards Track [Page 92]\n\f\nRFC 5246 TLS August 2008\n\n\n pre_master_secret) every time they communicate. To prevent the\n pre_master_secret from staying in memory any longer than necessary,\n it should be converted into the master_secret as soon as possible.\n Client Diffie-Hellman parameters must be compatible with those\n supplied by the server for the key exchange to work.\n\n If the client has a standard DSA or RSA certificate or is\n unauthenticated, it sends a set of temporary parameters to the server\n in the client key exchange message, then optionally uses a\n certificate verify message to authenticate itself.\n\n If the same DH keypair is to be used for multiple handshakes, either\n because the client or server has a certificate containing a fixed DH\n keypair or because the server is reusing DH keys, care must be taken\n to prevent small subgroup attacks. Implementations SHOULD follow the\n guidelines found in [SUBGROUP].\n\n Small subgroup attacks are most easily avoided by using one of the\n DHE cipher suites and generating a fresh DH private key (X) for each\n handshake. If a suitable base (such as 2) is chosen, g^X mod p can\n be computed very quickly; therefore, the performance cost is\n minimized. Additionally, using a fresh key for each handshake\n provides Perfect Forward Secrecy. Implementations SHOULD generate a\n new X for each handshake when using DHE cipher suites.\n\n Because TLS allows the server to provide arbitrary DH groups, the\n client should verify that the DH group is of suitable size as defined\n by local policy. The client SHOULD also verify that the DH public\n exponent appears to be of adequate size. [KEYSIZ] provides a useful\n guide to the strength of various group sizes. The server MAY choose\n to assist the client by providing a known group, such as those\n defined in [IKEALG] or [MODP]. These can be verified by simple\n comparison.\n\nF.1.2. Version Rollback Attacks\n\n Because TLS includes substantial improvements over SSL Version 2.0,\n attackers may try to make TLS-capable clients and servers fall back\n to Version 2.0. This attack can occur if (and only if) two TLS-\n capable parties use an SSL 2.0 handshake.\n\n Although the solution using non-random PKCS #1 block type 2 message\n padding is inelegant, it provides a reasonably secure way for Version\n 3.0 servers to detect the attack. This solution is not secure\n against attackers who can brute-force the key and substitute a new\n ENCRYPTED-KEY-DATA message containing the same key (but with normal\n padding) before the application-specified wait threshold has expired.\n Altering the padding of the least-significant 8 bytes of the PKCS\n\n\n\nDierks & Rescorla Standards Track [Page 93]\n\f\nRFC 5246 TLS August 2008\n\n\n padding does not impact security for the size of the signed hashes\n and RSA key lengths used in the protocol, since this is essentially\n equivalent to increasing the input block size by 8 bytes.\n\nF.1.3. Detecting Attacks Against the Handshake Protocol\n\n An attacker might try to influence the handshake exchange to make the\n parties select different encryption algorithms than they would\n normally choose.\n\n For this attack, an attacker must actively change one or more\n handshake messages. If this occurs, the client and server will\n compute different values for the handshake message hashes. As a\n result, the parties will not accept each others' Finished messages.\n Without the master_secret, the attacker cannot repair the Finished\n messages, so the attack will be discovered.\n\nF.1.4. Resuming Sessions\n\n When a connection is established by resuming a session, new\n ClientHello.random and ServerHello.random values are hashed with the\n session's master_secret. Provided that the master_secret has not\n been compromised and that the secure hash operations used to produce\n the encryption keys and MAC keys are secure, the connection should be\n secure and effectively independent from previous connections.\n Attackers cannot use known encryption keys or MAC secrets to\n compromise the master_secret without breaking the secure hash\n operations.\n\n Sessions cannot be resumed unless both the client and server agree.\n If either party suspects that the session may have been compromised,\n or that certificates may have expired or been revoked, it should\n force a full handshake. An upper limit of 24 hours is suggested for\n session ID lifetimes, since an attacker who obtains a master_secret\n may be able to impersonate the compromised party until the\n corresponding session ID is retired. Applications that may be run in\n relatively insecure environments should not write session IDs to\n stable storage.\n\nF.2. Protecting Application Data\n\n The master_secret is hashed with the ClientHello.random and\n ServerHello.random to produce unique data encryption keys and MAC\n secrets for each connection.\n\n Outgoing data is protected with a MAC before transmission. To\n prevent message replay or modification attacks, the MAC is computed\n from the MAC key, the sequence number, the message length, the\n\n\n\nDierks & Rescorla Standards Track [Page 94]\n\f\nRFC 5246 TLS August 2008\n\n\n message contents, and two fixed character strings. The message type\n field is necessary to ensure that messages intended for one TLS\n record layer client are not redirected to another. The sequence\n number ensures that attempts to delete or reorder messages will be\n detected. Since sequence numbers are 64 bits long, they should never\n overflow. Messages from one party cannot be inserted into the\n other's output, since they use independent MAC keys. Similarly, the\n server write and client write keys are independent, so stream cipher\n keys are used only once.\n\n If an attacker does break an encryption key, all messages encrypted\n with it can be read. Similarly, compromise of a MAC key can make\n message-modification attacks possible. Because MACs are also\n encrypted, message-alteration attacks generally require breaking the\n encryption algorithm as well as the MAC.\n\n Note: MAC keys may be larger than encryption keys, so messages can\n remain tamper resistant even if encryption keys are broken.\n\nF.3. Explicit IVs\n\n [CBCATT] describes a chosen plaintext attack on TLS that depends on\n knowing the IV for a record. Previous versions of TLS [TLS1.0] used\n the CBC residue of the previous record as the IV and therefore\n enabled this attack. This version uses an explicit IV in order to\n protect against this attack.\n\nF.4. Security of Composite Cipher Modes\n\n TLS secures transmitted application data via the use of symmetric\n encryption and authentication functions defined in the negotiated\n cipher suite. The objective is to protect both the integrity and\n confidentiality of the transmitted data from malicious actions by\n active attackers in the network. It turns out that the order in\n which encryption and authentication functions are applied to the data\n plays an important role for achieving this goal [ENCAUTH].\n\n The most robust method, called encrypt-then-authenticate, first\n applies encryption to the data and then applies a MAC to the\n ciphertext. This method ensures that the integrity and\n confidentiality goals are obtained with ANY pair of encryption and\n MAC functions, provided that the former is secure against chosen\n plaintext attacks and that the MAC is secure against chosen-message\n attacks. TLS uses another method, called authenticate-then-encrypt,\n in which first a MAC is computed on the plaintext and then the\n concatenation of plaintext and MAC is encrypted. This method has\n been proven secure for CERTAIN combinations of encryption functions\n and MAC functions, but it is not guaranteed to be secure in general.\n\n\n\nDierks & Rescorla Standards Track [Page 95]\n\f\nRFC 5246 TLS August 2008\n\n\n In particular, it has been shown that there exist perfectly secure\n encryption functions (secure even in the information-theoretic sense)\n that combined with any secure MAC function, fail to provide the\n confidentiality goal against an active attack. Therefore, new cipher\n suites and operation modes adopted into TLS need to be analyzed under\n the authenticate-then-encrypt method to verify that they achieve the\n stated integrity and confidentiality goals.\n\n Currently, the security of the authenticate-then-encrypt method has\n been proven for some important cases. One is the case of stream\n ciphers in which a computationally unpredictable pad of the length of\n the message, plus the length of the MAC tag, is produced using a\n pseudorandom generator and this pad is exclusive-ORed with the\n concatenation of plaintext and MAC tag. The other is the case of CBC\n mode using a secure block cipher. In this case, security can be\n shown if one applies one CBC encryption pass to the concatenation of\n plaintext and MAC and uses a new, independent, and unpredictable IV\n for each new pair of plaintext and MAC. In versions of TLS prior to\n 1.1, CBC mode was used properly EXCEPT that it used a predictable IV\n in the form of the last block of the previous ciphertext. This made\n TLS open to chosen plaintext attacks. This version of the protocol\n is immune to those attacks. For exact details in the encryption\n modes proven secure, see [ENCAUTH].\n\nF.5. Denial of Service\n\n TLS is susceptible to a number of denial-of-service (DoS) attacks.\n In particular, an attacker who initiates a large number of TCP\n connections can cause a server to consume large amounts of CPU for\n doing RSA decryption. However, because TLS is generally used over\n TCP, it is difficult for the attacker to hide his point of origin if\n proper TCP SYN randomization is used [SEQNUM] by the TCP stack.\n\n Because TLS runs over TCP, it is also susceptible to a number of DoS\n attacks on individual connections. In particular, attackers can\n forge RSTs, thereby terminating connections, or forge partial TLS\n records, thereby causing the connection to stall. These attacks\n cannot in general be defended against by a TCP-using protocol.\n Implementors or users who are concerned with this class of attack\n should use IPsec AH [AH] or ESP [ESP].\n\nF.6. Final Notes\n\n For TLS to be able to provide a secure connection, both the client\n and server systems, keys, and applications must be secure. In\n addition, the implementation must be free of security errors.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 96]\n\f\nRFC 5246 TLS August 2008\n\n\n The system is only as strong as the weakest key exchange and\n authentication algorithm supported, and only trustworthy\n cryptographic functions should be used. Short public keys and\n anonymous servers should be used with great caution. Implementations\n and users must be careful when deciding which certificates and\n certificate authorities are acceptable; a dishonest certificate\n authority can do tremendous damage.\n\nNormative References\n\n [AES] National Institute of Standards and Technology,\n \"Specification for the Advanced Encryption Standard (AES)\"\n FIPS 197. November 26, 2001.\n\n [3DES] National Institute of Standards and Technology,\n \"Recommendation for the Triple Data Encryption Algorithm\n (TDEA) Block Cipher\", NIST Special Publication 800-67, May\n 2004.\n\n [DSS] NIST FIPS PUB 186-2, \"Digital Signature Standard\",\n National Institute of Standards and Technology, U.S.\n Department of Commerce, 2000.\n\n [HMAC] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC: Keyed-\n Hashing for Message Authentication\", RFC 2104, February\n 1997.\n\n [MD5] Rivest, R., \"The MD5 Message-Digest Algorithm\", RFC 1321,\n April 1992.\n\n [PKCS1] Jonsson, J. and B. Kaliski, \"Public-Key Cryptography\n Standards (PKCS) #1: RSA Cryptography Specifications\n Version 2.1\", RFC 3447, February 2003.\n\n [PKIX] Housley, R., Polk, W., Ford, W., and D. Solo, \"Internet\n X.509 Public Key Infrastructure Certificate and\n Certificate Revocation List (CRL) Profile\", RFC 3280,\n April 2002.\n\n [SCH] B. Schneier. \"Applied Cryptography: Protocols, Algorithms,\n and Source Code in C, 2nd ed.\", Published by John Wiley &\n Sons, Inc. 1996.\n\n [SHS] NIST FIPS PUB 180-2, \"Secure Hash Standard\", National\n Institute of Standards and Technology, U.S. Department of\n Commerce, August 2002.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 97]\n\f\nRFC 5246 TLS August 2008\n\n\n [REQ] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2434] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 2434,\n October 1998.\n\n [X680] ITU-T Recommendation X.680 (2002) | ISO/IEC 8824-1:2002,\n Information technology - Abstract Syntax Notation One\n (ASN.1): Specification of basic notation.\n\n [X690] ITU-T Recommendation X.690 (2002) | ISO/IEC 8825-1:2002,\n Information technology - ASN.1 encoding Rules:\n Specification of Basic Encoding Rules (BER), Canonical\n Encoding Rules (CER) and Distinguished Encoding Rules\n (DER).\n\nInformative References\n\n [AEAD] McGrew, D., \"An Interface and Algorithms for Authenticated\n Encryption\", RFC 5116, January 2008.\n\n [AH] Kent, S., \"IP Authentication Header\", RFC 4302, December\n 2005.\n\n [BLEI] Bleichenbacher D., \"Chosen Ciphertext Attacks against\n Protocols Based on RSA Encryption Standard PKCS #1\" in\n Advances in Cryptology -- CRYPTO'98, LNCS vol. 1462,\n pages: 1-12, 1998.\n\n [CBCATT] Moeller, B., \"Security of CBC Ciphersuites in SSL/TLS:\n Problems and Countermeasures\",\n http://www.openssl.org/~bodo/tls-cbc.txt.\n\n [CBCTIME] Canvel, B., Hiltgen, A., Vaudenay, S., and M. Vuagnoux,\n \"Password Interception in a SSL/TLS Channel\", Advances in\n Cryptology -- CRYPTO 2003, LNCS vol. 2729, 2003.\n\n [CCM] \"NIST Special Publication 800-38C: The CCM Mode for\n Authentication and Confidentiality\",\n http://csrc.nist.gov/publications/nistpubs/800-38C/\n SP800-38C.pdf\n\n [DES] National Institute of Standards and Technology, \"Data\n Encryption Standard (DES)\", FIPS PUB 46-3, October 1999.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 98]\n\f\nRFC 5246 TLS August 2008\n\n\n [DSS-3] NIST FIPS PUB 186-3 Draft, \"Digital Signature Standard\",\n National Institute of Standards and Technology, U.S.\n Department of Commerce, 2006.\n\n [ECDSA] American National Standards Institute, \"Public Key\n Cryptography for the Financial Services Industry: The\n Elliptic Curve Digital Signature Algorithm (ECDSA)\", ANS\n X9.62-2005, November 2005.\n\n [ENCAUTH] Krawczyk, H., \"The Order of Encryption and Authentication\n for Protecting Communications (Or: How Secure is SSL?)\",\n Crypto 2001.\n\n [ESP] Kent, S., \"IP Encapsulating Security Payload (ESP)\", RFC\n 4303, December 2005.\n\n [FI06] Hal Finney, \"Bleichenbacher's RSA signature forgery based\n on implementation error\", [redacted-email] mailing\n list, 27 August 2006, http://www.imc.org/ietf-openpgp/\n mail-archive/msg14307.html.\n\n [GCM] Dworkin, M., NIST Special Publication 800-38D,\n \"Recommendation for Block Cipher Modes of Operation:\n Galois/Counter Mode (GCM) and GMAC\", November 2007.\n\n [IKEALG] Schiller, J., \"Cryptographic Algorithms for Use in the\n Internet Key Exchange Version 2 (IKEv2)\", RFC 4307,\n December 2005.\n\n [KEYSIZ] Orman, H. and P. Hoffman, \"Determining Strengths For\n Public Keys Used For Exchanging Symmetric Keys\", BCP 86,\n RFC 3766, April 2004.\n\n [KPR03] Klima, V., Pokorny, O., Rosa, T., \"Attacking RSA-based\n Sessions in SSL/TLS\", http://eprint.iacr.org/2003/052/,\n March 2003.\n\n [MODP] Kivinen, T. and M. Kojo, \"More Modular Exponential (MODP)\n Diffie-Hellman groups for Internet Key Exchange (IKE)\",\n RFC 3526, May 2003.\n\n [PKCS6] RSA Laboratories, \"PKCS #6: RSA Extended Certificate\n Syntax Standard\", version 1.5, November 1993.\n\n [PKCS7] RSA Laboratories, \"PKCS #7: RSA Cryptographic Message\n Syntax Standard\", version 1.5, November 1993.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 99]\n\f\nRFC 5246 TLS August 2008\n\n\n [RANDOM] Eastlake, D., 3rd, Schiller, J., and S. Crocker,\n \"Randomness Requirements for Security\", BCP 106, RFC 4086,\n June 2005.\n\n [RFC3749] Hollenbeck, S., \"Transport Layer Security Protocol\n Compression Methods\", RFC 3749, May 2004.\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, April 2006.\n\n [RSA] R. Rivest, A. Shamir, and L. M. Adleman, \"A Method for\n Obtaining Digital Signatures and Public-Key\n Cryptosystems\", Communications of the ACM, v. 21, n. 2,\n Feb 1978, pp. 120-126.\n\n [SEQNUM] Bellovin, S., \"Defending Against Sequence Number Attacks\",\n RFC 1948, May 1996.\n\n [SSL2] Hickman, Kipp, \"The SSL Protocol\", Netscape Communications\n Corp., Feb 9, 1995.\n\n [SSL3] A. Freier, P. Karlton, and P. Kocher, \"The SSL 3.0\n Protocol\", Netscape Communications Corp., Nov 18, 1996.\n\n [SUBGROUP] Zuccherato, R., \"Methods for Avoiding the \"Small-Subgroup\"\n Attacks on the Diffie-Hellman Key Agreement Method for\n S/MIME\", RFC 2785, March 2000.\n\n [TCP] Postel, J., \"Transmission Control Protocol\", STD 7, RFC\n 793, September 1981.\n\n [TIMING] Boneh, D., Brumley, D., \"Remote timing attacks are\n practical\", USENIX Security Symposium 2003.\n\n [TLSAES] Chown, P., \"Advanced Encryption Standard (AES)\n Ciphersuites for Transport Layer Security (TLS)\", RFC\n 3268, June 2002.\n\n [TLSECC] Blake-Wilson, S., Bolyard, N., Gupta, V., Hawk, C., and B.\n Moeller, \"Elliptic Curve Cryptography (ECC) Cipher Suites\n for Transport Layer Security (TLS)\", RFC 4492, May 2006.\n\n [TLSEXT] Eastlake, D., 3rd, \"Transport Layer Security (TLS)\n Extensions: Extension Definitions\", Work in Progress,\n February 2008.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 100]\n\f\nRFC 5246 TLS August 2008\n\n\n [TLSPGP] Mavrogiannopoulos, N., \"Using OpenPGP Keys for Transport\n Layer Security (TLS) Authentication\", RFC 5081, November\n 2007.\n\n [TLSPSK] Eronen, P., Ed., and H. Tschofenig, Ed., \"Pre-Shared Key\n Ciphersuites for Transport Layer Security (TLS)\", RFC\n 4279, December 2005.\n\n [TLS1.0] Dierks, T. and C. Allen, \"The TLS Protocol Version 1.0\",\n RFC 2246, January 1999.\n\n [TLS1.1] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346, April 2006.\n\n [X501] ITU-T Recommendation X.501: Information Technology - Open\n Systems Interconnection - The Directory: Models, 1993.\n\n [XDR] Eisler, M., Ed., \"XDR: External Data Representation\n Standard\", STD 67, RFC 4506, May 2006.\n\nWorking Group Information\n\n The discussion list for the IETF TLS working group is located at the\n e-mail address <[redacted-email]>. Information on the group and\n information on how to subscribe to the list is at\n \n\n Archives of the list can be found at:\n \n\nContributors\n\n Christopher Allen (co-editor of TLS 1.0)\n Alacrity Ventures\n [redacted-email]\n\n Martin Abadi\n University of California, Santa Cruz\n [redacted-email]\n\n Steven M. Bellovin\n Columbia University\n [redacted-email]\n\n Simon Blake-Wilson\n BCI\n [redacted-email]\n\n\n\n\nDierks & Rescorla Standards Track [Page 101]\n\f\nRFC 5246 TLS August 2008\n\n\n Ran Canetti\n IBM\n [redacted-email]\n\n Pete Chown\n Skygate Technology Ltd\n [redacted-email]\n\n Taher Elgamal\n [redacted-email]\n Securify\n\n Pasi Eronen\n [redacted-email]\n Nokia\n\n Anil Gangolli\n [redacted-email]\n\n Kipp Hickman\n\n Alfred Hoenes\n\n David Hopwood\n Independent Consultant\n [redacted-email]\n\n Phil Karlton (co-author of SSLv3)\n\n Paul Kocher (co-author of SSLv3)\n Cryptography Research\n [redacted-email]\n\n Hugo Krawczyk\n IBM\n [redacted-email]\n\n Jan Mikkelsen\n Transactionware\n [redacted-email]\n\n Magnus Nystrom\n RSA Security\n [redacted-email]\n\n Robert Relyea\n Netscape Communications\n [redacted-email]\n\n\n\nDierks & Rescorla Standards Track [Page 102]\n\f\nRFC 5246 TLS August 2008\n\n\n Jim Roskind\n Netscape Communications\n [redacted-email]\n\n Michael Sabin\n\n Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n Tom Weinstein\n\n Tim Wright\n Vodafone\n [redacted-email]\n\nEditors' Addresses\n\n Tim Dierks\n Independent\n EMail: [redacted-email]\n\n Eric Rescorla\n RTFM, Inc.\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 103]\n\f\nRFC 5246 TLS August 2008\n\n\nFull Copyright Statement\n\n Copyright (C) The IETF Trust (2008).\n\n This document is subject to the rights, licenses and restrictions\n contained in BCP 78, and except as set forth therein, the authors\n retain all their rights.\n\n This document and the information contained herein are provided on an\n \"AS IS\" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\n OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\n THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\n OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\n THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\n WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\n\nIntellectual Property\n\n The IETF takes no position regarding the validity or scope of any\n Intellectual Property Rights or other rights that might be claimed to\n pertain to the implementation or use of the technology described in\n this document or the extent to which any license under such rights\n might or might not be available; nor does it represent that it has\n made any independent effort to identify any such rights. Information\n on the procedures with respect to rights in RFC documents can be\n found in BCP 78 and BCP 79.\n\n Copies of IPR disclosures made to the IETF Secretariat and any\n assurances of licenses to be made available, or the result of an\n attempt made to obtain a general license or permission for the use of\n such proprietary rights by implementers or users of this\n specification can be obtained from the IETF on-line IPR repository at\n http://www.ietf.org/ipr.\n\n The IETF invites any interested party to bring to its attention any\n copyrights, patents or patent applications, or other proprietary\n rights that may cover technology that may be required to implement\n this standard. Please address the information to the IETF at\n ietf-ipr@ietf.org.\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 104]", "text_sha256": "0ba07ad7c1e04ded76e38500ef83eaa39b002af6cc226329ff75c3e8859fed85"}, {"draft_references": [], "excluded_rfc_relation_targets": [], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5705 RTFM, Inc.", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "5705"}], "kind": "rfc", "occurred_at": "2010-03-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 1, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a", "record_id": "ietf:rfc:5705", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "rfc_number": 5705, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a", "source_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "temporal_semantics": "source_event_time", "text": "Internet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 5705 RTFM, Inc.\nCategory: Standards Track March 2010\nISSN: 2070-1721\n\n\n Keying Material Exporters for Transport Layer Security (TLS)\n\nAbstract\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n to perform key establishment but then use some of the keying material\n for their own purposes. This document describes a general mechanism\n for allowing that.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/5705.\n\nCopyright Notice\n\n Copyright (c) 2010 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2\n 2. Conventions Used In This Document . . . . . . . . . . . . . . . 3\n 3. Binding to Application Contexts . . . . . . . . . . . . . . . . 3\n 4. Exporter Definition . . . . . . . . . . . . . . . . . . . . . . 4\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . . 5\n 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 6\n 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . 6\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 7\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 7\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 7\n\n1. Introduction\n\n Note: The mechanism described in this document was previously known\n as \"TLS Extractors\" but was changed to avoid a name conflict\n with the use of the term \"Extractor\" in the cryptographic\n community.\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n [RFC5246] or Datagram TLS (DTLS) [RFC4347] to perform key\n establishment but then use some of the keying material for their own\n purposes. A typical example is DTLS-SRTP [DTLS-SRTP], a key\n management scheme for the Secure Real-time Transport Protocol (SRTP)\n that uses DTLS to perform a key exchange and negotiate the SRTP\n [RFC3711] protection suite and then uses the DTLS master_secret to\n generate the SRTP keys.\n\n These applications imply a need to be able to export keying material\n (later called Exported Keying Material or EKM) from TLS/DTLS to an\n application or protocol residing at an upper layer, and to securely\n agree on the upper-layer context where the keying material will be\n used. The mechanism for exporting the keying material has the\n following requirements:\n\n o Both client and server need to be able to export the same EKM\n value.\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n o EKM values should be indistinguishable from random data to\n attackers who don't know the master_secret.\n\n o It should be possible to export multiple EKM values from the same\n TLS/DTLS association.\n\n o Knowing one EKM value should not reveal any useful information\n about the master_secret or about other EKM values.\n\n The mechanism described in this document is intended to fulfill these\n requirements. This mechanism is compatible with all versions of TLS.\n\n2. Conventions Used In This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Binding to Application Contexts\n\n In addition to using an exporter to obtain keying material, an\n application using the keying material has to securely establish the\n upper-layer context where the keying material will be used. The\n details of this context depend on the application, but it could\n include things such as algorithms and parameters that will be used\n with the keys, identifier(s) for the endpoint(s) who will use the\n keys, identifier(s) for the session(s) where the keys will be used,\n and the lifetime(s) for the context and/or keys. At a minimum, there\n should be some mechanism for signaling that an exporter will be used.\n\n This specification does not mandate a single mechanism for agreeing\n on such context; instead, there are several possibilities that can be\n used (and can complement each other). For example:\n\n o Information about the upper-layer context can be included in the\n optional data after the exporter label (see Section 4).\n\n o Information about the upper-layer context can be exchanged in TLS\n extensions included in the ClientHello and ServerHello messages.\n This approach is used in [DTLS-SRTP]. The handshake messages are\n protected by the Finished messages, so once the handshake\n completes, the peers will have the same view of the information.\n Extensions also allow a limited form of negotiation: for example,\n the TLS client could propose several alternatives for some context\n parameters, and the TLS server could select one of them.\n\n o The upper-layer protocol can include its own handshake, which can\n be protected using the keys exported by TLS.\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n No matter how the context is agreed, it is required that it has one\n part that indicates which application will use the exported keys.\n This part is the disambiguating label string (see Section 4).\n\n It is important to note that just embedding TLS messages in the\n upper-layer protocol may not automatically secure all the important\n context information, since the upper-layer messages are not covered\n by TLS Finished messages.\n\n4. Exporter Definition\n\n The output of the exporter is intended to be used in a single scope,\n which is associated with the TLS session, the label, and the context\n value.\n\n The exporter takes three input values:\n\n o a disambiguating label string,\n\n o a per-association context value provided by the application using\n the exporter, and\n\n o a length value.\n\n If no context is provided, it then computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random\n )[length]\n\n If context is provided, it computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random +\n context_value_length + context_value\n )[length]\n\n Where PRF is the TLS Pseudorandom Function in use for the session.\n The output is a pseudorandom bit string of length bytes generated\n from the master_secret. (This construction allows for\n interoperability with older exporter-type constructions which do not\n use context values, e.g., [RFC5281]).\n\n Labels here have the same definition as in TLS, i.e., an ASCII string\n with no terminating NULL. Label values beginning with \"EXPERIMENTAL\"\n MAY be used for private use without registration. All other label\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n values MUST be registered via Specification Required as described by\n RFC 5226 [RFC5226]. Note that exporter labels have the potential to\n collide with existing PRF labels. In order to prevent this, labels\n SHOULD begin with \"EXPORTER\". This is not a MUST because there are\n existing uses that have labels which do not begin with this prefix.\n\n The context value allows the application using the exporter to mix\n its own data with the TLS PRF for the exporter output. One example\n of where this might be useful is an authentication setting where the\n client credentials are valid for more than one identity; the context\n value could then be used to mix the expected identity into the keying\n material, thus preventing substitution attacks. The context value\n length is encoded as an unsigned, 16-bit quantity (uint16; see\n [RFC5246], Section 4.4) representing the length of the context value.\n The context MAY be zero length. Because the context value is mixed\n with the master_secret via the PRF, it is safe to mix confidential\n information into the exporter, provided that the master_secret will\n not be known to the attacker.\n\n5. Security Considerations\n\n The prime security requirement for exporter outputs is that they be\n independent. More formally, after a particular TLS session, if an\n adversary is allowed to choose multiple (label, context value) pairs\n and is given the output of the PRF for those values, the attacker is\n still unable to distinguish between the output of the PRF for a\n (label, context value) pair (different from the ones that it\n submitted) and a random value of the same length. In particular,\n there may be settings, such as the one described in Section 4, where\n the attacker can control the context value; such an attacker MUST NOT\n be able to predict the output of the exporter. Similarly, an\n attacker who does not know the master secret should not be able to\n distinguish valid exporter outputs from random values. The current\n set of TLS PRFs is believed to meet this objective, provided the\n master secret is randomly generated.\n\n Because an exporter produces the same value if applied twice with the\n same label to the same master_secret, it is critical that two EKM\n values generated with the same label not be used for two different\n purposes -- hence, the requirement for IANA registration. However,\n because exporters depend on the TLS PRF, it is not a threat to the\n use of an EKM value generated from one label to reveal an EKM value\n generated from another label.\n\n With certain TLS cipher suites, the TLS master secret is not\n necessarily unique to a single TLS session. In particular, with RSA\n key exchange, a malicious party acting as TLS server in one session\n and as TLS client in another session can cause those two sessions to\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n have the same TLS master secret (though the sessions must be\n established simultaneously to get adequate control of the Random\n values). Applications using the EKM need to consider this in how\n they use the EKM; in some cases, requiring the use of other cipher\n suites (such as those using a Diffie-Hellman key exchange) may be\n advisable.\n\n Designing a secure mechanism that uses exporters is not necessarily\n straightforward. This document only provides the exporter mechanism,\n but the problem of agreeing on the surrounding context and the\n meaning of the information passed to and from the exporter remains.\n Any new uses of the exporter mechanism should be subject to careful\n review.\n\n6. IANA Considerations\n\n IANA has created a TLS Exporter Label registry for this purpose. The\n initial contents of the registry are given below:\n\n Value Reference Note\n ----------------------------- --------- ----\n client finished [RFC5246] (1)\n server finished [RFC5246] (1)\n master secret [RFC5246] (1)\n key expansion [RFC5246] (1)\n client EAP encryption [RFC5216]\n ttls keying material [RFC5281]\n ttls challenge [RFC5281]\n\n Note: (1) These entries are reserved and MUST NOT be used for the\n purpose described in RFC 5705, in order to avoid confusion with\n similar, but distinct, use in RFC 5246.\n\n Future values are allocated via the RFC 5226 Specification Required\n policy. The label is a string consisting of printable ASCII\n characters. IANA MUST also verify that one label is not a prefix of\n any other label. For example, labels \"key\" or \"master secretary\" are\n forbidden.\n\n7. Acknowledgments\n\n Thanks to Pasi Eronen for valuable comments and for the contents of\n the IANA section and Section 3. Thanks to David McGrew for helpful\n discussion of the security considerations and to Vijay Gurbani and\n Alfred Hoenes for editorial comments.\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246,\n August 2008.\n\n8.2. Informative References\n\n [DTLS-SRTP] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for Secure\n Real-time Transport Protocol (SRTP)\", Work in Progress,\n February 2009.\n\n [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and\n K. Norrman, \"The Secure Real-time Transport Protocol\n (SRTP)\", RFC 3711, March 2004.\n\n [RFC4347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security\", RFC 4347, April 2006.\n\n [RFC5216] Simon, D., Aboba, B., and R. Hurst, \"The EAP-TLS\n Authentication Protocol\", RFC 5216, March 2008.\n\n [RFC5281] Funk, P. and S. Blake-Wilson, \"Extensible Authentication\n Protocol Tunneled Transport Layer Security Authenticated\n Protocol Version 0 (EAP-TTLSv0)\", RFC 5281, August 2008.\n\nAuthor's Address\n\n Eric Rescorla\n RTFM, Inc.\n 2064 Edgewood Drive\n Palo Alto, CA 94303\n USA\n\n EMail: [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 7]", "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"draft_references": [], "excluded_rfc_relation_targets": [{"kind": "obsoletes", "reason": "not_requested", "rfc_number": 2011}, {"kind": "obsoletes", "reason": "not_requested", "rfc_number": 4366}], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6066 Huawei", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "6066"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 4366 January 2011", "fact_id": "obsoletes:4366", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "4366"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 4366 January 2011", "fact_id": "obsoletes:2011", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "2011"}], "kind": "rfc", "occurred_at": "2011-01-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 3, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0", "record_id": "ietf:rfc:6066", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "rfc_number": 6066, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0", "source_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "temporal_semantics": "source_event_time", "text": "Internet Engineering Task Force (IETF) D. Eastlake 3rd\nRequest for Comments: 6066 Huawei\nObsoletes: 4366 January 2011\nCategory: Standards Track\nISSN: 2070-1721\n\n\n Transport Layer Security (TLS) Extensions: Extension Definitions\n\nAbstract\n\n This document provides specifications for existing TLS extensions.\n It is a companion document for RFC 5246, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\". The extensions specified are\n server_name, max_fragment_length, client_certificate_url,\n trusted_ca_keys, truncated_hmac, and status_request.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6066.\n\nCopyright Notice\n\n Copyright (c) 2011 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n\n\n\n\n\nEastlake Standards Track [Page 1]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction ....................................................3\n 1.1. Specific Extensions Covered ................................3\n 1.2. Conventions Used in This Document ..........................5\n 2. Extensions to the Handshake Protocol ............................5\n 3. Server Name Indication ..........................................6\n 4. Maximum Fragment Length Negotiation .............................8\n 5. Client Certificate URLs .........................................9\n 6. Trusted CA Indication ..........................................12\n 7. Truncated HMAC .................................................13\n 8. Certificate Status Request .....................................14\n 9. Error Alerts ...................................................16\n 10. IANA Considerations ...........................................17\n 10.1. pkipath MIME Type Registration ...........................17\n 10.2. Reference for TLS Alerts, TLS HandshakeTypes, and\n ExtensionTypes ...........................................19\n 11. Security Considerations .......................................19\n 11.1. Security Considerations for server_name ..................19\n 11.2. Security Considerations for max_fragment_length ..........20\n 11.3. Security Considerations for client_certificate_url .......20\n 11.4. Security Considerations for trusted_ca_keys ..............21\n 11.5. Security Considerations for truncated_hmac ...............21\n 11.6. Security Considerations for status_request ...............22\n 12. Normative References ..........................................22\n 13. Informative References ........................................23\n Appendix A. Changes from RFC 4366 .................................24\n Appendix B. Acknowledgements ......................................25\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 2]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1. Introduction\n\n The Transport Layer Security (TLS) Protocol Version 1.2 is specified\n in [RFC5246]. That specification includes the framework for\n extensions to TLS, considerations in designing such extensions (see\n Section 7.4.1.4 of [RFC5246]), and IANA Considerations for the\n allocation of new extension code points; however, it does not specify\n any particular extensions other than Signature Algorithms (see\n Section 7.4.1.4.1 of [RFC5246]).\n\n This document provides the specifications for existing TLS\n extensions. It is, for the most part, the adaptation and editing of\n material from RFC 4366, which covered TLS extensions for TLS 1.0 (RFC\n 2246) and TLS 1.1 (RFC 4346).\n\n1.1. Specific Extensions Covered\n\n The extensions described here focus on extending the functionality\n provided by the TLS protocol message formats. Other issues, such as\n the addition of new cipher suites, are deferred.\n\n The extension types defined in this document are:\n\n enum {\n server_name(0), max_fragment_length(1),\n client_certificate_url(2), trusted_ca_keys(3),\n truncated_hmac(4), status_request(5), (65535)\n } ExtensionType;\n\n Specifically, the extensions described in this document:\n\n - Allow TLS clients to provide to the TLS server the name of the\n server they are contacting. This functionality is desirable in\n order to facilitate secure connections to servers that host\n multiple 'virtual' servers at a single underlying network address.\n\n - Allow TLS clients and servers to negotiate the maximum fragment\n length to be sent. This functionality is desirable as a result of\n memory constraints among some clients, and bandwidth constraints\n among some access networks.\n\n - Allow TLS clients and servers to negotiate the use of client\n certificate URLs. This functionality is desirable in order to\n conserve memory on constrained clients.\n\n\n\n\n\n\n\nEastlake Standards Track [Page 3]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - Allow TLS clients to indicate to TLS servers which certification\n authority (CA) root keys they possess. This functionality is\n desirable in order to prevent multiple handshake failures\n involving TLS clients that are only able to store a small number\n of CA root keys due to memory limitations.\n\n - Allow TLS clients and servers to negotiate the use of truncated\n Message Authentication Codes (MACs). This functionality is\n desirable in order to conserve bandwidth in constrained access\n networks.\n\n - Allow TLS clients and servers to negotiate that the server sends\n the client certificate status information (e.g., an Online\n Certificate Status Protocol (OCSP) [RFC2560] response) during a\n TLS handshake. This functionality is desirable in order to avoid\n sending a Certificate Revocation List (CRL) over a constrained\n access network and therefore saving bandwidth.\n\n TLS clients and servers may use the extensions described in this\n document. The extensions are designed to be backwards compatible,\n meaning that TLS clients that support the extensions can talk to TLS\n servers that do not support the extensions, and vice versa.\n\n Note that any messages associated with these extensions that are sent\n during the TLS handshake MUST be included in the hash calculations\n involved in \"Finished\" messages.\n\n Note also that all the extensions defined in this document are\n relevant only when a session is initiated. A client that requests\n session resumption does not in general know whether the server will\n accept this request, and therefore it SHOULD send the same extensions\n as it would send if it were not attempting resumption. When a client\n includes one or more of the defined extension types in an extended\n client hello while requesting session resumption:\n\n - The server name indication extension MAY be used by the server\n when deciding whether or not to resume a session as described in\n Section 3.\n\n - If the resumption request is denied, the use of the extensions is\n negotiated as normal.\n\n - If, on the other hand, the older session is resumed, then the\n server MUST ignore the extensions and send a server hello\n containing none of the extension types. In this case, the\n functionality of these extensions negotiated during the original\n session initiation is applied to the resumed session.\n\n\n\n\nEastlake Standards Track [Page 4]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1.2. Conventions Used in This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n [RFC2119].\n\n2. Extensions to the Handshake Protocol\n\n This document specifies the use of two new handshake messages,\n \"CertificateURL\" and \"CertificateStatus\". These messages are\n described in Sections 5 and 8, respectively. The new handshake\n message structure therefore becomes:\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), certificate_url(21), certificate_status(22),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case certificate_url: CertificateURL;\n case certificate_status: CertificateStatus;\n } body;\n } Handshake;\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 5]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n3. Server Name Indication\n\n TLS does not provide a mechanism for a client to tell a server the\n name of the server it is contacting. It may be desirable for clients\n to provide this information to facilitate secure connections to\n servers that host multiple 'virtual' servers at a single underlying\n network address.\n\n In order to provide any of the server names, clients MAY include an\n extension of type \"server_name\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"ServerNameList\" where:\n\n struct {\n NameType name_type;\n select (name_type) {\n case host_name: HostName;\n } name;\n } ServerName;\n\n enum {\n host_name(0), (255)\n } NameType;\n\n opaque HostName<1..2^16-1>;\n\n struct {\n ServerName server_name_list<1..2^16-1>\n } ServerNameList;\n\n The ServerNameList MUST NOT contain more than one name of the same\n name_type. If the server understood the ClientHello extension but\n does not recognize the server name, the server SHOULD take one of two\n actions: either abort the handshake by sending a fatal-level\n unrecognized_name(112) alert or continue the handshake. It is NOT\n RECOMMENDED to send a warning-level unrecognized_name(112) alert,\n because the client's behavior in response to warning-level alerts is\n unpredictable. If there is a mismatch between the server name used\n by the client application and the server name of the credential\n chosen by the server, this mismatch will become apparent when the\n client application performs the server endpoint identification, at\n which point the client application will have to decide whether to\n proceed with the communication. TLS implementations are encouraged\n to make information available to application callers about warning-\n level alerts that were received or sent during a TLS handshake. Such\n information can be useful for diagnostic purposes.\n\n\n\n\n\nEastlake Standards Track [Page 6]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note: Earlier versions of this specification permitted multiple\n names of the same name_type. In practice, current client\n implementations only send one name, and the client cannot\n necessarily find out which name the server selected. Multiple\n names of the same name_type are therefore now prohibited.\n\n Currently, the only server names supported are DNS hostnames;\n however, this does not imply any dependency of TLS on DNS, and other\n name types may be added in the future (by an RFC that updates this\n document). The data structure associated with the host_name NameType\n is a variable-length vector that begins with a 16-bit length. For\n backward compatibility, all future data structures associated with\n new NameTypes MUST begin with a 16-bit length field. TLS MAY treat\n provided server names as opaque data and pass the names and types to\n the application.\n\n \"HostName\" contains the fully qualified DNS hostname of the server,\n as understood by the client. The hostname is represented as a byte\n string using ASCII encoding without a trailing dot. This allows the\n support of internationalized domain names through the use of A-labels\n defined in [RFC5890]. DNS hostnames are case-insensitive. The\n algorithm to compare hostnames is described in [RFC5890], Section\n 2.3.2.4.\n\n Literal IPv4 and IPv6 addresses are not permitted in \"HostName\".\n\n It is RECOMMENDED that clients include an extension of type\n \"server_name\" in the client hello whenever they locate a server by a\n supported name type.\n\n A server that receives a client hello containing the \"server_name\"\n extension MAY use the information contained in the extension to guide\n its selection of an appropriate certificate to return to the client,\n and/or other aspects of security policy. In this event, the server\n SHALL include an extension of type \"server_name\" in the (extended)\n server hello. The \"extension_data\" field of this extension SHALL be\n empty.\n\n When the server is deciding whether or not to accept a request to\n resume a session, the contents of a server_name extension MAY be used\n in the lookup of the session in the session cache. The client SHOULD\n include the same server_name extension in the session resumption\n request as it did in the full handshake that established the session.\n A server that implements this extension MUST NOT accept the request\n to resume the session if the server_name extension contains a\n different name. Instead, it proceeds with a full handshake to\n establish a new session. When resuming a session, the server MUST\n NOT include a server_name extension in the server hello.\n\n\n\nEastlake Standards Track [Page 7]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n If an application negotiates a server name using an application\n protocol and then upgrades to TLS, and if a server_name extension is\n sent, then the extension SHOULD contain the same name that was\n negotiated in the application protocol. If the server_name is\n established in the TLS session handshake, the client SHOULD NOT\n attempt to request a different server name at the application layer.\n\n4. Maximum Fragment Length Negotiation\n\n Without this extension, TLS specifies a fixed maximum plaintext\n fragment length of 2^14 bytes. It may be desirable for constrained\n clients to negotiate a smaller maximum fragment length due to memory\n limitations or bandwidth limitations.\n\n In order to negotiate smaller maximum fragment lengths, clients MAY\n include an extension of type \"max_fragment_length\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain:\n\n enum{\n 2^9(1), 2^10(2), 2^11(3), 2^12(4), (255)\n } MaxFragmentLength;\n\n whose value is the desired maximum fragment length. The allowed\n values for this field are: 2^9, 2^10, 2^11, and 2^12.\n\n Servers that receive an extended client hello containing a\n \"max_fragment_length\" extension MAY accept the requested maximum\n fragment length by including an extension of type\n \"max_fragment_length\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL contain a\n \"MaxFragmentLength\" whose value is the same as the requested maximum\n fragment length.\n\n If a server receives a maximum fragment length negotiation request\n for a value other than the allowed values, it MUST abort the\n handshake with an \"illegal_parameter\" alert. Similarly, if a client\n receives a maximum fragment length negotiation response that differs\n from the length it requested, it MUST also abort the handshake with\n an \"illegal_parameter\" alert.\n\n Once a maximum fragment length other than 2^14 has been successfully\n negotiated, the client and server MUST immediately begin fragmenting\n messages (including handshake messages) to ensure that no fragment\n larger than the negotiated length is sent. Note that TLS already\n requires clients and servers to support fragmentation of handshake\n messages.\n\n\n\n\nEastlake Standards Track [Page 8]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n The negotiated length applies for the duration of the session\n including session resumptions.\n\n The negotiated length limits the input that the record layer may\n process without fragmentation (that is, the maximum value of\n TLSPlaintext.length; see [RFC5246], Section 6.2.1). Note that the\n output of the record layer may be larger. For example, if the\n negotiated length is 2^9=512, then, when using currently defined\n cipher suites (those defined in [RFC5246] and [RFC2712]) and null\n compression, the record-layer output can be at most 805 bytes: 5\n bytes of headers, 512 bytes of application data, 256 bytes of\n padding, and 32 bytes of MAC. This means that in this event a TLS\n record-layer peer receiving a TLS record-layer message larger than\n 805 bytes MUST discard the message and send a \"record_overflow\"\n alert, without decrypting the message. When this extension is used\n with Datagram Transport Layer Security (DTLS), implementations SHOULD\n NOT generate record_overflow alerts unless the packet passes message\n authentication.\n\n5. Client Certificate URLs\n\n Without this extension, TLS specifies that when client authentication\n is performed, client certificates are sent by clients to servers\n during the TLS handshake. It may be desirable for constrained\n clients to send certificate URLs in place of certificates, so that\n they do not need to store their certificates and can therefore save\n memory.\n\n In order to negotiate sending certificate URLs to a server, clients\n MAY include an extension of type \"client_certificate_url\" in the\n (extended) client hello. The \"extension_data\" field of this\n extension SHALL be empty.\n\n (Note that it is necessary to negotiate the use of client certificate\n URLs in order to avoid \"breaking\" existing TLS servers.)\n\n Servers that receive an extended client hello containing a\n \"client_certificate_url\" extension MAY indicate that they are willing\n to accept certificate URLs by including an extension of type\n \"client_certificate_url\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n After negotiation of the use of client certificate URLs has been\n successfully completed (by exchanging hellos including\n \"client_certificate_url\" extensions), clients MAY send a\n \"CertificateURL\" message in place of a \"Certificate\" message as\n follows (see also Section 2):\n\n\n\n\nEastlake Standards Track [Page 9]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n enum {\n individual_certs(0), pkipath(1), (255)\n } CertChainType;\n\n struct {\n CertChainType type;\n URLAndHash url_and_hash_list<1..2^16-1>;\n } CertificateURL;\n\n struct {\n opaque url<1..2^16-1>;\n unint8 padding;\n opaque SHA1Hash[20];\n } URLAndHash;\n\n Here, \"url_and_hash_list\" contains a sequence of URLs and hashes.\n Each \"url\" MUST be an absolute URI reference according to [RFC3986]\n that can be immediately used to fetch the certificate(s).\n\n When X.509 certificates are used, there are two possibilities:\n\n - If CertificateURL.type is \"individual_certs\", each URL refers to a\n single DER-encoded X.509v3 certificate, with the URL for the\n client's certificate first.\n\n - If CertificateURL.type is \"pkipath\", the list contains a single\n URL referring to a DER-encoded certificate chain, using the type\n PkiPath described in Section 10.1.\n\n When any other certificate format is used, the specification that\n describes use of that format in TLS should define the encoding format\n of certificates or certificate chains, and any constraint on their\n ordering.\n\n The \"padding\" byte MUST be 0x01. It is present to make the structure\n backwards compatible.\n\n The hash corresponding to each URL is the SHA-1 hash of the\n certificate or certificate chain (in the case of X.509 certificates,\n the DER-encoded certificate or the DER-encoded PkiPath).\n\n Note that when a list of URLs for X.509 certificates is used, the\n ordering of URLs is the same as that used in the TLS Certificate\n message (see [RFC5246], Section 7.4.2), but opposite to the order in\n which certificates are encoded in PkiPath. In either case, the self-\n signed root certificate MAY be omitted from the chain, under the\n assumption that the server must already possess it in order to\n validate it.\n\n\n\nEastlake Standards Track [Page 10]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Servers receiving \"CertificateURL\" SHALL attempt to retrieve the\n client's certificate chain from the URLs and then process the\n certificate chain as usual. A cached copy of the content of any URL\n in the chain MAY be used, provided that the SHA-1 hash matches the\n hash of the cached copy.\n\n Servers that support this extension MUST support the 'http' URI\n scheme for certificate URLs and MAY support other schemes. Use of\n other schemes than 'http', 'https', or 'ftp' may create unexpected\n problems.\n\n If the protocol used is HTTP, then the HTTP server can be configured\n to use the Cache-Control and Expires directives described in\n [RFC2616] to specify whether and for how long certificates or\n certificate chains should be cached.\n\n The TLS server MUST NOT follow HTTP redirects when retrieving the\n certificates or certificate chain. The URLs used in this extension\n MUST NOT be chosen to depend on such redirects.\n\n If the protocol used to retrieve certificates or certificate chains\n returns a MIME-formatted response (as HTTP does), then the following\n MIME Content-Types SHALL be used: when a single X.509v3 certificate\n is returned, the Content-Type is \"application/pkix-cert\" [RFC2585],\n and when a chain of X.509v3 certificates is returned, the Content-\n Type is \"application/pkix-pkipath\" (Section 10.1).\n\n The server MUST check that the SHA-1 hash of the contents of the\n object retrieved from that URL (after decoding any MIME Content-\n Transfer-Encoding) matches the given hash. If any retrieved object\n does not have the correct SHA-1 hash, the server MUST abort the\n handshake with a bad_certificate_hash_value(114) alert. This alert\n is always fatal.\n\n Clients may choose to send either \"Certificate\" or \"CertificateURL\"\n after successfully negotiating the option to send certificate URLs.\n The option to send a certificate is included to provide flexibility\n to clients possessing multiple certificates.\n\n If a server is unable to obtain certificates in a given\n CertificateURL, it MUST send a fatal certificate_unobtainable(111)\n alert if it requires the certificates to complete the handshake. If\n the server does not require the certificates, then the server\n continues the handshake. The server MAY send a warning-level alert\n in this case. Clients receiving such an alert SHOULD log the alert\n and continue with the handshake if possible.\n\n\n\n\n\nEastlake Standards Track [Page 11]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n6. Trusted CA Indication\n\n Constrained clients that, due to memory limitations, possess only a\n small number of CA root keys may wish to indicate to servers which\n root keys they possess, in order to avoid repeated handshake\n failures.\n\n In order to indicate which CA root keys they possess, clients MAY\n include an extension of type \"trusted_ca_keys\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain \"TrustedAuthorities\" where:\n\n struct {\n TrustedAuthority trusted_authorities_list<0..2^16-1>;\n } TrustedAuthorities;\n\n struct {\n IdentifierType identifier_type;\n select (identifier_type) {\n case pre_agreed: struct {};\n case key_sha1_hash: SHA1Hash;\n case x509_name: DistinguishedName;\n case cert_sha1_hash: SHA1Hash;\n } identifier;\n } TrustedAuthority;\n\n enum {\n pre_agreed(0), key_sha1_hash(1), x509_name(2),\n cert_sha1_hash(3), (255)\n } IdentifierType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n Here, \"TrustedAuthorities\" provides a list of CA root key identifiers\n that the client possesses. Each CA root key is identified via\n either:\n\n - \"pre_agreed\": no CA root key identity supplied.\n\n - \"key_sha1_hash\": contains the SHA-1 hash of the CA root key. For\n Digital Signature Algorithm (DSA) and Elliptic Curve Digital\n Signature Algorithm (ECDSA) keys, this is the hash of the\n \"subjectPublicKey\" value. For RSA keys, the hash is of the big-\n endian byte string representation of the modulus without any\n initial zero-valued bytes. (This copies the key hash formats\n deployed in other environments.)\n\n\n\n\n\nEastlake Standards Track [Page 12]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - \"x509_name\": contains the DER-encoded X.509 DistinguishedName of\n the CA.\n\n - \"cert_sha1_hash\": contains the SHA-1 hash of a DER-encoded\n Certificate containing the CA root key.\n\n Note that clients may include none, some, or all of the CA root keys\n they possess in this extension.\n\n Note also that it is possible that a key hash or a Distinguished Name\n alone may not uniquely identify a certificate issuer (for example, if\n a particular CA has multiple key pairs). However, here we assume\n this is the case following the use of Distinguished Names to identify\n certificate issuers in TLS.\n\n The option to include no CA root keys is included to allow the client\n to indicate possession of some pre-defined set of CA root keys.\n\n Servers that receive a client hello containing the \"trusted_ca_keys\"\n extension MAY use the information contained in the extension to guide\n their selection of an appropriate certificate chain to return to the\n client. In this event, the server SHALL include an extension of type\n \"trusted_ca_keys\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n7. Truncated HMAC\n\n Currently defined TLS cipher suites use the MAC construction HMAC\n [RFC2104] to authenticate record-layer communications. In TLS, the\n entire output of the hash function is used as the MAC tag. However,\n it may be desirable in constrained environments to save bandwidth by\n truncating the output of the hash function to 80 bits when forming\n MAC tags.\n\n In order to negotiate the use of 80-bit truncated HMAC, clients MAY\n include an extension of type \"truncated_hmac\" in the extended client\n hello. The \"extension_data\" field of this extension SHALL be empty.\n\n Servers that receive an extended hello containing a \"truncated_hmac\"\n extension MAY agree to use a truncated HMAC by including an extension\n of type \"truncated_hmac\", with empty \"extension_data\", in the\n extended server hello.\n\n Note that if new cipher suites are added that do not use HMAC, and\n the session negotiates one of these cipher suites, this extension\n will have no effect. It is strongly recommended that any new cipher\n suites using other MACs consider the MAC size an integral part of the\n\n\n\n\nEastlake Standards Track [Page 13]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n cipher suite definition, taking into account both security and\n bandwidth considerations.\n\n If HMAC truncation has been successfully negotiated during a TLS\n handshake, and the negotiated cipher suite uses HMAC, both the client\n and the server pass this fact to the TLS record layer along with the\n other negotiated security parameters. Subsequently during the\n session, clients and servers MUST use truncated HMACs, calculated as\n specified in [RFC2104]. That is, SecurityParameters.mac_length is 10\n bytes, and only the first 10 bytes of the HMAC output are transmitted\n and checked. Note that this extension does not affect the\n calculation of the pseudo-random function (PRF) as part of\n handshaking or key derivation.\n\n The negotiated HMAC truncation size applies for the duration of the\n session including session resumptions.\n\n8. Certificate Status Request\n\n Constrained clients may wish to use a certificate-status protocol\n such as OCSP [RFC2560] to check the validity of server certificates,\n in order to avoid transmission of CRLs and therefore save bandwidth\n on constrained networks. This extension allows for such information\n to be sent in the TLS handshake, saving roundtrips and resources.\n\n In order to indicate their desire to receive certificate status\n information, clients MAY include an extension of type\n \"status_request\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"CertificateStatusRequest\" where:\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n } request;\n } CertificateStatusRequest;\n\n enum { ocsp(1), (255) } CertificateStatusType;\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n\n\n\nEastlake Standards Track [Page 14]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n In the OCSPStatusRequest, the \"ResponderIDs\" provides a list of OCSP\n responders that the client trusts. A zero-length \"responder_id_list\"\n sequence has the special meaning that the responders are implicitly\n known to the server, e.g., by prior arrangement. \"Extensions\" is a\n DER encoding of OCSP request extensions.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC2560]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a zero-length ASN.1 SEQUENCE, which is not\n valid for the \"Extensions\" type).\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement).\n\n Servers that receive a client hello containing the \"status_request\"\n extension MAY return a suitable certificate status response to the\n client along with their certificate. If OCSP is requested, they\n SHOULD use the information contained in the extension when selecting\n an OCSP responder and SHOULD include request_extensions in the OCSP\n request.\n\n Servers return a certificate response along with their certificate by\n sending a \"CertificateStatus\" message immediately after the\n \"Certificate\" message (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message, then the server MUST have included an\n extension of type \"status_request\" with empty \"extension_data\" in the\n extended server hello. The \"CertificateStatus\" message is conveyed\n using the handshake message type \"certificate_status\" as follows (see\n also Section 2):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<1..2^24-1>;\n\n An \"ocsp_response\" contains a complete, DER-encoded OCSP response\n (using the ASN.1 type OCSPResponse defined in [RFC2560]). Only one\n OCSP response may be sent.\n\n\n\n\nEastlake Standards Track [Page 15]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if has received a \"status_request\" extension in the\n client hello message and has sent a \"status_request\" extension in the\n server hello message.\n\n Note in addition that a server MUST NOT send the \"CertificateStatus\"\n message unless it received a \"status_request\" extension in the client\n hello message and sent a \"status_request\" extension in the server\n hello message.\n\n Clients requesting an OCSP response and receiving an OCSP response in\n a \"CertificateStatus\" message MUST check the OCSP response and abort\n the handshake if the response is not satisfactory with\n bad_certificate_status_response(113) alert. This alert is always\n fatal.\n\n9. Error Alerts\n\n Four new error alerts are defined for use with the TLS extensions\n defined in this document. To avoid \"breaking\" existing clients and\n servers, these alerts MUST NOT be sent unless the sending party has\n received an extended hello message from the party they are\n communicating with. These error alerts are conveyed using the\n following syntax. The new alerts are the last four, as indicated by\n the comments on the same line as the error alert number.\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n /* 41 is not defined, for historical reasons */\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction(60),\n protocol_version(70),\n insufficient_security(71),\n\n\n\nEastlake Standards Track [Page 16]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n certificate_unobtainable(111), /* new */\n unrecognized_name(112), /* new */\n bad_certificate_status_response(113), /* new */\n bad_certificate_hash_value(114), /* new */\n (255)\n } AlertDescription;\n\n \"certificate_unobtainable\" is described in Section 5.\n \"unrecognized_name\" is described in Section 3.\n \"bad_certificate_status_response\" is described in Section 8.\n \"bad_certificate_hash_value\" is described in Section 5.\n\n10. IANA Considerations\n\n IANA Considerations for TLS extensions and the creation of a registry\n are covered in Section 12 of [RFC5246] except for the registration of\n MIME type application/pkix-pkipath, which appears below.\n\n The IANA TLS extensions and MIME type application/pkix-pkipath\n registry entries that reference RFC 4366 have been updated to\n reference this document.\n\n10.1. pkipath MIME Type Registration\n\n MIME media type name: application\n MIME subtype name: pkix-pkipath\n Required parameters: none\n\n Optional parameters: version (default value is \"1\")\n\n Encoding considerations:\n Binary; this MIME type is a DER encoding of the ASN.1 type\n PkiPath, defined as follows:\n PkiPath ::= SEQUENCE OF Certificate\n PkiPath is used to represent a certification path. Within the\n sequence, the order of certificates is such that the subject of\n the first certificate is the issuer of the second certificate,\n etc.\n This is identical to the definition published in [X509-4th-TC1];\n note that it is different from that in [X509-4th].\n\n All Certificates MUST conform to [RFC5280]. (This should be\n interpreted as a requirement to encode only PKIX-conformant\n certificates using this type. It does not necessarily require\n\n\n\nEastlake Standards Track [Page 17]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n that all certificates that are not strictly PKIX-conformant must\n be rejected by relying parties, although the security consequences\n of accepting any such certificates should be considered\n carefully.)\n\n DER (as opposed to BER) encoding MUST be used. If this type is\n sent over a 7-bit transport, base64 encoding SHOULD be used.\n\n Security considerations:\n The security considerations of [X509-4th] and [RFC5280] (or any\n updates to them) apply, as well as those of any protocol that uses\n this type (e.g., TLS).\n\n Note that this type only specifies a certificate chain that can be\n assessed for validity according to the relying party's existing\n configuration of trusted CAs; it is not intended to be used to\n specify any change to that configuration.\n\n Interoperability considerations:\n No specific interoperability problems are known with this type,\n but for recommendations relating to X.509 certificates in general,\n see [RFC5280].\n\n Published specification: This document and [RFC5280].\n\n Applications that use this media type:\n TLS. It may also be used by other protocols or for general\n interchange of PKIX certificate chains.\n\n Additional information:\n Magic number(s): DER-encoded ASN.1 can be easily recognized.\n Further parsing is required to distinguish it from other ASN.1\n types.\n File extension(s): .pkipath\n Macintosh File Type Code(s): not specified\n\n Person & email address to contact for further information:\n Magnus Nystrom <[redacted-email]>\n\n Intended usage: COMMON\n\n Change controller: IESG <[redacted-email]>\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 18]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n10.2. Reference for TLS Alerts, TLS HandshakeTypes, and ExtensionTypes\n\n The following values in the TLS Alert Registry have been updated to\n reference this document:\n\n 111 certificate_unobtainable\n 112 unrecognized_name\n 113 bad_certificate_status_response\n 114 bad_certificate_hash_value\n\n The following values in the TLS HandshakeType Registry have been\n updated to reference this document:\n\n 21 certificate_url\n 22 certificate_status\n\n The following ExtensionType values have been updated to reference\n this document:\n\n 0 server_name\n 1 max_fragment_length\n 2 client_certificate_url\n 3 trusted_ca_keys\n 4 truncated_hmac\n 5 status_request\n\n11. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security Considerations for particular extensions\n specified in this document are given below.\n\n In general, implementers should continue to monitor the state of the\n art and address any weaknesses identified.\n\n11.1. Security Considerations for server_name\n\n If a single server hosts several domains, then clearly it is\n necessary for the owners of each domain to ensure that this satisfies\n their security needs. Apart from this, server_name does not appear\n to introduce significant security issues.\n\n Since it is possible for a client to present a different server_name\n in the application protocol, application server implementations that\n rely upon these names being the same MUST check to make sure the\n client did not present a different name in the application protocol.\n\n\n\n\n\nEastlake Standards Track [Page 19]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Implementations MUST ensure that a buffer overflow does not occur,\n whatever the values of the length fields in server_name.\n\n11.2. Security Considerations for max_fragment_length\n\n The maximum fragment length takes effect immediately, including for\n handshake messages. However, that does not introduce any security\n complications that are not already present in TLS, since TLS requires\n implementations to be able to handle fragmented handshake messages.\n\n Note that, as described in Section 4, once a non-null cipher suite\n has been activated, the effective maximum fragment length depends on\n the cipher suite and compression method, as well as on the negotiated\n max_fragment_length. This must be taken into account when sizing\n buffers and checking for buffer overflow.\n\n11.3. Security Considerations for client_certificate_url\n\n Support for client_certificate_url involves the server's acting as a\n client in another URI-scheme-dependent protocol. The server\n therefore becomes subject to many of the same security concerns that\n clients of the URI scheme are subject to, with the added concern that\n the client can attempt to prompt the server to connect to some\n (possibly weird-looking) URL.\n\n In general, this issue means that an attacker might use the server to\n indirectly attack another host that is vulnerable to some security\n flaw. It also introduces the possibility of denial-of-service\n attacks in which an attacker makes many connections to the server,\n each of which results in the server's attempting a connection to the\n target of the attack.\n\n Note that the server may be behind a firewall or otherwise able to\n access hosts that would not be directly accessible from the public\n Internet. This could exacerbate the potential security and denial-\n of-service problems described above, as well as allow the existence\n of internal hosts to be confirmed when they would otherwise be\n hidden.\n\n The detailed security concerns involved will depend on the URI\n schemes supported by the server. In the case of HTTP, the concerns\n are similar to those that apply to a publicly accessible HTTP proxy\n server. In the case of HTTPS, loops and deadlocks may be created,\n and this should be addressed. In the case of FTP, attacks arise that\n are similar to FTP bounce attacks.\n\n\n\n\n\n\nEastlake Standards Track [Page 20]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n As a result of this issue, it is RECOMMENDED that the\n client_certificate_url extension should have to be specifically\n enabled by a server administrator, rather than be enabled by default.\n It is also RECOMMENDED that URI schemes be enabled by the\n administrator individually, and only a minimal set of schemes be\n enabled. Unusual protocols that offer limited security or whose\n security is not well understood SHOULD be avoided.\n\n As discussed in [RFC3986], URLs that specify ports other than the\n default may cause problems, as may very long URLs (which are more\n likely to be useful in exploiting buffer overflow bugs).\n\n This extension continues to use SHA-1 (as in RFC 4366) and does not\n provide algorithm agility. The property required of SHA-1 in this\n case is second pre-image resistance, not collision resistance.\n Furthermore, even if second pre-image attacks against SHA-1 are found\n in the future, an attack against client_certificate_url would require\n a second pre-image that is accepted as a valid certificate by the\n server and contains the same public key.\n\n Also note that HTTP caching proxies are common on the Internet, and\n some proxies do not check for the latest version of an object\n correctly. If a request using HTTP (or another caching protocol)\n goes through a misconfigured or otherwise broken proxy, the proxy may\n return an out-of-date response.\n\n11.4. Security Considerations for trusted_ca_keys\n\n Potentially, the CA root keys a client possesses could be regarded as\n confidential information. As a result, the CA root key indication\n extension should be used with care.\n\n The use of the SHA-1 certificate hash alternative ensures that each\n certificate is specified unambiguously. This context does not\n require a cryptographic hash function, so the use of SHA-1 is\n considered acceptable, and no algorithm agility is provided.\n\n11.5. Security Considerations for truncated_hmac\n\n It is possible that truncated MACs are weaker than \"un-truncated\"\n MACs. However, no significant weaknesses are currently known or\n expected to exist for HMAC with MD5 or SHA-1, truncated to 80 bits.\n\n Note that the output length of a MAC need not be as long as the\n length of a symmetric cipher key, since forging of MAC values cannot\n be done off-line: in TLS, a single failed MAC guess will cause the\n immediate termination of the TLS session.\n\n\n\n\nEastlake Standards Track [Page 21]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Since the MAC algorithm only takes effect after all handshake\n messages that affect extension parameters have been authenticated by\n the hashes in the Finished messages, it is not possible for an active\n attacker to force negotiation of the truncated HMAC extension where\n it would not otherwise be used (to the extent that the handshake\n authentication is secure). Therefore, in the event that any security\n problems were found with truncated HMAC in the future, if either the\n client or the server for a given session were updated to take the\n problem into account, it would be able to veto use of this extension.\n\n11.6. Security Considerations for status_request\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC2560] for further details.\n\n12. Normative References\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC:\n Keyed-Hashing for Message Authentication\", RFC 2104,\n February 1997.\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and\n C. Adams, \"X.509 Internet Public Key Infrastructure\n Online Certificate Status Protocol - OCSP\", RFC 2560,\n June 1999.\n\n [RFC2585] Housley, R. and P. Hoffman, \"Internet X.509 Public Key\n Infrastructure Operational Protocols: FTP and HTTP\",\n RFC 2585, May 1999.\n\n [RFC2616] Fielding, R., Gettys, J., Mogul, J., Frystyk, H.,\n Masinter, L., Leach, P., and T. Berners-Lee,\n \"Hypertext Transfer Protocol -- HTTP/1.1\", RFC 2616,\n June 1999.\n\n [RFC3986] Berners-Lee, T., Fielding, R., and L. Masinter,\n \"Uniform Resource Identifier (URI): Generic Syntax\",\n STD 66, RFC 3986, January 2005.\n\n\n\nEastlake Standards Track [Page 22]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246, August\n 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation\n List (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC5890] Klensin, J., \"Internationalized Domain Names for\n Applications (IDNA): Definitions and Document\n Framework\", RFC 5890, August 2010.\n\n13. Informative References\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [X509-4th] ITU-T Recommendation X.509 (2000) | ISO/IEC\n 9594-8:2001, \"Information Systems - Open Systems\n Interconnection - The Directory: Public key and\n attribute certificate frameworks\".\n\n [X509-4th-TC1] ITU-T Recommendation X.509(2000) Corrigendum 1(2001) |\n ISO/IEC 9594-8:2001/Cor.1:2002, Technical Corrigendum\n 1 to ISO/IEC 9594:8:2001.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 23]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix A. Changes from RFC 4366\n\n The significant changes between RFC 4366 and this document are\n described below.\n\n RFC 4366 described both general extension mechanisms (for the TLS\n handshake and client and server hellos) as well as specific\n extensions. RFC 4366 was associated with RFC 4346, TLS 1.1. The\n client and server hello extension mechanisms have been moved into RFC\n 5246, TLS 1.2, so this document, which is associated with RFC 5246,\n includes only the handshake extension mechanisms and the specific\n extensions from RFC 4366. RFC 5246 also specifies the unknown\n extension error and new extension specification considerations, so\n that material has been removed from this document.\n\n The Server Name extension now specifies only ASCII representation,\n eliminating UTF-8. It is provided that the ServerNameList can\n contain more than only one name of any particular name_type. If a\n server name is provided but not recognized, the server should either\n continue the handshake without an error or send a fatal error.\n Sending a warning-level message is not recommended because client\n behavior will be unpredictable. Provision was added for the user\n using the server_name extension in deciding whether or not to resume\n a session. Furthermore, this extension should be the same in a\n session resumption request as it was in the full handshake that\n established the session. Such a resumption request must not be\n accepted if the server_name extension is different, but instead a\n full handshake must be done to possibly establish a new session.\n\n The Client Certificate URLs extension has been changed to make the\n presence of a hash mandatory.\n\n For the case of DTLS, the requirement to report an overflow of the\n negotiated maximum fragment length is made conditional on passing\n authentication.\n\n TLS servers are now prohibited from following HTTP redirects when\n retrieving certificates.\n\n The material was also re-organized in minor ways. For example,\n information as to which errors are fatal is moved from the \"Error\n Alerts\" section to the individual extension specifications.\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 24]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix B. Acknowledgements\n\n This document is based on material from RFC 4366 for which the\n authors were S. Blake-Wilson, M. Nystrom, D. Hopwood, J. Mikkelsen,\n and T. Wright. Other contributors include Joseph Salowey, Alexey\n Melnikov, Peter Saint-Andre, and Adrian Farrel.\n\nAuthor's Address\n\n Donald Eastlake 3rd\n Huawei\n 155 Beaver Street\n Milford, MA 01757 USA\n\n Phone: +1-508-333-2270\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 25]", "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}, {"draft_references": [], "excluded_rfc_relation_targets": [], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6961 June 2013", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "6961"}], "kind": "rfc", "occurred_at": "2013-06-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 1, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db", "record_id": "ietf:rfc:6961", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "rfc_number": 6961, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db", "source_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "temporal_semantics": "source_event_time", "text": "Internet Engineering Task Force (IETF) Y. Pettersen\nRequest for Comments: 6961 June 2013\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\n\nAbstract\n\n This document defines the Transport Layer Security (TLS) Certificate\n Status Version 2 Extension to allow clients to specify and support\n several certificate status methods. (The use of the Certificate\n Status extension is commonly referred to as \"OCSP stapling\".) Also\n defined is a new method based on the Online Certificate Status\n Protocol (OCSP) that servers can use to provide status information\n about not only the server's own certificate but also the status of\n intermediate certificates in the chain.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6961.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 1]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nCopyright Notice\n\n Copyright (c) 2013 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n1. Introduction\n\n The Transport Layer Security (TLS) Extension [RFC6066] framework\n defines, among other extensions, the Certificate Status extension\n (also referred to as \"OCSP stapling\") that clients can use to request\n the server's copy of the current status of its certificate. The\n benefits of this extension include a reduced number of roundtrips and\n network delays for the client to verify the status of the server's\n certificate and a reduced load on the certificate issuer's status\n response servers, thus solving a problem that can become significant\n when the issued certificate is presented by a frequently visited\n server.\n\n There are two problems with the existing Certificate Status\n extension. First, it does not provide functionality to request the\n status information about intermediate Certification Authority (CA)\n certificates, which means the client has to request status\n information through other methods, such as Certificate Revocation\n Lists (CRLs), introducing further delays. Second, the current format\n of the extension and requirements in the TLS protocol prevent a\n client from offering the server multiple status methods.\n\n\n\nPettersen Standards Track [Page 2]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n Many CAs are now issuing intermediate CA certificates that not only\n specify the publication point for their CRLs in a CRL Distribution\n Point [RFC5280] but also specify a URL for their OCSP [RFC6960]\n server in Authority Information Access [RFC5280]. Given that\n client-cached CRLs are frequently out of date, clients would benefit\n from using OCSP to access up-to-date status information about\n intermediate CA certificates. The benefit to the issuing CA is less\n clear, as providing the bandwidth for the OCSP responder can be\n costly, especially for CAs with many high-traffic subscriber sites,\n and this cost is a concern for many CAs. There are cases where OCSP\n requests for a single high-traffic site caused significant network\n problems for the issuing CA.\n\n Clients will benefit from the TLS server providing certificate status\n information regardless of type, not just for the server certificate\n but also for the intermediate CA certificates. Combining the status\n checks into one extension will reduce the roundtrips needed to\n complete the handshake by the client to just those needed for\n negotiating the TLS connection. Also, for the Certification\n Authorities, the load on their servers will depend on the number of\n certificates they have issued, not on the number of visitors to those\n sites. Additionally, using this extension significantly reduces\n privacy concerns around the clients informing the certificate issuer\n about which sites they are visiting.\n\n For such a new system to be introduced seamlessly, clients need to be\n able to indicate support for the existing OCSP Certificate Status\n method and a new multiple-OCSP mode.\n\n Unfortunately, the definition of the Certificate Status extension\n only allows a single Certificate Status extension to be defined in a\n single extension record in the handshake, and the TLS protocol\n [RFC5246] only allows a single record in the extension list for any\n given extension. This means that it is not possible for clients to\n indicate support for new methods while still supporting older\n methods, which would cause problems for interoperability between\n newer clients and older servers. This will not just be an issue for\n the multiple status request mode proposed above but also for any\n other future status methods that might be introduced. This will be\n true not just for the current PKIX infrastructure [RFC5280] but also\n for alternative PKI structures.\n\n The solution to this problem is to define a new extension,\n \"status_request_v2\", with an extended format that allows the client\n to indicate support for multiple status request methods. This is\n implemented using a list of CertificateStatusRequestItemV2 records in\n the extension record. As the server will select the single status\n\n\n\n\nPettersen Standards Track [Page 3]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n method based on the selected cipher suite and the certificate\n presented, no significant changes are needed in the server's\n extension format.\n\n1.1. Requirements Language\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [RFC2119].\n\n1.2. Presentation Language\n\n This document defines protocol structures using the same conventions\n and presentation language as defined in Section 4 of [RFC5246].\n\n2. Multiple Certificate Status Extension\n\n2.1. New Extension\n\n The extension defined by this document is indicated by\n \"status_request_v2\" in the ExtensionType enum (originally defined by\n [RFC6066]), which uses the following value:\n\n enum {\n status_request_v2(17), (65535)\n } ExtensionType;\n\n2.2. Multiple Certificate Status Request Record\n\n Clients that support a certificate status protocol like OCSP may send\n the \"status_request_v2\" extension to the server in order to use the\n TLS handshake to transfer such data instead of downloading it through\n separate connections. When using this extension, the\n \"extension_data\" field (defined in Section 7.4.1.4 of [RFC5246]) of\n the extension SHALL contain a CertificateStatusRequestListV2 where:\n\n struct {\n CertificateStatusType status_type;\n uint16 request_length; /* Length of request field in bytes */\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n case ocsp_multi: OCSPStatusRequest;\n } request;\n } CertificateStatusRequestItemV2;\n\n enum { ocsp(1), ocsp_multi(2), (255) } CertificateStatusType;\n\n\n\n\n\nPettersen Standards Track [Page 4]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n struct {\n CertificateStatusRequestItemV2\n certificate_status_req_list<1..2^16-1>;\n } CertificateStatusRequestListV2;\n\n In the OCSPStatusRequest (originally defined by [RFC6066]), the\n \"ResponderID\" provides a list of OCSP responders that the client\n trusts. A zero-length \"responder_id_list\" sequence has the special\n meaning that the responders are implicitly known to the server, e.g.,\n by prior arrangement, or are identified by the certificates used by\n the server. \"Extensions\" is a DER encoding [X.690] of the OCSP\n request extensions, and if the server supports the forwarding of OCSP\n request extensions, this value MUST be forwarded without\n modification.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC6960]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a DER-encoded zero-length ASN.1 SEQUENCE,\n which is not valid for the \"Extensions\" type).\n\n Servers that support a client's selection of responders using the\n ResponderID field could implement this selection by matching the\n responder ID values from the client's list with the ResponderIDs of\n known OCSP responders, either by using a binary compare of the values\n or a hash calculation and compare method.\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement). This\n has been clarified in [RFC6960].\n\n The items in the list of CertificateStatusRequestItemV2 entries are\n ordered according to the client's preference (favorite choice first).\n\n A server that receives a client hello containing the\n \"status_request_v2\" extension MAY return a suitable certificate\n status response message to the client along with the server's\n\n\n\nPettersen Standards Track [Page 5]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n certificate message. If OCSP is requested, it SHOULD use the\n information contained in the extension when selecting an OCSP\n responder and SHOULD include request_extensions in the OCSP request.\n\n The server returns a certificate status response along with its\n certificate by sending a \"CertificateStatus\" message (originally\n defined by [RFC6066]) immediately after the \"Certificate\" message\n (Section 7.4.2 of [RFC5246]) (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message in response to a \"status_request_v2\"\n request, then the server MUST have included an extension of type\n \"status_request_v2\" with empty \"extension_data\" in the extended\n server hello.\n\n The \"CertificateStatus\" message is conveyed using the handshake\n message type \"certificate_status\" (defined in [RFC6066]) as follows\n (updated from the definition in [RFC6066]):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n case ocsp_multi: OCSPResponseList;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<0..2^24-1>;\n\n struct {\n OCSPResponse ocsp_response_list<1..2^24-1>;\n } OCSPResponseList;\n\n An \"OCSPResponse\" element (originally defined by [RFC6066]) contains\n a complete, DER-encoded OCSP response (using the ASN.1 [X.680] type\n OCSPResponse defined in [RFC6960]). Only one OCSP response, with a\n length of at least one byte, may be sent for status_type \"ocsp\".\n\n An \"ocsp_response_list\" contains a list of \"OCSPResponse\" elements,\n as specified above, each containing the OCSP response for the\n matching corresponding certificate in the server's Certificate TLS\n handshake message. That is, the first entry is the OCSP response for\n the first certificate in the Certificate list, the second entry is\n the response for the second certificate, and so on. The list MAY\n contain fewer OCSP responses than there were certificates in the\n Certificate handshake message, but there MUST NOT be more responses\n than there were certificates in the list. Individual elements of the\n list MAY have a length of 0 (zero) bytes if the server does not have\n the OCSP response for that particular certificate stored, in which\n\n\n\nPettersen Standards Track [Page 6]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n case the client MUST act as if a response was not received for that\n particular certificate. If the client receives a\n \"ocsp_response_list\" that does not contain a response for one or more\n of the certificates in the completed certificate chain, the client\n SHOULD attempt to validate the certificate using an alternative\n retrieval method, such as downloading the relevant CRL; OCSP SHOULD\n in this situation only be used for the end-entity certificate, not\n intermediate CA certificates, for reasons stated above.\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if it has received a \"status_request_v2\" extension in\n the client hello message and has sent a \"status_request_v2\" extension\n in the server hello message. Additionally, note that a server MUST\n NOT send the \"CertificateStatus\" message unless it received either a\n \"status_request\" or \"status_request_v2\" extension in the client hello\n message and sent a corresponding \"status_request\" or\n \"status_request_v2\" extension in the server hello message.\n\n Clients requesting an OCSP response and receiving one or more OCSP\n responses in a \"CertificateStatus\" message MUST check the OCSP\n response(s) and abort the handshake if the response is a \"revoked\"\n status or other unacceptable responses (as determined by client\n policy) with a bad_certificate_status_response(113) alert. This\n alert is always fatal.\n\n If the OCSP response received from the server does not result in a\n definite \"good\" or \"revoked\" status, it is inconclusive. A TLS\n client in such a case MAY check the validity of the server\n certificate through other means, e.g., by directly querying the\n certificate issuer. If such processing still results in an\n inconclusive response, then the application using the TLS connection\n will have to decide whether to close the connection or not. Note\n that this problem cannot be decided by the generic TLS client code\n without information from the application. If the application doesn't\n provide any such information, then the client MUST abort the\n connection, since the server certificate has not been sufficiently\n validated.\n\n An example of where the application might wish to continue is with\n EAP-TLS (Extensible Authentication Protocol - TLS), where the\n application can use another mechanism to check the status of a\n certificate once it obtains network access. In this case, the\n application could have the client continue with the handshake, but it\n MUST NOT disclose a username and password until it has fully\n validated the server certificate.\n\n\n\n\n\n\nPettersen Standards Track [Page 7]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n3. IANA Considerations\n\n Section 2.1 defines the new TLS extension status_request_v2 (17)\n enum, which has been added to the \"ExtensionType Values\" list in the\n IANA \"Transport Layer Security (TLS) Extensions\" registry.\n\n Section 2.2 describes a TLS CertificateStatusType registry that is\n now maintained by IANA. The \"TLS Certificate Status Types\" registry\n has been created under the \"Transport Layer Security (TLS)\n Extensions\" registry. CertificateStatusType values are to be\n assigned via IETF Review as defined in [RFC5226]. The initial\n registry corresponds to the definition of \"CertificateStatusType\" in\n Section 2.2.\n\n Value Description Reference\n -----------------------------------------\n 0 Reserved [RFC6961]\n 1 ocsp [RFC6066] [RFC6961]\n 2 ocsp_multi [RFC6961]\n 3-255 Unassigned\n\n4. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security considerations for the particular extension\n specified in this document are given below. In general, implementers\n should continue to monitor the state of the art and address any\n weaknesses identified.\n\n4.1. Security Considerations for status_request_v2\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC6960] for further details.\n\n This extension allows the client to send arbitrary data to the\n server. The server implementers need to handle such data carefully\n to avoid introducing security vulnerabilities.\n\n The security considerations of [RFC6960] apply to OCSP requests and\n responses.\n\n\n\n\nPettersen Standards Track [Page 8]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n5. Acknowledgements\n\n This document is based on [RFC6066], authored by Donald Eastlake 3rd.\n\n6. References\n\n6.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246, August 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC6066] Eastlake, D., \"Transport Layer Security (TLS) Extensions:\n Extension Definitions\", RFC 6066, January 2011.\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, June 2013.\n\n [X.680] ITU-T Recommendation X.680 (2008) | ISO/IEC 8824-1:2008,\n \"Abstract Syntax Notation One (ASN.1): Specification of\n basic notation\", November 2008.\n\n [X.690] ITU-T Recommendation X.690 (2008) | ISO/IEC 8825-1:2008,\n \"ASN.1 encoding rules: Specification of Basic Encoding\n Rules (BER), Canonical Encoding Rules (CER) and\n Distinguished Encoding Rules (DER)\", November 2008.\n\n6.2. Informative References\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and C.\n Adams, \"X.509 Internet Public Key Infrastructure Online\n Certificate Status Protocol - OCSP\", RFC 2560, June 1999.\n\n\n\n\n\n\nPettersen Standards Track [Page 9]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nAuthor's Address\n\n Yngve N. Pettersen\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 10]", "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"draft_name": "draft-ietf-tls-tls13", "facts": [{"char_end": 454, "char_start": 431, "evidence_quote": "draft-ietf-tls-tls13-28", "fact_id": "draft_revision", "field": "draft_revision", "raw_char_end": 458, "raw_char_start": 435, "value": "draft-ietf-tls-tls13-28"}], "kind": "draft_revision", "occurred_at": "2018-03-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 85, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6", "record_id": "ietf:draft:draft-ietf-tls-tls13-28", "retrieval_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "revision": "28", "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6", "source_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "temporal_semantics": "source_event_time", "text": "Network Working Group E. Rescorla\nInternet-Draft RTFM, Inc.\nObsoletes: 5077, 5246, 6961 (if March 20, 2018\n approved)\nUpdates: 4492, 5705, 6066 (if approved)\nIntended status: Standards Track\nExpires: September 21, 2018\n\n\n The Transport Layer Security (TLS) Protocol Version 1.3\n draft-ietf-tls-tls13-28\n\nAbstract\n\n This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol. TLS allows client/server applications to communicate\n over the Internet in a way that is designed to prevent eavesdropping,\n tampering, and message forgery.\n\n This document updates RFCs 4492, 5705, and 6066 and it obsoletes RFCs\n 5077, 5246, and 6961. This document also specifies new requirements\n for TLS 1.2 implementations.\n\nStatus of This Memo\n\n This Internet-Draft is submitted in full conformance with the\n provisions of BCP 78 and BCP 79.\n\n Internet-Drafts are working documents of the Internet Engineering\n Task Force (IETF). Note that other groups may also distribute\n working documents as Internet-Drafts. The list of current Internet-\n Drafts is at http://datatracker.ietf.org/drafts/current/.\n\n Internet-Drafts are draft documents valid for a maximum of six months\n and may be updated, replaced, or obsoleted by other documents at any\n time. It is inappropriate to use Internet-Drafts as reference\n material or to cite them other than as \"work in progress.\"\n\n This Internet-Draft will expire on September 21, 2018.\n\nCopyright Notice\n\n Copyright (c) 2018 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n\n\n\nRescorla Expires September 21, 2018 [Page 1]\n\f\nInternet-Draft TLS March 2018\n\n\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 5\n 1.1. Conventions and Terminology . . . . . . . . . . . . . . . 6\n 1.2. Change Log . . . . . . . . . . . . . . . . . . . . . . . 7\n 1.3. Major Differences from TLS 1.2 . . . . . . . . . . . . . 16\n 1.4. Updates Affecting TLS 1.2 . . . . . . . . . . . . . . . . 17\n 2. Protocol Overview . . . . . . . . . . . . . . . . . . . . . . 18\n 2.1. Incorrect DHE Share . . . . . . . . . . . . . . . . . . . 21\n 2.2. Resumption and Pre-Shared Key (PSK) . . . . . . . . . . . 22\n 2.3. 0-RTT Data . . . . . . . . . . . . . . . . . . . . . . . 24\n 3. Presentation Language . . . . . . . . . . . . . . . . . . . . 26\n 3.1. Basic Block Size . . . . . . . . . . . . . . . . . . . . 26\n 3.2. Miscellaneous . . . . . . . . . . . . . . . . . . . . . . 26\n 3.3. Numbers . . . . . . . . . . . . . . . . . . . . . . . . . 27\n 3.4. Vectors . . . . . . . . . . . . . . . . . . . . . . . . . 27\n 3.5. Enumerateds . . . . . . . . . . . . . . . . . . . . . . . 28\n 3.6. Constructed Types . . . . . . . . . . . . . . . . . . . . 29\n 3.7. Constants . . . . . . . . . . . . . . . . . . . . . . . . 29\n 3.8. Variants . . . . . . . . . . . . . . . . . . . . . . . . 30\n 4. Handshake Protocol . . . . . . . . . . . . . . . . . . . . . 31\n 4.1. Key Exchange Messages . . . . . . . . . . . . . . . . . . 32\n 4.1.1. Cryptographic Negotiation . . . . . . . . . . . . . . 32\n 4.1.2. Client Hello . . . . . . . . . . . . . . . . . . . . 33\n 4.1.3. Server Hello . . . . . . . . . . . . . . . . . . . . 36\n 4.1.4. Hello Retry Request . . . . . . . . . . . . . . . . . 38\n 4.2. Extensions . . . . . . . . . . . . . . . . . . . . . . . 40\n 4.2.1. Supported Versions . . . . . . . . . . . . . . . . . 43\n 4.2.2. Cookie . . . . . . . . . . . . . . . . . . . . . . . 45\n\n\n\nRescorla Expires September 21, 2018 [Page 2]\n\f\nInternet-Draft TLS March 2018\n\n\n 4.2.3. Signature Algorithms . . . . . . . . . . . . . . . . 46\n 4.2.4. Certificate Authorities . . . . . . . . . . . . . . . 50\n 4.2.5. OID Filters . . . . . . . . . . . . . . . . . . . . . 50\n 4.2.6. Post-Handshake Client Authentication . . . . . . . . 51\n 4.2.7. Negotiated Groups . . . . . . . . . . . . . . . . . . 52\n 4.2.8. Key Share . . . . . . . . . . . . . . . . . . . . . . 53\n 4.2.9. Pre-Shared Key Exchange Modes . . . . . . . . . . . . 56\n 4.2.10. Early Data Indication . . . . . . . . . . . . . . . . 57\n 4.2.11. Pre-Shared Key Extension . . . . . . . . . . . . . . 60\n 4.3. Server Parameters . . . . . . . . . . . . . . . . . . . . 63\n 4.3.1. Encrypted Extensions . . . . . . . . . . . . . . . . 63\n 4.3.2. Certificate Request . . . . . . . . . . . . . . . . . 64\n 4.4. Authentication Messages . . . . . . . . . . . . . . . . . 65\n 4.4.1. The Transcript Hash . . . . . . . . . . . . . . . . . 66\n 4.4.2. Certificate . . . . . . . . . . . . . . . . . . . . . 67\n 4.4.3. Certificate Verify . . . . . . . . . . . . . . . . . 72\n 4.4.4. Finished . . . . . . . . . . . . . . . . . . . . . . 74\n 4.5. End of Early Data . . . . . . . . . . . . . . . . . . . . 75\n 4.6. Post-Handshake Messages . . . . . . . . . . . . . . . . . 76\n 4.6.1. New Session Ticket Message . . . . . . . . . . . . . 76\n 4.6.2. Post-Handshake Authentication . . . . . . . . . . . . 78\n 4.6.3. Key and IV Update . . . . . . . . . . . . . . . . . . 79\n 5. Record Protocol . . . . . . . . . . . . . . . . . . . . . . . 80\n 5.1. Record Layer . . . . . . . . . . . . . . . . . . . . . . 81\n 5.2. Record Payload Protection . . . . . . . . . . . . . . . . 83\n 5.3. Per-Record Nonce . . . . . . . . . . . . . . . . . . . . 85\n 5.4. Record Padding . . . . . . . . . . . . . . . . . . . . . 86\n 5.5. Limits on Key Usage . . . . . . . . . . . . . . . . . . . 87\n 6. Alert Protocol . . . . . . . . . . . . . . . . . . . . . . . 87\n 6.1. Closure Alerts . . . . . . . . . . . . . . . . . . . . . 89\n 6.2. Error Alerts . . . . . . . . . . . . . . . . . . . . . . 90\n 7. Cryptographic Computations . . . . . . . . . . . . . . . . . 93\n 7.1. Key Schedule . . . . . . . . . . . . . . . . . . . . . . 93\n 7.2. Updating Traffic Secrets . . . . . . . . . . . . . . . . 96\n 7.3. Traffic Key Calculation . . . . . . . . . . . . . . . . . 97\n 7.4. (EC)DHE Shared Secret Calculation . . . . . . . . . . . . 98\n 7.4.1. Finite Field Diffie-Hellman . . . . . . . . . . . . . 98\n 7.4.2. Elliptic Curve Diffie-Hellman . . . . . . . . . . . . 98\n 7.5. Exporters . . . . . . . . . . . . . . . . . . . . . . . . 99\n 8. 0-RTT and Anti-Replay . . . . . . . . . . . . . . . . . . . . 99\n 8.1. Single-Use Tickets . . . . . . . . . . . . . . . . . . . 101\n 8.2. Client Hello Recording . . . . . . . . . . . . . . . . . 101\n 8.3. Freshness Checks . . . . . . . . . . . . . . . . . . . . 102\n 9. Compliance Requirements . . . . . . . . . . . . . . . . . . . 104\n 9.1. Mandatory-to-Implement Cipher Suites . . . . . . . . . . 104\n 9.2. Mandatory-to-Implement Extensions . . . . . . . . . . . . 104\n 9.3. Protocol Invariants . . . . . . . . . . . . . . . . . . . 105\n 10. Security Considerations . . . . . . . . . . . . . . . . . . . 106\n\n\n\nRescorla Expires September 21, 2018 [Page 3]\n\f\nInternet-Draft TLS March 2018\n\n\n 11. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 107\n 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 108\n 12.1. Normative References . . . . . . . . . . . . . . . . . . 108\n 12.2. Informative References . . . . . . . . . . . . . . . . . 111\n Appendix A. State Machine . . . . . . . . . . . . . . . . . . . 119\n A.1. Client . . . . . . . . . . . . . . . . . . . . . . . . . 119\n A.2. Server . . . . . . . . . . . . . . . . . . . . . . . . . 120\n Appendix B. Protocol Data Structures and Constant Values . . . . 120\n B.1. Record Layer . . . . . . . . . . . . . . . . . . . . . . 121\n B.2. Alert Messages . . . . . . . . . . . . . . . . . . . . . 121\n B.3. Handshake Protocol . . . . . . . . . . . . . . . . . . . 123\n B.3.1. Key Exchange Messages . . . . . . . . . . . . . . . . 123\n B.3.2. Server Parameters Messages . . . . . . . . . . . . . 128\n B.3.3. Authentication Messages . . . . . . . . . . . . . . . 129\n B.3.4. Ticket Establishment . . . . . . . . . . . . . . . . 130\n B.3.5. Updating Keys . . . . . . . . . . . . . . . . . . . . 131\n B.4. Cipher Suites . . . . . . . . . . . . . . . . . . . . . . 131\n Appendix C. Implementation Notes . . . . . . . . . . . . . . . . 132\n C.1. Random Number Generation and Seeding . . . . . . . . . . 132\n C.2. Certificates and Authentication . . . . . . . . . . . . . 133\n C.3. Implementation Pitfalls . . . . . . . . . . . . . . . . . 133\n C.4. Client Tracking Prevention . . . . . . . . . . . . . . . 134\n C.5. Unauthenticated Operation . . . . . . . . . . . . . . . . 135\n Appendix D. Backward Compatibility . . . . . . . . . . . . . . . 135\n D.1. Negotiating with an older server . . . . . . . . . . . . 136\n D.2. Negotiating with an older client . . . . . . . . . . . . 137\n D.3. 0-RTT backwards compatibility . . . . . . . . . . . . . . 137\n D.4. Middlebox Compatibility Mode . . . . . . . . . . . . . . 137\n D.5. Backwards Compatibility Security Restrictions . . . . . . 138\n Appendix E. Overview of Security Properties . . . . . . . . . . 139\n E.1. Handshake . . . . . . . . . . . . . . . . . . . . . . . . 139\n E.1.1. Key Derivation and HKDF . . . . . . . . . . . . . . . 142\n E.1.2. Client Authentication . . . . . . . . . . . . . . . . 143\n E.1.3. 0-RTT . . . . . . . . . . . . . . . . . . . . . . . . 143\n E.1.4. Exporter Independence . . . . . . . . . . . . . . . . 143\n E.1.5. Post-Compromise Security . . . . . . . . . . . . . . 144\n E.1.6. External References . . . . . . . . . . . . . . . . . 144\n E.2. Record Layer . . . . . . . . . . . . . . . . . . . . . . 144\n E.2.1. External References . . . . . . . . . . . . . . . . . 145\n E.3. Traffic Analysis . . . . . . . . . . . . . . . . . . . . 145\n E.4. Side Channel Attacks . . . . . . . . . . . . . . . . . . 146\n E.5. Replay Attacks on 0-RTT . . . . . . . . . . . . . . . . . 147\n E.5.1. Replay and Exporters . . . . . . . . . . . . . . . . 148\n E.6. PSK Identity Exposure . . . . . . . . . . . . . . . . . . 149\n E.7. Attacks on Static RSA . . . . . . . . . . . . . . . . . . 149\n Appendix F. Working Group Information . . . . . . . . . . . . . 149\n Appendix G. Contributors . . . . . . . . . . . . . . . . . . . . 149\n Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 156\n\n\n\nRescorla Expires September 21, 2018 [Page 4]\n\f\nInternet-Draft TLS March 2018\n\n\n1. Introduction\n\n RFC EDITOR: PLEASE REMOVE THE FOLLOWING PARAGRAPH The source for this\n draft is maintained in GitHub. Suggested changes should be submitted\n as pull requests at https://github.com/tlswg/tls13-spec.\n Instructions are on that page as well. Editorial changes can be\n managed in GitHub, but any substantive change should be discussed on\n the TLS mailing list.\n\n The primary goal of TLS is to provide a secure channel between two\n communicating peers; the only requirement from the underlying\n transport is a reliable, in-order, data stream. Specifically, the\n secure channel should provide the following properties:\n\n - Authentication: The server side of the channel is always\n authenticated; the client side is optionally authenticated.\n Authentication can happen via asymmetric cryptography (e.g., RSA\n [RSA], ECDSA [ECDSA], EdDSA [RFC8032]) or a pre-shared key (PSK).\n\n - Confidentiality: Data sent over the channel after establishment is\n only visible to the endpoints. TLS does not hide the length of\n the data it transmits, though endpoints are able to pad TLS\n records in order to obscure lengths and improve protection against\n traffic analysis techniques.\n\n - Integrity: Data sent over the channel after establishment cannot\n be modified by attackers.\n\n These properties should be true even in the face of an attacker who\n has complete control of the network, as described in [RFC3552]. See\n Appendix E for a more complete statement of the relevant security\n properties.\n\n TLS consists of two primary components:\n\n - A handshake protocol (Section 4) that authenticates the\n communicating parties, negotiates cryptographic modes and\n parameters, and establishes shared keying material. The handshake\n protocol is designed to resist tampering; an active attacker\n should not be able to force the peers to negotiate different\n parameters than they would if the connection were not under\n attack.\n\n - A record protocol (Section 5) that uses the parameters established\n by the handshake protocol to protect traffic between the\n communicating peers. The record protocol divides traffic up into\n a series of records, each of which is independently protected\n using the traffic keys.\n\n\n\nRescorla Expires September 21, 2018 [Page 5]\n\f\nInternet-Draft TLS March 2018\n\n\n TLS is application protocol independent; higher-level protocols can\n layer on top of TLS transparently. The TLS standard, however, does\n not specify how protocols add security with TLS; how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n This document defines TLS version 1.3. While TLS 1.3 is not directly\n compatible with previous versions, all versions of TLS incorporate a\n versioning mechanism which allows clients and servers to\n interoperably negotiate a common version if one is supported by both\n peers.\n\n This document supersedes and obsoletes previous versions of TLS\n including version 1.2 [RFC5246]. It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2. Section 4.2.7 updates [RFC4492] by modifying\n the protocol attributes used to negotiate Elliptic Curves. Because\n TLS 1.3 changes the way keys are derived, it updates [RFC5705] as\n described in Section 7.5. It also changes how OCSP messages are\n carried and therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in section Section 4.4.2.1.\n\n1.1. Conventions and Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in BCP\n 14 [RFC2119] [RFC8174] when, and only when, they appear in all\n capitals, as shown here.\n\n The following terms are used:\n\n client: The endpoint initiating the TLS connection.\n\n connection: A transport-layer connection between two endpoints.\n\n endpoint: Either the client or server of the connection.\n\n handshake: An initial negotiation between client and server that\n establishes the parameters of their subsequent interactions within\n TLS.\n\n peer: An endpoint. When discussing a particular endpoint, \"peer\"\n refers to the endpoint that is not the primary subject of discussion.\n\n receiver: An endpoint that is receiving records.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 6]\n\f\nInternet-Draft TLS March 2018\n\n\n sender: An endpoint that is transmitting records.\n\n server: The endpoint which did not initiate the TLS connection.\n\n1.2. Change Log\n\n RFC EDITOR PLEASE DELETE THIS SECTION.\n\n (*) indicates changes to the wire protocol which may require\n implementations to update.\n\n draft-28\n\n Add a section on exposure of PSK identities.\n\n draft-27\n\n - SHOULD->MUST for being able to process \"supported_versions\"\n without 0x0304.\n\n - Much editorial cleanup.\n\n draft-26\n\n - Clarify that you can't negotiate pre-TLS 1.3 with\n supported_versions.\n\n draft-25\n\n - Add the header to additional data (*)\n\n - Minor clarifications.\n\n - IANA cleanup.\n\n draft-24\n\n - Require that CH2 have version 0303 (*)\n\n - Some clarifications\n\n draft-23\n\n - Renumber key_share (*)\n\n - Add a new extension and new code points to allow negotiating PSS\n separately for certificates and CertificateVerify (*)\n\n\n\n\nRescorla Expires September 21, 2018 [Page 7]\n\f\nInternet-Draft TLS March 2018\n\n\n - Slightly restrict when CCS must be accepted to make implementation\n easier.\n\n - Document protocol invariants\n\n - Add some text on the security of static RSA.\n\n draft-22\n\n - Implement changes for improved middlebox penetration (*)\n\n - Move server_certificate_type to encrypted extensions (*)\n\n - Allow resumption with a different SNI (*)\n\n - Padding extension can change on HRR (*)\n\n - Allow an empty ticket_nonce (*)\n\n - Remove requirement to immediately respond to close_notify with\n close_notify (allowing half-close)\n\n draft-21\n\n - Add a per-ticket nonce so that each ticket is associated with a\n different PSK (*).\n\n - Clarify that clients should send alerts with the handshake key if\n possible.\n\n - Update state machine to show rekeying events\n\n - Add discussion of 0-RTT and replay. Recommend that\n implementations implement some anti-replay mechanism.\n\n draft-20\n\n - Add \"post_handshake_auth\" extension to negotiate post-handshake\n authentication (*).\n\n - Shorten labels for HKDF-Expand-Label so that we can fit within one\n compression block (*).\n\n - Define how RFC 7250 works (*).\n\n - Re-enable post-handshake client authentication even when you do\n PSK. The previous prohibition was editorial error.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 8]\n\f\nInternet-Draft TLS March 2018\n\n\n - Remove cert_type and user_mapping, which don't work on TLS 1.3\n anyway.\n\n - Added the no_application_protocol alert from [RFC7301] to the list\n of extensions.\n\n - Added discussion of traffic analysis and side channel attacks.\n\n draft-19\n\n - Hash context_value input to Exporters (*)\n\n - Add an additional Derive-Secret stage to Exporters (*).\n\n - Hash ClientHello1 in the transcript when HRR is used. This\n reduces the state that needs to be carried in cookies. (*)\n\n - Restructure CertificateRequest to have the selectors in\n extensions. This also allowed defining a\n \"certificate_authorities\" extension which can be used by the\n client instead of trusted_ca_keys (*).\n\n - Tighten record framing requirements and require checking of them\n (*).\n\n - Consolidate \"ticket_early_data_info\" and \"early_data\" into a\n single extension (*).\n\n - Change end_of_early_data to be a handshake message (*).\n\n - Add pre-extract Derive-Secret stages to key schedule (*).\n\n - Remove spurious requirement to implement \"pre_shared_key\".\n\n - Clarify location of \"early_data\" from server (it goes in EE, as\n indicated by the table in S 10).\n\n - Require peer public key validation\n\n - Add state machine diagram.\n\n draft-18\n\n - Remove unnecessary resumption_psk which is the only thing expanded\n from the resumption master secret. (*).\n\n - Fix signature_algorithms entry in extensions table.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 9]\n\f\nInternet-Draft TLS March 2018\n\n\n - Restate rule from RFC 6066 that you can't resume unless SNI is the\n same.\n\n draft-17\n\n - Remove 0-RTT Finished and resumption_context, and replace with a\n psk_binder field in the PSK itself (*)\n\n - Restructure PSK key exchange negotiation modes (*)\n\n - Add max_early_data_size field to TicketEarlyDataInfo (*)\n\n - Add a 0-RTT exporter and change the transcript for the regular\n exporter (*)\n\n - Merge TicketExtensions and Extensions registry. Changes\n ticket_early_data_info code point (*)\n\n - Replace Client.key_shares in response to HRR (*)\n\n - Remove redundant labels for traffic key derivation (*)\n\n - Harmonize requirements about cipher suite matching: for resumption\n you need to match KDF but for 0-RTT you need whole cipher suite.\n This allows PSKs to actually negotiate cipher suites. (*)\n\n - Move SCT and OCSP into Certificate.extensions (*)\n\n - Explicitly allow non-offered extensions in NewSessionTicket\n\n - Explicitly allow predicting client Finished for NST\n\n - Clarify conditions for allowing 0-RTT with PSK\n\n draft-16\n\n - Revise version negotiation (*)\n\n - Change RSASSA-PSS and EdDSA SignatureScheme codepoints for better\n backwards compatibility (*)\n\n - Move HelloRetryRequest.selected_group to an extension (*)\n\n - Clarify the behavior of no exporter context and make it the same\n as an empty context.(*)\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 10]\n\f\nInternet-Draft TLS March 2018\n\n\n - New KeyUpdate format that allows for requesting/not-requesting an\n answer. This also means changes to the key schedule to support\n independent updates (*)\n\n - New certificate_required alert (*)\n\n - Forbid CertificateRequest with 0-RTT and PSK.\n\n - Relax requirement to check SNI for 0-RTT.\n\n draft-15\n\n - New negotiation syntax as discussed in Berlin (*)\n\n - Require CertificateRequest.context to be empty during handshake\n (*)\n\n - Forbid empty tickets (*)\n\n - Forbid application data messages in between post-handshake\n messages from the same flight (*)\n\n - Clean up alert guidance (*)\n\n - Clearer guidance on what is needed for TLS 1.2.\n\n - Guidance on 0-RTT time windows.\n\n - Rename a bunch of fields.\n\n - Remove old PRNG text.\n\n - Explicitly require checking that handshake records not span key\n changes.\n\n draft-14\n\n - Allow cookies to be longer (*)\n\n - Remove the \"context\" from EarlyDataIndication as it was undefined\n and nobody used it (*)\n\n - Remove 0-RTT EncryptedExtensions and replace the ticket_age\n extension with an obfuscated version. Also necessitates a change\n to NewSessionTicket (*).\n\n - Move the downgrade sentinel to the end of ServerHello.Random to\n accommodate tlsdate (*).\n\n\n\nRescorla Expires September 21, 2018 [Page 11]\n\f\nInternet-Draft TLS March 2018\n\n\n - Define ecdsa_sha1 (*).\n\n - Allow resumption even after fatal alerts. This matches current\n practice.\n\n - Remove non-closure warning alerts. Require treating unknown\n alerts as fatal.\n\n - Make the rules for accepting 0-RTT less restrictive.\n\n - Clarify 0-RTT backward-compatibility rules.\n\n - Clarify how 0-RTT and PSK identities interact.\n\n - Add a section describing the data limits for each cipher.\n\n - Major editorial restructuring.\n\n - Replace the Security Analysis section with a WIP draft.\n\n draft-13\n\n - Allow server to send SupportedGroups.\n\n - Remove 0-RTT client authentication\n\n - Remove (EC)DHE 0-RTT.\n\n - Flesh out 0-RTT PSK mode and shrink EarlyDataIndication\n\n - Turn PSK-resumption response into an index to save room\n\n - Move CertificateStatus to an extension\n\n - Extra fields in NewSessionTicket.\n\n - Restructure key schedule and add a resumption_context value.\n\n - Require DH public keys and secrets to be zero-padded to the size\n of the group.\n\n - Remove the redundant length fields in KeyShareEntry.\n\n - Define a cookie field for HRR.\n\n draft-12\n\n - Provide a list of the PSK cipher suites.\n\n\n\nRescorla Expires September 21, 2018 [Page 12]\n\f\nInternet-Draft TLS March 2018\n\n\n - Remove the ability for the ServerHello to have no extensions (this\n aligns the syntax with the text).\n\n - Clarify that the server can send application data after its first\n flight (0.5 RTT data)\n\n - Revise signature algorithm negotiation to group hash, signature\n algorithm, and curve together. This is backwards compatible.\n\n - Make ticket lifetime mandatory and limit it to a week.\n\n - Make the purpose strings lower-case. This matches how people are\n implementing for interop.\n\n - Define exporters.\n\n - Editorial cleanup\n\n draft-11\n\n - Port the CFRG curves & signatures work from RFC4492bis.\n\n - Remove sequence number and version from additional_data, which is\n now empty.\n\n - Reorder values in HkdfLabel.\n\n - Add support for version anti-downgrade mechanism.\n\n - Update IANA considerations section and relax some of the policies.\n\n - Unify authentication modes. Add post-handshake client\n authentication.\n\n - Remove early_handshake content type. Terminate 0-RTT data with an\n alert.\n\n - Reset sequence number upon key change (as proposed by Fournet et\n al.)\n\n draft-10\n\n - Remove ClientCertificateTypes field from CertificateRequest and\n add extensions.\n\n - Merge client and server key shares into a single extension.\n\n draft-09\n\n\n\nRescorla Expires September 21, 2018 [Page 13]\n\f\nInternet-Draft TLS March 2018\n\n\n - Change to RSA-PSS signatures for handshake messages.\n\n - Remove support for DSA.\n\n - Update key schedule per suggestions by Hugo, Hoeteck, and Bjoern\n Tackmann.\n\n - Add support for per-record padding.\n\n - Switch to encrypted record ContentType.\n\n - Change HKDF labeling to include protocol version and value\n lengths.\n\n - Shift the final decision to abort a handshake due to incompatible\n certificates to the client rather than having servers abort early.\n\n - Deprecate SHA-1 with signatures.\n\n - Add MTI algorithms.\n\n draft-08\n\n - Remove support for weak and lesser used named curves.\n\n - Remove support for MD5 and SHA-224 hashes with signatures.\n\n - Update lists of available AEAD cipher suites and error alerts.\n\n - Reduce maximum permitted record expansion for AEAD from 2048 to\n 256 octets.\n\n - Require digital signatures even when a previous configuration is\n used.\n\n - Merge EarlyDataIndication and KnownConfiguration.\n\n - Change code point for server_configuration to avoid collision with\n server_hello_done.\n\n - Relax certificate_list ordering requirement to match current\n practice.\n\n draft-07\n\n - Integration of semi-ephemeral DH proposal.\n\n - Add initial 0-RTT support.\n\n\n\nRescorla Expires September 21, 2018 [Page 14]\n\f\nInternet-Draft TLS March 2018\n\n\n - Remove resumption and replace with PSK + tickets.\n\n - Move ClientKeyShare into an extension.\n\n - Move to HKDF.\n\n draft-06\n\n - Prohibit RC4 negotiation for backwards compatibility.\n\n - Freeze & deprecate record layer version field.\n\n - Update format of signatures with context.\n\n - Remove explicit IV.\n\n draft-05\n\n - Prohibit SSL negotiation for backwards compatibility.\n\n - Fix which MS is used for exporters.\n\n draft-04\n\n - Modify key computations to include session hash.\n\n - Remove ChangeCipherSpec.\n\n - Renumber the new handshake messages to be somewhat more consistent\n with existing convention and to remove a duplicate registration.\n\n - Remove renegotiation.\n\n - Remove point format negotiation.\n\n draft-03\n\n - Remove GMT time.\n\n - Merge in support for ECC from RFC 4492 but without explicit\n curves.\n\n - Remove the unnecessary length field from the AD input to AEAD\n ciphers.\n\n - Rename {Client,Server}KeyExchange to {Client,Server}KeyShare.\n\n - Add an explicit HelloRetryRequest to reject the client's.\n\n\n\nRescorla Expires September 21, 2018 [Page 15]\n\f\nInternet-Draft TLS March 2018\n\n\n draft-02\n\n - Increment version number.\n\n - Rework handshake to provide 1-RTT mode.\n\n - Remove custom DHE groups.\n\n - Remove support for compression.\n\n - Remove support for static RSA and DH key exchange.\n\n - Remove support for non-AEAD ciphers.\n\n1.3. Major Differences from TLS 1.2\n\n The following is a list of the major functional differences between\n TLS 1.2 and TLS 1.3. It is not intended to be exhaustive and there\n are many minor differences.\n\n - The list of supported symmetric algorithms has been pruned of all\n algorithms that are considered legacy. Those that remain all use\n Authenticated Encryption with Associated Data (AEAD) algorithms.\n The ciphersuite concept has been changed to separate the\n authentication and key exchange mechanisms from the record\n protection algorithm (including secret key length) and a hash to\n be used with the key derivation function and HMAC.\n\n - A 0-RTT mode was added, saving a round-trip at connection setup\n for some application data, at the cost of certain security\n properties.\n\n - Static RSA and Diffie-Hellman cipher suites have been removed; all\n public-key based key exchange mechanisms now provide forward\n secrecy.\n\n - All handshake messages after the ServerHello are now encrypted.\n The newly introduced EncryptedExtension message allows various\n extensions previously sent in clear in the ServerHello to also\n enjoy confidentiality protection from active attackers.\n\n - The key derivation functions have been re-designed. The new\n design allows easier analysis by cryptographers due to their\n improved key separation properties. The HMAC-based Extract-and-\n Expand Key Derivation Function (HKDF) is used as an underlying\n primitive.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 16]\n\f\nInternet-Draft TLS March 2018\n\n\n - The handshake state machine has been significantly restructured to\n be more consistent and to remove superfluous messages such as\n ChangeCipherSpec (except when needed for middlebox compatibility).\n\n - Elliptic curve algorithms are now in the base spec and new\n signature algorithms, such as ed25519 and ed448, are included.\n TLS 1.3 removed point format negotiation in favor of a single\n point format for each curve.\n\n - Other cryptographic improvements including the removal of\n compression and custom DHE groups, changing the RSA padding to use\n RSASSA-PSS, and the removal of DSA.\n\n - The TLS 1.2 version negotiation mechanism has been deprecated in\n favor of a version list in an extension. This increases\n compatibility with existing servers that incorrectly implemented\n version negotiation.\n\n - Session resumption with and without server-side state as well as\n the PSK-based ciphersuites of earlier TLS versions have been\n replaced by a single new PSK exchange.\n\n - Updated references to point to the updated versions of RFCs, as\n appropriate (e.g., RFC 5280 rather than RFC 3280).\n\n1.4. Updates Affecting TLS 1.2\n\n This document defines several changes that optionally affect\n implementations of TLS 1.2, including those which do not also support\n TLS 1.3:\n\n - A version downgrade protection mechanism is described in\n Section 4.1.3.\n\n - RSASSA-PSS signature schemes are defined in Section 4.2.3.\n\n - The \"supported_versions\" ClientHello extension can be used to\n negotiate the version of TLS to use, in preference to the\n legacy_version field of the ClientHello.\n\n - The \"signature_algorithms_cert\" extension allows a client to\n indicate which signature algorithms it can validate in X.509\n certificates\n\n Additionally, this document clarifies some compliance requirements\n for earlier versions of TLS; see Section 9.3.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 17]\n\f\nInternet-Draft TLS March 2018\n\n\n2. Protocol Overview\n\n The cryptographic parameters used by the secure channel are produced\n by the TLS handshake protocol. This sub-protocol of TLS is used by\n the client and server when first communicating with each other. The\n handshake protocol allows peers to negotiate a protocol version,\n select cryptographic algorithms, optionally authenticate each other,\n and establish shared secret keying material. Once the handshake is\n complete, the peers use the established keys to protect the\n application layer traffic.\n\n A failure of the handshake or other protocol error triggers the\n termination of the connection, optionally preceded by an alert\n message (Section 6).\n\n TLS supports three basic key exchange modes:\n\n - (EC)DHE (Diffie-Hellman over either finite fields or elliptic\n curves)\n\n - PSK-only\n\n - PSK with (EC)DHE\n\n Figure 1 below shows the basic full TLS handshake:\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 18]\n\f\nInternet-Draft TLS March 2018\n\n\n Client Server\n\nKey ^ ClientHello\nExch | + key_share*\n | + signature_algorithms*\n | + psk_key_exchange_modes*\n v + pre_shared_key* -------->\n ServerHello ^ Key\n + key_share* | Exch\n + pre_shared_key* v\n {EncryptedExtensions} ^ Server\n {CertificateRequest*} v Params\n {Certificate*} ^\n {CertificateVerify*} | Auth\n {Finished} v\n <-------- [Application Data*]\n ^ {Certificate*}\nAuth | {CertificateVerify*}\n v {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N\n\n Figure 1: Message flow for full TLS Handshake\n\n The handshake can be thought of as having three phases (indicated in\n the diagram above):\n\n - Key Exchange: Establish shared keying material and select the\n cryptographic parameters. Everything after this phase is\n encrypted.\n\n - Server Parameters: Establish other handshake parameters (whether\n the client is authenticated, application layer protocol support,\n etc.).\n\n - Authentication: Authenticate the server (and optionally the\n client) and provide key confirmation and handshake integrity.\n\n\n\nRescorla Expires September 21, 2018 [Page 19]\n\f\nInternet-Draft TLS March 2018\n\n\n In the Key Exchange phase, the client sends the ClientHello\n (Section 4.1.2) message, which contains a random nonce\n (ClientHello.random); its offered protocol versions; a list of\n symmetric cipher/HKDF hash pairs; either a set of Diffie-Hellman key\n shares (in the \"key_share\" extension Section 4.2.8), a set of pre-\n shared key labels (in the \"pre_shared_key\" extension Section 4.2.11)\n or both; and potentially additional extensions. Additional fields\n and/or messages may also be present for middlebox compatibility.\n\n The server processes the ClientHello and determines the appropriate\n cryptographic parameters for the connection. It then responds with\n its own ServerHello (Section 4.1.3), which indicates the negotiated\n connection parameters. The combination of the ClientHello and the\n ServerHello determines the shared keys. If (EC)DHE key establishment\n is in use, then the ServerHello contains a \"key_share\" extension with\n the server's ephemeral Diffie-Hellman share; the server's share MUST\n be in the same group as one of the client's shares. If PSK key\n establishment is in use, then the ServerHello contains a\n \"pre_shared_key\" extension indicating which of the client's offered\n PSKs was selected. Note that implementations can use (EC)DHE and PSK\n together, in which case both extensions will be supplied.\n\n The server then sends two messages to establish the Server\n Parameters:\n\n EncryptedExtensions: responses to ClientHello extensions that are\n not required to determine the cryptographic parameters, other than\n those that are specific to individual certificates.\n [Section 4.3.1]\n\n CertificateRequest: if certificate-based client authentication is\n desired, the desired parameters for that certificate. This\n message is omitted if client authentication is not desired.\n [Section 4.3.2]\n\n Finally, the client and server exchange Authentication messages. TLS\n uses the same set of messages every time that certificate-based\n authentication is needed. (PSK-based authentication happens as a\n side effect of key exchange.) Specifically:\n\n Certificate: the certificate of the endpoint and any per-certificate\n extensions. This message is omitted by the server if not\n authenticating with a certificate and by the client if the server\n did not send CertificateRequest (thus indicating that the client\n should not authenticate with a certificate). Note that if raw\n public keys [RFC7250] or the cached information extension\n [RFC7924] are in use, then this message will not contain a\n\n\n\n\nRescorla Expires September 21, 2018 [Page 20]\n\f\nInternet-Draft TLS March 2018\n\n\n certificate but rather some other value corresponding to the\n server's long-term key. [Section 4.4.2]\n\n CertificateVerify: a signature over the entire handshake using the\n private key corresponding to the public key in the Certificate\n message. This message is omitted if the endpoint is not\n authenticating via a certificate. [Section 4.4.3]\n\n Finished: a MAC (Message Authentication Code) over the entire\n handshake. This message provides key confirmation, binds the\n endpoint's identity to the exchanged keys, and in PSK mode also\n authenticates the handshake. [Section 4.4.4]\n\n Upon receiving the server's messages, the client responds with its\n Authentication messages, namely Certificate and CertificateVerify (if\n requested), and Finished.\n\n At this point, the handshake is complete, and the client and server\n derive the keying material required by the record layer to exchange\n application-layer data protected through authenticated encryption.\n Application data MUST NOT be sent prior to sending the Finished\n message, except as specified in [Section 2.3]. Note that while the\n server may send application data prior to receiving the client's\n Authentication messages, any data sent at that point is, of course,\n being sent to an unauthenticated peer.\n\n2.1. Incorrect DHE Share\n\n If the client has not provided a sufficient \"key_share\" extension\n (e.g., it includes only DHE or ECDHE groups unacceptable to or\n unsupported by the server), the server corrects the mismatch with a\n HelloRetryRequest and the client needs to restart the handshake with\n an appropriate \"key_share\" extension, as shown in Figure 2. If no\n common cryptographic parameters can be negotiated, the server MUST\n abort the handshake with an appropriate alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 21]\n\f\nInternet-Draft TLS March 2018\n\n\n Client Server\n\n ClientHello\n + key_share -------->\n HelloRetryRequest\n <-------- + key_share\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 2: Message flow for a full handshake with mismatched\n parameters\n\n Note: The handshake transcript incorporates the initial ClientHello/\n HelloRetryRequest exchange; it is not reset with the new ClientHello.\n\n TLS also allows several optimized variants of the basic handshake, as\n described in the following sections.\n\n2.2. Resumption and Pre-Shared Key (PSK)\n\n Although TLS PSKs can be established out of band, PSKs can also be\n established in a previous connection and then used to establish a new\n connection (\"session resumption\" or \"resuming\" with a PSK). Once a\n handshake has completed, the server can send to the client a PSK\n identity that corresponds to a unique key derived from the initial\n handshake (see Section 4.6.1). The client can then use that PSK\n identity in future handshakes to negotiate the use of the associated\n PSK. If the server accepts the PSK, then the security context of the\n new connection is cryptographically tied to the original connection\n and the key derived from the initial handshake is used to bootstrap\n the cryptographic state instead of a full handshake. In TLS 1.2 and\n below, this functionality was provided by \"session IDs\" and \"session\n tickets\" [RFC5077]. Both mechanisms are obsoleted in TLS 1.3.\n\n PSKs can be used with (EC)DHE key exchange in order to provide\n forward secrecy in combination with shared keys, or can be used\n\n\n\nRescorla Expires September 21, 2018 [Page 22]\n\f\nInternet-Draft TLS March 2018\n\n\n alone, at the cost of losing forward secrecy for the application\n data.\n\n Figure 3 shows a pair of handshakes in which the first establishes a\n PSK and the second uses it:\n\n Client Server\n\n Initial Handshake:\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n <-------- [NewSessionTicket]\n [Application Data] <-------> [Application Data]\n\n\n Subsequent Handshake:\n ClientHello\n + key_share*\n + pre_shared_key -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n {Finished}\n <-------- [Application Data*]\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 3: Message flow for resumption and PSK\n\n As the server is authenticating via a PSK, it does not send a\n Certificate or a CertificateVerify message. When a client offers\n resumption via PSK, it SHOULD also supply a \"key_share\" extension to\n the server to allow the server to decline resumption and fall back to\n a full handshake, if needed. The server responds with a\n \"pre_shared_key\" extension to negotiate use of PSK key establishment\n\n\n\n\nRescorla Expires September 21, 2018 [Page 23]\n\f\nInternet-Draft TLS March 2018\n\n\n and can (as shown here) respond with a \"key_share\" extension to do\n (EC)DHE key establishment, thus providing forward secrecy.\n\n When PSKs are provisioned out of band, the PSK identity and the KDF\n hash algorithm to be used with the PSK MUST also be provisioned.\n\n Note: When using an out-of-band provisioned pre-shared secret, a\n critical consideration is using sufficient entropy during the key\n generation, as discussed in [RFC4086]. Deriving a shared secret\n from a password or other low-entropy sources is not secure. A\n low-entropy secret, or password, is subject to dictionary attacks\n based on the PSK binder. The specified PSK authentication is not\n a strong password-based authenticated key exchange even when used\n with Diffie-Hellman key establishment. Specifically, it does not\n prevent an attacker that can observe the handshake from performing\n a brute-force attack on the password/pre-shared key.\n\n2.3. 0-RTT Data\n\n When clients and servers share a PSK (either obtained externally or\n via a previous handshake), TLS 1.3 allows clients to send data on the\n first flight (\"early data\"). The client uses the PSK to authenticate\n the server and to encrypt the early data.\n\n As shown in Figure 4, the 0-RTT data is just added to the 1-RTT\n handshake in the first flight. The rest of the handshake uses the\n same messages as for a 1-RTT handshake with PSK resumption.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 24]\n\f\nInternet-Draft TLS March 2018\n\n\n Client Server\n\n ClientHello\n + early_data\n + key_share*\n + psk_key_exchange_modes\n + pre_shared_key\n (Application Data*) -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n + early_data*\n {Finished}\n <-------- [Application Data*]\n (EndOfEarlyData)\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n () Indicates messages protected using keys\n derived from client_early_traffic_secret.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N\n\n Figure 4: Message flow for a zero round trip handshake\n\n IMPORTANT NOTE: The security properties for 0-RTT data are weaker\n than those for other kinds of TLS data. Specifically:\n\n 1. This data is not forward secret, as it is encrypted solely under\n keys derived using the offered PSK.\n\n 2. There are no guarantees of non-replay between connections.\n Protection against replay for ordinary TLS 1.3 1-RTT data is\n provided via the server's Random value, but 0-RTT data does not\n depend on the ServerHello and therefore has weaker guarantees.\n This is especially relevant if the data is authenticated either\n with TLS client authentication or inside the application\n\n\n\nRescorla Expires September 21, 2018 [Page 25]\n\f\nInternet-Draft TLS March 2018\n\n\n protocol. The same warnings apply to any use of the\n early_exporter_master_secret.\n\n 0-RTT data cannot be duplicated within a connection (i.e., the server\n will not process the same data twice for the same connection) and an\n attacker will not be able to make 0-RTT data appear to be 1-RTT data\n (because it is protected with different keys.) Appendix E.5 contains\n a description of potential attacks and Section 8 describes mechanisms\n which the server can use to limit the impact of replay.\n\n3. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used.\n\n3.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n3.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" double\n brackets.\n\n Single-byte entities containing uninterpreted data are of type\n opaque.\n\n A type alias T' for an existing type T is defined by:\n\n T T';\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 26]\n\f\nInternet-Draft TLS March 2018\n\n\n3.3. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are formed from fixed-length series of bytes\n concatenated as described in Section 3.1 and are also unsigned. The\n following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are transmitted\n in network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n3.4. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* 3 consecutive 3-byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n\n\n\nRescorla Expires September 21, 2018 [Page 27]\n\f\nInternet-Draft TLS March 2018\n\n\n In the following example, mandatory is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 3.3). Similarly,\n longer can represent up to 800 bytes of data, or 400 uint16 elements,\n and it may be empty. Its encoding will include a two-byte actual\n length field prepended to the vector. The length of an encoded\n vector must be an exact multiple of the length of a single element\n (e.g., a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is 2 bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n3.5. Enumerateds\n\n An additional sparse data type is available called enum or\n enumerated. Each definition is a different type. Only enumerateds\n of the same type may be assigned or compared. Every element of an\n enumerated must be assigned a value, as demonstrated in the following\n example. Since the elements of the enumerated are not ordered, they\n can be assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n Future extensions or additions to the protocol may define new values.\n Implementations need to be able to parse and ignore unknown values\n unless the definition of the field states otherwise.\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4 in the current\n version of the protocol.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n\n\n\nRescorla Expires September 21, 2018 [Page 28]\n\f\nInternet-Draft TLS March 2018\n\n\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n The names assigned to enumerateds do not need to be unique. The\n numerical value can describe a range over which the same name\n applies. The value includes the minimum and maximum inclusive values\n in that range, separated by two period characters. This is\n principally useful for reserving regions of the space.\n\n enum { sad(0), meh(1..254), happy(255) } Mood;\n\n3.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax for definition is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } T;\n\n Fixed- and variable-length vector fields are allowed using the\n standard vector syntax. Structures V1 and V2 in the variants example\n below demonstrate this.\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n\n3.7. Constants\n\n Fields and variables may be assigned a fixed value using \"=\", as in:\n\n struct {\n T1 f1 = 8; /* T.f1 must always be 8 */\n T2 f2;\n } T;\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 29]\n\f\nInternet-Draft TLS March 2018\n\n\n3.8. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. Each\n arm of the select specifies the type of that variant's field and an\n optional field label. The mechanism by which the variant is selected\n at runtime is not prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1 [[fe1]];\n case e2: Te2 [[fe2]];\n ....\n case en: Ten [[fen]];\n };\n } Tv;\n\n For example:\n\n enum { apple(0), orange(1) } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n VariantTag type;\n select (VariantRecord.type) {\n case apple: V1;\n case orange: V2;\n };\n } VariantRecord;\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 30]\n\f\nInternet-Draft TLS March 2018\n\n\n4. Handshake Protocol\n\n The handshake protocol is used to negotiate the security parameters\n of a connection. Handshake messages are supplied to the TLS record\n layer, where they are encapsulated within one or more TLSPlaintext or\n TLSCiphertext structures, which are processed and transmitted as\n specified by the current active connection state.\n\n enum {\n client_hello(1),\n server_hello(2),\n new_session_ticket(4),\n end_of_early_data(5),\n encrypted_extensions(8),\n certificate(11),\n certificate_request(13),\n certificate_verify(15),\n finished(20),\n key_update(24),\n message_hash(254),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (Handshake.msg_type) {\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case end_of_early_data: EndOfEarlyData;\n case encrypted_extensions: EncryptedExtensions;\n case certificate_request: CertificateRequest;\n case certificate: Certificate;\n case certificate_verify: CertificateVerify;\n case finished: Finished;\n case new_session_ticket: NewSessionTicket;\n case key_update: KeyUpdate;\n };\n } Handshake;\n\n Protocol messages MUST be sent in the order defined in Section 4.4.1\n and shown in the diagrams in Section 2. A peer which receives a\n handshake message in an unexpected order MUST abort the handshake\n with an \"unexpected_message\" alert.\n\n New handshake message types are assigned by IANA as described in\n Section 11.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 31]\n\f\nInternet-Draft TLS March 2018\n\n\n4.1. Key Exchange Messages\n\n The key exchange messages are used to determine the security\n capabilities of the client and the server and to establish shared\n secrets including the traffic keys used to protect the rest of the\n handshake and the data.\n\n4.1.1. Cryptographic Negotiation\n\n In TLS, the cryptographic negotiation proceeds by the client offering\n the following four sets of options in its ClientHello:\n\n - A list of cipher suites which indicates the AEAD algorithm/HKDF\n hash pairs which the client supports.\n\n - A \"supported_groups\" (Section 4.2.7) extension which indicates the\n (EC)DHE groups which the client supports and a \"key_share\"\n (Section 4.2.8) extension which contains (EC)DHE shares for some\n or all of these groups.\n\n - A \"signature_algorithms\" (Section 4.2.3) extension which indicates\n the signature algorithms which the client can accept.\n\n - A \"pre_shared_key\" (Section 4.2.11) extension which contains a\n list of symmetric key identities known to the client and a\n \"psk_key_exchange_modes\" (Section 4.2.9) extension which indicates\n the key exchange modes that may be used with PSKs.\n\n If the server does not select a PSK, then the first three of these\n options are entirely orthogonal: the server independently selects a\n cipher suite, an (EC)DHE group and key share for key establishment,\n and a signature algorithm/certificate pair to authenticate itself to\n the client. If there is no overlap between the received\n \"supported_groups\" and the groups supported by the server then the\n server MUST abort the handshake with a \"handshake_failure\" or an\n \"insufficient_security\" alert.\n\n If the server selects a PSK, then it MUST also select a key\n establishment mode from the set indicated by client's\n \"psk_key_exchange_modes\" extension (at present, PSK alone or with\n (EC)DHE). Note that if the PSK can be used without (EC)DHE then non-\n overlap in the \"supported_groups\" parameters need not be fatal, as it\n is in the non-PSK case discussed in the previous paragraph.\n\n If the server selects an (EC)DHE group and the client did not offer a\n compatible \"key_share\" extension in the initial ClientHello, the\n server MUST respond with a HelloRetryRequest (Section 4.1.4) message.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 32]\n\f\nInternet-Draft TLS March 2018\n\n\n If the server successfully selects parameters and does not require a\n HelloRetryRequest, it indicates the selected parameters in the\n ServerHello as follows:\n\n - If PSK is being used, then the server will send a \"pre_shared_key\"\n extension indicating the selected key.\n\n - If PSK is not being used, then (EC)DHE and certificate-based\n authentication are always used.\n\n - When (EC)DHE is in use, the server will also provide a \"key_share\"\n extension.\n\n - When authenticating via a certificate, the server will send the\n Certificate (Section 4.4.2) and CertificateVerify (Section 4.4.3)\n messages. In TLS 1.3 as defined by this document, either a PSK or\n a certificate is always used, but not both. Future documents may\n define how to use them together.\n\n If the server is unable to negotiate a supported set of parameters\n (i.e., there is no overlap between the client and server parameters),\n it MUST abort the handshake with either a \"handshake_failure\" or\n \"insufficient_security\" fatal alert (see Section 6).\n\n4.1.2. Client Hello\n\n When a client first connects to a server, it is REQUIRED to send the\n ClientHello as its first TLS message. The client will also send a\n ClientHello when the server has responded to its ClientHello with a\n HelloRetryRequest. In that case, the client MUST send the same\n ClientHello without modification, except:\n\n - If a \"key_share\" extension was supplied in the HelloRetryRequest,\n replacing the list of shares with a list containing a single\n KeyShareEntry from the indicated group.\n\n - Removing the \"early_data\" extension (Section 4.2.10) if one was\n present. Early data is not permitted after HelloRetryRequest.\n\n - Including a \"cookie\" extension if one was provided in the\n HelloRetryRequest.\n\n - Updating the \"pre_shared_key\" extension if present by recomputing\n the \"obfuscated_ticket_age\" and binder values and (optionally)\n removing any PSKs which are incompatible with the server's\n indicated cipher suite.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 33]\n\f\nInternet-Draft TLS March 2018\n\n\n - Optionally adding, removing, or changing the length of the\n \"padding\" extension [RFC7685].\n\n - Other modifications that may be allowed by an extension defined in\n the future and present in the HelloRetryRequest.\n\n Because TLS 1.3 forbids renegotiation, if a server has negotiated TLS\n 1.3 and receives a ClientHello at any other time, it MUST terminate\n the connection with an \"unexpected_message\" alert.\n\n If a server established a TLS connection with a previous version of\n TLS and receives a TLS 1.3 ClientHello in a renegotiation, it MUST\n retain the previous protocol version. In particular, it MUST NOT\n negotiate TLS 1.3.\n\n Structure of this message:\n\n uint16 ProtocolVersion;\n opaque Random[32];\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id<0..32>;\n CipherSuite cipher_suites<2..2^16-2>;\n opaque legacy_compression_methods<1..2^8-1>;\n Extension extensions<8..2^16-1>;\n } ClientHello;\n\n legacy_version In previous versions of TLS, this field was used for\n version negotiation and represented the highest version number\n supported by the client. Experience has shown that many servers\n do not properly implement version negotiation, leading to \"version\n intolerance\" in which the server rejects an otherwise acceptable\n ClientHello with a version number higher than it supports. In TLS\n 1.3, the client indicates its version preferences in the\n \"supported_versions\" extension (Section 4.2.1) and the\n legacy_version field MUST be set to 0x0303, which is the version\n number for TLS 1.2. (See Appendix D for details about backward\n compatibility.)\n\n random 32 bytes generated by a secure random number generator. See\n Appendix C for additional information.\n\n legacy_session_id Versions of TLS before TLS 1.3 supported a\n \"session resumption\" feature which has been merged with Pre-Shared\n\n\n\nRescorla Expires September 21, 2018 [Page 34]\n\f\nInternet-Draft TLS March 2018\n\n\n Keys in this version (see Section 2.2). A client which has a\n cached session ID set by a pre-TLS 1.3 server SHOULD set this\n field to that value. In compatibility mode (see Appendix D.4)\n this field MUST be non-empty, so a client not offering a pre-TLS\n 1.3 session MUST generate a new 32-byte value. This value need\n not be random but SHOULD be unpredictable to avoid implementations\n fixating on a specific value (also known as ossification).\n Otherwise, it MUST be set as a zero length vector (i.e., a single\n zero byte length field).\n\n cipher_suites This is a list of the symmetric cipher options\n supported by the client, specifically the record protection\n algorithm (including secret key length) and a hash to be used with\n HKDF, in descending order of client preference. If the list\n contains cipher suites that the server does not recognize, support\n or wish to use, the server MUST ignore those cipher suites and\n process the remaining ones as usual. Values are defined in\n Appendix B.4. If the client is attempting a PSK key\n establishment, it SHOULD advertise at least one cipher suite\n indicating a Hash associated with the PSK.\n\n legacy_compression_methods Versions of TLS before 1.3 supported\n compression with the list of supported compression methods being\n sent in this field. For every TLS 1.3 ClientHello, this vector\n MUST contain exactly one byte, set to zero, which corresponds to\n the \"null\" compression method in prior versions of TLS. If a TLS\n 1.3 ClientHello is received with any other value in this field,\n the server MUST abort the handshake with an \"illegal_parameter\"\n alert. Note that TLS 1.3 servers might receive TLS 1.2 or prior\n ClientHellos which contain other compression methods and (if\n negotiating such a prior version) MUST follow the procedures for\n the appropriate prior version of TLS. TLS 1.3 ClientHellos are\n identified as having a legacy_version of 0x0303 and a\n supported_versions extension present with 0x0304 as the highest\n version indicated therein.\n\n extensions Clients request extended functionality from servers by\n sending data in the extensions field. The actual \"Extension\"\n format is defined in Section 4.2. In TLS 1.3, use of certain\n extensions is mandatory, as functionality is moved into extensions\n to preserve ClientHello compatibility with previous versions of\n TLS. Servers MUST ignore unrecognized extensions.\n\n All versions of TLS allow an extensions field to optionally follow\n the compression_methods field. TLS 1.3 ClientHello messages always\n contain extensions (minimally \"supported_versions\", otherwise they\n will be interpreted as TLS 1.2 ClientHello messages). However, TLS\n 1.3 servers might receive ClientHello messages without an extensions\n\n\n\nRescorla Expires September 21, 2018 [Page 35]\n\f\nInternet-Draft TLS March 2018\n\n\n field from prior versions of TLS. The presence of extensions can be\n detected by determining whether there are bytes following the\n compression_methods field at the end of the ClientHello. Note that\n this method of detecting optional data differs from the normal TLS\n method of having a variable-length field, but it is used for\n compatibility with TLS before extensions were defined. TLS 1.3\n servers will need to perform this check first and only attempt to\n negotiate TLS 1.3 if the \"supported_versions\" extension is present.\n If negotiating a version of TLS prior to 1.3, a server MUST check\n that the message either contains no data after\n legacy_compression_methods or that it contains a valid extensions\n block with no data following. If not, then it MUST abort the\n handshake with a \"decode_error\" alert.\n\n In the event that a client requests additional functionality using\n extensions, and this functionality is not supplied by the server, the\n client MAY abort the handshake.\n\n After sending the ClientHello message, the client waits for a\n ServerHello or HelloRetryRequest message. If early data is in use,\n the client may transmit early application data (Section 2.3) while\n waiting for the next handshake message.\n\n4.1.3. Server Hello\n\n The server will send this message in response to a ClientHello\n message to proceed with the handshake if it is able to negotiate an\n acceptable set of handshake parameters based on the ClientHello.\n\n Structure of this message:\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id_echo<0..32>;\n CipherSuite cipher_suite;\n uint8 legacy_compression_method = 0;\n Extension extensions<6..2^16-1>;\n } ServerHello;\n\n legacy_version In previous versions of TLS, this field was used for\n version negotiation and represented the selected version number\n for the connection. Unfortunately, some middleboxes fail when\n presented with new values. In TLS 1.3, the TLS server indicates\n its version using the \"supported_versions\" extension\n (Section 4.2.1), and the legacy_version field MUST be set to\n 0x0303, which is the version number for TLS 1.2. (See Appendix D\n for details about backward compatibility.)\n\n\n\nRescorla Expires September 21, 2018 [Page 36]\n\f\nInternet-Draft TLS March 2018\n\n\n random 32 bytes generated by a secure random number generator. See\n Appendix C for additional information. The last eight bytes MUST\n be overwritten as described below if negotiating TLS 1.2 or TLS\n 1.1, but the remaining bytes MUST be random. This structure is\n generated by the server and MUST be generated independently of the\n ClientHello.random.\n\n legacy_session_id_echo The contents of the client's\n legacy_session_id field. Note that this field is echoed even if\n the client's value corresponded to a cached pre-TLS 1.3 session\n which the server has chosen not to resume. A client which\n receives a legacy_session_id_echo field that does not match what\n it sent in the ClientHello MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n cipher_suite The single cipher suite selected by the server from the\n list in ClientHello.cipher_suites. A client which receives a\n cipher suite that was not offered MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n legacy_compression_method A single byte which MUST have the value 0.\n\n extensions A list of extensions. The ServerHello MUST only include\n extensions which are required to establish the cryptographic\n context and negotiate the protocol version. All TLS 1.3\n ServerHello messages MUST contain the \"supported_versions\"\n extension. Current ServerHello messages additionally contain\n either the \"pre_shared_key\" or \"key_share\" extensions, or both\n when using a PSK with (EC)DHE key establishment. Other extensions\n are sent separately in the EncryptedExtensions message.\n\n For reasons of backward compatibility with middleboxes (see\n Appendix D.4) the HelloRetryRequest message uses the same structure\n as the ServerHello, but with Random set to the special value of the\n SHA-256 of \"HelloRetryRequest\":\n\n CF 21 AD 74 E5 9A 61 11 BE 1D 8C 02 1E 65 B8 91\n C2 A2 11 16 7A BB 8C 5E 07 9E 09 E2 C8 A8 33 9C\n\n Upon receiving a message with type server_hello, implementations MUST\n first examine the Random value and if it matches this value, process\n it as described in Section 4.1.4).\n\n TLS 1.3 has a downgrade protection mechanism embedded in the server's\n random value. TLS 1.3 servers which negotiate TLS 1.2 or below in\n response to a ClientHello MUST set the last eight bytes of their\n Random value specially.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 37]\n\f\nInternet-Draft TLS March 2018\n\n\n If negotiating TLS 1.2, TLS 1.3 servers MUST set the last eight bytes\n of their Random value to the bytes:\n\n 44 4F 57 4E 47 52 44 01\n\n If negotiating TLS 1.1 or below, TLS 1.3 servers MUST and TLS 1.2\n servers SHOULD set the last eight bytes of their Random value to the\n bytes:\n\n 44 4F 57 4E 47 52 44 00\n\n TLS 1.3 clients receiving a ServerHello indicating TLS 1.2 or below\n MUST check that the last eight bytes are not equal to either of these\n values. TLS 1.2 clients SHOULD also check that the last eight bytes\n are not equal to the second value if the ServerHello indicates TLS\n 1.1 or below. If a match is found, the client MUST abort the\n handshake with an \"illegal_parameter\" alert. This mechanism provides\n limited protection against downgrade attacks over and above what is\n provided by the Finished exchange: because the ServerKeyExchange, a\n message present in TLS 1.2 and below, includes a signature over both\n random values, it is not possible for an active attacker to modify\n the random values without detection as long as ephemeral ciphers are\n used. It does not provide downgrade protection when static RSA is\n used.\n\n Note: This is a change from [RFC5246], so in practice many TLS 1.2\n clients and servers will not behave as specified above.\n\n A legacy TLS client performing renegotiation with TLS 1.2 or prior\n and which receives a TLS 1.3 ServerHello during renegotiation MUST\n abort the handshake with a \"protocol_version\" alert. Note that\n renegotiation is not possible when TLS 1.3 has been negotiated.\n\n RFC EDITOR: PLEASE REMOVE THE FOLLOWING PARAGRAPH Implementations of\n draft versions (see Section 4.2.1.1) of this specification SHOULD NOT\n implement this mechanism on either client and server. A pre-RFC\n client connecting to RFC servers, or vice versa, will appear to\n downgrade to TLS 1.2. With the mechanism enabled, this will cause an\n interoperability failure.\n\n4.1.4. Hello Retry Request\n\n The server will send this message in response to a ClientHello\n message if it is able to find an acceptable set of parameters but the\n ClientHello does not contain sufficient information to proceed with\n the handshake. As discussed in Section 4.1.3, the HelloRetryRequest\n has the same format as a ServerHello message, and the legacy_version,\n legacy_session_id_echo, cipher_suite, and legacy_compression methods\n\n\n\nRescorla Expires September 21, 2018 [Page 38]\n\f\nInternet-Draft TLS March 2018\n\n\n fields have the same meaning. However, for convenience we discuss\n HelloRetryRequest throughout this document as if it were a distinct\n message.\n\n The server's extensions MUST contain \"supported_versions\" and\n otherwise the server SHOULD send only the extensions necessary for\n the client to generate a correct ClientHello pair. As with\n ServerHello, a HelloRetryRequest MUST NOT contain any extensions that\n were not first offered by the client in its ClientHello, with the\n exception of optionally the \"cookie\" (see Section 4.2.2) extension.\n\n Upon receipt of a HelloRetryRequest, the client MUST check the\n legacy_version, legacy_session_id_echo, cipher_suite, and\n legacy_compression_method as specified in Section 4.1.3 and then\n process the extensions, starting with determining the version using\n \"supported_versions\". Clients MUST abort the handshake with an\n \"illegal_parameter\" alert if the HelloRetryRequest would not result\n in any change in the ClientHello. If a client receives a second\n HelloRetryRequest in the same connection (i.e., where the ClientHello\n was itself in response to a HelloRetryRequest), it MUST abort the\n handshake with an \"unexpected_message\" alert.\n\n Otherwise, the client MUST process all extensions in the\n HelloRetryRequest and send a second updated ClientHello. The\n HelloRetryRequest extensions defined in this specification are:\n\n - supported_versions (see Section 4.2.1)\n\n - cookie (see Section 4.2.2)\n\n - key_share (see Section 4.2.8)\n\n In addition, in its updated ClientHello, the client SHOULD NOT offer\n any pre-shared keys associated with a hash other than that of the\n selected cipher suite. This allows the client to avoid having to\n compute partial hash transcripts for multiple hashes in the second\n ClientHello. A client which receives a cipher suite that was not\n offered MUST abort the handshake. Servers MUST ensure that they\n negotiate the same cipher suite when receiving a conformant updated\n ClientHello (if the server selects the cipher suite as the first step\n in the negotiation, then this will happen automatically). Upon\n receiving the ServerHello, clients MUST check that the cipher suite\n supplied in the ServerHello is the same as that in the\n HelloRetryRequest and otherwise abort the handshake with an\n \"illegal_parameter\" alert.\n\n The value of selected_version in the HelloRetryRequest\n \"supported_versions\" extension MUST be retained in the ServerHello,\n\n\n\nRescorla Expires September 21, 2018 [Page 39]\n\f\nInternet-Draft TLS March 2018\n\n\n and a client MUST abort the handshake with an \"illegal_parameter\"\n alert if the value changes.\n\n4.2. Extensions\n\n A number of TLS messages contain tag-length-value encoded extensions\n structures.\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n server_name(0), /* RFC 6066 */\n max_fragment_length(1), /* RFC 6066 */\n status_request(5), /* RFC 6066 */\n supported_groups(10), /* RFC 4492, 7919 */\n signature_algorithms(13), /* [[this document]] */\n use_srtp(14), /* RFC 5764 */\n heartbeat(15), /* RFC 6520 */\n application_layer_protocol_negotiation(16), /* RFC 7301 */\n signed_certificate_timestamp(18), /* RFC 6962 */\n client_certificate_type(19), /* RFC 7250 */\n server_certificate_type(20), /* RFC 7250 */\n padding(21), /* RFC 7685 */\n pre_shared_key(41), /* [[this document]] */\n early_data(42), /* [[this document]] */\n supported_versions(43), /* [[this document]] */\n cookie(44), /* [[this document]] */\n psk_key_exchange_modes(45), /* [[this document]] */\n certificate_authorities(47), /* [[this document]] */\n oid_filters(48), /* [[this document]] */\n post_handshake_auth(49), /* [[this document]] */\n signature_algorithms_cert(50), /* [[this document]] */\n key_share(51), /* [[this document]] */\n (65535)\n } ExtensionType;\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The list of extension types is maintained by IANA as described in\n Section 11.\n\n\n\nRescorla Expires September 21, 2018 [Page 40]\n\f\nInternet-Draft TLS March 2018\n\n\n Extensions are generally structured in a request/response fashion,\n though some extensions are just indications with no corresponding\n response. The client sends its extension requests in the ClientHello\n message and the server sends its extension responses in the\n ServerHello, EncryptedExtensions, HelloRetryRequest and Certificate\n messages. The server sends extension requests in the\n CertificateRequest message which a client MAY respond to with a\n Certificate message. The server MAY also send unsolicited extensions\n in the NewSessionTicket, though the client does not respond directly\n to these.\n\n Implementations MUST NOT send extension responses if the remote\n endpoint did not send the corresponding extension requests, with the\n exception of the \"cookie\" extension in HelloRetryRequest. Upon\n receiving such an extension, an endpoint MUST abort the handshake\n with an \"unsupported_extension\" alert.\n\n The table below indicates the messages where a given extension may\n appear, using the following notation: CH (ClientHello), SH\n (ServerHello), EE (EncryptedExtensions), CT (Certificate), CR\n (CertificateRequest), NST (NewSessionTicket) and HRR\n (HelloRetryRequest). If an implementation receives an extension\n which it recognizes and which is not specified for the message in\n which it appears it MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 41]\n\f\nInternet-Draft TLS March 2018\n\n\n +--------------------------------------------------+-------------+\n | Extension | TLS 1.3 |\n +--------------------------------------------------+-------------+\n | server_name [RFC6066] | CH, EE |\n | | |\n | max_fragment_length [RFC6066] | CH, EE |\n | | |\n | status_request [RFC6066] | CH, CR, CT |\n | | |\n | supported_groups [RFC7919] | CH, EE |\n | | |\n | signature_algorithms [RFC5246] | CH, CR |\n | | |\n | use_srtp [RFC5764] | CH, EE |\n | | |\n | heartbeat [RFC6520] | CH, EE |\n | | |\n | application_layer_protocol_negotiation [RFC7301] | CH, EE |\n | | |\n | signed_certificate_timestamp [RFC6962] | CH, CR, CT |\n | | |\n | client_certificate_type [RFC7250] | CH, EE |\n | | |\n | server_certificate_type [RFC7250] | CH, EE |\n | | |\n | padding [RFC7685] | CH |\n | | |\n | key_share [[this document]] | CH, SH, HRR |\n | | |\n | pre_shared_key [[this document]] | CH, SH |\n | | |\n | psk_key_exchange_modes [[this document]] | CH |\n | | |\n | early_data [[this document]] | CH, EE, NST |\n | | |\n | cookie [[this document]] | CH, HRR |\n | | |\n | supported_versions [[this document]] | CH, SH, HRR |\n | | |\n | certificate_authorities [[this document]] | CH, CR |\n | | |\n | oid_filters [[this document]] | CR |\n | | |\n | post_handshake_auth [[this document]] | CH |\n | | |\n | signature_algorithms_cert [[this document]] | CH, CR |\n +--------------------------------------------------+-------------+\n\n\n\n\nRescorla Expires September 21, 2018 [Page 42]\n\f\nInternet-Draft TLS March 2018\n\n\n When multiple extensions of different types are present, the\n extensions MAY appear in any order, with the exception of\n \"pre_shared_key\" Section 4.2.11 which MUST be the last extension in\n the ClientHello. There MUST NOT be more than one extension of the\n same type in a given extension block.\n\n In TLS 1.3, unlike TLS 1.2, extensions are negotiated for each\n handshake even when in resumption-PSK mode. However, 0-RTT\n parameters are those negotiated in the previous handshake; mismatches\n may require rejecting 0-RTT (see Section 4.2.10).\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions (e.g., the handshake cannot continue), and some are\n simply refusals to support particular features. In general, error\n alerts should be used for the former and a field in the server\n extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem. Often the fact that the extension fields are\n included in the inputs to the Finished message hashes will be\n sufficient, but extreme care is needed when the extension changes\n the meaning of messages sent in the handshake phase. Designers\n and implementors should be aware of the fact that until the\n handshake has been authenticated, active attackers can modify\n messages and insert, remove, or replace extensions.\n\n4.2.1. Supported Versions\n\n struct {\n select (Handshake.msg_type) {\n case client_hello:\n ProtocolVersion versions<2..254>;\n\n case server_hello: /* and HelloRetryRequest */\n ProtocolVersion selected_version;\n };\n } SupportedVersions;\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 43]\n\f\nInternet-Draft TLS March 2018\n\n\n The \"supported_versions\" extension is used by the client to indicate\n which versions of TLS it supports and by the server to indicate which\n version it is using. The extension contains a list of supported\n versions in preference order, with the most preferred version first.\n Implementations of this specification MUST send this extension in the\n ClientHello containing all versions of TLS which they are prepared to\n negotiate (for this specification, that means minimally 0x0304, but\n if previous versions of TLS are allowed to be negotiated, they MUST\n be present as well).\n\n If this extension is not present, servers which are compliant with\n this specification, and which also support TLS 1.2, MUST negotiate\n TLS 1.2 or prior as specified in [RFC5246], even if\n ClientHello.legacy_version is 0x0304 or later. Servers MAY abort the\n handshake upon receiving a ClientHello with legacy_version 0x0304 or\n later.\n\n If this extension is present in the ClientHello, servers MUST NOT use\n the ClientHello.legacy_version value for version negotiation and MUST\n use only the \"supported_versions\" extension to determine client\n preferences. Servers MUST only select a version of TLS present in\n that extension and MUST ignore any unknown versions that are present\n in that extension. Note that this mechanism makes it possible to\n negotiate a version prior to TLS 1.2 if one side supports a sparse\n range. Implementations of TLS 1.3 which choose to support prior\n versions of TLS SHOULD support TLS 1.2. Servers MUST be prepared to\n receive ClientHellos that include this extension but do not include\n 0x0304 in the list of versions.\n\n A server which negotiates a version of TLS prior to TLS 1.3 MUST set\n ServerHello.version and MUST NOT send the \"supported_versions\"\n extension. A server which negotiates TLS 1.3 MUST respond by sending\n a \"supported_versions\" extension containing the selected version\n value (0x0304). It MUST set the ServerHello.legacy_version field to\n 0x0303 (TLS 1.2). Clients MUST check for this extension prior to\n processing the rest of the ServerHello (although they will have to\n parse the ServerHello in order to read the extension). If this\n extension is present, clients MUST ignore the\n ServerHello.legacy_version value and MUST use only the\n \"supported_versions\" extension to determine the selected version. If\n the \"supported_versions\" extension in the ServerHello contains a\n version not offered by the client or contains a version prior to TLS\n 1.3, the client MUST abort the handshake with an \"illegal_parameter\"\n alert.\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 44]\n\f\nInternet-Draft TLS March 2018\n\n\n4.2.1.1. Draft Version Indicator\n\n RFC EDITOR: PLEASE REMOVE THIS SECTION\n\n While the eventual version indicator for the RFC version of TLS 1.3\n will be 0x0304, implementations of draft versions of this\n specification SHOULD instead advertise 0x7f00 | draft_version in the\n ServerHello and HelloRetryRequest \"supported_versions\" extension.\n For instance, draft-17 would be encoded as the 0x7f11. This allows\n pre-RFC implementations to safely negotiate with each other, even if\n they would otherwise be incompatible.\n\n4.2.2. Cookie\n\n struct {\n opaque cookie<1..2^16-1>;\n } Cookie;\n\n Cookies serve two primary purposes:\n\n - Allowing the server to force the client to demonstrate\n reachability at their apparent network address (thus providing a\n measure of DoS protection). This is primarily useful for non-\n connection-oriented transports (see [RFC6347] for an example of\n this).\n\n - Allowing the server to offload state to the client, thus allowing\n it to send a HelloRetryRequest without storing any state. The\n server can do this by storing the hash of the ClientHello in the\n HelloRetryRequest cookie (protected with some suitable integrity\n algorithm).\n\n When sending a HelloRetryRequest, the server MAY provide a \"cookie\"\n extension to the client (this is an exception to the usual rule that\n the only extensions that may be sent are those that appear in the\n ClientHello). When sending the new ClientHello, the client MUST copy\n the contents of the extension received in the HelloRetryRequest into\n a \"cookie\" extension in the new ClientHello. Clients MUST NOT use\n cookies in their initial ClientHello in subsequent connections.\n\n When a server is operating statelessly it may receive an unprotected\n record of type change_cipher_spec between the first and second\n ClientHello (see Section 5). Since the server is not storing any\n state this will appear as if it were the first message to be\n received. Servers operating statelessly MUST ignore these records.\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 45]\n\f\nInternet-Draft TLS March 2018\n\n\n4.2.3. Signature Algorithms\n\n TLS 1.3 provides two extensions for indicating which signature\n algorithms may be used in digital signatures. The\n \"signature_algorithms_cert\" extension applies to signatures in\n certificates and the \"signature_algorithms\" extension, which\n originally appeared in TLS 1.2, applies to signatures in\n CertificateVerify messages. The keys found in certificates MUST also\n be of appropriate type for the signature algorithms they are used\n with. This is a particular issue for RSA keys and PSS signatures, as\n described below. If no \"signature_algorithms_cert\" extension is\n present, then the \"signature_algorithms\" extension also applies to\n signatures appearing in certificates. Clients which desire the\n server to authenticate itself via a certificate MUST send\n \"signature_algorithms\". If a server is authenticating via a\n certificate and the client has not sent a \"signature_algorithms\"\n extension, then the server MUST abort the handshake with a\n \"missing_extension\" alert (see Section 9.2).\n\n The \"signature_algorithms_cert\" extension was added to allow\n implementations which supported different sets of algorithms for\n certificates and in TLS itself to clearly signal their capabilities.\n TLS 1.2 implementations SHOULD also process this extension.\n Implementations which have the same policy in both cases MAY omit the\n \"signature_algorithms_cert\" extension.\n\n The \"extension_data\" field of these extensions contains a\n SignatureSchemeList value:\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 46]\n\f\nInternet-Draft TLS March 2018\n\n\n enum {\n /* RSASSA-PKCS1-v1_5 algorithms */\n rsa_pkcs1_sha256(0x0401),\n rsa_pkcs1_sha384(0x0501),\n rsa_pkcs1_sha512(0x0601),\n\n /* ECDSA algorithms */\n ecdsa_secp256r1_sha256(0x0403),\n ecdsa_secp384r1_sha384(0x0503),\n ecdsa_secp521r1_sha512(0x0603),\n\n /* RSASSA-PSS algorithms with public key OID rsaEncryption */\n rsa_pss_rsae_sha256(0x0804),\n rsa_pss_rsae_sha384(0x0805),\n rsa_pss_rsae_sha512(0x0806),\n\n /* EdDSA algorithms */\n ed25519(0x0807),\n ed448(0x0808),\n\n /* RSASSA-PSS algorithms with public key OID RSASSA-PSS */\n rsa_pss_pss_sha256(0x0809),\n rsa_pss_pss_sha384(0x080a),\n rsa_pss_pss_sha512(0x080b),\n\n /* Legacy algorithms */\n rsa_pkcs1_sha1(0x0201),\n ecdsa_sha1(0x0203),\n\n /* Reserved Code Points */\n private_use(0xFE00..0xFFFF),\n (0xFFFF)\n } SignatureScheme;\n\n struct {\n SignatureScheme supported_signature_algorithms<2..2^16-2>;\n } SignatureSchemeList;\n\n Note: This enum is named \"SignatureScheme\" because there is already a\n \"SignatureAlgorithm\" type in TLS 1.2, which this replaces. We use\n the term \"signature algorithm\" throughout the text.\n\n Each SignatureScheme value lists a single signature algorithm that\n the client is willing to verify. The values are indicated in\n descending order of preference. Note that a signature algorithm\n takes as input an arbitrary-length message, rather than a digest.\n Algorithms which traditionally act on a digest should be defined in\n TLS to first hash the input with a specified hash algorithm and then\n\n\n\nRescorla Expires September 21, 2018 [Page 47]\n\f\nInternet-Draft TLS March 2018\n\n\n proceed as usual. The code point groups listed above have the\n following meanings:\n\n RSASSA-PKCS1-v1_5 algorithms Indicates a signature algorithm using\n RSASSA-PKCS1-v1_5 [RFC8017] with the corresponding hash algorithm\n as defined in [SHS]. These values refer solely to signatures\n which appear in certificates (see Section 4.4.2.2) and are not\n defined for use in signed TLS handshake messages, although they\n MAY appear in \"signature_algorithms\" and\n \"signature_algorithms_cert\" for backward compatibility with TLS\n 1.2,\n\n ECDSA algorithms Indicates a signature algorithm using ECDSA\n [ECDSA], the corresponding curve as defined in ANSI X9.62 [X962]\n and FIPS 186-4 [DSS], and the corresponding hash algorithm as\n defined in [SHS]. The signature is represented as a DER-encoded\n [X690] ECDSA-Sig-Value structure.\n\n RSASSA-PSS RSAE algorithms Indicates a signature algorithm using\n RSASSA-PSS [RFC8017] with mask generation function 1. The digest\n used in the mask generation function and the digest being signed\n are both the corresponding hash algorithm as defined in [SHS].\n The length of the salt MUST be equal to the length of the output\n of the digest algorithm. If the public key is carried in an X.509\n certificate, it MUST use the rsaEncryption OID [RFC5280].\n\n EdDSA algorithms Indicates a signature algorithm using EdDSA as\n defined in [RFC8032] or its successors. Note that these\n correspond to the \"PureEdDSA\" algorithms and not the \"prehash\"\n variants.\n\n RSASSA-PSS PSS algorithms Indicates a signature algorithm using\n RSASSA-PSS [RFC8017] with mask generation function 1. The digest\n used in the mask generation function and the digest being signed\n are both the corresponding hash algorithm as defined in [SHS].\n The length of the salt MUST be equal to the length of the digest\n algorithm. If the public key is carried in an X.509 certificate,\n it MUST use the RSASSA-PSS OID [RFC5756]. When used in\n certificate signatures, the algorithm parameters MUST be DER\n encoded. If the corresponding public key's parameters are\n present, then the parameters in the signature MUST be identical to\n those in the public key.\n\n Legacy algorithms Indicates algorithms which are being deprecated\n because they use algorithms with known weaknesses, specifically\n SHA-1 which is used in this context with either with RSA using\n RSASSA-PKCS1-v1_5 or ECDSA. These values refer solely to\n signatures which appear in certificates (see Section 4.4.2.2) and\n\n\n\nRescorla Expires September 21, 2018 [Page 48]\n\f\nInternet-Draft TLS March 2018\n\n\n are not defined for use in signed TLS handshake messages, although\n they MAY appear in \"signature_algorithms\" and\n \"signature_algorithms_cert\" for backward compatibility with TLS\n 1.2, Endpoints SHOULD NOT negotiate these algorithms but are\n permitted to do so solely for backward compatibility. Clients\n offering these values MUST list them as the lowest priority\n (listed after all other algorithms in SignatureSchemeList). TLS\n 1.3 servers MUST NOT offer a SHA-1 signed certificate unless no\n valid certificate chain can be produced without it (see\n Section 4.4.2.2).\n\n The signatures on certificates that are self-signed or certificates\n that are trust anchors are not validated since they begin a\n certification path (see [RFC5280], Section 3.2). A certificate that\n begins a certification path MAY use a signature algorithm that is not\n advertised as being supported in the \"signature_algorithms\"\n extension.\n\n Note that TLS 1.2 defines this extension differently. TLS 1.3\n implementations willing to negotiate TLS 1.2 MUST behave in\n accordance with the requirements of [RFC5246] when negotiating that\n version. In particular:\n\n - TLS 1.2 ClientHellos MAY omit this extension.\n\n - In TLS 1.2, the extension contained hash/signature pairs. The\n pairs are encoded in two octets, so SignatureScheme values have\n been allocated to align with TLS 1.2's encoding. Some legacy\n pairs are left unallocated. These algorithms are deprecated as of\n TLS 1.3. They MUST NOT be offered or negotiated by any\n implementation. In particular, MD5 [SLOTH], SHA-224, and DSA MUST\n NOT be used.\n\n - ECDSA signature schemes align with TLS 1.2's ECDSA hash/signature\n pairs. However, the old semantics did not constrain the signing\n curve. If TLS 1.2 is negotiated, implementations MUST be prepared\n to accept a signature that uses any curve that they advertised in\n the \"supported_groups\" extension.\n\n - Implementations that advertise support for RSASSA-PSS (which is\n mandatory in TLS 1.3), MUST be prepared to accept a signature\n using that scheme even when TLS 1.2 is negotiated. In TLS 1.2,\n RSASSA-PSS is used with RSA cipher suites.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 49]\n\f\nInternet-Draft TLS March 2018\n\n\n4.2.4. Certificate Authorities\n\n The \"certificate_authorities\" extension is used to indicate the\n certificate authorities which an endpoint supports and which SHOULD\n be used by the receiving endpoint to guide certificate selection.\n\n The body of the \"certificate_authorities\" extension consists of a\n CertificateAuthoritiesExtension structure.\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n DistinguishedName authorities<3..2^16-1>;\n } CertificateAuthoritiesExtension;\n\n authorities A list of the distinguished names [X501] of acceptable\n certificate authorities, represented in DER-encoded [X690] format.\n These distinguished names specify a desired distinguished name for\n trust anchor or subordinate CA; thus, this message can be used to\n describe known trust anchors as well as a desired authorization\n space.\n\n The client MAY send the \"certificate_authorities\" extension in the\n ClientHello message. The server MAY send it in the\n CertificateRequest message.\n\n The \"trusted_ca_keys\" extension, which serves a similar purpose\n [RFC6066], but is more complicated, is not used in TLS 1.3 (although\n it may appear in ClientHello messages from clients which are offering\n prior versions of TLS).\n\n4.2.5. OID Filters\n\n The \"oid_filters\" extension allows servers to provide a set of OID/\n value pairs which it would like the client's certificate to match.\n This extension, if provided by the server, MUST only be sent in the\n CertificateRequest message.\n\n struct {\n opaque certificate_extension_oid<1..2^8-1>;\n opaque certificate_extension_values<0..2^16-1>;\n } OIDFilter;\n\n struct {\n OIDFilter filters<0..2^16-1>;\n } OIDFilterExtension;\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 50]\n\f\nInternet-Draft TLS March 2018\n\n\n filters A list of certificate extension OIDs [RFC5280] with their\n allowed value(s) and represented in DER-encoded [X690] format.\n Some certificate extension OIDs allow multiple values (e.g.,\n Extended Key Usage). If the server has included a non-empty\n filters list, the client certificate included in the response MUST\n contain all of the specified extension OIDs that the client\n recognizes. For each extension OID recognized by the client, all\n of the specified values MUST be present in the client certificate\n (but the certificate MAY have other values as well). However, the\n client MUST ignore and skip any unrecognized certificate extension\n OIDs. If the client ignored some of the required certificate\n extension OIDs and supplied a certificate that does not satisfy\n the request, the server MAY at its discretion either continue the\n connection without client authentication, or abort the handshake\n with an \"unsupported_certificate\" alert. Any given OID MUST NOT\n appear more than once in the filters list.\n\n PKIX RFCs define a variety of certificate extension OIDs and their\n corresponding value types. Depending on the type, matching\n certificate extension values are not necessarily bitwise-equal. It\n is expected that TLS implementations will rely on their PKI libraries\n to perform certificate selection using certificate extension OIDs.\n\n This document defines matching rules for two standard certificate\n extensions defined in [RFC5280]:\n\n - The Key Usage extension in a certificate matches the request when\n all key usage bits asserted in the request are also asserted in\n the Key Usage certificate extension.\n\n - The Extended Key Usage extension in a certificate matches the\n request when all key purpose OIDs present in the request are also\n found in the Extended Key Usage certificate extension. The\n special anyExtendedKeyUsage OID MUST NOT be used in the request.\n\n Separate specifications may define matching rules for other\n certificate extensions.\n\n4.2.6. Post-Handshake Client Authentication\n\n The \"post_handshake_auth\" extension is used to indicate that a client\n is willing to perform post-handshake authentication (Section 4.6.2).\n Servers MUST NOT send a post-handshake CertificateRequest to clients\n which do not offer this extension. Servers MUST NOT send this\n extension.\n\n struct {} PostHandshakeAuth;\n\n\n\n\nRescorla Expires September 21, 2018 [Page 51]\n\f\nInternet-Draft TLS March 2018\n\n\n The \"extension_data\" field of the \"post_handshake_auth\" extension is\n zero length.\n\n4.2.7. Negotiated Groups\n\n When sent by the client, the \"supported_groups\" extension indicates\n the named groups which the client supports for key exchange, ordered\n from most preferred to least preferred.\n\n Note: In versions of TLS prior to TLS 1.3, this extension was named\n \"elliptic_curves\" and only contained elliptic curve groups. See\n [RFC4492] and [RFC7919]. This extension was also used to negotiate\n ECDSA curves. Signature algorithms are now negotiated independently\n (see Section 4.2.3).\n\n The \"extension_data\" field of this extension contains a\n \"NamedGroupList\" value:\n\n enum {\n\n /* Elliptic Curve Groups (ECDHE) */\n secp256r1(0x0017), secp384r1(0x0018), secp521r1(0x0019),\n x25519(0x001D), x448(0x001E),\n\n /* Finite Field Groups (DHE) */\n ffdhe2048(0x0100), ffdhe3072(0x0101), ffdhe4096(0x0102),\n ffdhe6144(0x0103), ffdhe8192(0x0104),\n\n /* Reserved Code Points */\n ffdhe_private_use(0x01FC..0x01FF),\n ecdhe_private_use(0xFE00..0xFEFF),\n (0xFFFF)\n } NamedGroup;\n\n struct {\n NamedGroup named_group_list<2..2^16-1>;\n } NamedGroupList;\n\n Elliptic Curve Groups (ECDHE) Indicates support for the\n corresponding named curve, defined either in FIPS 186-4 [DSS] or\n in [RFC7748]. Values 0xFE00 through 0xFEFF are reserved for\n private use.\n\n Finite Field Groups (DHE) Indicates support of the corresponding\n finite field group, defined in [RFC7919]. Values 0x01FC through\n 0x01FF are reserved for private use.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 52]\n\f\nInternet-Draft TLS March 2018\n\n\n Items in named_group_list are ordered according to the client's\n preferences (most preferred choice first).\n\n As of TLS 1.3, servers are permitted to send the \"supported_groups\"\n extension to the client. Clients MUST NOT act upon any information\n found in \"supported_groups\" prior to successful completion of the\n handshake but MAY use the information learned from a successfully\n completed handshake to change what groups they use in their\n \"key_share\" extension in subsequent connections. If the server has a\n group it prefers to the ones in the \"key_share\" extension but is\n still willing to accept the ClientHello, it SHOULD send\n \"supported_groups\" to update the client's view of its preferences;\n this extension SHOULD contain all groups the server supports,\n regardless of whether they are currently supported by the client.\n\n4.2.8. Key Share\n\n The \"key_share\" extension contains the endpoint's cryptographic\n parameters.\n\n Clients MAY send an empty client_shares vector in order to request\n group selection from the server at the cost of an additional round\n trip. (see Section 4.1.4)\n\n struct {\n NamedGroup group;\n opaque key_exchange<1..2^16-1>;\n } KeyShareEntry;\n\n group The named group for the key being exchanged.\n\n key_exchange Key exchange information. The contents of this field\n are determined by the specified group and its corresponding\n definition. Finite Field Diffie-Hellman [DH] parameters are\n described in Section 4.2.8.1; Elliptic Curve Diffie-Hellman\n parameters are described in Section 4.2.8.2.\n\n In the ClientHello message, the \"extension_data\" field of this\n extension contains a \"KeyShareClientHello\" value:\n\n struct {\n KeyShareEntry client_shares<0..2^16-1>;\n } KeyShareClientHello;\n\n client_shares A list of offered KeyShareEntry values in descending\n order of client preference.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 53]\n\f\nInternet-Draft TLS March 2018\n\n\n This vector MAY be empty if the client is requesting a\n HelloRetryRequest. Each KeyShareEntry value MUST correspond to a\n group offered in the \"supported_groups\" extension and MUST appear in\n the same order. However, the values MAY be a non-contiguous subset\n of the \"supported_groups\" extension and MAY omit the most preferred\n groups. Such a situation could arise if the most preferred groups\n are new and unlikely to be supported in enough places to make\n pregenerating key shares for them efficient.\n\n Clients can offer as many KeyShareEntry values as the number of\n supported groups it is offering, each representing a single set of\n key exchange parameters. For instance, a client might offer shares\n for several elliptic curves or multiple FFDHE groups. The\n key_exchange values for each KeyShareEntry MUST be generated\n independently. Clients MUST NOT offer multiple KeyShareEntry values\n for the same group. Clients MUST NOT offer any KeyShareEntry values\n for groups not listed in the client's \"supported_groups\" extension.\n Servers MAY check for violations of these rules and abort the\n handshake with an \"illegal_parameter\" alert if one is violated.\n\n In a HelloRetryRequest message, the \"extension_data\" field of this\n extension contains a KeyShareHelloRetryRequest value:\n\n struct {\n NamedGroup selected_group;\n } KeyShareHelloRetryRequest;\n\n selected_group The mutually supported group the server intends to\n negotiate and is requesting a retried ClientHello/KeyShare for.\n\n Upon receipt of this extension in a HelloRetryRequest, the client\n MUST verify that (1) the selected_group field corresponds to a group\n which was provided in the \"supported_groups\" extension in the\n original ClientHello; and (2) the selected_group field does not\n correspond to a group which was provided in the \"key_share\" extension\n in the original ClientHello. If either of these checks fails, then\n the client MUST abort the handshake with an \"illegal_parameter\"\n alert. Otherwise, when sending the new ClientHello, the client MUST\n replace the original \"key_share\" extension with one containing only a\n new KeyShareEntry for the group indicated in the selected_group field\n of the triggering HelloRetryRequest.\n\n In a ServerHello message, the \"extension_data\" field of this\n extension contains a KeyShareServerHello value:\n\n struct {\n KeyShareEntry server_share;\n } KeyShareServerHello;\n\n\n\nRescorla Expires September 21, 2018 [Page 54]\n\f\nInternet-Draft TLS March 2018\n\n\n server_share A single KeyShareEntry value that is in the same group\n as one of the client's shares.\n\n If using (EC)DHE key establishment, servers offer exactly one\n KeyShareEntry in the ServerHello. This value MUST be in the same\n group as the KeyShareEntry value offered by the client that the\n server has selected for the negotiated key exchange. Servers MUST\n NOT send a KeyShareEntry for any group not indicated in the\n \"supported_groups\" extension and MUST NOT send a KeyShareEntry when\n using the \"psk_ke\" PskKeyExchangeMode. If using (EC)DHE key\n establishment, and a HelloRetryRequest containing a \"key_share\"\n extension was received by the client, the client MUST verify that the\n selected NamedGroup in the ServerHello is the same as that in the\n HelloRetryRequest. If this check fails, the client MUST abort the\n handshake with an \"illegal_parameter\" alert.\n\n4.2.8.1. Diffie-Hellman Parameters\n\n Diffie-Hellman [DH] parameters for both clients and servers are\n encoded in the opaque key_exchange field of a KeyShareEntry in a\n KeyShare structure. The opaque value contains the Diffie-Hellman\n public value (Y = g^X mod p) for the specified group (see [RFC7919]\n for group definitions) encoded as a big-endian integer and padded to\n the left with zeros to the size of p in bytes.\n\n Note: For a given Diffie-Hellman group, the padding results in all\n public keys having the same length.\n\n Peers MUST validate each other's public key Y by ensuring that 1 < Y\n < p-1. This check ensures that the remote peer is properly behaved\n and isn't forcing the local system into a small subgroup.\n\n4.2.8.2. ECDHE Parameters\n\n ECDHE parameters for both clients and servers are encoded in the\n opaque key_exchange field of a KeyShareEntry in a KeyShare structure.\n\n For secp256r1, secp384r1 and secp521r1, the contents are the\n serialized value of the following struct:\n\n struct {\n uint8 legacy_form = 4;\n opaque X[coordinate_length];\n opaque Y[coordinate_length];\n } UncompressedPointRepresentation;\n\n X and Y respectively are the binary representations of the x and y\n values in network byte order. There are no internal length markers,\n\n\n\nRescorla Expires September 21, 2018 [Page 55]\n\f\nInternet-Draft TLS March 2018\n\n\n so each number representation occupies as many octets as implied by\n the curve parameters. For P-256 this means that each of X and Y use\n 32 octets, padded on the left by zeros if necessary. For P-384 they\n take 48 octets each, and for P-521 they take 66 octets each.\n\n For the curves secp256r1, secp384r1 and secp521r1, peers MUST\n validate each other's public value Q by ensuring that the point is a\n valid point on the elliptic curve. The appropriate validation\n procedures are defined in Section 4.3.7 of [X962] and alternatively\n in Section 5.6.2.3 of [KEYAGREEMENT]. This process consists of three\n steps: (1) verify that Q is not the point at infinity (O), (2) verify\n that for Q = (x, y) both integers x and y are in the correct\n interval, (3) ensure that (x, y) is a correct solution to the\n elliptic curve equation. For these curves, implementers do not need\n to verify membership in the correct subgroup.\n\n For X25519 and X448, the contents of the public value are the byte\n string inputs and outputs of the corresponding functions defined in\n [RFC7748], 32 bytes for X25519 and 56 bytes for X448.\n\n Note: Versions of TLS prior to 1.3 permitted point format\n negotiation; TLS 1.3 removes this feature in favor of a single point\n format for each curve.\n\n4.2.9. Pre-Shared Key Exchange Modes\n\n In order to use PSKs, clients MUST also send a\n \"psk_key_exchange_modes\" extension. The semantics of this extension\n are that the client only supports the use of PSKs with these modes,\n which restricts both the use of PSKs offered in this ClientHello and\n those which the server might supply via NewSessionTicket.\n\n A client MUST provide a \"psk_key_exchange_modes\" extension if it\n offers a \"pre_shared_key\" extension. If clients offer\n \"pre_shared_key\" without a \"psk_key_exchange_modes\" extension,\n servers MUST abort the handshake. Servers MUST NOT select a key\n exchange mode that is not listed by the client. This extension also\n restricts the modes for use with PSK resumption; servers SHOULD NOT\n send NewSessionTicket with tickets that are not compatible with the\n advertised modes; however, if a server does so, the impact will just\n be that the client's attempts at resumption fail.\n\n The server MUST NOT send a \"psk_key_exchange_modes\" extension.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 56]\n\f\nInternet-Draft TLS March 2018\n\n\n enum { psk_ke(0), psk_dhe_ke(1), (255) } PskKeyExchangeMode;\n\n struct {\n PskKeyExchangeMode ke_modes<1..255>;\n } PskKeyExchangeModes;\n\n psk_ke PSK-only key establishment. In this mode, the server MUST\n NOT supply a \"key_share\" value.\n\n psk_dhe_ke PSK with (EC)DHE key establishment. In this mode, the\n client and server MUST supply \"key_share\" values as described in\n Section 4.2.8.\n\n Any future values that are allocated must ensure that the transmitted\n protocol messages unambiguously identify which mode was selected by\n the server; at present, this is indicated by the presence of the\n \"key_share\" in the ServerHello.\n\n4.2.10. Early Data Indication\n\n When a PSK is used and early data is allowed for that PSK, the client\n can send application data in its first flight of messages. If the\n client opts to do so, it MUST supply both the \"early_data\" extension\n as well as the \"pre_shared_key\" extension.\n\n The \"extension_data\" field of this extension contains an\n \"EarlyDataIndication\" value.\n\n struct {} Empty;\n\n struct {\n select (Handshake.msg_type) {\n case new_session_ticket: uint32 max_early_data_size;\n case client_hello: Empty;\n case encrypted_extensions: Empty;\n };\n } EarlyDataIndication;\n\n See Section 4.6.1 for the use of the max_early_data_size field.\n\n The parameters for the 0-RTT data (version, symmetric cipher suite,\n ALPN protocol, etc.) are those associated with the PSK in use. For\n externally provisioned PSKs, the associated values are those\n provisioned along with the key. For PSKs established via a\n NewSessionTicket message, the associated values are those which were\n negotiated in the connection which established the PSK. The PSK used\n to encrypt the early data MUST be the first PSK listed in the\n client's \"pre_shared_key\" extension.\n\n\n\nRescorla Expires September 21, 2018 [Page 57]\n\f\nInternet-Draft TLS March 2018\n\n\n For PSKs provisioned via NewSessionTicket, a server MUST validate\n that the ticket age for the selected PSK identity (computed by\n subtracting ticket_age_add from PskIdentity.obfuscated_ticket_age\n modulo 2^32) is within a small tolerance of the time since the ticket\n was issued (see Section 8). If it is not, the server SHOULD proceed\n with the handshake but reject 0-RTT, and SHOULD NOT take any other\n action that assumes that this ClientHello is fresh.\n\n 0-RTT messages sent in the first flight have the same (encrypted)\n content types as messages of the same type sent in other flights\n (handshake and application_data) but are protected under different\n keys. After receiving the server's Finished message, if the server\n has accepted early data, an EndOfEarlyData message will be sent to\n indicate the key change. This message will be encrypted with the\n 0-RTT traffic keys.\n\n A server which receives an \"early_data\" extension MUST behave in one\n of three ways:\n\n - Ignore the extension and return a regular 1-RTT response. The\n server then skips past early data by attempting to deprotect\n received records using the handshake traffic key, discarding\n records which fail deprotection (up to the configured\n max_early_data_size). Once a record is deprotected successfully,\n it is treated as the start of the client's second flight and the\n the server proceeds as with an ordinary 1-RTT handshake.\n\n - Request that the client send another ClientHello by responding\n with a HelloRetryRequest. A client MUST NOT include the\n \"early_data\" extension in its followup ClientHello. The server\n then ignores early data by skipping all records with external\n content type of \"application_data\" (indicating that they are\n encrypted), up to the configured max_early_data_size.\n\n - Return its own \"early_data\" extension in EncryptedExtensions,\n indicating that it intends to process the early data. It is not\n possible for the server to accept only a subset of the early data\n messages. Even though the server sends a message accepting early\n data, the actual early data itself may already be in flight by the\n time the server generates this message.\n\n In order to accept early data, the server MUST have accepted a PSK\n cipher suite and selected the first key offered in the client's\n \"pre_shared_key\" extension. In addition, it MUST verify that the\n following values are the same as those associated with the selected\n PSK:\n\n - The TLS version number\n\n\n\nRescorla Expires September 21, 2018 [Page 58]\n\f\nInternet-Draft TLS March 2018\n\n\n - The selected cipher suite\n\n - The selected ALPN [RFC7301] protocol, if any\n\n These requirements are a superset of those needed to perform a 1-RTT\n handshake using the PSK in question. For externally established\n PSKs, the associated values are those provisioned along with the key.\n For PSKs established via a NewSessionTicket message, the associated\n values are those negotiated in the connection during which the ticket\n was established.\n\n Future extensions MUST define their interaction with 0-RTT.\n\n If any of these checks fail, the server MUST NOT respond with the\n extension and must discard all the first flight data using one of the\n first two mechanisms listed above (thus falling back to 1-RTT or\n 2-RTT). If the client attempts a 0-RTT handshake but the server\n rejects it, the server will generally not have the 0-RTT record\n protection keys and must instead use trial decryption (either with\n the 1-RTT handshake keys or by looking for a cleartext ClientHello in\n the case of HelloRetryRequest) to find the first non-0-RTT message.\n\n If the server chooses to accept the \"early_data\" extension, then it\n MUST comply with the same error handling requirements specified for\n all records when processing early data records. Specifically, if the\n server fails to decrypt a 0-RTT record following an accepted\n \"early_data\" extension it MUST terminate the connection with a\n \"bad_record_mac\" alert as per Section 5.2.\n\n If the server rejects the \"early_data\" extension, the client\n application MAY opt to retransmit the application data previously\n sent in early data once the handshake has been completed. Note that\n automatic re-transmission of early data could result in assumptions\n about the status of the connection being incorrect. For instance,\n when the negotiated connection selects a different ALPN protocol from\n what was used for the early data, an application might need to\n construct different messages. Similarly, if early data assumes\n anything about the connection state, it might be sent in error after\n the handshake completes.\n\n A TLS implementation SHOULD NOT automatically re-send early data;\n applications are in a better position to decide when re-transmission\n is appropriate. A TLS implementation MUST NOT automatically re-send\n early data unless the negotiated connection selects the same ALPN\n protocol.\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 59]\n\f\nInternet-Draft TLS March 2018\n\n\n4.2.11. Pre-Shared Key Extension\n\n The \"pre_shared_key\" extension is used to negotiate the identity of\n the pre-shared key to be used with a given handshake in association\n with PSK key establishment.\n\n The \"extension_data\" field of this extension contains a\n \"PreSharedKeyExtension\" value:\n\n struct {\n opaque identity<1..2^16-1>;\n uint32 obfuscated_ticket_age;\n } PskIdentity;\n\n opaque PskBinderEntry<32..255>;\n\n struct {\n PskIdentity identities<7..2^16-1>;\n PskBinderEntry binders<33..2^16-1>;\n } OfferedPsks;\n\n struct {\n select (Handshake.msg_type) {\n case client_hello: OfferedPsks;\n case server_hello: uint16 selected_identity;\n };\n } PreSharedKeyExtension;\n\n identity A label for a key. For instance, a ticket defined in\n Appendix B.3.4 or a label for a pre-shared key established\n externally.\n\n obfuscated_ticket_age An obfuscated version of the age of the key.\n Section 4.2.11.1 describes how to form this value for identities\n established via the NewSessionTicket message. For identities\n established externally an obfuscated_ticket_age of 0 SHOULD be\n used, and servers MUST ignore the value.\n\n identities A list of the identities that the client is willing to\n negotiate with the server. If sent alongside the \"early_data\"\n extension (see Section 4.2.10), the first identity is the one used\n for 0-RTT data.\n\n binders A series of HMAC values, one for each PSK offered in the\n \"pre_shared_keys\" extension and in the same order, computed as\n described below.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 60]\n\f\nInternet-Draft TLS March 2018\n\n\n selected_identity The server's chosen identity expressed as a\n (0-based) index into the identities in the client's list.\n\n Each PSK is associated with a single Hash algorithm. For PSKs\n established via the ticket mechanism (Section 4.6.1), this is the KDF\n Hash algorithm on the connection where the ticket was established.\n For externally established PSKs, the Hash algorithm MUST be set when\n the PSK is established, or default to SHA-256 if no such algorithm is\n defined. The server MUST ensure that it selects a compatible PSK (if\n any) and cipher suite.\n\n In TLS versions prior to TLS 1.3, the Server Name Identification\n (SNI) value was intended to be associated with the session (Section 3\n of [RFC6066]), with the server being required to enforce that the SNI\n value associated with the session matches the one specified in the\n resumption handshake. However, in reality the implementations were\n not consistent on which of two supplied SNI values they would use,\n leading to the consistency requirement being de-facto enforced by the\n clients. In TLS 1.3, the SNI value is always explicitly specified in\n the resumption handshake, and there is no need for the server to\n associate an SNI value with the ticket. Clients, however, SHOULD\n store the SNI with the PSK to fulfill the requirements of\n Section 4.6.1.\n\n Implementor's note: when session resumption is the primary use case\n of PSKs the most straightforward way to implement the PSK/cipher\n suite matching requirements is to negotiate the cipher suite first\n and then exclude any incompatible PSKs. Any unknown PSKs (e.g., they\n are not in the PSK database or are encrypted with an unknown key)\n SHOULD simply be ignored. If no acceptable PSKs are found, the\n server SHOULD perform a non-PSK handshake if possible. If backwards\n compatibility is important, client provided, externally established\n PSKs SHOULD influence cipher suite selection.\n\n Prior to accepting PSK key establishment, the server MUST validate\n the corresponding binder value (see Section 4.2.11.2 below). If this\n value is not present or does not validate, the server MUST abort the\n handshake. Servers SHOULD NOT attempt to validate multiple binders;\n rather they SHOULD select a single PSK and validate solely the binder\n that corresponds to that PSK. See [Section 8.2] and [Appendix E.6]\n for the security rationale for this requirement. In order to accept\n PSK key establishment, the server sends a \"pre_shared_key\" extension\n indicating the selected identity.\n\n Clients MUST verify that the server's selected_identity is within the\n range supplied by the client, that the server selected a cipher suite\n indicating a Hash associated with the PSK and that a server\n \"key_share\" extension is present if required by the ClientHello\n\n\n\nRescorla Expires September 21, 2018 [Page 61]\n\f\nInternet-Draft TLS March 2018\n\n\n \"psk_key_exchange_modes\". If these values are not consistent the\n client MUST abort the handshake with an \"illegal_parameter\" alert.\n\n If the server supplies an \"early_data\" extension, the client MUST\n verify that the server's selected_identity is 0. If any other value\n is returned, the client MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n The \"pre_shared_key\" extension MUST be the last extension in the\n ClientHello (this facilitates implementation as described below).\n Servers MUST check that it is the last extension and otherwise fail\n the handshake with an \"illegal_parameter\" alert.\n\n4.2.11.1. Ticket Age\n\n The client's view of the age of a ticket is the time since the\n receipt of the NewSessionTicket message. Clients MUST NOT attempt to\n use tickets which have ages greater than the \"ticket_lifetime\" value\n which was provided with the ticket. The \"obfuscated_ticket_age\"\n field of each PskIdentity contains an obfuscated version of the\n ticket age formed by taking the age in milliseconds and adding the\n \"ticket_age_add\" value that was included with the ticket (see\n Section 4.6.1), modulo 2^32. This addition prevents passive\n observers from correlating connections unless tickets are reused.\n Note that the \"ticket_lifetime\" field in the NewSessionTicket message\n is in seconds but the \"obfuscated_ticket_age\" is in milliseconds.\n Because ticket lifetimes are restricted to a week, 32 bits is enough\n to represent any plausible age, even in milliseconds.\n\n4.2.11.2. PSK Binder\n\n The PSK binder value forms a binding between a PSK and the current\n handshake, as well as a binding between the handshake in which the\n PSK was generated (if via a NewSessionTicket message) and the current\n handshake. Each entry in the binders list is computed as an HMAC\n over a transcript hash (see Section 4.4.1) containing a partial\n ClientHello up to and including the PreSharedKeyExtension.identities\n field. That is, it includes all of the ClientHello but not the\n binders list itself. The length fields for the message (including\n the overall length, the length of the extensions block, and the\n length of the \"pre_shared_key\" extension) are all set as if binders\n of the correct lengths were present.\n\n The PskBinderEntry is computed in the same way as the Finished\n message (Section 4.4.4) but with the BaseKey being the binder_key\n derived via the key schedule from the corresponding PSK which is\n being offered (see Section 7.1).\n\n\n\n\nRescorla Expires September 21, 2018 [Page 62]\n\f\nInternet-Draft TLS March 2018\n\n\n If the handshake includes a HelloRetryRequest, the initial\n ClientHello and HelloRetryRequest are included in the transcript\n along with the new ClientHello. For instance, if the client sends\n ClientHello1, its binder will be computed over:\n\n Transcript-Hash(Truncate(ClientHello1))\n\n Where Truncate() removes the binders list from the ClientHello.\n\n If the server responds with HelloRetryRequest, and the client then\n sends ClientHello2, its binder will be computed over:\n\n Transcript-Hash(ClientHello1,\n HelloRetryRequest,\n Truncate(ClientHello2))\n\n The full ClientHello1/ClientHello2 is included in all other handshake\n hash computations. Note that in the first flight,\n Truncate(ClientHello1) is hashed directly, but in the second flight,\n ClientHello1 is hashed and then reinjected as a \"message_hash\"\n message, as described in Section 4.4.1.\n\n4.2.11.3. Processing Order\n\n Clients are permitted to \"stream\" 0-RTT data until they receive the\n server's Finished, only then sending the EndOfEarlyData message,\n followed by the rest of the handshake. In order to avoid deadlocks,\n when accepting \"early_data\", servers MUST process the client's\n ClientHello and then immediately send their flight of messages,\n rather than waiting for the client's EndOfEarlyData message before\n sending its ServerHello.\n\n4.3. Server Parameters\n\n The next two messages from the server, EncryptedExtensions and\n CertificateRequest, contain information from the server that\n determines the rest of the handshake. These messages are encrypted\n with keys derived from the server_handshake_traffic_secret.\n\n4.3.1. Encrypted Extensions\n\n In all handshakes, the server MUST send the EncryptedExtensions\n message immediately after the ServerHello message. This is the first\n message that is encrypted under keys derived from the\n server_handshake_traffic_secret.\n\n The EncryptedExtensions message contains extensions that can be\n protected, i.e., any which are not needed to establish the\n\n\n\nRescorla Expires September 21, 2018 [Page 63]\n\f\nInternet-Draft TLS March 2018\n\n\n cryptographic context, but which are not associated with individual\n certificates. The client MUST check EncryptedExtensions for the\n presence of any forbidden extensions and if any are found MUST abort\n the handshake with an \"illegal_parameter\" alert.\n\n Structure of this message:\n\n struct {\n Extension extensions<0..2^16-1>;\n } EncryptedExtensions;\n\n extensions A list of extensions. For more information, see the\n table in Section 4.2.\n\n4.3.2. Certificate Request\n\n A server which is authenticating with a certificate MAY optionally\n request a certificate from the client. This message, if sent, MUST\n follow EncryptedExtensions.\n\n Structure of this message:\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n Extension extensions<2..2^16-1>;\n } CertificateRequest;\n\n certificate_request_context An opaque string which identifies the\n certificate request and which will be echoed in the client's\n Certificate message. The certificate_request_context MUST be\n unique within the scope of this connection (thus preventing replay\n of client CertificateVerify messages). This field SHALL be zero\n length unless used for the post-handshake authentication exchanges\n described in Section 4.6.2. When requesting post-handshake\n authentication, the server SHOULD make the context unpredictable\n to the client (e.g., by randomly generating it) in order to\n prevent an attacker who has temporary access to the client's\n private key from pre-computing valid CertificateVerify messages.\n\n extensions A set of extensions describing the parameters of the\n certificate being requested. The \"signature_algorithms\" extension\n MUST be specified, and other extensions may optionally be included\n if defined for this message. Clients MUST ignore unrecognized\n extensions.\n\n In prior versions of TLS, the CertificateRequest message carried a\n list of signature algorithms and certificate authorities which the\n server would accept. In TLS 1.3 the former is expressed by sending\n\n\n\nRescorla Expires September 21, 2018 [Page 64]\n\f\nInternet-Draft TLS March 2018\n\n\n the \"signature_algorithms\" and optionally \"signature_algorithms_cert\"\n extensions. The latter is expressed by sending the\n \"certificate_authorities\" extension (see Section 4.2.4).\n\n Servers which are authenticating with a PSK MUST NOT send the\n CertificateRequest message in the main handshake, though they MAY\n send it in post-handshake authentication (see Section 4.6.2) provided\n that the client has sent the \"post_handshake_auth\" extension (see\n Section 4.2.6).\n\n4.4. Authentication Messages\n\n As discussed in Section 2, TLS generally uses a common set of\n messages for authentication, key confirmation, and handshake\n integrity: Certificate, CertificateVerify, and Finished. (The\n PreSharedKey binders also perform key confirmation, in a similar\n fashion.) These three messages are always sent as the last messages\n in their handshake flight. The Certificate and CertificateVerify\n messages are only sent under certain circumstances, as defined below.\n The Finished message is always sent as part of the Authentication\n block. These messages are encrypted under keys derived from\n [sender]_handshake_traffic_secret.\n\n The computations for the Authentication messages all uniformly take\n the following inputs:\n\n - The certificate and signing key to be used.\n\n - A Handshake Context consisting of the set of messages to be\n included in the transcript hash.\n\n - A base key to be used to compute a MAC key.\n\n Based on these inputs, the messages then contain:\n\n Certificate The certificate to be used for authentication, and any\n supporting certificates in the chain. Note that certificate-based\n client authentication is not available in PSK (including 0-RTT)\n flows.\n\n CertificateVerify A signature over the value Transcript-\n Hash(Handshake Context, Certificate)\n\n Finished A MAC over the value Transcript-Hash(Handshake Context,\n Certificate, CertificateVerify) using a MAC key derived from the\n base key.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 65]\n\f\nInternet-Draft TLS March 2018\n\n\n The following table defines the Handshake Context and MAC Base Key\n for each scenario:\n\n +-----------+----------------------------+--------------------------+\n | Mode | Handshake Context | Base Key |\n +-----------+----------------------------+--------------------------+\n | Server | ClientHello ... later of E | server_handshake_traffic |\n | | ncryptedExtensions/Certifi | _secret |\n | | cateRequest | |\n | | | |\n | Client | ClientHello ... later of | client_handshake_traffic |\n | | server | _secret |\n | | Finished/EndOfEarlyData | |\n | | | |\n | Post- | ClientHello ... client | client_application_traff |\n | Handshake | Finished + | ic_secret_N |\n | | CertificateRequest | |\n +-----------+----------------------------+--------------------------+\n\n4.4.1. The Transcript Hash\n\n Many of the cryptographic computations in TLS make use of a\n transcript hash. This value is computed by hashing the concatenation\n of each included handshake message, including the handshake message\n header carrying the handshake message type and length fields, but not\n including record layer headers. I.e.,\n\n Transcript-Hash(M1, M2, ... Mn) = Hash(M1 || M2 || ... || Mn)\n\n As an exception to this general rule, when the server responds to a\n ClientHello with a HelloRetryRequest, the value of ClientHello1 is\n replaced with a special synthetic handshake message of handshake type\n \"message_hash\" containing Hash(ClientHello1). I.e.,\n\n Transcript-Hash(ClientHello1, HelloRetryRequest, ... Mn) =\n Hash(message_hash || /* Handshake type */\n 00 00 Hash.length || /* Handshake message length (bytes) */\n Hash(ClientHello1) || /* Hash of ClientHello1 */\n HelloRetryRequest || ... || Mn)\n\n The reason for this construction is to allow the server to do a\n stateless HelloRetryRequest by storing just the hash of ClientHello1\n in the cookie, rather than requiring it to export the entire\n intermediate hash state (see Section 4.2.2).\n\n For concreteness, the transcript hash is always taken from the\n following sequence of handshake messages, starting at the first\n ClientHello and including only those messages that were sent:\n\n\n\nRescorla Expires September 21, 2018 [Page 66]\n\f\nInternet-Draft TLS March 2018\n\n\n ClientHello, HelloRetryRequest, ClientHello, ServerHello,\n EncryptedExtensions, server CertificateRequest, server Certificate,\n server CertificateVerify, server Finished, EndOfEarlyData, client\n Certificate, client CertificateVerify, client Finished.\n\n In general, implementations can implement the transcript by keeping a\n running transcript hash value based on the negotiated hash. Note,\n however, that subsequent post-handshake authentications do not\n include each other, just the messages through the end of the main\n handshake.\n\n4.4.2. Certificate\n\n This message conveys the endpoint's certificate chain to the peer.\n\n The server MUST send a Certificate message whenever the agreed-upon\n key exchange method uses certificates for authentication (this\n includes all key exchange methods defined in this document except\n PSK).\n\n The client MUST send a Certificate message if and only if the server\n has requested client authentication via a CertificateRequest message\n (Section 4.3.2). If the server requests client authentication but no\n suitable certificate is available, the client MUST send a Certificate\n message containing no certificates (i.e., with the \"certificate_list\"\n field having length 0). A Finished message MUST be sent regardless\n of whether the Certificate message is empty.\n\n Structure of this message:\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 67]\n\f\nInternet-Draft TLS March 2018\n\n\n /* Managed by IANA */\n enum {\n X509(0),\n RawPublicKey(2),\n (255)\n } CertificateType;\n\n struct {\n select (certificate_type) {\n case RawPublicKey:\n /* From RFC 7250 ASN.1_subjectPublicKeyInfo */\n opaque ASN1_subjectPublicKeyInfo<1..2^24-1>;\n\n case X509:\n opaque cert_data<1..2^24-1>;\n };\n Extension extensions<0..2^16-1>;\n } CertificateEntry;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n CertificateEntry certificate_list<0..2^24-1>;\n } Certificate;\n\n certificate_request_context If this message is in response to a\n CertificateRequest, the value of certificate_request_context in\n that message. Otherwise (in the case of server authentication),\n this field SHALL be zero length.\n\n certificate_list This is a sequence (chain) of CertificateEntry\n structures, each containing a single certificate and set of\n extensions.\n\n extensions: A set of extension values for the CertificateEntry. The\n \"Extension\" format is defined in Section 4.2. Valid extensions\n for server certificates at present include OCSP Status extension\n ([RFC6066]) and SignedCertificateTimestamps ([RFC6962]); future\n extensions may be defined for this message as well. Extensions in\n the Certificate message from the server MUST correspond to ones\n from the ClientHello message. Extensions in the Certificate from\n the client MUST correspond with extensions in the\n CertificateRequest message from the server. If an extension\n applies to the entire chain, it SHOULD be included in the first\n CertificateEntry.\n\n If the corresponding certificate type extension\n (\"server_certificate_type\" or \"client_certificate_type\") was not\n negotiated in Encrypted Extensions, or the X.509 certificate type was\n\n\n\nRescorla Expires September 21, 2018 [Page 68]\n\f\nInternet-Draft TLS March 2018\n\n\n negotiated, then each CertificateEntry contains a DER-encoded X.509\n certificate. The sender's certificate MUST come in the first\n CertificateEntry in the list. Each following certificate SHOULD\n directly certify the one immediately preceding it. Because\n certificate validation requires that trust anchors be distributed\n independently, a certificate that specifies a trust anchor MAY be\n omitted from the chain, provided that supported peers are known to\n possess any omitted certificates.\n\n Note: Prior to TLS 1.3, \"certificate_list\" ordering required each\n certificate to certify the one immediately preceding it; however,\n some implementations allowed some flexibility. Servers sometimes\n send both a current and deprecated intermediate for transitional\n purposes, and others are simply configured incorrectly, but these\n cases can nonetheless be validated properly. For maximum\n compatibility, all implementations SHOULD be prepared to handle\n potentially extraneous certificates and arbitrary orderings from any\n TLS version, with the exception of the end-entity certificate which\n MUST be first.\n\n If the RawPublicKey certificate type was negotiated, then the\n certificate_list MUST contain no more than one CertificateEntry,\n which contains an ASN1_subjectPublicKeyInfo value as defined in\n [RFC7250], Section 3.\n\n The OpenPGP certificate type [RFC6091] MUST NOT be used with TLS 1.3.\n\n The server's certificate_list MUST always be non-empty. A client\n will send an empty certificate_list if it does not have an\n appropriate certificate to send in response to the server's\n authentication request.\n\n4.4.2.1. OCSP Status and SCT Extensions\n\n [RFC6066] and [RFC6961] provide extensions to negotiate the server\n sending OCSP responses to the client. In TLS 1.2 and below, the\n server replies with an empty extension to indicate negotiation of\n this extension and the OCSP information is carried in a\n CertificateStatus message. In TLS 1.3, the server's OCSP information\n is carried in an extension in the CertificateEntry containing the\n associated certificate. Specifically: The body of the\n \"status_request\" extension from the server MUST be a\n CertificateStatus structure as defined in [RFC6066], which is\n interpreted as defined in [RFC6960].\n\n Note: status_request_v2 extension ([RFC6961]) is deprecated. TLS 1.3\n servers MUST NOT act upon its presence or information in it when\n processing Client Hello, in particular they MUST NOT send the\n\n\n\nRescorla Expires September 21, 2018 [Page 69]\n\f\nInternet-Draft TLS March 2018\n\n\n status_request_v2 extension in the Encrypted Extensions, Certificate\n Request or the Certificate messages. TLS 1.3 servers MUST be able to\n process Client Hello messages that include it, as it MAY be sent by\n clients that wish to use it in earlier protocol versions.\n\n A server MAY request that a client present an OCSP response with its\n certificate by sending an empty \"status_request\" extension in its\n CertificateRequest message. If the client opts to send an OCSP\n response, the body of its \"status_request\" extension MUST be a\n CertificateStatus structure as defined in [RFC6066].\n\n Similarly, [RFC6962] provides a mechanism for a server to send a\n Signed Certificate Timestamp (SCT) as an extension in the ServerHello\n in TLS 1.2 and below. In TLS 1.3, the server's SCT information is\n carried in an extension in CertificateEntry.\n\n4.4.2.2. Server Certificate Selection\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3 [RFC5280], unless explicitly\n negotiated otherwise (e.g., [RFC7250]).\n\n - The server's end-entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected authentication\n algorithm from the client's \"signature_algorithms\" extension\n (currently RSA, ECDSA, or EdDSA).\n\n - The certificate MUST allow the key to be used for signing (i.e.,\n the digitalSignature bit MUST be set if the Key Usage extension is\n present) with a signature scheme indicated in the client's\n \"signature_algorithms\"/\"signature_algorithms_cert\" extensions (see\n Section 4.2.3).\n\n - The \"server_name\" [RFC6066] and \"certificate_authorities\"\n extensions are used to guide certificate selection. As servers\n MAY require the presence of the \"server_name\" extension, clients\n SHOULD send this extension, when applicable.\n\n All certificates provided by the server MUST be signed by a signature\n algorithm advertised by the client, if it is able to provide such a\n chain (see Section 4.2.3). Certificates that are self-signed or\n certificates that are expected to be trust anchors are not validated\n as part of the chain and therefore MAY be signed with any algorithm.\n\n If the server cannot produce a certificate chain that is signed only\n via the indicated supported algorithms, then it SHOULD continue the\n handshake by sending the client a certificate chain of its choice\n\n\n\nRescorla Expires September 21, 2018 [Page 70]\n\f\nInternet-Draft TLS March 2018\n\n\n that may include algorithms that are not known to be supported by the\n client. This fallback chain SHOULD NOT use the deprecated SHA-1 hash\n algorithm in general, but MAY do so if the client's advertisement\n permits it, and MUST NOT do so otherwise.\n\n If the client cannot construct an acceptable chain using the provided\n certificates and decides to abort the handshake, then it MUST abort\n the handshake with an appropriate certificate-related alert (by\n default, \"unsupported_certificate\"; see Section 6.2 for more).\n\n If the server has multiple certificates, it chooses one of them based\n on the above-mentioned criteria (in addition to other criteria, such\n as transport layer endpoint, local configuration and preferences).\n\n4.4.2.3. Client Certificate Selection\n\n The following rules apply to certificates sent by the client:\n\n - The certificate type MUST be X.509v3 [RFC5280], unless explicitly\n negotiated otherwise (e.g., [RFC7250]).\n\n - If the \"certificate_authorities\" extension in the\n CertificateRequest message was present, at least one of the\n certificates in the certificate chain SHOULD be issued by one of\n the listed CAs.\n\n - The certificates MUST be signed using an acceptable signature\n algorithm, as described in Section 4.3.2. Note that this relaxes\n the constraints on certificate-signing algorithms found in prior\n versions of TLS.\n\n - If the CertificateRequest message contained a non-empty\n \"oid_filters\" extension, the end-entity certificate MUST match the\n extension OIDs that are recognized by the client, as described in\n Section 4.2.5.\n\n Note that, as with the server certificate, there are certificates\n that use algorithm combinations that cannot be currently used with\n TLS.\n\n4.4.2.4. Receiving a Certificate Message\n\n In general, detailed certificate validation procedures are out of\n scope for TLS (see [RFC5280]). This section provides TLS-specific\n requirements.\n\n If the server supplies an empty Certificate message, the client MUST\n abort the handshake with a \"decode_error\" alert.\n\n\n\nRescorla Expires September 21, 2018 [Page 71]\n\f\nInternet-Draft TLS March 2018\n\n\n If the client does not send any certificates (i.e., it sends an empty\n Certificate message), the server MAY at its discretion either\n continue the handshake without client authentication, or abort the\n handshake with a \"certificate_required\" alert. Also, if some aspect\n of the certificate chain was unacceptable (e.g., it was not signed by\n a known, trusted CA), the server MAY at its discretion either\n continue the handshake (considering the client unauthenticated) or\n abort the handshake.\n\n Any endpoint receiving any certificate which it would need to\n validate using any signature algorithm using an MD5 hash MUST abort\n the handshake with a \"bad_certificate\" alert. SHA-1 is deprecated\n and it is RECOMMENDED that any endpoint receiving any certificate\n which it would need to validate using any signature algorithm using a\n SHA-1 hash abort the handshake with a \"bad_certificate\" alert. For\n clarity, this means that endpoints MAY accept these algorithms for\n certificates that are self-signed or are trust anchors.\n\n All endpoints are RECOMMENDED to transition to SHA-256 or better as\n soon as possible to maintain interoperability with implementations\n currently in the process of phasing out SHA-1 support.\n\n Note that a certificate containing a key for one signature algorithm\n MAY be signed using a different signature algorithm (for instance, an\n RSA key signed with an ECDSA key).\n\n4.4.3. Certificate Verify\n\n This message is used to provide explicit proof that an endpoint\n possesses the private key corresponding to its certificate. The\n CertificateVerify message also provides integrity for the handshake\n up to this point. Servers MUST send this message when authenticating\n via a certificate. Clients MUST send this message whenever\n authenticating via a certificate (i.e., when the Certificate message\n is non-empty). When sent, this message MUST appear immediately after\n the Certificate message and immediately prior to the Finished\n message.\n\n Structure of this message:\n\n struct {\n SignatureScheme algorithm;\n opaque signature<0..2^16-1>;\n } CertificateVerify;\n\n The algorithm field specifies the signature algorithm used (see\n Section 4.2.3 for the definition of this field). The signature is a\n digital signature using that algorithm. The content that is covered\n\n\n\nRescorla Expires September 21, 2018 [Page 72]\n\f\nInternet-Draft TLS March 2018\n\n\n under the signature is the hash output as described in Section 4.4.1,\n namely:\n\n Transcript-Hash(Handshake Context, Certificate)\n\n The digital signature is then computed over the concatenation of:\n\n - A string that consists of octet 32 (0x20) repeated 64 times\n\n - The context string\n\n - A single 0 byte which serves as the separator\n\n - The content to be signed\n\n This structure is intended to prevent an attack on previous versions\n of TLS in which the ServerKeyExchange format meant that attackers\n could obtain a signature of a message with a chosen 32-byte prefix\n (ClientHello.random). The initial 64-byte pad clears that prefix\n along with the server-controlled ServerHello.random.\n\n The context string for a server signature is: \"TLS 1.3, server\n CertificateVerify\" The context string for a client signature is: \"TLS\n 1.3, client CertificateVerify\" It is used to provide separation\n between signatures made in different contexts, helping against\n potential cross-protocol attacks.\n\n For example, if the transcript hash was 32 bytes of 01 (this length\n would make sense for SHA-256), the content covered by the digital\n signature for a server CertificateVerify would be:\n\n 2020202020202020202020202020202020202020202020202020202020202020\n 2020202020202020202020202020202020202020202020202020202020202020\n 544c5320312e332c207365727665722043657274696669636174655665726966\n 79\n 00\n 0101010101010101010101010101010101010101010101010101010101010101\n\n On the sender side the process for computing the signature field of\n the CertificateVerify message takes as input:\n\n - The content covered by the digital signature\n\n - The private signing key corresponding to the certificate sent in\n the previous message\n\n If the CertificateVerify message is sent by a server, the signature\n algorithm MUST be one offered in the client's \"signature_algorithms\"\n\n\n\nRescorla Expires September 21, 2018 [Page 73]\n\f\nInternet-Draft TLS March 2018\n\n\n extension unless no valid certificate chain can be produced without\n unsupported algorithms (see Section 4.2.3).\n\n If sent by a client, the signature algorithm used in the signature\n MUST be one of those present in the supported_signature_algorithms\n field of the \"signature_algorithms\" extension in the\n CertificateRequest message.\n\n In addition, the signature algorithm MUST be compatible with the key\n in the sender's end-entity certificate. RSA signatures MUST use an\n RSASSA-PSS algorithm, regardless of whether RSASSA-PKCS1-v1_5\n algorithms appear in \"signature_algorithms\". The SHA-1 algorithm\n MUST NOT be used in any signatures of CertificateVerify messages.\n All SHA-1 signature algorithms in this specification are defined\n solely for use in legacy certificates and are not valid for\n CertificateVerify signatures.\n\n The receiver of a CertificateVerify message MUST verify the signature\n field. The verification process takes as input:\n\n - The content covered by the digital signature\n\n - The public key contained in the end-entity certificate found in\n the associated Certificate message.\n\n - The digital signature received in the signature field of the\n CertificateVerify message\n\n If the verification fails, the receiver MUST terminate the handshake\n with a \"decrypt_error\" alert.\n\n4.4.4. Finished\n\n The Finished message is the final message in the authentication\n block. It is essential for providing authentication of the handshake\n and of the computed keys.\n\n Recipients of Finished messages MUST verify that the contents are\n correct and if incorrect MUST terminate the connection with a\n \"decrypt_error\" alert.\n\n Once a side has sent its Finished message and received and validated\n the Finished message from its peer, it may begin to send and receive\n application data over the connection. There are two settings in\n which it is permitted to send data prior to receiving the peer's\n Finished:\n\n 1. Clients sending 0-RTT data as described in Section 4.2.10.\n\n\n\nRescorla Expires September 21, 2018 [Page 74]\n\f\nInternet-Draft TLS March 2018\n\n\n 2. Servers MAY send data after sending their first flight, but\n because the handshake is not yet complete, they have no assurance\n of either the peer's identity or of its liveness (i.e., the\n ClientHello might have been replayed).\n\n The key used to compute the Finished message is computed from the\n Base key defined in Section 4.4 using HKDF (see Section 7.1).\n Specifically:\n\n finished_key =\n HKDF-Expand-Label(BaseKey, \"finished\", \"\", Hash.length)\n\n Structure of this message:\n\n struct {\n opaque verify_data[Hash.length];\n } Finished;\n\n The verify_data value is computed as follows:\n\n verify_data =\n HMAC(finished_key,\n Transcript-Hash(Handshake Context,\n Certificate*, CertificateVerify*))\n\n * Only included if present.\n\n HMAC [RFC2104] uses the Hash algorithm for the handshake. As noted\n above, the HMAC input can generally be implemented by a running hash,\n i.e., just the handshake hash at this point.\n\n In previous versions of TLS, the verify_data was always 12 octets\n long. In TLS 1.3, it is the size of the HMAC output for the Hash\n used for the handshake.\n\n Note: Alerts and any other record types are not handshake messages\n and are not included in the hash computations.\n\n Any records following a Finished message MUST be encrypted under the\n appropriate application traffic key as described in Section 7.2. In\n particular, this includes any alerts sent by the server in response\n to client Certificate and CertificateVerify messages.\n\n4.5. End of Early Data\n\n struct {} EndOfEarlyData;\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 75]\n\f\nInternet-Draft TLS March 2018\n\n\n If the server sent an \"early_data\" extension, the client MUST send an\n EndOfEarlyData message after receiving the server Finished. If the\n server does not send an \"early_data\" extension, then the client MUST\n NOT send an EndOfEarlyData message. This message indicates that all\n 0-RTT application_data messages, if any, have been transmitted and\n that the following records are protected under handshake traffic\n keys. Servers MUST NOT send this message and clients receiving it\n MUST terminate the connection with an \"unexpected_message\" alert.\n This message is encrypted under keys derived from the\n client_early_traffic_secret.\n\n4.6. Post-Handshake Messages\n\n TLS also allows other messages to be sent after the main handshake.\n These messages use a handshake content type and are encrypted under\n the appropriate application traffic key.\n\n4.6.1. New Session Ticket Message\n\n At any time after the server has received the client Finished\n message, it MAY send a NewSessionTicket message. This message\n creates a unique association between the ticket value and a secret\n PSK derived from the resumption master secret (see Section 7.\n\n The client MAY use this PSK for future handshakes by including the\n ticket value in the \"pre_shared_key\" extension in its ClientHello\n (Section 4.2.11). Servers MAY send multiple tickets on a single\n connection, either immediately after each other or after specific\n events (see Appendix C.4). For instance, the server might send a new\n ticket after post-handshake authentication in order to encapsulate\n the additional client authentication state. Multiple tickets are\n useful for clients for a variety of purposes, including:\n\n - Opening multiple parallel HTTP connections.\n\n - Performing connection racing across interfaces and address\n families via, e.g., Happy Eyeballs [RFC8305] or related\n techniques.\n\n Any ticket MUST only be resumed with a cipher suite that has the same\n KDF hash algorithm as that used to establish the original connection.\n\n Clients MUST only resume if the new SNI value is valid for the server\n certificate presented in the original session, and SHOULD only resume\n if the SNI value matches the one used in the original session. The\n latter is a performance optimization: normally, there is no reason to\n expect that different servers covered by a single certificate would\n be able to accept each other's tickets, hence attempting resumption\n\n\n\nRescorla Expires September 21, 2018 [Page 76]\n\f\nInternet-Draft TLS March 2018\n\n\n in that case would waste a single-use ticket. If such an indication\n is provided (externally or by any other means), clients MAY resume\n with a different SNI value.\n\n On resumption, if reporting an SNI value to the calling application,\n implementations MUST use the value sent in the resumption ClientHello\n rather than the value sent in the previous session. Note that if a\n server implementation declines all PSK identities with different SNI\n values, these two values are always the same.\n\n Note: Although the resumption master secret depends on the client's\n second flight, servers which do not request client authentication MAY\n compute the remainder of the transcript independently and then send a\n NewSessionTicket immediately upon sending its Finished rather than\n waiting for the client Finished. This might be appropriate in cases\n where the client is expected to open multiple TLS connections in\n parallel and would benefit from the reduced overhead of a resumption\n handshake, for example.\n\n struct {\n uint32 ticket_lifetime;\n uint32 ticket_age_add;\n opaque ticket_nonce<0..255>;\n opaque ticket<1..2^16-1>;\n Extension extensions<0..2^16-2>;\n } NewSessionTicket;\n\n ticket_lifetime Indicates the lifetime in seconds as a 32-bit\n unsigned integer in network byte order from the time of ticket\n issuance. Servers MUST NOT use any value greater than 604800\n seconds (7 days). The value of zero indicates that the ticket\n should be discarded immediately. Clients MUST NOT cache tickets\n for longer than 7 days, regardless of the ticket_lifetime, and MAY\n delete tickets earlier based on local policy. A server MAY treat\n a ticket as valid for a shorter period of time than what is stated\n in the ticket_lifetime.\n\n ticket_age_add A securely generated, random 32-bit value that is\n used to obscure the age of the ticket that the client includes in\n the \"pre_shared_key\" extension. The client-side ticket age is\n added to this value modulo 2^32 to obtain the value that is\n transmitted by the client. The server MUST generate a fresh value\n for each ticket it sends.\n\n ticket_nonce A per-ticket value that is unique across all tickets\n issued on this connection.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 77]\n\f\nInternet-Draft TLS March 2018\n\n\n ticket The value of the ticket to be used as the PSK identity. The\n ticket itself is an opaque label. It MAY either be a database\n lookup key or a self-encrypted and self-authenticated value.\n Section 4 of [RFC5077] describes a recommended ticket construction\n mechanism.\n\n extensions A set of extension values for the ticket. The\n \"Extension\" format is defined in Section 4.2. Clients MUST ignore\n unrecognized extensions.\n\n The sole extension currently defined for NewSessionTicket is\n \"early_data\", indicating that the ticket may be used to send 0-RTT\n data (Section 4.2.10)). It contains the following value:\n\n max_early_data_size The maximum amount of 0-RTT data that the client\n is allowed to send when using this ticket, in bytes. Only\n Application Data payload (i.e., plaintext but not padding or the\n inner content type byte) is counted. A server receiving more than\n max_early_data_size bytes of 0-RTT data SHOULD terminate the\n connection with an \"unexpected_message\" alert. Note that servers\n that reject early data due to lack of cryptographic material will\n be unable to differentiate padding from content, so clients SHOULD\n NOT depend on being able to send large quantities of padding in\n early data records.\n\n The PSK associated with the ticket is computed as:\n\n HKDF-Expand-Label(resumption_master_secret,\n \"resumption\", ticket_nonce, Hash.length)\n\n Because the ticket_nonce value is distinct for each NewSessionTicket\n message, a different PSK will be derived for each ticket.\n\n Note that in principle it is possible to continue issuing new tickets\n which indefinitely extend the lifetime of the keying material\n originally derived from an initial non-PSK handshake (which was most\n likely tied to the peer's certificate). It is RECOMMENDED that\n implementations place limits on the total lifetime of such keying\n material; these limits should take into account the lifetime of the\n peer's certificate, the likelihood of intervening revocation, and the\n time since the peer's online CertificateVerify signature.\n\n4.6.2. Post-Handshake Authentication\n\n When the client has sent the \"post_handshake_auth\" extension (see\n Section 4.2.6), a server MAY request client authentication at any\n time after the handshake has completed by sending a\n CertificateRequest message. The client MUST respond with the\n\n\n\nRescorla Expires September 21, 2018 [Page 78]\n\f\nInternet-Draft TLS March 2018\n\n\n appropriate Authentication messages (see Section 4.4). If the client\n chooses to authenticate, it MUST send Certificate, CertificateVerify,\n and Finished. If it declines, it MUST send a Certificate message\n containing no certificates followed by Finished. All of the client's\n messages for a given response MUST appear consecutively on the wire\n with no intervening messages of other types.\n\n A client that receives a CertificateRequest message without having\n sent the \"post_handshake_auth\" extension MUST send an\n \"unexpected_message\" fatal alert.\n\n Note: Because client authentication could involve prompting the user,\n servers MUST be prepared for some delay, including receiving an\n arbitrary number of other messages between sending the\n CertificateRequest and receiving a response. In addition, clients\n which receive multiple CertificateRequests in close succession MAY\n respond to them in a different order than they were received (the\n certificate_request_context value allows the server to disambiguate\n the responses).\n\n4.6.3. Key and IV Update\n\n enum {\n update_not_requested(0), update_requested(1), (255)\n } KeyUpdateRequest;\n\n struct {\n KeyUpdateRequest request_update;\n } KeyUpdate;\n\n request_update Indicates whether the recipient of the KeyUpdate\n should respond with its own KeyUpdate. If an implementation\n receives any other value, it MUST terminate the connection with an\n \"illegal_parameter\" alert.\n\n The KeyUpdate handshake message is used to indicate that the sender\n is updating its sending cryptographic keys. This message can be sent\n by either peer after it has sent a Finished message. Implementations\n that receive a KeyUpdate message prior to receiving a Finished\n message MUST terminate the connection with an \"unexpected_message\"\n alert. After sending a KeyUpdate message, the sender SHALL send all\n its traffic using the next generation of keys, computed as described\n in Section 7.2. Upon receiving a KeyUpdate, the receiver MUST update\n its receiving keys.\n\n If the request_update field is set to \"update_requested\" then the\n receiver MUST send a KeyUpdate of its own with request_update set to\n \"update_not_requested\" prior to sending its next application data\n\n\n\nRescorla Expires September 21, 2018 [Page 79]\n\f\nInternet-Draft TLS March 2018\n\n\n record. This mechanism allows either side to force an update to the\n entire connection, but causes an implementation which receives\n multiple KeyUpdates while it is silent to respond with a single\n update. Note that implementations may receive an arbitrary number of\n messages between sending a KeyUpdate with request_update set to\n update_requested and receiving the peer's KeyUpdate, because those\n messages may already be in flight. However, because send and receive\n keys are derived from independent traffic secrets, retaining the\n receive traffic secret does not threaten the forward secrecy of data\n sent before the sender changed keys.\n\n If implementations independently send their own KeyUpdates with\n request_update set to \"update_requested\", and they cross in flight,\n then each side will also send a response, with the result that each\n side increments by two generations.\n\n Both sender and receiver MUST encrypt their KeyUpdate messages with\n the old keys. Additionally, both sides MUST enforce that a KeyUpdate\n with the old key is received before accepting any messages encrypted\n with the new key. Failure to do so may allow message truncation\n attacks.\n\n5. Record Protocol\n\n The TLS record protocol takes messages to be transmitted, fragments\n the data into manageable blocks, protects the records, and transmits\n the result. Received data is verified, decrypted, reassembled, and\n then delivered to higher-level clients.\n\n TLS records are typed, which allows multiple higher-level protocols\n to be multiplexed over the same record layer. This document\n specifies four content types: handshake, application data, alert, and\n change_cipher_spec. The change_cipher_spec record is used only for\n compatibility purposes (see Appendix D.4).\n\n An implementation may receive an unencrypted record of type\n change_cipher_spec consisting of the single byte value 0x01 at any\n time after the first ClientHello message has been sent or received\n and before the peer's Finished message has been received and MUST\n simply drop it without further processing. Note that this record may\n appear at a point at the handshake where the implementation is\n expecting protected records and so it is necessary to detect this\n condition prior to attempting to deprotect the record. An\n implementation which receives any other change_cipher_spec value or\n which receives a protected change_cipher_spec record MUST abort the\n handshake with an \"unexpected_message\" alert. A change_cipher_spec\n record received before the first ClientHello message or after the\n peer's Finished message MUST be treated as an unexpected record type\n\n\n\nRescorla Expires September 21, 2018 [Page 80]\n\f\nInternet-Draft TLS March 2018\n\n\n (though stateless servers may not be able to distinguish these cases\n from allowed cases).\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST terminate\n the connection with an \"unexpected_message\" alert. New record\n content type values are assigned by IANA in the TLS Content Type\n Registry as described in Section 11.\n\n5.1. Record Layer\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Message\n boundaries are handled differently depending on the underlying\n ContentType. Any future content types MUST specify appropriate\n rules. Note that these rules are stricter than what was enforced in\n TLS 1.2.\n\n Handshake messages MAY be coalesced into a single TLSPlaintext record\n or fragmented across several records, provided that:\n\n - Handshake messages MUST NOT be interleaved with other record\n types. That is, if a handshake message is split over two or more\n records, there MUST NOT be any other records between them.\n\n - Handshake messages MUST NOT span key changes. Implementations\n MUST verify that all messages immediately preceding a key change\n align with a record boundary; if not, then they MUST terminate the\n connection with an \"unexpected_message\" alert. Because the\n ClientHello, EndOfEarlyData, ServerHello, Finished, and KeyUpdate\n messages can immediately precede a key change, implementations\n MUST send these messages in alignment with a record boundary.\n\n Implementations MUST NOT send zero-length fragments of Handshake\n types, even if those fragments contain padding.\n\n Alert messages (Section 6) MUST NOT be fragmented across records and\n multiple Alert messages MUST NOT be coalesced into a single\n TLSPlaintext record. In other words, a record with an Alert type\n MUST contain exactly one message.\n\n Application Data messages contain data that is opaque to TLS.\n Application Data messages are always protected. Zero-length\n fragments of Application Data MAY be sent as they are potentially\n useful as a traffic analysis countermeasure. Application Data\n fragments MAY be split across multiple records or coalesced into a\n single record.\n\n\n\nRescorla Expires September 21, 2018 [Page 81]\n\f\nInternet-Draft TLS March 2018\n\n\n enum {\n invalid(0),\n change_cipher_spec(20),\n alert(21),\n handshake(22),\n application_data(23),\n (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion legacy_record_version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type The higher-level protocol used to process the enclosed\n fragment.\n\n legacy_record_version This value MUST be set to 0x0303 for all\n records generated by a TLS 1.3 implementation other than an\n initial ClientHello (i.e., one not generated after a\n HelloRetryRequest), where it MAY also be 0x0301 for compatibility\n purposes. This field is deprecated and MUST be ignored for all\n purposes. Previous versions of TLS would use other values in this\n field under some circumstances.\n\n length The length (in bytes) of the following TLSPlaintext.fragment.\n The length MUST NOT exceed 2^14 bytes. An endpoint that receives\n a record that exceeds this length MUST terminate the connection\n with a \"record_overflow\" alert.\n\n fragment The data being transmitted. This value is transparent and\n is treated as an independent block to be dealt with by the higher-\n level protocol specified by the type field.\n\n This document describes TLS 1.3, which uses the version 0x0304. This\n version value is historical, deriving from the use of 0x0301 for TLS\n 1.0 and 0x0300 for SSL 3.0. In order to maximize backwards\n compatibility, records containing an initial ClientHello SHOULD have\n version 0x0301 and a record containing a second ClientHello or a\n ServerHello MUST have version 0x0303, reflecting TLS 1.0 and TLS 1.2\n respectively. When negotiating prior versions of TLS, endpoints\n follow the procedure and requirements in Appendix D.\n\n When record protection has not yet been engaged, TLSPlaintext\n structures are written directly onto the wire. Once record\n protection has started, TLSPlaintext records are protected and sent\n\n\n\nRescorla Expires September 21, 2018 [Page 82]\n\f\nInternet-Draft TLS March 2018\n\n\n as described in the following section. Note that application data\n records MUST NOT be written to the wire unprotected (see Section 2\n for details).\n\n5.2. Record Payload Protection\n\n The record protection functions translate a TLSPlaintext structure\n into a TLSCiphertext. The deprotection functions reverse the\n process. In TLS 1.3, as opposed to previous versions of TLS, all\n ciphers are modeled as \"Authenticated Encryption with Additional\n Data\" (AEAD) [RFC5116]. AEAD functions provide an unified encryption\n and authentication operation which turns plaintext into authenticated\n ciphertext and back again. Each encrypted record consists of a\n plaintext header followed by an encrypted body, which itself contains\n a type and optional padding.\n\n struct {\n opaque content[TLSPlaintext.length];\n ContentType type;\n uint8 zeros[length_of_padding];\n } TLSInnerPlaintext;\n\n struct {\n ContentType opaque_type = application_data; /* 23 */\n ProtocolVersion legacy_record_version = 0x0303; /* TLS v1.2 */\n uint16 length;\n opaque encrypted_record[TLSCiphertext.length];\n } TLSCiphertext;\n\n content The TLSPLaintext.fragment value, containing the byte\n encoding of a handshake or an alert message, or the raw bytes of\n the application's data to send.\n\n type The TLSPlaintext.type value containing the content type of the\n record.\n\n zeros An arbitrary-length run of zero-valued bytes may appear in the\n cleartext after the type field. This provides an opportunity for\n senders to pad any TLS record by a chosen amount as long as the\n total stays within record size limits. See Section 5.4 for more\n details.\n\n opaque_type The outer opaque_type field of a TLSCiphertext record is\n always set to the value 23 (application_data) for outward\n compatibility with middleboxes accustomed to parsing previous\n versions of TLS. The actual content type of the record is found\n in TLSInnerPlaintext.type after decryption.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 83]\n\f\nInternet-Draft TLS March 2018\n\n\n legacy_record_version The legacy_record_version field is always\n 0x0303. TLS 1.3 TLSCiphertexts are not generated until after TLS\n 1.3 has been negotiated, so there are no historical compatibility\n concerns where other values might be received. Note that the\n handshake protocol including the ClientHello and ServerHello\n messages authenticates the protocol version, so this value is\n redundant.\n\n length The length (in bytes) of the following\n TLSCiphertext.encrypted_record, which is the sum of the lengths of\n the content and the padding, plus one for the inner content type,\n plus any expansion added by the AEAD algorithm. The length MUST\n NOT exceed 2^14 + 256 bytes. An endpoint that receives a record\n that exceeds this length MUST terminate the connection with a\n \"record_overflow\" alert.\n\n encrypted_record The AEAD-encrypted form of the serialized\n TLSInnerPlaintext structure.\n\n AEAD algorithms take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [RFC5116]. The key is either the\n client_write_key or the server_write_key, the nonce is derived from\n the sequence number and the client_write_iv or server_write_iv (see\n Section 5.3), and the additional data input is the record header.\n I.e.,\n\n additional_data = TLSCiphertext.opaque_type ||\n TLSCiphertext.legacy_record_version ||\n TLSCiphertext.length\n\n The plaintext input to the AEAD algorithm is the encoded\n TLSInnerPlaintext structure. Derivation of traffic keys is defined\n in Section 7.3.\n\n The AEAD output consists of the ciphertext output from the AEAD\n encryption operation. The length of the plaintext is greater than\n the corresponding TLSPlaintext.length due to the inclusion of\n TLSInnerPlaintext.type and any padding supplied by the sender. The\n length of the AEAD output will generally be larger than the\n plaintext, but by an amount that varies with the AEAD algorithm.\n Since the ciphers might incorporate padding, the amount of overhead\n could vary with different lengths of plaintext. Symbolically,\n\n AEADEncrypted =\n AEAD-Encrypt(write_key, nonce, additional_data, plaintext)\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 84]\n\f\nInternet-Draft TLS March 2018\n\n\n Then the encrypted_record field of TLSCiphertext is set to\n AEADEncrypted.\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, additional data, and the AEADEncrypted value. The output is\n either the plaintext or an error indicating that the decryption\n failed. There is no separate integrity check. That is:\n\n plaintext of encrypted_record =\n AEAD-Decrypt(peer_write_key, nonce, additional_data, AEADEncrypted)\n\n If the decryption fails, the receiver MUST terminate the connection\n with a \"bad_record_mac\" alert.\n\n An AEAD algorithm used in TLS 1.3 MUST NOT produce an expansion\n greater than 255 octets. An endpoint that receives a record from its\n peer with TLSCiphertext.length larger than 2^14 + 256 octets MUST\n terminate the connection with a \"record_overflow\" alert. This limit\n is derived from the maximum TLSInnerPlaintext length of 2^14 octets +\n 1 octet for ContentType + the maximum AEAD expansion of 255 octets.\n\n5.3. Per-Record Nonce\n\n A 64-bit sequence number is maintained separately for reading and\n writing records. The appropriate sequence number is incremented by\n one after reading or writing each record. Each sequence number is\n set to zero at the beginning of a connection and whenever the key is\n changed; the first record transmitted under a particular traffic key\n MUST use sequence number 0.\n\n Because the size of sequence numbers is 64-bit, they should not wrap.\n If a TLS implementation would need to wrap a sequence number, it MUST\n either re-key (Section 4.6.3) or terminate the connection.\n\n Each AEAD algorithm will specify a range of possible lengths for the\n per-record nonce, from N_MIN bytes to N_MAX bytes of input\n ([RFC5116]). The length of the TLS per-record nonce (iv_length) is\n set to the larger of 8 bytes and N_MIN for the AEAD algorithm (see\n [RFC5116] Section 4). An AEAD algorithm where N_MAX is less than 8\n bytes MUST NOT be used with TLS. The per-record nonce for the AEAD\n construction is formed as follows:\n\n 1. The 64-bit record sequence number is encoded in network byte\n order and padded to the left with zeros to iv_length.\n\n 2. The padded sequence number is XORed with the static\n client_write_iv or server_write_iv, depending on the role.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 85]\n\f\nInternet-Draft TLS March 2018\n\n\n The resulting quantity (of length iv_length) is used as the per-\n record nonce.\n\n Note: This is a different construction from that in TLS 1.2, which\n specified a partially explicit nonce.\n\n5.4. Record Padding\n\n All encrypted TLS records can be padded to inflate the size of the\n TLSCiphertext. This allows the sender to hide the size of the\n traffic from an observer.\n\n When generating a TLSCiphertext record, implementations MAY choose to\n pad. An unpadded record is just a record with a padding length of\n zero. Padding is a string of zero-valued bytes appended to the\n ContentType field before encryption. Implementations MUST set the\n padding octets to all zeros before encrypting.\n\n Application Data records may contain a zero-length\n TLSInnerPlaintext.content if the sender desires. This permits\n generation of plausibly-sized cover traffic in contexts where the\n presence or absence of activity may be sensitive. Implementations\n MUST NOT send Handshake or Alert records that have a zero-length\n TLSInnerPlaintext.content; if such a message is received, the\n receiving implementation MUST terminate the connection with an\n \"unexpected_message\" alert.\n\n The padding sent is automatically verified by the record protection\n mechanism; upon successful decryption of a\n TLSCiphertext.encrypted_record, the receiving implementation scans\n the field from the end toward the beginning until it finds a non-zero\n octet. This non-zero octet is the content type of the message. This\n padding scheme was selected because it allows padding of any\n encrypted TLS record by an arbitrary size (from zero up to TLS record\n size limits) without introducing new content types. The design also\n enforces all-zero padding octets, which allows for quick detection of\n padding errors.\n\n Implementations MUST limit their scanning to the cleartext returned\n from the AEAD decryption. If a receiving implementation does not\n find a non-zero octet in the cleartext, it MUST terminate the\n connection with an \"unexpected_message\" alert.\n\n The presence of padding does not change the overall record size\n limitations - the full encoded TLSInnerPlaintext MUST NOT exceed 2^14\n + 1 octets. If the maximum fragment length is reduced, as for\n example by the max_fragment_length extension from [RFC6066], then the\n\n\n\n\nRescorla Expires September 21, 2018 [Page 86]\n\f\nInternet-Draft TLS March 2018\n\n\n reduced limit applies to the full plaintext, including the content\n type and padding.\n\n Selecting a padding policy that suggests when and how much to pad is\n a complex topic and is beyond the scope of this specification. If\n the application layer protocol on top of TLS has its own padding, it\n may be preferable to pad application_data TLS records within the\n application layer. Padding for encrypted handshake and alert TLS\n records must still be handled at the TLS layer, though. Later\n documents may define padding selection algorithms or define a padding\n policy request mechanism through TLS extensions or some other means.\n\n5.5. Limits on Key Usage\n\n There are cryptographic limits on the amount of plaintext which can\n be safely encrypted under a given set of keys. [AEAD-LIMITS]\n provides an analysis of these limits under the assumption that the\n underlying primitive (AES or ChaCha20) has no weaknesses.\n Implementations SHOULD do a key update as described in Section 4.6.3\n prior to reaching these limits.\n\n For AES-GCM, up to 2^24.5 full-size records (about 24 million) may be\n encrypted on a given connection while keeping a safety margin of\n approximately 2^-57 for Authenticated Encryption (AE) security. For\n ChaCha20/Poly1305, the record sequence number would wrap before the\n safety limit is reached.\n\n6. Alert Protocol\n\n One of the content types supported by the TLS record layer is the\n alert type. Like other messages, alert messages are encrypted as\n specified by the current connection state.\n\n Alert messages convey a description of the alert and a legacy field\n that conveyed the severity of the message in previous versions of\n TLS. Alerts are divided into two classes: closure alerts and error\n alerts. In TLS 1.3, the severity is implicit in the type of alert\n being sent, and the 'level' field can safely be ignored. The\n \"close_notify\" alert is used to indicate orderly closure of one\n direction of the connection. Upon receiving such an alert, the TLS\n implementation SHOULD indicate end-of-data to the application.\n\n Error alerts indicate abortive closure of the connection (see\n Section 6.2). Upon receiving an error alert, the TLS implementation\n SHOULD indicate an error to the application and MUST NOT allow any\n further data to be sent or received on the connection. Servers and\n clients MUST forget the secret values and keys established in failed\n\n\n\n\nRescorla Expires September 21, 2018 [Page 87]\n\f\nInternet-Draft TLS March 2018\n\n\n connections, with the exception of the PSKs associated with session\n tickets, which SHOULD be discarded if possible.\n\n All the alerts listed in Section 6.2 MUST be sent with\n AlertLevel=fatal and MUST be treated as error alerts regardless of\n the AlertLevel in the message. Unknown alert types MUST be treated\n as error alerts.\n\n Note: TLS defines two generic alerts (see Section 6) to use upon\n failure to parse a message. Peers which receive a message which\n cannot be parsed according to the syntax (e.g., have a length\n extending beyond the message boundary or contain an out-of-range\n length) MUST terminate the connection with a \"decode_error\" alert.\n Peers which receive a message which is syntactically correct but\n semantically invalid (e.g., a DHE share of p - 1, or an invalid enum)\n MUST terminate the connection with an \"illegal_parameter\" alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 88]\n\f\nInternet-Draft TLS March 2018\n\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n record_overflow(22),\n handshake_failure(40),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n inappropriate_fallback(86),\n user_canceled(90),\n missing_extension(109),\n unsupported_extension(110),\n unrecognized_name(112),\n bad_certificate_status_response(113),\n unknown_psk_identity(115),\n certificate_required(116),\n no_application_protocol(120),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n6.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack.\n\n close_notify This alert notifies the recipient that the sender will\n not send any more messages on this connection. Any data received\n after a closure alert has been received MUST be ignored.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 89]\n\f\nInternet-Draft TLS March 2018\n\n\n user_canceled This alert notifies the recipient that the sender is\n canceling the handshake for some reason unrelated to a protocol\n failure. If a user cancels an operation after the handshake is\n complete, just closing the connection by sending a \"close_notify\"\n is more appropriate. This alert SHOULD be followed by a\n \"close_notify\". This alert generally has AlertLevel=warning.\n\n Either party MAY initiate a close of its write side of the connection\n by sending a \"close_notify\" alert. Any data received after a closure\n alert has been received MUST be ignored. If a transport-level close\n is received prior to a \"close_notify\", the receiver cannot know that\n all the data that was sent has been received.\n\n Each party MUST send a \"close_notify\" alert before closing its write\n side of the connection, unless it has already sent some error alert.\n This does not have any effect on its read side of the connection.\n Note that this is a change from versions of TLS prior to TLS 1.3 in\n which implementations were required to react to a \"close_notify\" by\n discarding pending writes and sending an immediate \"close_notify\"\n alert of their own. That previous requirement could cause truncation\n in the read side. Both parties need not wait to receive a\n \"close_notify\" alert before closing their read side of the\n connection, though doing so would introduce the possibility of\n truncation.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation MUST receive a \"close_notify\" alert\n before indicating end-of-data to the application-layer. No part of\n this standard should be taken to dictate the manner in which a usage\n profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing the write side of a connection\n reliably delivers pending data before destroying the transport.\n\n6.2. Error Alerts\n\n Error handling in the TLS Handshake Protocol is very simple. When an\n error is detected, the detecting party sends a message to its peer.\n Upon transmission or receipt of a fatal alert message, both parties\n MUST immediately close the connection.\n\n Whenever an implementation encounters a fatal error condition, it\n SHOULD send an appropriate fatal alert and MUST close the connection\n without sending or receiving any additional data. In the rest of\n this specification, when the phrases \"terminate the connection\" and\n \"abort the handshake\" are used without a specific alert it means that\n\n\n\nRescorla Expires September 21, 2018 [Page 90]\n\f\nInternet-Draft TLS March 2018\n\n\n the implementation SHOULD send the alert indicated by the\n descriptions below. The phrases \"terminate the connection with a X\n alert\" and \"abort the handshake with a X alert\" mean that the\n implementation MUST send alert X if it sends any alert. All alerts\n defined in this section below, as well as all unknown alerts, are\n universally considered fatal as of TLS 1.3 (see Section 6). The\n implementation SHOULD provide a way to facilitate logging the sending\n and receiving of alerts.\n\n The following error alerts are defined:\n\n unexpected_message An inappropriate message (e.g., the wrong\n handshake message, premature application data, etc.) was received.\n This alert should never be observed in communication between\n proper implementations.\n\n bad_record_mac This alert is returned if a record is received which\n cannot be deprotected. Because AEAD algorithms combine decryption\n and verification, and also to avoid side channel attacks, this\n alert is used for all deprotection failures. This alert should\n never be observed in communication between proper implementations,\n except when messages were corrupted in the network.\n\n record_overflow A TLSCiphertext record was received that had a\n length more than 2^14 + 256 bytes, or a record decrypted to a\n TLSPlaintext record with more than 2^14 bytes (or some other\n negotiated limit). This alert should never be observed in\n communication between proper implementations, except when messages\n were corrupted in the network.\n\n handshake_failure Receipt of a \"handshake_failure\" alert message\n indicates that the sender was unable to negotiate an acceptable\n set of security parameters given the options available.\n\n bad_certificate A certificate was corrupt, contained signatures that\n did not verify correctly, etc.\n\n unsupported_certificate A certificate was of an unsupported type.\n\n certificate_revoked A certificate was revoked by its signer.\n\n certificate_expired A certificate has expired or is not currently\n valid.\n\n certificate_unknown Some other (unspecified) issue arose in\n processing the certificate, rendering it unacceptable.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 91]\n\f\nInternet-Draft TLS March 2018\n\n\n illegal_parameter A field in the handshake was incorrect or\n inconsistent with other fields. This alert is used for errors\n which conform to the formal protocol syntax but are otherwise\n incorrect.\n\n unknown_ca A valid certificate chain or partial chain was received,\n but the certificate was not accepted because the CA certificate\n could not be located or could not be matched with a known trust\n anchor.\n\n access_denied A valid certificate or PSK was received, but when\n access control was applied, the sender decided not to proceed with\n negotiation.\n\n decode_error A message could not be decoded because some field was\n out of the specified range or the length of the message was\n incorrect. This alert is used for errors where the message does\n not conform to the formal protocol syntax. This alert should\n never be observed in communication between proper implementations,\n except when messages were corrupted in the network.\n\n decrypt_error A handshake (not record-layer) cryptographic operation\n failed, including being unable to correctly verify a signature or\n validate a Finished message or a PSK binder.\n\n protocol_version The protocol version the peer has attempted to\n negotiate is recognized but not supported. (see Appendix D)\n\n insufficient_security Returned instead of \"handshake_failure\" when a\n negotiation has failed specifically because the server requires\n parameters more secure than those supported by the client.\n\n internal_error An internal error unrelated to the peer or the\n correctness of the protocol (such as a memory allocation failure)\n makes it impossible to continue.\n\n inappropriate_fallback Sent by a server in response to an invalid\n connection retry attempt from a client (see [RFC7507]).\n\n missing_extension Sent by endpoints that receive a handshake message\n not containing an extension that is mandatory to send for the\n offered TLS version or other negotiated parameters.\n\n unsupported_extension Sent by endpoints receiving any handshake\n message containing an extension known to be prohibited for\n inclusion in the given handshake message, or including any\n extensions in a ServerHello or Certificate not first offered in\n the corresponding ClientHello.\n\n\n\nRescorla Expires September 21, 2018 [Page 92]\n\f\nInternet-Draft TLS March 2018\n\n\n unrecognized_name Sent by servers when no server exists identified\n by the name provided by the client via the \"server_name\" extension\n (see [RFC6066]).\n\n bad_certificate_status_response Sent by clients when an invalid or\n unacceptable OCSP response is provided by the server via the\n \"status_request\" extension (see [RFC6066]).\n\n unknown_psk_identity Sent by servers when PSK key establishment is\n desired but no acceptable PSK identity is provided by the client.\n Sending this alert is OPTIONAL; servers MAY instead choose to send\n a \"decrypt_error\" alert to merely indicate an invalid PSK\n identity.\n\n certificate_required Sent by servers when a client certificate is\n desired but none was provided by the client.\n\n no_application_protocol Sent by servers when a client\n \"application_layer_protocol_negotiation\" extension advertises only\n protocols that the server does not support (see [RFC7301]).\n\n New Alert values are assigned by IANA as described in Section 11.\n\n7. Cryptographic Computations\n\n The TLS handshake establishes one or more input secrets which are\n combined to create the actual working keying material, as detailed\n below. The key derivation process incorporates both the input\n secrets and the handshake transcript. Note that because the\n handshake transcript includes the random values from the Hello\n messages, any given handshake will have different traffic secrets,\n even if the same input secrets are used, as is the case when the same\n PSK is used for multiple connections.\n\n7.1. Key Schedule\n\n The key derivation process makes use of the HKDF-Extract and HKDF-\n Expand functions as defined for HKDF [RFC5869], as well as the\n functions defined below:\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 93]\n\f\nInternet-Draft TLS March 2018\n\n\n HKDF-Expand-Label(Secret, Label, Context, Length) =\n HKDF-Expand(Secret, HkdfLabel, Length)\n\n Where HkdfLabel is specified as:\n\n struct {\n uint16 length = Length;\n opaque label<7..255> = \"tls13 \" + Label;\n opaque context<0..255> = Context;\n } HkdfLabel;\n\n Derive-Secret(Secret, Label, Messages) =\n HKDF-Expand-Label(Secret, Label,\n Transcript-Hash(Messages), Hash.length)\n\n The Hash function used by Transcript-Hash and HKDF is the cipher\n suite hash algorithm. Hash.length is its output length in bytes.\n Messages is the concatenation of the indicated handshake messages,\n including the handshake message type and length fields, but not\n including record layer headers. Note that in some cases a zero-\n length Context (indicated by \"\") is passed to HKDF-Expand-Label. The\n Labels specified in this document are all ASCII strings, and do not\n include a trailing NUL byte.\n\n Note: with common hash functions, any label longer than 12 characters\n requires an additional iteration of the hash function to compute.\n The labels in this specification have all been chosen to fit within\n this limit.\n\n Keys are derived from two input secrets using the HKDF-Extract and\n Derive-Secret functions. The general pattern for adding a new secret\n is to use HKDF-Extract with the salt being the current secret state\n and the IKM being the new secret to be added. In this version of TLS\n 1.3, the two input secrets are:\n\n - PSK (a pre-shared key established externally or derived from the\n resumption_master_secret value from a previous connection)\n\n - (EC)DHE shared secret (Section 7.4)\n\n This produces a full key derivation schedule shown in the diagram\n below. In this diagram, the following formatting conventions apply:\n\n - HKDF-Extract is drawn as taking the Salt argument from the top and\n the IKM argument from the left, with its output to the bottom and\n the name of the output on the right.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 94]\n\f\nInternet-Draft TLS March 2018\n\n\n - Derive-Secret's Secret argument is indicated by the incoming\n arrow. For instance, the Early Secret is the Secret for\n generating the client_early_traffic_secret.\n\n - \"0\" indicates a string of Hash-lengths bytes set to 0.\n\n 0\n |\n v\n PSK -> HKDF-Extract = Early Secret\n |\n +-----> Derive-Secret(.,\n | \"ext binder\" |\n | \"res binder\",\n | \"\")\n | = binder_key\n |\n +-----> Derive-Secret(., \"c e traffic\",\n | ClientHello)\n | = client_early_traffic_secret\n |\n +-----> Derive-Secret(., \"e exp master\",\n | ClientHello)\n | = early_exporter_master_secret\n v\n Derive-Secret(., \"derived\", \"\")\n |\n v\n(EC)DHE -> HKDF-Extract = Handshake Secret\n |\n +-----> Derive-Secret(., \"c hs traffic\",\n | ClientHello...ServerHello)\n | = client_handshake_traffic_secret\n |\n +-----> Derive-Secret(., \"s hs traffic\",\n | ClientHello...ServerHello)\n | = server_handshake_traffic_secret\n v\n Derive-Secret(., \"derived\", \"\")\n |\n v\n 0 -> HKDF-Extract = Master Secret\n |\n +-----> Derive-Secret(., \"c ap traffic\",\n | ClientHello...server Finished)\n | = client_application_traffic_secret_0\n |\n +-----> Derive-Secret(., \"s ap traffic\",\n\n\n\nRescorla Expires September 21, 2018 [Page 95]\n\f\nInternet-Draft TLS March 2018\n\n\n | ClientHello...server Finished)\n | = server_application_traffic_secret_0\n |\n +-----> Derive-Secret(., \"exp master\",\n | ClientHello...server Finished)\n | = exporter_master_secret\n |\n +-----> Derive-Secret(., \"res master\",\n ClientHello...client Finished)\n = resumption_master_secret\n\n The general pattern here is that the secrets shown down the left side\n of the diagram are just raw entropy without context, whereas the\n secrets down the right side include handshake context and therefore\n can be used to derive working keys without additional context. Note\n that the different calls to Derive-Secret may take different Messages\n arguments, even with the same secret. In a 0-RTT exchange, Derive-\n Secret is called with four distinct transcripts; in a 1-RTT-only\n exchange with three distinct transcripts.\n\n If a given secret is not available, then the 0-value consisting of a\n string of Hash.length bytes set to zeros is used. Note that this\n does not mean skipping rounds, so if PSK is not in use Early Secret\n will still be HKDF-Extract(0, 0). For the computation of the\n binder_secret, the label is \"ext binder\" for external PSKs (those\n provisioned outside of TLS) and \"res binder\" for resumption PSKs\n (those provisioned as the resumption master secret of a previous\n handshake). The different labels prevent the substitution of one\n type of PSK for the other.\n\n There are multiple potential Early Secret values depending on which\n PSK the server ultimately selects. The client will need to compute\n one for each potential PSK; if no PSK is selected, it will then need\n to compute the early secret corresponding to the zero PSK.\n\n Once all the values which are to be derived from a given secret have\n been computed, that secret SHOULD be erased.\n\n7.2. Updating Traffic Secrets\n\n Once the handshake is complete, it is possible for either side to\n update its sending traffic keys using the KeyUpdate handshake message\n defined in Section 4.6.3. The next generation of traffic keys is\n computed by generating client_/server_application_traffic_secret_N+1\n from client_/server_application_traffic_secret_N as described in this\n section then re-deriving the traffic keys as described in\n Section 7.3.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 96]\n\f\nInternet-Draft TLS March 2018\n\n\n The next-generation application_traffic_secret is computed as:\n\n application_traffic_secret_N+1 =\n HKDF-Expand-Label(application_traffic_secret_N,\n \"traffic upd\", \"\", Hash.length)\n\n Once client/server_application_traffic_secret_N+1 and its associated\n traffic keys have been computed, implementations SHOULD delete\n client_/server_application_traffic_secret_N and its associated\n traffic keys.\n\n7.3. Traffic Key Calculation\n\n The traffic keying material is generated from the following input\n values:\n\n - A secret value\n\n - A purpose value indicating the specific value being generated\n\n - The length of the key being generated\n\n The traffic keying material is generated from an input traffic secret\n value using:\n\n [sender]_write_key = HKDF-Expand-Label(Secret, \"key\", \"\", key_length)\n [sender]_write_iv = HKDF-Expand-Label(Secret, \"iv\" , \"\", iv_length)\n\n [sender] denotes the sending side. The Secret value for each record\n type is shown in the table below.\n\n +-------------------+---------------------------------------+\n | Record Type | Secret |\n +-------------------+---------------------------------------+\n | 0-RTT Application | client_early_traffic_secret |\n | | |\n | Handshake | [sender]_handshake_traffic_secret |\n | | |\n | Application Data | [sender]_application_traffic_secret_N |\n +-------------------+---------------------------------------+\n\n All the traffic keying material is recomputed whenever the underlying\n Secret changes (e.g., when changing from the handshake to application\n data keys or upon a key update).\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 97]\n\f\nInternet-Draft TLS March 2018\n\n\n7.4. (EC)DHE Shared Secret Calculation\n\n7.4.1. Finite Field Diffie-Hellman\n\n For finite field groups, a conventional Diffie-Hellman [DH76]\n computation is performed. The negotiated key (Z) is converted to a\n byte string by encoding in big-endian and left padded with zeros up\n to the size of the prime. This byte string is used as the shared\n secret in the key schedule as specified above.\n\n Note that this construction differs from previous versions of TLS\n which remove leading zeros.\n\n7.4.2. Elliptic Curve Diffie-Hellman\n\n For secp256r1, secp384r1 and secp521r1, ECDH calculations (including\n parameter and key generation as well as the shared secret\n calculation) are performed according to [IEEE1363] using the ECKAS-\n DH1 scheme with the identity map as key derivation function (KDF), so\n that the shared secret is the x-coordinate of the ECDH shared secret\n elliptic curve point represented as an octet string. Note that this\n octet string (Z in IEEE 1363 terminology) as output by FE2OSP, the\n Field Element to Octet String Conversion Primitive, has constant\n length for any given field; leading zeros found in this octet string\n MUST NOT be truncated.\n\n (Note that this use of the identity KDF is a technicality. The\n complete picture is that ECDH is employed with a non-trivial KDF\n because TLS does not directly use this secret for anything other than\n for computing other secrets.)\n\n ECDH functions are used as follows:\n\n - The public key to put into the KeyShareEntry.key_exchange\n structure is the result of applying the ECDH scalar multiplication\n function to the secret key of appropriate length (into scalar\n input) and the standard public basepoint (into u-coordinate point\n input).\n\n - The ECDH shared secret is the result of applying the ECDH scalar\n multiplication function to the secret key (into scalar input) and\n the peer's public key (into u-coordinate point input). The output\n is used raw, with no processing.\n\n For X25519 and X448, implementations SHOULD use the approach\n specified in [RFC7748] to calculate the Diffie-Hellman shared secret.\n Implementations MUST check whether the computed Diffie-Hellman shared\n secret is the all-zero value and abort if so, as described in\n\n\n\nRescorla Expires September 21, 2018 [Page 98]\n\f\nInternet-Draft TLS March 2018\n\n\n Section 6 of [RFC7748]. If implementors use an alternative\n implementation of these elliptic curves, they SHOULD perform the\n additional checks specified in Section 7 of [RFC7748].\n\n7.5. Exporters\n\n [RFC5705] defines keying material exporters for TLS in terms of the\n TLS pseudorandom function (PRF). This document replaces the PRF with\n HKDF, thus requiring a new construction. The exporter interface\n remains the same.\n\n The exporter value is computed as:\n\n TLS-Exporter(label, context_value, key_length) =\n HKDF-Expand-Label(Derive-Secret(Secret, label, \"\"),\n \"exporter\", Hash(context_value), key_length)\n\n Where Secret is either the early_exporter_master_secret or the\n exporter_master_secret. Implementations MUST use the\n exporter_master_secret unless explicitly specified by the\n application. The early_exporter_master_secret is defined for use in\n settings where an exporter is needed for 0-RTT data. A separate\n interface for the early exporter is RECOMMENDED; this avoids the\n exporter user accidentally using an early exporter when a regular one\n is desired or vice versa.\n\n If no context is provided, the context_value is zero-length.\n Consequently, providing no context computes the same value as\n providing an empty context. This is a change from previous versions\n of TLS where an empty context produced a different output to an\n absent context. As of this document's publication, no allocated\n exporter label is used both with and without a context. Future\n specifications MUST NOT define a use of exporters that permit both an\n empty context and no context with the same label. New uses of\n exporters SHOULD provide a context in all exporter computations,\n though the value could be empty.\n\n Requirements for the format of exporter labels are defined in section\n 4 of [RFC5705].\n\n8. 0-RTT and Anti-Replay\n\n As noted in Section 2.3 and Appendix E.5, TLS does not provide\n inherent replay protections for 0-RTT data. There are two potential\n threats to be concerned with:\n\n - Network attackers who mount a replay attack by simply duplicating\n a flight of 0-RTT data.\n\n\n\nRescorla Expires September 21, 2018 [Page 99]\n\f\nInternet-Draft TLS March 2018\n\n\n - Network attackers who take advantage of client retry behavior to\n arrange for the server to receive multiple copies of an\n application message. This threat already exists to some extent\n because clients that value robustness respond to network errors by\n attempting to retry requests. However, 0-RTT adds an additional\n dimension for any server system which does not maintain globally\n consistent server state. Specifically, if a server system has\n multiple zones where tickets from zone A will not be accepted in\n zone B, then an attacker can duplicate a ClientHello and early\n data intended for A to both A and B. At A, the data will be\n accepted in 0-RTT, but at B the server will reject 0-RTT data and\n instead force a full handshake. If the attacker blocks the\n ServerHello from A, then the client will complete the handshake\n with B and probably retry the request, leading to duplication on\n the server system as a whole.\n\n The first class of attack can be prevented by sharing state to\n guarantee that the 0-RTT data is accepted at most once. Servers\n SHOULD provide that level of replay safety, by implementing one of\n the methods described in this section or by equivalent means. It is\n understood, however, that due to operational concerns not all\n deployments will maintain state at that level. Therefore, in normal\n operation, clients will not know which, if any, of these mechanisms\n servers actually implement and hence MUST only send early data which\n they deem safe to be replayed.\n\n In addition to the direct effects of replays, there is a class of\n attacks where even operations normally considered idempotent could be\n exploited by a large number of replays (timing attacks, resource\n limit exhaustion and others described in Appendix E.5). Those can be\n mitigated by ensuring that every 0-RTT payload can be replayed only a\n limited number of times. The server MUST ensure that any instance of\n it (be it a machine, a thread or any other entity within the relevant\n serving infrastructure) would accept 0-RTT for the same 0-RTT\n handshake at most once; this limits the number of replays to the\n number of server instances in the deployment. Such a guarantee can\n be accomplished by locally recording data from recently-received\n ClientHellos and rejecting repeats, or by any other method that\n provides the same or a stronger guarantee. The \"at most once per\n server instance\" guarantee is a minimum requirement; servers SHOULD\n limit 0-RTT replays further when feasible.\n\n The second class of attack cannot be prevented at the TLS layer and\n MUST be dealt with by any application. Note that any application\n whose clients implement any kind of retry behavior already needs to\n implement some sort of anti-replay defense.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 100]\n\f\nInternet-Draft TLS March 2018\n\n\n8.1. Single-Use Tickets\n\n The simplest form of anti-replay defense is for the server to only\n allow each session ticket to be used once. For instance, the server\n can maintain a database of all outstanding valid tickets; deleting\n each ticket from the database as it is used. If an unknown ticket is\n provided, the server would then fall back to a full handshake.\n\n If the tickets are not self-contained but rather are database keys,\n and the corresponding PSKs are deleted upon use, then connections\n established using PSKs enjoy forward secrecy. This improves security\n for all 0-RTT data and PSK usage when PSK is used without (EC)DHE.\n\n Because this mechanism requires sharing the session database between\n server nodes in environments with multiple distributed servers, it\n may be hard to achieve high rates of successful PSK 0-RTT connections\n when compared to self-encrypted tickets. Unlike session databases,\n session tickets can successfully do PSK-based session establishment\n even without consistent storage, though when 0-RTT is allowed they\n still require consistent storage for anti-replay of 0-RTT data, as\n detailed in the following section.\n\n8.2. Client Hello Recording\n\n An alternative form of anti-replay is to record a unique value\n derived from the ClientHello (generally either the random value or\n the PSK binder) and reject duplicates. Recording all ClientHellos\n causes state to grow without bound, but a server can instead record\n ClientHellos within a given time window and use the\n \"obfuscated_ticket_age\" to ensure that tickets aren't reused outside\n that window.\n\n In order to implement this, when a ClientHello is received, the\n server first verifies the PSK binder as described Section 4.2.11. It\n then computes the expected_arrival_time as described in the next\n section and rejects 0-RTT if it is outside the recording window,\n falling back to the 1-RTT handshake.\n\n If the expected arrival time is in the window, then the server checks\n to see if it has recorded a matching ClientHello. If one is found,\n it either aborts the handshake with an \"illegal_parameter\" alert or\n accepts the PSK but reject 0-RTT. If no matching ClientHello is\n found, then it accepts 0-RTT and then stores the ClientHello for as\n long as the expected_arrival_time is inside the window. Servers MAY\n also implement data stores with false positives, such as Bloom\n filters, in which case they MUST respond to apparent replay by\n rejecting 0-RTT but MUST NOT abort the handshake.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 101]\n\f\nInternet-Draft TLS March 2018\n\n\n The server MUST derive the storage key only from validated sections\n of the ClientHello. If the ClientHello contains multiple PSK\n identities, then an attacker can create multiple ClientHellos with\n different binder values for the less-preferred identity on the\n assumption that the server will not verify it, as recommended by\n Section 4.2.11. I.e., if the client sends PSKs A and B but the\n server prefers A, then the attacker can change the binder for B\n without affecting the binder for A. If the binder for B is part of\n the storage key, then this ClientHello will not appear as a\n duplicate, which will cause the ClientHello to be accepted, and may\n cause side effects such as replay cache pollution, although any 0-RTT\n data will not be decryptable because it will use different keys. If\n the validated binder or the ClientHello.random are used as the\n storage key, then this attack is not possible.\n\n Because this mechanism does not require storing all outstanding\n tickets, it may be easier to implement in distributed systems with\n high rates of resumption and 0-RTT, at the cost of potentially weaker\n anti-replay defense because of the difficulty of reliably storing and\n retrieving the received ClientHello messages. In many such systems,\n it is impractical to have globally consistent storage of all the\n received ClientHellos. In this case, the best anti-replay protection\n is provided by having a single storage zone be authoritative for a\n given ticket and refusing 0-RTT for that ticket in any other zone.\n This approach prevents simple replay by the attacker because only one\n zone will accept 0-RTT data. A weaker design is to implement\n separate storage for each zone but allow 0-RTT in any zone. This\n approach limits the number of replays to once per zone. Application\n message duplication of course remains possible with either design.\n\n When implementations are freshly started, they SHOULD reject 0-RTT as\n long as any portion of their recording window overlaps the startup\n time. Otherwise, they run the risk of accepting replays which were\n originally sent during that period.\n\n Note: If the client's clock is running much faster than the server's\n then a ClientHello may be received that is outside the window in the\n future, in which case it might be accepted for 1-RTT, causing a\n client retry, and then acceptable later for 0-RTT. This is another\n variant of the second form of attack described above.\n\n8.3. Freshness Checks\n\n Because the ClientHello indicates the time at which the client sent\n it, it is possible to efficiently determine whether a ClientHello was\n likely sent reasonably recently and only accept 0-RTT for such a\n ClientHello, otherwise falling back to a 1-RTT handshake. This is\n necessary for the ClientHello storage mechanism described in\n\n\n\nRescorla Expires September 21, 2018 [Page 102]\n\f\nInternet-Draft TLS March 2018\n\n\n Section 8.2 because otherwise the server needs to store an unlimited\n number of ClientHellos and is a useful optimization for self-\n contained single-use tickets because it allows efficient rejection of\n ClientHellos which cannot be used for 0-RTT.\n\n In order to implement this mechanism, a server needs to store the\n time that the server generated the session ticket, offset by an\n estimate of the round trip time between client and server. I.e.,\n\n adjusted_creation_time = creation_time + estimated_RTT\n\n This value can be encoded in the ticket, thus avoiding the need to\n keep state for each outstanding ticket. The server can determine the\n client's view of the age of the ticket by subtracting the ticket's\n \"ticket_age_add value\" from the \"obfuscated_ticket_age\" parameter in\n the client's \"pre_shared_key\" extension. The server can determine\n the \"expected arrival time\" of the ClientHello as:\n\n expected_arrival_time = adjusted_creation_time + clients_ticket_age\n\n When a new ClientHello is received, the expected_arrival_time is then\n compared against the current server wall clock time and if they\n differ by more than a certain amount, 0-RTT is rejected, though the\n 1-RTT handshake can be allowed to complete.\n\n There are several potential sources of error that might cause\n mismatches between the expected arrival time and the measured time.\n Variations in client and server clock rates are likely to be minimal,\n though potentially the absolute times may be off by large values.\n Network propagation delays are the most likely causes of a mismatch\n in legitimate values for elapsed time. Both the NewSessionTicket and\n ClientHello messages might be retransmitted and therefore delayed,\n which might be hidden by TCP. For clients on the Internet, this\n implies windows on the order of ten seconds to account for errors in\n clocks and variations in measurements; other deployment scenarios may\n have different needs. Clock skew distributions are not symmetric, so\n the optimal tradeoff may involve an asymmetric range of permissible\n mismatch values.\n\n Note that freshness checking alone is not sufficient to prevent\n replays because it does not detect them during the error window,\n which, depending on bandwidth and system capacity could include\n billions of replays in real-world settings. In addition, this\n freshness checking is only done at the time the ClientHello is\n received, and not when later early application data records are\n received. After early data is accepted, records may continue to be\n streamed to the server over a longer time period.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 103]\n\f\nInternet-Draft TLS March 2018\n\n\n9. Compliance Requirements\n\n9.1. Mandatory-to-Implement Cipher Suites\n\n In the absence of an application profile standard specifying\n otherwise, a TLS-compliant application MUST implement the\n TLS_AES_128_GCM_SHA256 [GCM] cipher suite and SHOULD implement the\n TLS_AES_256_GCM_SHA384 [GCM] and TLS_CHACHA20_POLY1305_SHA256\n [RFC7539] cipher suites. (see Appendix B.4)\n\n A TLS-compliant application MUST support digital signatures with\n rsa_pkcs1_sha256 (for certificates), rsa_pss_rsae_sha256 (for\n CertificateVerify and certificates), and ecdsa_secp256r1_sha256. A\n TLS-compliant application MUST support key exchange with secp256r1\n (NIST P-256) and SHOULD support key exchange with X25519 [RFC7748].\n\n9.2. Mandatory-to-Implement Extensions\n\n In the absence of an application profile standard specifying\n otherwise, a TLS-compliant application MUST implement the following\n TLS extensions:\n\n - Supported Versions (\"supported_versions\"; Section 4.2.1)\n\n - Cookie (\"cookie\"; Section 4.2.2)\n\n - Signature Algorithms (\"signature_algorithms\"; Section 4.2.3)\n\n - Signature Algorithms Certificate (\"signature_algorithms_cert\";\n Section 4.2.3)\n\n - Negotiated Groups (\"supported_groups\"; Section 4.2.7)\n\n - Key Share (\"key_share\"; Section 4.2.8)\n\n - Server Name Indication (\"server_name\"; Section 3 of [RFC6066])\n\n All implementations MUST send and use these extensions when offering\n applicable features:\n\n - \"supported_versions\" is REQUIRED for all ClientHello, ServerHello\n and HelloRetryRequest messages.\n\n - \"signature_algorithms\" is REQUIRED for certificate authentication.\n\n - \"supported_groups\" is REQUIRED for ClientHello messages using DHE\n or ECDHE key exchange.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 104]\n\f\nInternet-Draft TLS March 2018\n\n\n - \"key_share\" is REQUIRED for DHE or ECDHE key exchange.\n\n - \"pre_shared_key\" is REQUIRED for PSK key agreement.\n\n - \"psk_key_exchange_modes\" is REQUIRED for PSK key agreement.\n\n A client is considered to be attempting to negotiate using this\n specification if the ClientHello contains a \"supported_versions\"\n extension with 0x0304 contained in its body. Such a ClientHello\n message MUST meet the following requirements:\n\n - If not containing a \"pre_shared_key\" extension, it MUST contain\n both a \"signature_algorithms\" extension and a \"supported_groups\"\n extension.\n\n - If containing a \"supported_groups\" extension, it MUST also contain\n a \"key_share\" extension, and vice versa. An empty\n KeyShare.client_shares vector is permitted.\n\n Servers receiving a ClientHello which does not conform to these\n requirements MUST abort the handshake with a \"missing_extension\"\n alert.\n\n Additionally, all implementations MUST support use of the\n \"server_name\" extension with applications capable of using it.\n Servers MAY require clients to send a valid \"server_name\" extension.\n Servers requiring this extension SHOULD respond to a ClientHello\n lacking a \"server_name\" extension by terminating the connection with\n a \"missing_extension\" alert.\n\n9.3. Protocol Invariants\n\n This section describes invariants that TLS endpoints and middleboxes\n MUST follow. It also applies to earlier versions of TLS.\n\n TLS is designed to be securely and compatibly extensible. Newer\n clients or servers, when communicating with newer peers, should\n negotiate the most preferred common parameters. The TLS handshake\n provides downgrade protection: Middleboxes passing traffic between a\n newer client and newer server without terminating TLS should be\n unable to influence the handshake (see Appendix E.1). At the same\n time, deployments update at different rates, so a newer client or\n server MAY continue to support older parameters, which would allow it\n to interoperate with older endpoints.\n\n For this to work, implementations MUST correctly handle extensible\n fields:\n\n\n\n\nRescorla Expires September 21, 2018 [Page 105]\n\f\nInternet-Draft TLS March 2018\n\n\n - A client sending a ClientHello MUST support all parameters\n advertised in it. Otherwise, the server may fail to interoperate\n by selecting one of those parameters.\n\n - A server receiving a ClientHello MUST correctly ignore all\n unrecognized cipher suites, extensions, and other parameters.\n Otherwise, it may fail to interoperate with newer clients. In TLS\n 1.3, a client receiving a CertificateRequest or NewSessionTicket\n MUST also ignore all unrecognized extensions.\n\n - A middlebox which terminates a TLS connection MUST behave as a\n compliant TLS server (to the original client), including having a\n certificate which the client is willing to accept, and as a\n compliant TLS client (to the original server), including verifying\n the original server's certificate. In particular, it MUST\n generate its own ClientHello containing only parameters it\n understands, and it MUST generate a fresh ServerHello random\n value, rather than forwarding the endpoint's value.\n\n Note that TLS's protocol requirements and security analysis only\n apply to the two connections separately. Safely deploying a TLS\n terminator requires additional security considerations which are\n beyond the scope of this document.\n\n - An middlebox which forwards ClientHello parameters it does not\n understand MUST NOT process any messages beyond that ClientHello.\n It MUST forward all subsequent traffic unmodified. Otherwise, it\n may fail to interoperate with newer clients and servers.\n\n Forwarded ClientHellos may contain advertisements for features not\n supported by the middlebox, so the response may include future TLS\n additions the middlebox does not recognize. These additions MAY\n change any message beyond the ClientHello arbitrarily. In\n particular, the values sent in the ServerHello might change, the\n ServerHello format might change, and the TLSCiphertext format\n might change.\n\n The design of TLS 1.3 was constrained by widely-deployed non-\n compliant TLS middleboxes (see Appendix D.4), however it does not\n relax the invariants. Those middleboxes continue to be non-\n compliant.\n\n10. Security Considerations\n\n Security issues are discussed throughout this memo, especially in\n Appendix C, Appendix D, and Appendix E.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 106]\n\f\nInternet-Draft TLS March 2018\n\n\n11. IANA Considerations\n\n This document uses several registries that were originally created in\n [RFC4346]. IANA [SHALL update/has updated] these to reference this\n document. The registries and their allocation policies are below:\n\n - TLS Cipher Suite Registry: values with the first byte in the range\n 0-254 (decimal) are assigned via Specification Required [RFC8126].\n Values with the first byte 255 (decimal) are reserved for Private\n Use [RFC8126].\n\n IANA [SHALL add/has added] the cipher suites listed in\n Appendix B.4 to the registry. The \"Value\" and \"Description\"\n columns are taken from the table. The \"DTLS-OK\" and \"Recommended\"\n columns are both marked as \"Yes\" for each new cipher suite.\n [[This assumes [I-D.ietf-tls-iana-registry-updates] has been\n applied.]]\n\n - TLS ContentType Registry: Future values are allocated via\n Standards Action [RFC8126].\n\n - TLS Alert Registry: Future values are allocated via Standards\n Action [RFC8126]. IANA [SHALL update/has updated] this registry\n to include values for \"missing_extension\" and\n \"certificate_required\". The \"DTLS-OK\" column is marked as \"Yes\"\n for each new alert.\n\n - TLS HandshakeType Registry: Future values are allocated via\n Standards Action [RFC8126]. IANA [SHALL update/has updated] this\n registry to rename item 4 from \"NewSessionTicket\" to\n \"new_session_ticket\" and to add the\n \"hello_retry_request_RESERVED\", \"encrypted_extensions\",\n \"end_of_early_data\", \"key_update\", and \"message_hash\" values. The\n \"DTLS-OK\" are marked as \"Yes\" for each of these additions.\n\n This document also uses the TLS ExtensionType Registry originally\n created in [RFC4366]. IANA has updated it to reference this\n document. Changes to the registry follow:\n\n - IANA [SHALL update/has updated] the registration policy as\n follows:\n\n Values with the first byte in the range 0-254 (decimal) are\n assigned via Specification Required [RFC8126]. Values with the\n first byte 255 (decimal) are reserved for Private Use [RFC8126].\n\n - IANA [SHALL update/has updated] this registry to include the\n \"key_share\", \"pre_shared_key\", \"psk_key_exchange_modes\",\n\n\n\nRescorla Expires September 21, 2018 [Page 107]\n\f\nInternet-Draft TLS March 2018\n\n\n \"early_data\", \"cookie\", \"supported_versions\",\n \"certificate_authorities\", \"oid_filters\", \"post_handshake_auth\",\n and \"signature_algorithms_cert\", extensions with the values\n defined in this document and the Recommended value of \"Yes\".\n\n - IANA [SHALL update/has updated] this registry to include a \"TLS\n 1.3\" column which lists the messages in which the extension may\n appear. This column [SHALL be/has been] initially populated from\n the table in Section 4.2 with any extension not listed there\n marked as \"-\" to indicate that it is not used by TLS 1.3.\n\n In addition, this document defines two new registries to be\n maintained by IANA:\n\n - TLS SignatureScheme Registry: Values with the first byte in the\n range 0-253 (decimal) are assigned via Specification Required\n [RFC8126]. Values with the first byte 254 or 255 (decimal) are\n reserved for Private Use [RFC8126]. Values with the first byte in\n the range 0-6 or with the second byte in the range 0-3 that are\n not currently allocated are reserved for backwards compatibility.\n This registry SHALL have a \"Recommended\" column. The registry\n [shall be/ has been] initially populated with the values described\n in Section 4.2.3. The following values SHALL be marked as\n \"Recommended\": ecdsa_secp256r1_sha256, ecdsa_secp384r1_sha384,\n rsa_pss_rsae_sha256, rsa_pss_rsae_sha384, rsa_pss_rsae_sha512,\n rsa_pss_pss_sha256, rsa_pss_pss_sha384, rsa_pss_pss_sha512, and\n ed25519.\n\n - TLS PskKeyExchangeMode Registry: Values in the range 0-253\n (decimal) are assigned via Specification Required [RFC8126].\n Values with the first byte 254 or 255 (decimal) are reserved for\n Private Use [RFC8126]. This registry SHALL have a \"Recommended\"\n column. The registry [shall be/ has been] initially populated\n psk_ke (0) and psk_dhe_ke (1). Both SHALL be marked as\n \"Recommended\".\n\n12. References\n\n12.1. Normative References\n\n [DH] Diffie, W. and M. Hellman, \"New Directions in\n Cryptography\", IEEE Transactions on Information Theory,\n V.IT-22 n.6 , June 1977.\n\n [DH76] Diffie, W. and M. Hellman, \"New directions in\n cryptography\", IEEE Transactions on Information\n Theory Vol. 22, pp. 644-654, DOI 10.1109/tit.1976.1055638,\n November 1976.\n\n\n\nRescorla Expires September 21, 2018 [Page 108]\n\f\nInternet-Draft TLS March 2018\n\n\n [GCM] Dworkin, M., \"Recommendation for Block Cipher Modes of\n Operation: Galois/Counter Mode (GCM) and GMAC\",\n NIST Special Publication 800-38D, November 2007.\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC: Keyed-\n Hashing for Message Authentication\", RFC 2104,\n DOI 10.17487/RFC2104, February 1997, .\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119,\n DOI 10.17487/RFC2119, March 1997, .\n\n [RFC5116] McGrew, D., \"An Interface and Algorithms for Authenticated\n Encryption\", RFC 5116, DOI 10.17487/RFC5116, January 2008,\n .\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, DOI 10.17487/RFC5280, May 2008,\n .\n\n [RFC5705] Rescorla, E., \"Keying Material Exporters for Transport\n Layer Security (TLS)\", RFC 5705, DOI 10.17487/RFC5705,\n March 2010, .\n\n [RFC5756] Turner, S., Brown, D., Yiu, K., Housley, R., and T. Polk,\n \"Updates for RSAES-OAEP and RSASSA-PSS Algorithm\n Parameters\", RFC 5756, DOI 10.17487/RFC5756, January 2010,\n .\n\n [RFC5869] Krawczyk, H. and P. Eronen, \"HMAC-based Extract-and-Expand\n Key Derivation Function (HKDF)\", RFC 5869,\n DOI 10.17487/RFC5869, May 2010, .\n\n [RFC6066] Eastlake 3rd, D., \"Transport Layer Security (TLS)\n Extensions: Extension Definitions\", RFC 6066,\n DOI 10.17487/RFC6066, January 2011, .\n\n [RFC6655] McGrew, D. and D. Bailey, \"AES-CCM Cipher Suites for\n Transport Layer Security (TLS)\", RFC 6655,\n DOI 10.17487/RFC6655, July 2012, .\n\n\n\n\nRescorla Expires September 21, 2018 [Page 109]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, DOI 10.17487/RFC6960, June 2013,\n .\n\n [RFC6961] Pettersen, Y., \"The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\", RFC 6961,\n DOI 10.17487/RFC6961, June 2013, .\n\n [RFC6962] Laurie, B., Langley, A., and E. Kasper, \"Certificate\n Transparency\", RFC 6962, DOI 10.17487/RFC6962, June 2013,\n .\n\n [RFC6979] Pornin, T., \"Deterministic Usage of the Digital Signature\n Algorithm (DSA) and Elliptic Curve Digital Signature\n Algorithm (ECDSA)\", RFC 6979, DOI 10.17487/RFC6979, August\n 2013, .\n\n [RFC7301] Friedl, S., Popov, A., Langley, A., and E. Stephan,\n \"Transport Layer Security (TLS) Application-Layer Protocol\n Negotiation Extension\", RFC 7301, DOI 10.17487/RFC7301,\n July 2014, .\n\n [RFC7507] Moeller, B. and A. Langley, \"TLS Fallback Signaling Cipher\n Suite Value (SCSV) for Preventing Protocol Downgrade\n Attacks\", RFC 7507, DOI 10.17487/RFC7507, April 2015,\n .\n\n [RFC7539] Nir, Y. and A. Langley, \"ChaCha20 and Poly1305 for IETF\n Protocols\", RFC 7539, DOI 10.17487/RFC7539, May 2015,\n .\n\n [RFC7748] Langley, A., Hamburg, M., and S. Turner, \"Elliptic Curves\n for Security\", RFC 7748, DOI 10.17487/RFC7748, January\n 2016, .\n\n [RFC7919] Gillmor, D., \"Negotiated Finite Field Diffie-Hellman\n Ephemeral Parameters for Transport Layer Security (TLS)\",\n RFC 7919, DOI 10.17487/RFC7919, August 2016,\n .\n\n [RFC8017] Moriarty, K., Ed., Kaliski, B., Jonsson, J., and A. Rusch,\n \"PKCS #1: RSA Cryptography Specifications Version 2.2\",\n RFC 8017, DOI 10.17487/RFC8017, November 2016,\n .\n\n\n\n\nRescorla Expires September 21, 2018 [Page 110]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC8032] Josefsson, S. and I. Liusvaara, \"Edwards-Curve Digital\n Signature Algorithm (EdDSA)\", RFC 8032,\n DOI 10.17487/RFC8032, January 2017, .\n\n [RFC8126] Cotton, M., Leiba, B., and T. Narten, \"Guidelines for\n Writing an IANA Considerations Section in RFCs\", BCP 26,\n RFC 8126, DOI 10.17487/RFC8126, June 2017,\n .\n\n [RFC8174] Leiba, B., \"Ambiguity of Uppercase vs Lowercase in RFC\n 2119 Key Words\", BCP 14, RFC 8174, DOI 10.17487/RFC8174,\n May 2017, .\n\n [SHS] Dang, Q., \"Secure Hash Standard\", National Institute of\n Standards and Technology report,\n DOI 10.6028/nist.fips.180-4, July 2015.\n\n [X690] ITU-T, \"Information technology - ASN.1 encoding Rules:\n Specification of Basic Encoding Rules (BER), Canonical\n Encoding Rules (CER) and Distinguished Encoding Rules\n (DER)\", ISO/IEC 8825-1:2002, 2002.\n\n [X962] ANSI, \"Public Key Cryptography For The Financial Services\n Industry: The Elliptic Curve Digital Signature Algorithm\n (ECDSA)\", ANSI X9.62, 1998.\n\n12.2. Informative References\n\n [AEAD-LIMITS]\n Luykx, A. and K. Paterson, \"Limits on Authenticated\n Encryption Use in TLS\", 2016,\n .\n\n [Anon18] Anonymous, A., \"Secure Channels for Multiplexed Data\n Streams: Analyzing the TLS 1.3 Record Layer Without\n Elision\", In submission to CRYPTO 2018. RFC EDITOR: PLEASE\n UPDATE THIS REFERENCE AFTER FINAL NOTIFICATION\n (2018-4-29). , 2018.\n\n [BBFKZG16]\n Bhargavan, K., Brzuska, C., Fournet, C., Kohlweiss, M.,\n Zanella-Beguelin, S., and M. Green, \"Downgrade Resilience\n in Key-Exchange Protocols\", Proceedings of IEEE Symposium\n on Security and Privacy (Oakland) 2016 , 2016.\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 111]\n\f\nInternet-Draft TLS March 2018\n\n\n [BBK17] Bhargavan, K., Blanchet, B., and N. Kobeissi, \"Verified\n Models and Reference Implementations for the TLS 1.3\n Standard Candidate\", Proceedings of IEEE Symposium on\n Security and Privacy (Oakland) 2017 , 2017.\n\n [BDFKPPRSZZ16]\n Bhargavan, K., Delignat-Lavaud, A., Fournet, C.,\n Kohlweiss, M., Pan, J., Protzenko, J., Rastogi, A., Swamy,\n N., Zanella-Beguelin, S., and J. Zinzindohoue,\n \"Implementing and Proving the TLS 1.3 Record Layer\",\n Proceedings of IEEE Symposium on Security and Privacy\n (Oakland) 2017 , December 2016,\n .\n\n [Ben17a] Benjamin, D., \"Presentation before the TLS WG at IETF\n 100\", 2017,\n .\n\n [Ben17b] Benjamin, D., \"Additional TLS 1.3 results from Chrome\",\n 2017, .\n\n [BMMT15] Badertscher, C., Matt, C., Maurer, U., and B. Tackmann,\n \"Augmented Secure Channels and the Goal of the TLS 1.3\n Record Layer\", ProvSec 2015 , September 2015,\n .\n\n [BT16] Bellare, M. and B. Tackmann, \"The Multi-User Security of\n Authenticated Encryption: AES-GCM in TLS 1.3\", Proceedings\n of CRYPTO 2016 , 2016, .\n\n [CCG16] Cohn-Gordon, K., Cremers, C., and L. Garratt, \"On Post-\n Compromise Security\", IEEE Computer Security Foundations\n Symposium , 2015.\n\n [CHECKOWAY]\n Checkoway, S., Shacham, H., Maskiewicz, J., Garman, C.,\n Fried, J., Cohney, S., Green, M., Heninger, N., Weinmann,\n R., and E. Rescorla, \"A Systematic Analysis of the Juniper\n Dual EC Incident\", Proceedings of the 2016 ACM SIGSAC\n Conference on Computer and Communications Security\n - CCS'16, DOI 10.1145/2976749.2978395, 2016.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 112]\n\f\nInternet-Draft TLS March 2018\n\n\n [CHHSV17] Cremers, C., Horvat, M., Hoyland, J., van der Merwe, T.,\n and S. Scott, \"Awkward Handshake: Possible mismatch of\n client/server view on client authentication in post-\n handshake mode in Revision 18\", 2017,\n .\n\n [CHSV16] Cremers, C., Horvat, M., Scott, S., and T. van der Merwe,\n \"Automated Analysis and Verification of TLS 1.3: 0-RTT,\n Resumption and Delayed Authentication\", Proceedings of\n IEEE Symposium on Security and Privacy (Oakland) 2016 ,\n 2016, .\n\n [CK01] Canetti, R. and H. Krawczyk, \"Analysis of Key-Exchange\n Protocols and Their Use for Building Secure Channels\",\n Proceedings of Eurocrypt 2001 , 2001.\n\n [CLINIC] Miller, B., Huang, L., Joseph, A., and J. Tygar, \"I Know\n Why You Went to the Clinic: Risks and Realization of HTTPS\n Traffic Analysis\", Privacy Enhancing Technologies pp.\n 143-163, DOI 10.1007/978-3-319-08506-7_8, 2014.\n\n [DFGS15] Dowling, B., Fischlin, M., Guenther, F., and D. Stebila,\n \"A Cryptographic Analysis of the TLS 1.3 draft-10 Full and\n Pre-shared Key Handshake Protocol\", Proceedings of ACM CCS\n 2015 , 2015, .\n\n [DFGS16] Dowling, B., Fischlin, M., Guenther, F., and D. Stebila,\n \"A Cryptographic Analysis of the TLS 1.3 draft-10 Full and\n Pre-shared Key Handshake Protocol\", TRON 2016 , 2016,\n .\n\n [DOW92] Diffie, W., van Oorschot, P., and M. Wiener,\n \"Authentication and authenticated key exchanges\", Designs,\n Codes and Cryptography , 1992.\n\n [DSS] National Institute of Standards and Technology, U.S.\n Department of Commerce, \"Digital Signature Standard,\n version 4\", NIST FIPS PUB 186-4, 2013.\n\n [ECDSA] American National Standards Institute, \"Public Key\n Cryptography for the Financial Services Industry: The\n Elliptic Curve Digital Signature Algorithm (ECDSA)\",\n ANSI ANS X9.62-2005, November 2005.\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 113]\n\f\nInternet-Draft TLS March 2018\n\n\n [FG17] Fischlin, M. and F. Guenther, \"Replay Attacks on Zero\n Round-Trip Time: The Case of the TLS 1.3 Handshake\n Candidates\", Proceedings of Euro S\"P 2017 , 2017,\n .\n\n [FGSW16] Fischlin, M., Guenther, F., Schmidt, B., and B. Warinschi,\n \"Key Confirmation in Key Exchange: A Formal Treatment and\n Implications for TLS 1.3\", Proceedings of IEEE Symposium\n on Security and Privacy (Oakland) 2016 , 2016,\n .\n\n [FW15] Florian Weimer, ., \"Factoring RSA Keys With TLS Perfect\n Forward Secrecy\", September 2015.\n\n [HCJ16] Husak, M., Čermak, M., Jirsik, T., and P.\n Čeleda, \"HTTPS traffic analysis and client\n identification using passive SSL/TLS fingerprinting\",\n EURASIP Journal on Information Security Vol. 2016,\n DOI 10.1186/s13635-016-0030-7, February 2016.\n\n [HGFS15] Hlauschek, C., Gruber, M., Fankhauser, F., and C. Schanes,\n \"Prying Open Pandora's Box: KCI Attacks against TLS\",\n Proceedings of USENIX Workshop on Offensive Technologies ,\n 2015.\n\n [I-D.ietf-tls-iana-registry-updates]\n Salowey, J. and S. Turner, \"IANA Registry Updates for TLS\n and DTLS\", draft-ietf-tls-iana-registry-updates-04 (work\n in progress), February 2018.\n\n [I-D.ietf-tls-tls13-vectors]\n Thomson, M., \"Example Handshake Traces for TLS 1.3\",\n draft-ietf-tls-tls13-vectors-03 (work in progress),\n December 2017.\n\n [IEEE1363]\n IEEE, \"Standard Specifications for Public Key\n Cryptography\", IEEE 1363 , 2000.\n\n [JSS15] Jager, T., Schwenk, J., and J. Somorovsky, \"On the\n Security of TLS 1.3 and QUIC Against Weaknesses in PKCS#1\n v1.5 Encryption\", Proceedings of ACM CCS 2015 , 2015,\n .\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 114]\n\f\nInternet-Draft TLS March 2018\n\n\n [KEYAGREEMENT]\n Barker, E., Chen, L., Roginsky, A., and M. Smid,\n \"Recommendation for Pair-Wise Key Establishment Schemes\n Using Discrete Logarithm Cryptography\", National Institute\n of Standards and Technology report,\n DOI 10.6028/nist.sp.800-56ar2, May 2013.\n\n [Kraw10] Krawczyk, H., \"Cryptographic Extraction and Key\n Derivation: The HKDF Scheme\", Proceedings of CRYPTO 2010 ,\n 2010, .\n\n [Kraw16] Krawczyk, H., \"A Unilateral-to-Mutual Authentication\n Compiler for Key Exchange (with Applications to Client\n Authentication in TLS 1.3\", Proceedings of ACM CCS 2016 ,\n 2016, .\n\n [KW16] Krawczyk, H. and H. Wee, \"The OPTLS Protocol and TLS 1.3\",\n Proceedings of Euro S\"P 2016 , 2016,\n .\n\n [LXZFH16] Li, X., Xu, J., Feng, D., Zhang, Z., and H. Hu, \"Multiple\n Handshakes Security of TLS 1.3 Candidates\", Proceedings of\n IEEE Symposium on Security and Privacy (Oakland) 2016 ,\n 2016, .\n\n [Mac17] MacCarthaigh, C., \"Security Review of TLS1.3 0-RTT\", 2017,\n .\n\n [PSK-FINISHED]\n Cremers, C., Horvat, M., van der Merwe, T., and S. Scott,\n \"Revision 10: possible attack if client authentication is\n allowed during PSK\", 2015, .\n\n [REKEY] Abdalla, M. and M. Bellare, \"Increasing the Lifetime of a\n Key: A Comparative Analysis of the Security of Re-keying\n Techniques\", ASIACRYPT2000 , October 2000.\n\n [Res17a] Rescorla, E., \"Preliminary data on Firefox TLS 1.3\n Middlebox experiment\", 2017, .\n\n [Res17b] Rescorla, E., \"More compatibility measurement results\",\n 2017, .\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 115]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC3552] Rescorla, E. and B. Korver, \"Guidelines for Writing RFC\n Text on Security Considerations\", BCP 72, RFC 3552,\n DOI 10.17487/RFC3552, July 2003, .\n\n [RFC4086] Eastlake 3rd, D., Schiller, J., and S. Crocker,\n \"Randomness Requirements for Security\", BCP 106, RFC 4086,\n DOI 10.17487/RFC4086, June 2005, .\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346,\n DOI 10.17487/RFC4346, April 2006, .\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, DOI 10.17487/RFC4366, April 2006,\n .\n\n [RFC4492] Blake-Wilson, S., Bolyard, N., Gupta, V., Hawk, C., and B.\n Moeller, \"Elliptic Curve Cryptography (ECC) Cipher Suites\n for Transport Layer Security (TLS)\", RFC 4492,\n DOI 10.17487/RFC4492, May 2006, .\n\n [RFC5077] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 5077, DOI 10.17487/RFC5077,\n January 2008, .\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246,\n DOI 10.17487/RFC5246, August 2008, .\n\n [RFC5764] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for the Secure\n Real-time Transport Protocol (SRTP)\", RFC 5764,\n DOI 10.17487/RFC5764, May 2010, .\n\n [RFC5929] Altman, J., Williams, N., and L. Zhu, \"Channel Bindings\n for TLS\", RFC 5929, DOI 10.17487/RFC5929, July 2010,\n .\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 116]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC6091] Mavrogiannopoulos, N. and D. Gillmor, \"Using OpenPGP Keys\n for Transport Layer Security (TLS) Authentication\",\n RFC 6091, DOI 10.17487/RFC6091, February 2011,\n .\n\n [RFC6176] Turner, S. and T. Polk, \"Prohibiting Secure Sockets Layer\n (SSL) Version 2.0\", RFC 6176, DOI 10.17487/RFC6176, March\n 2011, .\n\n [RFC6347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security Version 1.2\", RFC 6347, DOI 10.17487/RFC6347,\n January 2012, .\n\n [RFC6520] Seggelmann, R., Tuexen, M., and M. Williams, \"Transport\n Layer Security (TLS) and Datagram Transport Layer Security\n (DTLS) Heartbeat Extension\", RFC 6520,\n DOI 10.17487/RFC6520, February 2012, .\n\n [RFC7230] Fielding, R., Ed. and J. Reschke, Ed., \"Hypertext Transfer\n Protocol (HTTP/1.1): Message Syntax and Routing\",\n RFC 7230, DOI 10.17487/RFC7230, June 2014,\n .\n\n [RFC7250] Wouters, P., Ed., Tschofenig, H., Ed., Gilmore, J.,\n Weiler, S., and T. Kivinen, \"Using Raw Public Keys in\n Transport Layer Security (TLS) and Datagram Transport\n Layer Security (DTLS)\", RFC 7250, DOI 10.17487/RFC7250,\n June 2014, .\n\n [RFC7465] Popov, A., \"Prohibiting RC4 Cipher Suites\", RFC 7465,\n DOI 10.17487/RFC7465, February 2015, .\n\n [RFC7568] Barnes, R., Thomson, M., Pironti, A., and A. Langley,\n \"Deprecating Secure Sockets Layer Version 3.0\", RFC 7568,\n DOI 10.17487/RFC7568, June 2015, .\n\n [RFC7627] Bhargavan, K., Ed., Delignat-Lavaud, A., Pironti, A.,\n Langley, A., and M. Ray, \"Transport Layer Security (TLS)\n Session Hash and Extended Master Secret Extension\",\n RFC 7627, DOI 10.17487/RFC7627, September 2015,\n .\n\n [RFC7685] Langley, A., \"A Transport Layer Security (TLS) ClientHello\n Padding Extension\", RFC 7685, DOI 10.17487/RFC7685,\n October 2015, .\n\n\n\nRescorla Expires September 21, 2018 [Page 117]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC7924] Santesson, S. and H. Tschofenig, \"Transport Layer Security\n (TLS) Cached Information Extension\", RFC 7924,\n DOI 10.17487/RFC7924, July 2016, .\n\n [RFC8305] Schinazi, D. and T. Pauly, \"Happy Eyeballs Version 2:\n Better Connectivity Using Concurrency\", RFC 8305,\n DOI 10.17487/RFC8305, December 2017, .\n\n [RSA] Rivest, R., Shamir, A., and L. Adleman, \"A Method for\n Obtaining Digital Signatures and Public-Key\n Cryptosystems\", Communications of the ACM v. 21, n. 2, pp.\n 120-126., February 1978.\n\n [SIGMA] Krawczyk, H., \"SIGMA: the 'SIGn-and-MAc' approach to\n authenticated Diffie-Hellman and its use in the IKE\n protocols\", Proceedings of CRYPTO 2003 , 2003.\n\n [SLOTH] Bhargavan, K. and G. Leurent, \"Transcript Collision\n Attacks: Breaking Authentication in TLS, IKE, and SSH\",\n Network and Distributed System Security Symposium (NDSS\n 2016) , 2016.\n\n [SSL2] Hickman, K., \"The SSL Protocol\", February 1995.\n\n [SSL3] Freier, A., Karlton, P., and P. Kocher, \"The SSL 3.0\n Protocol\", November 1996.\n\n [TIMING] Boneh, D. and D. Brumley, \"Remote timing attacks are\n practical\", USENIX Security Symposium, 2003.\n\n [X501] \"Information Technology - Open Systems Interconnection -\n The Directory: Models\", ITU-T X.501, 1993.\n\n12.3. URIs\n\n [1] mailto:[redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 118]\n\f\nInternet-Draft TLS March 2018\n\n\nAppendix A. State Machine\n\n This section provides a summary of the legal state transitions for\n the client and server handshakes. State names (in all capitals,\n e.g., START) have no formal meaning but are provided for ease of\n comprehension. Actions which are taken only in certain circumstances\n are indicated in []. The notation \"K_{send,recv} = foo\" means \"set\n the send/recv key to the given key\".\n\nA.1. Client\n\n START <----+\n Send ClientHello | | Recv HelloRetryRequest\n [K_send = early data] | |\n v |\n / WAIT_SH ----+\n | | Recv ServerHello\n | | K_recv = handshake\n Can | V\n send | WAIT_EE\n early | | Recv EncryptedExtensions\n data | +--------+--------+\n | Using | | Using certificate\n | PSK | v\n | | WAIT_CERT_CR\n | | Recv | | Recv CertificateRequest\n | | Certificate | v\n | | | WAIT_CERT\n | | | | Recv Certificate\n | | v v\n | | WAIT_CV\n | | | Recv CertificateVerify\n | +> WAIT_FINISHED <+\n | | Recv Finished\n \\ | [Send EndOfEarlyData]\n | K_send = handshake\n | [Send Certificate [+ CertificateVerify]]\n Can send | Send Finished\n app data --> | K_send = K_recv = application\n after here v\n CONNECTED\n\n Note that with the transitions as shown above, clients may send\n alerts that derive from post-ServerHello messages in the clear or\n with the early data keys. If clients need to send such alerts, they\n SHOULD first rekey to the handshake keys if possible.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 119]\n\f\nInternet-Draft TLS March 2018\n\n\nA.2. Server\n\n START <-----+\n Recv ClientHello | | Send HelloRetryRequest\n v |\n RECVD_CH ----+\n | Select parameters\n v\n NEGOTIATED\n | Send ServerHello\n | K_send = handshake\n | Send EncryptedExtensions\n | [Send CertificateRequest]\n Can send | [Send Certificate + CertificateVerify]\n app data | Send Finished\n after --> | K_send = application\n here +--------+--------+\n No 0-RTT | | 0-RTT\n | |\n K_recv = handshake | | K_recv = early data\n [Skip decrypt errors] | +------> WAIT_EOED -+\n | | Recv | | Recv EndOfEarlyData\n | | early data | | K_recv = handshake\n | +------------+ |\n | |\n +> WAIT_FLIGHT2 <--------+\n |\n +--------+--------+\n No auth | | Client auth\n | |\n | v\n | WAIT_CERT\n | Recv | | Recv Certificate\n | empty | v\n | Certificate | WAIT_CV\n | | | Recv\n | v | CertificateVerify\n +-> WAIT_FINISHED <---+\n | Recv Finished\n | K_recv = application\n v\n CONNECTED\n\nAppendix B. Protocol Data Structures and Constant Values\n\n This section provides the normative protocol types and constants\n definitions. Values listed as _RESERVED were used in previous\n versions of TLS and are listed here for completeness. TLS 1.3\n\n\n\nRescorla Expires September 21, 2018 [Page 120]\n\f\nInternet-Draft TLS March 2018\n\n\n implementations MUST NOT send them but might receive them from older\n TLS implementations.\n\nB.1. Record Layer\n\n enum {\n invalid(0),\n change_cipher_spec(20),\n alert(21),\n handshake(22),\n application_data(23),\n (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion legacy_record_version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n struct {\n opaque content[TLSPlaintext.length];\n ContentType type;\n uint8 zeros[length_of_padding];\n } TLSInnerPlaintext;\n\n struct {\n ContentType opaque_type = application_data; /* 23 */\n ProtocolVersion legacy_record_version = 0x0303; /* TLS v1.2 */\n uint16 length;\n opaque encrypted_record[TLSCiphertext.length];\n } TLSCiphertext;\n\nB.2. Alert Messages\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 121]\n\f\nInternet-Draft TLS March 2018\n\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure_RESERVED(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n inappropriate_fallback(86),\n user_canceled(90),\n no_renegotiation_RESERVED(100),\n missing_extension(109),\n unsupported_extension(110),\n certificate_unobtainable_RESERVED(111),\n unrecognized_name(112),\n bad_certificate_status_response(113),\n bad_certificate_hash_value_RESERVED(114),\n unknown_psk_identity(115),\n certificate_required(116),\n no_application_protocol(120),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 122]\n\f\nInternet-Draft TLS March 2018\n\n\nB.3. Handshake Protocol\n\n enum {\n hello_request_RESERVED(0),\n client_hello(1),\n server_hello(2),\n hello_verify_request_RESERVED(3),\n new_session_ticket(4),\n end_of_early_data(5),\n hello_retry_request_RESERVED(6),\n encrypted_extensions(8),\n certificate(11),\n server_key_exchange_RESERVED(12),\n certificate_request(13),\n server_hello_done_RESERVED(14),\n certificate_verify(15),\n client_key_exchange_RESERVED(16),\n finished(20),\n key_update(24),\n message_hash(254),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (Handshake.msg_type) {\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case end_of_early_data: EndOfEarlyData;\n case encrypted_extensions: EncryptedExtensions;\n case certificate_request: CertificateRequest;\n case certificate: Certificate;\n case certificate_verify: CertificateVerify;\n case finished: Finished;\n case new_session_ticket: NewSessionTicket;\n case key_update: KeyUpdate;\n };\n } Handshake;\n\nB.3.1. Key Exchange Messages\n\n uint16 ProtocolVersion;\n opaque Random[32];\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n struct {\n\n\n\nRescorla Expires September 21, 2018 [Page 123]\n\f\nInternet-Draft TLS March 2018\n\n\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id<0..32>;\n CipherSuite cipher_suites<2..2^16-2>;\n opaque legacy_compression_methods<1..2^8-1>;\n Extension extensions<8..2^16-1>;\n } ClientHello;\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id_echo<0..32>;\n CipherSuite cipher_suite;\n uint8 legacy_compression_method = 0;\n Extension extensions<6..2^16-1>;\n } ServerHello;\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n server_name(0), /* RFC 6066 */\n max_fragment_length(1), /* RFC 6066 */\n status_request(5), /* RFC 6066 */\n supported_groups(10), /* RFC 4492, 7919 */\n signature_algorithms(13), /* [[this document]] */\n use_srtp(14), /* RFC 5764 */\n heartbeat(15), /* RFC 6520 */\n application_layer_protocol_negotiation(16), /* RFC 7301 */\n signed_certificate_timestamp(18), /* RFC 6962 */\n client_certificate_type(19), /* RFC 7250 */\n server_certificate_type(20), /* RFC 7250 */\n padding(21), /* RFC 7685 */\n RESERVED(40), /* Used but never assigned */\n pre_shared_key(41), /* [[this document]] */\n early_data(42), /* [[this document]] */\n supported_versions(43), /* [[this document]] */\n cookie(44), /* [[this document]] */\n psk_key_exchange_modes(45), /* [[this document]] */\n RESERVED(46), /* Used but never assigned */\n certificate_authorities(47), /* [[this document]] */\n oid_filters(48), /* [[this document]] */\n post_handshake_auth(49), /* [[this document]] */\n signature_algorithms_cert(50), /* [[this document]] */\n key_share(51), /* [[this document]] */\n (65535)\n\n\n\nRescorla Expires September 21, 2018 [Page 124]\n\f\nInternet-Draft TLS March 2018\n\n\n } ExtensionType;\n\n struct {\n NamedGroup group;\n opaque key_exchange<1..2^16-1>;\n } KeyShareEntry;\n\n struct {\n KeyShareEntry client_shares<0..2^16-1>;\n } KeyShareClientHello;\n\n struct {\n NamedGroup selected_group;\n } KeyShareHelloRetryRequest;\n\n struct {\n KeyShareEntry server_share;\n } KeyShareServerHello;\n\n struct {\n uint8 legacy_form = 4;\n opaque X[coordinate_length];\n opaque Y[coordinate_length];\n } UncompressedPointRepresentation;\n\n enum { psk_ke(0), psk_dhe_ke(1), (255) } PskKeyExchangeMode;\n\n struct {\n PskKeyExchangeMode ke_modes<1..255>;\n } PskKeyExchangeModes;\n\n struct {} Empty;\n\n struct {\n select (Handshake.msg_type) {\n case new_session_ticket: uint32 max_early_data_size;\n case client_hello: Empty;\n case encrypted_extensions: Empty;\n };\n } EarlyDataIndication;\n\n struct {\n opaque identity<1..2^16-1>;\n uint32 obfuscated_ticket_age;\n } PskIdentity;\n\n opaque PskBinderEntry<32..255>;\n\n\n\n\nRescorla Expires September 21, 2018 [Page 125]\n\f\nInternet-Draft TLS March 2018\n\n\n struct {\n PskIdentity identities<7..2^16-1>;\n PskBinderEntry binders<33..2^16-1>;\n } OfferedPsks;\n\n struct {\n select (Handshake.msg_type) {\n case client_hello: OfferedPsks;\n case server_hello: uint16 selected_identity;\n };\n } PreSharedKeyExtension;\n\nB.3.1.1. Version Extension\n\n struct {\n select (Handshake.msg_type) {\n case client_hello:\n ProtocolVersion versions<2..254>;\n\n case server_hello: /* and HelloRetryRequest */\n ProtocolVersion selected_version;\n };\n } SupportedVersions;\n\nB.3.1.2. Cookie Extension\n\n struct {\n opaque cookie<1..2^16-1>;\n } Cookie;\n\nB.3.1.3. Signature Algorithm Extension\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 126]\n\f\nInternet-Draft TLS March 2018\n\n\n enum {\n /* RSASSA-PKCS1-v1_5 algorithms */\n rsa_pkcs1_sha256(0x0401),\n rsa_pkcs1_sha384(0x0501),\n rsa_pkcs1_sha512(0x0601),\n\n /* ECDSA algorithms */\n ecdsa_secp256r1_sha256(0x0403),\n ecdsa_secp384r1_sha384(0x0503),\n ecdsa_secp521r1_sha512(0x0603),\n\n /* RSASSA-PSS algorithms with public key OID rsaEncryption */\n rsa_pss_rsae_sha256(0x0804),\n rsa_pss_rsae_sha384(0x0805),\n rsa_pss_rsae_sha512(0x0806),\n\n /* EdDSA algorithms */\n ed25519(0x0807),\n ed448(0x0808),\n\n /* RSASSA-PSS algorithms with public key OID RSASSA-PSS */\n rsa_pss_pss_sha256(0x0809),\n rsa_pss_pss_sha384(0x080a),\n rsa_pss_pss_sha512(0x080b),\n\n /* Legacy algorithms */\n rsa_pkcs1_sha1(0x0201),\n ecdsa_sha1(0x0203),\n\n /* Reserved Code Points */\n obsolete_RESERVED(0x0000..0x0200),\n dsa_sha1_RESERVED(0x0202),\n obsolete_RESERVED(0x0204..0x0400),\n dsa_sha256_RESERVED(0x0402),\n obsolete_RESERVED(0x0404..0x0500),\n dsa_sha384_RESERVED(0x0502),\n obsolete_RESERVED(0x0504..0x0600),\n dsa_sha512_RESERVED(0x0602),\n obsolete_RESERVED(0x0604..0x06FF),\n private_use(0xFE00..0xFFFF),\n (0xFFFF)\n } SignatureScheme;\n\n struct {\n SignatureScheme supported_signature_algorithms<2..2^16-2>;\n } SignatureSchemeList;\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 127]\n\f\nInternet-Draft TLS March 2018\n\n\nB.3.1.4. Supported Groups Extension\n\n enum {\n unallocated_RESERVED(0x0000),\n\n /* Elliptic Curve Groups (ECDHE) */\n obsolete_RESERVED(0x0001..0x0016),\n secp256r1(0x0017), secp384r1(0x0018), secp521r1(0x0019),\n obsolete_RESERVED(0x001A..0x001C),\n x25519(0x001D), x448(0x001E),\n\n /* Finite Field Groups (DHE) */\n ffdhe2048(0x0100), ffdhe3072(0x0101), ffdhe4096(0x0102),\n ffdhe6144(0x0103), ffdhe8192(0x0104),\n\n /* Reserved Code Points */\n ffdhe_private_use(0x01FC..0x01FF),\n ecdhe_private_use(0xFE00..0xFEFF),\n obsolete_RESERVED(0xFF01..0xFF02),\n (0xFFFF)\n } NamedGroup;\n\n struct {\n NamedGroup named_group_list<2..2^16-1>;\n } NamedGroupList;\n\n Values within \"obsolete_RESERVED\" ranges are used in previous\n versions of TLS and MUST NOT be offered or negotiated by TLS 1.3\n implementations. The obsolete curves have various known/theoretical\n weaknesses or have had very little usage, in some cases only due to\n unintentional server configuration issues. They are no longer\n considered appropriate for general use and should be assumed to be\n potentially unsafe. The set of curves specified here is sufficient\n for interoperability with all currently deployed and properly\n configured TLS implementations.\n\nB.3.2. Server Parameters Messages\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 128]\n\f\nInternet-Draft TLS March 2018\n\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n DistinguishedName authorities<3..2^16-1>;\n } CertificateAuthoritiesExtension;\n\n struct {\n opaque certificate_extension_oid<1..2^8-1>;\n opaque certificate_extension_values<0..2^16-1>;\n } OIDFilter;\n\n struct {\n OIDFilter filters<0..2^16-1>;\n } OIDFilterExtension;\n\n struct {} PostHandshakeAuth;\n\n struct {\n Extension extensions<0..2^16-1>;\n } EncryptedExtensions;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n Extension extensions<2..2^16-1>;\n } CertificateRequest;\n\nB.3.3. Authentication Messages\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 129]\n\f\nInternet-Draft TLS March 2018\n\n\n /* Managed by IANA */\n enum {\n X509(0),\n OpenPGP_RESERVED(1),\n RawPublicKey(2),\n (255)\n } CertificateType;\n\n struct {\n select (certificate_type) {\n case RawPublicKey:\n /* From RFC 7250 ASN.1_subjectPublicKeyInfo */\n opaque ASN1_subjectPublicKeyInfo<1..2^24-1>;\n\n case X509:\n opaque cert_data<1..2^24-1>;\n };\n Extension extensions<0..2^16-1>;\n } CertificateEntry;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n CertificateEntry certificate_list<0..2^24-1>;\n } Certificate;\n\n struct {\n SignatureScheme algorithm;\n opaque signature<0..2^16-1>;\n } CertificateVerify;\n\n struct {\n opaque verify_data[Hash.length];\n } Finished;\n\nB.3.4. Ticket Establishment\n\n struct {\n uint32 ticket_lifetime;\n uint32 ticket_age_add;\n opaque ticket_nonce<0..255>;\n opaque ticket<1..2^16-1>;\n Extension extensions<0..2^16-2>;\n } NewSessionTicket;\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 130]\n\f\nInternet-Draft TLS March 2018\n\n\nB.3.5. Updating Keys\n\n struct {} EndOfEarlyData;\n\n enum {\n update_not_requested(0), update_requested(1), (255)\n } KeyUpdateRequest;\n\n struct {\n KeyUpdateRequest request_update;\n } KeyUpdate;\n\nB.4. Cipher Suites\n\n A symmetric cipher suite defines the pair of the AEAD algorithm and\n hash algorithm to be used with HKDF. Cipher suite names follow the\n naming convention:\n\n CipherSuite TLS_AEAD_HASH = VALUE;\n\n +-----------+------------------------------------------------+\n | Component | Contents |\n +-----------+------------------------------------------------+\n | TLS | The string \"TLS\" |\n | | |\n | AEAD | The AEAD algorithm used for record protection |\n | | |\n | HASH | The hash algorithm used with HKDF |\n | | |\n | VALUE | The two byte ID assigned for this cipher suite |\n +-----------+------------------------------------------------+\n\n This specification defines the following cipher suites for use with\n TLS 1.3.\n\n +------------------------------+-------------+\n | Description | Value |\n +------------------------------+-------------+\n | TLS_AES_128_GCM_SHA256 | {0x13,0x01} |\n | | |\n | TLS_AES_256_GCM_SHA384 | {0x13,0x02} |\n | | |\n | TLS_CHACHA20_POLY1305_SHA256 | {0x13,0x03} |\n | | |\n | TLS_AES_128_CCM_SHA256 | {0x13,0x04} |\n | | |\n | TLS_AES_128_CCM_8_SHA256 | {0x13,0x05} |\n +------------------------------+-------------+\n\n\n\nRescorla Expires September 21, 2018 [Page 131]\n\f\nInternet-Draft TLS March 2018\n\n\n The corresponding AEAD algorithms AEAD_AES_128_GCM, AEAD_AES_256_GCM,\n and AEAD_AES_128_CCM are defined in [RFC5116].\n AEAD_CHACHA20_POLY1305 is defined in [RFC7539]. AEAD_AES_128_CCM_8\n is defined in [RFC6655]. The corresponding hash algorithms are\n defined in [SHS].\n\n Although TLS 1.3 uses the same cipher suite space as previous\n versions of TLS, TLS 1.3 cipher suites are defined differently, only\n specifying the symmetric ciphers, and cannot be used for TLS 1.2.\n Similarly, TLS 1.2 and lower cipher suites cannot be used with TLS\n 1.3.\n\n New cipher suite values are assigned by IANA as described in\n Section 11.\n\nAppendix C. Implementation Notes\n\n The TLS protocol cannot prevent many common security mistakes. This\n section provides several recommendations to assist implementors.\n [I-D.ietf-tls-tls13-vectors] provides test vectors for TLS 1.3\n handshakes.\n\nC.1. Random Number Generation and Seeding\n\n TLS requires a cryptographically secure pseudorandom number generator\n (CSPRNG). In most cases, the operating system provides an\n appropriate facility such as /dev/urandom, which should be used\n absent other (performance) concerns. It is RECOMMENDED to use an\n existing CSPRNG implementation in preference to crafting a new one.\n Many adequate cryptographic libraries are already available under\n favorable license terms. Should those prove unsatisfactory,\n [RFC4086] provides guidance on the generation of random values.\n\n TLS uses random values both in public protocol fields such as the\n public Random values in the ClientHello and ServerHello and to\n generate keying material. With a properly functioning CSPRNG, this\n does not present a security problem as it is not feasible to\n determine the CSPRNG state from its output. However, with a broken\n CSPRNG, it may be possible for an attacker to use the public output\n to determine the CSPRNG internal state and thereby predict the keying\n material, as documented in [CHECKOWAY]. Implementations can provide\n extra security against this form of attack by using separate CSPRNGs\n to generate public and private values.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 132]\n\f\nInternet-Draft TLS March 2018\n\n\nC.2. Certificates and Authentication\n\n Implementations are responsible for verifying the integrity of\n certificates and should generally support certificate revocation\n messages. Absent a specific indication from an application profile,\n Certificates should always be verified to ensure proper signing by a\n trusted Certificate Authority (CA). The selection and addition of\n trust anchors should be done very carefully. Users should be able to\n view information about the certificate and trust anchor.\n Applications SHOULD also enforce minimum and maximum key sizes. For\n example, certification paths containing keys or signatures weaker\n than 2048-bit RSA or 224-bit ECDSA are not appropriate for secure\n applications.\n\nC.3. Implementation Pitfalls\n\n Implementation experience has shown that certain parts of earlier TLS\n specifications are not easy to understand and have been a source of\n interoperability and security problems. Many of these areas have\n been clarified in this document but this appendix contains a short\n list of the most important things that require special attention from\n implementors.\n\n TLS protocol issues:\n\n - Do you correctly handle handshake messages that are fragmented to\n multiple TLS records (see Section 5.1)? Including corner cases\n like a ClientHello that is split to several small fragments? Do\n you fragment handshake messages that exceed the maximum fragment\n size? In particular, the Certificate and CertificateRequest\n handshake messages can be large enough to require fragmentation.\n\n - Do you ignore the TLS record layer version number in all\n unencrypted TLS records? (see Appendix D)\n\n - Have you ensured that all support for SSL, RC4, EXPORT ciphers,\n and MD5 (via the \"signature_algorithms\" extension) is completely\n removed from all possible configurations that support TLS 1.3 or\n later, and that attempts to use these obsolete capabilities fail\n correctly? (see Appendix D)\n\n - Do you handle TLS extensions in ClientHello correctly, including\n unknown extensions?\n\n - When the server has requested a client certificate, but no\n suitable certificate is available, do you correctly send an empty\n Certificate message, instead of omitting the whole message (see\n Section 4.4.2.3)?\n\n\n\nRescorla Expires September 21, 2018 [Page 133]\n\f\nInternet-Draft TLS March 2018\n\n\n - When processing the plaintext fragment produced by AEAD-Decrypt\n and scanning from the end for the ContentType, do you avoid\n scanning past the start of the cleartext in the event that the\n peer has sent a malformed plaintext of all-zeros?\n\n - Do you properly ignore unrecognized cipher suites (Section 4.1.2),\n hello extensions (Section 4.2), named groups (Section 4.2.7), key\n shares (Section 4.2.8), supported versions (Section 4.2.1), and\n signature algorithms (Section 4.2.3) in the ClientHello?\n\n - As a server, do you send a HelloRetryRequest to clients which\n support a compatible (EC)DHE group but do not predict it in the\n \"key_share\" extension? As a client, do you correctly handle a\n HelloRetryRequest from the server?\n\n Cryptographic details:\n\n - What countermeasures do you use to prevent timing attacks\n [TIMING]?\n\n - When using Diffie-Hellman key exchange, do you correctly preserve\n leading zero bytes in the negotiated key (see Section 7.4.1)?\n\n - Does your TLS client check that the Diffie-Hellman parameters sent\n by the server are acceptable, (see Section 4.2.8.1)?\n\n - Do you use a strong and, most importantly, properly seeded random\n number generator (see Appendix C.1) when generating Diffie-Hellman\n private values, the ECDSA \"k\" parameter, and other security-\n critical values? It is RECOMMENDED that implementations implement\n \"deterministic ECDSA\" as specified in [RFC6979].\n\n - Do you zero-pad Diffie-Hellman public key values to the group size\n (see Section 4.2.8.1)?\n\n - Do you verify signatures after making them to protect against RSA-\n CRT key leaks? [FW15]\n\nC.4. Client Tracking Prevention\n\n Clients SHOULD NOT reuse a ticket for multiple connections. Reuse of\n a ticket allows passive observers to correlate different connections.\n Servers that issue tickets SHOULD offer at least as many tickets as\n the number of connections that a client might use; for example, a web\n browser using HTTP/1.1 [RFC7230] might open six connections to a\n server. Servers SHOULD issue new tickets with every connection.\n This ensures that clients are always able to use a new ticket when\n creating a new connection.\n\n\n\nRescorla Expires September 21, 2018 [Page 134]\n\f\nInternet-Draft TLS March 2018\n\n\nC.5. Unauthenticated Operation\n\n Previous versions of TLS offered explicitly unauthenticated cipher\n suites based on anonymous Diffie-Hellman. These modes have been\n deprecated in TLS 1.3. However, it is still possible to negotiate\n parameters that do not provide verifiable server authentication by\n several methods, including:\n\n - Raw public keys [RFC7250].\n\n - Using a public key contained in a certificate but without\n validation of the certificate chain or any of its contents.\n\n Either technique used alone is vulnerable to man-in-the-middle\n attacks and therefore unsafe for general use. However, it is also\n possible to bind such connections to an external authentication\n mechanism via out-of-band validation of the server's public key,\n trust on first use, or a mechanism such as channel bindings (though\n the channel bindings described in [RFC5929] are not defined for TLS\n 1.3). If no such mechanism is used, then the connection has no\n protection against active man-in-the-middle attack; applications MUST\n NOT use TLS in such a way absent explicit configuration or a specific\n application profile.\n\nAppendix D. Backward Compatibility\n\n The TLS protocol provides a built-in mechanism for version\n negotiation between endpoints potentially supporting different\n versions of TLS.\n\n TLS 1.x and SSL 3.0 use compatible ClientHello messages. Servers can\n also handle clients trying to use future versions of TLS as long as\n the ClientHello format remains compatible and there is at least one\n protocol version supported by both the client and the server.\n\n Prior versions of TLS used the record layer version number\n (TLSPlaintext.legacy_record_version and\n TLSCiphertext.legacy_record_version) for various purposes. As of TLS\n 1.3, this field is deprecated. The value of\n TLSPlaintext.legacy_record_version MUST be ignored by all\n implementations. The value of TLSCiphertext.legacy_record_version is\n included in the additional data for deprotection but MAY otherwise be\n ignored or MAY be validated to match the fixed constant value.\n Version negotiation is performed using only the handshake versions\n (ClientHello.legacy_version, ServerHello.legacy_version, as well as\n the ClientHello, HelloRetryRequest and ServerHello\n \"supported_versions\" extensions). In order to maximize\n interoperability with older endpoints, implementations that negotiate\n\n\n\nRescorla Expires September 21, 2018 [Page 135]\n\f\nInternet-Draft TLS March 2018\n\n\n the use of TLS 1.0-1.2 SHOULD set the record layer version number to\n the negotiated version for the ServerHello and all records\n thereafter.\n\n For maximum compatibility with previously non-standard behavior and\n misconfigured deployments, all implementations SHOULD support\n validation of certification paths based on the expectations in this\n document, even when handling prior TLS versions' handshakes. (see\n Section 4.4.2.2)\n\n TLS 1.2 and prior supported an \"Extended Master Secret\" [RFC7627]\n extension which digested large parts of the handshake transcript into\n the master secret. Because TLS 1.3 always hashes in the transcript\n up to the server CertificateVerify, implementations which support\n both TLS 1.3 and earlier versions SHOULD indicate the use of the\n Extended Master Secret extension in their APIs whenever TLS 1.3 is\n used.\n\nD.1. Negotiating with an older server\n\n A TLS 1.3 client who wishes to negotiate with servers that do not\n support TLS 1.3 will send a normal TLS 1.3 ClientHello containing\n 0x0303 (TLS 1.2) in ClientHello.legacy_version but with the correct\n version(s) in the \"supported_versions\" extension. If the server does\n not support TLS 1.3 it will respond with a ServerHello containing an\n older version number. If the client agrees to use this version, the\n negotiation will proceed as appropriate for the negotiated protocol.\n A client using a ticket for resumption SHOULD initiate the connection\n using the version that was previously negotiated.\n\n Note that 0-RTT data is not compatible with older servers and SHOULD\n NOT be sent absent knowledge that the server supports TLS 1.3. See\n Appendix D.3.\n\n If the version chosen by the server is not supported by the client\n (or not acceptable), the client MUST abort the handshake with a\n \"protocol_version\" alert.\n\n Some legacy server implementations are known to not implement the TLS\n specification properly and might abort connections upon encountering\n TLS extensions or versions which they are not aware of.\n Interoperability with buggy servers is a complex topic beyond the\n scope of this document. Multiple connection attempts may be required\n in order to negotiate a backwards compatible connection; however,\n this practice is vulnerable to downgrade attacks and is NOT\n RECOMMENDED.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 136]\n\f\nInternet-Draft TLS March 2018\n\n\nD.2. Negotiating with an older client\n\n A TLS server can also receive a ClientHello indicating a version\n number smaller than its highest supported version. If the\n \"supported_versions\" extension is present, the server MUST negotiate\n using that extension as described in Section 4.2.1. If the\n \"supported_versions\" extension is not present, the server MUST\n negotiate the minimum of ClientHello.legacy_version and TLS 1.2. For\n example, if the server supports TLS 1.0, 1.1, and 1.2, and\n legacy_version is TLS 1.0, the server will proceed with a TLS 1.0\n ServerHello. If the \"supported_versions\" extension is absent and the\n server only supports versions greater than\n ClientHello.legacy_version, the server MUST abort the handshake with\n a \"protocol_version\" alert.\n\n Note that earlier versions of TLS did not clearly specify the record\n layer version number value in all cases\n (TLSPlaintext.legacy_record_version). Servers will receive various\n TLS 1.x versions in this field, but its value MUST always be ignored.\n\nD.3. 0-RTT backwards compatibility\n\n 0-RTT data is not compatible with older servers. An older server\n will respond to the ClientHello with an older ServerHello, but it\n will not correctly skip the 0-RTT data and will fail to complete the\n handshake. This can cause issues when a client attempts to use\n 0-RTT, particularly against multi-server deployments. For example, a\n deployment could deploy TLS 1.3 gradually with some servers\n implementing TLS 1.3 and some implementing TLS 1.2, or a TLS 1.3\n deployment could be downgraded to TLS 1.2.\n\n A client that attempts to send 0-RTT data MUST fail a connection if\n it receives a ServerHello with TLS 1.2 or older. A client that\n attempts to repair this error SHOULD NOT send a TLS 1.2 ClientHello,\n but instead send a TLS 1.3 ClientHello without 0-RTT data.\n\n To avoid this error condition, multi-server deployments SHOULD ensure\n a uniform and stable deployment of TLS 1.3 without 0-RTT prior to\n enabling 0-RTT.\n\nD.4. Middlebox Compatibility Mode\n\n Field measurements [Ben17a], [Ben17b], [Res17a], [Res17b] have found\n that a significant number of middleboxes misbehave when a TLS client/\n server pair negotiates TLS 1.3. Implementations can increase the\n chance of making connections through those middleboxes by making the\n TLS 1.3 handshake look more like a TLS 1.2 handshake:\n\n\n\n\nRescorla Expires September 21, 2018 [Page 137]\n\f\nInternet-Draft TLS March 2018\n\n\n - The client always provides a non-empty session ID in the\n ClientHello, as described in the legacy_session_id section of\n Section 4.1.2.\n\n - If not offering early data, the client sends a dummy\n change_cipher_spec record (see the third paragraph of Section 5.1)\n immediately before its second flight. This may either be before\n its second ClientHello or before its encrypted handshake flight.\n If offering early data, the record is placed immediately after the\n first ClientHello.\n\n - The server sends a dummy change_cipher_spec record immediately\n after its first handshake message. This may either be after a\n ServerHello or a HelloRetryRequest.\n\n When put together, these changes make the TLS 1.3 handshake resemble\n TLS 1.2 session resumption, which improves the chance of successfully\n connecting through middleboxes. This \"compatibility mode\" is\n partially negotiated: The client can opt to provide a session ID or\n not and the server has to echo it. Either side can send\n change_cipher_spec at any time during the handshake, as they must be\n ignored by the peer, but if the client sends a non-empty session ID,\n the server MUST send the change_cipher_spec as described in this\n section.\n\nD.5. Backwards Compatibility Security Restrictions\n\n Implementations negotiating use of older versions of TLS SHOULD\n prefer forward secret and AEAD cipher suites, when available.\n\n The security of RC4 cipher suites is considered insufficient for the\n reasons cited in [RFC7465]. Implementations MUST NOT offer or\n negotiate RC4 cipher suites for any version of TLS for any reason.\n\n Old versions of TLS permitted the use of very low strength ciphers.\n Ciphers with a strength less than 112 bits MUST NOT be offered or\n negotiated for any version of TLS for any reason.\n\n The security of SSL 3.0 [SSL3] is considered insufficient for the\n reasons enumerated in [RFC7568], and it MUST NOT be negotiated for\n any reason.\n\n The security of SSL 2.0 [SSL2] is considered insufficient for the\n reasons enumerated in [RFC6176], and it MUST NOT be negotiated for\n any reason.\n\n Implementations MUST NOT send an SSL version 2.0 compatible CLIENT-\n HELLO. Implementations MUST NOT negotiate TLS 1.3 or later using an\n\n\n\nRescorla Expires September 21, 2018 [Page 138]\n\f\nInternet-Draft TLS March 2018\n\n\n SSL version 2.0 compatible CLIENT-HELLO. Implementations are NOT\n RECOMMENDED to accept an SSL version 2.0 compatible CLIENT-HELLO in\n order to negotiate older versions of TLS.\n\n Implementations MUST NOT send a ClientHello.legacy_version or\n ServerHello.legacy_version set to 0x0300 or less. Any endpoint\n receiving a Hello message with ClientHello.legacy_version or\n ServerHello.legacy_version set to 0x0300 MUST abort the handshake\n with a \"protocol_version\" alert.\n\n Implementations MUST NOT send any records with a version less than\n 0x0300. Implementations SHOULD NOT accept any records with a version\n less than 0x0300 (but may inadvertently do so if the record version\n number is ignored completely).\n\n Implementations MUST NOT use the Truncated HMAC extension, defined in\n Section 7 of [RFC6066], as it is not applicable to AEAD algorithms\n and has been shown to be insecure in some scenarios.\n\nAppendix E. Overview of Security Properties\n\n A complete security analysis of TLS is outside the scope of this\n document. In this section, we provide an informal description the\n desired properties as well as references to more detailed work in the\n research literature which provides more formal definitions.\n\n We cover properties of the handshake separately from those of the\n record layer.\n\nE.1. Handshake\n\n The TLS handshake is an Authenticated Key Exchange (AKE) protocol\n which is intended to provide both one-way authenticated (server-only)\n and mutually authenticated (client and server) functionality. At the\n completion of the handshake, each side outputs its view of the\n following values:\n\n - A set of \"session keys\" (the various secrets derived from the\n master secret) from which can be derived a set of working keys.\n\n - A set of cryptographic parameters (algorithms, etc.)\n\n - The identities of the communicating parties.\n\n We assume the attacker to be an active network attacker, which means\n it has complete control over the network used to communicate between\n the parties [RFC3552]. Even under these conditions, the handshake\n should provide the properties listed below. Note that these\n\n\n\nRescorla Expires September 21, 2018 [Page 139]\n\f\nInternet-Draft TLS March 2018\n\n\n properties are not necessarily independent, but reflect the protocol\n consumers' needs.\n\n Establishing the same session keys. The handshake needs to output\n the same set of session keys on both sides of the handshake,\n provided that it completes successfully on each endpoint (See\n [CK01]; defn 1, part 1).\n\n Secrecy of the session keys. The shared session keys should be known\n only to the communicating parties and not to the attacker (See\n [CK01]; defn 1, part 2). Note that in a unilaterally\n authenticated connection, the attacker can establish its own\n session keys with the server, but those session keys are distinct\n from those established by the client.\n\n Peer Authentication. The client's view of the peer identity should\n reflect the server's identity. If the client is authenticated,\n the server's view of the peer identity should match the client's\n identity.\n\n Uniqueness of the session keys: Any two distinct handshakes should\n produce distinct, unrelated session keys. Individual session keys\n produced by a handshake should also be distinct and independent.\n\n Downgrade protection. The cryptographic parameters should be the\n same on both sides and should be the same as if the peers had been\n communicating in the absence of an attack (See [BBFKZG16]; defns 8\n and 9}).\n\n Forward secret with respect to long-term keys If the long-term\n keying material (in this case the signature keys in certificate-\n based authentication modes or the external/resumption PSK in PSK\n with (EC)DHE modes) is compromised after the handshake is\n complete, this does not compromise the security of the session key\n (See [DOW92]), as long as the session key itself has been erased.\n The forward secrecy property is not satisfied when PSK is used in\n the \"psk_ke\" PskKeyExchangeMode.\n\n Key Compromise Impersonation (KCI) resistance In a mutually-\n authenticated connection with certificates, compromising the long-\n term secret of one actor should not break that actor's\n authentication of their peer in the given connection (see\n [HGFS15]). For example, if a client's signature key is\n compromised, it should not be possible to impersonate arbitrary\n servers to that client in subsequent handshakes.\n\n Protection of endpoint identities. The server's identity\n (certificate) should be protected against passive attackers. The\n\n\n\nRescorla Expires September 21, 2018 [Page 140]\n\f\nInternet-Draft TLS March 2018\n\n\n client's identity should be protected against both passive and\n active attackers.\n\n Informally, the signature-based modes of TLS 1.3 provide for the\n establishment of a unique, secret, shared key established by an\n (EC)DHE key exchange and authenticated by the server's signature over\n the handshake transcript, as well as tied to the server's identity by\n a MAC. If the client is authenticated by a certificate, it also\n signs over the handshake transcript and provides a MAC tied to both\n identities. [SIGMA] describes the design and analysis of this type\n of key exchange protocol. If fresh (EC)DHE keys are used for each\n connection, then the output keys are forward secret.\n\n The external PSK and resumption PSK bootstrap from a long-term shared\n secret into a unique per-connection set of short-term session keys.\n This secret may have been established in a previous handshake. If\n PSK with (EC)DHE key establishment is used, these session keys will\n also be forward secret. The resumption PSK has been designed so that\n the resumption master secret computed by connection N and needed to\n form connection N+1 is separate from the traffic keys used by\n connection N, thus providing forward secrecy between the connections.\n In addition, if multiple tickets are established on the same\n connection, they are associated with different keys, so compromise of\n the PSK associated with one ticket does not lead to the compromise of\n connections established with PSKs associated with other tickets.\n This property is most interesting if tickets are stored in a database\n (and so can be deleted) rather than if they are self-encrypted.\n\n The PSK binder value forms a binding between a PSK and the current\n handshake, as well as between the session where the PSK was\n established and the current session. This binding transitively\n includes the original handshake transcript, because that transcript\n is digested into the values which produce the Resumption Master\n Secret. This requires that both the KDF used to produce the\n resumption master secret and the MAC used to compute the binder be\n collision resistant. See Appendix E.1.1 for more on this. Note: The\n binder does not cover the binder values from other PSKs, though they\n are included in the Finished MAC.\n\n Note: TLS does not currently permit the server to send a\n certificate_request message in non-certificate-based handshakes\n (e.g., PSK). If this restriction were to be relaxed in future, the\n client's signature would not cover the server's certificate directly.\n However, if the PSK was established through a NewSessionTicket, the\n client's signature would transitively cover the server's certificate\n through the PSK binder. [PSK-FINISHED] describes a concrete attack\n on constructions that do not bind to the server's certificate (see\n also [Kraw16]). It is unsafe to use certificate-based client\n\n\n\nRescorla Expires September 21, 2018 [Page 141]\n\f\nInternet-Draft TLS March 2018\n\n\n authentication when the client might potentially share the same PSK/\n key-id pair with two different endpoints. Implementations MUST NOT\n combine external PSKs with certificate-based authentication of either\n the client or the server unless negotiated by some extension.\n\n If an exporter is used, then it produces values which are unique and\n secret (because they are generated from a unique session key).\n Exporters computed with different labels and contexts are\n computationally independent, so it is not feasible to compute one\n from another or the session secret from the exported value. Note:\n exporters can produce arbitrary-length values. If exporters are to\n be used as channel bindings, the exported value MUST be large enough\n to provide collision resistance. The exporters provided in TLS 1.3\n are derived from the same handshake contexts as the early traffic\n keys and the application traffic keys respectively, and thus have\n similar security properties. Note that they do not include the\n client's certificate; future applications which wish to bind to the\n client's certificate may need to define a new exporter that includes\n the full handshake transcript.\n\n For all handshake modes, the Finished MAC (and where present, the\n signature), prevents downgrade attacks. In addition, the use of\n certain bytes in the random nonces as described in Section 4.1.3\n allows the detection of downgrade to previous TLS versions. See\n [BBFKZG16] for more detail on TLS 1.3 and downgrade.\n\n As soon as the client and the server have exchanged enough\n information to establish shared keys, the remainder of the handshake\n is encrypted, thus providing protection against passive attackers,\n even if the computed shared key is not authenticated. Because the\n server authenticates before the client, the client can ensure that if\n it authenticates to the server, it only reveals its identity to an\n authenticated server. Note that implementations must use the\n provided record padding mechanism during the handshake to avoid\n leaking information about the identities due to length. The client's\n proposed PSK identities are not encrypted, nor is the one that the\n server selects.\n\nE.1.1. Key Derivation and HKDF\n\n Key derivation in TLS 1.3 uses the HKDF function defined in [RFC5869]\n and its two components, HKDF-Extract and HKDF-Expand. The full\n rationale for the HKDF construction can be found in [Kraw10] and the\n rationale for the way it is used in TLS 1.3 in [KW16]. Throughout\n this document, each application of HKDF-Extract is followed by one or\n more invocations of HKDF-Expand. This ordering should always be\n followed (including in future revisions of this document), in\n particular, one SHOULD NOT use an output of HKDF-Extract as an input\n\n\n\nRescorla Expires September 21, 2018 [Page 142]\n\f\nInternet-Draft TLS March 2018\n\n\n to another application of HKDF-Extract without an HKDF-Expand in\n between. Consecutive applications of HKDF-Expand are allowed as long\n as these are differentiated via the key and/or the labels.\n\n Note that HKDF-Expand implements a pseudorandom function (PRF) with\n both inputs and outputs of variable length. In some of the uses of\n HKDF in this document (e.g., for generating exporters and the\n resumption_master_secret), it is necessary that the application of\n HKDF-Expand be collision-resistant, namely, it should be infeasible\n to find two different inputs to HKDF-Expand that output the same\n value. This requires the underlying hash function to be collision\n resistant and the output length from HKDF-Expand to be of size at\n least 256 bits (or as much as needed for the hash function to prevent\n finding collisions).\n\nE.1.2. Client Authentication\n\n A client that has sent authentication data to a server, either during\n the handshake or in post-handshake authentication, cannot be sure if\n the server afterwards considers the client to be authenticated or\n not. If the client needs to determine if the server considers the\n connection to be unilaterally or mutually authenticated, this has to\n be provisioned by the application layer. See [CHHSV17] for details.\n In addition, the analysis of post-handshake authentication from\n [Kraw16] shows that the client identified by the certificate sent in\n the post-handshake phase possesses the traffic key. This party is\n therefore the client that participated in the original handshake or\n one to whom the original client delegated the traffic key (assuming\n that the traffic key has not been compromised).\n\nE.1.3. 0-RTT\n\n The 0-RTT mode of operation generally provides similar security\n properties as 1-RTT data, with the two exceptions that the 0-RTT\n encryption keys do not provide full forward secrecy and that the\n server is not able to guarantee uniqueness of the handshake (non-\n replayability) without keeping potentially undue amounts of state.\n See Section 8 for mechanisms to limit the exposure to replay.\n\nE.1.4. Exporter Independence\n\n The exporter_master_secret and early_exporter_master_secret are\n derived to be independent of the traffic keys and therefore do not\n represent a threat to the security of traffic encrypted with those\n keys. However, because these secrets can be used to compute any\n exporter value, they SHOULD be erased as soon as possible. If the\n total set of exporter labels is known, then implementations SHOULD\n pre-compute the inner Derive-Secret stage of the exporter computation\n\n\n\nRescorla Expires September 21, 2018 [Page 143]\n\f\nInternet-Draft TLS March 2018\n\n\n for all those labels, then erase the [early_]exporter_master_secret,\n followed by each inner values as soon as it is known that it will not\n be needed again.\n\nE.1.5. Post-Compromise Security\n\n TLS does not provide security for handshakes which take place after\n the peer's long-term secret (signature key or external PSK) is\n compromised. It therefore does not provide post-compromise security\n [CCG16], sometimes also referred to as backwards or future secrecy.\n This is in contrast to KCI resistance, which describes the security\n guarantees that a party has after its own long-term secret has been\n compromised.\n\nE.1.6. External References\n\n The reader should refer to the following references for analysis of\n the TLS handshake: [DFGS15] [CHSV16] [DFGS16] [KW16] [Kraw16]\n [FGSW16] [LXZFH16] [FG17] [BBK17].\n\nE.2. Record Layer\n\n The record layer depends on the handshake producing strong traffic\n secrets which can be used to derive bidirectional encryption keys and\n nonces. Assuming that is true, and the keys are used for no more\n data than indicated in Section 5.5 then the record layer should\n provide the following guarantees:\n\n Confidentiality. An attacker should not be able to determine the\n plaintext contents of a given record.\n\n Integrity. An attacker should not be able to craft a new record\n which is different from an existing record which will be accepted\n by the receiver.\n\n Order protection/non-replayability An attacker should not be able to\n cause the receiver to accept a record which it has already\n accepted or cause the receiver to accept record N+1 without having\n first processed record N.\n\n Length concealment. Given a record with a given external length, the\n attacker should not be able to determine the amount of the record\n that is content versus padding.\n\n Forward secrecy after key change. If the traffic key update\n mechanism described in Section 4.6.3 has been used and the\n previous generation key is deleted, an attacker who compromises\n\n\n\n\nRescorla Expires September 21, 2018 [Page 144]\n\f\nInternet-Draft TLS March 2018\n\n\n the endpoint should not be able to decrypt traffic encrypted with\n the old key.\n\n Informally, TLS 1.3 provides these properties by AEAD-protecting the\n plaintext with a strong key. AEAD encryption [RFC5116] provides\n confidentiality and integrity for the data. Non-replayability is\n provided by using a separate nonce for each record, with the nonce\n being derived from the record sequence number (Section 5.3), with the\n sequence number being maintained independently at both sides thus\n records which are delivered out of order result in AEAD deprotection\n failures. In order to prevent mass cryptanalysis when the same\n plaintext is repeatedly encrypted by different users under the same\n key (as is commonly the case for HTTP), the nonce is formed by mixing\n the sequence number with a secret per-connection initialization\n vector derived along with the traffic keys. See [BT16] for analysis\n of this construction.\n\n The re-keying technique in TLS 1.3 (see Section 7.2) follows the\n construction of the serial generator in [REKEY], which shows that re-\n keying can allow keys to be used for a larger number of encryptions\n than without re-keying. This relies on the security of the HKDF-\n Expand-Label function as a pseudorandom function (PRF). In addition,\n as long as this function is truly one way, it is not possible to\n compute traffic keys from prior to a key change (forward secrecy).\n\n TLS does not provide security for data which is communicated on a\n connection after a traffic secret of that connection is compromised.\n That is, TLS does not provide post-compromise security/future\n secrecy/backward secrecy with respect to the traffic secret. Indeed,\n an attacker who learns a traffic secret can compute all future\n traffic secrets on that connection. Systems which want such\n guarantees need to do a fresh handshake and establish a new\n connection with an (EC)DHE exchange.\n\nE.2.1. External References\n\n The reader should refer to the following references for analysis of\n the TLS record layer: [BMMT15] [BT16] [BDFKPPRSZZ16] [BBK17]\n [Anon18].\n\nE.3. Traffic Analysis\n\n TLS is susceptible to a variety of traffic analysis attacks based on\n observing the length and timing of encrypted packets [CLINIC]\n [HCJ16]. This is particularly easy when there is a small set of\n possible messages to be distinguished, such as for a video server\n hosting a fixed corpus of content, but still provides usable\n information even in more complicated scenarios.\n\n\n\nRescorla Expires September 21, 2018 [Page 145]\n\f\nInternet-Draft TLS March 2018\n\n\n TLS does not provide any specific defenses against this form of\n attack but does include a padding mechanism for use by applications:\n The plaintext protected by the AEAD function consists of content plus\n variable-length padding, which allows the application to produce\n arbitrary length encrypted records as well as padding-only cover\n traffic to conceal the difference between periods of transmission and\n periods of silence. Because the padding is encrypted alongside the\n actual content, an attacker cannot directly determine the length of\n the padding, but may be able to measure it indirectly by the use of\n timing channels exposed during record processing (i.e., seeing how\n long it takes to process a record or trickling in records to see\n which ones elicit a response from the server). In general, it is not\n known how to remove all of these channels because even a constant\n time padding removal function will likely feed the content into data-\n dependent functions. At minimum, a fully constant time server or\n client would require close cooperation with the application layer\n protocol implementation, including making that higher level protocol\n constant time.\n\n Note: Robust traffic analysis defences will likely lead to inferior\n performance due to delay in transmitting packets and increased\n traffic volume.\n\nE.4. Side Channel Attacks\n\n In general, TLS does not have specific defenses against side-channel\n attacks (i.e., those which attack the communications via secondary\n channels such as timing) leaving those to the implementation of the\n relevant cryptographic primitives. However, certain features of TLS\n are designed to make it easier to write side-channel resistant code:\n\n - Unlike previous versions of TLS which used a composite MAC-then-\n encrypt structure, TLS 1.3 only uses AEAD algorithms, allowing\n implementations to use self-contained constant-time\n implementations of those primitives.\n\n - TLS uses a uniform \"bad_record_mac\" alert for all decryption\n errors, which is intended to prevent an attacker from gaining\n piecewise insight into portions of the message. Additional\n resistance is provided by terminating the connection on such\n errors; a new connection will have different cryptographic\n material, preventing attacks against the cryptographic primitives\n that require multiple trials.\n\n Information leakage through side channels can occur at layers above\n TLS, in application protocols and the applications that use them.\n Resistance to side-channel attacks depends on applications and\n\n\n\n\nRescorla Expires September 21, 2018 [Page 146]\n\f\nInternet-Draft TLS March 2018\n\n\n application protocols separately ensuring that confidential\n information is not inadvertently leaked.\n\nE.5. Replay Attacks on 0-RTT\n\n Replayable 0-RTT data presents a number of security threats to TLS-\n using applications, unless those applications are specifically\n engineered to be safe under replay (minimally, this means idempotent,\n but in many cases may also require other stronger conditions, such as\n constant-time response). Potential attacks include:\n\n - Duplication of actions which cause side effects (e.g., purchasing\n an item or transferring money) to be duplicated, thus harming the\n site or the user.\n\n - Attackers can store and replay 0-RTT messages in order to re-order\n them with respect to other messages (e.g., moving a delete to\n after a create).\n\n - Exploiting cache timing behavior to discover the content of 0-RTT\n messages by replaying a 0-RTT message to a different cache node\n and then using a separate connection to measure request latency,\n to see if the two requests address the same resource.\n\n If data can be replayed a large number of times, additional attacks\n become possible, such as making repeated measurements of the speed of\n cryptographic operations. In addition, they may be able to overload\n rate-limiting systems. For further description of these attacks, see\n [Mac17].\n\n Ultimately, servers have the responsibility to protect themselves\n against attacks employing 0-RTT data replication. The mechanisms\n described in Section 8 are intended to prevent replay at the TLS\n layer but do not provide complete protection against receiving\n multiple copies of client data. TLS 1.3 falls back to the 1-RTT\n handshake when the server does not have any information about the\n client, e.g., because it is in a different cluster which does not\n share state or because the ticket has been deleted as described in\n Section 8.1. If the application layer protocol retransmits data in\n this setting, then it is possible for an attacker to induce message\n duplication by sending the ClientHello to both the original cluster\n (which processes the data immediately) and another cluster which will\n fall back to 1-RTT and process the data upon application layer\n replay. The scale of this attack is limited by the client's\n willingness to retry transactions and therefore only allows a limited\n amount of duplication, with each copy appearing as a new connection\n at the server.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 147]\n\f\nInternet-Draft TLS March 2018\n\n\n If implemented correctly, the mechanisms described in Section 8.1 and\n Section 8.2 prevent a replayed ClientHello and its associated 0-RTT\n data from being accepted multiple times by any cluster with\n consistent state; for servers which limit the use of 0-RTT to one\n cluster for a single ticket, then a given ClientHello and its\n associated 0-RTT data will only be accepted once. However, if state\n is not completely consistent, then an attacker might be able to have\n multiple copies of the data be accepted during the replication\n window. Because clients do not know the exact details of server\n behavior, they MUST NOT send messages in early data which are not\n safe to have replayed and which they would not be willing to retry\n across multiple 1-RTT connections.\n\n Application protocols MUST NOT use 0-RTT data without a profile that\n defines its use. That profile needs to identify which messages or\n interactions are safe to use with 0-RTT and how to handle the\n situation when the server rejects 0-RTT and falls back to 1-RTT.\n\n In addition, to avoid accidental misuse, TLS implementations MUST NOT\n enable 0-RTT (either sending or accepting) unless specifically\n requested by the application and MUST NOT automatically resend 0-RTT\n data if it is rejected by the server unless instructed by the\n application. Server-side applications may wish to implement special\n processing for 0-RTT data for some kinds of application traffic\n (e.g., abort the connection, request that data be resent at the\n application layer, or delay processing until the handshake\n completes). In order to allow applications to implement this kind of\n processing, TLS implementations MUST provide a way for the\n application to determine if the handshake has completed.\n\nE.5.1. Replay and Exporters\n\n Replays of the ClientHello produce the same early exporter, thus\n requiring additional care by applications which use these exporters.\n In particular, if these exporters are used as an authentication\n channel binding (e.g., by signing the output of the exporter) an\n attacker who compromises the PSK can transplant authenticators\n between connections without compromising the authentication key.\n\n In addition, the early exporter SHOULD NOT be used to generate\n server-to-client encryption keys because that would entail the reuse\n of those keys. This parallels the use of the early application\n traffic keys only in the client-to-server direction.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 148]\n\f\nInternet-Draft TLS March 2018\n\n\nE.6. PSK Identity Exposure\n\n Because implementations respond to an invalid PSK binder by aborting\n the handshake, it may be possible for an attacker to verify whether a\n given PSK identity is valid. Specifically, if a server accepts both\n external PSK and certificate-based handshakes, a valid PSK identity\n will result in a failed handshake, whereas an invalid identity will\n just be skipped and result in a successful certificate handshake.\n Servers which solely support PSK handshakes may be able to resist\n this form of attack by treating the cases where there is no valid PSK\n identity and where there is an identity but it has an invalid binder\n identically.\n\nE.7. Attacks on Static RSA\n\n Although TLS 1.3 does not use RSA key transport and so is not\n directly susceptible to Bleichenbacher-type attacks, if TLS 1.3\n servers also support static RSA in the context of previous versions\n of TLS, then it may be possible to impersonate the server for TLS 1.3\n connections [JSS15]. TLS 1.3 implementations can prevent this attack\n by disabling support for static RSA across all versions of TLS. In\n principle, implementations might also be able to separate\n certificates with different keyUsage bits for static RSA decryption\n and RSA signature, but this technique relies on clients refusing to\n accept signatures using keys in certificates that do not have the\n digitalSignature bit set, and many clients do not enforce this\n restriction.\n\nAppendix F. Working Group Information\n\n The discussion list for the IETF TLS working group is located at the\n e-mail address [redacted-email] [1]. Information on the group and\n information on how to subscribe to the list is at\n https://www.ietf.org/mailman/listinfo/tls\n\n Archives of the list can be found at: https://www.ietf.org/mail-\n archive/web/tls/current/index.html\n\nAppendix G. Contributors\n\n - Martin Abadi\n University of California, Santa Cruz\n [redacted-email]\n\n - Christopher Allen (co-editor of TLS 1.0)\n Alacrity Ventures\n [redacted-email]\n\n\n\n\nRescorla Expires September 21, 2018 [Page 149]\n\f\nInternet-Draft TLS March 2018\n\n\n - Richard Barnes\n Cisco\n [redacted-email]\n\n - Steven M. Bellovin\n Columbia University\n [redacted-email]\n\n - David Benjamin\n Google\n [redacted-email]\n\n - Benjamin Beurdouche\n INRIA & Microsoft Research\n [redacted-email]\n\n - Karthikeyan Bhargavan (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n - Simon Blake-Wilson (co-author of [RFC4492])\n BCI\n [redacted-email]\n\n - Nelson Bolyard (co-author of [RFC4492])\n Sun Microsystems, Inc.\n [redacted-email]\n\n - Ran Canetti\n IBM\n [redacted-email]\n\n - Matt Caswell\n OpenSSL\n [redacted-email]\n\n - Stephen Checkoway\n University of Illinois at Chicago\n [redacted-email]\n\n - Pete Chown\n Skygate Technology Ltd\n [redacted-email]\n\n - Katriel Cohn-Gordon\n University of Oxford\n [redacted-email]\n\n\n\n\nRescorla Expires September 21, 2018 [Page 150]\n\f\nInternet-Draft TLS March 2018\n\n\n - Cas Cremers\n University of Oxford\n [redacted-email]\n\n - Antoine Delignat-Lavaud (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n - Tim Dierks (co-editor of TLS 1.0, 1.1, and 1.2)\n Independent\n [redacted-email]\n\n - Roelof DuToit\n Symantec Corporation\n [redacted-email]\n\n - Taher Elgamal\n Securify\n [redacted-email]\n\n - Pasi Eronen\n Nokia\n [redacted-email]\n\n - Cedric Fournet\n Microsoft\n [redacted-email]\n\n - Anil Gangolli\n [redacted-email]\n\n - David M. Garrett\n [redacted-email]\n\n - Illya Gerasymchuk\n Independent\n [redacted-email]\n\n - Alessandro Ghedini\n Cloudflare Inc.\n [redacted-email]\n\n - Daniel Kahn Gillmor\n ACLU\n [redacted-email]\n\n - Matthew Green\n Johns Hopkins University\n\n\n\nRescorla Expires September 21, 2018 [Page 151]\n\f\nInternet-Draft TLS March 2018\n\n\n [redacted-email]\n\n - Jens Guballa\n ETAS\n [redacted-email]\n\n - Felix Guenther\n TU Darmstadt\n [redacted-email]\n\n - Vipul Gupta (co-author of [RFC4492])\n Sun Microsystems Laboratories\n [redacted-email]\n\n - Chris Hawk (co-author of [RFC4492])\n Corriente Networks LLC\n [redacted-email]\n\n - Kipp Hickman\n\n - Alfred Hoenes\n\n - David Hopwood\n Independent Consultant\n [redacted-email]\n\n - Marko Horvat\n MPI-SWS\n [redacted-email]\n\n - Jonathan Hoyland\n Royal Holloway, University of London [redacted-email]\n\n - Subodh Iyengar\n Facebook\n [redacted-email]\n\n - Benjamin Kaduk\n Akamai\n [redacted-email]\n\n - Hubert Kario\n Red Hat Inc.\n [redacted-email]\n\n - Phil Karlton (co-author of SSL 3.0)\n\n - Leon Klingele\n\n\n\nRescorla Expires September 21, 2018 [Page 152]\n\f\nInternet-Draft TLS March 2018\n\n\n Independent\n [redacted-email]\n\n - Paul Kocher (co-author of SSL 3.0)\n Cryptography Research\n [redacted-email]\n\n - Hugo Krawczyk\n IBM\n [redacted-email]\n\n - Adam Langley (co-author of [RFC7627])\n Google\n [redacted-email]\n\n - Olivier Levillain\n ANSSI\n [redacted-email]\n\n - Xiaoyin Liu\n University of North Carolina at Chapel Hill\n [redacted-email]\n\n - Ilari Liusvaara\n Independent\n [redacted-email]\n\n - Atul Luykx\n K.U. Leuven\n [redacted-email]\n\n - Colm MacCarthaigh\n Amazon Web Services\n [redacted-email]\n\n - Carl Mehner\n USAA\n [redacted-email]\n\n - Jan Mikkelsen\n Transactionware\n [redacted-email]\n\n - Bodo Moeller (co-author of [RFC4492])\n Google\n [redacted-email]\n\n - Kyle Nekritz\n\n\n\nRescorla Expires September 21, 2018 [Page 153]\n\f\nInternet-Draft TLS March 2018\n\n\n Facebook\n [redacted-email]\n\n - Erik Nygren\n Akamai Technologies\n [redacted-email]\n\n - Magnus Nystrom\n Microsoft\n [redacted-email]\n\n - Kazuho Oku\n DeNA Co., Ltd.\n [redacted-email]\n\n - Kenny Paterson\n Royal Holloway, University of London\n [redacted-email]\n\n - Alfredo Pironti (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n - Andrei Popov\n Microsoft\n [redacted-email]\n\n - Marsh Ray (co-author of [RFC7627])\n Microsoft\n [redacted-email]\n\n - Robert Relyea\n Netscape Communications\n [redacted-email]\n\n - Kyle Rose\n Akamai Technologies\n [redacted-email]\n\n - Jim Roskind\n Amazon\n [redacted-email]\n\n - Michael Sabin\n\n - Joe Salowey\n Tableau Software\n [redacted-email]\n\n\n\nRescorla Expires September 21, 2018 [Page 154]\n\f\nInternet-Draft TLS March 2018\n\n\n - Rich Salz\n Akamai\n [redacted-email]\n\n - David Schinazi\n Apple Inc.\n [redacted-email]\n\n - Sam Scott\n Royal Holloway, University of London\n [redacted-email]\n\n - Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n - Brian Smith\n Independent\n [redacted-email]\n\n - Brian Sniffen\n Akamai Technologies\n [redacted-email]\n\n - Nick Sullivan\n Cloudflare Inc.\n [redacted-email]\n\n - Bjoern Tackmann\n University of California, San Diego\n [redacted-email]\n\n - Tim Taubert\n Mozilla\n [redacted-email]\n\n - Martin Thomson\n Mozilla\n [redacted-email]\n\n - Sean Turner\n sn3rd\n [redacted-email]\n\n - Steven Valdez\n Google\n [redacted-email]\n\n\n\n\nRescorla Expires September 21, 2018 [Page 155]\n\f\nInternet-Draft TLS March 2018\n\n\n - Filippo Valsorda\n Cloudflare Inc.\n [redacted-email]\n\n - Thyla van der Merwe\n Royal Holloway, University of London\n [redacted-email]\n\n - Victor Vasiliev\n Google\n [redacted-email]\n\n - Tom Weinstein\n\n - Hoeteck Wee\n Ecole Normale Superieure, Paris\n [redacted-email]\n\n - David Wong\n NCC Group\n [redacted-email]\n\n - Christopher A. Wood\n Apple Inc.\n [redacted-email]\n\n - Tim Wright\n Vodafone\n [redacted-email]\n\n - Peter Wu\n Independent\n [redacted-email]\n\n - Kazu Yamamoto\n Internet Initiative Japan Inc.\n [redacted-email]\n\nAuthor's Address\n\n Eric Rescorla\n RTFM, Inc.\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 156]", "text_sha256": "a7cda7524edf08714e498aaf4b2a684ee1b5fdf0c9648505eb54649b7cfb0b92"}, {"draft_references": ["draft-ietf-tls-tls13-vectors-06"], "excluded_rfc_relation_targets": [{"kind": "obsoletes", "reason": "not_requested", "rfc_number": 2018}], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "8446"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_id": "obsoletes:5077", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "5077"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_id": "obsoletes:5246", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "5246"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_id": "obsoletes:6961", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "6961"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_id": "obsoletes:2018", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "2018"}, {"char_end": 238, "char_start": 219, "evidence_quote": "Updates: 5705, 6066", "fact_id": "updates:5705", "field": "updates", "raw_char_end": 244, "raw_char_start": 225, "value": "5705"}, {"char_end": 238, "char_start": 219, "evidence_quote": "Updates: 5705, 6066", "fact_id": "updates:6066", "field": "updates", "raw_char_end": 244, "raw_char_start": 225, "value": "6066"}, {"char_end": 261161, "char_start": 261130, "evidence_quote": "draft-ietf-tls-tls13-vectors-06", "fact_id": "draft_reference:draft-ietf-tls-tls13-vectors-06", "field": "draft_reference", "raw_char_end": 261167, "raw_char_start": 261136, "value": "draft-ietf-tls-tls13-vectors-06"}], "kind": "rfc", "occurred_at": "2018-08-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 86, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd", "record_id": "ietf:rfc:8446", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "rfc_number": 8446, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd", "source_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "temporal_semantics": "source_event_time", "text": "Internet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 8446 Mozilla\nObsoletes: 5077, 5246, 6961 August 2018\nUpdates: 5705, 6066\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS) Protocol Version 1.3\n\nAbstract\n\n This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol. TLS allows client/server applications to communicate\n over the Internet in a way that is designed to prevent eavesdropping,\n tampering, and message forgery.\n\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077,\n 5246, and 6961. This document also specifies new requirements for\n TLS 1.2 implementations.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 7841.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n https://www.rfc-editor.org/info/rfc8446.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 8446 TLS August 2018\n\n\nCopyright Notice\n\n Copyright (c) 2018 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (https://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 8446 TLS August 2018\n\n\nTable of Contents\n\n 1. Introduction ....................................................6\n 1.1. Conventions and Terminology ................................7\n 1.2. Major Differences from TLS 1.2 .............................8\n 1.3. Updates Affecting TLS 1.2 ..................................9\n 2. Protocol Overview ..............................................10\n 2.1. Incorrect DHE Share .......................................14\n 2.2. Resumption and Pre-Shared Key (PSK) .......................15\n 2.3. 0-RTT Data ................................................17\n 3. Presentation Language ..........................................19\n 3.1. Basic Block Size ..........................................19\n 3.2. Miscellaneous .............................................20\n 3.3. Numbers ...................................................20\n 3.4. Vectors ...................................................20\n 3.5. Enumerateds ...............................................21\n 3.6. Constructed Types .........................................22\n 3.7. Constants .................................................23\n 3.8. Variants ..................................................23\n 4. Handshake Protocol .............................................24\n 4.1. Key Exchange Messages .....................................25\n 4.1.1. Cryptographic Negotiation ..........................26\n 4.1.2. Client Hello .......................................27\n 4.1.3. Server Hello .......................................31\n 4.1.4. Hello Retry Request ................................33\n 4.2. Extensions ................................................35\n 4.2.1. Supported Versions .................................39\n 4.2.2. Cookie .............................................40\n 4.2.3. Signature Algorithms ...............................41\n 4.2.4. Certificate Authorities ............................45\n 4.2.5. OID Filters ........................................45\n 4.2.6. Post-Handshake Client Authentication ...............47\n 4.2.7. Supported Groups ...................................47\n 4.2.8. Key Share ..........................................48\n 4.2.9. Pre-Shared Key Exchange Modes ......................51\n 4.2.10. Early Data Indication .............................52\n 4.2.11. Pre-Shared Key Extension ..........................55\n 4.3. Server Parameters .........................................59\n 4.3.1. Encrypted Extensions ...............................60\n 4.3.2. Certificate Request ................................60\n 4.4. Authentication Messages ...................................61\n 4.4.1. The Transcript Hash ................................63\n 4.4.2. Certificate ........................................64\n 4.4.3. Certificate Verify .................................69\n 4.4.4. Finished ...........................................71\n 4.5. End of Early Data .........................................72\n\n\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 8446 TLS August 2018\n\n\n 4.6. Post-Handshake Messages ...................................73\n 4.6.1. New Session Ticket Message .........................73\n 4.6.2. Post-Handshake Authentication ......................75\n 4.6.3. Key and Initialization Vector Update ...............76\n 5. Record Protocol ................................................77\n 5.1. Record Layer ..............................................78\n 5.2. Record Payload Protection .................................80\n 5.3. Per-Record Nonce ..........................................82\n 5.4. Record Padding ............................................83\n 5.5. Limits on Key Usage .......................................84\n 6. Alert Protocol .................................................85\n 6.1. Closure Alerts ............................................87\n 6.2. Error Alerts ..............................................88\n 7. Cryptographic Computations .....................................90\n 7.1. Key Schedule ..............................................91\n 7.2. Updating Traffic Secrets ..................................94\n 7.3. Traffic Key Calculation ...................................95\n 7.4. (EC)DHE Shared Secret Calculation .........................95\n 7.4.1. Finite Field Diffie-Hellman ........................95\n 7.4.2. Elliptic Curve Diffie-Hellman ......................96\n 7.5. Exporters .................................................97\n 8. 0-RTT and Anti-Replay ..........................................98\n 8.1. Single-Use Tickets ........................................99\n 8.2. Client Hello Recording ....................................99\n 8.3. Freshness Checks .........................................101\n 9. Compliance Requirements .......................................102\n 9.1. Mandatory-to-Implement Cipher Suites .....................102\n 9.2. Mandatory-to-Implement Extensions ........................103\n 9.3. Protocol Invariants ......................................104\n 10. Security Considerations ......................................106\n 11. IANA Considerations ..........................................106\n 12. References ...................................................109\n 12.1. Normative References ....................................109\n 12.2. Informative References ..................................112\n Appendix A. State Machine ........................................120\n A.1. Client ....................................................120\n A.2. Server ....................................................121\n Appendix B. Protocol Data Structures and Constant Values .........122\n B.1. Record Layer ..............................................122\n B.2. Alert Messages ............................................123\n B.3. Handshake Protocol ........................................124\n B.3.1. Key Exchange Messages .................................125\n B.3.2. Server Parameters Messages ............................131\n B.3.3. Authentication Messages ...............................132\n B.3.4. Ticket Establishment ..................................132\n B.3.5. Updating Keys .........................................133\n B.4. Cipher Suites .............................................133\n\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 8446 TLS August 2018\n\n\n Appendix C. Implementation Notes .................................134\n C.1. Random Number Generation and Seeding ......................134\n C.2. Certificates and Authentication ...........................135\n C.3. Implementation Pitfalls ...................................135\n C.4. Client Tracking Prevention ................................137\n C.5. Unauthenticated Operation .................................137\n Appendix D. Backward Compatibility ...............................138\n D.1. Negotiating with an Older Server ..........................139\n D.2. Negotiating with an Older Client ..........................139\n D.3. 0-RTT Backward Compatibility ..............................140\n D.4. Middlebox Compatibility Mode ..............................140\n D.5. Security Restrictions Related to Backward Compatibility ...141\n Appendix E. Overview of Security Properties ......................142\n E.1. Handshake .................................................142\n E.1.1. Key Derivation and HKDF ...............................145\n E.1.2. Client Authentication .................................146\n E.1.3. 0-RTT .................................................146\n E.1.4. Exporter Independence .................................146\n E.1.5. Post-Compromise Security ..............................146\n E.1.6. External References ...................................147\n E.2. Record Layer ..............................................147\n E.2.1. External References ...................................148\n E.3. Traffic Analysis ..........................................148\n E.4. Side-Channel Attacks ......................................149\n E.5. Replay Attacks on 0-RTT ...................................150\n E.5.1. Replay and Exporters ..................................151\n E.6. PSK Identity Exposure .....................................152\n E.7. Sharing PSKs ..............................................152\n E.8. Attacks on Static RSA .....................................152\n Contributors .....................................................153\n Author's Address .................................................160\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 8446 TLS August 2018\n\n\n1. Introduction\n\n The primary goal of TLS is to provide a secure channel between two\n communicating peers; the only requirement from the underlying\n transport is a reliable, in-order data stream. Specifically, the\n secure channel should provide the following properties:\n\n - Authentication: The server side of the channel is always\n authenticated; the client side is optionally authenticated.\n Authentication can happen via asymmetric cryptography (e.g., RSA\n [RSA], the Elliptic Curve Digital Signature Algorithm (ECDSA)\n [ECDSA], or the Edwards-Curve Digital Signature Algorithm (EdDSA)\n [RFC8032]) or a symmetric pre-shared key (PSK).\n\n - Confidentiality: Data sent over the channel after establishment is\n only visible to the endpoints. TLS does not hide the length of\n the data it transmits, though endpoints are able to pad TLS\n records in order to obscure lengths and improve protection against\n traffic analysis techniques.\n\n - Integrity: Data sent over the channel after establishment cannot\n be modified by attackers without detection.\n\n These properties should be true even in the face of an attacker who\n has complete control of the network, as described in [RFC3552]. See\n Appendix E for a more complete statement of the relevant security\n properties.\n\n TLS consists of two primary components:\n\n - A handshake protocol (Section 4) that authenticates the\n communicating parties, negotiates cryptographic modes and\n parameters, and establishes shared keying material. The handshake\n protocol is designed to resist tampering; an active attacker\n should not be able to force the peers to negotiate different\n parameters than they would if the connection were not under\n attack.\n\n - A record protocol (Section 5) that uses the parameters established\n by the handshake protocol to protect traffic between the\n communicating peers. The record protocol divides traffic up into\n a series of records, each of which is independently protected\n using the traffic keys.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 8446 TLS August 2018\n\n\n TLS is application protocol independent; higher-level protocols can\n layer on top of TLS transparently. The TLS standard, however, does\n not specify how protocols add security with TLS; how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n This document defines TLS version 1.3. While TLS 1.3 is not directly\n compatible with previous versions, all versions of TLS incorporate a\n versioning mechanism which allows clients and servers to\n interoperably negotiate a common version if one is supported by both\n peers.\n\n This document supersedes and obsoletes previous versions of TLS,\n including version 1.2 [RFC5246]. It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2. Because TLS 1.3 changes the way keys are\n derived, it updates [RFC5705] as described in Section 7.5. It also\n changes how Online Certificate Status Protocol (OCSP) messages are\n carried and therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in Section 4.4.2.1.\n\n1.1. Conventions and Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all\n capitals, as shown here.\n\n The following terms are used:\n\n client: The endpoint initiating the TLS connection.\n\n connection: A transport-layer connection between two endpoints.\n\n endpoint: Either the client or server of the connection.\n\n handshake: An initial negotiation between client and server that\n establishes the parameters of their subsequent interactions\n within TLS.\n\n peer: An endpoint. When discussing a particular endpoint, \"peer\"\n refers to the endpoint that is not the primary subject of\n discussion.\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\f\nRFC 8446 TLS August 2018\n\n\n receiver: An endpoint that is receiving records.\n\n sender: An endpoint that is transmitting records.\n\n server: The endpoint that did not initiate the TLS connection.\n\n1.2. Major Differences from TLS 1.2\n\n The following is a list of the major functional differences between\n TLS 1.2 and TLS 1.3. It is not intended to be exhaustive, and there\n are many minor differences.\n\n - The list of supported symmetric encryption algorithms has been\n pruned of all algorithms that are considered legacy. Those that\n remain are all Authenticated Encryption with Associated Data\n (AEAD) algorithms. The cipher suite concept has been changed to\n separate the authentication and key exchange mechanisms from the\n record protection algorithm (including secret key length) and a\n hash to be used with both the key derivation function and\n handshake message authentication code (MAC).\n\n - A zero round-trip time (0-RTT) mode was added, saving a round trip\n at connection setup for some application data, at the cost of\n certain security properties.\n\n - Static RSA and Diffie-Hellman cipher suites have been removed; all\n public-key based key exchange mechanisms now provide forward\n secrecy.\n\n - All handshake messages after the ServerHello are now encrypted.\n The newly introduced EncryptedExtensions message allows various\n extensions previously sent in the clear in the ServerHello to also\n enjoy confidentiality protection.\n\n - The key derivation functions have been redesigned. The new design\n allows easier analysis by cryptographers due to their improved key\n separation properties. The HMAC-based Extract-and-Expand Key\n Derivation Function (HKDF) is used as an underlying primitive.\n\n - The handshake state machine has been significantly restructured to\n be more consistent and to remove superfluous messages such as\n ChangeCipherSpec (except when needed for middlebox compatibility).\n\n - Elliptic curve algorithms are now in the base spec, and new\n signature algorithms, such as EdDSA, are included. TLS 1.3\n removed point format negotiation in favor of a single point format\n for each curve.\n\n\n\n\nRescorla Standards Track [Page 8]\n\f\nRFC 8446 TLS August 2018\n\n\n - Other cryptographic improvements were made, including changing the\n RSA padding to use the RSA Probabilistic Signature Scheme\n (RSASSA-PSS), and the removal of compression, the Digital\n Signature Algorithm (DSA), and custom Ephemeral Diffie-Hellman\n (DHE) groups.\n\n - The TLS 1.2 version negotiation mechanism has been deprecated in\n favor of a version list in an extension. This increases\n compatibility with existing servers that incorrectly implemented\n version negotiation.\n\n - Session resumption with and without server-side state as well as\n the PSK-based cipher suites of earlier TLS versions have been\n replaced by a single new PSK exchange.\n\n - References have been updated to point to the updated versions of\n RFCs, as appropriate (e.g., RFC 5280 rather than RFC 3280).\n\n1.3. Updates Affecting TLS 1.2\n\n This document defines several changes that optionally affect\n implementations of TLS 1.2, including those which do not also support\n TLS 1.3:\n\n - A version downgrade protection mechanism is described in\n Section 4.1.3.\n\n - RSASSA-PSS signature schemes are defined in Section 4.2.3.\n\n - The \"supported_versions\" ClientHello extension can be used to\n negotiate the version of TLS to use, in preference to the\n legacy_version field of the ClientHello.\n\n - The \"signature_algorithms_cert\" extension allows a client to\n indicate which signature algorithms it can validate in X.509\n certificates.\n\n Additionally, this document clarifies some compliance requirements\n for earlier versions of TLS; see Section 9.3.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 9]\n\f\nRFC 8446 TLS August 2018\n\n\n2. Protocol Overview\n\n The cryptographic parameters used by the secure channel are produced\n by the TLS handshake protocol. This sub-protocol of TLS is used by\n the client and server when first communicating with each other. The\n handshake protocol allows peers to negotiate a protocol version,\n select cryptographic algorithms, optionally authenticate each other,\n and establish shared secret keying material. Once the handshake is\n complete, the peers use the established keys to protect the\n application-layer traffic.\n\n A failure of the handshake or other protocol error triggers the\n termination of the connection, optionally preceded by an alert\n message (Section 6).\n\n TLS supports three basic key exchange modes:\n\n - (EC)DHE (Diffie-Hellman over either finite fields or elliptic\n curves)\n\n - PSK-only\n\n - PSK with (EC)DHE\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 10]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 1 below shows the basic full TLS handshake:\n\n Client Server\n\nKey ^ ClientHello\nExch | + key_share*\n | + signature_algorithms*\n | + psk_key_exchange_modes*\n v + pre_shared_key* -------->\n ServerHello ^ Key\n + key_share* | Exch\n + pre_shared_key* v\n {EncryptedExtensions} ^ Server\n {CertificateRequest*} v Params\n {Certificate*} ^\n {CertificateVerify*} | Auth\n {Finished} v\n <-------- [Application Data*]\n ^ {Certificate*}\nAuth | {CertificateVerify*}\n v {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 1: Message Flow for Full TLS Handshake\n\n The handshake can be thought of as having three phases (indicated in\n the diagram above):\n\n - Key Exchange: Establish shared keying material and select the\n cryptographic parameters. Everything after this phase is\n encrypted.\n\n - Server Parameters: Establish other handshake parameters\n (whether the client is authenticated, application-layer protocol\n support, etc.).\n\n\n\n\nRescorla Standards Track [Page 11]\n\f\nRFC 8446 TLS August 2018\n\n\n - Authentication: Authenticate the server (and, optionally, the\n client) and provide key confirmation and handshake integrity.\n\n In the Key Exchange phase, the client sends the ClientHello\n (Section 4.1.2) message, which contains a random nonce\n (ClientHello.random); its offered protocol versions; a list of\n symmetric cipher/HKDF hash pairs; either a set of Diffie-Hellman key\n shares (in the \"key_share\" (Section 4.2.8) extension), a set of\n pre-shared key labels (in the \"pre_shared_key\" (Section 4.2.11)\n extension), or both; and potentially additional extensions.\n Additional fields and/or messages may also be present for middlebox\n compatibility.\n\n The server processes the ClientHello and determines the appropriate\n cryptographic parameters for the connection. It then responds with\n its own ServerHello (Section 4.1.3), which indicates the negotiated\n connection parameters. The combination of the ClientHello and the\n ServerHello determines the shared keys. If (EC)DHE key establishment\n is in use, then the ServerHello contains a \"key_share\" extension with\n the server's ephemeral Diffie-Hellman share; the server's share MUST\n be in the same group as one of the client's shares. If PSK key\n establishment is in use, then the ServerHello contains a\n \"pre_shared_key\" extension indicating which of the client's offered\n PSKs was selected. Note that implementations can use (EC)DHE and PSK\n together, in which case both extensions will be supplied.\n\n The server then sends two messages to establish the Server\n Parameters:\n\n EncryptedExtensions: responses to ClientHello extensions that are\n not required to determine the cryptographic parameters, other than\n those that are specific to individual certificates.\n [Section 4.3.1]\n\n CertificateRequest: if certificate-based client authentication is\n desired, the desired parameters for that certificate. This\n message is omitted if client authentication is not desired.\n [Section 4.3.2]\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 12]\n\f\nRFC 8446 TLS August 2018\n\n\n Finally, the client and server exchange Authentication messages. TLS\n uses the same set of messages every time that certificate-based\n authentication is needed. (PSK-based authentication happens as a\n side effect of key exchange.) Specifically:\n\n Certificate: The certificate of the endpoint and any per-certificate\n extensions. This message is omitted by the server if not\n authenticating with a certificate and by the client if the server\n did not send CertificateRequest (thus indicating that the client\n should not authenticate with a certificate). Note that if raw\n public keys [RFC7250] or the cached information extension\n [RFC7924] are in use, then this message will not contain a\n certificate but rather some other value corresponding to the\n server's long-term key. [Section 4.4.2]\n\n CertificateVerify: A signature over the entire handshake using the\n private key corresponding to the public key in the Certificate\n message. This message is omitted if the endpoint is not\n authenticating via a certificate. [Section 4.4.3]\n\n Finished: A MAC (Message Authentication Code) over the entire\n handshake. This message provides key confirmation, binds the\n endpoint's identity to the exchanged keys, and in PSK mode also\n authenticates the handshake. [Section 4.4.4]\n\n Upon receiving the server's messages, the client responds with its\n Authentication messages, namely Certificate and CertificateVerify (if\n requested), and Finished.\n\n At this point, the handshake is complete, and the client and server\n derive the keying material required by the record layer to exchange\n application-layer data protected through authenticated encryption.\n Application Data MUST NOT be sent prior to sending the Finished\n message, except as specified in Section 2.3. Note that while the\n server may send Application Data prior to receiving the client's\n Authentication messages, any data sent at that point is, of course,\n being sent to an unauthenticated peer.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 13]\n\f\nRFC 8446 TLS August 2018\n\n\n2.1. Incorrect DHE Share\n\n If the client has not provided a sufficient \"key_share\" extension\n (e.g., it includes only DHE or ECDHE groups unacceptable to or\n unsupported by the server), the server corrects the mismatch with a\n HelloRetryRequest and the client needs to restart the handshake with\n an appropriate \"key_share\" extension, as shown in Figure 2. If no\n common cryptographic parameters can be negotiated, the server MUST\n abort the handshake with an appropriate alert.\n\n Client Server\n\n ClientHello\n + key_share -------->\n HelloRetryRequest\n <-------- + key_share\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 2: Message Flow for a Full Handshake with\n Mismatched Parameters\n\n Note: The handshake transcript incorporates the initial\n ClientHello/HelloRetryRequest exchange; it is not reset with the\n new ClientHello.\n\n TLS also allows several optimized variants of the basic handshake, as\n described in the following sections.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 14]\n\f\nRFC 8446 TLS August 2018\n\n\n2.2. Resumption and Pre-Shared Key (PSK)\n\n Although TLS PSKs can be established out of band, PSKs can also be\n established in a previous connection and then used to establish a new\n connection (\"session resumption\" or \"resuming\" with a PSK). Once a\n handshake has completed, the server can send the client a PSK\n identity that corresponds to a unique key derived from the initial\n handshake (see Section 4.6.1). The client can then use that PSK\n identity in future handshakes to negotiate the use of the associated\n PSK. If the server accepts the PSK, then the security context of the\n new connection is cryptographically tied to the original connection\n and the key derived from the initial handshake is used to bootstrap\n the cryptographic state instead of a full handshake. In TLS 1.2 and\n below, this functionality was provided by \"session IDs\" and \"session\n tickets\" [RFC5077]. Both mechanisms are obsoleted in TLS 1.3.\n\n PSKs can be used with (EC)DHE key exchange in order to provide\n forward secrecy in combination with shared keys, or can be used\n alone, at the cost of losing forward secrecy for the application\n data.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 15]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 3 shows a pair of handshakes in which the first handshake\n establishes a PSK and the second handshake uses it:\n\n Client Server\n\n Initial Handshake:\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n <-------- [NewSessionTicket]\n [Application Data] <-------> [Application Data]\n\n\n Subsequent Handshake:\n ClientHello\n + key_share*\n + pre_shared_key -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n {Finished}\n <-------- [Application Data*]\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 3: Message Flow for Resumption and PSK\n\n As the server is authenticating via a PSK, it does not send a\n Certificate or a CertificateVerify message. When a client offers\n resumption via a PSK, it SHOULD also supply a \"key_share\" extension\n to the server to allow the server to decline resumption and fall back\n to a full handshake, if needed. The server responds with a\n \"pre_shared_key\" extension to negotiate the use of PSK key\n establishment and can (as shown here) respond with a \"key_share\"\n extension to do (EC)DHE key establishment, thus providing forward\n secrecy.\n\n\n\n\nRescorla Standards Track [Page 16]\n\f\nRFC 8446 TLS August 2018\n\n\n When PSKs are provisioned out of band, the PSK identity and the KDF\n hash algorithm to be used with the PSK MUST also be provisioned.\n\n Note: When using an out-of-band provisioned pre-shared secret, a\n critical consideration is using sufficient entropy during the key\n generation, as discussed in [RFC4086]. Deriving a shared secret\n from a password or other low-entropy sources is not secure. A\n low-entropy secret, or password, is subject to dictionary attacks\n based on the PSK binder. The specified PSK authentication is not\n a strong password-based authenticated key exchange even when used\n with Diffie-Hellman key establishment. Specifically, it does not\n prevent an attacker that can observe the handshake from performing\n a brute-force attack on the password/pre-shared key.\n\n2.3. 0-RTT Data\n\n When clients and servers share a PSK (either obtained externally or\n via a previous handshake), TLS 1.3 allows clients to send data on the\n first flight (\"early data\"). The client uses the PSK to authenticate\n the server and to encrypt the early data.\n\n As shown in Figure 4, the 0-RTT data is just added to the 1-RTT\n handshake in the first flight. The rest of the handshake uses the\n same messages as for a 1-RTT handshake with PSK resumption.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 17]\n\f\nRFC 8446 TLS August 2018\n\n\n Client Server\n\n ClientHello\n + early_data\n + key_share*\n + psk_key_exchange_modes\n + pre_shared_key\n (Application Data*) -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n + early_data*\n {Finished}\n <-------- [Application Data*]\n (EndOfEarlyData)\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n () Indicates messages protected using keys\n derived from a client_early_traffic_secret.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 4: Message Flow for a 0-RTT Handshake\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 18]\n\f\nRFC 8446 TLS August 2018\n\n\n IMPORTANT NOTE: The security properties for 0-RTT data are weaker\n than those for other kinds of TLS data. Specifically:\n\n 1. This data is not forward secret, as it is encrypted solely under\n keys derived using the offered PSK.\n\n 2. There are no guarantees of non-replay between connections.\n Protection against replay for ordinary TLS 1.3 1-RTT data is\n provided via the server's Random value, but 0-RTT data does not\n depend on the ServerHello and therefore has weaker guarantees.\n This is especially relevant if the data is authenticated either\n with TLS client authentication or inside the application\n protocol. The same warnings apply to any use of the\n early_exporter_master_secret.\n\n 0-RTT data cannot be duplicated within a connection (i.e., the server\n will not process the same data twice for the same connection), and an\n attacker will not be able to make 0-RTT data appear to be 1-RTT data\n (because it is protected with different keys). Appendix E.5 contains\n a description of potential attacks, and Section 8 describes\n mechanisms which the server can use to limit the impact of replay.\n\n3. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used.\n\n3.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple-byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n following example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 19]\n\f\nRFC 8446 TLS August 2018\n\n\n3.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" (double\n brackets).\n\n Single-byte entities containing uninterpreted data are of\n type opaque.\n\n A type alias T' for an existing type T is defined by:\n\n T T';\n\n3.3. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are constructed from a fixed-length series of\n bytes concatenated as described in Section 3.1 and are also unsigned.\n The following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are transmitted\n in network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n3.4. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 20]\n\f\nRFC 8446 TLS August 2018\n\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* three consecutive 3-byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n In the following example, \"mandatory\" is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 3.3). Similarly,\n \"longer\" can represent up to 800 bytes of data, or 400 uint16\n elements, and it may be empty. Its encoding will include a two-byte\n actual length field prepended to the vector. The length of an\n encoded vector must be an exact multiple of the length of a single\n element (e.g., a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is two bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n3.5. Enumerateds\n\n An additional sparse data type, called \"enum\" or \"enumerated\", is\n available. Each definition is a different type. Only enumerateds of\n the same type may be assigned or compared. Every element of an\n enumerated must be assigned a value, as demonstrated in the following\n example. Since the elements of the enumerated are not ordered, they\n can be assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n Future extensions or additions to the protocol may define new values.\n Implementations need to be able to parse and ignore unknown values\n unless the definition of the field states otherwise.\n\n\n\n\n\nRescorla Standards Track [Page 21]\n\f\nRFC 8446 TLS August 2018\n\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4 in the current\n version of the protocol.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n The names assigned to enumerateds do not need to be unique. The\n numerical value can describe a range over which the same name\n applies. The value includes the minimum and maximum inclusive values\n in that range, separated by two period characters. This is\n principally useful for reserving regions of the space.\n\n enum { sad(0), meh(1..254), happy(255) } Mood;\n\n3.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax used for definitions is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } T;\n\n Fixed- and variable-length vector fields are allowed using the\n standard vector syntax. Structures V1 and V2 in the variants example\n (Section 3.8) demonstrate this.\n\n\n\nRescorla Standards Track [Page 22]\n\f\nRFC 8446 TLS August 2018\n\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n\n3.7. Constants\n\n Fields and variables may be assigned a fixed value using \"=\", as in:\n\n struct {\n T1 f1 = 8; /* T.f1 must always be 8 */\n T2 f2;\n } T;\n\n3.8. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. Each\n arm of the select (below) specifies the type of that variant's field\n and an optional field label. The mechanism by which the variant is\n selected at runtime is not prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1 [[fe1]];\n case e2: Te2 [[fe2]];\n ....\n case en: Ten [[fen]];\n };\n } Tv;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 23]\n\f\nRFC 8446 TLS August 2018\n\n\n For example:\n\n enum { apple(0), orange(1) } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n VariantTag type;\n select (VariantRecord.type) {\n case apple: V1;\n case orange: V2;\n };\n } VariantRecord;\n\n4. Handshake Protocol\n\n The handshake protocol is used to negotiate the security parameters\n of a connection. Handshake messages are supplied to the TLS record\n layer, where they are encapsulated within one or more TLSPlaintext or\n TLSCiphertext structures which are processed and transmitted as\n specified by the current active connection state.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 24]\n\f\nRFC 8446 TLS August 2018\n\n\n enum {\n client_hello(1),\n server_hello(2),\n new_session_ticket(4),\n end_of_early_data(5),\n encrypted_extensions(8),\n certificate(11),\n certificate_request(13),\n certificate_verify(15),\n finished(20),\n key_update(24),\n message_hash(254),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* remaining bytes in message */\n select (Handshake.msg_type) {\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case end_of_early_data: EndOfEarlyData;\n case encrypted_extensions: EncryptedExtensions;\n case certificate_request: CertificateRequest;\n case certificate: Certificate;\n case certificate_verify: CertificateVerify;\n case finished: Finished;\n case new_session_ticket: NewSessionTicket;\n case key_update: KeyUpdate;\n };\n } Handshake;\n\n Protocol messages MUST be sent in the order defined in Section 4.4.1\n and shown in the diagrams in Section 2. A peer which receives a\n handshake message in an unexpected order MUST abort the handshake\n with an \"unexpected_message\" alert.\n\n New handshake message types are assigned by IANA as described in\n Section 11.\n\n4.1. Key Exchange Messages\n\n The key exchange messages are used to determine the security\n capabilities of the client and the server and to establish shared\n secrets, including the traffic keys used to protect the rest of the\n handshake and the data.\n\n\n\n\n\nRescorla Standards Track [Page 25]\n\f\nRFC 8446 TLS August 2018\n\n\n4.1.1. Cryptographic Negotiation\n\n In TLS, the cryptographic negotiation proceeds by the client offering\n the following four sets of options in its ClientHello:\n\n - A list of cipher suites which indicates the AEAD algorithm/HKDF\n hash pairs which the client supports.\n\n - A \"supported_groups\" (Section 4.2.7) extension which indicates the\n (EC)DHE groups which the client supports and a \"key_share\"\n (Section 4.2.8) extension which contains (EC)DHE shares for some\n or all of these groups.\n\n - A \"signature_algorithms\" (Section 4.2.3) extension which indicates\n the signature algorithms which the client can accept. A\n \"signature_algorithms_cert\" extension (Section 4.2.3) may also be\n added to indicate certificate-specific signature algorithms.\n\n - A \"pre_shared_key\" (Section 4.2.11) extension which contains a\n list of symmetric key identities known to the client and a\n \"psk_key_exchange_modes\" (Section 4.2.9) extension which indicates\n the key exchange modes that may be used with PSKs.\n\n If the server does not select a PSK, then the first three of these\n options are entirely orthogonal: the server independently selects a\n cipher suite, an (EC)DHE group and key share for key establishment,\n and a signature algorithm/certificate pair to authenticate itself to\n the client. If there is no overlap between the received\n \"supported_groups\" and the groups supported by the server, then the\n server MUST abort the handshake with a \"handshake_failure\" or an\n \"insufficient_security\" alert.\n\n If the server selects a PSK, then it MUST also select a key\n establishment mode from the set indicated by the client's\n \"psk_key_exchange_modes\" extension (at present, PSK alone or with\n (EC)DHE). Note that if the PSK can be used without (EC)DHE, then\n non-overlap in the \"supported_groups\" parameters need not be fatal,\n as it is in the non-PSK case discussed in the previous paragraph.\n\n If the server selects an (EC)DHE group and the client did not offer a\n compatible \"key_share\" extension in the initial ClientHello, the\n server MUST respond with a HelloRetryRequest (Section 4.1.4) message.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 26]\n\f\nRFC 8446 TLS August 2018\n\n\n If the server successfully selects parameters and does not require a\n HelloRetryRequest, it indicates the selected parameters in the\n ServerHello as follows:\n\n - If PSK is being used, then the server will send a \"pre_shared_key\"\n extension indicating the selected key.\n\n - When (EC)DHE is in use, the server will also provide a \"key_share\"\n extension. If PSK is not being used, then (EC)DHE and\n certificate-based authentication are always used.\n\n - When authenticating via a certificate, the server will send the\n Certificate (Section 4.4.2) and CertificateVerify (Section 4.4.3)\n messages. In TLS 1.3 as defined by this document, either a PSK or\n a certificate is always used, but not both. Future documents may\n define how to use them together.\n\n If the server is unable to negotiate a supported set of parameters\n (i.e., there is no overlap between the client and server parameters),\n it MUST abort the handshake with either a \"handshake_failure\" or\n \"insufficient_security\" fatal alert (see Section 6).\n\n4.1.2. Client Hello\n\n When a client first connects to a server, it is REQUIRED to send the\n ClientHello as its first TLS message. The client will also send a\n ClientHello when the server has responded to its ClientHello with a\n HelloRetryRequest. In that case, the client MUST send the same\n ClientHello without modification, except as follows:\n\n - If a \"key_share\" extension was supplied in the HelloRetryRequest,\n replacing the list of shares with a list containing a single\n KeyShareEntry from the indicated group.\n\n - Removing the \"early_data\" extension (Section 4.2.10) if one was\n present. Early data is not permitted after a HelloRetryRequest.\n\n - Including a \"cookie\" extension if one was provided in the\n HelloRetryRequest.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 27]\n\f\nRFC 8446 TLS August 2018\n\n\n - Updating the \"pre_shared_key\" extension if present by recomputing\n the \"obfuscated_ticket_age\" and binder values and (optionally)\n removing any PSKs which are incompatible with the server's\n indicated cipher suite.\n\n - Optionally adding, removing, or changing the length of the\n \"padding\" extension [RFC7685].\n\n - Other modifications that may be allowed by an extension defined in\n the future and present in the HelloRetryRequest.\n\n Because TLS 1.3 forbids renegotiation, if a server has negotiated\n TLS 1.3 and receives a ClientHello at any other time, it MUST\n terminate the connection with an \"unexpected_message\" alert.\n\n If a server established a TLS connection with a previous version of\n TLS and receives a TLS 1.3 ClientHello in a renegotiation, it MUST\n retain the previous protocol version. In particular, it MUST NOT\n negotiate TLS 1.3.\n\n Structure of this message:\n\n uint16 ProtocolVersion;\n opaque Random[32];\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id<0..32>;\n CipherSuite cipher_suites<2..2^16-2>;\n opaque legacy_compression_methods<1..2^8-1>;\n Extension extensions<8..2^16-1>;\n } ClientHello;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 28]\n\f\nRFC 8446 TLS August 2018\n\n\n legacy_version: In previous versions of TLS, this field was used for\n version negotiation and represented the highest version number\n supported by the client. Experience has shown that many servers\n do not properly implement version negotiation, leading to \"version\n intolerance\" in which the server rejects an otherwise acceptable\n ClientHello with a version number higher than it supports. In\n TLS 1.3, the client indicates its version preferences in the\n \"supported_versions\" extension (Section 4.2.1) and the\n legacy_version field MUST be set to 0x0303, which is the version\n number for TLS 1.2. TLS 1.3 ClientHellos are identified as having\n a legacy_version of 0x0303 and a supported_versions extension\n present with 0x0304 as the highest version indicated therein.\n (See Appendix D for details about backward compatibility.)\n\n random: 32 bytes generated by a secure random number generator. See\n Appendix C for additional information.\n\n legacy_session_id: Versions of TLS before TLS 1.3 supported a\n \"session resumption\" feature which has been merged with pre-shared\n keys in this version (see Section 2.2). A client which has a\n cached session ID set by a pre-TLS 1.3 server SHOULD set this\n field to that value. In compatibility mode (see Appendix D.4),\n this field MUST be non-empty, so a client not offering a\n pre-TLS 1.3 session MUST generate a new 32-byte value. This value\n need not be random but SHOULD be unpredictable to avoid\n implementations fixating on a specific value (also known as\n ossification). Otherwise, it MUST be set as a zero-length vector\n (i.e., a zero-valued single byte length field).\n\n cipher_suites: A list of the symmetric cipher options supported by\n the client, specifically the record protection algorithm\n (including secret key length) and a hash to be used with HKDF, in\n descending order of client preference. Values are defined in\n Appendix B.4. If the list contains cipher suites that the server\n does not recognize, support, or wish to use, the server MUST\n ignore those cipher suites and process the remaining ones as\n usual. If the client is attempting a PSK key establishment, it\n SHOULD advertise at least one cipher suite indicating a Hash\n associated with the PSK.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 29]\n\f\nRFC 8446 TLS August 2018\n\n\n legacy_compression_methods: Versions of TLS before 1.3 supported\n compression with the list of supported compression methods being\n sent in this field. For every TLS 1.3 ClientHello, this vector\n MUST contain exactly one byte, set to zero, which corresponds to\n the \"null\" compression method in prior versions of TLS. If a\n TLS 1.3 ClientHello is received with any other value in this\n field, the server MUST abort the handshake with an\n \"illegal_parameter\" alert. Note that TLS 1.3 servers might\n receive TLS 1.2 or prior ClientHellos which contain other\n compression methods and (if negotiating such a prior version) MUST\n follow the procedures for the appropriate prior version of TLS.\n\n extensions: Clients request extended functionality from servers by\n sending data in the extensions field. The actual \"Extension\"\n format is defined in Section 4.2. In TLS 1.3, the use of certain\n extensions is mandatory, as functionality has moved into\n extensions to preserve ClientHello compatibility with previous\n versions of TLS. Servers MUST ignore unrecognized extensions.\n\n All versions of TLS allow an extensions field to optionally follow\n the compression_methods field. TLS 1.3 ClientHello messages always\n contain extensions (minimally \"supported_versions\", otherwise, they\n will be interpreted as TLS 1.2 ClientHello messages). However,\n TLS 1.3 servers might receive ClientHello messages without an\n extensions field from prior versions of TLS. The presence of\n extensions can be detected by determining whether there are bytes\n following the compression_methods field at the end of the\n ClientHello. Note that this method of detecting optional data\n differs from the normal TLS method of having a variable-length field,\n but it is used for compatibility with TLS before extensions were\n defined. TLS 1.3 servers will need to perform this check first and\n only attempt to negotiate TLS 1.3 if the \"supported_versions\"\n extension is present. If negotiating a version of TLS prior to 1.3,\n a server MUST check that the message either contains no data after\n legacy_compression_methods or that it contains a valid extensions\n block with no data following. If not, then it MUST abort the\n handshake with a \"decode_error\" alert.\n\n In the event that a client requests additional functionality using\n extensions and this functionality is not supplied by the server, the\n client MAY abort the handshake.\n\n After sending the ClientHello message, the client waits for a\n ServerHello or HelloRetryRequest message. If early data is in use,\n the client may transmit early Application Data (Section 2.3) while\n waiting for the next handshake message.\n\n\n\n\n\nRescorla Standards Track [Page 30]\n\f\nRFC 8446 TLS August 2018\n\n\n4.1.3. Server Hello\n\n The server will send this message in response to a ClientHello\n message to proceed with the handshake if it is able to negotiate an\n acceptable set of handshake parameters based on the ClientHello.\n\n Structure of this message:\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id_echo<0..32>;\n CipherSuite cipher_suite;\n uint8 legacy_compression_method = 0;\n Extension extensions<6..2^16-1>;\n } ServerHello;\n\n legacy_version: In previous versions of TLS, this field was used for\n version negotiation and represented the selected version number\n for the connection. Unfortunately, some middleboxes fail when\n presented with new values. In TLS 1.3, the TLS server indicates\n its version using the \"supported_versions\" extension\n (Section 4.2.1), and the legacy_version field MUST be set to\n 0x0303, which is the version number for TLS 1.2. (See Appendix D\n for details about backward compatibility.)\n\n random: 32 bytes generated by a secure random number generator. See\n Appendix C for additional information. The last 8 bytes MUST be\n overwritten as described below if negotiating TLS 1.2 or TLS 1.1,\n but the remaining bytes MUST be random. This structure is\n generated by the server and MUST be generated independently of the\n ClientHello.random.\n\n legacy_session_id_echo: The contents of the client's\n legacy_session_id field. Note that this field is echoed even if\n the client's value corresponded to a cached pre-TLS 1.3 session\n which the server has chosen not to resume. A client which\n receives a legacy_session_id_echo field that does not match what\n it sent in the ClientHello MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n cipher_suite: The single cipher suite selected by the server from\n the list in ClientHello.cipher_suites. A client which receives a\n cipher suite that was not offered MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n legacy_compression_method: A single byte which MUST have the\n value 0.\n\n\n\nRescorla Standards Track [Page 31]\n\f\nRFC 8446 TLS August 2018\n\n\n extensions: A list of extensions. The ServerHello MUST only include\n extensions which are required to establish the cryptographic\n context and negotiate the protocol version. All TLS 1.3\n ServerHello messages MUST contain the \"supported_versions\"\n extension. Current ServerHello messages additionally contain\n either the \"pre_shared_key\" extension or the \"key_share\"\n extension, or both (when using a PSK with (EC)DHE key\n establishment). Other extensions (see Section 4.2) are sent\n separately in the EncryptedExtensions message.\n\n For reasons of backward compatibility with middleboxes (see\n Appendix D.4), the HelloRetryRequest message uses the same structure\n as the ServerHello, but with Random set to the special value of the\n SHA-256 of \"HelloRetryRequest\":\n\n CF 21 AD 74 E5 9A 61 11 BE 1D 8C 02 1E 65 B8 91\n C2 A2 11 16 7A BB 8C 5E 07 9E 09 E2 C8 A8 33 9C\n\n Upon receiving a message with type server_hello, implementations MUST\n first examine the Random value and, if it matches this value, process\n it as described in Section 4.1.4).\n\n TLS 1.3 has a downgrade protection mechanism embedded in the server's\n random value. TLS 1.3 servers which negotiate TLS 1.2 or below in\n response to a ClientHello MUST set the last 8 bytes of their Random\n value specially in their ServerHello.\n\n If negotiating TLS 1.2, TLS 1.3 servers MUST set the last 8 bytes of\n their Random value to the bytes:\n\n 44 4F 57 4E 47 52 44 01\n\n If negotiating TLS 1.1 or below, TLS 1.3 servers MUST, and TLS 1.2\n servers SHOULD, set the last 8 bytes of their ServerHello.Random\n value to the bytes:\n\n 44 4F 57 4E 47 52 44 00\n\n TLS 1.3 clients receiving a ServerHello indicating TLS 1.2 or below\n MUST check that the last 8 bytes are not equal to either of these\n values. TLS 1.2 clients SHOULD also check that the last 8 bytes are\n not equal to the second value if the ServerHello indicates TLS 1.1 or\n below. If a match is found, the client MUST abort the handshake with\n an \"illegal_parameter\" alert. This mechanism provides limited\n protection against downgrade attacks over and above what is provided\n by the Finished exchange: because the ServerKeyExchange, a message\n present in TLS 1.2 and below, includes a signature over both random\n values, it is not possible for an active attacker to modify the\n\n\n\nRescorla Standards Track [Page 32]\n\f\nRFC 8446 TLS August 2018\n\n\n random values without detection as long as ephemeral ciphers are\n used. It does not provide downgrade protection when static RSA\n is used.\n\n Note: This is a change from [RFC5246], so in practice many TLS 1.2\n clients and servers will not behave as specified above.\n\n A legacy TLS client performing renegotiation with TLS 1.2 or prior\n and which receives a TLS 1.3 ServerHello during renegotiation MUST\n abort the handshake with a \"protocol_version\" alert. Note that\n renegotiation is not possible when TLS 1.3 has been negotiated.\n\n4.1.4. Hello Retry Request\n\n The server will send this message in response to a ClientHello\n message if it is able to find an acceptable set of parameters but the\n ClientHello does not contain sufficient information to proceed with\n the handshake. As discussed in Section 4.1.3, the HelloRetryRequest\n has the same format as a ServerHello message, and the legacy_version,\n legacy_session_id_echo, cipher_suite, and legacy_compression_method\n fields have the same meaning. However, for convenience we discuss\n \"HelloRetryRequest\" throughout this document as if it were a distinct\n message.\n\n The server's extensions MUST contain \"supported_versions\".\n Additionally, it SHOULD contain the minimal set of extensions\n necessary for the client to generate a correct ClientHello pair. As\n with the ServerHello, a HelloRetryRequest MUST NOT contain any\n extensions that were not first offered by the client in its\n ClientHello, with the exception of optionally the \"cookie\" (see\n Section 4.2.2) extension.\n\n Upon receipt of a HelloRetryRequest, the client MUST check the\n legacy_version, legacy_session_id_echo, cipher_suite, and\n legacy_compression_method as specified in Section 4.1.3 and then\n process the extensions, starting with determining the version using\n \"supported_versions\". Clients MUST abort the handshake with an\n \"illegal_parameter\" alert if the HelloRetryRequest would not result\n in any change in the ClientHello. If a client receives a second\n HelloRetryRequest in the same connection (i.e., where the ClientHello\n was itself in response to a HelloRetryRequest), it MUST abort the\n handshake with an \"unexpected_message\" alert.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 33]\n\f\nRFC 8446 TLS August 2018\n\n\n Otherwise, the client MUST process all extensions in the\n HelloRetryRequest and send a second updated ClientHello. The\n HelloRetryRequest extensions defined in this specification are:\n\n - supported_versions (see Section 4.2.1)\n\n - cookie (see Section 4.2.2)\n\n - key_share (see Section 4.2.8)\n\n A client which receives a cipher suite that was not offered MUST\n abort the handshake. Servers MUST ensure that they negotiate the\n same cipher suite when receiving a conformant updated ClientHello (if\n the server selects the cipher suite as the first step in the\n negotiation, then this will happen automatically). Upon receiving\n the ServerHello, clients MUST check that the cipher suite supplied in\n the ServerHello is the same as that in the HelloRetryRequest and\n otherwise abort the handshake with an \"illegal_parameter\" alert.\n\n In addition, in its updated ClientHello, the client SHOULD NOT offer\n any pre-shared keys associated with a hash other than that of the\n selected cipher suite. This allows the client to avoid having to\n compute partial hash transcripts for multiple hashes in the second\n ClientHello.\n\n The value of selected_version in the HelloRetryRequest\n \"supported_versions\" extension MUST be retained in the ServerHello,\n and a client MUST abort the handshake with an \"illegal_parameter\"\n alert if the value changes.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 34]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2. Extensions\n\n A number of TLS messages contain tag-length-value encoded extensions\n structures.\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n server_name(0), /* RFC 6066 */\n max_fragment_length(1), /* RFC 6066 */\n status_request(5), /* RFC 6066 */\n supported_groups(10), /* RFC 8422, 7919 */\n signature_algorithms(13), /* RFC 8446 */\n use_srtp(14), /* RFC 5764 */\n heartbeat(15), /* RFC 6520 */\n application_layer_protocol_negotiation(16), /* RFC 7301 */\n signed_certificate_timestamp(18), /* RFC 6962 */\n client_certificate_type(19), /* RFC 7250 */\n server_certificate_type(20), /* RFC 7250 */\n padding(21), /* RFC 7685 */\n pre_shared_key(41), /* RFC 8446 */\n early_data(42), /* RFC 8446 */\n supported_versions(43), /* RFC 8446 */\n cookie(44), /* RFC 8446 */\n psk_key_exchange_modes(45), /* RFC 8446 */\n certificate_authorities(47), /* RFC 8446 */\n oid_filters(48), /* RFC 8446 */\n post_handshake_auth(49), /* RFC 8446 */\n signature_algorithms_cert(50), /* RFC 8446 */\n key_share(51), /* RFC 8446 */\n (65535)\n } ExtensionType;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 35]\n\f\nRFC 8446 TLS August 2018\n\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The list of extension types is maintained by IANA as described in\n Section 11.\n\n Extensions are generally structured in a request/response fashion,\n though some extensions are just indications with no corresponding\n response. The client sends its extension requests in the ClientHello\n message, and the server sends its extension responses in the\n ServerHello, EncryptedExtensions, HelloRetryRequest, and Certificate\n messages. The server sends extension requests in the\n CertificateRequest message which a client MAY respond to with a\n Certificate message. The server MAY also send unsolicited extensions\n in the NewSessionTicket, though the client does not respond directly\n to these.\n\n Implementations MUST NOT send extension responses if the remote\n endpoint did not send the corresponding extension requests, with the\n exception of the \"cookie\" extension in the HelloRetryRequest. Upon\n receiving such an extension, an endpoint MUST abort the handshake\n with an \"unsupported_extension\" alert.\n\n The table below indicates the messages where a given extension may\n appear, using the following notation: CH (ClientHello),\n SH (ServerHello), EE (EncryptedExtensions), CT (Certificate),\n CR (CertificateRequest), NST (NewSessionTicket), and\n HRR (HelloRetryRequest). If an implementation receives an extension\n which it recognizes and which is not specified for the message in\n which it appears, it MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 36]\n\f\nRFC 8446 TLS August 2018\n\n\n +--------------------------------------------------+-------------+\n | Extension | TLS 1.3 |\n +--------------------------------------------------+-------------+\n | server_name [RFC6066] | CH, EE |\n | | |\n | max_fragment_length [RFC6066] | CH, EE |\n | | |\n | status_request [RFC6066] | CH, CR, CT |\n | | |\n | supported_groups [RFC7919] | CH, EE |\n | | |\n | signature_algorithms (RFC 8446) | CH, CR |\n | | |\n | use_srtp [RFC5764] | CH, EE |\n | | |\n | heartbeat [RFC6520] | CH, EE |\n | | |\n | application_layer_protocol_negotiation [RFC7301] | CH, EE |\n | | |\n | signed_certificate_timestamp [RFC6962] | CH, CR, CT |\n | | |\n | client_certificate_type [RFC7250] | CH, EE |\n | | |\n | server_certificate_type [RFC7250] | CH, EE |\n | | |\n | padding [RFC7685] | CH |\n | | |\n | key_share (RFC 8446) | CH, SH, HRR |\n | | |\n | pre_shared_key (RFC 8446) | CH, SH |\n | | |\n | psk_key_exchange_modes (RFC 8446) | CH |\n | | |\n | early_data (RFC 8446) | CH, EE, NST |\n | | |\n | cookie (RFC 8446) | CH, HRR |\n | | |\n | supported_versions (RFC 8446) | CH, SH, HRR |\n | | |\n | certificate_authorities (RFC 8446) | CH, CR |\n | | |\n | oid_filters (RFC 8446) | CR |\n | | |\n | post_handshake_auth (RFC 8446) | CH |\n | | |\n | signature_algorithms_cert (RFC 8446) | CH, CR |\n +--------------------------------------------------+-------------+\n\n\n\n\nRescorla Standards Track [Page 37]\n\f\nRFC 8446 TLS August 2018\n\n\n When multiple extensions of different types are present, the\n extensions MAY appear in any order, with the exception of\n \"pre_shared_key\" (Section 4.2.11) which MUST be the last extension in\n the ClientHello (but can appear anywhere in the ServerHello\n extensions block). There MUST NOT be more than one extension of the\n same type in a given extension block.\n\n In TLS 1.3, unlike TLS 1.2, extensions are negotiated for each\n handshake even when in resumption-PSK mode. However, 0-RTT\n parameters are those negotiated in the previous handshake; mismatches\n may require rejecting 0-RTT (see Section 4.2.10).\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions (e.g., the handshake cannot continue), and some are\n simply refusals to support particular features. In general, error\n alerts should be used for the former and a field in the server\n extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem. Often the fact that the extension fields are\n included in the inputs to the Finished message hashes will be\n sufficient, but extreme care is needed when the extension changes\n the meaning of messages sent in the handshake phase. Designers\n and implementors should be aware of the fact that until the\n handshake has been authenticated, active attackers can modify\n messages and insert, remove, or replace extensions.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 38]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.1. Supported Versions\n\n struct {\n select (Handshake.msg_type) {\n case client_hello:\n ProtocolVersion versions<2..254>;\n\n case server_hello: /* and HelloRetryRequest */\n ProtocolVersion selected_version;\n };\n } SupportedVersions;\n\n The \"supported_versions\" extension is used by the client to indicate\n which versions of TLS it supports and by the server to indicate which\n version it is using. The extension contains a list of supported\n versions in preference order, with the most preferred version first.\n Implementations of this specification MUST send this extension in the\n ClientHello containing all versions of TLS which they are prepared to\n negotiate (for this specification, that means minimally 0x0304, but\n if previous versions of TLS are allowed to be negotiated, they MUST\n be present as well).\n\n If this extension is not present, servers which are compliant with\n this specification and which also support TLS 1.2 MUST negotiate\n TLS 1.2 or prior as specified in [RFC5246], even if\n ClientHello.legacy_version is 0x0304 or later. Servers MAY abort the\n handshake upon receiving a ClientHello with legacy_version 0x0304 or\n later.\n\n If this extension is present in the ClientHello, servers MUST NOT use\n the ClientHello.legacy_version value for version negotiation and MUST\n use only the \"supported_versions\" extension to determine client\n preferences. Servers MUST only select a version of TLS present in\n that extension and MUST ignore any unknown versions that are present\n in that extension. Note that this mechanism makes it possible to\n negotiate a version prior to TLS 1.2 if one side supports a sparse\n range. Implementations of TLS 1.3 which choose to support prior\n versions of TLS SHOULD support TLS 1.2. Servers MUST be prepared to\n receive ClientHellos that include this extension but do not include\n 0x0304 in the list of versions.\n\n A server which negotiates a version of TLS prior to TLS 1.3 MUST set\n ServerHello.version and MUST NOT send the \"supported_versions\"\n extension. A server which negotiates TLS 1.3 MUST respond by sending\n a \"supported_versions\" extension containing the selected version\n value (0x0304). It MUST set the ServerHello.legacy_version field to\n 0x0303 (TLS 1.2). Clients MUST check for this extension prior to\n processing the rest of the ServerHello (although they will have to\n\n\n\nRescorla Standards Track [Page 39]\n\f\nRFC 8446 TLS August 2018\n\n\n parse the ServerHello in order to read the extension). If this\n extension is present, clients MUST ignore the\n ServerHello.legacy_version value and MUST use only the\n \"supported_versions\" extension to determine the selected version. If\n the \"supported_versions\" extension in the ServerHello contains a\n version not offered by the client or contains a version prior to\n TLS 1.3, the client MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n4.2.2. Cookie\n\n struct {\n opaque cookie<1..2^16-1>;\n } Cookie;\n\n Cookies serve two primary purposes:\n\n - Allowing the server to force the client to demonstrate\n reachability at their apparent network address (thus providing a\n measure of DoS protection). This is primarily useful for\n non-connection-oriented transports (see [RFC6347] for an example\n of this).\n\n - Allowing the server to offload state to the client, thus allowing\n it to send a HelloRetryRequest without storing any state. The\n server can do this by storing the hash of the ClientHello in the\n HelloRetryRequest cookie (protected with some suitable integrity\n protection algorithm).\n\n When sending a HelloRetryRequest, the server MAY provide a \"cookie\"\n extension to the client (this is an exception to the usual rule that\n the only extensions that may be sent are those that appear in the\n ClientHello). When sending the new ClientHello, the client MUST copy\n the contents of the extension received in the HelloRetryRequest into\n a \"cookie\" extension in the new ClientHello. Clients MUST NOT use\n cookies in their initial ClientHello in subsequent connections.\n\n When a server is operating statelessly, it may receive an unprotected\n record of type change_cipher_spec between the first and second\n ClientHello (see Section 5). Since the server is not storing any\n state, this will appear as if it were the first message to be\n received. Servers operating statelessly MUST ignore these records.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 40]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.3. Signature Algorithms\n\n TLS 1.3 provides two extensions for indicating which signature\n algorithms may be used in digital signatures. The\n \"signature_algorithms_cert\" extension applies to signatures in\n certificates, and the \"signature_algorithms\" extension, which\n originally appeared in TLS 1.2, applies to signatures in\n CertificateVerify messages. The keys found in certificates MUST also\n be of appropriate type for the signature algorithms they are used\n with. This is a particular issue for RSA keys and PSS signatures, as\n described below. If no \"signature_algorithms_cert\" extension is\n present, then the \"signature_algorithms\" extension also applies to\n signatures appearing in certificates. Clients which desire the\n server to authenticate itself via a certificate MUST send the\n \"signature_algorithms\" extension. If a server is authenticating via\n a certificate and the client has not sent a \"signature_algorithms\"\n extension, then the server MUST abort the handshake with a\n \"missing_extension\" alert (see Section 9.2).\n\n The \"signature_algorithms_cert\" extension was added to allow\n implementations which supported different sets of algorithms for\n certificates and in TLS itself to clearly signal their capabilities.\n TLS 1.2 implementations SHOULD also process this extension.\n Implementations which have the same policy in both cases MAY omit the\n \"signature_algorithms_cert\" extension.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 41]\n\f\nRFC 8446 TLS August 2018\n\n\n The \"extension_data\" field of these extensions contains a\n SignatureSchemeList value:\n\n enum {\n /* RSASSA-PKCS1-v1_5 algorithms */\n rsa_pkcs1_sha256(0x0401),\n rsa_pkcs1_sha384(0x0501),\n rsa_pkcs1_sha512(0x0601),\n\n /* ECDSA algorithms */\n ecdsa_secp256r1_sha256(0x0403),\n ecdsa_secp384r1_sha384(0x0503),\n ecdsa_secp521r1_sha512(0x0603),\n\n /* RSASSA-PSS algorithms with public key OID rsaEncryption */\n rsa_pss_rsae_sha256(0x0804),\n rsa_pss_rsae_sha384(0x0805),\n rsa_pss_rsae_sha512(0x0806),\n\n /* EdDSA algorithms */\n ed25519(0x0807),\n ed448(0x0808),\n\n /* RSASSA-PSS algorithms with public key OID RSASSA-PSS */\n rsa_pss_pss_sha256(0x0809),\n rsa_pss_pss_sha384(0x080a),\n rsa_pss_pss_sha512(0x080b),\n\n /* Legacy algorithms */\n rsa_pkcs1_sha1(0x0201),\n ecdsa_sha1(0x0203),\n\n /* Reserved Code Points */\n private_use(0xFE00..0xFFFF),\n (0xFFFF)\n } SignatureScheme;\n\n struct {\n SignatureScheme supported_signature_algorithms<2..2^16-2>;\n } SignatureSchemeList;\n\n Note: This enum is named \"SignatureScheme\" because there is already a\n \"SignatureAlgorithm\" type in TLS 1.2, which this replaces. We use\n the term \"signature algorithm\" throughout the text.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 42]\n\f\nRFC 8446 TLS August 2018\n\n\n Each SignatureScheme value lists a single signature algorithm that\n the client is willing to verify. The values are indicated in\n descending order of preference. Note that a signature algorithm\n takes as input an arbitrary-length message, rather than a digest.\n Algorithms which traditionally act on a digest should be defined in\n TLS to first hash the input with a specified hash algorithm and then\n proceed as usual. The code point groups listed above have the\n following meanings:\n\n RSASSA-PKCS1-v1_5 algorithms: Indicates a signature algorithm using\n RSASSA-PKCS1-v1_5 [RFC8017] with the corresponding hash algorithm\n as defined in [SHS]. These values refer solely to signatures\n which appear in certificates (see Section 4.4.2.2) and are not\n defined for use in signed TLS handshake messages, although they\n MAY appear in \"signature_algorithms\" and\n \"signature_algorithms_cert\" for backward compatibility with\n TLS 1.2.\n\n ECDSA algorithms: Indicates a signature algorithm using ECDSA\n [ECDSA], the corresponding curve as defined in ANSI X9.62 [ECDSA]\n and FIPS 186-4 [DSS], and the corresponding hash algorithm as\n defined in [SHS]. The signature is represented as a DER-encoded\n [X690] ECDSA-Sig-Value structure.\n\n RSASSA-PSS RSAE algorithms: Indicates a signature algorithm using\n RSASSA-PSS [RFC8017] with mask generation function 1. The digest\n used in the mask generation function and the digest being signed\n are both the corresponding hash algorithm as defined in [SHS].\n The length of the Salt MUST be equal to the length of the output\n of the digest algorithm. If the public key is carried in an X.509\n certificate, it MUST use the rsaEncryption OID [RFC5280].\n\n EdDSA algorithms: Indicates a signature algorithm using EdDSA as\n defined in [RFC8032] or its successors. Note that these\n correspond to the \"PureEdDSA\" algorithms and not the \"prehash\"\n variants.\n\n RSASSA-PSS PSS algorithms: Indicates a signature algorithm using\n RSASSA-PSS [RFC8017] with mask generation function 1. The digest\n used in the mask generation function and the digest being signed\n are both the corresponding hash algorithm as defined in [SHS].\n The length of the Salt MUST be equal to the length of the digest\n algorithm. If the public key is carried in an X.509 certificate,\n it MUST use the RSASSA-PSS OID [RFC5756]. When used in\n certificate signatures, the algorithm parameters MUST be DER\n encoded. If the corresponding public key's parameters are\n present, then the parameters in the signature MUST be identical to\n those in the public key.\n\n\n\nRescorla Standards Track [Page 43]\n\f\nRFC 8446 TLS August 2018\n\n\n Legacy algorithms: Indicates algorithms which are being deprecated\n because they use algorithms with known weaknesses, specifically\n SHA-1 which is used in this context with either (1) RSA using\n RSASSA-PKCS1-v1_5 or (2) ECDSA. These values refer solely to\n signatures which appear in certificates (see Section 4.4.2.2) and\n are not defined for use in signed TLS handshake messages, although\n they MAY appear in \"signature_algorithms\" and\n \"signature_algorithms_cert\" for backward compatibility with\n TLS 1.2. Endpoints SHOULD NOT negotiate these algorithms but are\n permitted to do so solely for backward compatibility. Clients\n offering these values MUST list them as the lowest priority\n (listed after all other algorithms in SignatureSchemeList).\n TLS 1.3 servers MUST NOT offer a SHA-1 signed certificate unless\n no valid certificate chain can be produced without it (see\n Section 4.4.2.2).\n\n The signatures on certificates that are self-signed or certificates\n that are trust anchors are not validated, since they begin a\n certification path (see [RFC5280], Section 3.2). A certificate that\n begins a certification path MAY use a signature algorithm that is not\n advertised as being supported in the \"signature_algorithms\"\n extension.\n\n Note that TLS 1.2 defines this extension differently. TLS 1.3\n implementations willing to negotiate TLS 1.2 MUST behave in\n accordance with the requirements of [RFC5246] when negotiating that\n version. In particular:\n\n - TLS 1.2 ClientHellos MAY omit this extension.\n\n - In TLS 1.2, the extension contained hash/signature pairs. The\n pairs are encoded in two octets, so SignatureScheme values have\n been allocated to align with TLS 1.2's encoding. Some legacy\n pairs are left unallocated. These algorithms are deprecated as of\n TLS 1.3. They MUST NOT be offered or negotiated by any\n implementation. In particular, MD5 [SLOTH], SHA-224, and DSA\n MUST NOT be used.\n\n - ECDSA signature schemes align with TLS 1.2's ECDSA hash/signature\n pairs. However, the old semantics did not constrain the signing\n curve. If TLS 1.2 is negotiated, implementations MUST be prepared\n to accept a signature that uses any curve that they advertised in\n the \"supported_groups\" extension.\n\n - Implementations that advertise support for RSASSA-PSS (which is\n mandatory in TLS 1.3) MUST be prepared to accept a signature using\n that scheme even when TLS 1.2 is negotiated. In TLS 1.2,\n RSASSA-PSS is used with RSA cipher suites.\n\n\n\nRescorla Standards Track [Page 44]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.4. Certificate Authorities\n\n The \"certificate_authorities\" extension is used to indicate the\n certificate authorities (CAs) which an endpoint supports and which\n SHOULD be used by the receiving endpoint to guide certificate\n selection.\n\n The body of the \"certificate_authorities\" extension consists of a\n CertificateAuthoritiesExtension structure.\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n DistinguishedName authorities<3..2^16-1>;\n } CertificateAuthoritiesExtension;\n\n authorities: A list of the distinguished names [X501] of acceptable\n certificate authorities, represented in DER-encoded [X690] format.\n These distinguished names specify a desired distinguished name for\n a trust anchor or subordinate CA; thus, this message can be used\n to describe known trust anchors as well as a desired authorization\n space.\n\n The client MAY send the \"certificate_authorities\" extension in the\n ClientHello message. The server MAY send it in the\n CertificateRequest message.\n\n The \"trusted_ca_keys\" extension [RFC6066], which serves a similar\n purpose but is more complicated, is not used in TLS 1.3 (although it\n may appear in ClientHello messages from clients which are offering\n prior versions of TLS).\n\n4.2.5. OID Filters\n\n The \"oid_filters\" extension allows servers to provide a set of\n OID/value pairs which it would like the client's certificate to\n match. This extension, if provided by the server, MUST only be sent\n in the CertificateRequest message.\n\n struct {\n opaque certificate_extension_oid<1..2^8-1>;\n opaque certificate_extension_values<0..2^16-1>;\n } OIDFilter;\n\n struct {\n OIDFilter filters<0..2^16-1>;\n } OIDFilterExtension;\n\n\n\n\nRescorla Standards Track [Page 45]\n\f\nRFC 8446 TLS August 2018\n\n\n filters: A list of certificate extension OIDs [RFC5280] with their\n allowed value(s) and represented in DER-encoded [X690] format.\n Some certificate extension OIDs allow multiple values (e.g.,\n Extended Key Usage). If the server has included a non-empty\n filters list, the client certificate included in the response MUST\n contain all of the specified extension OIDs that the client\n recognizes. For each extension OID recognized by the client, all\n of the specified values MUST be present in the client certificate\n (but the certificate MAY have other values as well). However, the\n client MUST ignore and skip any unrecognized certificate extension\n OIDs. If the client ignored some of the required certificate\n extension OIDs and supplied a certificate that does not satisfy\n the request, the server MAY at its discretion either continue the\n connection without client authentication or abort the handshake\n with an \"unsupported_certificate\" alert. Any given OID MUST NOT\n appear more than once in the filters list.\n\n PKIX RFCs define a variety of certificate extension OIDs and their\n corresponding value types. Depending on the type, matching\n certificate extension values are not necessarily bitwise-equal. It\n is expected that TLS implementations will rely on their PKI libraries\n to perform certificate selection using certificate extension OIDs.\n\n This document defines matching rules for two standard certificate\n extensions defined in [RFC5280]:\n\n - The Key Usage extension in a certificate matches the request when\n all key usage bits asserted in the request are also asserted in\n the Key Usage certificate extension.\n\n - The Extended Key Usage extension in a certificate matches the\n request when all key purpose OIDs present in the request are also\n found in the Extended Key Usage certificate extension. The\n special anyExtendedKeyUsage OID MUST NOT be used in the request.\n\n Separate specifications may define matching rules for other\n certificate extensions.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 46]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.6. Post-Handshake Client Authentication\n\n The \"post_handshake_auth\" extension is used to indicate that a client\n is willing to perform post-handshake authentication (Section 4.6.2).\n Servers MUST NOT send a post-handshake CertificateRequest to clients\n which do not offer this extension. Servers MUST NOT send this\n extension.\n\n struct {} PostHandshakeAuth;\n\n The \"extension_data\" field of the \"post_handshake_auth\" extension is\n zero length.\n\n4.2.7. Supported Groups\n\n When sent by the client, the \"supported_groups\" extension indicates\n the named groups which the client supports for key exchange, ordered\n from most preferred to least preferred.\n\n Note: In versions of TLS prior to TLS 1.3, this extension was named\n \"elliptic_curves\" and only contained elliptic curve groups. See\n [RFC8422] and [RFC7919]. This extension was also used to negotiate\n ECDSA curves. Signature algorithms are now negotiated independently\n (see Section 4.2.3).\n\n The \"extension_data\" field of this extension contains a\n \"NamedGroupList\" value:\n\n enum {\n\n /* Elliptic Curve Groups (ECDHE) */\n secp256r1(0x0017), secp384r1(0x0018), secp521r1(0x0019),\n x25519(0x001D), x448(0x001E),\n\n /* Finite Field Groups (DHE) */\n ffdhe2048(0x0100), ffdhe3072(0x0101), ffdhe4096(0x0102),\n ffdhe6144(0x0103), ffdhe8192(0x0104),\n\n /* Reserved Code Points */\n ffdhe_private_use(0x01FC..0x01FF),\n ecdhe_private_use(0xFE00..0xFEFF),\n (0xFFFF)\n } NamedGroup;\n\n struct {\n NamedGroup named_group_list<2..2^16-1>;\n } NamedGroupList;\n\n\n\n\nRescorla Standards Track [Page 47]\n\f\nRFC 8446 TLS August 2018\n\n\n Elliptic Curve Groups (ECDHE): Indicates support for the\n corresponding named curve, defined in either FIPS 186-4 [DSS] or\n [RFC7748]. Values 0xFE00 through 0xFEFF are reserved for\n Private Use [RFC8126].\n\n Finite Field Groups (DHE): Indicates support for the corresponding\n finite field group, defined in [RFC7919]. Values 0x01FC through\n 0x01FF are reserved for Private Use.\n\n Items in named_group_list are ordered according to the sender's\n preferences (most preferred choice first).\n\n As of TLS 1.3, servers are permitted to send the \"supported_groups\"\n extension to the client. Clients MUST NOT act upon any information\n found in \"supported_groups\" prior to successful completion of the\n handshake but MAY use the information learned from a successfully\n completed handshake to change what groups they use in their\n \"key_share\" extension in subsequent connections. If the server has a\n group it prefers to the ones in the \"key_share\" extension but is\n still willing to accept the ClientHello, it SHOULD send\n \"supported_groups\" to update the client's view of its preferences;\n this extension SHOULD contain all groups the server supports,\n regardless of whether they are currently supported by the client.\n\n4.2.8. Key Share\n\n The \"key_share\" extension contains the endpoint's cryptographic\n parameters.\n\n Clients MAY send an empty client_shares vector in order to request\n group selection from the server, at the cost of an additional round\n trip (see Section 4.1.4).\n\n struct {\n NamedGroup group;\n opaque key_exchange<1..2^16-1>;\n } KeyShareEntry;\n\n group: The named group for the key being exchanged.\n\n key_exchange: Key exchange information. The contents of this field\n are determined by the specified group and its corresponding\n definition. Finite Field Diffie-Hellman [DH76] parameters are\n described in Section 4.2.8.1; Elliptic Curve Diffie-Hellman\n parameters are described in Section 4.2.8.2.\n\n\n\n\n\n\nRescorla Standards Track [Page 48]\n\f\nRFC 8446 TLS August 2018\n\n\n In the ClientHello message, the \"extension_data\" field of this\n extension contains a \"KeyShareClientHello\" value:\n\n struct {\n KeyShareEntry client_shares<0..2^16-1>;\n } KeyShareClientHello;\n\n client_shares: A list of offered KeyShareEntry values in descending\n order of client preference.\n\n This vector MAY be empty if the client is requesting a\n HelloRetryRequest. Each KeyShareEntry value MUST correspond to a\n group offered in the \"supported_groups\" extension and MUST appear in\n the same order. However, the values MAY be a non-contiguous subset\n of the \"supported_groups\" extension and MAY omit the most preferred\n groups. Such a situation could arise if the most preferred groups\n are new and unlikely to be supported in enough places to make\n pregenerating key shares for them efficient.\n\n Clients can offer as many KeyShareEntry values as the number of\n supported groups it is offering, each representing a single set of\n key exchange parameters. For instance, a client might offer shares\n for several elliptic curves or multiple FFDHE groups. The\n key_exchange values for each KeyShareEntry MUST be generated\n independently. Clients MUST NOT offer multiple KeyShareEntry values\n for the same group. Clients MUST NOT offer any KeyShareEntry values\n for groups not listed in the client's \"supported_groups\" extension.\n Servers MAY check for violations of these rules and abort the\n handshake with an \"illegal_parameter\" alert if one is violated.\n\n In a HelloRetryRequest message, the \"extension_data\" field of this\n extension contains a KeyShareHelloRetryRequest value:\n\n struct {\n NamedGroup selected_group;\n } KeyShareHelloRetryRequest;\n\n selected_group: The mutually supported group the server intends to\n negotiate and is requesting a retried ClientHello/KeyShare for.\n\n Upon receipt of this extension in a HelloRetryRequest, the client\n MUST verify that (1) the selected_group field corresponds to a group\n which was provided in the \"supported_groups\" extension in the\n original ClientHello and (2) the selected_group field does not\n correspond to a group which was provided in the \"key_share\" extension\n in the original ClientHello. If either of these checks fails, then\n the client MUST abort the handshake with an \"illegal_parameter\"\n alert. Otherwise, when sending the new ClientHello, the client MUST\n\n\n\nRescorla Standards Track [Page 49]\n\f\nRFC 8446 TLS August 2018\n\n\n replace the original \"key_share\" extension with one containing only a\n new KeyShareEntry for the group indicated in the selected_group field\n of the triggering HelloRetryRequest.\n\n In a ServerHello message, the \"extension_data\" field of this\n extension contains a KeyShareServerHello value:\n\n struct {\n KeyShareEntry server_share;\n } KeyShareServerHello;\n\n server_share: A single KeyShareEntry value that is in the same group\n as one of the client's shares.\n\n If using (EC)DHE key establishment, servers offer exactly one\n KeyShareEntry in the ServerHello. This value MUST be in the same\n group as the KeyShareEntry value offered by the client that the\n server has selected for the negotiated key exchange. Servers\n MUST NOT send a KeyShareEntry for any group not indicated in the\n client's \"supported_groups\" extension and MUST NOT send a\n KeyShareEntry when using the \"psk_ke\" PskKeyExchangeMode. If using\n (EC)DHE key establishment and a HelloRetryRequest containing a\n \"key_share\" extension was received by the client, the client MUST\n verify that the selected NamedGroup in the ServerHello is the same as\n that in the HelloRetryRequest. If this check fails, the client MUST\n abort the handshake with an \"illegal_parameter\" alert.\n\n4.2.8.1. Diffie-Hellman Parameters\n\n Diffie-Hellman [DH76] parameters for both clients and servers are\n encoded in the opaque key_exchange field of a KeyShareEntry in a\n KeyShare structure. The opaque value contains the Diffie-Hellman\n public value (Y = g^X mod p) for the specified group (see [RFC7919]\n for group definitions) encoded as a big-endian integer and padded to\n the left with zeros to the size of p in bytes.\n\n Note: For a given Diffie-Hellman group, the padding results in all\n public keys having the same length.\n\n Peers MUST validate each other's public key Y by ensuring that 1 < Y\n < p-1. This check ensures that the remote peer is properly behaved\n and isn't forcing the local system into a small subgroup.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 50]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.8.2. ECDHE Parameters\n\n ECDHE parameters for both clients and servers are encoded in the\n opaque key_exchange field of a KeyShareEntry in a KeyShare structure.\n\n For secp256r1, secp384r1, and secp521r1, the contents are the\n serialized value of the following struct:\n\n struct {\n uint8 legacy_form = 4;\n opaque X[coordinate_length];\n opaque Y[coordinate_length];\n } UncompressedPointRepresentation;\n\n X and Y, respectively, are the binary representations of the x and y\n values in network byte order. There are no internal length markers,\n so each number representation occupies as many octets as implied by\n the curve parameters. For P-256, this means that each of X and Y use\n 32 octets, padded on the left by zeros if necessary. For P-384, they\n take 48 octets each. For P-521, they take 66 octets each.\n\n For the curves secp256r1, secp384r1, and secp521r1, peers MUST\n validate each other's public value Q by ensuring that the point is a\n valid point on the elliptic curve. The appropriate validation\n procedures are defined in Section 4.3.7 of [ECDSA] and alternatively\n in Section 5.6.2.3 of [KEYAGREEMENT]. This process consists of three\n steps: (1) verify that Q is not the point at infinity (O), (2) verify\n that for Q = (x, y) both integers x and y are in the correct\n interval, and (3) ensure that (x, y) is a correct solution to the\n elliptic curve equation. For these curves, implementors do not need\n to verify membership in the correct subgroup.\n\n For X25519 and X448, the contents of the public value are the byte\n string inputs and outputs of the corresponding functions defined in\n [RFC7748]: 32 bytes for X25519 and 56 bytes for X448.\n\n Note: Versions of TLS prior to 1.3 permitted point format\n negotiation; TLS 1.3 removes this feature in favor of a single point\n format for each curve.\n\n4.2.9. Pre-Shared Key Exchange Modes\n\n In order to use PSKs, clients MUST also send a\n \"psk_key_exchange_modes\" extension. The semantics of this extension\n are that the client only supports the use of PSKs with these modes,\n which restricts both the use of PSKs offered in this ClientHello and\n those which the server might supply via NewSessionTicket.\n\n\n\n\nRescorla Standards Track [Page 51]\n\f\nRFC 8446 TLS August 2018\n\n\n A client MUST provide a \"psk_key_exchange_modes\" extension if it\n offers a \"pre_shared_key\" extension. If clients offer\n \"pre_shared_key\" without a \"psk_key_exchange_modes\" extension,\n servers MUST abort the handshake. Servers MUST NOT select a key\n exchange mode that is not listed by the client. This extension also\n restricts the modes for use with PSK resumption. Servers SHOULD NOT\n send NewSessionTicket with tickets that are not compatible with the\n advertised modes; however, if a server does so, the impact will just\n be that the client's attempts at resumption fail.\n\n The server MUST NOT send a \"psk_key_exchange_modes\" extension.\n\n enum { psk_ke(0), psk_dhe_ke(1), (255) } PskKeyExchangeMode;\n\n struct {\n PskKeyExchangeMode ke_modes<1..255>;\n } PskKeyExchangeModes;\n\n psk_ke: PSK-only key establishment. In this mode, the server\n MUST NOT supply a \"key_share\" value.\n\n psk_dhe_ke: PSK with (EC)DHE key establishment. In this mode, the\n client and server MUST supply \"key_share\" values as described in\n Section 4.2.8.\n\n Any future values that are allocated must ensure that the transmitted\n protocol messages unambiguously identify which mode was selected by\n the server; at present, this is indicated by the presence of the\n \"key_share\" in the ServerHello.\n\n4.2.10. Early Data Indication\n\n When a PSK is used and early data is allowed for that PSK, the client\n can send Application Data in its first flight of messages. If the\n client opts to do so, it MUST supply both the \"pre_shared_key\" and\n \"early_data\" extensions.\n\n The \"extension_data\" field of this extension contains an\n \"EarlyDataIndication\" value.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 52]\n\f\nRFC 8446 TLS August 2018\n\n\n struct {} Empty;\n\n struct {\n select (Handshake.msg_type) {\n case new_session_ticket: uint32 max_early_data_size;\n case client_hello: Empty;\n case encrypted_extensions: Empty;\n };\n } EarlyDataIndication;\n\n See Section 4.6.1 for details regarding the use of the\n max_early_data_size field.\n\n The parameters for the 0-RTT data (version, symmetric cipher suite,\n Application-Layer Protocol Negotiation (ALPN) [RFC7301] protocol,\n etc.) are those associated with the PSK in use. For externally\n provisioned PSKs, the associated values are those provisioned along\n with the key. For PSKs established via a NewSessionTicket message,\n the associated values are those which were negotiated in the\n connection which established the PSK. The PSK used to encrypt the\n early data MUST be the first PSK listed in the client's\n \"pre_shared_key\" extension.\n\n For PSKs provisioned via NewSessionTicket, a server MUST validate\n that the ticket age for the selected PSK identity (computed by\n subtracting ticket_age_add from PskIdentity.obfuscated_ticket_age\n modulo 2^32) is within a small tolerance of the time since the ticket\n was issued (see Section 8). If it is not, the server SHOULD proceed\n with the handshake but reject 0-RTT, and SHOULD NOT take any other\n action that assumes that this ClientHello is fresh.\n\n 0-RTT messages sent in the first flight have the same (encrypted)\n content types as messages of the same type sent in other flights\n (handshake and application_data) but are protected under different\n keys. After receiving the server's Finished message, if the server\n has accepted early data, an EndOfEarlyData message will be sent to\n indicate the key change. This message will be encrypted with the\n 0-RTT traffic keys.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 53]\n\f\nRFC 8446 TLS August 2018\n\n\n A server which receives an \"early_data\" extension MUST behave in one\n of three ways:\n\n - Ignore the extension and return a regular 1-RTT response. The\n server then skips past early data by attempting to deprotect\n received records using the handshake traffic key, discarding\n records which fail deprotection (up to the configured\n max_early_data_size). Once a record is deprotected successfully,\n it is treated as the start of the client's second flight and the\n server proceeds as with an ordinary 1-RTT handshake.\n\n - Request that the client send another ClientHello by responding\n with a HelloRetryRequest. A client MUST NOT include the\n \"early_data\" extension in its followup ClientHello. The server\n then ignores early data by skipping all records with an external\n content type of \"application_data\" (indicating that they are\n encrypted), up to the configured max_early_data_size.\n\n - Return its own \"early_data\" extension in EncryptedExtensions,\n indicating that it intends to process the early data. It is not\n possible for the server to accept only a subset of the early data\n messages. Even though the server sends a message accepting early\n data, the actual early data itself may already be in flight by the\n time the server generates this message.\n\n In order to accept early data, the server MUST have accepted a PSK\n cipher suite and selected the first key offered in the client's\n \"pre_shared_key\" extension. In addition, it MUST verify that the\n following values are the same as those associated with the\n selected PSK:\n\n - The TLS version number\n\n - The selected cipher suite\n\n - The selected ALPN [RFC7301] protocol, if any\n\n These requirements are a superset of those needed to perform a 1-RTT\n handshake using the PSK in question. For externally established\n PSKs, the associated values are those provisioned along with the key.\n For PSKs established via a NewSessionTicket message, the associated\n values are those negotiated in the connection during which the ticket\n was established.\n\n Future extensions MUST define their interaction with 0-RTT.\n\n\n\n\n\n\nRescorla Standards Track [Page 54]\n\f\nRFC 8446 TLS August 2018\n\n\n If any of these checks fail, the server MUST NOT respond with the\n extension and must discard all the first-flight data using one of the\n first two mechanisms listed above (thus falling back to 1-RTT or\n 2-RTT). If the client attempts a 0-RTT handshake but the server\n rejects it, the server will generally not have the 0-RTT record\n protection keys and must instead use trial decryption (either with\n the 1-RTT handshake keys or by looking for a cleartext ClientHello in\n the case of a HelloRetryRequest) to find the first non-0-RTT message.\n\n If the server chooses to accept the \"early_data\" extension, then it\n MUST comply with the same error-handling requirements specified for\n all records when processing early data records. Specifically, if the\n server fails to decrypt a 0-RTT record following an accepted\n \"early_data\" extension, it MUST terminate the connection with a\n \"bad_record_mac\" alert as per Section 5.2.\n\n If the server rejects the \"early_data\" extension, the client\n application MAY opt to retransmit the Application Data previously\n sent in early data once the handshake has been completed. Note that\n automatic retransmission of early data could result in incorrect\n assumptions regarding the status of the connection. For instance,\n when the negotiated connection selects a different ALPN protocol from\n what was used for the early data, an application might need to\n construct different messages. Similarly, if early data assumes\n anything about the connection state, it might be sent in error after\n the handshake completes.\n\n A TLS implementation SHOULD NOT automatically resend early data;\n applications are in a better position to decide when retransmission\n is appropriate. A TLS implementation MUST NOT automatically resend\n early data unless the negotiated connection selects the same ALPN\n protocol.\n\n4.2.11. Pre-Shared Key Extension\n\n The \"pre_shared_key\" extension is used to negotiate the identity of\n the pre-shared key to be used with a given handshake in association\n with PSK key establishment.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 55]\n\f\nRFC 8446 TLS August 2018\n\n\n The \"extension_data\" field of this extension contains a\n \"PreSharedKeyExtension\" value:\n\n struct {\n opaque identity<1..2^16-1>;\n uint32 obfuscated_ticket_age;\n } PskIdentity;\n\n opaque PskBinderEntry<32..255>;\n\n struct {\n PskIdentity identities<7..2^16-1>;\n PskBinderEntry binders<33..2^16-1>;\n } OfferedPsks;\n\n struct {\n select (Handshake.msg_type) {\n case client_hello: OfferedPsks;\n case server_hello: uint16 selected_identity;\n };\n } PreSharedKeyExtension;\n\n identity: A label for a key. For instance, a ticket (as defined in\n Appendix B.3.4) or a label for a pre-shared key established\n externally.\n\n obfuscated_ticket_age: An obfuscated version of the age of the key.\n Section 4.2.11.1 describes how to form this value for identities\n established via the NewSessionTicket message. For identities\n established externally, an obfuscated_ticket_age of 0 SHOULD be\n used, and servers MUST ignore the value.\n\n identities: A list of the identities that the client is willing to\n negotiate with the server. If sent alongside the \"early_data\"\n extension (see Section 4.2.10), the first identity is the one used\n for 0-RTT data.\n\n binders: A series of HMAC values, one for each value in the\n identities list and in the same order, computed as described\n below.\n\n selected_identity: The server's chosen identity expressed as a\n (0-based) index into the identities in the client's list.\n\n Each PSK is associated with a single Hash algorithm. For PSKs\n established via the ticket mechanism (Section 4.6.1), this is the KDF\n Hash algorithm on the connection where the ticket was established.\n For externally established PSKs, the Hash algorithm MUST be set when\n\n\n\nRescorla Standards Track [Page 56]\n\f\nRFC 8446 TLS August 2018\n\n\n the PSK is established or default to SHA-256 if no such algorithm is\n defined. The server MUST ensure that it selects a compatible PSK\n (if any) and cipher suite.\n\n In TLS versions prior to TLS 1.3, the Server Name Identification\n (SNI) value was intended to be associated with the session (Section 3\n of [RFC6066]), with the server being required to enforce that the SNI\n value associated with the session matches the one specified in the\n resumption handshake. However, in reality the implementations were\n not consistent on which of two supplied SNI values they would use,\n leading to the consistency requirement being de facto enforced by the\n clients. In TLS 1.3, the SNI value is always explicitly specified in\n the resumption handshake, and there is no need for the server to\n associate an SNI value with the ticket. Clients, however, SHOULD\n store the SNI with the PSK to fulfill the requirements of\n Section 4.6.1.\n\n Implementor's note: When session resumption is the primary use case\n of PSKs, the most straightforward way to implement the PSK/cipher\n suite matching requirements is to negotiate the cipher suite first\n and then exclude any incompatible PSKs. Any unknown PSKs (e.g., ones\n not in the PSK database or encrypted with an unknown key) SHOULD\n simply be ignored. If no acceptable PSKs are found, the server\n SHOULD perform a non-PSK handshake if possible. If backward\n compatibility is important, client-provided, externally established\n PSKs SHOULD influence cipher suite selection.\n\n Prior to accepting PSK key establishment, the server MUST validate\n the corresponding binder value (see Section 4.2.11.2 below). If this\n value is not present or does not validate, the server MUST abort the\n handshake. Servers SHOULD NOT attempt to validate multiple binders;\n rather, they SHOULD select a single PSK and validate solely the\n binder that corresponds to that PSK. See Section 8.2 and\n Appendix E.6 for the security rationale for this requirement. In\n order to accept PSK key establishment, the server sends a\n \"pre_shared_key\" extension indicating the selected identity.\n\n Clients MUST verify that the server's selected_identity is within the\n range supplied by the client, that the server selected a cipher suite\n indicating a Hash associated with the PSK, and that a server\n \"key_share\" extension is present if required by the ClientHello\n \"psk_key_exchange_modes\" extension. If these values are not\n consistent, the client MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 57]\n\f\nRFC 8446 TLS August 2018\n\n\n If the server supplies an \"early_data\" extension, the client MUST\n verify that the server's selected_identity is 0. If any other value\n is returned, the client MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n The \"pre_shared_key\" extension MUST be the last extension in the\n ClientHello (this facilitates implementation as described below).\n Servers MUST check that it is the last extension and otherwise fail\n the handshake with an \"illegal_parameter\" alert.\n\n4.2.11.1. Ticket Age\n\n The client's view of the age of a ticket is the time since the\n receipt of the NewSessionTicket message. Clients MUST NOT attempt to\n use tickets which have ages greater than the \"ticket_lifetime\" value\n which was provided with the ticket. The \"obfuscated_ticket_age\"\n field of each PskIdentity contains an obfuscated version of the\n ticket age formed by taking the age in milliseconds and adding the\n \"ticket_age_add\" value that was included with the ticket (see\n Section 4.6.1), modulo 2^32. This addition prevents passive\n observers from correlating connections unless tickets are reused.\n Note that the \"ticket_lifetime\" field in the NewSessionTicket message\n is in seconds but the \"obfuscated_ticket_age\" is in milliseconds.\n Because ticket lifetimes are restricted to a week, 32 bits is enough\n to represent any plausible age, even in milliseconds.\n\n4.2.11.2. PSK Binder\n\n The PSK binder value forms a binding between a PSK and the current\n handshake, as well as a binding between the handshake in which the\n PSK was generated (if via a NewSessionTicket message) and the current\n handshake. Each entry in the binders list is computed as an HMAC\n over a transcript hash (see Section 4.4.1) containing a partial\n ClientHello up to and including the PreSharedKeyExtension.identities\n field. That is, it includes all of the ClientHello but not the\n binders list itself. The length fields for the message (including\n the overall length, the length of the extensions block, and the\n length of the \"pre_shared_key\" extension) are all set as if binders\n of the correct lengths were present.\n\n The PskBinderEntry is computed in the same way as the Finished\n message (Section 4.4.4) but with the BaseKey being the binder_key\n derived via the key schedule from the corresponding PSK which is\n being offered (see Section 7.1).\n\n\n\n\n\n\n\nRescorla Standards Track [Page 58]\n\f\nRFC 8446 TLS August 2018\n\n\n If the handshake includes a HelloRetryRequest, the initial\n ClientHello and HelloRetryRequest are included in the transcript\n along with the new ClientHello. For instance, if the client sends\n ClientHello1, its binder will be computed over:\n\n Transcript-Hash(Truncate(ClientHello1))\n\n Where Truncate() removes the binders list from the ClientHello.\n\n If the server responds with a HelloRetryRequest and the client then\n sends ClientHello2, its binder will be computed over:\n\n Transcript-Hash(ClientHello1,\n HelloRetryRequest,\n Truncate(ClientHello2))\n\n The full ClientHello1/ClientHello2 is included in all other handshake\n hash computations. Note that in the first flight,\n Truncate(ClientHello1) is hashed directly, but in the second flight,\n ClientHello1 is hashed and then reinjected as a \"message_hash\"\n message, as described in Section 4.4.1.\n\n4.2.11.3. Processing Order\n\n Clients are permitted to \"stream\" 0-RTT data until they receive the\n server's Finished, only then sending the EndOfEarlyData message,\n followed by the rest of the handshake. In order to avoid deadlocks,\n when accepting \"early_data\", servers MUST process the client's\n ClientHello and then immediately send their flight of messages,\n rather than waiting for the client's EndOfEarlyData message before\n sending its ServerHello.\n\n4.3. Server Parameters\n\n The next two messages from the server, EncryptedExtensions and\n CertificateRequest, contain information from the server that\n determines the rest of the handshake. These messages are encrypted\n with keys derived from the server_handshake_traffic_secret.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 59]\n\f\nRFC 8446 TLS August 2018\n\n\n4.3.1. Encrypted Extensions\n\n In all handshakes, the server MUST send the EncryptedExtensions\n message immediately after the ServerHello message. This is the first\n message that is encrypted under keys derived from the\n server_handshake_traffic_secret.\n\n The EncryptedExtensions message contains extensions that can be\n protected, i.e., any which are not needed to establish the\n cryptographic context but which are not associated with individual\n certificates. The client MUST check EncryptedExtensions for the\n presence of any forbidden extensions and if any are found MUST abort\n the handshake with an \"illegal_parameter\" alert.\n\n Structure of this message:\n\n struct {\n Extension extensions<0..2^16-1>;\n } EncryptedExtensions;\n\n extensions: A list of extensions. For more information, see the\n table in Section 4.2.\n\n4.3.2. Certificate Request\n\n A server which is authenticating with a certificate MAY optionally\n request a certificate from the client. This message, if sent, MUST\n follow EncryptedExtensions.\n\n Structure of this message:\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n Extension extensions<2..2^16-1>;\n } CertificateRequest;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 60]\n\f\nRFC 8446 TLS August 2018\n\n\n certificate_request_context: An opaque string which identifies the\n certificate request and which will be echoed in the client's\n Certificate message. The certificate_request_context MUST be\n unique within the scope of this connection (thus preventing replay\n of client CertificateVerify messages). This field SHALL be zero\n length unless used for the post-handshake authentication exchanges\n described in Section 4.6.2. When requesting post-handshake\n authentication, the server SHOULD make the context unpredictable\n to the client (e.g., by randomly generating it) in order to\n prevent an attacker who has temporary access to the client's\n private key from pre-computing valid CertificateVerify messages.\n\n extensions: A set of extensions describing the parameters of the\n certificate being requested. The \"signature_algorithms\" extension\n MUST be specified, and other extensions may optionally be included\n if defined for this message. Clients MUST ignore unrecognized\n extensions.\n\n In prior versions of TLS, the CertificateRequest message carried a\n list of signature algorithms and certificate authorities which the\n server would accept. In TLS 1.3, the former is expressed by sending\n the \"signature_algorithms\" and optionally \"signature_algorithms_cert\"\n extensions. The latter is expressed by sending the\n \"certificate_authorities\" extension (see Section 4.2.4).\n\n Servers which are authenticating with a PSK MUST NOT send the\n CertificateRequest message in the main handshake, though they MAY\n send it in post-handshake authentication (see Section 4.6.2) provided\n that the client has sent the \"post_handshake_auth\" extension (see\n Section 4.2.6).\n\n4.4. Authentication Messages\n\n As discussed in Section 2, TLS generally uses a common set of\n messages for authentication, key confirmation, and handshake\n integrity: Certificate, CertificateVerify, and Finished. (The PSK\n binders also perform key confirmation, in a similar fashion.) These\n three messages are always sent as the last messages in their\n handshake flight. The Certificate and CertificateVerify messages are\n only sent under certain circumstances, as defined below. The\n Finished message is always sent as part of the Authentication Block.\n These messages are encrypted under keys derived from the\n [sender]_handshake_traffic_secret.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 61]\n\f\nRFC 8446 TLS August 2018\n\n\n The computations for the Authentication messages all uniformly take\n the following inputs:\n\n - The certificate and signing key to be used.\n\n - A Handshake Context consisting of the set of messages to be\n included in the transcript hash.\n\n - A Base Key to be used to compute a MAC key.\n\n Based on these inputs, the messages then contain:\n\n Certificate: The certificate to be used for authentication, and any\n supporting certificates in the chain. Note that certificate-based\n client authentication is not available in PSK handshake flows\n (including 0-RTT).\n\n CertificateVerify: A signature over the value\n Transcript-Hash(Handshake Context, Certificate).\n\n Finished: A MAC over the value Transcript-Hash(Handshake Context,\n Certificate, CertificateVerify) using a MAC key derived from the\n Base Key.\n\n The following table defines the Handshake Context and MAC Base Key\n for each scenario:\n\n +-----------+-------------------------+-----------------------------+\n | Mode | Handshake Context | Base Key |\n +-----------+-------------------------+-----------------------------+\n | Server | ClientHello ... later | server_handshake_traffic_ |\n | | of EncryptedExtensions/ | secret |\n | | CertificateRequest | |\n | | | |\n | Client | ClientHello ... later | client_handshake_traffic_ |\n | | of server | secret |\n | | Finished/EndOfEarlyData | |\n | | | |\n | Post- | ClientHello ... client | client_application_traffic_ |\n | Handshake | Finished + | secret_N |\n | | CertificateRequest | |\n +-----------+-------------------------+-----------------------------+\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 62]\n\f\nRFC 8446 TLS August 2018\n\n\n4.4.1. The Transcript Hash\n\n Many of the cryptographic computations in TLS make use of a\n transcript hash. This value is computed by hashing the concatenation\n of each included handshake message, including the handshake message\n header carrying the handshake message type and length fields, but not\n including record layer headers. I.e.,\n\n Transcript-Hash(M1, M2, ... Mn) = Hash(M1 || M2 || ... || Mn)\n\n As an exception to this general rule, when the server responds to a\n ClientHello with a HelloRetryRequest, the value of ClientHello1 is\n replaced with a special synthetic handshake message of handshake type\n \"message_hash\" containing Hash(ClientHello1). I.e.,\n\n Transcript-Hash(ClientHello1, HelloRetryRequest, ... Mn) =\n Hash(message_hash || /* Handshake type */\n 00 00 Hash.length || /* Handshake message length (bytes) */\n Hash(ClientHello1) || /* Hash of ClientHello1 */\n HelloRetryRequest || ... || Mn)\n\n The reason for this construction is to allow the server to do a\n stateless HelloRetryRequest by storing just the hash of ClientHello1\n in the cookie, rather than requiring it to export the entire\n intermediate hash state (see Section 4.2.2).\n\n For concreteness, the transcript hash is always taken from the\n following sequence of handshake messages, starting at the first\n ClientHello and including only those messages that were sent:\n ClientHello, HelloRetryRequest, ClientHello, ServerHello,\n EncryptedExtensions, server CertificateRequest, server Certificate,\n server CertificateVerify, server Finished, EndOfEarlyData, client\n Certificate, client CertificateVerify, client Finished.\n\n In general, implementations can implement the transcript by keeping a\n running transcript hash value based on the negotiated hash. Note,\n however, that subsequent post-handshake authentications do not\n include each other, just the messages through the end of the main\n handshake.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 63]\n\f\nRFC 8446 TLS August 2018\n\n\n4.4.2. Certificate\n\n This message conveys the endpoint's certificate chain to the peer.\n\n The server MUST send a Certificate message whenever the agreed-upon\n key exchange method uses certificates for authentication (this\n includes all key exchange methods defined in this document\n except PSK).\n\n The client MUST send a Certificate message if and only if the server\n has requested client authentication via a CertificateRequest message\n (Section 4.3.2). If the server requests client authentication but no\n suitable certificate is available, the client MUST send a Certificate\n message containing no certificates (i.e., with the \"certificate_list\"\n field having length 0). A Finished message MUST be sent regardless\n of whether the Certificate message is empty.\n\n Structure of this message:\n\n enum {\n X509(0),\n RawPublicKey(2),\n (255)\n } CertificateType;\n\n struct {\n select (certificate_type) {\n case RawPublicKey:\n /* From RFC 7250 ASN.1_subjectPublicKeyInfo */\n opaque ASN1_subjectPublicKeyInfo<1..2^24-1>;\n\n case X509:\n opaque cert_data<1..2^24-1>;\n };\n Extension extensions<0..2^16-1>;\n } CertificateEntry;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n CertificateEntry certificate_list<0..2^24-1>;\n } Certificate;\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 64]\n\f\nRFC 8446 TLS August 2018\n\n\n certificate_request_context: If this message is in response to a\n CertificateRequest, the value of certificate_request_context in\n that message. Otherwise (in the case of server authentication),\n this field SHALL be zero length.\n\n certificate_list: A sequence (chain) of CertificateEntry structures,\n each containing a single certificate and set of extensions.\n\n extensions: A set of extension values for the CertificateEntry. The\n \"Extension\" format is defined in Section 4.2. Valid extensions\n for server certificates at present include the OCSP Status\n extension [RFC6066] and the SignedCertificateTimestamp extension\n [RFC6962]; future extensions may be defined for this message as\n well. Extensions in the Certificate message from the server MUST\n correspond to ones from the ClientHello message. Extensions in\n the Certificate message from the client MUST correspond to\n extensions in the CertificateRequest message from the server. If\n an extension applies to the entire chain, it SHOULD be included in\n the first CertificateEntry.\n\n If the corresponding certificate type extension\n (\"server_certificate_type\" or \"client_certificate_type\") was not\n negotiated in EncryptedExtensions, or the X.509 certificate type was\n negotiated, then each CertificateEntry contains a DER-encoded X.509\n certificate. The sender's certificate MUST come in the first\n CertificateEntry in the list. Each following certificate SHOULD\n directly certify the one immediately preceding it. Because\n certificate validation requires that trust anchors be distributed\n independently, a certificate that specifies a trust anchor MAY be\n omitted from the chain, provided that supported peers are known to\n possess any omitted certificates.\n\n Note: Prior to TLS 1.3, \"certificate_list\" ordering required each\n certificate to certify the one immediately preceding it; however,\n some implementations allowed some flexibility. Servers sometimes\n send both a current and deprecated intermediate for transitional\n purposes, and others are simply configured incorrectly, but these\n cases can nonetheless be validated properly. For maximum\n compatibility, all implementations SHOULD be prepared to handle\n potentially extraneous certificates and arbitrary orderings from any\n TLS version, with the exception of the end-entity certificate which\n MUST be first.\n\n If the RawPublicKey certificate type was negotiated, then the\n certificate_list MUST contain no more than one CertificateEntry,\n which contains an ASN1_subjectPublicKeyInfo value as defined in\n [RFC7250], Section 3.\n\n\n\n\nRescorla Standards Track [Page 65]\n\f\nRFC 8446 TLS August 2018\n\n\n The OpenPGP certificate type [RFC6091] MUST NOT be used with TLS 1.3.\n\n The server's certificate_list MUST always be non-empty. A client\n will send an empty certificate_list if it does not have an\n appropriate certificate to send in response to the server's\n authentication request.\n\n4.4.2.1. OCSP Status and SCT Extensions\n\n [RFC6066] and [RFC6961] provide extensions to negotiate the server\n sending OCSP responses to the client. In TLS 1.2 and below, the\n server replies with an empty extension to indicate negotiation of\n this extension and the OCSP information is carried in a\n CertificateStatus message. In TLS 1.3, the server's OCSP information\n is carried in an extension in the CertificateEntry containing the\n associated certificate. Specifically, the body of the\n \"status_request\" extension from the server MUST be a\n CertificateStatus structure as defined in [RFC6066], which is\n interpreted as defined in [RFC6960].\n\n Note: The status_request_v2 extension [RFC6961] is deprecated.\n TLS 1.3 servers MUST NOT act upon its presence or information in it\n when processing ClientHello messages; in particular, they MUST NOT\n send the status_request_v2 extension in the EncryptedExtensions,\n CertificateRequest, or Certificate messages. TLS 1.3 servers MUST be\n able to process ClientHello messages that include it, as it MAY be\n sent by clients that wish to use it in earlier protocol versions.\n\n A server MAY request that a client present an OCSP response with its\n certificate by sending an empty \"status_request\" extension in its\n CertificateRequest message. If the client opts to send an OCSP\n response, the body of its \"status_request\" extension MUST be a\n CertificateStatus structure as defined in [RFC6066].\n\n Similarly, [RFC6962] provides a mechanism for a server to send a\n Signed Certificate Timestamp (SCT) as an extension in the ServerHello\n in TLS 1.2 and below. In TLS 1.3, the server's SCT information is\n carried in an extension in the CertificateEntry.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 66]\n\f\nRFC 8446 TLS August 2018\n\n\n4.4.2.2. Server Certificate Selection\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3 [RFC5280], unless explicitly\n negotiated otherwise (e.g., [RFC7250]).\n\n - The server's end-entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected authentication\n algorithm from the client's \"signature_algorithms\" extension\n (currently RSA, ECDSA, or EdDSA).\n\n - The certificate MUST allow the key to be used for signing (i.e.,\n the digitalSignature bit MUST be set if the Key Usage extension is\n present) with a signature scheme indicated in the client's\n \"signature_algorithms\"/\"signature_algorithms_cert\" extensions (see\n Section 4.2.3).\n\n - The \"server_name\" [RFC6066] and \"certificate_authorities\"\n extensions are used to guide certificate selection. As servers\n MAY require the presence of the \"server_name\" extension, clients\n SHOULD send this extension, when applicable.\n\n All certificates provided by the server MUST be signed by a signature\n algorithm advertised by the client if it is able to provide such a\n chain (see Section 4.2.3). Certificates that are self-signed or\n certificates that are expected to be trust anchors are not validated\n as part of the chain and therefore MAY be signed with any algorithm.\n\n If the server cannot produce a certificate chain that is signed only\n via the indicated supported algorithms, then it SHOULD continue the\n handshake by sending the client a certificate chain of its choice\n that may include algorithms that are not known to be supported by the\n client. This fallback chain SHOULD NOT use the deprecated SHA-1 hash\n algorithm in general, but MAY do so if the client's advertisement\n permits it, and MUST NOT do so otherwise.\n\n If the client cannot construct an acceptable chain using the provided\n certificates and decides to abort the handshake, then it MUST abort\n the handshake with an appropriate certificate-related alert (by\n default, \"unsupported_certificate\"; see Section 6.2 for more\n information).\n\n If the server has multiple certificates, it chooses one of them based\n on the above-mentioned criteria (in addition to other criteria, such\n as transport-layer endpoint, local configuration, and preferences).\n\n\n\n\n\nRescorla Standards Track [Page 67]\n\f\nRFC 8446 TLS August 2018\n\n\n4.4.2.3. Client Certificate Selection\n\n The following rules apply to certificates sent by the client:\n\n - The certificate type MUST be X.509v3 [RFC5280], unless explicitly\n negotiated otherwise (e.g., [RFC7250]).\n\n - If the \"certificate_authorities\" extension in the\n CertificateRequest message was present, at least one of the\n certificates in the certificate chain SHOULD be issued by one of\n the listed CAs.\n\n - The certificates MUST be signed using an acceptable signature\n algorithm, as described in Section 4.3.2. Note that this relaxes\n the constraints on certificate-signing algorithms found in prior\n versions of TLS.\n\n - If the CertificateRequest message contained a non-empty\n \"oid_filters\" extension, the end-entity certificate MUST match the\n extension OIDs that are recognized by the client, as described in\n Section 4.2.5.\n\n4.4.2.4. Receiving a Certificate Message\n\n In general, detailed certificate validation procedures are out of\n scope for TLS (see [RFC5280]). This section provides TLS-specific\n requirements.\n\n If the server supplies an empty Certificate message, the client MUST\n abort the handshake with a \"decode_error\" alert.\n\n If the client does not send any certificates (i.e., it sends an empty\n Certificate message), the server MAY at its discretion either\n continue the handshake without client authentication or abort the\n handshake with a \"certificate_required\" alert. Also, if some aspect\n of the certificate chain was unacceptable (e.g., it was not signed by\n a known, trusted CA), the server MAY at its discretion either\n continue the handshake (considering the client unauthenticated) or\n abort the handshake.\n\n Any endpoint receiving any certificate which it would need to\n validate using any signature algorithm using an MD5 hash MUST abort\n the handshake with a \"bad_certificate\" alert. SHA-1 is deprecated,\n and it is RECOMMENDED that any endpoint receiving any certificate\n which it would need to validate using any signature algorithm using a\n SHA-1 hash abort the handshake with a \"bad_certificate\" alert. For\n clarity, this means that endpoints can accept these algorithms for\n certificates that are self-signed or are trust anchors.\n\n\n\nRescorla Standards Track [Page 68]\n\f\nRFC 8446 TLS August 2018\n\n\n All endpoints are RECOMMENDED to transition to SHA-256 or better as\n soon as possible to maintain interoperability with implementations\n currently in the process of phasing out SHA-1 support.\n\n Note that a certificate containing a key for one signature algorithm\n MAY be signed using a different signature algorithm (for instance, an\n RSA key signed with an ECDSA key).\n\n4.4.3. Certificate Verify\n\n This message is used to provide explicit proof that an endpoint\n possesses the private key corresponding to its certificate. The\n CertificateVerify message also provides integrity for the handshake\n up to this point. Servers MUST send this message when authenticating\n via a certificate. Clients MUST send this message whenever\n authenticating via a certificate (i.e., when the Certificate message\n is non-empty). When sent, this message MUST appear immediately after\n the Certificate message and immediately prior to the Finished\n message.\n\n Structure of this message:\n\n struct {\n SignatureScheme algorithm;\n opaque signature<0..2^16-1>;\n } CertificateVerify;\n\n The algorithm field specifies the signature algorithm used (see\n Section 4.2.3 for the definition of this type). The signature is a\n digital signature using that algorithm. The content that is covered\n under the signature is the hash output as described in Section 4.4.1,\n namely:\n\n Transcript-Hash(Handshake Context, Certificate)\n\n The digital signature is then computed over the concatenation of:\n\n - A string that consists of octet 32 (0x20) repeated 64 times\n\n - The context string\n\n - A single 0 byte which serves as the separator\n\n - The content to be signed\n\n\n\n\n\n\n\nRescorla Standards Track [Page 69]\n\f\nRFC 8446 TLS August 2018\n\n\n This structure is intended to prevent an attack on previous versions\n of TLS in which the ServerKeyExchange format meant that attackers\n could obtain a signature of a message with a chosen 32-byte prefix\n (ClientHello.random). The initial 64-byte pad clears that prefix\n along with the server-controlled ServerHello.random.\n\n The context string for a server signature is\n \"TLS 1.3, server CertificateVerify\". The context string for a\n client signature is \"TLS 1.3, client CertificateVerify\". It is\n used to provide separation between signatures made in different\n contexts, helping against potential cross-protocol attacks.\n\n For example, if the transcript hash was 32 bytes of 01 (this length\n would make sense for SHA-256), the content covered by the digital\n signature for a server CertificateVerify would be:\n\n 2020202020202020202020202020202020202020202020202020202020202020\n 2020202020202020202020202020202020202020202020202020202020202020\n 544c5320312e332c207365727665722043657274696669636174655665726966\n 79\n 00\n 0101010101010101010101010101010101010101010101010101010101010101\n\n On the sender side, the process for computing the signature field of\n the CertificateVerify message takes as input:\n\n - The content covered by the digital signature\n\n - The private signing key corresponding to the certificate sent in\n the previous message\n\n If the CertificateVerify message is sent by a server, the signature\n algorithm MUST be one offered in the client's \"signature_algorithms\"\n extension unless no valid certificate chain can be produced without\n unsupported algorithms (see Section 4.2.3).\n\n If sent by a client, the signature algorithm used in the signature\n MUST be one of those present in the supported_signature_algorithms\n field of the \"signature_algorithms\" extension in the\n CertificateRequest message.\n\n In addition, the signature algorithm MUST be compatible with the key\n in the sender's end-entity certificate. RSA signatures MUST use an\n RSASSA-PSS algorithm, regardless of whether RSASSA-PKCS1-v1_5\n algorithms appear in \"signature_algorithms\". The SHA-1 algorithm\n MUST NOT be used in any signatures of CertificateVerify messages.\n\n\n\n\n\nRescorla Standards Track [Page 70]\n\f\nRFC 8446 TLS August 2018\n\n\n All SHA-1 signature algorithms in this specification are defined\n solely for use in legacy certificates and are not valid for\n CertificateVerify signatures.\n\n The receiver of a CertificateVerify message MUST verify the signature\n field. The verification process takes as input:\n\n - The content covered by the digital signature\n\n - The public key contained in the end-entity certificate found in\n the associated Certificate message\n\n - The digital signature received in the signature field of the\n CertificateVerify message\n\n If the verification fails, the receiver MUST terminate the handshake\n with a \"decrypt_error\" alert.\n\n4.4.4. Finished\n\n The Finished message is the final message in the Authentication\n Block. It is essential for providing authentication of the handshake\n and of the computed keys.\n\n Recipients of Finished messages MUST verify that the contents are\n correct and if incorrect MUST terminate the connection with a\n \"decrypt_error\" alert.\n\n Once a side has sent its Finished message and has received and\n validated the Finished message from its peer, it may begin to send\n and receive Application Data over the connection. There are two\n settings in which it is permitted to send data prior to receiving the\n peer's Finished:\n\n 1. Clients sending 0-RTT data as described in Section 4.2.10.\n\n 2. Servers MAY send data after sending their first flight, but\n because the handshake is not yet complete, they have no assurance\n of either the peer's identity or its liveness (i.e., the\n ClientHello might have been replayed).\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 71]\n\f\nRFC 8446 TLS August 2018\n\n\n The key used to compute the Finished message is computed from the\n Base Key defined in Section 4.4 using HKDF (see Section 7.1).\n Specifically:\n\n finished_key =\n HKDF-Expand-Label(BaseKey, \"finished\", \"\", Hash.length)\n\n Structure of this message:\n\n struct {\n opaque verify_data[Hash.length];\n } Finished;\n\n The verify_data value is computed as follows:\n\n verify_data =\n HMAC(finished_key,\n Transcript-Hash(Handshake Context,\n Certificate*, CertificateVerify*))\n\n * Only included if present.\n\n HMAC [RFC2104] uses the Hash algorithm for the handshake. As noted\n above, the HMAC input can generally be implemented by a running hash,\n i.e., just the handshake hash at this point.\n\n In previous versions of TLS, the verify_data was always 12 octets\n long. In TLS 1.3, it is the size of the HMAC output for the Hash\n used for the handshake.\n\n Note: Alerts and any other non-handshake record types are not\n handshake messages and are not included in the hash computations.\n\n Any records following a Finished message MUST be encrypted under the\n appropriate application traffic key as described in Section 7.2. In\n particular, this includes any alerts sent by the server in response\n to client Certificate and CertificateVerify messages.\n\n4.5. End of Early Data\n\n struct {} EndOfEarlyData;\n\n If the server sent an \"early_data\" extension in EncryptedExtensions,\n the client MUST send an EndOfEarlyData message after receiving the\n server Finished. If the server does not send an \"early_data\"\n extension in EncryptedExtensions, then the client MUST NOT send an\n EndOfEarlyData message. This message indicates that all 0-RTT\n application_data messages, if any, have been transmitted and that the\n\n\n\nRescorla Standards Track [Page 72]\n\f\nRFC 8446 TLS August 2018\n\n\n following records are protected under handshake traffic keys.\n Servers MUST NOT send this message, and clients receiving it MUST\n terminate the connection with an \"unexpected_message\" alert. This\n message is encrypted under keys derived from the\n client_early_traffic_secret.\n\n4.6. Post-Handshake Messages\n\n TLS also allows other messages to be sent after the main handshake.\n These messages use a handshake content type and are encrypted under\n the appropriate application traffic key.\n\n4.6.1. New Session Ticket Message\n\n At any time after the server has received the client Finished\n message, it MAY send a NewSessionTicket message. This message\n creates a unique association between the ticket value and a secret\n PSK derived from the resumption master secret (see Section 7).\n\n The client MAY use this PSK for future handshakes by including the\n ticket value in the \"pre_shared_key\" extension in its ClientHello\n (Section 4.2.11). Servers MAY send multiple tickets on a single\n connection, either immediately after each other or after specific\n events (see Appendix C.4). For instance, the server might send a new\n ticket after post-handshake authentication in order to encapsulate\n the additional client authentication state. Multiple tickets are\n useful for clients for a variety of purposes, including:\n\n - Opening multiple parallel HTTP connections.\n\n - Performing connection racing across interfaces and address\n families via (for example) Happy Eyeballs [RFC8305] or related\n techniques.\n\n Any ticket MUST only be resumed with a cipher suite that has the same\n KDF hash algorithm as that used to establish the original connection.\n\n Clients MUST only resume if the new SNI value is valid for the server\n certificate presented in the original session and SHOULD only resume\n if the SNI value matches the one used in the original session. The\n latter is a performance optimization: normally, there is no reason to\n expect that different servers covered by a single certificate would\n be able to accept each other's tickets; hence, attempting resumption\n in that case would waste a single-use ticket. If such an indication\n is provided (externally or by any other means), clients MAY resume\n with a different SNI value.\n\n\n\n\n\nRescorla Standards Track [Page 73]\n\f\nRFC 8446 TLS August 2018\n\n\n On resumption, if reporting an SNI value to the calling application,\n implementations MUST use the value sent in the resumption ClientHello\n rather than the value sent in the previous session. Note that if a\n server implementation declines all PSK identities with different SNI\n values, these two values are always the same.\n\n Note: Although the resumption master secret depends on the client's\n second flight, a server which does not request client authentication\n MAY compute the remainder of the transcript independently and then\n send a NewSessionTicket immediately upon sending its Finished rather\n than waiting for the client Finished. This might be appropriate in\n cases where the client is expected to open multiple TLS connections\n in parallel and would benefit from the reduced overhead of a\n resumption handshake, for example.\n\n struct {\n uint32 ticket_lifetime;\n uint32 ticket_age_add;\n opaque ticket_nonce<0..255>;\n opaque ticket<1..2^16-1>;\n Extension extensions<0..2^16-2>;\n } NewSessionTicket;\n\n ticket_lifetime: Indicates the lifetime in seconds as a 32-bit\n unsigned integer in network byte order from the time of ticket\n issuance. Servers MUST NOT use any value greater than\n 604800 seconds (7 days). The value of zero indicates that the\n ticket should be discarded immediately. Clients MUST NOT cache\n tickets for longer than 7 days, regardless of the ticket_lifetime,\n and MAY delete tickets earlier based on local policy. A server\n MAY treat a ticket as valid for a shorter period of time than what\n is stated in the ticket_lifetime.\n\n ticket_age_add: A securely generated, random 32-bit value that is\n used to obscure the age of the ticket that the client includes in\n the \"pre_shared_key\" extension. The client-side ticket age is\n added to this value modulo 2^32 to obtain the value that is\n transmitted by the client. The server MUST generate a fresh value\n for each ticket it sends.\n\n ticket_nonce: A per-ticket value that is unique across all tickets\n issued on this connection.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 74]\n\f\nRFC 8446 TLS August 2018\n\n\n ticket: The value of the ticket to be used as the PSK identity. The\n ticket itself is an opaque label. It MAY be either a database\n lookup key or a self-encrypted and self-authenticated value.\n\n extensions: A set of extension values for the ticket. The\n \"Extension\" format is defined in Section 4.2. Clients MUST ignore\n unrecognized extensions.\n\n The sole extension currently defined for NewSessionTicket is\n \"early_data\", indicating that the ticket may be used to send 0-RTT\n data (Section 4.2.10). It contains the following value:\n\n max_early_data_size: The maximum amount of 0-RTT data that the\n client is allowed to send when using this ticket, in bytes. Only\n Application Data payload (i.e., plaintext but not padding or the\n inner content type byte) is counted. A server receiving more than\n max_early_data_size bytes of 0-RTT data SHOULD terminate the\n connection with an \"unexpected_message\" alert. Note that servers\n that reject early data due to lack of cryptographic material will\n be unable to differentiate padding from content, so clients\n SHOULD NOT depend on being able to send large quantities of\n padding in early data records.\n\n The PSK associated with the ticket is computed as:\n\n HKDF-Expand-Label(resumption_master_secret,\n \"resumption\", ticket_nonce, Hash.length)\n\n Because the ticket_nonce value is distinct for each NewSessionTicket\n message, a different PSK will be derived for each ticket.\n\n Note that in principle it is possible to continue issuing new tickets\n which indefinitely extend the lifetime of the keying material\n originally derived from an initial non-PSK handshake (which was most\n likely tied to the peer's certificate). It is RECOMMENDED that\n implementations place limits on the total lifetime of such keying\n material; these limits should take into account the lifetime of the\n peer's certificate, the likelihood of intervening revocation, and the\n time since the peer's online CertificateVerify signature.\n\n4.6.2. Post-Handshake Authentication\n\n When the client has sent the \"post_handshake_auth\" extension (see\n Section 4.2.6), a server MAY request client authentication at any\n time after the handshake has completed by sending a\n CertificateRequest message. The client MUST respond with the\n appropriate Authentication messages (see Section 4.4). If the client\n chooses to authenticate, it MUST send Certificate, CertificateVerify,\n\n\n\nRescorla Standards Track [Page 75]\n\f\nRFC 8446 TLS August 2018\n\n\n and Finished. If it declines, it MUST send a Certificate message\n containing no certificates followed by Finished. All of the client's\n messages for a given response MUST appear consecutively on the wire\n with no intervening messages of other types.\n\n A client that receives a CertificateRequest message without having\n sent the \"post_handshake_auth\" extension MUST send an\n \"unexpected_message\" fatal alert.\n\n Note: Because client authentication could involve prompting the user,\n servers MUST be prepared for some delay, including receiving an\n arbitrary number of other messages between sending the\n CertificateRequest and receiving a response. In addition, clients\n which receive multiple CertificateRequests in close succession MAY\n respond to them in a different order than they were received (the\n certificate_request_context value allows the server to disambiguate\n the responses).\n\n4.6.3. Key and Initialization Vector Update\n\n The KeyUpdate handshake message is used to indicate that the sender\n is updating its sending cryptographic keys. This message can be sent\n by either peer after it has sent a Finished message. Implementations\n that receive a KeyUpdate message prior to receiving a Finished\n message MUST terminate the connection with an \"unexpected_message\"\n alert. After sending a KeyUpdate message, the sender SHALL send all\n its traffic using the next generation of keys, computed as described\n in Section 7.2. Upon receiving a KeyUpdate, the receiver MUST update\n its receiving keys.\n\n enum {\n update_not_requested(0), update_requested(1), (255)\n } KeyUpdateRequest;\n\n struct {\n KeyUpdateRequest request_update;\n } KeyUpdate;\n\n request_update: Indicates whether the recipient of the KeyUpdate\n should respond with its own KeyUpdate. If an implementation\n receives any other value, it MUST terminate the connection with an\n \"illegal_parameter\" alert.\n\n If the request_update field is set to \"update_requested\", then the\n receiver MUST send a KeyUpdate of its own with request_update set to\n \"update_not_requested\" prior to sending its next Application Data\n record. This mechanism allows either side to force an update to the\n entire connection, but causes an implementation which receives\n\n\n\nRescorla Standards Track [Page 76]\n\f\nRFC 8446 TLS August 2018\n\n\n multiple KeyUpdates while it is silent to respond with a single\n update. Note that implementations may receive an arbitrary number of\n messages between sending a KeyUpdate with request_update set to\n \"update_requested\" and receiving the peer's KeyUpdate, because those\n messages may already be in flight. However, because send and receive\n keys are derived from independent traffic secrets, retaining the\n receive traffic secret does not threaten the forward secrecy of data\n sent before the sender changed keys.\n\n If implementations independently send their own KeyUpdates with\n request_update set to \"update_requested\" and they cross in flight,\n then each side will also send a response, with the result that each\n side increments by two generations.\n\n Both sender and receiver MUST encrypt their KeyUpdate messages with\n the old keys. Additionally, both sides MUST enforce that a KeyUpdate\n with the old key is received before accepting any messages encrypted\n with the new key. Failure to do so may allow message truncation\n attacks.\n\n5. Record Protocol\n\n The TLS record protocol takes messages to be transmitted, fragments\n the data into manageable blocks, protects the records, and transmits\n the result. Received data is verified, decrypted, reassembled, and\n then delivered to higher-level clients.\n\n TLS records are typed, which allows multiple higher-level protocols\n to be multiplexed over the same record layer. This document\n specifies four content types: handshake, application_data, alert, and\n change_cipher_spec. The change_cipher_spec record is used only for\n compatibility purposes (see Appendix D.4).\n\n An implementation may receive an unencrypted record of type\n change_cipher_spec consisting of the single byte value 0x01 at any\n time after the first ClientHello message has been sent or received\n and before the peer's Finished message has been received and MUST\n simply drop it without further processing. Note that this record may\n appear at a point at the handshake where the implementation is\n expecting protected records, and so it is necessary to detect this\n condition prior to attempting to deprotect the record. An\n implementation which receives any other change_cipher_spec value or\n which receives a protected change_cipher_spec record MUST abort the\n handshake with an \"unexpected_message\" alert. If an implementation\n detects a change_cipher_spec record received before the first\n ClientHello message or after the peer's Finished message, it MUST be\n treated as an unexpected record type (though stateless servers may\n not be able to distinguish these cases from allowed cases).\n\n\n\nRescorla Standards Track [Page 77]\n\f\nRFC 8446 TLS August 2018\n\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST terminate\n the connection with an \"unexpected_message\" alert. New record\n content type values are assigned by IANA in the TLS ContentType\n registry as described in Section 11.\n\n5.1. Record Layer\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Message\n boundaries are handled differently depending on the underlying\n ContentType. Any future content types MUST specify appropriate\n rules. Note that these rules are stricter than what was enforced in\n TLS 1.2.\n\n Handshake messages MAY be coalesced into a single TLSPlaintext record\n or fragmented across several records, provided that:\n\n - Handshake messages MUST NOT be interleaved with other record\n types. That is, if a handshake message is split over two or more\n records, there MUST NOT be any other records between them.\n\n - Handshake messages MUST NOT span key changes. Implementations\n MUST verify that all messages immediately preceding a key change\n align with a record boundary; if not, then they MUST terminate the\n connection with an \"unexpected_message\" alert. Because the\n ClientHello, EndOfEarlyData, ServerHello, Finished, and KeyUpdate\n messages can immediately precede a key change, implementations\n MUST send these messages in alignment with a record boundary.\n\n Implementations MUST NOT send zero-length fragments of Handshake\n types, even if those fragments contain padding.\n\n Alert messages (Section 6) MUST NOT be fragmented across records, and\n multiple alert messages MUST NOT be coalesced into a single\n TLSPlaintext record. In other words, a record with an Alert type\n MUST contain exactly one message.\n\n Application Data messages contain data that is opaque to TLS.\n Application Data messages are always protected. Zero-length\n fragments of Application Data MAY be sent, as they are potentially\n useful as a traffic analysis countermeasure. Application Data\n fragments MAY be split across multiple records or coalesced into a\n single record.\n\n\n\n\n\n\nRescorla Standards Track [Page 78]\n\f\nRFC 8446 TLS August 2018\n\n\n enum {\n invalid(0),\n change_cipher_spec(20),\n alert(21),\n handshake(22),\n application_data(23),\n (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion legacy_record_version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type: The higher-level protocol used to process the enclosed\n fragment.\n\n legacy_record_version: MUST be set to 0x0303 for all records\n generated by a TLS 1.3 implementation other than an initial\n ClientHello (i.e., one not generated after a HelloRetryRequest),\n where it MAY also be 0x0301 for compatibility purposes. This\n field is deprecated and MUST be ignored for all purposes.\n Previous versions of TLS would use other values in this field\n under some circumstances.\n\n length: The length (in bytes) of the following\n TLSPlaintext.fragment. The length MUST NOT exceed 2^14 bytes. An\n endpoint that receives a record that exceeds this length MUST\n terminate the connection with a \"record_overflow\" alert.\n\n fragment: The data being transmitted. This value is transparent and\n is treated as an independent block to be dealt with by the higher-\n level protocol specified by the type field.\n\n This document describes TLS 1.3, which uses the version 0x0304. This\n version value is historical, deriving from the use of 0x0301 for\n TLS 1.0 and 0x0300 for SSL 3.0. In order to maximize backward\n compatibility, a record containing an initial ClientHello SHOULD have\n version 0x0301 (reflecting TLS 1.0) and a record containing a second\n ClientHello or a ServerHello MUST have version 0x0303 (reflecting\n TLS 1.2). When negotiating prior versions of TLS, endpoints follow\n the procedure and requirements provided in Appendix D.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 79]\n\f\nRFC 8446 TLS August 2018\n\n\n When record protection has not yet been engaged, TLSPlaintext\n structures are written directly onto the wire. Once record\n protection has started, TLSPlaintext records are protected and sent\n as described in the following section. Note that Application Data\n records MUST NOT be written to the wire unprotected (see Section 2\n for details).\n\n5.2. Record Payload Protection\n\n The record protection functions translate a TLSPlaintext structure\n into a TLSCiphertext structure. The deprotection functions reverse\n the process. In TLS 1.3, as opposed to previous versions of TLS, all\n ciphers are modeled as \"Authenticated Encryption with Associated\n Data\" (AEAD) [RFC5116]. AEAD functions provide a unified encryption\n and authentication operation which turns plaintext into authenticated\n ciphertext and back again. Each encrypted record consists of a\n plaintext header followed by an encrypted body, which itself contains\n a type and optional padding.\n\n struct {\n opaque content[TLSPlaintext.length];\n ContentType type;\n uint8 zeros[length_of_padding];\n } TLSInnerPlaintext;\n\n struct {\n ContentType opaque_type = application_data; /* 23 */\n ProtocolVersion legacy_record_version = 0x0303; /* TLS v1.2 */\n uint16 length;\n opaque encrypted_record[TLSCiphertext.length];\n } TLSCiphertext;\n\n content: The TLSPlaintext.fragment value, containing the byte\n encoding of a handshake or an alert message, or the raw bytes of\n the application's data to send.\n\n type: The TLSPlaintext.type value containing the content type of the\n record.\n\n zeros: An arbitrary-length run of zero-valued bytes may appear in\n the cleartext after the type field. This provides an opportunity\n for senders to pad any TLS record by a chosen amount as long as\n the total stays within record size limits. See Section 5.4 for\n more details.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 80]\n\f\nRFC 8446 TLS August 2018\n\n\n opaque_type: The outer opaque_type field of a TLSCiphertext record\n is always set to the value 23 (application_data) for outward\n compatibility with middleboxes accustomed to parsing previous\n versions of TLS. The actual content type of the record is found\n in TLSInnerPlaintext.type after decryption.\n\n legacy_record_version: The legacy_record_version field is always\n 0x0303. TLS 1.3 TLSCiphertexts are not generated until after\n TLS 1.3 has been negotiated, so there are no historical\n compatibility concerns where other values might be received. Note\n that the handshake protocol, including the ClientHello and\n ServerHello messages, authenticates the protocol version, so this\n value is redundant.\n\n length: The length (in bytes) of the following\n TLSCiphertext.encrypted_record, which is the sum of the lengths of\n the content and the padding, plus one for the inner content type,\n plus any expansion added by the AEAD algorithm. The length\n MUST NOT exceed 2^14 + 256 bytes. An endpoint that receives a\n record that exceeds this length MUST terminate the connection with\n a \"record_overflow\" alert.\n\n encrypted_record: The AEAD-encrypted form of the serialized\n TLSInnerPlaintext structure.\n\n AEAD algorithms take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [RFC5116]. The key is either the\n client_write_key or the server_write_key, the nonce is derived from\n the sequence number and the client_write_iv or server_write_iv (see\n Section 5.3), and the additional data input is the record header.\n\n I.e.,\n\n additional_data = TLSCiphertext.opaque_type ||\n TLSCiphertext.legacy_record_version ||\n TLSCiphertext.length\n\n The plaintext input to the AEAD algorithm is the encoded\n TLSInnerPlaintext structure. Derivation of traffic keys is defined\n in Section 7.3.\n\n The AEAD output consists of the ciphertext output from the AEAD\n encryption operation. The length of the plaintext is greater than\n the corresponding TLSPlaintext.length due to the inclusion of\n TLSInnerPlaintext.type and any padding supplied by the sender. The\n length of the AEAD output will generally be larger than the\n plaintext, but by an amount that varies with the AEAD algorithm.\n\n\n\nRescorla Standards Track [Page 81]\n\f\nRFC 8446 TLS August 2018\n\n\n Since the ciphers might incorporate padding, the amount of overhead\n could vary with different lengths of plaintext. Symbolically,\n\n AEADEncrypted =\n AEAD-Encrypt(write_key, nonce, additional_data, plaintext)\n\n The encrypted_record field of TLSCiphertext is set to AEADEncrypted.\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, additional data, and the AEADEncrypted value. The output is\n either the plaintext or an error indicating that the decryption\n failed. There is no separate integrity check. Symbolically,\n\n plaintext of encrypted_record =\n AEAD-Decrypt(peer_write_key, nonce,\n additional_data, AEADEncrypted)\n\n If the decryption fails, the receiver MUST terminate the connection\n with a \"bad_record_mac\" alert.\n\n An AEAD algorithm used in TLS 1.3 MUST NOT produce an expansion\n greater than 255 octets. An endpoint that receives a record from its\n peer with TLSCiphertext.length larger than 2^14 + 256 octets MUST\n terminate the connection with a \"record_overflow\" alert. This limit\n is derived from the maximum TLSInnerPlaintext length of 2^14 octets +\n 1 octet for ContentType + the maximum AEAD expansion of 255 octets.\n\n5.3. Per-Record Nonce\n\n A 64-bit sequence number is maintained separately for reading and\n writing records. The appropriate sequence number is incremented by\n one after reading or writing each record. Each sequence number is\n set to zero at the beginning of a connection and whenever the key is\n changed; the first record transmitted under a particular traffic key\n MUST use sequence number 0.\n\n Because the size of sequence numbers is 64-bit, they should not wrap.\n If a TLS implementation would need to wrap a sequence number, it MUST\n either rekey (Section 4.6.3) or terminate the connection.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 82]\n\f\nRFC 8446 TLS August 2018\n\n\n Each AEAD algorithm will specify a range of possible lengths for the\n per-record nonce, from N_MIN bytes to N_MAX bytes of input [RFC5116].\n The length of the TLS per-record nonce (iv_length) is set to the\n larger of 8 bytes and N_MIN for the AEAD algorithm (see [RFC5116],\n Section 4). An AEAD algorithm where N_MAX is less than 8 bytes\n MUST NOT be used with TLS. The per-record nonce for the AEAD\n construction is formed as follows:\n\n 1. The 64-bit record sequence number is encoded in network byte\n order and padded to the left with zeros to iv_length.\n\n 2. The padded sequence number is XORed with either the static\n client_write_iv or server_write_iv (depending on the role).\n\n The resulting quantity (of length iv_length) is used as the\n per-record nonce.\n\n Note: This is a different construction from that in TLS 1.2, which\n specified a partially explicit nonce.\n\n5.4. Record Padding\n\n All encrypted TLS records can be padded to inflate the size of the\n TLSCiphertext. This allows the sender to hide the size of the\n traffic from an observer.\n\n When generating a TLSCiphertext record, implementations MAY choose to\n pad. An unpadded record is just a record with a padding length of\n zero. Padding is a string of zero-valued bytes appended to the\n ContentType field before encryption. Implementations MUST set the\n padding octets to all zeros before encrypting.\n\n Application Data records may contain a zero-length\n TLSInnerPlaintext.content if the sender desires. This permits\n generation of plausibly sized cover traffic in contexts where the\n presence or absence of activity may be sensitive. Implementations\n MUST NOT send Handshake and Alert records that have a zero-length\n TLSInnerPlaintext.content; if such a message is received, the\n receiving implementation MUST terminate the connection with an\n \"unexpected_message\" alert.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 83]\n\f\nRFC 8446 TLS August 2018\n\n\n The padding sent is automatically verified by the record protection\n mechanism; upon successful decryption of a\n TLSCiphertext.encrypted_record, the receiving implementation scans\n the field from the end toward the beginning until it finds a non-zero\n octet. This non-zero octet is the content type of the message. This\n padding scheme was selected because it allows padding of any\n encrypted TLS record by an arbitrary size (from zero up to TLS record\n size limits) without introducing new content types. The design also\n enforces all-zero padding octets, which allows for quick detection of\n padding errors.\n\n Implementations MUST limit their scanning to the cleartext returned\n from the AEAD decryption. If a receiving implementation does not\n find a non-zero octet in the cleartext, it MUST terminate the\n connection with an \"unexpected_message\" alert.\n\n The presence of padding does not change the overall record size\n limitations: the full encoded TLSInnerPlaintext MUST NOT exceed 2^14\n + 1 octets. If the maximum fragment length is reduced -- as, for\n example, by the record_size_limit extension from [RFC8449] -- then\n the reduced limit applies to the full plaintext, including the\n content type and padding.\n\n Selecting a padding policy that suggests when and how much to pad is\n a complex topic and is beyond the scope of this specification. If\n the application-layer protocol on top of TLS has its own padding, it\n may be preferable to pad Application Data TLS records within the\n application layer. Padding for encrypted Handshake or Alert records\n must still be handled at the TLS layer, though. Later documents may\n define padding selection algorithms or define a padding policy\n request mechanism through TLS extensions or some other means.\n\n5.5. Limits on Key Usage\n\n There are cryptographic limits on the amount of plaintext which can\n be safely encrypted under a given set of keys. [AEAD-LIMITS]\n provides an analysis of these limits under the assumption that the\n underlying primitive (AES or ChaCha20) has no weaknesses.\n Implementations SHOULD do a key update as described in Section 4.6.3\n prior to reaching these limits.\n\n For AES-GCM, up to 2^24.5 full-size records (about 24 million) may be\n encrypted on a given connection while keeping a safety margin of\n approximately 2^-57 for Authenticated Encryption (AE) security. For\n ChaCha20/Poly1305, the record sequence number would wrap before the\n safety limit is reached.\n\n\n\n\n\nRescorla Standards Track [Page 84]\n\f\nRFC 8446 TLS August 2018\n\n\n6. Alert Protocol\n\n TLS provides an Alert content type to indicate closure information\n and errors. Like other messages, alert messages are encrypted as\n specified by the current connection state.\n\n Alert messages convey a description of the alert and a legacy field\n that conveyed the severity level of the message in previous versions\n of TLS. Alerts are divided into two classes: closure alerts and\n error alerts. In TLS 1.3, the severity is implicit in the type of\n alert being sent, and the \"level\" field can safely be ignored. The\n \"close_notify\" alert is used to indicate orderly closure of one\n direction of the connection. Upon receiving such an alert, the TLS\n implementation SHOULD indicate end-of-data to the application.\n\n Error alerts indicate abortive closure of the connection (see\n Section 6.2). Upon receiving an error alert, the TLS implementation\n SHOULD indicate an error to the application and MUST NOT allow any\n further data to be sent or received on the connection. Servers and\n clients MUST forget the secret values and keys established in failed\n connections, with the exception of the PSKs associated with session\n tickets, which SHOULD be discarded if possible.\n\n All the alerts listed in Section 6.2 MUST be sent with\n AlertLevel=fatal and MUST be treated as error alerts when received\n regardless of the AlertLevel in the message. Unknown Alert types\n MUST be treated as error alerts.\n\n Note: TLS defines two generic alerts (see Section 6) to use upon\n failure to parse a message. Peers which receive a message which\n cannot be parsed according to the syntax (e.g., have a length\n extending beyond the message boundary or contain an out-of-range\n length) MUST terminate the connection with a \"decode_error\" alert.\n Peers which receive a message which is syntactically correct but\n semantically invalid (e.g., a DHE share of p - 1, or an invalid enum)\n MUST terminate the connection with an \"illegal_parameter\" alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 85]\n\f\nRFC 8446 TLS August 2018\n\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n record_overflow(22),\n handshake_failure(40),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n inappropriate_fallback(86),\n user_canceled(90),\n missing_extension(109),\n unsupported_extension(110),\n unrecognized_name(112),\n bad_certificate_status_response(113),\n unknown_psk_identity(115),\n certificate_required(116),\n no_application_protocol(120),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 86]\n\f\nRFC 8446 TLS August 2018\n\n\n6.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack.\n\n close_notify: This alert notifies the recipient that the sender will\n not send any more messages on this connection. Any data received\n after a closure alert has been received MUST be ignored.\n\n user_canceled: This alert notifies the recipient that the sender is\n canceling the handshake for some reason unrelated to a protocol\n failure. If a user cancels an operation after the handshake is\n complete, just closing the connection by sending a \"close_notify\"\n is more appropriate. This alert SHOULD be followed by a\n \"close_notify\". This alert generally has AlertLevel=warning.\n\n Either party MAY initiate a close of its write side of the connection\n by sending a \"close_notify\" alert. Any data received after a closure\n alert has been received MUST be ignored. If a transport-level close\n is received prior to a \"close_notify\", the receiver cannot know that\n all the data that was sent has been received.\n\n Each party MUST send a \"close_notify\" alert before closing its write\n side of the connection, unless it has already sent some error alert.\n This does not have any effect on its read side of the connection.\n Note that this is a change from versions of TLS prior to TLS 1.3 in\n which implementations were required to react to a \"close_notify\" by\n discarding pending writes and sending an immediate \"close_notify\"\n alert of their own. That previous requirement could cause truncation\n in the read side. Both parties need not wait to receive a\n \"close_notify\" alert before closing their read side of the\n connection, though doing so would introduce the possibility of\n truncation.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation MUST receive a \"close_notify\" alert\n before indicating end-of-data to the application layer. No part of\n this standard should be taken to dictate the manner in which a usage\n profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing the write side of a connection\n reliably delivers pending data before destroying the transport.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 87]\n\f\nRFC 8446 TLS August 2018\n\n\n6.2. Error Alerts\n\n Error handling in TLS is very simple. When an error is detected, the\n detecting party sends a message to its peer. Upon transmission or\n receipt of a fatal alert message, both parties MUST immediately close\n the connection.\n\n Whenever an implementation encounters a fatal error condition, it\n SHOULD send an appropriate fatal alert and MUST close the connection\n without sending or receiving any additional data. In the rest of\n this specification, when the phrases \"terminate the connection\" and\n \"abort the handshake\" are used without a specific alert it means that\n the implementation SHOULD send the alert indicated by the\n descriptions below. The phrases \"terminate the connection with an X\n alert\" and \"abort the handshake with an X alert\" mean that the\n implementation MUST send alert X if it sends any alert. All alerts\n defined below in this section, as well as all unknown alerts, are\n universally considered fatal as of TLS 1.3 (see Section 6). The\n implementation SHOULD provide a way to facilitate logging the sending\n and receiving of alerts.\n\n The following error alerts are defined:\n\n unexpected_message: An inappropriate message (e.g., the wrong\n handshake message, premature Application Data, etc.) was received.\n This alert should never be observed in communication between\n proper implementations.\n\n bad_record_mac: This alert is returned if a record is received which\n cannot be deprotected. Because AEAD algorithms combine decryption\n and verification, and also to avoid side-channel attacks, this\n alert is used for all deprotection failures. This alert should\n never be observed in communication between proper implementations,\n except when messages were corrupted in the network.\n\n record_overflow: A TLSCiphertext record was received that had a\n length more than 2^14 + 256 bytes, or a record decrypted to a\n TLSPlaintext record with more than 2^14 bytes (or some other\n negotiated limit). This alert should never be observed in\n communication between proper implementations, except when messages\n were corrupted in the network.\n\n handshake_failure: Receipt of a \"handshake_failure\" alert message\n indicates that the sender was unable to negotiate an acceptable\n set of security parameters given the options available.\n\n bad_certificate: A certificate was corrupt, contained signatures\n that did not verify correctly, etc.\n\n\n\nRescorla Standards Track [Page 88]\n\f\nRFC 8446 TLS August 2018\n\n\n unsupported_certificate: A certificate was of an unsupported type.\n\n certificate_revoked: A certificate was revoked by its signer.\n\n certificate_expired: A certificate has expired or is not currently\n valid.\n\n certificate_unknown: Some other (unspecified) issue arose in\n processing the certificate, rendering it unacceptable.\n\n illegal_parameter: A field in the handshake was incorrect or\n inconsistent with other fields. This alert is used for errors\n which conform to the formal protocol syntax but are otherwise\n incorrect.\n\n unknown_ca: A valid certificate chain or partial chain was received,\n but the certificate was not accepted because the CA certificate\n could not be located or could not be matched with a known trust\n anchor.\n\n access_denied: A valid certificate or PSK was received, but when\n access control was applied, the sender decided not to proceed with\n negotiation.\n\n decode_error: A message could not be decoded because some field was\n out of the specified range or the length of the message was\n incorrect. This alert is used for errors where the message does\n not conform to the formal protocol syntax. This alert should\n never be observed in communication between proper implementations,\n except when messages were corrupted in the network.\n\n decrypt_error: A handshake (not record layer) cryptographic\n operation failed, including being unable to correctly verify a\n signature or validate a Finished message or a PSK binder.\n\n protocol_version: The protocol version the peer has attempted to\n negotiate is recognized but not supported (see Appendix D).\n\n insufficient_security: Returned instead of \"handshake_failure\" when\n a negotiation has failed specifically because the server requires\n parameters more secure than those supported by the client.\n\n internal_error: An internal error unrelated to the peer or the\n correctness of the protocol (such as a memory allocation failure)\n makes it impossible to continue.\n\n inappropriate_fallback: Sent by a server in response to an invalid\n connection retry attempt from a client (see [RFC7507]).\n\n\n\nRescorla Standards Track [Page 89]\n\f\nRFC 8446 TLS August 2018\n\n\n missing_extension: Sent by endpoints that receive a handshake\n message not containing an extension that is mandatory to send for\n the offered TLS version or other negotiated parameters.\n\n unsupported_extension: Sent by endpoints receiving any handshake\n message containing an extension known to be prohibited for\n inclusion in the given handshake message, or including any\n extensions in a ServerHello or Certificate not first offered in\n the corresponding ClientHello or CertificateRequest.\n\n unrecognized_name: Sent by servers when no server exists identified\n by the name provided by the client via the \"server_name\" extension\n (see [RFC6066]).\n\n bad_certificate_status_response: Sent by clients when an invalid or\n unacceptable OCSP response is provided by the server via the\n \"status_request\" extension (see [RFC6066]).\n\n unknown_psk_identity: Sent by servers when PSK key establishment is\n desired but no acceptable PSK identity is provided by the client.\n Sending this alert is OPTIONAL; servers MAY instead choose to send\n a \"decrypt_error\" alert to merely indicate an invalid PSK\n identity.\n\n certificate_required: Sent by servers when a client certificate is\n desired but none was provided by the client.\n\n no_application_protocol: Sent by servers when a client\n \"application_layer_protocol_negotiation\" extension advertises only\n protocols that the server does not support (see [RFC7301]).\n\n New Alert values are assigned by IANA as described in Section 11.\n\n7. Cryptographic Computations\n\n The TLS handshake establishes one or more input secrets which are\n combined to create the actual working keying material, as detailed\n below. The key derivation process incorporates both the input\n secrets and the handshake transcript. Note that because the\n handshake transcript includes the random values from the Hello\n messages, any given handshake will have different traffic secrets,\n even if the same input secrets are used, as is the case when the same\n PSK is used for multiple connections.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 90]\n\f\nRFC 8446 TLS August 2018\n\n\n7.1. Key Schedule\n\n The key derivation process makes use of the HKDF-Extract and\n HKDF-Expand functions as defined for HKDF [RFC5869], as well as the\n functions defined below:\n\n HKDF-Expand-Label(Secret, Label, Context, Length) =\n HKDF-Expand(Secret, HkdfLabel, Length)\n\n Where HkdfLabel is specified as:\n\n struct {\n uint16 length = Length;\n opaque label<7..255> = \"tls13 \" + Label;\n opaque context<0..255> = Context;\n } HkdfLabel;\n\n Derive-Secret(Secret, Label, Messages) =\n HKDF-Expand-Label(Secret, Label,\n Transcript-Hash(Messages), Hash.length)\n\n The Hash function used by Transcript-Hash and HKDF is the cipher\n suite hash algorithm. Hash.length is its output length in bytes.\n Messages is the concatenation of the indicated handshake messages,\n including the handshake message type and length fields, but not\n including record layer headers. Note that in some cases a zero-\n length Context (indicated by \"\") is passed to HKDF-Expand-Label. The\n labels specified in this document are all ASCII strings and do not\n include a trailing NUL byte.\n\n Note: With common hash functions, any label longer than 12 characters\n requires an additional iteration of the hash function to compute.\n The labels in this specification have all been chosen to fit within\n this limit.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 91]\n\f\nRFC 8446 TLS August 2018\n\n\n Keys are derived from two input secrets using the HKDF-Extract and\n Derive-Secret functions. The general pattern for adding a new secret\n is to use HKDF-Extract with the Salt being the current secret state\n and the Input Keying Material (IKM) being the new secret to be added.\n In this version of TLS 1.3, the two input secrets are:\n\n - PSK (a pre-shared key established externally or derived from the\n resumption_master_secret value from a previous connection)\n\n - (EC)DHE shared secret (Section 7.4)\n\n This produces a full key derivation schedule shown in the diagram\n below. In this diagram, the following formatting conventions apply:\n\n - HKDF-Extract is drawn as taking the Salt argument from the top and\n the IKM argument from the left, with its output to the bottom and\n the name of the output on the right.\n\n - Derive-Secret's Secret argument is indicated by the incoming\n arrow. For instance, the Early Secret is the Secret for\n generating the client_early_traffic_secret.\n\n - \"0\" indicates a string of Hash.length bytes set to zero.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 92]\n\f\nRFC 8446 TLS August 2018\n\n\n 0\n |\n v\n PSK -> HKDF-Extract = Early Secret\n |\n +-----> Derive-Secret(., \"ext binder\" | \"res binder\", \"\")\n | = binder_key\n |\n +-----> Derive-Secret(., \"c e traffic\", ClientHello)\n | = client_early_traffic_secret\n |\n +-----> Derive-Secret(., \"e exp master\", ClientHello)\n | = early_exporter_master_secret\n v\n Derive-Secret(., \"derived\", \"\")\n |\n v\n (EC)DHE -> HKDF-Extract = Handshake Secret\n |\n +-----> Derive-Secret(., \"c hs traffic\",\n | ClientHello...ServerHello)\n | = client_handshake_traffic_secret\n |\n +-----> Derive-Secret(., \"s hs traffic\",\n | ClientHello...ServerHello)\n | = server_handshake_traffic_secret\n v\n Derive-Secret(., \"derived\", \"\")\n |\n v\n 0 -> HKDF-Extract = Master Secret\n |\n +-----> Derive-Secret(., \"c ap traffic\",\n | ClientHello...server Finished)\n | = client_application_traffic_secret_0\n |\n +-----> Derive-Secret(., \"s ap traffic\",\n | ClientHello...server Finished)\n | = server_application_traffic_secret_0\n |\n +-----> Derive-Secret(., \"exp master\",\n | ClientHello...server Finished)\n | = exporter_master_secret\n |\n +-----> Derive-Secret(., \"res master\",\n ClientHello...client Finished)\n = resumption_master_secret\n\n\n\n\nRescorla Standards Track [Page 93]\n\f\nRFC 8446 TLS August 2018\n\n\n The general pattern here is that the secrets shown down the left side\n of the diagram are just raw entropy without context, whereas the\n secrets down the right side include Handshake Context and therefore\n can be used to derive working keys without additional context. Note\n that the different calls to Derive-Secret may take different Messages\n arguments, even with the same secret. In a 0-RTT exchange,\n Derive-Secret is called with four distinct transcripts; in a\n 1-RTT-only exchange, it is called with three distinct transcripts.\n\n If a given secret is not available, then the 0-value consisting of a\n string of Hash.length bytes set to zeros is used. Note that this\n does not mean skipping rounds, so if PSK is not in use, Early Secret\n will still be HKDF-Extract(0, 0). For the computation of the\n binder_key, the label is \"ext binder\" for external PSKs (those\n provisioned outside of TLS) and \"res binder\" for resumption PSKs\n (those provisioned as the resumption master secret of a previous\n handshake). The different labels prevent the substitution of one\n type of PSK for the other.\n\n There are multiple potential Early Secret values, depending on which\n PSK the server ultimately selects. The client will need to compute\n one for each potential PSK; if no PSK is selected, it will then need\n to compute the Early Secret corresponding to the zero PSK.\n\n Once all the values which are to be derived from a given secret have\n been computed, that secret SHOULD be erased.\n\n7.2. Updating Traffic Secrets\n\n Once the handshake is complete, it is possible for either side to\n update its sending traffic keys using the KeyUpdate handshake message\n defined in Section 4.6.3. The next generation of traffic keys is\n computed by generating client_/server_application_traffic_secret_N+1\n from client_/server_application_traffic_secret_N as described in this\n section and then re-deriving the traffic keys as described in\n Section 7.3.\n\n The next-generation application_traffic_secret is computed as:\n\n application_traffic_secret_N+1 =\n HKDF-Expand-Label(application_traffic_secret_N,\n \"traffic upd\", \"\", Hash.length)\n\n Once client_/server_application_traffic_secret_N+1 and its associated\n traffic keys have been computed, implementations SHOULD delete\n client_/server_application_traffic_secret_N and its associated\n traffic keys.\n\n\n\n\nRescorla Standards Track [Page 94]\n\f\nRFC 8446 TLS August 2018\n\n\n7.3. Traffic Key Calculation\n\n The traffic keying material is generated from the following input\n values:\n\n - A secret value\n\n - A purpose value indicating the specific value being generated\n\n - The length of the key being generated\n\n The traffic keying material is generated from an input traffic secret\n value using:\n\n [sender]_write_key = HKDF-Expand-Label(Secret, \"key\", \"\", key_length)\n [sender]_write_iv = HKDF-Expand-Label(Secret, \"iv\", \"\", iv_length)\n\n [sender] denotes the sending side. The value of Secret for each\n record type is shown in the table below.\n\n +-------------------+---------------------------------------+\n | Record Type | Secret |\n +-------------------+---------------------------------------+\n | 0-RTT Application | client_early_traffic_secret |\n | | |\n | Handshake | [sender]_handshake_traffic_secret |\n | | |\n | Application Data | [sender]_application_traffic_secret_N |\n +-------------------+---------------------------------------+\n\n All the traffic keying material is recomputed whenever the underlying\n Secret changes (e.g., when changing from the handshake to Application\n Data keys or upon a key update).\n\n7.4. (EC)DHE Shared Secret Calculation\n\n7.4.1. Finite Field Diffie-Hellman\n\n For finite field groups, a conventional Diffie-Hellman [DH76]\n computation is performed. The negotiated key (Z) is converted to a\n byte string by encoding in big-endian form and left-padded with zeros\n up to the size of the prime. This byte string is used as the shared\n secret in the key schedule as specified above.\n\n Note that this construction differs from previous versions of TLS\n which removed leading zeros.\n\n\n\n\n\nRescorla Standards Track [Page 95]\n\f\nRFC 8446 TLS August 2018\n\n\n7.4.2. Elliptic Curve Diffie-Hellman\n\n For secp256r1, secp384r1, and secp521r1, ECDH calculations (including\n parameter and key generation as well as the shared secret\n calculation) are performed according to [IEEE1363] using the\n ECKAS-DH1 scheme with the identity map as the key derivation function\n (KDF), so that the shared secret is the x-coordinate of the ECDH\n shared secret elliptic curve point represented as an octet string.\n Note that this octet string (\"Z\" in IEEE 1363 terminology) as output\n by FE2OSP (the Field Element to Octet String Conversion Primitive)\n has constant length for any given field; leading zeros found in this\n octet string MUST NOT be truncated.\n\n (Note that this use of the identity KDF is a technicality. The\n complete picture is that ECDH is employed with a non-trivial KDF\n because TLS does not directly use this secret for anything other than\n for computing other secrets.)\n\n For X25519 and X448, the ECDH calculations are as follows:\n\n - The public key to put into the KeyShareEntry.key_exchange\n structure is the result of applying the ECDH scalar multiplication\n function to the secret key of appropriate length (into scalar\n input) and the standard public basepoint (into u-coordinate point\n input).\n\n - The ECDH shared secret is the result of applying the ECDH scalar\n multiplication function to the secret key (into scalar input) and\n the peer's public key (into u-coordinate point input). The output\n is used raw, with no processing.\n\n For these curves, implementations SHOULD use the approach specified\n in [RFC7748] to calculate the Diffie-Hellman shared secret.\n Implementations MUST check whether the computed Diffie-Hellman shared\n secret is the all-zero value and abort if so, as described in\n Section 6 of [RFC7748]. If implementors use an alternative\n implementation of these elliptic curves, they SHOULD perform the\n additional checks specified in Section 7 of [RFC7748].\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 96]\n\f\nRFC 8446 TLS August 2018\n\n\n7.5. Exporters\n\n [RFC5705] defines keying material exporters for TLS in terms of the\n TLS pseudorandom function (PRF). This document replaces the PRF with\n HKDF, thus requiring a new construction. The exporter interface\n remains the same.\n\n The exporter value is computed as:\n\n TLS-Exporter(label, context_value, key_length) =\n HKDF-Expand-Label(Derive-Secret(Secret, label, \"\"),\n \"exporter\", Hash(context_value), key_length)\n\n Where Secret is either the early_exporter_master_secret or the\n exporter_master_secret. Implementations MUST use the\n exporter_master_secret unless explicitly specified by the\n application. The early_exporter_master_secret is defined for use in\n settings where an exporter is needed for 0-RTT data. A separate\n interface for the early exporter is RECOMMENDED; this avoids the\n exporter user accidentally using an early exporter when a regular one\n is desired or vice versa.\n\n If no context is provided, the context_value is zero length.\n Consequently, providing no context computes the same value as\n providing an empty context. This is a change from previous versions\n of TLS where an empty context produced a different output than an\n absent context. As of this document's publication, no allocated\n exporter label is used both with and without a context. Future\n specifications MUST NOT define a use of exporters that permit both an\n empty context and no context with the same label. New uses of\n exporters SHOULD provide a context in all exporter computations,\n though the value could be empty.\n\n Requirements for the format of exporter labels are defined in\n Section 4 of [RFC5705].\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 97]\n\f\nRFC 8446 TLS August 2018\n\n\n8. 0-RTT and Anti-Replay\n\n As noted in Section 2.3 and Appendix E.5, TLS does not provide\n inherent replay protections for 0-RTT data. There are two potential\n threats to be concerned with:\n\n - Network attackers who mount a replay attack by simply duplicating\n a flight of 0-RTT data.\n\n - Network attackers who take advantage of client retry behavior to\n arrange for the server to receive multiple copies of an\n application message. This threat already exists to some extent\n because clients that value robustness respond to network errors by\n attempting to retry requests. However, 0-RTT adds an additional\n dimension for any server system which does not maintain globally\n consistent server state. Specifically, if a server system has\n multiple zones where tickets from zone A will not be accepted in\n zone B, then an attacker can duplicate a ClientHello and early\n data intended for A to both A and B. At A, the data will be\n accepted in 0-RTT, but at B the server will reject 0-RTT data and\n instead force a full handshake. If the attacker blocks the\n ServerHello from A, then the client will complete the handshake\n with B and probably retry the request, leading to duplication on\n the server system as a whole.\n\n The first class of attack can be prevented by sharing state to\n guarantee that the 0-RTT data is accepted at most once. Servers\n SHOULD provide that level of replay safety by implementing one of the\n methods described in this section or by equivalent means. It is\n understood, however, that due to operational concerns not all\n deployments will maintain state at that level. Therefore, in normal\n operation, clients will not know which, if any, of these mechanisms\n servers actually implement and hence MUST only send early data which\n they deem safe to be replayed.\n\n In addition to the direct effects of replays, there is a class of\n attacks where even operations normally considered idempotent could be\n exploited by a large number of replays (timing attacks, resource\n limit exhaustion and others, as described in Appendix E.5). Those\n can be mitigated by ensuring that every 0-RTT payload can be replayed\n only a limited number of times. The server MUST ensure that any\n instance of it (be it a machine, a thread, or any other entity within\n the relevant serving infrastructure) would accept 0-RTT for the same\n 0-RTT handshake at most once; this limits the number of replays to\n the number of server instances in the deployment. Such a guarantee\n can be accomplished by locally recording data from recently received\n ClientHellos and rejecting repeats, or by any other method that\n\n\n\n\nRescorla Standards Track [Page 98]\n\f\nRFC 8446 TLS August 2018\n\n\n provides the same or a stronger guarantee. The \"at most once per\n server instance\" guarantee is a minimum requirement; servers SHOULD\n limit 0-RTT replays further when feasible.\n\n The second class of attack cannot be prevented at the TLS layer and\n MUST be dealt with by any application. Note that any application\n whose clients implement any kind of retry behavior already needs to\n implement some sort of anti-replay defense.\n\n8.1. Single-Use Tickets\n\n The simplest form of anti-replay defense is for the server to only\n allow each session ticket to be used once. For instance, the server\n can maintain a database of all outstanding valid tickets, deleting\n each ticket from the database as it is used. If an unknown ticket is\n provided, the server would then fall back to a full handshake.\n\n If the tickets are not self-contained but rather are database keys,\n and the corresponding PSKs are deleted upon use, then connections\n established using PSKs enjoy forward secrecy. This improves security\n for all 0-RTT data and PSK usage when PSK is used without (EC)DHE.\n\n Because this mechanism requires sharing the session database between\n server nodes in environments with multiple distributed servers, it\n may be hard to achieve high rates of successful PSK 0-RTT connections\n when compared to self-encrypted tickets. Unlike session databases,\n session tickets can successfully do PSK-based session establishment\n even without consistent storage, though when 0-RTT is allowed they\n still require consistent storage for anti-replay of 0-RTT data, as\n detailed in the following section.\n\n8.2. Client Hello Recording\n\n An alternative form of anti-replay is to record a unique value\n derived from the ClientHello (generally either the random value or\n the PSK binder) and reject duplicates. Recording all ClientHellos\n causes state to grow without bound, but a server can instead record\n ClientHellos within a given time window and use the\n \"obfuscated_ticket_age\" to ensure that tickets aren't reused outside\n that window.\n\n In order to implement this, when a ClientHello is received, the\n server first verifies the PSK binder as described in Section 4.2.11.\n It then computes the expected_arrival_time as described in the next\n section and rejects 0-RTT if it is outside the recording window,\n falling back to the 1-RTT handshake.\n\n\n\n\n\nRescorla Standards Track [Page 99]\n\f\nRFC 8446 TLS August 2018\n\n\n If the expected_arrival_time is in the window, then the server checks\n to see if it has recorded a matching ClientHello. If one is found,\n it either aborts the handshake with an \"illegal_parameter\" alert or\n accepts the PSK but rejects 0-RTT. If no matching ClientHello is\n found, then it accepts 0-RTT and then stores the ClientHello for as\n long as the expected_arrival_time is inside the window. Servers MAY\n also implement data stores with false positives, such as Bloom\n filters, in which case they MUST respond to apparent replay by\n rejecting 0-RTT but MUST NOT abort the handshake.\n\n The server MUST derive the storage key only from validated sections\n of the ClientHello. If the ClientHello contains multiple PSK\n identities, then an attacker can create multiple ClientHellos with\n different binder values for the less-preferred identity on the\n assumption that the server will not verify it (as recommended by\n Section 4.2.11). I.e., if the client sends PSKs A and B but the\n server prefers A, then the attacker can change the binder for B\n without affecting the binder for A. If the binder for B is part of\n the storage key, then this ClientHello will not appear as a\n duplicate, which will cause the ClientHello to be accepted, and may\n cause side effects such as replay cache pollution, although any 0-RTT\n data will not be decryptable because it will use different keys. If\n the validated binder or the ClientHello.random is used as the storage\n key, then this attack is not possible.\n\n Because this mechanism does not require storing all outstanding\n tickets, it may be easier to implement in distributed systems with\n high rates of resumption and 0-RTT, at the cost of potentially weaker\n anti-replay defense because of the difficulty of reliably storing and\n retrieving the received ClientHello messages. In many such systems,\n it is impractical to have globally consistent storage of all the\n received ClientHellos. In this case, the best anti-replay protection\n is provided by having a single storage zone be authoritative for a\n given ticket and refusing 0-RTT for that ticket in any other zone.\n This approach prevents simple replay by the attacker because only one\n zone will accept 0-RTT data. A weaker design is to implement\n separate storage for each zone but allow 0-RTT in any zone. This\n approach limits the number of replays to once per zone. Application\n message duplication of course remains possible with either design.\n\n When implementations are freshly started, they SHOULD reject 0-RTT as\n long as any portion of their recording window overlaps the startup\n time. Otherwise, they run the risk of accepting replays which were\n originally sent during that period.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 100]\n\f\nRFC 8446 TLS August 2018\n\n\n Note: If the client's clock is running much faster than the server's,\n then a ClientHello may be received that is outside the window in the\n future, in which case it might be accepted for 1-RTT, causing a\n client retry, and then acceptable later for 0-RTT. This is another\n variant of the second form of attack described in Section 8.\n\n8.3. Freshness Checks\n\n Because the ClientHello indicates the time at which the client sent\n it, it is possible to efficiently determine whether a ClientHello was\n likely sent reasonably recently and only accept 0-RTT for such a\n ClientHello, otherwise falling back to a 1-RTT handshake. This is\n necessary for the ClientHello storage mechanism described in\n Section 8.2 because otherwise the server needs to store an unlimited\n number of ClientHellos, and is a useful optimization for self-\n contained single-use tickets because it allows efficient rejection of\n ClientHellos which cannot be used for 0-RTT.\n\n In order to implement this mechanism, a server needs to store the\n time that the server generated the session ticket, offset by an\n estimate of the round-trip time between client and server. I.e.,\n\n adjusted_creation_time = creation_time + estimated_RTT\n\n This value can be encoded in the ticket, thus avoiding the need to\n keep state for each outstanding ticket. The server can determine the\n client's view of the age of the ticket by subtracting the ticket's\n \"ticket_age_add\" value from the \"obfuscated_ticket_age\" parameter in\n the client's \"pre_shared_key\" extension. The server can determine\n the expected_arrival_time of the ClientHello as:\n\n expected_arrival_time = adjusted_creation_time + clients_ticket_age\n\n When a new ClientHello is received, the expected_arrival_time is then\n compared against the current server wall clock time and if they\n differ by more than a certain amount, 0-RTT is rejected, though the\n 1-RTT handshake can be allowed to complete.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 101]\n\f\nRFC 8446 TLS August 2018\n\n\n There are several potential sources of error that might cause\n mismatches between the expected_arrival_time and the measured time.\n Variations in client and server clock rates are likely to be minimal,\n though potentially the absolute times may be off by large values.\n Network propagation delays are the most likely causes of a mismatch\n in legitimate values for elapsed time. Both the NewSessionTicket and\n ClientHello messages might be retransmitted and therefore delayed,\n which might be hidden by TCP. For clients on the Internet, this\n implies windows on the order of ten seconds to account for errors in\n clocks and variations in measurements; other deployment scenarios may\n have different needs. Clock skew distributions are not symmetric, so\n the optimal tradeoff may involve an asymmetric range of permissible\n mismatch values.\n\n Note that freshness checking alone is not sufficient to prevent\n replays because it does not detect them during the error window,\n which -- depending on bandwidth and system capacity -- could include\n billions of replays in real-world settings. In addition, this\n freshness checking is only done at the time the ClientHello is\n received and not when subsequent early Application Data records are\n received. After early data is accepted, records may continue to be\n streamed to the server over a longer time period.\n\n9. Compliance Requirements\n\n9.1. Mandatory-to-Implement Cipher Suites\n\n In the absence of an application profile standard specifying\n otherwise:\n\n A TLS-compliant application MUST implement the TLS_AES_128_GCM_SHA256\n [GCM] cipher suite and SHOULD implement the TLS_AES_256_GCM_SHA384\n [GCM] and TLS_CHACHA20_POLY1305_SHA256 [RFC8439] cipher suites (see\n Appendix B.4).\n\n A TLS-compliant application MUST support digital signatures with\n rsa_pkcs1_sha256 (for certificates), rsa_pss_rsae_sha256 (for\n CertificateVerify and certificates), and ecdsa_secp256r1_sha256. A\n TLS-compliant application MUST support key exchange with secp256r1\n (NIST P-256) and SHOULD support key exchange with X25519 [RFC7748].\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 102]\n\f\nRFC 8446 TLS August 2018\n\n\n9.2. Mandatory-to-Implement Extensions\n\n In the absence of an application profile standard specifying\n otherwise, a TLS-compliant application MUST implement the following\n TLS extensions:\n\n - Supported Versions (\"supported_versions\"; Section 4.2.1)\n\n - Cookie (\"cookie\"; Section 4.2.2)\n\n - Signature Algorithms (\"signature_algorithms\"; Section 4.2.3)\n\n - Signature Algorithms Certificate (\"signature_algorithms_cert\";\n Section 4.2.3)\n\n - Negotiated Groups (\"supported_groups\"; Section 4.2.7)\n\n - Key Share (\"key_share\"; Section 4.2.8)\n\n - Server Name Indication (\"server_name\"; Section 3 of [RFC6066])\n\n All implementations MUST send and use these extensions when offering\n applicable features:\n\n - \"supported_versions\" is REQUIRED for all ClientHello, ServerHello,\n and HelloRetryRequest messages.\n\n - \"signature_algorithms\" is REQUIRED for certificate authentication.\n\n - \"supported_groups\" is REQUIRED for ClientHello messages using DHE\n or ECDHE key exchange.\n\n - \"key_share\" is REQUIRED for DHE or ECDHE key exchange.\n\n - \"pre_shared_key\" is REQUIRED for PSK key agreement.\n\n - \"psk_key_exchange_modes\" is REQUIRED for PSK key agreement.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 103]\n\f\nRFC 8446 TLS August 2018\n\n\n A client is considered to be attempting to negotiate using this\n specification if the ClientHello contains a \"supported_versions\"\n extension with 0x0304 contained in its body. Such a ClientHello\n message MUST meet the following requirements:\n\n - If not containing a \"pre_shared_key\" extension, it MUST contain\n both a \"signature_algorithms\" extension and a \"supported_groups\"\n extension.\n\n - If containing a \"supported_groups\" extension, it MUST also contain\n a \"key_share\" extension, and vice versa. An empty\n KeyShare.client_shares vector is permitted.\n\n Servers receiving a ClientHello which does not conform to these\n requirements MUST abort the handshake with a \"missing_extension\"\n alert.\n\n Additionally, all implementations MUST support the use of the\n \"server_name\" extension with applications capable of using it.\n Servers MAY require clients to send a valid \"server_name\" extension.\n Servers requiring this extension SHOULD respond to a ClientHello\n lacking a \"server_name\" extension by terminating the connection with\n a \"missing_extension\" alert.\n\n9.3. Protocol Invariants\n\n This section describes invariants that TLS endpoints and middleboxes\n MUST follow. It also applies to earlier versions of TLS.\n\n TLS is designed to be securely and compatibly extensible. Newer\n clients or servers, when communicating with newer peers, should\n negotiate the most preferred common parameters. The TLS handshake\n provides downgrade protection: Middleboxes passing traffic between a\n newer client and newer server without terminating TLS should be\n unable to influence the handshake (see Appendix E.1). At the same\n time, deployments update at different rates, so a newer client or\n server MAY continue to support older parameters, which would allow it\n to interoperate with older endpoints.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 104]\n\f\nRFC 8446 TLS August 2018\n\n\n For this to work, implementations MUST correctly handle extensible\n fields:\n\n - A client sending a ClientHello MUST support all parameters\n advertised in it. Otherwise, the server may fail to interoperate\n by selecting one of those parameters.\n\n - A server receiving a ClientHello MUST correctly ignore all\n unrecognized cipher suites, extensions, and other parameters.\n Otherwise, it may fail to interoperate with newer clients. In\n TLS 1.3, a client receiving a CertificateRequest or\n NewSessionTicket MUST also ignore all unrecognized extensions.\n\n - A middlebox which terminates a TLS connection MUST behave as a\n compliant TLS server (to the original client), including having a\n certificate which the client is willing to accept, and also as a\n compliant TLS client (to the original server), including verifying\n the original server's certificate. In particular, it MUST\n generate its own ClientHello containing only parameters it\n understands, and it MUST generate a fresh ServerHello random\n value, rather than forwarding the endpoint's value.\n\n Note that TLS's protocol requirements and security analysis only\n apply to the two connections separately. Safely deploying a TLS\n terminator requires additional security considerations which are\n beyond the scope of this document.\n\n - A middlebox which forwards ClientHello parameters it does not\n understand MUST NOT process any messages beyond that ClientHello.\n It MUST forward all subsequent traffic unmodified. Otherwise, it\n may fail to interoperate with newer clients and servers.\n\n Forwarded ClientHellos may contain advertisements for features not\n supported by the middlebox, so the response may include future TLS\n additions the middlebox does not recognize. These additions MAY\n change any message beyond the ClientHello arbitrarily. In\n particular, the values sent in the ServerHello might change, the\n ServerHello format might change, and the TLSCiphertext format\n might change.\n\n The design of TLS 1.3 was constrained by widely deployed\n non-compliant TLS middleboxes (see Appendix D.4); however, it does\n not relax the invariants. Those middleboxes continue to be\n non-compliant.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 105]\n\f\nRFC 8446 TLS August 2018\n\n\n10. Security Considerations\n\n Security issues are discussed throughout this memo, especially in\n Appendices C, D, and E.\n\n11. IANA Considerations\n\n This document uses several registries that were originally created in\n [RFC4346] and updated in [RFC8447]. IANA has updated these to\n reference this document. The registries and their allocation\n policies are below:\n\n - TLS Cipher Suites registry: values with the first byte in the\n range 0-254 (decimal) are assigned via Specification Required\n [RFC8126]. Values with the first byte 255 (decimal) are reserved\n for Private Use [RFC8126].\n\n IANA has added the cipher suites listed in Appendix B.4 to the\n registry. The \"Value\" and \"Description\" columns are taken from\n the table. The \"DTLS-OK\" and \"Recommended\" columns are both\n marked as \"Y\" for each new cipher suite.\n\n - TLS ContentType registry: Future values are allocated via\n Standards Action [RFC8126].\n\n - TLS Alerts registry: Future values are allocated via Standards\n Action [RFC8126]. IANA has populated this registry with the\n values from Appendix B.2. The \"DTLS-OK\" column is marked as \"Y\"\n for all such values. Values marked as \"_RESERVED\" have comments\n describing their previous usage.\n\n - TLS HandshakeType registry: Future values are allocated via\n Standards Action [RFC8126]. IANA has updated this registry to\n rename item 4 from \"NewSessionTicket\" to \"new_session_ticket\" and\n populated this registry with the values from Appendix B.3. The\n \"DTLS-OK\" column is marked as \"Y\" for all such values. Values\n marked \"_RESERVED\" have comments describing their previous or\n temporary usage.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 106]\n\f\nRFC 8446 TLS August 2018\n\n\n This document also uses the TLS ExtensionType Values registry\n originally created in [RFC4366]. IANA has updated it to reference\n this document. Changes to the registry follow:\n\n - IANA has updated the registration policy as follows:\n\n Values with the first byte in the range 0-254 (decimal) are\n assigned via Specification Required [RFC8126]. Values with the\n first byte 255 (decimal) are reserved for Private Use [RFC8126].\n\n - IANA has updated this registry to include the \"key_share\",\n \"pre_shared_key\", \"psk_key_exchange_modes\", \"early_data\",\n \"cookie\", \"supported_versions\", \"certificate_authorities\",\n \"oid_filters\", \"post_handshake_auth\", and\n \"signature_algorithms_cert\" extensions with the values defined in\n this document and the \"Recommended\" value of \"Y\".\n\n - IANA has updated this registry to include a \"TLS 1.3\" column which\n lists the messages in which the extension may appear. This column\n has been initially populated from the table in Section 4.2, with\n any extension not listed there marked as \"-\" to indicate that it\n is not used by TLS 1.3.\n\n This document updates an entry in the TLS Certificate Types registry\n originally created in [RFC6091] and updated in [RFC8447]. IANA has\n updated the entry for value 1 to have the name \"OpenPGP_RESERVED\",\n \"Recommended\" value \"N\", and comment \"Used in TLS versions prior\n to 1.3.\"\n\n This document updates an entry in the TLS Certificate Status Types\n registry originally created in [RFC6961]. IANA has updated the entry\n for value 2 to have the name \"ocsp_multi_RESERVED\" and comment \"Used\n in TLS versions prior to 1.3\".\n\n This document updates two entries in the TLS Supported Groups\n registry (created under a different name by [RFC4492]; now maintained\n by [RFC8422]) and updated by [RFC7919] and [RFC8447]. The entries\n for values 29 and 30 (x25519 and x448) have been updated to also\n refer to this document.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 107]\n\f\nRFC 8446 TLS August 2018\n\n\n In addition, this document defines two new registries that are\n maintained by IANA:\n\n - TLS SignatureScheme registry: Values with the first byte in the\n range 0-253 (decimal) are assigned via Specification Required\n [RFC8126]. Values with the first byte 254 or 255 (decimal) are\n reserved for Private Use [RFC8126]. Values with the first byte in\n the range 0-6 or with the second byte in the range 0-3 that are\n not currently allocated are reserved for backward compatibility.\n This registry has a \"Recommended\" column. The registry has been\n initially populated with the values described in Section 4.2.3.\n The following values are marked as \"Recommended\":\n ecdsa_secp256r1_sha256, ecdsa_secp384r1_sha384,\n rsa_pss_rsae_sha256, rsa_pss_rsae_sha384, rsa_pss_rsae_sha512,\n rsa_pss_pss_sha256, rsa_pss_pss_sha384, rsa_pss_pss_sha512, and\n ed25519. The \"Recommended\" column is assigned a value of \"N\"\n unless explicitly requested, and adding a value with a\n \"Recommended\" value of \"Y\" requires Standards Action [RFC8126].\n IESG Approval is REQUIRED for a Y->N transition.\n\n - TLS PskKeyExchangeMode registry: Values in the range 0-253\n (decimal) are assigned via Specification Required [RFC8126].\n The values 254 and 255 (decimal) are reserved for Private Use\n [RFC8126]. This registry has a \"Recommended\" column. The\n registry has been initially populated with psk_ke (0) and\n psk_dhe_ke (1). Both are marked as \"Recommended\". The\n \"Recommended\" column is assigned a value of \"N\" unless explicitly\n requested, and adding a value with a \"Recommended\" value of \"Y\"\n requires Standards Action [RFC8126]. IESG Approval is REQUIRED\n for a Y->N transition.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 108]\n\f\nRFC 8446 TLS August 2018\n\n\n12. References\n\n12.1. Normative References\n\n [DH76] Diffie, W. and M. Hellman, \"New directions in\n cryptography\", IEEE Transactions on Information\n Theory, Vol. 22 No. 6, pp. 644-654,\n DOI 10.1109/TIT.1976.1055638, November 1976.\n\n [ECDSA] American National Standards Institute, \"Public Key\n Cryptography for the Financial Services Industry: The\n Elliptic Curve Digital Signature Algorithm (ECDSA)\",\n ANSI ANS X9.62-2005, November 2005.\n\n [GCM] Dworkin, M., \"Recommendation for Block Cipher Modes of\n Operation: Galois/Counter Mode (GCM) and GMAC\",\n NIST Special Publication 800-38D,\n DOI 10.6028/NIST.SP.800-38D, November 2007.\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC: Keyed-\n Hashing for Message Authentication\", RFC 2104,\n DOI 10.17487/RFC2104, February 1997,\n .\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119,\n DOI 10.17487/RFC2119, March 1997,\n .\n\n [RFC5116] McGrew, D., \"An Interface and Algorithms for Authenticated\n Encryption\", RFC 5116, DOI 10.17487/RFC5116, January 2008,\n .\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, DOI 10.17487/RFC5280, May 2008,\n .\n\n [RFC5705] Rescorla, E., \"Keying Material Exporters for Transport\n Layer Security (TLS)\", RFC 5705, DOI 10.17487/RFC5705,\n March 2010, .\n\n [RFC5756] Turner, S., Brown, D., Yiu, K., Housley, R., and T. Polk,\n \"Updates for RSAES-OAEP and RSASSA-PSS Algorithm\n Parameters\", RFC 5756, DOI 10.17487/RFC5756, January 2010,\n .\n\n\n\n\nRescorla Standards Track [Page 109]\n\f\nRFC 8446 TLS August 2018\n\n\n [RFC5869] Krawczyk, H. and P. Eronen, \"HMAC-based Extract-and-Expand\n Key Derivation Function (HKDF)\", RFC 5869,\n DOI 10.17487/RFC5869, May 2010,\n .\n\n [RFC6066] Eastlake 3rd, D., \"Transport Layer Security (TLS)\n Extensions: Extension Definitions\", RFC 6066,\n DOI 10.17487/RFC6066, January 2011,\n .\n\n [RFC6655] McGrew, D. and D. Bailey, \"AES-CCM Cipher Suites for\n Transport Layer Security (TLS)\", RFC 6655,\n DOI 10.17487/RFC6655, July 2012,\n .\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, DOI 10.17487/RFC6960, June 2013,\n .\n\n [RFC6961] Pettersen, Y., \"The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\", RFC 6961,\n DOI 10.17487/RFC6961, June 2013,\n .\n\n [RFC6962] Laurie, B., Langley, A., and E. Kasper, \"Certificate\n Transparency\", RFC 6962, DOI 10.17487/RFC6962, June 2013,\n .\n\n [RFC6979] Pornin, T., \"Deterministic Usage of the Digital Signature\n Algorithm (DSA) and Elliptic Curve Digital Signature\n Algorithm (ECDSA)\", RFC 6979, DOI 10.17487/RFC6979,\n August 2013, .\n\n [RFC7301] Friedl, S., Popov, A., Langley, A., and E. Stephan,\n \"Transport Layer Security (TLS) Application-Layer Protocol\n Negotiation Extension\", RFC 7301, DOI 10.17487/RFC7301,\n July 2014, .\n\n [RFC7507] Moeller, B. and A. Langley, \"TLS Fallback Signaling Cipher\n Suite Value (SCSV) for Preventing Protocol Downgrade\n Attacks\", RFC 7507, DOI 10.17487/RFC7507, April 2015,\n .\n\n [RFC7748] Langley, A., Hamburg, M., and S. Turner, \"Elliptic Curves\n for Security\", RFC 7748, DOI 10.17487/RFC7748,\n January 2016, .\n\n\n\nRescorla Standards Track [Page 110]\n\f\nRFC 8446 TLS August 2018\n\n\n [RFC7919] Gillmor, D., \"Negotiated Finite Field Diffie-Hellman\n Ephemeral Parameters for Transport Layer Security (TLS)\",\n RFC 7919, DOI 10.17487/RFC7919, August 2016,\n .\n\n [RFC8017] Moriarty, K., Ed., Kaliski, B., Jonsson, J., and A. Rusch,\n \"PKCS #1: RSA Cryptography Specifications Version 2.2\",\n RFC 8017, DOI 10.17487/RFC8017, November 2016,\n .\n\n [RFC8032] Josefsson, S. and I. Liusvaara, \"Edwards-Curve Digital\n Signature Algorithm (EdDSA)\", RFC 8032,\n DOI 10.17487/RFC8032, January 2017,\n .\n\n [RFC8126] Cotton, M., Leiba, B., and T. Narten, \"Guidelines for\n Writing an IANA Considerations Section in RFCs\", BCP 26,\n RFC 8126, DOI 10.17487/RFC8126, June 2017,\n .\n\n [RFC8174] Leiba, B., \"Ambiguity of Uppercase vs Lowercase in\n RFC 2119 Key Words\", BCP 14, RFC 8174,\n DOI 10.17487/RFC8174, May 2017,\n .\n\n [RFC8439] Nir, Y. and A. Langley, \"ChaCha20 and Poly1305 for IETF\n Protocols\", RFC 8439, DOI 10.17487/RFC8439, June 2018,\n .\n\n [SHS] Dang, Q., \"Secure Hash Standard (SHS)\", National Institute\n of Standards and Technology report,\n DOI 10.6028/NIST.FIPS.180-4, August 2015.\n\n [X690] ITU-T, \"Information technology -- ASN.1 encoding rules:\n Specification of Basic Encoding Rules (BER), Canonical\n Encoding Rules (CER) and Distinguished Encoding Rules\n (DER)\", ISO/IEC 8825-1:2015, November 2015.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 111]\n\f\nRFC 8446 TLS August 2018\n\n\n12.2. Informative References\n\n [AEAD-LIMITS]\n Luykx, A. and K. Paterson, \"Limits on Authenticated\n Encryption Use in TLS\", August 2017,\n .\n\n [BBFGKZ16]\n Bhargavan, K., Brzuska, C., Fournet, C., Green, M.,\n Kohlweiss, M., and S. Zanella-Beguelin, \"Downgrade\n Resilience in Key-Exchange Protocols\", Proceedings of IEEE\n Symposium on Security and Privacy (San Jose),\n DOI 10.1109/SP.2016.37, May 2016.\n\n [BBK17] Bhargavan, K., Blanchet, B., and N. Kobeissi, \"Verified\n Models and Reference Implementations for the TLS 1.3\n Standard Candidate\", Proceedings of IEEE Symposium on\n Security and Privacy (San Jose), DOI 10.1109/SP.2017.26,\n May 2017.\n\n [BDFKPPRSZZ16]\n Bhargavan, K., Delignat-Lavaud, A., Fournet, C.,\n Kohlweiss, M., Pan, J., Protzenko, J., Rastogi, A., Swamy,\n N., Zanella-Beguelin, S., and J. Zinzindohoue,\n \"Implementing and Proving the TLS 1.3 Record Layer\",\n Proceedings of IEEE Symposium on Security and Privacy (San\n Jose), May 2017, .\n\n [Ben17a] Benjamin, D., \"Presentation before the TLS WG at\n IETF 100\", November 2017,\n .\n\n [Ben17b] Benjamin, D., \"Additional TLS 1.3 results from Chrome\",\n message to the TLS mailing list, 18 December 2017,\n .\n\n [Blei98] Bleichenbacher, D., \"Chosen Ciphertext Attacks against\n Protocols Based on RSA Encryption Standard PKCS #1\",\n Proceedings of CRYPTO '98, 1998.\n\n [BMMRT15] Badertscher, C., Matt, C., Maurer, U., Rogaway, P., and B.\n Tackmann, \"Augmented Secure Channels and the Goal of the\n TLS 1.3 Record Layer\", ProvSec 2015, September 2015,\n .\n\n\n\n\n\nRescorla Standards Track [Page 112]\n\f\nRFC 8446 TLS August 2018\n\n\n [BT16] Bellare, M. and B. Tackmann, \"The Multi-User Security of\n Authenticated Encryption: AES-GCM in TLS 1.3\", Proceedings\n of CRYPTO 2016, July 2016,\n .\n\n [CCG16] Cohn-Gordon, K., Cremers, C., and L. Garratt, \"On\n Post-compromise Security\", IEEE Computer Security\n Foundations Symposium, DOI 10.1109/CSF.2016.19, July 2015.\n\n [CHECKOWAY]\n Checkoway, S., Maskiewicz, J., Garman, C., Fried, J.,\n Cohney, S., Green, M., Heninger, N., Weinmann, R.,\n Rescorla, E., and H. Shacham, \"A Systematic Analysis of\n the Juniper Dual EC Incident\", Proceedings of the 2016 ACM\n SIGSAC Conference on Computer and Communications Security\n - CCS '16, DOI 10.1145/2976749.2978395, October 2016.\n\n [CHHSV17] Cremers, C., Horvat, M., Hoyland, J., Scott, S., and T.\n van der Merwe, \"Awkward Handshake: Possible mismatch of\n client/server view on client authentication in\n post-handshake mode in Revision 18\", message to the TLS\n mailing list, 10 February 2017, .\n\n [CHSV16] Cremers, C., Horvat, M., Scott, S., and T. van der Merwe,\n \"Automated Analysis and Verification of TLS 1.3: 0-RTT,\n Resumption and Delayed Authentication\", Proceedings of\n IEEE Symposium on Security and Privacy (San Jose),\n DOI 10.1109/SP.2016.35, May 2016,\n .\n\n [CK01] Canetti, R. and H. Krawczyk, \"Analysis of Key-Exchange\n Protocols and Their Use for Building Secure Channels\",\n Proceedings of Eurocrypt 2001,\n DOI 10.1007/3-540-44987-6_28, April 2001.\n\n [CLINIC] Miller, B., Huang, L., Joseph, A., and J. Tygar, \"I Know\n Why You Went to the Clinic: Risks and Realization of HTTPS\n Traffic Analysis\", Privacy Enhancing Technologies, pp.\n 143-163, DOI 10.1007/978-3-319-08506-7_8, 2014.\n\n [DFGS15] Dowling, B., Fischlin, M., Guenther, F., and D. Stebila,\n \"A Cryptographic Analysis of the TLS 1.3 Handshake\n Protocol Candidates\", Proceedings of ACM CCS 2015,\n October 2015, .\n\n\n\n\n\n\nRescorla Standards Track [Page 113]\n\f\nRFC 8446 TLS August 2018\n\n\n [DFGS16] Dowling, B., Fischlin, M., Guenther, F., and D. Stebila,\n \"A Cryptographic Analysis of the TLS 1.3 Full and\n Pre-shared Key Handshake Protocol\", TRON 2016,\n February 2016, .\n\n [DOW92] Diffie, W., van Oorschot, P., and M. Wiener,\n \"Authentication and authenticated key exchanges\", Designs,\n Codes and Cryptography, DOI 10.1007/BF00124891, June 1992.\n\n [DSS] National Institute of Standards and Technology, U.S.\n Department of Commerce, \"Digital Signature Standard\n (DSS)\", NIST FIPS PUB 186-4, DOI 10.6028/NIST.FIPS.186-4,\n July 2013.\n\n [FG17] Fischlin, M. and F. Guenther, \"Replay Attacks on Zero\n Round-Trip Time: The Case of the TLS 1.3 Handshake\n Candidates\", Proceedings of EuroS&P 2017, April 2017,\n .\n\n [FGSW16] Fischlin, M., Guenther, F., Schmidt, B., and B. Warinschi,\n \"Key Confirmation in Key Exchange: A Formal Treatment and\n Implications for TLS 1.3\", Proceedings of IEEE Symposium\n on Security and Privacy (San Jose),\n DOI 10.1109/SP.2016.34, May 2016,\n .\n\n [FW15] Weimer, F., \"Factoring RSA Keys With TLS Perfect Forward\n Secrecy\", September 2015.\n\n [HCJC16] Husak, M., Cermak, M., Jirsik, T., and P. Celeda, \"HTTPS\n traffic analysis and client identification using passive\n SSL/TLS fingerprinting\", EURASIP Journal on Information\n Security, Vol. 2016, DOI 10.1186/s13635-016-0030-7,\n February 2016.\n\n [HGFS15] Hlauschek, C., Gruber, M., Fankhauser, F., and C. Schanes,\n \"Prying Open Pandora's Box: KCI Attacks against TLS\",\n Proceedings of USENIX Workshop on Offensive Technologies,\n August 2015.\n\n [IEEE1363]\n IEEE, \"IEEE Standard Specifications for Public Key\n Cryptography\", IEEE Std. 1363-2000,\n DOI 10.1109/IEEESTD.2000.92292.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 114]\n\f\nRFC 8446 TLS August 2018\n\n\n [JSS15] Jager, T., Schwenk, J., and J. Somorovsky, \"On the\n Security of TLS 1.3 and QUIC Against Weaknesses in PKCS#1\n v1.5 Encryption\", Proceedings of ACM CCS 2015,\n DOI 10.1145/2810103.2813657, October 2015,\n .\n\n [KEYAGREEMENT]\n Barker, E., Chen, L., Roginsky, A., Vassilev, A., and R.\n Davis, \"Recommendation for Pair-Wise Key Establishment\n Schemes Using Discrete Logarithm Cryptography\", National\n Institute of Standards and Technology,\n DOI 10.6028/NIST.SP.800-56Ar3, April 2018.\n\n [Kraw10] Krawczyk, H., \"Cryptographic Extraction and Key\n Derivation: The HKDF Scheme\", Proceedings of CRYPTO 2010,\n August 2010, .\n\n [Kraw16] Krawczyk, H., \"A Unilateral-to-Mutual Authentication\n Compiler for Key Exchange (with Applications to Client\n Authentication in TLS 1.3\", Proceedings of ACM CCS 2016,\n October 2016, .\n\n [KW16] Krawczyk, H. and H. Wee, \"The OPTLS Protocol and TLS 1.3\",\n Proceedings of EuroS&P 2016, March 2016,\n .\n\n [LXZFH16] Li, X., Xu, J., Zhang, Z., Feng, D., and H. Hu, \"Multiple\n Handshakes Security of TLS 1.3 Candidates\", Proceedings of\n IEEE Symposium on Security and Privacy (San Jose),\n DOI 10.1109/SP.2016.36, May 2016,\n .\n\n [Mac17] MacCarthaigh, C., \"Security Review of TLS1.3 0-RTT\",\n March 2017, .\n\n [PS18] Patton, C. and T. Shrimpton, \"Partially specified\n channels: The TLS 1.3 record layer without elision\", 2018,\n .\n\n [PSK-FINISHED]\n Scott, S., Cremers, C., Horvat, M., and T. van der Merwe,\n \"Revision 10: possible attack if client authentication is\n allowed during PSK\", message to the TLS mailing list,\n 31 October 2015, .\n\n\n\n\nRescorla Standards Track [Page 115]\n\f\nRFC 8446 TLS August 2018\n\n\n [REKEY] Abdalla, M. and M. Bellare, \"Increasing the Lifetime of a\n Key: A Comparative Analysis of the Security of Re-keying\n Techniques\", ASIACRYPT 2000, DOI 10.1007/3-540-44448-3_42,\n October 2000.\n\n [Res17a] Rescorla, E., \"Preliminary data on Firefox TLS 1.3\n Middlebox experiment\", message to the TLS mailing list,\n 5 December 2017, .\n\n [Res17b] Rescorla, E., \"More compatibility measurement results\",\n message to the TLS mailing list, 22 December 2017,\n .\n\n [RFC3552] Rescorla, E. and B. Korver, \"Guidelines for Writing RFC\n Text on Security Considerations\", BCP 72, RFC 3552,\n DOI 10.17487/RFC3552, July 2003,\n .\n\n [RFC4086] Eastlake 3rd, D., Schiller, J., and S. Crocker,\n \"Randomness Requirements for Security\", BCP 106, RFC 4086,\n DOI 10.17487/RFC4086, June 2005,\n .\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346,\n DOI 10.17487/RFC4346, April 2006,\n .\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, DOI 10.17487/RFC4366, April 2006,\n .\n\n [RFC4492] Blake-Wilson, S., Bolyard, N., Gupta, V., Hawk, C., and B.\n Moeller, \"Elliptic Curve Cryptography (ECC) Cipher Suites\n for Transport Layer Security (TLS)\", RFC 4492,\n DOI 10.17487/RFC4492, May 2006,\n .\n\n [RFC5077] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 5077, DOI 10.17487/RFC5077,\n January 2008, .\n\n\n\n\n\n\nRescorla Standards Track [Page 116]\n\f\nRFC 8446 TLS August 2018\n\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246,\n DOI 10.17487/RFC5246, August 2008,\n .\n\n [RFC5764] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for the Secure\n Real-time Transport Protocol (SRTP)\", RFC 5764,\n DOI 10.17487/RFC5764, May 2010,\n .\n\n [RFC5929] Altman, J., Williams, N., and L. Zhu, \"Channel Bindings\n for TLS\", RFC 5929, DOI 10.17487/RFC5929, July 2010,\n .\n\n [RFC6091] Mavrogiannopoulos, N. and D. Gillmor, \"Using OpenPGP Keys\n for Transport Layer Security (TLS) Authentication\",\n RFC 6091, DOI 10.17487/RFC6091, February 2011,\n .\n\n [RFC6101] Freier, A., Karlton, P., and P. Kocher, \"The Secure\n Sockets Layer (SSL) Protocol Version 3.0\", RFC 6101,\n DOI 10.17487/RFC6101, August 2011,\n .\n\n [RFC6176] Turner, S. and T. Polk, \"Prohibiting Secure Sockets Layer\n (SSL) Version 2.0\", RFC 6176, DOI 10.17487/RFC6176,\n March 2011, .\n\n [RFC6347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security Version 1.2\", RFC 6347, DOI 10.17487/RFC6347,\n January 2012, .\n\n [RFC6520] Seggelmann, R., Tuexen, M., and M. Williams, \"Transport\n Layer Security (TLS) and Datagram Transport Layer Security\n (DTLS) Heartbeat Extension\", RFC 6520,\n DOI 10.17487/RFC6520, February 2012,\n .\n\n [RFC7230] Fielding, R., Ed. and J. Reschke, Ed., \"Hypertext Transfer\n Protocol (HTTP/1.1): Message Syntax and Routing\",\n RFC 7230, DOI 10.17487/RFC7230, June 2014,\n .\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 117]\n\f\nRFC 8446 TLS August 2018\n\n\n [RFC7250] Wouters, P., Ed., Tschofenig, H., Ed., Gilmore, J.,\n Weiler, S., and T. Kivinen, \"Using Raw Public Keys in\n Transport Layer Security (TLS) and Datagram Transport\n Layer Security (DTLS)\", RFC 7250, DOI 10.17487/RFC7250,\n June 2014, .\n\n [RFC7465] Popov, A., \"Prohibiting RC4 Cipher Suites\", RFC 7465,\n DOI 10.17487/RFC7465, February 2015,\n .\n\n [RFC7568] Barnes, R., Thomson, M., Pironti, A., and A. Langley,\n \"Deprecating Secure Sockets Layer Version 3.0\", RFC 7568,\n DOI 10.17487/RFC7568, June 2015,\n .\n\n [RFC7627] Bhargavan, K., Ed., Delignat-Lavaud, A., Pironti, A.,\n Langley, A., and M. Ray, \"Transport Layer Security (TLS)\n Session Hash and Extended Master Secret Extension\",\n RFC 7627, DOI 10.17487/RFC7627, September 2015,\n .\n\n [RFC7685] Langley, A., \"A Transport Layer Security (TLS) ClientHello\n Padding Extension\", RFC 7685, DOI 10.17487/RFC7685,\n October 2015, .\n\n [RFC7924] Santesson, S. and H. Tschofenig, \"Transport Layer Security\n (TLS) Cached Information Extension\", RFC 7924,\n DOI 10.17487/RFC7924, July 2016,\n .\n\n [RFC8305] Schinazi, D. and T. Pauly, \"Happy Eyeballs Version 2:\n Better Connectivity Using Concurrency\", RFC 8305,\n DOI 10.17487/RFC8305, December 2017,\n .\n\n [RFC8422] Nir, Y., Josefsson, S., and M. Pegourie-Gonnard, \"Elliptic\n Curve Cryptography (ECC) Cipher Suites for Transport Layer\n Security (TLS) Versions 1.2 and Earlier\", RFC 8422,\n DOI 10.17487/RFC8422, August 2018,\n .\n\n [RFC8447] Salowey, J. and S. Turner, \"IANA Registry Updates for TLS\n and DTLS\", RFC 8447, DOI 10.17487/RFC8447, August 2018,\n .\n\n [RFC8449] Thomson, M., \"Record Size Limit Extension for TLS\",\n RFC 8449, DOI 10.17487/RFC8449, August 2018,\n .\n\n\n\nRescorla Standards Track [Page 118]\n\f\nRFC 8446 TLS August 2018\n\n\n [RSA] Rivest, R., Shamir, A., and L. Adleman, \"A Method for\n Obtaining Digital Signatures and Public-Key\n Cryptosystems\", Communications of the ACM, Vol. 21 No. 2,\n pp. 120-126, DOI 10.1145/359340.359342, February 1978.\n\n [SIGMA] Krawczyk, H., \"SIGMA: The 'SIGn-and-MAc' Approach to\n Authenticated Diffie-Hellman and its Use in the IKE\n Protocols\", Proceedings of CRYPTO 2003,\n DOI 10.1007/978-3-540-45146-4_24, August 2003.\n\n [SLOTH] Bhargavan, K. and G. Leurent, \"Transcript Collision\n Attacks: Breaking Authentication in TLS, IKE, and SSH\",\n Network and Distributed System Security\n Symposium (NDSS 2016), DOI 10.14722/ndss.2016.23418,\n February 2016.\n\n [SSL2] Hickman, K., \"The SSL Protocol\", February 1995.\n\n [TIMING] Boneh, D. and D. Brumley, \"Remote Timing Attacks Are\n Practical\", USENIX Security Symposium, August 2003.\n\n [TLS13-TRACES]\n Thomson, M., \"Example Handshake Traces for TLS 1.3\", Work\n in Progress, draft-ietf-tls-tls13-vectors-06, July 2018.\n\n [X501] ITU-T, \"Information Technology - Open Systems\n Interconnection - The Directory: Models\", ITU-T X.501,\n October 2016, .\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 119]\n\f\nRFC 8446 TLS August 2018\n\n\nAppendix A. State Machine\n\n This appendix provides a summary of the legal state transitions for\n the client and server handshakes. State names (in all capitals,\n e.g., START) have no formal meaning but are provided for ease of\n comprehension. Actions which are taken only in certain circumstances\n are indicated in []. The notation \"K_{send,recv} = foo\" means \"set\n the send/recv key to the given key\".\n\nA.1. Client\n\n START <----+\n Send ClientHello | | Recv HelloRetryRequest\n [K_send = early data] | |\n v |\n / WAIT_SH ----+\n | | Recv ServerHello\n | | K_recv = handshake\n Can | V\n send | WAIT_EE\n early | | Recv EncryptedExtensions\n data | +--------+--------+\n | Using | | Using certificate\n | PSK | v\n | | WAIT_CERT_CR\n | | Recv | | Recv CertificateRequest\n | | Certificate | v\n | | | WAIT_CERT\n | | | | Recv Certificate\n | | v v\n | | WAIT_CV\n | | | Recv CertificateVerify\n | +> WAIT_FINISHED <+\n | | Recv Finished\n \\ | [Send EndOfEarlyData]\n | K_send = handshake\n | [Send Certificate [+ CertificateVerify]]\n Can send | Send Finished\n app data --> | K_send = K_recv = application\n after here v\n CONNECTED\n\n Note that with the transitions as shown above, clients may send\n alerts that derive from post-ServerHello messages in the clear or\n with the early data keys. If clients need to send such alerts, they\n SHOULD first rekey to the handshake keys if possible.\n\n\n\n\n\nRescorla Standards Track [Page 120]\n\f\nRFC 8446 TLS August 2018\n\n\nA.2. Server\n\n START <-----+\n Recv ClientHello | | Send HelloRetryRequest\n v |\n RECVD_CH ----+\n | Select parameters\n v\n NEGOTIATED\n | Send ServerHello\n | K_send = handshake\n | Send EncryptedExtensions\n | [Send CertificateRequest]\n Can send | [Send Certificate + CertificateVerify]\n app data | Send Finished\n after --> | K_send = application\n here +--------+--------+\n No 0-RTT | | 0-RTT\n | |\n K_recv = handshake | | K_recv = early data\n [Skip decrypt errors] | +------> WAIT_EOED -+\n | | Recv | | Recv EndOfEarlyData\n | | early data | | K_recv = handshake\n | +------------+ |\n | |\n +> WAIT_FLIGHT2 <--------+\n |\n +--------+--------+\n No auth | | Client auth\n | |\n | v\n | WAIT_CERT\n | Recv | | Recv Certificate\n | empty | v\n | Certificate | WAIT_CV\n | | | Recv\n | v | CertificateVerify\n +-> WAIT_FINISHED <---+\n | Recv Finished\n | K_recv = application\n v\n CONNECTED\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 121]\n\f\nRFC 8446 TLS August 2018\n\n\nAppendix B. Protocol Data Structures and Constant Values\n\n This appendix provides the normative protocol types and the\n definitions for constants. Values listed as \"_RESERVED\" were used in\n previous versions of TLS and are listed here for completeness.\n TLS 1.3 implementations MUST NOT send them but might receive them\n from older TLS implementations.\n\nB.1. Record Layer\n\n enum {\n invalid(0),\n change_cipher_spec(20),\n alert(21),\n handshake(22),\n application_data(23),\n heartbeat(24), /* RFC 6520 */\n (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion legacy_record_version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n struct {\n opaque content[TLSPlaintext.length];\n ContentType type;\n uint8 zeros[length_of_padding];\n } TLSInnerPlaintext;\n\n struct {\n ContentType opaque_type = application_data; /* 23 */\n ProtocolVersion legacy_record_version = 0x0303; /* TLS v1.2 */\n uint16 length;\n opaque encrypted_record[TLSCiphertext.length];\n } TLSCiphertext;\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 122]\n\f\nRFC 8446 TLS August 2018\n\n\nB.2. Alert Messages\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure_RESERVED(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n inappropriate_fallback(86),\n user_canceled(90),\n no_renegotiation_RESERVED(100),\n missing_extension(109),\n unsupported_extension(110),\n certificate_unobtainable_RESERVED(111),\n unrecognized_name(112),\n bad_certificate_status_response(113),\n bad_certificate_hash_value_RESERVED(114),\n unknown_psk_identity(115),\n certificate_required(116),\n no_application_protocol(120),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n\n\n\n\nRescorla Standards Track [Page 123]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3. Handshake Protocol\n\n enum {\n hello_request_RESERVED(0),\n client_hello(1),\n server_hello(2),\n hello_verify_request_RESERVED(3),\n new_session_ticket(4),\n end_of_early_data(5),\n hello_retry_request_RESERVED(6),\n encrypted_extensions(8),\n certificate(11),\n server_key_exchange_RESERVED(12),\n certificate_request(13),\n server_hello_done_RESERVED(14),\n certificate_verify(15),\n client_key_exchange_RESERVED(16),\n finished(20),\n certificate_url_RESERVED(21),\n certificate_status_RESERVED(22),\n supplemental_data_RESERVED(23),\n key_update(24),\n message_hash(254),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (Handshake.msg_type) {\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case end_of_early_data: EndOfEarlyData;\n case encrypted_extensions: EncryptedExtensions;\n case certificate_request: CertificateRequest;\n case certificate: Certificate;\n case certificate_verify: CertificateVerify;\n case finished: Finished;\n case new_session_ticket: NewSessionTicket;\n case key_update: KeyUpdate;\n };\n } Handshake;\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 124]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.1. Key Exchange Messages\n\n uint16 ProtocolVersion;\n opaque Random[32];\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id<0..32>;\n CipherSuite cipher_suites<2..2^16-2>;\n opaque legacy_compression_methods<1..2^8-1>;\n Extension extensions<8..2^16-1>;\n } ClientHello;\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id_echo<0..32>;\n CipherSuite cipher_suite;\n uint8 legacy_compression_method = 0;\n Extension extensions<6..2^16-1>;\n } ServerHello;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 125]\n\f\nRFC 8446 TLS August 2018\n\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n server_name(0), /* RFC 6066 */\n max_fragment_length(1), /* RFC 6066 */\n status_request(5), /* RFC 6066 */\n supported_groups(10), /* RFC 8422, 7919 */\n signature_algorithms(13), /* RFC 8446 */\n use_srtp(14), /* RFC 5764 */\n heartbeat(15), /* RFC 6520 */\n application_layer_protocol_negotiation(16), /* RFC 7301 */\n signed_certificate_timestamp(18), /* RFC 6962 */\n client_certificate_type(19), /* RFC 7250 */\n server_certificate_type(20), /* RFC 7250 */\n padding(21), /* RFC 7685 */\n RESERVED(40), /* Used but never\n assigned */\n pre_shared_key(41), /* RFC 8446 */\n early_data(42), /* RFC 8446 */\n supported_versions(43), /* RFC 8446 */\n cookie(44), /* RFC 8446 */\n psk_key_exchange_modes(45), /* RFC 8446 */\n RESERVED(46), /* Used but never\n assigned */\n certificate_authorities(47), /* RFC 8446 */\n oid_filters(48), /* RFC 8446 */\n post_handshake_auth(49), /* RFC 8446 */\n signature_algorithms_cert(50), /* RFC 8446 */\n key_share(51), /* RFC 8446 */\n (65535)\n } ExtensionType;\n\n struct {\n NamedGroup group;\n opaque key_exchange<1..2^16-1>;\n } KeyShareEntry;\n\n struct {\n KeyShareEntry client_shares<0..2^16-1>;\n } KeyShareClientHello;\n\n struct {\n NamedGroup selected_group;\n } KeyShareHelloRetryRequest;\n\n\n\n\nRescorla Standards Track [Page 126]\n\f\nRFC 8446 TLS August 2018\n\n\n struct {\n KeyShareEntry server_share;\n } KeyShareServerHello;\n\n struct {\n uint8 legacy_form = 4;\n opaque X[coordinate_length];\n opaque Y[coordinate_length];\n } UncompressedPointRepresentation;\n\n enum { psk_ke(0), psk_dhe_ke(1), (255) } PskKeyExchangeMode;\n\n struct {\n PskKeyExchangeMode ke_modes<1..255>;\n } PskKeyExchangeModes;\n\n struct {} Empty;\n\n struct {\n select (Handshake.msg_type) {\n case new_session_ticket: uint32 max_early_data_size;\n case client_hello: Empty;\n case encrypted_extensions: Empty;\n };\n } EarlyDataIndication;\n\n struct {\n opaque identity<1..2^16-1>;\n uint32 obfuscated_ticket_age;\n } PskIdentity;\n\n opaque PskBinderEntry<32..255>;\n\n struct {\n PskIdentity identities<7..2^16-1>;\n PskBinderEntry binders<33..2^16-1>;\n } OfferedPsks;\n\n struct {\n select (Handshake.msg_type) {\n case client_hello: OfferedPsks;\n case server_hello: uint16 selected_identity;\n };\n } PreSharedKeyExtension;\n\n\n\n\n\n\n\nRescorla Standards Track [Page 127]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.1.1. Version Extension\n\n struct {\n select (Handshake.msg_type) {\n case client_hello:\n ProtocolVersion versions<2..254>;\n\n case server_hello: /* and HelloRetryRequest */\n ProtocolVersion selected_version;\n };\n } SupportedVersions;\n\nB.3.1.2. Cookie Extension\n\n struct {\n opaque cookie<1..2^16-1>;\n } Cookie;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 128]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.1.3. Signature Algorithm Extension\n\n enum {\n /* RSASSA-PKCS1-v1_5 algorithms */\n rsa_pkcs1_sha256(0x0401),\n rsa_pkcs1_sha384(0x0501),\n rsa_pkcs1_sha512(0x0601),\n\n /* ECDSA algorithms */\n ecdsa_secp256r1_sha256(0x0403),\n ecdsa_secp384r1_sha384(0x0503),\n ecdsa_secp521r1_sha512(0x0603),\n\n /* RSASSA-PSS algorithms with public key OID rsaEncryption */\n rsa_pss_rsae_sha256(0x0804),\n rsa_pss_rsae_sha384(0x0805),\n rsa_pss_rsae_sha512(0x0806),\n\n /* EdDSA algorithms */\n ed25519(0x0807),\n ed448(0x0808),\n\n /* RSASSA-PSS algorithms with public key OID RSASSA-PSS */\n rsa_pss_pss_sha256(0x0809),\n rsa_pss_pss_sha384(0x080a),\n rsa_pss_pss_sha512(0x080b),\n\n /* Legacy algorithms */\n rsa_pkcs1_sha1(0x0201),\n ecdsa_sha1(0x0203),\n\n /* Reserved Code Points */\n obsolete_RESERVED(0x0000..0x0200),\n dsa_sha1_RESERVED(0x0202),\n obsolete_RESERVED(0x0204..0x0400),\n dsa_sha256_RESERVED(0x0402),\n obsolete_RESERVED(0x0404..0x0500),\n dsa_sha384_RESERVED(0x0502),\n obsolete_RESERVED(0x0504..0x0600),\n dsa_sha512_RESERVED(0x0602),\n obsolete_RESERVED(0x0604..0x06FF),\n private_use(0xFE00..0xFFFF),\n (0xFFFF)\n } SignatureScheme;\n\n struct {\n SignatureScheme supported_signature_algorithms<2..2^16-2>;\n } SignatureSchemeList;\n\n\n\nRescorla Standards Track [Page 129]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.1.4. Supported Groups Extension\n\n enum {\n unallocated_RESERVED(0x0000),\n\n /* Elliptic Curve Groups (ECDHE) */\n obsolete_RESERVED(0x0001..0x0016),\n secp256r1(0x0017), secp384r1(0x0018), secp521r1(0x0019),\n obsolete_RESERVED(0x001A..0x001C),\n x25519(0x001D), x448(0x001E),\n\n /* Finite Field Groups (DHE) */\n ffdhe2048(0x0100), ffdhe3072(0x0101), ffdhe4096(0x0102),\n ffdhe6144(0x0103), ffdhe8192(0x0104),\n\n /* Reserved Code Points */\n ffdhe_private_use(0x01FC..0x01FF),\n ecdhe_private_use(0xFE00..0xFEFF),\n obsolete_RESERVED(0xFF01..0xFF02),\n (0xFFFF)\n } NamedGroup;\n\n struct {\n NamedGroup named_group_list<2..2^16-1>;\n } NamedGroupList;\n\n Values within \"obsolete_RESERVED\" ranges are used in previous\n versions of TLS and MUST NOT be offered or negotiated by TLS 1.3\n implementations. The obsolete curves have various known/theoretical\n weaknesses or have had very little usage, in some cases only due to\n unintentional server configuration issues. They are no longer\n considered appropriate for general use and should be assumed to be\n potentially unsafe. The set of curves specified here is sufficient\n for interoperability with all currently deployed and properly\n configured TLS implementations.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 130]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.2. Server Parameters Messages\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n DistinguishedName authorities<3..2^16-1>;\n } CertificateAuthoritiesExtension;\n\n struct {\n opaque certificate_extension_oid<1..2^8-1>;\n opaque certificate_extension_values<0..2^16-1>;\n } OIDFilter;\n\n struct {\n OIDFilter filters<0..2^16-1>;\n } OIDFilterExtension;\n\n struct {} PostHandshakeAuth;\n\n struct {\n Extension extensions<0..2^16-1>;\n } EncryptedExtensions;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n Extension extensions<2..2^16-1>;\n } CertificateRequest;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 131]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.3. Authentication Messages\n\n enum {\n X509(0),\n OpenPGP_RESERVED(1),\n RawPublicKey(2),\n (255)\n } CertificateType;\n\n struct {\n select (certificate_type) {\n case RawPublicKey:\n /* From RFC 7250 ASN.1_subjectPublicKeyInfo */\n opaque ASN1_subjectPublicKeyInfo<1..2^24-1>;\n\n case X509:\n opaque cert_data<1..2^24-1>;\n };\n Extension extensions<0..2^16-1>;\n } CertificateEntry;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n CertificateEntry certificate_list<0..2^24-1>;\n } Certificate;\n\n struct {\n SignatureScheme algorithm;\n opaque signature<0..2^16-1>;\n } CertificateVerify;\n\n struct {\n opaque verify_data[Hash.length];\n } Finished;\n\nB.3.4. Ticket Establishment\n\n struct {\n uint32 ticket_lifetime;\n uint32 ticket_age_add;\n opaque ticket_nonce<0..255>;\n opaque ticket<1..2^16-1>;\n Extension extensions<0..2^16-2>;\n } NewSessionTicket;\n\n\n\n\n\n\n\nRescorla Standards Track [Page 132]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.5. Updating Keys\n\n struct {} EndOfEarlyData;\n\n enum {\n update_not_requested(0), update_requested(1), (255)\n } KeyUpdateRequest;\n\n struct {\n KeyUpdateRequest request_update;\n } KeyUpdate;\n\nB.4. Cipher Suites\n\n A symmetric cipher suite defines the pair of the AEAD algorithm and\n hash algorithm to be used with HKDF. Cipher suite names follow the\n naming convention:\n\n CipherSuite TLS_AEAD_HASH = VALUE;\n\n +-----------+------------------------------------------------+\n | Component | Contents |\n +-----------+------------------------------------------------+\n | TLS | The string \"TLS\" |\n | | |\n | AEAD | The AEAD algorithm used for record protection |\n | | |\n | HASH | The hash algorithm used with HKDF |\n | | |\n | VALUE | The two-byte ID assigned for this cipher suite |\n +-----------+------------------------------------------------+\n\n This specification defines the following cipher suites for use with\n TLS 1.3.\n\n +------------------------------+-------------+\n | Description | Value |\n +------------------------------+-------------+\n | TLS_AES_128_GCM_SHA256 | {0x13,0x01} |\n | | |\n | TLS_AES_256_GCM_SHA384 | {0x13,0x02} |\n | | |\n | TLS_CHACHA20_POLY1305_SHA256 | {0x13,0x03} |\n | | |\n | TLS_AES_128_CCM_SHA256 | {0x13,0x04} |\n | | |\n | TLS_AES_128_CCM_8_SHA256 | {0x13,0x05} |\n +------------------------------+-------------+\n\n\n\nRescorla Standards Track [Page 133]\n\f\nRFC 8446 TLS August 2018\n\n\n The corresponding AEAD algorithms AEAD_AES_128_GCM, AEAD_AES_256_GCM,\n and AEAD_AES_128_CCM are defined in [RFC5116].\n AEAD_CHACHA20_POLY1305 is defined in [RFC8439]. AEAD_AES_128_CCM_8\n is defined in [RFC6655]. The corresponding hash algorithms are\n defined in [SHS].\n\n Although TLS 1.3 uses the same cipher suite space as previous\n versions of TLS, TLS 1.3 cipher suites are defined differently, only\n specifying the symmetric ciphers, and cannot be used for TLS 1.2.\n Similarly, cipher suites for TLS 1.2 and lower cannot be used with\n TLS 1.3.\n\n New cipher suite values are assigned by IANA as described in\n Section 11.\n\nAppendix C. Implementation Notes\n\n The TLS protocol cannot prevent many common security mistakes. This\n appendix provides several recommendations to assist implementors.\n [TLS13-TRACES] provides test vectors for TLS 1.3 handshakes.\n\nC.1. Random Number Generation and Seeding\n\n TLS requires a cryptographically secure pseudorandom number generator\n (CSPRNG). In most cases, the operating system provides an\n appropriate facility such as /dev/urandom, which should be used\n absent other (e.g., performance) concerns. It is RECOMMENDED to use\n an existing CSPRNG implementation in preference to crafting a new\n one. Many adequate cryptographic libraries are already available\n under favorable license terms. Should those prove unsatisfactory,\n [RFC4086] provides guidance on the generation of random values.\n\n TLS uses random values (1) in public protocol fields such as the\n public Random values in the ClientHello and ServerHello and (2) to\n generate keying material. With a properly functioning CSPRNG, this\n does not present a security problem, as it is not feasible to\n determine the CSPRNG state from its output. However, with a broken\n CSPRNG, it may be possible for an attacker to use the public output\n to determine the CSPRNG internal state and thereby predict the keying\n material, as documented in [CHECKOWAY]. Implementations can provide\n extra security against this form of attack by using separate CSPRNGs\n to generate public and private values.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 134]\n\f\nRFC 8446 TLS August 2018\n\n\nC.2. Certificates and Authentication\n\n Implementations are responsible for verifying the integrity of\n certificates and should generally support certificate revocation\n messages. Absent a specific indication from an application profile,\n certificates should always be verified to ensure proper signing by a\n trusted certificate authority (CA). The selection and addition of\n trust anchors should be done very carefully. Users should be able to\n view information about the certificate and trust anchor.\n Applications SHOULD also enforce minimum and maximum key sizes. For\n example, certification paths containing keys or signatures weaker\n than 2048-bit RSA or 224-bit ECDSA are not appropriate for secure\n applications.\n\nC.3. Implementation Pitfalls\n\n Implementation experience has shown that certain parts of earlier TLS\n specifications are not easy to understand and have been a source of\n interoperability and security problems. Many of these areas have\n been clarified in this document, but this appendix contains a short\n list of the most important things that require special attention from\n implementors.\n\n TLS protocol issues:\n\n - Do you correctly handle handshake messages that are fragmented to\n multiple TLS records (see Section 5.1)? Do you correctly handle\n corner cases like a ClientHello that is split into several small\n fragments? Do you fragment handshake messages that exceed the\n maximum fragment size? In particular, the Certificate and\n CertificateRequest handshake messages can be large enough to\n require fragmentation.\n\n - Do you ignore the TLS record layer version number in all\n unencrypted TLS records (see Appendix D)?\n\n - Have you ensured that all support for SSL, RC4, EXPORT ciphers,\n and MD5 (via the \"signature_algorithms\" extension) is completely\n removed from all possible configurations that support TLS 1.3 or\n later, and that attempts to use these obsolete capabilities fail\n correctly (see Appendix D)?\n\n - Do you handle TLS extensions in ClientHellos correctly, including\n unknown extensions?\n\n\n\n\n\n\n\nRescorla Standards Track [Page 135]\n\f\nRFC 8446 TLS August 2018\n\n\n - When the server has requested a client certificate but no suitable\n certificate is available, do you correctly send an empty\n Certificate message, instead of omitting the whole message (see\n Section 4.4.2)?\n\n - When processing the plaintext fragment produced by AEAD-Decrypt\n and scanning from the end for the ContentType, do you avoid\n scanning past the start of the cleartext in the event that the\n peer has sent a malformed plaintext of all zeros?\n\n - Do you properly ignore unrecognized cipher suites (Section 4.1.2),\n hello extensions (Section 4.2), named groups (Section 4.2.7), key\n shares (Section 4.2.8), supported versions (Section 4.2.1), and\n signature algorithms (Section 4.2.3) in the ClientHello?\n\n - As a server, do you send a HelloRetryRequest to clients which\n support a compatible (EC)DHE group but do not predict it in the\n \"key_share\" extension? As a client, do you correctly handle a\n HelloRetryRequest from the server?\n\n Cryptographic details:\n\n - What countermeasures do you use to prevent timing attacks\n [TIMING]?\n\n - When using Diffie-Hellman key exchange, do you correctly preserve\n leading zero bytes in the negotiated key (see Section 7.4.1)?\n\n - Does your TLS client check that the Diffie-Hellman parameters sent\n by the server are acceptable (see Section 4.2.8.1)?\n\n - Do you use a strong and, most importantly, properly seeded random\n number generator (see Appendix C.1) when generating Diffie-Hellman\n private values, the ECDSA \"k\" parameter, and other security-\n critical values? It is RECOMMENDED that implementations implement\n \"deterministic ECDSA\" as specified in [RFC6979].\n\n - Do you zero-pad Diffie-Hellman public key values and shared\n secrets to the group size (see Section 4.2.8.1 and Section 7.4.1)?\n\n - Do you verify signatures after making them, to protect against\n RSA-CRT key leaks [FW15]?\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 136]\n\f\nRFC 8446 TLS August 2018\n\n\nC.4. Client Tracking Prevention\n\n Clients SHOULD NOT reuse a ticket for multiple connections. Reuse of\n a ticket allows passive observers to correlate different connections.\n Servers that issue tickets SHOULD offer at least as many tickets as\n the number of connections that a client might use; for example, a web\n browser using HTTP/1.1 [RFC7230] might open six connections to a\n server. Servers SHOULD issue new tickets with every connection.\n This ensures that clients are always able to use a new ticket when\n creating a new connection.\n\nC.5. Unauthenticated Operation\n\n Previous versions of TLS offered explicitly unauthenticated cipher\n suites based on anonymous Diffie-Hellman. These modes have been\n deprecated in TLS 1.3. However, it is still possible to negotiate\n parameters that do not provide verifiable server authentication by\n several methods, including:\n\n - Raw public keys [RFC7250].\n\n - Using a public key contained in a certificate but without\n validation of the certificate chain or any of its contents.\n\n Either technique used alone is vulnerable to man-in-the-middle\n attacks and therefore unsafe for general use. However, it is also\n possible to bind such connections to an external authentication\n mechanism via out-of-band validation of the server's public key,\n trust on first use, or a mechanism such as channel bindings (though\n the channel bindings described in [RFC5929] are not defined for\n TLS 1.3). If no such mechanism is used, then the connection has no\n protection against active man-in-the-middle attack; applications\n MUST NOT use TLS in such a way absent explicit configuration or a\n specific application profile.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 137]\n\f\nRFC 8446 TLS August 2018\n\n\nAppendix D. Backward Compatibility\n\n The TLS protocol provides a built-in mechanism for version\n negotiation between endpoints potentially supporting different\n versions of TLS.\n\n TLS 1.x and SSL 3.0 use compatible ClientHello messages. Servers can\n also handle clients trying to use future versions of TLS as long as\n the ClientHello format remains compatible and there is at least one\n protocol version supported by both the client and the server.\n\n Prior versions of TLS used the record layer version number\n (TLSPlaintext.legacy_record_version and\n TLSCiphertext.legacy_record_version) for various purposes. As of\n TLS 1.3, this field is deprecated. The value of\n TLSPlaintext.legacy_record_version MUST be ignored by all\n implementations. The value of TLSCiphertext.legacy_record_version is\n included in the additional data for deprotection but MAY otherwise be\n ignored or MAY be validated to match the fixed constant value.\n Version negotiation is performed using only the handshake versions\n (ClientHello.legacy_version and ServerHello.legacy_version, as well\n as the ClientHello, HelloRetryRequest, and ServerHello\n \"supported_versions\" extensions). In order to maximize\n interoperability with older endpoints, implementations that negotiate\n the use of TLS 1.0-1.2 SHOULD set the record layer version number to\n the negotiated version for the ServerHello and all records\n thereafter.\n\n For maximum compatibility with previously non-standard behavior and\n misconfigured deployments, all implementations SHOULD support\n validation of certification paths based on the expectations in this\n document, even when handling prior TLS versions' handshakes (see\n Section 4.4.2.2).\n\n TLS 1.2 and prior supported an \"Extended Master Secret\" [RFC7627]\n extension which digested large parts of the handshake transcript into\n the master secret. Because TLS 1.3 always hashes in the transcript\n up to the server Finished, implementations which support both TLS 1.3\n and earlier versions SHOULD indicate the use of the Extended Master\n Secret extension in their APIs whenever TLS 1.3 is used.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 138]\n\f\nRFC 8446 TLS August 2018\n\n\nD.1. Negotiating with an Older Server\n\n A TLS 1.3 client who wishes to negotiate with servers that do not\n support TLS 1.3 will send a normal TLS 1.3 ClientHello containing\n 0x0303 (TLS 1.2) in ClientHello.legacy_version but with the correct\n version(s) in the \"supported_versions\" extension. If the server does\n not support TLS 1.3, it will respond with a ServerHello containing an\n older version number. If the client agrees to use this version, the\n negotiation will proceed as appropriate for the negotiated protocol.\n A client using a ticket for resumption SHOULD initiate the connection\n using the version that was previously negotiated.\n\n Note that 0-RTT data is not compatible with older servers and\n SHOULD NOT be sent absent knowledge that the server supports TLS 1.3.\n See Appendix D.3.\n\n If the version chosen by the server is not supported by the client\n (or is not acceptable), the client MUST abort the handshake with a\n \"protocol_version\" alert.\n\n Some legacy server implementations are known to not implement the TLS\n specification properly and might abort connections upon encountering\n TLS extensions or versions which they are not aware of.\n Interoperability with buggy servers is a complex topic beyond the\n scope of this document. Multiple connection attempts may be required\n in order to negotiate a backward-compatible connection; however, this\n practice is vulnerable to downgrade attacks and is NOT RECOMMENDED.\n\nD.2. Negotiating with an Older Client\n\n A TLS server can also receive a ClientHello indicating a version\n number smaller than its highest supported version. If the\n \"supported_versions\" extension is present, the server MUST negotiate\n using that extension as described in Section 4.2.1. If the\n \"supported_versions\" extension is not present, the server MUST\n negotiate the minimum of ClientHello.legacy_version and TLS 1.2. For\n example, if the server supports TLS 1.0, 1.1, and 1.2, and\n legacy_version is TLS 1.0, the server will proceed with a TLS 1.0\n ServerHello. If the \"supported_versions\" extension is absent and the\n server only supports versions greater than\n ClientHello.legacy_version, the server MUST abort the handshake with\n a \"protocol_version\" alert.\n\n Note that earlier versions of TLS did not clearly specify the record\n layer version number value in all cases\n (TLSPlaintext.legacy_record_version). Servers will receive various\n TLS 1.x versions in this field, but its value MUST always be ignored.\n\n\n\n\nRescorla Standards Track [Page 139]\n\f\nRFC 8446 TLS August 2018\n\n\nD.3. 0-RTT Backward Compatibility\n\n 0-RTT data is not compatible with older servers. An older server\n will respond to the ClientHello with an older ServerHello, but it\n will not correctly skip the 0-RTT data and will fail to complete the\n handshake. This can cause issues when a client attempts to use\n 0-RTT, particularly against multi-server deployments. For example, a\n deployment could deploy TLS 1.3 gradually with some servers\n implementing TLS 1.3 and some implementing TLS 1.2, or a TLS 1.3\n deployment could be downgraded to TLS 1.2.\n\n A client that attempts to send 0-RTT data MUST fail a connection if\n it receives a ServerHello with TLS 1.2 or older. It can then retry\n the connection with 0-RTT disabled. To avoid a downgrade attack, the\n client SHOULD NOT disable TLS 1.3, only 0-RTT.\n\n To avoid this error condition, multi-server deployments SHOULD ensure\n a uniform and stable deployment of TLS 1.3 without 0-RTT prior to\n enabling 0-RTT.\n\nD.4. Middlebox Compatibility Mode\n\n Field measurements [Ben17a] [Ben17b] [Res17a] [Res17b] have found\n that a significant number of middleboxes misbehave when a TLS\n client/server pair negotiates TLS 1.3. Implementations can increase\n the chance of making connections through those middleboxes by making\n the TLS 1.3 handshake look more like a TLS 1.2 handshake:\n\n - The client always provides a non-empty session ID in the\n ClientHello, as described in the legacy_session_id section of\n Section 4.1.2.\n\n - If not offering early data, the client sends a dummy\n change_cipher_spec record (see the third paragraph of Section 5)\n immediately before its second flight. This may either be before\n its second ClientHello or before its encrypted handshake flight.\n If offering early data, the record is placed immediately after the\n first ClientHello.\n\n - The server sends a dummy change_cipher_spec record immediately\n after its first handshake message. This may either be after a\n ServerHello or a HelloRetryRequest.\n\n When put together, these changes make the TLS 1.3 handshake resemble\n TLS 1.2 session resumption, which improves the chance of successfully\n connecting through middleboxes. This \"compatibility mode\" is\n partially negotiated: the client can opt to provide a session ID or\n not, and the server has to echo it. Either side can send\n\n\n\nRescorla Standards Track [Page 140]\n\f\nRFC 8446 TLS August 2018\n\n\n change_cipher_spec at any time during the handshake, as they must be\n ignored by the peer, but if the client sends a non-empty session ID,\n the server MUST send the change_cipher_spec as described in this\n appendix.\n\nD.5. Security Restrictions Related to Backward Compatibility\n\n Implementations negotiating the use of older versions of TLS SHOULD\n prefer forward secret and AEAD cipher suites, when available.\n\n The security of RC4 cipher suites is considered insufficient for the\n reasons cited in [RFC7465]. Implementations MUST NOT offer or\n negotiate RC4 cipher suites for any version of TLS for any reason.\n\n Old versions of TLS permitted the use of very low strength ciphers.\n Ciphers with a strength less than 112 bits MUST NOT be offered or\n negotiated for any version of TLS for any reason.\n\n The security of SSL 3.0 [RFC6101] is considered insufficient for the\n reasons enumerated in [RFC7568], and it MUST NOT be negotiated for\n any reason.\n\n The security of SSL 2.0 [SSL2] is considered insufficient for the\n reasons enumerated in [RFC6176], and it MUST NOT be negotiated for\n any reason.\n\n Implementations MUST NOT send an SSL version 2.0 compatible\n CLIENT-HELLO. Implementations MUST NOT negotiate TLS 1.3 or later\n using an SSL version 2.0 compatible CLIENT-HELLO. Implementations\n are NOT RECOMMENDED to accept an SSL version 2.0 compatible\n CLIENT-HELLO in order to negotiate older versions of TLS.\n\n Implementations MUST NOT send a ClientHello.legacy_version or\n ServerHello.legacy_version set to 0x0300 or less. Any endpoint\n receiving a Hello message with ClientHello.legacy_version or\n ServerHello.legacy_version set to 0x0300 MUST abort the handshake\n with a \"protocol_version\" alert.\n\n Implementations MUST NOT send any records with a version less than\n 0x0300. Implementations SHOULD NOT accept any records with a version\n less than 0x0300 (but may inadvertently do so if the record version\n number is ignored completely).\n\n Implementations MUST NOT use the Truncated HMAC extension, defined in\n Section 7 of [RFC6066], as it is not applicable to AEAD algorithms\n and has been shown to be insecure in some scenarios.\n\n\n\n\n\nRescorla Standards Track [Page 141]\n\f\nRFC 8446 TLS August 2018\n\n\nAppendix E. Overview of Security Properties\n\n A complete security analysis of TLS is outside the scope of this\n document. In this appendix, we provide an informal description of\n the desired properties as well as references to more detailed work in\n the research literature which provides more formal definitions.\n\n We cover properties of the handshake separately from those of the\n record layer.\n\nE.1. Handshake\n\n The TLS handshake is an Authenticated Key Exchange (AKE) protocol\n which is intended to provide both one-way authenticated (server-only)\n and mutually authenticated (client and server) functionality. At the\n completion of the handshake, each side outputs its view of the\n following values:\n\n - A set of \"session keys\" (the various secrets derived from the\n master secret) from which can be derived a set of working keys.\n\n - A set of cryptographic parameters (algorithms, etc.).\n\n - The identities of the communicating parties.\n\n We assume the attacker to be an active network attacker, which means\n it has complete control over the network used to communicate between\n the parties [RFC3552]. Even under these conditions, the handshake\n should provide the properties listed below. Note that these\n properties are not necessarily independent, but reflect the protocol\n consumers' needs.\n\n Establishing the same session keys: The handshake needs to output\n the same set of session keys on both sides of the handshake,\n provided that it completes successfully on each endpoint (see\n [CK01], Definition 1, part 1).\n\n Secrecy of the session keys: The shared session keys should be known\n only to the communicating parties and not to the attacker (see\n [CK01], Definition 1, part 2). Note that in a unilaterally\n authenticated connection, the attacker can establish its own\n session keys with the server, but those session keys are distinct\n from those established by the client.\n\n Peer authentication: The client's view of the peer identity should\n reflect the server's identity. If the client is authenticated,\n the server's view of the peer identity should match the client's\n identity.\n\n\n\nRescorla Standards Track [Page 142]\n\f\nRFC 8446 TLS August 2018\n\n\n Uniqueness of the session keys: Any two distinct handshakes should\n produce distinct, unrelated session keys. Individual session keys\n produced by a handshake should also be distinct and independent.\n\n Downgrade protection: The cryptographic parameters should be the\n same on both sides and should be the same as if the peers had been\n communicating in the absence of an attack (see [BBFGKZ16],\n Definitions 8 and 9).\n\n Forward secret with respect to long-term keys: If the long-term\n keying material (in this case the signature keys in certificate-\n based authentication modes or the external/resumption PSK in PSK\n with (EC)DHE modes) is compromised after the handshake is\n complete, this does not compromise the security of the session key\n (see [DOW92]), as long as the session key itself has been erased.\n The forward secrecy property is not satisfied when PSK is used in\n the \"psk_ke\" PskKeyExchangeMode.\n\n Key Compromise Impersonation (KCI) resistance: In a mutually\n authenticated connection with certificates, compromising the\n long-term secret of one actor should not break that actor's\n authentication of their peer in the given connection (see\n [HGFS15]). For example, if a client's signature key is\n compromised, it should not be possible to impersonate arbitrary\n servers to that client in subsequent handshakes.\n\n Protection of endpoint identities: The server's identity\n (certificate) should be protected against passive attackers. The\n client's identity should be protected against both passive and\n active attackers.\n\n Informally, the signature-based modes of TLS 1.3 provide for the\n establishment of a unique, secret, shared key established by an\n (EC)DHE key exchange and authenticated by the server's signature over\n the handshake transcript, as well as tied to the server's identity by\n a MAC. If the client is authenticated by a certificate, it also\n signs over the handshake transcript and provides a MAC tied to both\n identities. [SIGMA] describes the design and analysis of this type\n of key exchange protocol. If fresh (EC)DHE keys are used for each\n connection, then the output keys are forward secret.\n\n The external PSK and resumption PSK bootstrap from a long-term shared\n secret into a unique per-connection set of short-term session keys.\n This secret may have been established in a previous handshake. If\n PSK with (EC)DHE key establishment is used, these session keys will\n also be forward secret. The resumption PSK has been designed so that\n the resumption master secret computed by connection N and needed to\n form connection N+1 is separate from the traffic keys used by\n\n\n\nRescorla Standards Track [Page 143]\n\f\nRFC 8446 TLS August 2018\n\n\n connection N, thus providing forward secrecy between the connections.\n In addition, if multiple tickets are established on the same\n connection, they are associated with different keys, so compromise of\n the PSK associated with one ticket does not lead to the compromise of\n connections established with PSKs associated with other tickets.\n This property is most interesting if tickets are stored in a database\n (and so can be deleted) rather than if they are self-encrypted.\n\n The PSK binder value forms a binding between a PSK and the current\n handshake, as well as between the session where the PSK was\n established and the current session. This binding transitively\n includes the original handshake transcript, because that transcript\n is digested into the values which produce the resumption master\n secret. This requires that both the KDF used to produce the\n resumption master secret and the MAC used to compute the binder be\n collision resistant. See Appendix E.1.1 for more on this. Note: The\n binder does not cover the binder values from other PSKs, though they\n are included in the Finished MAC.\n\n TLS does not currently permit the server to send a\n certificate_request message in non-certificate-based handshakes\n (e.g., PSK). If this restriction were to be relaxed in future, the\n client's signature would not cover the server's certificate directly.\n However, if the PSK was established through a NewSessionTicket, the\n client's signature would transitively cover the server's certificate\n through the PSK binder. [PSK-FINISHED] describes a concrete attack\n on constructions that do not bind to the server's certificate (see\n also [Kraw16]). It is unsafe to use certificate-based client\n authentication when the client might potentially share the same\n PSK/key-id pair with two different endpoints. Implementations\n MUST NOT combine external PSKs with certificate-based authentication\n of either the client or the server unless negotiated by some\n extension.\n\n If an exporter is used, then it produces values which are unique and\n secret (because they are generated from a unique session key).\n Exporters computed with different labels and contexts are\n computationally independent, so it is not feasible to compute one\n from another or the session secret from the exported value.\n Note: Exporters can produce arbitrary-length values; if exporters are\n to be used as channel bindings, the exported value MUST be large\n enough to provide collision resistance. The exporters provided in\n TLS 1.3 are derived from the same Handshake Contexts as the early\n traffic keys and the application traffic keys, respectively, and thus\n have similar security properties. Note that they do not include the\n client's certificate; future applications which wish to bind to the\n client's certificate may need to define a new exporter that includes\n the full handshake transcript.\n\n\n\nRescorla Standards Track [Page 144]\n\f\nRFC 8446 TLS August 2018\n\n\n For all handshake modes, the Finished MAC (and, where present, the\n signature) prevents downgrade attacks. In addition, the use of\n certain bytes in the random nonces as described in Section 4.1.3\n allows the detection of downgrade to previous TLS versions. See\n [BBFGKZ16] for more details on TLS 1.3 and downgrade.\n\n As soon as the client and the server have exchanged enough\n information to establish shared keys, the remainder of the handshake\n is encrypted, thus providing protection against passive attackers,\n even if the computed shared key is not authenticated. Because the\n server authenticates before the client, the client can ensure that if\n it authenticates to the server, it only reveals its identity to an\n authenticated server. Note that implementations must use the\n provided record-padding mechanism during the handshake to avoid\n leaking information about the identities due to length. The client's\n proposed PSK identities are not encrypted, nor is the one that the\n server selects.\n\nE.1.1. Key Derivation and HKDF\n\n Key derivation in TLS 1.3 uses HKDF as defined in [RFC5869] and its\n two components, HKDF-Extract and HKDF-Expand. The full rationale for\n the HKDF construction can be found in [Kraw10] and the rationale for\n the way it is used in TLS 1.3 in [KW16]. Throughout this document,\n each application of HKDF-Extract is followed by one or more\n invocations of HKDF-Expand. This ordering should always be followed\n (including in future revisions of this document); in particular, one\n SHOULD NOT use an output of HKDF-Extract as an input to another\n application of HKDF-Extract without an HKDF-Expand in between.\n Multiple applications of HKDF-Expand to some of the same inputs are\n allowed as long as these are differentiated via the key and/or the\n labels.\n\n Note that HKDF-Expand implements a pseudorandom function (PRF) with\n both inputs and outputs of variable length. In some of the uses of\n HKDF in this document (e.g., for generating exporters and the\n resumption_master_secret), it is necessary that the application of\n HKDF-Expand be collision resistant; namely, it should be infeasible\n to find two different inputs to HKDF-Expand that output the same\n value. This requires the underlying hash function to be collision\n resistant and the output length from HKDF-Expand to be of size at\n least 256 bits (or as much as needed for the hash function to prevent\n finding collisions).\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 145]\n\f\nRFC 8446 TLS August 2018\n\n\nE.1.2. Client Authentication\n\n A client that has sent authentication data to a server, either during\n the handshake or in post-handshake authentication, cannot be sure\n whether the server afterwards considers the client to be\n authenticated or not. If the client needs to determine if the server\n considers the connection to be unilaterally or mutually\n authenticated, this has to be provisioned by the application layer.\n See [CHHSV17] for details. In addition, the analysis of\n post-handshake authentication from [Kraw16] shows that the client\n identified by the certificate sent in the post-handshake phase\n possesses the traffic key. This party is therefore the client that\n participated in the original handshake or one to whom the original\n client delegated the traffic key (assuming that the traffic key has\n not been compromised).\n\nE.1.3. 0-RTT\n\n The 0-RTT mode of operation generally provides security properties\n similar to those of 1-RTT data, with the two exceptions that the\n 0-RTT encryption keys do not provide full forward secrecy and that\n the server is not able to guarantee uniqueness of the handshake\n (non-replayability) without keeping potentially undue amounts of\n state. See Section 8 for mechanisms to limit the exposure to replay.\n\nE.1.4. Exporter Independence\n\n The exporter_master_secret and early_exporter_master_secret are\n derived to be independent of the traffic keys and therefore do not\n represent a threat to the security of traffic encrypted with those\n keys. However, because these secrets can be used to compute any\n exporter value, they SHOULD be erased as soon as possible. If the\n total set of exporter labels is known, then implementations SHOULD\n pre-compute the inner Derive-Secret stage of the exporter computation\n for all those labels, then erase the [early_]exporter_master_secret,\n followed by each inner value as soon as it is known that it will not\n be needed again.\n\nE.1.5. Post-Compromise Security\n\n TLS does not provide security for handshakes which take place after\n the peer's long-term secret (signature key or external PSK) is\n compromised. It therefore does not provide post-compromise security\n [CCG16], sometimes also referred to as backward or future secrecy.\n This is in contrast to KCI resistance, which describes the security\n guarantees that a party has after its own long-term secret has been\n compromised.\n\n\n\n\nRescorla Standards Track [Page 146]\n\f\nRFC 8446 TLS August 2018\n\n\nE.1.6. External References\n\n The reader should refer to the following references for analysis of\n the TLS handshake: [DFGS15], [CHSV16], [DFGS16], [KW16], [Kraw16],\n [FGSW16], [LXZFH16], [FG17], and [BBK17].\n\nE.2. Record Layer\n\n The record layer depends on the handshake producing strong traffic\n secrets which can be used to derive bidirectional encryption keys and\n nonces. Assuming that is true, and the keys are used for no more\n data than indicated in Section 5.5, then the record layer should\n provide the following guarantees:\n\n Confidentiality: An attacker should not be able to determine the\n plaintext contents of a given record.\n\n Integrity: An attacker should not be able to craft a new record\n which is different from an existing record which will be accepted\n by the receiver.\n\n Order protection/non-replayability: An attacker should not be able\n to cause the receiver to accept a record which it has already\n accepted or cause the receiver to accept record N+1 without having\n first processed record N.\n\n Length concealment: Given a record with a given external length, the\n attacker should not be able to determine the amount of the record\n that is content versus padding.\n\n Forward secrecy after key change: If the traffic key update\n mechanism described in Section 4.6.3 has been used and the\n previous generation key is deleted, an attacker who compromises\n the endpoint should not be able to decrypt traffic encrypted with\n the old key.\n\n Informally, TLS 1.3 provides these properties by AEAD-protecting the\n plaintext with a strong key. AEAD encryption [RFC5116] provides\n confidentiality and integrity for the data. Non-replayability is\n provided by using a separate nonce for each record, with the nonce\n being derived from the record sequence number (Section 5.3), with the\n sequence number being maintained independently at both sides; thus,\n records which are delivered out of order result in AEAD deprotection\n failures. In order to prevent mass cryptanalysis when the same\n plaintext is repeatedly encrypted by different users under the same\n key (as is commonly the case for HTTP), the nonce is formed by mixing\n\n\n\n\n\nRescorla Standards Track [Page 147]\n\f\nRFC 8446 TLS August 2018\n\n\n the sequence number with a secret per-connection initialization\n vector derived along with the traffic keys. See [BT16] for analysis\n of this construction.\n\n The rekeying technique in TLS 1.3 (see Section 7.2) follows the\n construction of the serial generator as discussed in [REKEY], which\n shows that rekeying can allow keys to be used for a larger number of\n encryptions than without rekeying. This relies on the security of\n the HKDF-Expand-Label function as a pseudorandom function (PRF). In\n addition, as long as this function is truly one way, it is not\n possible to compute traffic keys from prior to a key change (forward\n secrecy).\n\n TLS does not provide security for data which is communicated on a\n connection after a traffic secret of that connection is compromised.\n That is, TLS does not provide post-compromise security/future\n secrecy/backward secrecy with respect to the traffic secret. Indeed,\n an attacker who learns a traffic secret can compute all future\n traffic secrets on that connection. Systems which want such\n guarantees need to do a fresh handshake and establish a new\n connection with an (EC)DHE exchange.\n\nE.2.1. External References\n\n The reader should refer to the following references for analysis of\n the TLS record layer: [BMMRT15], [BT16], [BDFKPPRSZZ16], [BBK17], and\n [PS18].\n\nE.3. Traffic Analysis\n\n TLS is susceptible to a variety of traffic analysis attacks based on\n observing the length and timing of encrypted packets [CLINIC]\n [HCJC16]. This is particularly easy when there is a small set of\n possible messages to be distinguished, such as for a video server\n hosting a fixed corpus of content, but still provides usable\n information even in more complicated scenarios.\n\n TLS does not provide any specific defenses against this form of\n attack but does include a padding mechanism for use by applications:\n The plaintext protected by the AEAD function consists of content plus\n variable-length padding, which allows the application to produce\n arbitrary-length encrypted records as well as padding-only cover\n traffic to conceal the difference between periods of transmission and\n periods of silence. Because the padding is encrypted alongside the\n actual content, an attacker cannot directly determine the length of\n the padding but may be able to measure it indirectly by the use of\n timing channels exposed during record processing (i.e., seeing how\n long it takes to process a record or trickling in records to see\n\n\n\nRescorla Standards Track [Page 148]\n\f\nRFC 8446 TLS August 2018\n\n\n which ones elicit a response from the server). In general, it is not\n known how to remove all of these channels because even a\n constant-time padding removal function will likely feed the content\n into data-dependent functions. At minimum, a fully constant-time\n server or client would require close cooperation with the\n application-layer protocol implementation, including making that\n higher-level protocol constant time.\n\n Note: Robust traffic analysis defenses will likely lead to inferior\n performance due to delays in transmitting packets and increased\n traffic volume.\n\nE.4. Side-Channel Attacks\n\n In general, TLS does not have specific defenses against side-channel\n attacks (i.e., those which attack the communications via secondary\n channels such as timing), leaving those to the implementation of the\n relevant cryptographic primitives. However, certain features of TLS\n are designed to make it easier to write side-channel resistant code:\n\n - Unlike previous versions of TLS which used a composite MAC-then-\n encrypt structure, TLS 1.3 only uses AEAD algorithms, allowing\n implementations to use self-contained constant-time\n implementations of those primitives.\n\n - TLS uses a uniform \"bad_record_mac\" alert for all decryption\n errors, which is intended to prevent an attacker from gaining\n piecewise insight into portions of the message. Additional\n resistance is provided by terminating the connection on such\n errors; a new connection will have different cryptographic\n material, preventing attacks against the cryptographic primitives\n that require multiple trials.\n\n Information leakage through side channels can occur at layers above\n TLS, in application protocols and the applications that use them.\n Resistance to side-channel attacks depends on applications and\n application protocols separately ensuring that confidential\n information is not inadvertently leaked.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 149]\n\f\nRFC 8446 TLS August 2018\n\n\nE.5. Replay Attacks on 0-RTT\n\n Replayable 0-RTT data presents a number of security threats to TLS-\n using applications, unless those applications are specifically\n engineered to be safe under replay (minimally, this means idempotent,\n but in many cases may also require other stronger conditions, such as\n constant-time response). Potential attacks include:\n\n - Duplication of actions which cause side effects (e.g., purchasing\n an item or transferring money) to be duplicated, thus harming the\n site or the user.\n\n - Attackers can store and replay 0-RTT messages in order to reorder\n them with respect to other messages (e.g., moving a delete to\n after a create).\n\n - Exploiting cache timing behavior to discover the content of 0-RTT\n messages by replaying a 0-RTT message to a different cache node\n and then using a separate connection to measure request latency,\n to see if the two requests address the same resource.\n\n If data can be replayed a large number of times, additional attacks\n become possible, such as making repeated measurements of the speed of\n cryptographic operations. In addition, they may be able to overload\n rate-limiting systems. For a further description of these attacks,\n see [Mac17].\n\n Ultimately, servers have the responsibility to protect themselves\n against attacks employing 0-RTT data replication. The mechanisms\n described in Section 8 are intended to prevent replay at the TLS\n layer but do not provide complete protection against receiving\n multiple copies of client data. TLS 1.3 falls back to the 1-RTT\n handshake when the server does not have any information about the\n client, e.g., because it is in a different cluster which does not\n share state or because the ticket has been deleted as described in\n Section 8.1. If the application-layer protocol retransmits data in\n this setting, then it is possible for an attacker to induce message\n duplication by sending the ClientHello to both the original cluster\n (which processes the data immediately) and another cluster which will\n fall back to 1-RTT and process the data upon application-layer\n replay. The scale of this attack is limited by the client's\n willingness to retry transactions and therefore only allows a limited\n amount of duplication, with each copy appearing as a new connection\n at the server.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 150]\n\f\nRFC 8446 TLS August 2018\n\n\n If implemented correctly, the mechanisms described in Sections 8.1\n and 8.2 prevent a replayed ClientHello and its associated 0-RTT data\n from being accepted multiple times by any cluster with consistent\n state; for servers which limit the use of 0-RTT to one cluster for a\n single ticket, then a given ClientHello and its associated 0-RTT data\n will only be accepted once. However, if state is not completely\n consistent, then an attacker might be able to have multiple copies of\n the data be accepted during the replication window. Because clients\n do not know the exact details of server behavior, they MUST NOT send\n messages in early data which are not safe to have replayed and which\n they would not be willing to retry across multiple 1-RTT connections.\n\n Application protocols MUST NOT use 0-RTT data without a profile that\n defines its use. That profile needs to identify which messages or\n interactions are safe to use with 0-RTT and how to handle the\n situation when the server rejects 0-RTT and falls back to 1-RTT.\n\n In addition, to avoid accidental misuse, TLS implementations MUST NOT\n enable 0-RTT (either sending or accepting) unless specifically\n requested by the application and MUST NOT automatically resend 0-RTT\n data if it is rejected by the server unless instructed by the\n application. Server-side applications may wish to implement special\n processing for 0-RTT data for some kinds of application traffic\n (e.g., abort the connection, request that data be resent at the\n application layer, or delay processing until the handshake\n completes). In order to allow applications to implement this kind of\n processing, TLS implementations MUST provide a way for the\n application to determine if the handshake has completed.\n\nE.5.1. Replay and Exporters\n\n Replays of the ClientHello produce the same early exporter, thus\n requiring additional care by applications which use these exporters.\n In particular, if these exporters are used as an authentication\n channel binding (e.g., by signing the output of the exporter), an\n attacker who compromises the PSK can transplant authenticators\n between connections without compromising the authentication key.\n\n In addition, the early exporter SHOULD NOT be used to generate\n server-to-client encryption keys because that would entail the reuse\n of those keys. This parallels the use of the early application\n traffic keys only in the client-to-server direction.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 151]\n\f\nRFC 8446 TLS August 2018\n\n\nE.6. PSK Identity Exposure\n\n Because implementations respond to an invalid PSK binder by aborting\n the handshake, it may be possible for an attacker to verify whether a\n given PSK identity is valid. Specifically, if a server accepts both\n external-PSK handshakes and certificate-based handshakes, a valid PSK\n identity will result in a failed handshake, whereas an invalid\n identity will just be skipped and result in a successful certificate\n handshake. Servers which solely support PSK handshakes may be able\n to resist this form of attack by treating the cases where there is no\n valid PSK identity and where there is an identity but it has an\n invalid binder identically.\n\nE.7. Sharing PSKs\n\n TLS 1.3 takes a conservative approach to PSKs by binding them to a\n specific KDF. By contrast, TLS 1.2 allows PSKs to be used with any\n hash function and the TLS 1.2 PRF. Thus, any PSK which is used with\n both TLS 1.2 and TLS 1.3 must be used with only one hash in TLS 1.3,\n which is less than optimal if users want to provision a single PSK.\n The constructions in TLS 1.2 and TLS 1.3 are different, although they\n are both based on HMAC. While there is no known way in which the\n same PSK might produce related output in both versions, only limited\n analysis has been done. Implementations can ensure safety from\n cross-protocol related output by not reusing PSKs between TLS 1.3 and\n TLS 1.2.\n\nE.8. Attacks on Static RSA\n\n Although TLS 1.3 does not use RSA key transport and so is not\n directly susceptible to Bleichenbacher-type attacks [Blei98], if TLS\n 1.3 servers also support static RSA in the context of previous\n versions of TLS, then it may be possible to impersonate the server\n for TLS 1.3 connections [JSS15]. TLS 1.3 implementations can prevent\n this attack by disabling support for static RSA across all versions\n of TLS. In principle, implementations might also be able to separate\n certificates with different keyUsage bits for static RSA decryption\n and RSA signature, but this technique relies on clients refusing to\n accept signatures using keys in certificates that do not have the\n digitalSignature bit set, and many clients do not enforce this\n restriction.\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 152]\n\f\nRFC 8446 TLS August 2018\n\n\nContributors\n\n Martin Abadi\n University of California, Santa Cruz\n [redacted-email]\n\n Christopher Allen\n (co-editor of TLS 1.0)\n Alacrity Ventures\n [redacted-email]\n\n Richard Barnes\n Cisco\n [redacted-email]\n\n Steven M. Bellovin\n Columbia University\n [redacted-email]\n\n David Benjamin\n Google\n [redacted-email]\n\n Benjamin Beurdouche\n INRIA & Microsoft Research\n [redacted-email]\n\n Karthikeyan Bhargavan\n (editor of [RFC7627])\n INRIA\n [redacted-email]\n\n Simon Blake-Wilson\n (co-author of [RFC4492])\n BCI\n [redacted-email]\n\n Nelson Bolyard\n (co-author of [RFC4492])\n Sun Microsystems, Inc.\n [redacted-email]\n\n Ran Canetti\n IBM\n [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 153]\n\f\nRFC 8446 TLS August 2018\n\n\n Matt Caswell\n OpenSSL\n [redacted-email]\n\n Stephen Checkoway\n University of Illinois at Chicago\n [redacted-email]\n\n Pete Chown\n Skygate Technology Ltd\n [redacted-email]\n\n Katriel Cohn-Gordon\n University of Oxford\n [redacted-email]\n\n Cas Cremers\n University of Oxford\n [redacted-email]\n\n Antoine Delignat-Lavaud\n (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n Tim Dierks\n (co-author of TLS 1.0, co-editor of TLS 1.1 and 1.2)\n Independent\n [redacted-email]\n\n Roelof DuToit\n Symantec Corporation\n [redacted-email]\n\n Taher Elgamal\n Securify\n [redacted-email]\n\n Pasi Eronen\n Nokia\n [redacted-email]\n\n Cedric Fournet\n Microsoft\n [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 154]\n\f\nRFC 8446 TLS August 2018\n\n\n Anil Gangolli\n [redacted-email]\n\n David M. Garrett\n [redacted-email]\n\n Illya Gerasymchuk\n Independent\n [redacted-email]\n\n Alessandro Ghedini\n Cloudflare Inc.\n [redacted-email]\n\n Daniel Kahn Gillmor\n ACLU\n [redacted-email]\n\n Matthew Green\n Johns Hopkins University\n [redacted-email]\n\n Jens Guballa\n ETAS\n [redacted-email]\n\n Felix Guenther\n TU Darmstadt\n [redacted-email]\n\n Vipul Gupta\n (co-author of [RFC4492])\n Sun Microsystems Laboratories\n [redacted-email]\n\n Chris Hawk\n (co-author of [RFC4492])\n Corriente Networks LLC\n [redacted-email]\n\n Kipp Hickman\n\n Alfred Hoenes\n\n David Hopwood\n Independent Consultant\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 155]\n\f\nRFC 8446 TLS August 2018\n\n\n Marko Horvat\n MPI-SWS\n [redacted-email]\n\n Jonathan Hoyland\n Royal Holloway, University of London\n [redacted-email]\n\n Subodh Iyengar\n Facebook\n [redacted-email]\n\n Benjamin Kaduk\n Akamai Technologies\n [redacted-email]\n\n Hubert Kario\n Red Hat Inc.\n [redacted-email]\n\n Phil Karlton\n (co-author of SSL 3.0)\n\n Leon Klingele\n Independent\n [redacted-email]\n\n Paul Kocher\n (co-author of SSL 3.0)\n Cryptography Research\n [redacted-email]\n\n Hugo Krawczyk\n IBM\n [redacted-email]\n\n Adam Langley\n (co-author of [RFC7627])\n Google\n [redacted-email]\n\n Olivier Levillain\n ANSSI\n [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 156]\n\f\nRFC 8446 TLS August 2018\n\n\n Xiaoyin Liu\n University of North Carolina at Chapel Hill\n [redacted-email]\n\n Ilari Liusvaara\n Independent\n [redacted-email]\n\n Atul Luykx\n K.U. Leuven\n [redacted-email]\n\n Colm MacCarthaigh\n Amazon Web Services\n [redacted-email]\n\n Carl Mehner\n USAA\n [redacted-email]\n\n Jan Mikkelsen\n Transactionware\n [redacted-email]\n\n Bodo Moeller\n (co-author of [RFC4492])\n Google\n [redacted-email]\n\n Kyle Nekritz\n Facebook\n [redacted-email]\n\n Erik Nygren\n Akamai Technologies\n [redacted-email]\n\n Magnus Nystrom\n Microsoft\n [redacted-email]\n\n Kazuho Oku\n DeNA Co., Ltd.\n [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 157]\n\f\nRFC 8446 TLS August 2018\n\n\n Kenny Paterson\n Royal Holloway, University of London\n [redacted-email]\n\n Christopher Patton\n University of Florida\n [redacted-email]\n\n Alfredo Pironti\n (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n Andrei Popov\n Microsoft\n [redacted-email]\n\n Marsh Ray\n (co-author of [RFC7627])\n Microsoft\n [redacted-email]\n\n Robert Relyea\n Netscape Communications\n [redacted-email]\n\n Kyle Rose\n Akamai Technologies\n [redacted-email]\n\n Jim Roskind\n Amazon\n [redacted-email]\n\n Michael Sabin\n\n Joe Salowey\n Tableau Software\n [redacted-email]\n\n Rich Salz\n Akamai\n [redacted-email]\n\n David Schinazi\n Apple Inc.\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 158]\n\f\nRFC 8446 TLS August 2018\n\n\n Sam Scott\n Royal Holloway, University of London\n [redacted-email]\n\n Thomas Shrimpton\n University of Florida\n [redacted-email]\n\n Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n Brian Smith\n Independent\n [redacted-email]\n\n Brian Sniffen\n Akamai Technologies\n [redacted-email]\n\n Nick Sullivan\n Cloudflare Inc.\n [redacted-email]\n\n Bjoern Tackmann\n University of California, San Diego\n [redacted-email]\n\n Tim Taubert\n Mozilla\n [redacted-email]\n\n Martin Thomson\n Mozilla\n [redacted-email]\n\n Hannes Tschofenig\n Arm Limited\n [redacted-email]\n\n Sean Turner\n sn3rd\n [redacted-email]\n\n Steven Valdez\n Google\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 159]\n\f\nRFC 8446 TLS August 2018\n\n\n Filippo Valsorda\n Cloudflare Inc.\n [redacted-email]\n\n Thyla van der Merwe\n Royal Holloway, University of London\n [redacted-email]\n\n Victor Vasiliev\n Google\n [redacted-email]\n\n Hoeteck Wee\n Ecole Normale Superieure, Paris\n [redacted-email]\n\n Tom Weinstein\n\n David Wong\n NCC Group\n [redacted-email]\n\n Christopher A. Wood\n Apple Inc.\n [redacted-email]\n\n Tim Wright\n Vodafone\n [redacted-email]\n\n Peter Wu\n Independent\n [redacted-email]\n\n Kazu Yamamoto\n Internet Initiative Japan Inc.\n [redacted-email]\n\nAuthor's Address\n\n Eric Rescorla\n Mozilla\n\n Email: [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 160]", "text_sha256": "d596886622bceee8500b3f3c9d869d947b4e70a2f05ea65fdd75666b46debd81"}], "relations": [{"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5077 H. Zhou", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5077", "source_sha256": "e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef"}, {"char_end": 3675, "char_start": 3419, "evidence_quote": "{\"id\": 1269006, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269006/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}", "fact_ids": ["informative_reference:5077"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "informative_reference", "relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5077"}, {"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5246 Independent", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5246", "source_sha256": "2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb"}, {"char_end": 3417, "char_start": 3161, "evidence_quote": "{\"id\": 1269005, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269005/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}", "fact_ids": ["informative_reference:5246"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "informative_reference", "relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5246"}, {"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5705 RTFM, Inc.", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5705", "source_sha256": "29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a"}, {"char_end": 5223, "char_start": 4967, "evidence_quote": "{\"id\": 1269016, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269016/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}", "fact_ids": ["normative_reference:5705"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "normative_reference", "relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5705"}, {"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6066 Huawei", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:6066", "source_sha256": "6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0"}, {"char_end": 12717, "char_start": 12466, "evidence_quote": "{\"id\": 1591472, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591472/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}", "fact_ids": ["normative_reference:6066"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "normative_reference", "relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:6066"}, {"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6961 June 2013", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:6961", "source_sha256": "c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db"}, {"char_end": 10425, "char_start": 10169, "evidence_quote": "{\"id\": 1269043, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269043/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}", "fact_ids": ["normative_reference:6961"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "normative_reference", "relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:6961"}, {"evidence": [{"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_ids": ["obsoletes:5077"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5077 H. Zhou", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5077", "source_sha256": "e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef"}], "kind": "obsoletes", "relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5077"}, {"evidence": [{"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_ids": ["obsoletes:5246"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5246 Independent", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5246", "source_sha256": "2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb"}], "kind": "obsoletes", "relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5246"}, {"evidence": [{"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_ids": ["obsoletes:6961"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6961 June 2013", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:6961", "source_sha256": "c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db"}], "kind": "obsoletes", "relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:6961"}, {"evidence": [{"char_end": 454, "char_start": 431, "evidence_quote": "draft-ietf-tls-tls13-28", "fact_ids": ["draft_revision"], "record_id": "ietf:draft:draft-ietf-tls-tls13-28", "source_sha256": "9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 13945, "char_start": 13673, "evidence_quote": "{\"id\": 1295893, \"originaltargetaliasname\": \"rfc8446\", \"relationship\": \"/api/v1/name/docrelationshipname/became_rfc/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1295893/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8446/\"}", "fact_ids": ["became_rfc"], "record_id": "ietf:datatracker-relation:draft-ietf-tls-tls13:rfc8446", "source_sha256": "3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131"}], "kind": "published_as", "relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "target_record_id": "ietf:rfc:8446"}, {"evidence": [{"char_end": 238, "char_start": 219, "evidence_quote": "Updates: 5705, 6066", "fact_ids": ["updates:5705"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5705 RTFM, Inc.", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5705", "source_sha256": "29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a"}], "kind": "updates", "relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5705"}, {"evidence": [{"char_end": 238, "char_start": 219, "evidence_quote": "Updates: 5705, 6066", "fact_ids": ["updates:6066"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6066 Huawei", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:6066", "source_sha256": "6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0"}], "kind": "updates", "relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:6066"}], "schema_version": "longworld.ietf-workflow-manifest.v1", "semantic_facts_train_ready": false, "source_status": "public_api_export", "supporting_records": [{"draft_name": "draft-ietf-tls-tls13", "facts": [{"char_end": 673, "char_start": 643, "evidence_quote": "\"name\": \"draft-ietf-tls-tls13\"", "fact_id": "draft_name", "field": "draft_name", "raw_char_end": 673, "raw_char_start": 643, "value": "draft-ietf-tls-tls13"}], "kind": "datatracker_document", "occurred_at": "2020-03-07T08:26:41Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 2, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:a632cb66584c8f5d364d909b76eed121c1a9d00e341908463e019efad0faee0f", "record_id": "ietf:datatracker:draft-ietf-tls-tls13", "retrieval_url": "https://datatracker.ietf.org/api/v1/doc/document/draft-ietf-tls-tls13/", "revision": "28", "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "a632cb66584c8f5d364d909b76eed121c1a9d00e341908463e019efad0faee0f", "source_url": "https://datatracker.ietf.org/api/v1/doc/document/draft-ietf-tls-tls13/", "temporal_semantics": "source_event_time", "text": "{\"abstract\": \"This document specifies version 1.3 of the Transport Layer Security (TLS) protocol. TLS allows client/server applications to communicate over the Internet in a way that is designed to prevent eavesdropping, tampering, and message forgery.\\n\\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077, 5246, and 6961. This document also specifies new requirements for TLS 1.2 implementations.\", \"ad\": \"/api/v1/person/person/104851/\", \"expires\": \"2018-09-21T21:26:15Z\", \"external_url\": \"\", \"group\": \"/api/v1/group/group/1326/\", \"id\": 55251, \"intended_std_level\": \"/api/v1/name/intendedstdlevelname/ps/\", \"keywords\": \"[]\", \"name\": \"draft-ietf-tls-tls13\", \"note\": \"\", \"notify\": \"Sean Turner <[redacted-email]>\", \"pages\": 160, \"resource_uri\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"rev\": \"28\", \"rfc\": null, \"rfc_number\": null, \"shepherd\": \"/api/v1/person/email/[redacted-email]/\", \"states\": [\"/api/v1/doc/state/3/\", \"/api/v1/doc/state/108/\", \"/api/v1/doc/state/112/\", \"/api/v1/doc/state/7/\", \"/api/v1/doc/state/44/\"], \"std_level\": \"/api/v1/name/stdlevelname/ps/\", \"stream\": \"/api/v1/name/streamname/ietf/\", \"submissions\": [\"/api/v1/submit/submission/59397/\", \"/api/v1/submit/submission/59398/\", \"/api/v1/submit/submission/61729/\", \"/api/v1/submit/submission/64520/\", \"/api/v1/submit/submission/65644/\", \"/api/v1/submit/submission/67695/\", \"/api/v1/submit/submission/70123/\", \"/api/v1/submit/submission/70989/\", \"/api/v1/submit/submission/71776/\", \"/api/v1/submit/submission/72481/\", \"/api/v1/submit/submission/73392/\", \"/api/v1/submit/submission/74860/\", \"/api/v1/submit/submission/77260/\", \"/api/v1/submit/submission/78266/\", \"/api/v1/submit/submission/79977/\", \"/api/v1/submit/submission/80556/\", \"/api/v1/submit/submission/81099/\", \"/api/v1/submit/submission/81687/\", \"/api/v1/submit/submission/81857/\", \"/api/v1/submit/submission/84956/\", \"/api/v1/submit/submission/86252/\", \"/api/v1/submit/submission/92891/\", \"/api/v1/submit/submission/88268/\", \"/api/v1/submit/submission/91422/\", \"/api/v1/submit/submission/92066/\", \"/api/v1/submit/submission/93465/\", \"/api/v1/submit/submission/93544/\", \"/api/v1/submit/submission/94182/\", \"/api/v1/submit/submission/94319/\"], \"tags\": [], \"time\": \"2020-03-07T08:26:41Z\", \"title\": \"The Transport Layer Security (TLS) Protocol Version 1.3\", \"type\": \"/api/v1/name/doctypename/draft/\", \"uploaded_filename\": \"\", \"words\": 35738}", "text_sha256": "e51b82d004e563f85becf3bc398f5dea916940ec55b92cf7b728f78a625a293b"}, {"facts": [{"char_end": 708, "char_start": 691, "evidence_quote": "\"name\": \"rfc8446\"", "fact_id": "rfc_name", "field": "rfc_name", "raw_char_end": 708, "raw_char_start": 691, "value": "rfc8446"}], "kind": "datatracker_document", "occurred_at": "2026-05-20T15:34:35Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 0, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:89b01010d1ff76dc260539d5a453dde0b116d857db43c9e09d4a3dc062c9ce31", "record_id": "ietf:datatracker:rfc8446", "retrieval_url": "https://datatracker.ietf.org/api/v1/doc/document/rfc8446/", "rfc_number": 8446, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "89b01010d1ff76dc260539d5a453dde0b116d857db43c9e09d4a3dc062c9ce31", "source_url": "https://datatracker.ietf.org/api/v1/doc/document/rfc8446/", "temporal_semantics": "source_event_time", "text": "{\"abstract\": \"This document specifies version 1.3 of the Transport Layer Security (TLS) protocol. TLS allows client/server applications to communicate over the Internet in a way that is designed to prevent eavesdropping, tampering, and message forgery.\\n\\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077, 5246, and 6961. This document also specifies new requirements for TLS 1.2 implementations.\", \"ad\": \"/api/v1/person/person/104851/\", \"expires\": null, \"external_url\": \"\", \"group\": \"/api/v1/group/group/1326/\", \"id\": 130263, \"intended_std_level\": null, \"keywords\": \"['international data algorithm', 'symmetric', 'transport protocol layer', 'authentication', 'privacy']\", \"name\": \"rfc8446\", \"note\": \"\", \"notify\": \"\", \"pages\": 160, \"resource_uri\": \"/api/v1/doc/document/rfc8446/\", \"rev\": \"\", \"rfc\": \"8446\", \"rfc_number\": 8446, \"shepherd\": null, \"states\": [\"/api/v1/doc/state/177/\"], \"std_level\": \"/api/v1/name/stdlevelname/ps/\", \"stream\": \"/api/v1/name/streamname/ietf/\", \"submissions\": [], \"tags\": [\"/api/v1/name/doctagname/errata/\", \"/api/v1/name/doctagname/verified-errata/\"], \"time\": \"2026-05-20T15:34:35Z\", \"title\": \"The Transport Layer Security (TLS) Protocol Version 1.3\", \"type\": \"/api/v1/name/doctypename/rfc/\", \"uploaded_filename\": \"\", \"words\": 35738}", "text_sha256": "89b01010d1ff76dc260539d5a453dde0b116d857db43c9e09d4a3dc062c9ce31"}, {"draft_name": "draft-ietf-tls-tls13", "facts": [{"char_end": 13945, "char_start": 13673, "evidence_quote": "{\"id\": 1295893, \"originaltargetaliasname\": \"rfc8446\", \"relationship\": \"/api/v1/name/docrelationshipname/became_rfc/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1295893/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8446/\"}", "fact_id": "became_rfc", "field": "became_rfc", "raw_char_end": 13945, "raw_char_start": 13673, "value": "draft-ietf-tls-tls13->RFC 8446"}, {"char_end": 3140, "char_start": 2873, "evidence_quote": "{\"id\": 630806, \"originaltargetaliasname\": \"rfc6066\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630806/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}", "fact_id": "informative_reference:6066", "field": "informative_reference", "raw_char_end": 3140, "raw_char_start": 2873, "value": "/api/v1/doc/document/rfc6066/"}, {"char_end": 3409, "char_start": 3142, "evidence_quote": "{\"id\": 630807, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630807/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}", "fact_id": "informative_reference:5246", "field": "informative_reference", "raw_char_end": 3409, "raw_char_start": 3142, "value": "/api/v1/doc/document/rfc5246/"}, {"char_end": 3678, "char_start": 3411, "evidence_quote": "{\"id\": 630805, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630805/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}", "fact_id": "informative_reference:5077", "field": "informative_reference", "raw_char_end": 3678, "raw_char_start": 3411, "value": "/api/v1/doc/document/rfc5077/"}, {"char_end": 6098, "char_start": 5831, "evidence_quote": "{\"id\": 630795, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630795/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}", "fact_id": "normative_reference:5705", "field": "normative_reference", "raw_char_end": 6098, "raw_char_start": 5831, "value": "/api/v1/doc/document/rfc5705/"}, {"char_end": 12864, "char_start": 12597, "evidence_quote": "{\"id\": 630768, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630768/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}", "fact_id": "normative_reference:6961", "field": "normative_reference", "raw_char_end": 12864, "raw_char_start": 12597, "value": "/api/v1/doc/document/rfc6961/"}], "kind": "datatracker_relation", "occurred_at": "2026-09-07T04:14:32.295913Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 0, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131", "record_id": "ietf:datatracker-relation:draft-ietf-tls-tls13:rfc8446", "retrieval_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=draft-ietf-tls-tls13&limit=100", "rfc_number": 8446, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131", "source_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=draft-ietf-tls-tls13&limit=100", "temporal_semantics": "retrieval_observation_only", "text": "{\"meta\": {\"limit\": 100, \"next\": null, \"offset\": 0, \"previous\": null, \"total_count\": 51}, \"objects\": [{\"id\": 630809, \"originaltargetaliasname\": \"bcp72\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630809/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/bcp72/\"}, {\"id\": 630802, \"originaltargetaliasname\": \"bcp26\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630802/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/bcp26/\"}, {\"id\": 630782, \"originaltargetaliasname\": \"bcp106\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630782/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/bcp106/\"}, {\"id\": 630774, \"originaltargetaliasname\": \"bcp14\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630774/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/bcp14/\"}, {\"id\": 282421, \"originaltargetaliasname\": \"draft-ietf-tls-rfc5246-bis\", \"relationship\": \"/api/v1/name/docrelationshipname/replaces/\", \"resource_uri\": \"/api/v1/doc/relateddocument/282421/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/draft-ietf-tls-rfc5246-bis/\"}, {\"id\": 630813, \"originaltargetaliasname\": \"rfc7568\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630813/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7568/\"}, {\"id\": 630810, \"originaltargetaliasname\": \"draft-ietf-tls-iana-registry-updates\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630810/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/draft-ietf-tls-iana-registry-updates/\"}, {\"id\": 630812, \"originaltargetaliasname\": \"rfc5116\", \"relationship\": \"/api/v1/name/docrelationshipname/refunk/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630812/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5116/\"}, {\"id\": 630811, \"originaltargetaliasname\": \"rfc5280\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630811/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5280/\"}, {\"id\": 630808, \"originaltargetaliasname\": \"rfc5756\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630808/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5756/\"}, {\"id\": 630806, \"originaltargetaliasname\": \"rfc6066\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630806/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}, {\"id\": 630807, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630807/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}, {\"id\": 630805, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630805/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}, {\"id\": 630804, \"originaltargetaliasname\": \"rfc8174\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630804/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8174/\"}, {\"id\": 630803, \"originaltargetaliasname\": \"rfc8032\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630803/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8032/\"}, {\"id\": 630801, \"originaltargetaliasname\": \"rfc2119\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630801/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc2119/\"}, {\"id\": 630800, \"originaltargetaliasname\": \"rfc6347\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630800/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6347/\"}, {\"id\": 630799, \"originaltargetaliasname\": \"rfc4492\", \"relationship\": \"/api/v1/name/docrelationshipname/refunk/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630799/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc4492/\"}, {\"id\": 630798, \"originaltargetaliasname\": \"rfc7539\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630798/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7539/\"}, {\"id\": 630797, \"originaltargetaliasname\": \"rfc7301\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630797/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7301/\"}, {\"id\": 630796, \"originaltargetaliasname\": \"rfc6176\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630796/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6176/\"}, {\"id\": 630795, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630795/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}, {\"id\": 630794, \"originaltargetaliasname\": \"rfc6979\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630794/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6979/\"}, {\"id\": 630793, \"originaltargetaliasname\": \"rfc5764\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630793/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5764/\"}, {\"id\": 630792, \"originaltargetaliasname\": \"rfc8017\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630792/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8017/\"}, {\"id\": 630791, \"originaltargetaliasname\": \"draft-ietf-tls-tls13-vectors\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630791/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/draft-ietf-tls-tls13-vectors/\"}, {\"id\": 630790, \"originaltargetaliasname\": \"rfc6520\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630790/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6520/\"}, {\"id\": 630789, \"originaltargetaliasname\": \"rfc5869\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630789/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5869/\"}, {\"id\": 630788, \"originaltargetaliasname\": \"rfc6960\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630788/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6960/\"}, {\"id\": 630787, \"originaltargetaliasname\": \"rfc7748\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630787/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7748/\"}, {\"id\": 630786, \"originaltargetaliasname\": \"rfc8126\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630786/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8126/\"}, {\"id\": 630785, \"originaltargetaliasname\": \"rfc7924\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630785/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7924/\"}, {\"id\": 630784, \"originaltargetaliasname\": \"rfc5929\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630784/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5929/\"}, {\"id\": 630783, \"originaltargetaliasname\": \"rfc8305\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630783/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8305/\"}, {\"id\": 630781, \"originaltargetaliasname\": \"rfc6091\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630781/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6091/\"}, {\"id\": 630780, \"originaltargetaliasname\": \"rfc4086\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630780/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc4086/\"}, {\"id\": 630779, \"originaltargetaliasname\": \"rfc7627\", \"relationship\": \"/api/v1/name/docrelationshipname/refunk/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630779/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7627/\"}, {\"id\": 630778, \"originaltargetaliasname\": \"rfc6655\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630778/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6655/\"}, {\"id\": 630777, \"originaltargetaliasname\": \"rfc7250\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630777/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7250/\"}, {\"id\": 630776, \"originaltargetaliasname\": \"rfc7507\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630776/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7507/\"}, {\"id\": 630775, \"originaltargetaliasname\": \"rfc7685\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630775/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7685/\"}, {\"id\": 630773, \"originaltargetaliasname\": \"rfc2104\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630773/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc2104/\"}, {\"id\": 630771, \"originaltargetaliasname\": \"rfc4346\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630771/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc4346/\"}, {\"id\": 630772, \"originaltargetaliasname\": \"rfc4366\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630772/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc4366/\"}, {\"id\": 630770, \"originaltargetaliasname\": \"rfc6962\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630770/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6962/\"}, {\"id\": 630769, \"originaltargetaliasname\": \"rfc3552\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630769/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc3552/\"}, {\"id\": 630768, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630768/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}, {\"id\": 630767, \"originaltargetaliasname\": \"rfc7919\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630767/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7919/\"}, {\"id\": 630766, \"originaltargetaliasname\": \"rfc7465\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630766/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7465/\"}, {\"id\": 630814, \"originaltargetaliasname\": \"rfc7230\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630814/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7230/\"}, {\"id\": 1295893, \"originaltargetaliasname\": \"rfc8446\", \"relationship\": \"/api/v1/name/docrelationshipname/became_rfc/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1295893/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8446/\"}]}", "text_sha256": "3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131"}, {"facts": [{"char_end": 3417, "char_start": 3161, "evidence_quote": "{\"id\": 1269005, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269005/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}", "fact_id": "informative_reference:5246", "field": "informative_reference", "raw_char_end": 3417, "raw_char_start": 3161, "value": "/api/v1/doc/document/rfc5246/"}, {"char_end": 3675, "char_start": 3419, "evidence_quote": "{\"id\": 1269006, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269006/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}", "fact_id": "informative_reference:5077", "field": "informative_reference", "raw_char_end": 3675, "raw_char_start": 3419, "value": "/api/v1/doc/document/rfc5077/"}, {"char_end": 5223, "char_start": 4967, "evidence_quote": "{\"id\": 1269016, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269016/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}", "fact_id": "normative_reference:5705", "field": "normative_reference", "raw_char_end": 5223, "raw_char_start": 4967, "value": "/api/v1/doc/document/rfc5705/"}, {"char_end": 10425, "char_start": 10169, "evidence_quote": "{\"id\": 1269043, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269043/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}", "fact_id": "normative_reference:6961", "field": "normative_reference", "raw_char_end": 10425, "raw_char_start": 10169, "value": "/api/v1/doc/document/rfc6961/"}, {"char_end": 12717, "char_start": 12466, "evidence_quote": "{\"id\": 1591472, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591472/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}", "fact_id": "normative_reference:6066", "field": "normative_reference", "raw_char_end": 12717, "raw_char_start": 12466, "value": "/api/v1/doc/document/rfc6066/"}], "kind": "datatracker_relation", "occurred_at": "2026-09-07T04:14:41.148823Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 0, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7", "record_id": "ietf:datatracker-relation:rfc8446", "retrieval_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=rfc8446&limit=100", "rfc_number": 8446, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7", "source_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=rfc8446&limit=100", "temporal_semantics": "retrieval_observation_only", "text": "{\"meta\": {\"limit\": 100, \"next\": null, \"offset\": 0, \"previous\": null, \"total_count\": 57}, \"objects\": [{\"id\": 1269002, \"originaltargetaliasname\": \"bcp72\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269002/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/bcp72/\"}, {\"id\": 1269009, \"originaltargetaliasname\": \"bcp26\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269009/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/bcp26/\"}, {\"id\": 1269029, \"originaltargetaliasname\": \"bcp106\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269029/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/bcp106/\"}, {\"id\": 1269037, \"originaltargetaliasname\": \"bcp14\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269037/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/bcp14/\"}, {\"id\": 666521, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/obs/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666521/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}, {\"id\": 666523, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/updates/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666523/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}, {\"id\": 666520, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/obs/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666520/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}, {\"id\": 666524, \"originaltargetaliasname\": \"rfc6066\", \"relationship\": \"/api/v1/name/docrelationshipname/updates/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666524/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}, {\"id\": 666522, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/obs/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666522/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}, {\"id\": 1268998, \"originaltargetaliasname\": \"rfc7568\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1268998/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7568/\"}, {\"id\": 1269001, \"originaltargetaliasname\": \"rfc5280\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269001/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5280/\"}, {\"id\": 1269003, \"originaltargetaliasname\": \"rfc5756\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269003/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5756/\"}, {\"id\": 1269005, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269005/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}, {\"id\": 1269006, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269006/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}, {\"id\": 1269007, \"originaltargetaliasname\": \"rfc8174\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269007/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8174/\"}, {\"id\": 1269008, \"originaltargetaliasname\": \"rfc8032\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269008/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8032/\"}, {\"id\": 1269010, \"originaltargetaliasname\": \"rfc2119\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269010/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc2119/\"}, {\"id\": 1269011, \"originaltargetaliasname\": \"rfc6347\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269011/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6347/\"}, {\"id\": 1269015, \"originaltargetaliasname\": \"rfc6176\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269015/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6176/\"}, {\"id\": 1269016, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269016/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}, {\"id\": 1269018, \"originaltargetaliasname\": \"rfc5764\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269018/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5764/\"}, {\"id\": 1269019, \"originaltargetaliasname\": \"rfc8017\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269019/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8017/\"}, {\"id\": 1269020, \"originaltargetaliasname\": \"draft-ietf-tls-tls13-vectors\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269020/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/draft-ietf-tls-tls13-vectors/\"}, {\"id\": 1269021, \"originaltargetaliasname\": \"rfc6520\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269021/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6520/\"}, {\"id\": 1269023, \"originaltargetaliasname\": \"rfc6960\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269023/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6960/\"}, {\"id\": 1269024, \"originaltargetaliasname\": \"rfc7748\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269024/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7748/\"}, {\"id\": 1269025, \"originaltargetaliasname\": \"rfc8126\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269025/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8126/\"}, {\"id\": 1269026, \"originaltargetaliasname\": \"rfc7924\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269026/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7924/\"}, {\"id\": 1269027, \"originaltargetaliasname\": \"rfc5929\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269027/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5929/\"}, {\"id\": 1269028, \"originaltargetaliasname\": \"rfc8305\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269028/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8305/\"}, {\"id\": 1269030, \"originaltargetaliasname\": \"rfc6091\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269030/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6091/\"}, {\"id\": 1269031, \"originaltargetaliasname\": \"rfc4086\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269031/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc4086/\"}, {\"id\": 1269034, \"originaltargetaliasname\": \"rfc7250\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269034/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7250/\"}, {\"id\": 1269035, \"originaltargetaliasname\": \"rfc7507\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269035/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7507/\"}, {\"id\": 1269036, \"originaltargetaliasname\": \"rfc7685\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269036/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7685/\"}, {\"id\": 1269038, \"originaltargetaliasname\": \"rfc2104\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269038/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc2104/\"}, {\"id\": 1269039, \"originaltargetaliasname\": \"rfc4346\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269039/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc4346/\"}, {\"id\": 1269040, \"originaltargetaliasname\": \"rfc4366\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269040/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc4366/\"}, {\"id\": 1269042, \"originaltargetaliasname\": \"rfc3552\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269042/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc3552/\"}, {\"id\": 1269043, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269043/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}, {\"id\": 1269044, \"originaltargetaliasname\": \"rfc7919\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269044/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7919/\"}, {\"id\": 1269045, \"originaltargetaliasname\": \"rfc7465\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269045/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7465/\"}, {\"id\": 1269046, \"originaltargetaliasname\": \"rfc7230\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269046/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7230/\"}, {\"id\": 1591467, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591467/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5869/\"}, {\"id\": 1591468, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591468/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6655/\"}, {\"id\": 1591469, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591469/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7301/\"}, {\"id\": 1591470, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591470/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6962/\"}, {\"id\": 1591471, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591471/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6979/\"}, {\"id\": 1591472, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591472/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}, {\"id\": 1591473, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591473/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8449/\"}, {\"id\": 1591474, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591474/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5116/\"}, {\"id\": 1591475, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591475/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8422/\"}, {\"id\": 1591476, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591476/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc4492/\"}, {\"id\": 1591477, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591477/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8439/\"}, {\"id\": 1591478, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591478/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8447/\"}, {\"id\": 1591479, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591479/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7627/\"}, {\"id\": 1591480, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591480/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6101/\"}]}", "text_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}]}, "source_manifest_sha256": "287ca4031b223ad818d21781c575905d5ed30eb7e4e0e38fef196cf1ca1865b4"}, "length_bucket": "64k", "local_probe_trust_isolation": "non_independent_local_diagnostic", "motif": "tls13_handshake_succession+obsoletes_updates_closure", "pipeline_capabilities": {"generic_promotion": true, "generic_strict_replay": true}, "production_eligible": false, "promoted": true, "promotion": {"candidate_sha256": "6f2957d4a31198559985b79b011d9558f53a691eb1b4d276ff292762c80fb06f", "dense_audit_sha256": "4a828a66915b2c1c09806c53c22e3083ef9b4af5134fc9c2dff1fe4e79b96afd", "dense_chunking": {"aggregation": "max_similarity", "max_tokens": 192, "overlap_tokens": 32, "strategy": "tokenizer_token_windows"}, "dense_model_backend": "sentence-transformers-6.0.0", "dense_model_id": "sentence-transformers/all-MiniLM-L6-v2", "dense_model_provider": "huggingface", "dense_model_revision": "1110a243fdf4706b3f48f1d95db1a4f5529b4d41", "dense_ranking_sha256": "f6f5ee8a3d729bacde9a1680bd7689f3606e3835489081031ff1c2a345631171", "dense_score_metric": "dot_product", "dense_top_k": 3, "real_source_verified": true, "release_selection_sha256": "46c412e55a5e391ac46af172f4a11623a01c552075d5ba2d80f6604f48b5f89d", "schema_version": "train-ready-promotion-v2", "strict_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "strict_replay_revision": "longworld-strict-replay-v5", "task_candidate_content_commitment": {"content_sha256": "add87b971ebf4b92f49bc25547b07bb1eb21000961785d5bd3f2e5b6b41f7342", "length_bucket": "64k", "view": "cf", "world_id": "ietf-tls13-handshake-succession-v4"}, "task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "sidecar_schema_version": "longworld.task-replay-sidecar.v3"}, "task_semantic_commitment_sha256": "d0ba9fce5d7a14d72aa724fb05a7f9c6c81ffcf19090647fc0fca5aa6cfe21a5", "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb"}, "promotion_eligible": true, "query_id": "ietf-tls13-handshake-succession-v4:64k:standard:cf", "query_timing": "first", "question": "Resolve the effective TLS 1.3 handshake protocol succession at 2026-01-31T00:00:00Z for this scenario (canonical JSON): {\"protocol\":\"tls\",\"publication\":\"rfc8446\",\"succession\":\"obsoletes_and_updates\",\"version\":\"1.3\"}. Return exactly one JSON object with these keys in this order: [\"current_protocol\",\"obsoletes_session_tickets\",\"obsoletes_tls12\",\"obsoletes_ocsp_stapling\",\"updates_exporters\",\"updates_extensions\"]. Use this exact per-field output codebook (canonical JSON): {\"current_protocol\":{\"code\":\"TLS_1_3_RFC8446\",\"meaning\":\"RFC 8446 is TLS protocol version 1.3\"},\"obsoletes_ocsp_stapling\":{\"code\":\"RFC6961\",\"meaning\":\"RFC 8446 obsoletes multiple certificate status requests\"},\"obsoletes_session_tickets\":{\"code\":\"RFC5077\",\"meaning\":\"RFC 8446 obsoletes TLS session resumption tickets\"},\"obsoletes_tls12\":{\"code\":\"RFC5246\",\"meaning\":\"RFC 8446 obsoletes TLS protocol version 1.2\"},\"updates_exporters\":{\"code\":\"RFC5705\",\"meaning\":\"RFC 8446 updates TLS keying material exporters\"},\"updates_extensions\":{\"code\":\"RFC6066\",\"meaning\":\"RFC 8446 updates TLS extension definitions\"}}. Resolve each field independently from the supplied RFC graph; use the string \"UNKNOWN\" only for a field whose required evidence or relation is absent.", "real_source_family_ids": ["ietf_standards"], "real_source_token_ratio": 0.8702304063029637, "real_source_verified": true, "real_source_workflow_ids": ["task-source:73e262c5a2939b145819"], "schema_version": "longworld.ietf-tls13-generation-candidate.v1", "semantic_base_task_id": "250a7cfbf1fb2c0a24b9", "semantic_growth_group_id": "ietf-tls13-handshake-succession-v4|ietf-tls13-handshake-succession", "semantic_tokens": {"causal_supporting": 16788, "event_bearing": 48321, "generic_background": 0, "internal": 65114, "measurement_basis": "exact_pinned_tokenizer_replay_relation_path_whole_artifact_upper_bound", "proof_bearing": 48321}, "source_binding": {"signed_manifest_sha256": "287ca4031b223ad818d21781c575905d5ed30eb7e4e0e38fef196cf1ca1865b4"}, "source_family_ids": ["ietf_standards"], "source_origins": ["real_public", "synthetic_counterfactual"], "source_record_ids": ["ietf:draft:draft-ietf-tls-tls13-28", "ietf:rfc:5077", "ietf:rfc:5246", "ietf:rfc:5705", "ietf:rfc:6066", "ietf:rfc:6961", "ietf:rfc:8446"], "source_record_ids_by_artifact": {"ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2": ["ietf:draft:draft-ietf-tls-tls13-28"], "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8": ["ietf:rfc:5077"], "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681": ["ietf:rfc:5246"], "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1": ["ietf:rfc:5246"], "ietf:rfc:5246:chars:0072013-0109924:596a422d1201": ["ietf:rfc:5246"], "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44": ["ietf:rfc:5705"], "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44": ["ietf:rfc:6066"], "ietf:rfc:6961:chars:0000000-0021461:bde82964b134": ["ietf:rfc:6961"], "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4": ["ietf:rfc:8446"], "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093": ["ietf:rfc:8446"]}, "source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2018-03-01T00:00:00Z", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_relation_kind": "published_as"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "updates"}], "source_relation_id": "6ee046ba0ec12a545607", "source_relation_ids": ["ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "ietf:updates:ietf:rfc:8446:ietf:rfc:6066"], "split": "train", "split_strategy": "world", "strict_replay_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "strict_support_event_count": 5, "task_proof_receipt": {"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "artifact_aligned_windows": {"16k": {"enumeration_sha256": "176dd7ff0b642099e09047b3cdbb45efc68edb43dd0eb6792ae30e102a6d56b8", "max_artifact_count": 3, "max_span_tokens": 16365, "status": "insufficient", "strict_window_count": 20, "window_tokens": 16384}, "4k": {"enumeration_sha256": "c17059cf0a6db37f54932b489e8b206cba9297c08835f29fb38c17d5547143a3", "max_artifact_count": 1, "max_span_tokens": 4053, "status": "insufficient", "strict_window_count": 3, "window_tokens": 4096}, "8k": {"enumeration_sha256": "17c2a9b44867d08be887c7aa164a1f5a325af0400cd974e3d29c316e3e490030", "max_artifact_count": 1, "max_span_tokens": 8187, "status": "insufficient", "strict_window_count": 8, "window_tokens": 8192}}, "artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "end_token": 9837, "start_token": 0, "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "end_token": 10302, "start_token": 9841, "text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "end_token": 18494, "start_token": 10307, "text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "end_token": 26619, "start_token": 18498, "text_sha256": "f9892786fca5daa1aca587f899ae9a4d74912e13b89740492729f5c3fe986064"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "end_token": 34756, "start_token": 26623, "text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "end_token": 34863, "start_token": 34760, "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "end_token": 43009, "start_token": 34867, "text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "end_token": 48194, "start_token": 43013, "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "end_token": 52252, "start_token": 48199, "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "end_token": 65114, "start_token": 52257, "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}], "checks": {"answer_surface_free": true, "artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_top3_prefixes_insufficient": true, "body_and_classification_valid": true, "counterfactual_changes_answer": true, "empty_selection_insufficient": true, "exact_token_count_recomputed": true, "full_counterfactual_sufficient": true, "full_replay_sufficient": true, "lexical_tfidf_top3_prefixes_insufficient": true, "minimal_counterfactual_sufficient": true, "minimal_replay_sufficient": true, "minimum_complexity_met": true, "no_shortcut": true, "raw_token_executable_windows_insufficient": true, "remove_one_fails": true, "single_essential_insufficient": true, "source_relations_replayed": true}, "essential_artifact_ids": ["ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"], "proof_input_sha256": "6f2957d4a31198559985b79b011d9558f53a691eb1b4d276ff292762c80fb06f", "raw_token_offset_windows": {"16k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "159605de9d850aaa5e65289a749ab780b61cb505a50fe9f8a06abe85ea650063", "max_artifact_count": 5, "monotonicity_proven": false, "semantic_state_count": 16, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "44a4330b371decef1948dfc2a46a5eac9baa66e8ce2c93d0105748d586b4e537", "exact_raw_executable": {"enumeration_sha256": "44a4330b371decef1948dfc2a46a5eac9baa66e8ce2c93d0105748d586b4e537", "exhaustive_start_count": 48731, "semantic_state_count": 18, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 18, "window_tokens": 16384}, "4k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "57865553c94928c8b274e05a7294df1411c0d970a68bbdccd0f02f1916766cf6", "max_artifact_count": 3, "monotonicity_proven": false, "semantic_state_count": 19, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "6c49fd92d918b166dc6519a3ec1bbb751453b4e2b032ed36b69e633b48a42914", "exact_raw_executable": {"enumeration_sha256": "6c49fd92d918b166dc6519a3ec1bbb751453b4e2b032ed36b69e633b48a42914", "exhaustive_start_count": 61019, "semantic_state_count": 4, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 4, "window_tokens": 4096}, "8k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "1c8147135bb36caef0215ba3a71e1280b3adc7557e4d8e6578451b1f411afb4a", "max_artifact_count": 3, "monotonicity_proven": false, "semantic_state_count": 19, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "520eda04c4f0e40b887017e30f18ba13c85a092576996e26b44a5a8b767c9c6b", "exact_raw_executable": {"enumeration_sha256": "520eda04c4f0e40b887017e30f18ba13c85a092576996e26b44a5a8b767c9c6b", "exhaustive_start_count": 56923, "semantic_state_count": 9, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 9, "window_tokens": 8192}}, "receipt_sha256": "da66ada2a0f5addaa6922c37d1b27e4c41d5db85ebfb4d98bbd79da4a26b0b67", "replay": {"counterfactual_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "empty_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "full_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "minimal_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "remove_one": [{"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}], "single_essential": [{"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}]}, "retrieval": {"bm25": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "score": "117.51975065902222"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "score": "85.654555114976148"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "score": "73.28241578146303"}]}, "lexical_tfidf": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "score": "0.42605918578853391"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "score": "0.41657976134259173"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "score": "0.32400943442123614"}]}}, "schema_version": "longworld.task-proof-receipt.v5", "sidecar_schema_version": "longworld.task-replay-sidecar.v3", "task_replay_sidecar_sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "tokenizer": {"asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "document_context_tokens": 65114, "model_id": "Qwen/Qwen3.5-4B", "revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "serialized_context_tokens": 65493}, "window_scope": "exact_raw_slice_replay_with_separate_intersection_upper_bound"}, "task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "sidecar_schema_version": "longworld.task-replay-sidecar.v3"}, "task_view_projection": {"adapter_context_header": null, "adapter_replay_answer_sha256": "a03626479b4898da51b3abf1fe4f8ded86d258c20d6cbf7cc0b222ad8dd8ad57", "artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "text_sha256": "f9892786fca5daa1aca587f899ae9a4d74912e13b89740492729f5c3fe986064"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}], "chronology": [], "derivation_revision": "longworld.task-view-derivation.v4", "document_context_sha256": "6634ed81a76db995b5cc71072f5cbbdcd99c2e3222e320cb9fa4f9e4e69854b5", "dossier_id": "a00388a585cd2f4d89b3", "materialized_replay_answer_sha256": "a03626479b4898da51b3abf1fe4f8ded86d258c20d6cbf7cc0b222ad8dd8ad57", "parent_artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "provenance_id": "source-span-sha256:e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef:0:41965:80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49", "source_origin": "real_public", "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "provenance_id": "source-span-sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd:335784:337410:0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5", "source_origin": "real_public", "text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:0:35597:ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6", "source_origin": "real_public", "text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "provenance_id": "source-span-sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd:0:38698:61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e", "source_origin": "real_public", "text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:35597:72013:3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69", "source_origin": "real_public", "text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "provenance_id": "source-span-sha256:9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6:0:342:d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b", "source_origin": "real_public", "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:72013:109924:596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef", "source_origin": "real_public", "text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "provenance_id": "source-span-sha256:c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db:0:21461:bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a", "source_origin": "real_public", "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "provenance_id": "source-span-sha256:29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a:0:16341:4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473", "source_origin": "real_public", "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "provenance_id": "source-span-sha256:6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0:0:55067:a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813", "source_origin": "real_public", "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}], "parent_candidate_sha256": "3a56e24f453b328fb4635d85eaa9c95fe023fe8be199a806064111960bea07df", "parent_source_binding_sha256": "fc34a690fcb8ffffb64df92f1df724ea75d43ce88c10daf432fb085fee09ddbe", "parent_task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "e77fe08f470f227b433656f0fac14e677653bdb1f53faa939175c5c883608e76", "sidecar_schema_version": "longworld.task-replay-sidecar.v1"}, "schema_version": "longworld.task-view-projection.v1", "source_token_measurement_receipt": {"final_prompt_tokens": 65493, "measurement_basis": "final_prompt_real_source_marginal", "parent_artifact_token_contributions": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "parent_text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49", "retained_as_real": true, "token_contribution": 9837}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "parent_text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6", "retained_as_real": true, "token_contribution": 8192}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "parent_text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69", "retained_as_real": true, "token_contribution": 8137}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "parent_text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef", "retained_as_real": true, "token_contribution": 8146}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "parent_text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473", "retained_as_real": true, "token_contribution": 4057}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "parent_text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813", "retained_as_real": true, "token_contribution": 12862}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "parent_text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a", "retained_as_real": true, "token_contribution": 5186}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "parent_text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b", "retained_as_real": true, "token_contribution": 108}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "parent_text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e", "retained_as_real": false, "token_contribution": 8145}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "parent_text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5", "retained_as_real": true, "token_contribution": 464}], "parent_document_context_tokens": 65134, "parent_real_source_token_ratio": 1.0, "projected_document_context_tokens": 65114, "real_source_marginal_tokens": 56994, "real_source_token_ratio": 0.8702304063029637, "retained_parent_document_tokens": 56989, "schema_version": "longworld.source-token-measurement-receipt.v2", "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "tokenizer_model_id": "Qwen/Qwen3.5-4B", "tokenizer_revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "without_real_prompt_tokens": 8499}, "view": "cf"}, "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "tokenizer_context_tokens": 65493, "tokenizer_model_id": "Qwen/Qwen3.5-4B", "tokenizer_revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "train_ready": true, "training_objective": "sft", "trust_scope": "local_probe", "trust_valid_for_production": false, "verification": {"artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_topk_insufficient": true, "candidate_mode": false, "closed_book_unsolved": true, "contiguous_windows_insufficient": true, "counterfactual_changes_answer": true, "counterfactual_replay_sufficient": true, "distractor_invariance_gold": true, "embedding_topk_insufficient": true, "essential_single_doc_insufficient": true, "essential_surface_gold_free": true, "essential_text_grounded": true, "full_sufficient": true, "lexical_tfidf_topk_insufficient": true, "local_window_insufficient": true, "min_complexity": true, "minimal_sufficient": true, "no_shortcut": true, "production_mode": true, "question_only_unsolved": true, "remove_one_fails": true, "schema_ok": true, "semantic_sufficient": true, "strict_executable_sufficient": true, "surface_match": true}, "verified_derived_order_relation_edges": [{"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2018-03-01T00:00:00Z", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_relation_kind": "published_as"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "updates"}], "view": "cf", "view_verification": {"classification_ok": true, "content_gate_eligible": true, "essential_present": true, "expected_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "global_proof_green": true, "production_eligible": false, "semantic_text_grounded": true, "strict_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}"}, "workflow_ids": ["task-source:73e262c5a2939b145819"], "workflow_kinds": ["hybrid_causal", "real_source_derived"], "world_id": "ietf-tls13-handshake-succession-v4"} +{"actual_context_tokens": 65513, "answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "answer_program_id": "ietf.tls13_handshake_succession.v1", "artifact_classification": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef:0:41965:80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49", "source_char_end": 41965, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:5077", "source_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "evidence_role": "causal_supporting", "provenance_id": "source-span-sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd:335784:337410:0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5", "source_char_end": 337410, "source_char_start": 335784, "source_origin": "real_public", "source_record_id": "ietf:rfc:8446", "source_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "evidence_role": "causal_supporting", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:0:35597:ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6", "source_char_end": 35597, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:5246", "source_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd:0:38698:61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e", "source_char_end": 38698, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:8446", "source_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "evidence_role": "causal_supporting", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:35597:72013:3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69", "source_char_end": 72013, "source_char_start": 35597, "source_origin": "real_public", "source_record_id": "ietf:rfc:5246", "source_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6:0:342:d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b", "source_char_end": 342, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "source_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:72013:109924:596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef", "source_char_end": 109924, "source_char_start": 72013, "source_origin": "real_public", "source_record_id": "ietf:rfc:5246", "source_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db:0:21461:bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a", "source_char_end": 21461, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:6961", "source_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a:0:16341:4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473", "source_char_end": 16341, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:5705", "source_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0:0:55067:a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813", "source_char_end": 55067, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:6066", "source_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}], "attestation": {"digest": "16743f42bd297d2d1bc7ea832d007547f5a68f3abead82ca141ce3d491aa2725", "environment": "probe", "key_id": "probe-promotion-c9a5dab33e62d6c983d4696be812c27f", "purpose": "sft_row", "role": "promotion", "scheme": "hmac-sha256-v2"}, "authentic_source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}], "authentic_source_relation_id": "e0dcb1ad1463850449f0", "base_task_id": "776021e7d8e7f94c336a", "cf_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "composition_method": "same_case_dossier", "content_gate_eligible": true, "context": "Question:\nResolve the effective TLS 1.3 handshake protocol succession at 2026-01-31T00:00:00Z for this scenario (canonical JSON): {\"protocol\":\"tls\",\"publication\":\"rfc8446\",\"succession\":\"obsoletes_and_updates\",\"version\":\"1.3\"}. Return exactly one JSON object with these keys in this order: [\"current_protocol\",\"obsoletes_session_tickets\",\"obsoletes_tls12\",\"obsoletes_ocsp_stapling\",\"updates_exporters\",\"updates_extensions\"]. Use this exact per-field output codebook (canonical JSON): {\"current_protocol\":{\"code\":\"TLS_1_3_RFC8446\",\"meaning\":\"RFC 8446 is TLS protocol version 1.3\"},\"obsoletes_ocsp_stapling\":{\"code\":\"RFC6961\",\"meaning\":\"RFC 8446 obsoletes multiple certificate status requests\"},\"obsoletes_session_tickets\":{\"code\":\"RFC5077\",\"meaning\":\"RFC 8446 obsoletes TLS session resumption tickets\"},\"obsoletes_tls12\":{\"code\":\"RFC5246\",\"meaning\":\"RFC 8446 obsoletes TLS protocol version 1.2\"},\"updates_exporters\":{\"code\":\"RFC5705\",\"meaning\":\"RFC 8446 updates TLS keying material exporters\"},\"updates_extensions\":{\"code\":\"RFC6066\",\"meaning\":\"RFC 8446 updates TLS extension definitions\"}}. Resolve each field independently from the supplied RFC graph; use the string \"UNKNOWN\" only for a field whose required evidence or relation is absent.\n\nContext (internal records):\nNetwork Working Group J. Salowey\nRequest for Comments: 5077 H. Zhou\nObsoletes: 4507 Cisco Systems\nCategory: Standards Track P. Eronen\n Nokia\n H. Tschofenig\n Nokia Siemens Networks\n January 2008\n\n\n Transport Layer Security (TLS) Session Resumption without\n Server-Side State\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document describes a mechanism that enables the Transport Layer\n Security (TLS) server to resume sessions and avoid keeping per-client\n session state. The TLS server encapsulates the session state into a\n ticket and forwards it to the client. The client can subsequently\n resume a session using the obtained ticket. This document obsoletes\n RFC 4507.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 1]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3. Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . 4\n 3.2. SessionTicket TLS Extension . . . . . . . . . . . . . . . 7\n 3.3. NewSessionTicket Handshake Message . . . . . . . . . . . . 8\n 3.4. Interaction with TLS Session ID . . . . . . . . . . . . . 9\n 4. Recommended Ticket Construction . . . . . . . . . . . . . . . 10\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . 12\n 5.1. Invalidating Sessions . . . . . . . . . . . . . . . . . . 12\n 5.2. Stolen Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.3. Forged Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.4. Denial of Service Attacks . . . . . . . . . . . . . . . . 12\n 5.5. Ticket Protection Key Management . . . . . . . . . . . . . 13\n 5.6. Ticket Lifetime . . . . . . . . . . . . . . . . . . . . . 13\n 5.7. Alternate Ticket Formats and Distribution Schemes . . . . 13\n 5.8. Identity Privacy, Anonymity, and Unlinkability . . . . . . 14\n 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 14\n 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 15\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 15\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 15\n Appendix A. Discussion of Changes to RFC 4507 . . . . . . . . . . 17\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 2]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n1. Introduction\n\n This document defines a way to resume a Transport Layer Security\n (TLS) session without requiring session-specific state at the TLS\n server. This mechanism may be used with any TLS ciphersuite. This\n document applies to both TLS 1.0 defined in [RFC2246], and TLS 1.1\n defined in [RFC4346]. The mechanism makes use of TLS extensions\n defined in [RFC4366] and defines a new TLS message type.\n\n This mechanism is useful in the following situations:\n\n 1. servers that handle a large number of transactions from different\n users\n\n 2. servers that desire to cache sessions for a long time\n\n 3. ability to load balance requests across servers\n\n 4. embedded servers with little memory\n\n This document obsoletes RFC 4507 [RFC4507] to correct an error in the\n encoding that caused the specification to differ from deployed\n implementations. At the time of this writing, there are no known\n implementations that follow the encoding specified in RFC 4507. This\n update to RFC 4507 aligns the document with currently deployed\n implementations. More details of the change are given in Appendix A.\n\n2. Terminology\n\n Within this document, the term 'ticket' refers to a cryptographically\n protected data structure that is created and consumed by the server\n to rebuild session-specific state.\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Protocol\n\n This specification describes a mechanism to distribute encrypted\n session-state information to the client in the form of a ticket and a\n mechanism to present the ticket back to the server. The ticket is\n created by a TLS server and sent to a TLS client. The TLS client\n presents the ticket to the TLS server to resume a session.\n Implementations of this specification are expected to support both\n mechanisms. Other specifications can take advantage of the session\n tickets, perhaps specifying alternative means for distribution or\n selection. For example, a separate specification may describe an\n\n\n\nSalowey, et al. Standards Track [Page 3]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n alternate way to distribute a ticket and use the TLS extension in\n this document to resume the session. This behavior is beyond the\n scope of the document and would need to be described in a separate\n specification.\n\n3.1. Overview\n\n The client indicates that it supports this mechanism by including a\n SessionTicket TLS extension in the ClientHello message. The\n extension will be empty if the client does not already possess a\n ticket for the server. The server sends an empty SessionTicket\n extension to indicate that it will send a new session ticket using\n the NewSessionTicket handshake message. The extension is described\n in Section 3.2.\n\n If the server wants to use this mechanism, it stores its session\n state (such as ciphersuite and master secret) to a ticket that is\n encrypted and integrity-protected by a key known only to the server.\n The ticket is distributed to the client using the NewSessionTicket\n TLS handshake message described in Section 3.3. This message is sent\n during the TLS handshake before the ChangeCipherSpec message, after\n the server has successfully verified the client's Finished message.\n\n Client Server\n\n ClientHello\n (empty SessionTicket extension)-------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1: Message Flow for Full Handshake Issuing New Session Ticket\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 4]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n The client caches this ticket along with the master secret and other\n parameters associated with the current session. When the client\n wishes to resume the session, it includes the ticket in the\n SessionTicket extension within the ClientHello message. Appendix A\n provides a detailed description of the encoding of the extension and\n changes from RFC 4507. The server then decrypts the received ticket,\n verifies the ticket's validity, retrieves the session state from the\n contents of the ticket, and uses this state to resume the session.\n The interaction with the TLS Session ID is described in Section 3.4.\n If the server successfully verifies the client's ticket, then it may\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello.\n\n Client Server\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2: Message Flow for Abbreviated Handshake Using New Session\n Ticket\n\n A recommended ticket format is given in Section 4.\n\n If the server cannot or does not want to honor the ticket, then it\n can initiate a full handshake with the client.\n\n In the case that the server does not wish to issue a new ticket at\n this time, it just completes the handshake without including a\n SessionTicket extension or NewSessionTicket handshake message. This\n is shown below (this flow is identical to Figure 1 in RFC 4346,\n except for the SessionTicket extension in the first message):\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 5]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 3: Message Flow for Server Completing Full Handshake Without\n Issuing New Session Ticket\n\n It is also permissible to have an exchange similar to Figure 3 using\n the abbreviated handshake defined in Figure 2 of RFC 4346, where the\n client uses the SessionTicket extension to resume the session, but\n the server does not wish to issue a new ticket, and therefore does\n not send a SessionTicket extension.\n\n If the server rejects the ticket, it may still wish to issue a new\n ticket after performing the full handshake as shown below (this flow\n is identical to Figure 1, except the SessionTicket extension in the\n ClientHello is not empty):\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 6]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 4: Message Flow for Server Rejecting Ticket, Performing Full\n Handshake, and Issuing New Session Ticket\n\n3.2. SessionTicket TLS Extension\n\n The SessionTicket TLS extension is based on [RFC4366]. The format of\n the ticket is an opaque structure used to carry session-specific\n state information. This extension may be sent in the ClientHello and\n ServerHello.\n\n If the client possesses a ticket that it wants to use to resume a\n session, then it includes the ticket in the SessionTicket extension\n in the ClientHello. If the client does not have a ticket and is\n prepared to receive one in the NewSessionTicket handshake message,\n then it MUST include a zero-length ticket in the SessionTicket\n extension. If the client is not prepared to receive a ticket in the\n NewSessionTicket handshake message, then it MUST NOT include a\n SessionTicket extension unless it is sending a non-empty ticket it\n received through some other means from the server.\n\n The server uses a zero-length SessionTicket extension to indicate to\n the client that it will send a new session ticket using the\n NewSessionTicket handshake message described in Section 3.3. The\n server MUST send this extension in the ServerHello if it wishes to\n issue a new ticket to the client using the NewSessionTicket handshake\n message. The server MUST NOT send this extension if it does not\n receive one in the ClientHello.\n\n\n\n\nSalowey, et al. Standards Track [Page 7]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n If the server fails to verify the ticket, then it falls back to\n performing a full handshake. If the ticket is accepted by the server\n but the handshake fails, the client SHOULD delete the ticket.\n\n The SessionTicket extension has been assigned the number 35. The\n extension_data field of SessionTicket extension contains the ticket.\n\n3.3. NewSessionTicket Handshake Message\n\n This message is sent by the server during the TLS handshake before\n the ChangeCipherSpec message. This message MUST be sent if the\n server included a SessionTicket extension in the ServerHello. This\n message MUST NOT be sent if the server did not include a\n SessionTicket extension in the ServerHello. This message is included\n in the hash used to create and verify the Finished message. In the\n case of a full handshake, the server MUST verify the client's\n Finished message before sending the ticket. The client MUST NOT\n treat the ticket as valid until it has verified the server's Finished\n message. If the server determines that it does not want to include a\n ticket after it has included the SessionTicket extension in the\n ServerHello, then it sends a zero-length ticket in the\n NewSessionTicket handshake message.\n\n If the server successfully verifies the client's ticket, then it MAY\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello in the abbreviated handshake. The client\n should start using the new ticket as soon as possible after it\n verifies the server's Finished message for new connections. Note\n that since the updated ticket is issued before the handshake\n completes, it is possible that the client may not put the new ticket\n into use before it initiates new connections. The server MUST NOT\n assume that the client actually received the updated ticket until it\n successfully verifies the client's Finished message.\n\n The NewSessionTicket handshake message has been assigned the number 4\n and its definition is given at the end of this section. The\n ticket_lifetime_hint field contains a hint from the server about how\n long the ticket should be stored. The value indicates the lifetime\n in seconds as a 32-bit unsigned integer in network byte order\n relative to when the ticket is received. A value of zero is reserved\n to indicate that the lifetime of the ticket is unspecified. A client\n SHOULD delete the ticket and associated state when the time expires.\n It MAY delete the ticket earlier based on local policy. A server MAY\n treat a ticket as valid for a shorter or longer period of time than\n what is stated in the ticket_lifetime_hint.\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 8]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n HandshakeType msg_type;\n uint24 length;\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case session_ticket: NewSessionTicket; /* NEW */\n } body;\n } Handshake;\n\n\n struct {\n uint32 ticket_lifetime_hint;\n opaque ticket<0..2^16-1>;\n } NewSessionTicket;\n\n3.4. Interaction with TLS Session ID\n\n If a server is planning on issuing a session ticket to a client that\n does not present one, it SHOULD include an empty Session ID in the\n ServerHello. If the server rejects the ticket and falls back to the\n full handshake then it may include a non-empty Session ID to indicate\n its support for stateful session resumption. If the client receives\n a session ticket from the server, then it discards any Session ID\n that was sent in the ServerHello.\n\n When presenting a ticket, the client MAY generate and include a\n Session ID in the TLS ClientHello. If the server accepts the ticket\n and the Session ID is not empty, then it MUST respond with the same\n Session ID present in the ClientHello. This allows the client to\n easily differentiate when the server is resuming a session from when\n it is falling back to a full handshake. Since the client generates a\n Session ID, the server MUST NOT rely upon the Session ID having a\n particular value when validating the ticket. If a ticket is\n presented by the client, the server MUST NOT attempt to use the\n Session ID in the ClientHello for stateful session resumption.\n Alternatively, the client MAY include an empty Session ID in the\n ClientHello. In this case, the client ignores the Session ID sent in\n the ServerHello and determines if the server is resuming a session by\n the subsequent handshake messages.\n\n\n\nSalowey, et al. Standards Track [Page 9]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n4. Recommended Ticket Construction\n\n This section describes a recommended format and protection for the\n ticket. Note that the ticket is opaque to the client, so the\n structure is not subject to interoperability concerns, and\n implementations may diverge from this format. If implementations do\n diverge from this format, they must take security concerns seriously.\n Clients MUST NOT examine the ticket under the assumption that it\n complies with this document.\n\n The server uses two different keys: one 128-bit key for Advanced\n Encryption Standard (AES) [AES] in Cipher Block Chaining (CBC) mode\n [CBC] encryption and one 256-bit key for HMAC-SHA-256 [RFC4634].\n\n The ticket is structured as follows:\n\n struct {\n opaque key_name[16];\n opaque iv[16];\n opaque encrypted_state<0..2^16-1>;\n opaque mac[32];\n } ticket;\n\n Here, key_name serves to identify a particular set of keys used to\n protect the ticket. It enables the server to easily recognize\n tickets it has issued. The key_name should be randomly generated to\n avoid collisions between servers. One possibility is to generate new\n random keys and key_name every time the server is started.\n\n The actual state information in encrypted_state is encrypted using\n 128-bit AES in CBC mode with the given IV. The Message\n Authentication Code (MAC) is calculated using HMAC-SHA-256 over\n key_name (16 octets) and IV (16 octets), followed by the length of\n the encrypted_state field (2 octets) and its contents (variable\n length).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 10]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n ProtocolVersion protocol_version;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n opaque master_secret[48];\n ClientIdentity client_identity;\n uint32 timestamp;\n } StatePlaintext;\n\n enum {\n anonymous(0),\n certificate_based(1),\n psk(2)\n } ClientAuthenticationType;\n\n struct {\n ClientAuthenticationType client_authentication_type;\n select (ClientAuthenticationType) {\n case anonymous: struct {};\n case certificate_based:\n ASN.1Cert certificate_list<0..2^24-1>;\n case psk:\n opaque psk_identity<0..2^16-1>; /* from [RFC4279] */\n };\n } ClientIdentity;\n\n The structure StatePlaintext stores the TLS session state including\n the master_secret. The timestamp within this structure allows the\n TLS server to expire tickets. To cover the authentication and key\n exchange protocols provided by TLS, the ClientIdentity structure\n contains the authentication type of the client used in the initial\n exchange (see ClientAuthenticationType). To offer the TLS server\n with the same capabilities for authentication and authorization, a\n certificate list is included in case of public-key-based\n authentication. The TLS server is therefore able to inspect a number\n of different attributes within these certificates. A specific\n implementation might choose to store a subset of this information or\n additional information. Other authentication mechanisms, such as\n Kerberos [RFC2712], would require different client identity data.\n Other TLS extensions may require the inclusion of additional data in\n the StatePlaintext structure.\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 11]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5. Security Considerations\n\n This section addresses security issues related to the usage of a\n ticket. Tickets must be authenticated and encrypted to prevent\n modification or eavesdropping by an attacker. Several attacks\n described below will be possible if this is not carefully done.\n\n Implementations should take care to ensure that the processing of\n tickets does not increase the chance of denial of service as\n described below.\n\n5.1. Invalidating Sessions\n\n The TLS specification requires that TLS sessions be invalidated when\n errors occur. [CSSC] discusses the security implications of this in\n detail. In the analysis within this paper, failure to invalidate\n sessions does not pose a security risk. This is because the TLS\n handshake uses a non-reversible function to derive keys for a session\n so information about one session does not provide an advantage to\n attack the master secret or a different session. If a session\n invalidation scheme is used, the implementation should verify the\n integrity of the ticket before using the contents to invalidate a\n session to ensure that an attacker cannot invalidate a chosen\n session.\n\n5.2. Stolen Tickets\n\n An eavesdropper or man-in-the-middle may obtain the ticket and\n attempt to use it to establish a session with the server; however,\n since the ticket is encrypted and the attacker does not know the\n secret key, a stolen ticket does not help an attacker resume a\n session. A TLS server MUST use strong encryption and integrity\n protection for the ticket to prevent an attacker from using a brute\n force mechanism to obtain the ticket's contents.\n\n5.3. Forged Tickets\n\n A malicious user could forge or alter a ticket in order to resume a\n session, to extend its lifetime, to impersonate another user, or to\n gain additional privileges. This attack is not possible if the\n ticket is protected using a strong integrity protection algorithm\n such as a keyed HMAC-SHA-256.\n\n5.4. Denial of Service Attacks\n\n The key_name field defined in the recommended ticket format helps the\n server efficiently reject tickets that it did not issue. However, an\n adversary could store or generate a large number of tickets to send\n\n\n\nSalowey, et al. Standards Track [Page 12]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n to the TLS server for verification. To minimize the possibility of a\n denial of service, the verification of the ticket should be\n lightweight (e.g., using efficient symmetric key cryptographic\n algorithms).\n\n5.5. Ticket Protection Key Management\n\n A full description of the management of the keys used to protect the\n ticket is beyond the scope of this document. A list of RECOMMENDED\n practices is given below.\n\n o The keys should be generated securely following the randomness\n recommendations in [RFC4086].\n\n o The keys and cryptographic protection algorithms should be at\n least 128 bits in strength. Some ciphersuites and applications\n may require cryptographic protection greater than 128 bits in\n strength.\n\n o The keys should not be used for any purpose other than generating\n and verifying tickets.\n\n o The keys should be changed regularly.\n\n o The keys should be changed if the ticket format or cryptographic\n protection algorithms change.\n\n5.6. Ticket Lifetime\n\n The TLS server controls the lifetime of the ticket. Servers\n determine the acceptable lifetime based on the operational and\n security requirements of the environments in which they are deployed.\n The ticket lifetime may be longer than the 24-hour lifetime\n recommended in [RFC4346]. TLS clients may be given a hint of the\n lifetime of the ticket. Since the lifetime of a ticket may be\n unspecified, a client has its own local policy that determines when\n it discards tickets.\n\n5.7. Alternate Ticket Formats and Distribution Schemes\n\n If the ticket format or distribution scheme defined in this document\n is not used, then great care must be taken in analyzing the security\n of the solution. In particular, if confidential information, such as\n a secret key, is transferred to the client, it MUST be done using\n secure communication so as to prevent attackers from obtaining or\n modifying the key. Also, the ticket MUST have its integrity and\n confidentiality protected with strong cryptographic techniques to\n prevent a breach in the security of the system.\n\n\n\nSalowey, et al. Standards Track [Page 13]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5.8. Identity Privacy, Anonymity, and Unlinkability\n\n This document mandates that the content of the ticket is\n confidentiality protected in order to avoid leakage of its content,\n such as user-relevant information. As such, it prevents disclosure\n of potentially sensitive information carried within the ticket.\n\n The initial handshake exchange, which was used to obtain the ticket,\n might not provide identity confidentiality of the client based on the\n properties of TLS. Another relevant security threat is the ability\n for an on-path adversary to observe multiple TLS handshakes where the\n same ticket is used, therefore concluding they belong to the same\n communication endpoints. Application designers that use the ticket\n mechanism described in this document should consider that\n unlinkability [ANON] is not necessarily provided.\n\n While a full discussion of these topics is beyond the scope of this\n document, it should be noted that it is possible to issue a ticket\n using a TLS renegotiation handshake that occurs after a secure tunnel\n has been established by a previous handshake. This may help address\n some privacy and unlinkability issues in some environments.\n\n6. Acknowledgements\n\n The authors would like to thank the following people for their help\n with preparing and reviewing this document: Eric Rescorla, Mohamad\n Badra, Tim Dierks, Nelson Bolyard, Nancy Cam-Winget, David McGrew,\n Rob Dugal, Russ Housley, Amir Herzberg, Bernard Aboba, and members of\n the TLS working group.\n\n [CSSC] describes a solution that is very similar to the one described\n in this document and gives a detailed analysis of the security\n considerations involved. [RFC2712] describes a mechanism for using\n Kerberos [RFC4120] in TLS ciphersuites, which helped inspire the use\n of tickets to avoid server state. [RFC4851] makes use of a similar\n mechanism to avoid maintaining server state for the cryptographic\n tunnel. [SC97] also investigates the concept of stateless sessions.\n\n The authors would also like to thank Jan Nordqvist, who found the\n encoding error in RFC 4507, corrected by this document. In addition\n Nagendra Modadugu, Wan-Teh Chang, and Michael D'Errico provided\n useful feedback during the review of this document.\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 14]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n7. IANA Considerations\n\n IANA has assigned a TLS extension number of 35 to the SessionTicket\n TLS extension from the TLS registry of ExtensionType values defined\n in [RFC4366].\n\n IANA has assigned a TLS HandshakeType number 4 to the\n NewSessionTicket handshake type from the TLS registry of\n HandshakeType values defined in [RFC4346].\n\n This document does not require any actions or assignments from IANA.\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2246] Dierks, T. and C. Allen, \"The TLS Protocol Version 1.0\",\n RFC 2246, January 1999.\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346, April 2006.\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, April 2006.\n\n [RFC4507] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 4507, May 2006.\n\n8.2. Informative References\n\n [AES] National Institute of Standards and Technology, \"Advanced\n Encryption Standard (AES)\", Federal Information Processing\n Standards (FIPS) Publication 197, November 2001.\n\n [ANON] Pfitzmann, A. and M. Hansen, \"Anonymity, Unlinkability,\n Unobservability, Pseudonymity, and Identity Management - A\n Consolidated Proposal for Terminology\", http://\n dud.inf.tu-dresden.de/literatur/\n Anon_Terminology_v0.26-1.pdf Version 0.26, December 2005.\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 15]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n [CBC] National Institute of Standards and Technology,\n \"Recommendation for Block Cipher Modes of Operation -\n Methods and Techniques\", NIST Special Publication 800-38A,\n December 2001.\n\n [CSSC] Shacham, H., Boneh, D., and E. Rescorla, \"Client-side\n caching for TLS\", Transactions on Information and System\n Security (TISSEC) , Volume 7, Issue 4, November 2004.\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [RFC4086] Eastlake, D., Schiller, J., and S. Crocker, \"Randomness\n Requirements for Security\", BCP 106, RFC 4086, June 2005.\n\n [RFC4120] Neuman, C., Yu, T., Hartman, S., and K. Raeburn, \"The\n Kerberos Network Authentication Service (V5)\", RFC 4120,\n July 2005.\n\n [RFC4279] Eronen, P. and H. Tschofenig, \"Pre-Shared Key Ciphersuites\n for Transport Layer Security (TLS)\", RFC 4279,\n December 2005.\n\n [RFC4634] Eastlake, D. and T. Hansen, \"US Secure Hash Algorithms\n (SHA and HMAC-SHA)\", RFC 4634, July 2006.\n\n [RFC4851] Cam-Winget, N., McGrew, D., Salowey, J., and H. Zhou, \"The\n Flexible Authentication via Secure Tunneling Extensible\n Authentication Protocol Method (EAP-FAST)\", RFC 4851,\n May 2007.\n\n [SC97] Aura, T. and P. Nikander, \"Stateless Connections\",\n Proceedings of the First International Conference on\n Information and Communication Security (ICICS '97) , 1997.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 16]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAppendix A. Discussion of Changes to RFC 4507\n\n RFC 4507 [RFC4507] defines a mechanism to resume a TLS session\n without maintaining server side state by specifying an encrypted\n ticket that is maintained on the client. The client presents this\n ticket to the server in a SessionTicket hello extension. The\n encoding in RFC 4507 used the XDR style encoding specified in TLS\n [RFC4346].\n\n An error in the encoding caused the specification to differ from\n deployed implementations. At the time of this writing there are no\n known implementations that follow the encoding specified in RFC 4507.\n This update to RFC 4507 aligns the document with these currently\n deployed implementations.\n\n Erroneous encoding in RFC 4507 resulted in two length fields; one for\n the extension contents and one for the ticket itself. Hence, for a\n ticket that is 256 bytes long and begins with the hex value FF FF,\n the encoding of the extension would be as follows according to RFC\n 4507:\n\n 00 23 Ticket Extension type 35\n 01 02 Length of extension contents\n 01 00 Length of ticket\n FF FF .. .. Actual ticket\n\n The update proposed in this document reflects what implementations\n actually encode, namely it removes the redundant length field. So,\n for a ticket that is 256 bytes long and begins with the hex value FF\n FF, the encoding of the extension would be as follows according to\n this update:\n\n 00 23 Extension type 35\n 01 00 Length of extension contents (ticket)\n FF FF .. .. Actual ticket\n\n A server implemented according to RFC 4507 receiving a ticket\n extension from a client conforming to this document would interpret\n the first two bytes of the ticket as the length of this ticket. This\n will result in either an inconsistent length field or in the\n processing of a ticket missing the first two bytes. In the first\n case, the server should reject the request based on a malformed\n length. In the second case, the server should reject the ticket\n based on a malformed ticket, incorrect key version, or failed\n decryption. A server implementation based on this update receiving\n an RFC 4507 extension would interpret the first length field as the\n\n\n\n\n\nSalowey, et al. Standards Track [Page 17]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n length of the ticket and include the second two length bytes as the\n first bytes in the ticket, resulting in the ticket being rejected\n based on a malformed ticket, incorrect key version, or failed\n decryption.\n\n Note that the encoding of an empty SessionTicket extension was\n ambiguous in RFC 4507. An RFC 4507 implementation may have encoded\n it as:\n\n 00 23 Extension type 35\n 00 02 Length of extension contents\n 00 00 Length of ticket\n\n or it may have encoded it the same way as this update:\n\n 00 23 Extension type 35\n 00 00 Length of extension contents\n\n A server wishing to support RFC 4507 clients should respond to an\n empty SessionTicket extension encoded the same way as it received it.\n\n A server implementation can construct tickets such that it can detect\n an RFC 4507 implementation, if one existed, by including a cookie at\n the beginning of the tickets that can be differentiated from a valid\n length. For example, if an implementation constructed tickets to\n start with the hex values FF FF, then it could determine where the\n ticket begins and determine the length correctly from the type of\n length fields present.\n\n This document makes a few additional changes to RFC 4507 listed\n below.\n\n o Clarifying that the server can allow session resumption using a\n ticket without issuing a new ticket in Section 3.1.\n\n o Clarifying that the lifetime is relative to when the ticket is\n received in section 3.3.\n\n o Clarifying that the NewSessionTicket handshake message is included\n in the hash generated for the Finished messages in Section 3.3.\n\n o Clarifying the interaction with TLS Session ID in Section 3.4.\n\n o Recommending the use of SHA-256 for the integrity protection of\n the ticket in Section 4.\n\n o Clarifying that additional data can be included in the\n StatePlaintext structure in Section 4.\n\n\n\nSalowey, et al. Standards Track [Page 18]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAuthors' Addresses\n\n Joseph Salowey\n Cisco Systems\n 2901 3rd Ave\n Seattle, WA 98121\n US\n\n EMail: [redacted-email]\n\n\n Hao Zhou\n Cisco Systems\n 4125 Highlander Parkway\n Richfield, OH 44286\n US\n\n EMail: [redacted-email]\n\n\n Pasi Eronen\n Nokia Research Center\n P.O. Box 407\n FIN-00045 Nokia Group\n Finland\n\n EMail: [redacted-email]\n\n\n Hannes Tschofenig\n Nokia Siemens Networks\n Otto-Hahn-Ring 6\n Munich, Bayern 81739\n Germany\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 19]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nFull Copyright Statement\n\n Copyright (C) The IETF Trust (2008).\n\n This document is subject to the rights, licenses and restrictions\n contained in BCP 78, and except as set forth therein, the authors\n retain all their rights.\n\n This document and the information contained herein are provided on an\n \"AS IS\" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\n OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\n THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\n OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\n THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\n WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\n\nIntellectual Property\n\n The IETF takes no position regarding the validity or scope of any\n Intellectual Property Rights or other rights that might be claimed to\n pertain to the implementation or use of the technology described in\n this document or the extent to which any license under such rights\n might or might not be available; nor does it represent that it has\n made any independent effort to identify any such rights. Information\n on the procedures with respect to rights in RFC documents can be\n found in BCP 78 and BCP 79.\n\n Copies of IPR disclosures made to the IETF Secretariat and any\n assurances of licenses to be made available, or the result of an\n attempt made to obtain a general license or permission for the use of\n such proprietary rights by implementers or users of this\n specification can be obtained from the IETF on-line IPR repository at\n http://www.ietf.org/ipr.\n\n The IETF invites any interested party to bring to its attention any\n copyrights, patents or patent applications, or other proprietary\n rights that may cover technology that may be required to implement\n this standard. Please address the information to the IETF at\n ietf-ipr@ietf.org.\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 20]\n\n===== DOCUMENT =====\n\n\nRFC 8446 TLS August 2018\n\n\n Sam Scott\n Royal Holloway, University of London\n [redacted-email]\n\n Thomas Shrimpton\n University of Florida\n [redacted-email]\n\n Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n Brian Smith\n Independent\n [redacted-email]\n\n Brian Sniffen\n Akamai Technologies\n [redacted-email]\n\n Nick Sullivan\n Cloudflare Inc.\n [redacted-email]\n\n Bjoern Tackmann\n University of California, San Diego\n [redacted-email]\n\n Tim Taubert\n Mozilla\n [redacted-email]\n\n Martin Thomson\n Mozilla\n [redacted-email]\n\n Hannes Tschofenig\n Arm Limited\n [redacted-email]\n\n Sean Turner\n sn3rd\n [redacted-email]\n\n Steven Valdez\n Google\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 159]\n\f\nRFC 8446 TLS August 2018\n\n\n Filippo Valsorda\n Cloudflare Inc.\n [redacted-email]\n\n Thyla van der Merwe\n Royal Holloway, University of London\n [redacted-email]\n\n Victor Vasiliev\n Google\n [redacted-email]\n\n Hoeteck Wee\n Ecole Normale Superieure, Paris\n [redacted-email]\n\n Tom Weinstein\n\n David Wong\n NCC Group\n [redacted-email]\n\n Christopher A. Wood\n Apple Inc.\n [redacted-email]\n\n Tim Wright\n Vodafone\n [redacted-email]\n\n Peter Wu\n Independent\n [redacted-email]\n\n Kazu Yamamoto\n Internet Initiative Japan Inc.\n [redacted-email]\n\nAuthor's Address\n\n Eric Rescorla\n Mozilla\n\n Email: [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 160]\n\n===== DOCUMENT =====\n\nNetwork Working Group T. Dierks\nRequest for Comments: 5246 Independent\nObsoletes: 3268, 4346, 4366 E. Rescorla\nUpdates: 4492 RTFM, Inc.\nCategory: Standards Track August 2008\n\n\n The Transport Layer Security (TLS) Protocol\n Version 1.2\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document specifies Version 1.2 of the Transport Layer Security\n (TLS) protocol. The TLS protocol provides communications security\n over the Internet. The protocol allows client/server applications to\n communicate in a way that is designed to prevent eavesdropping,\n tampering, or message forgery.\n\nTable of Contents\n\n 1. Introduction ....................................................4\n 1.1. Requirements Terminology ...................................5\n 1.2. Major Differences from TLS 1.1 .............................5\n 2. Goals ...........................................................6\n 3. Goals of This Document ..........................................7\n 4. Presentation Language ...........................................7\n 4.1. Basic Block Size ...........................................7\n 4.2. Miscellaneous ..............................................8\n 4.3. Vectors ....................................................8\n 4.4. Numbers ....................................................9\n 4.5. Enumerateds ................................................9\n 4.6. Constructed Types .........................................10\n 4.6.1. Variants ...........................................10\n 4.7. Cryptographic Attributes ..................................12\n 4.8. Constants .................................................14\n 5. HMAC and the Pseudorandom Function .............................14\n 6. The TLS Record Protocol ........................................15\n 6.1. Connection States .........................................16\n 6.2. Record Layer ..............................................19\n 6.2.1. Fragmentation ......................................19\n\n\n\nDierks & Rescorla Standards Track [Page 1]\n\f\nRFC 5246 TLS August 2008\n\n\n 6.2.2. Record Compression and Decompression ...............20\n 6.2.3. Record Payload Protection ..........................21\n 6.2.3.1. Null or Standard Stream Cipher ............22\n 6.2.3.2. CBC Block Cipher ..........................22\n 6.2.3.3. AEAD Ciphers ..............................24\n 6.3. Key Calculation ...........................................25\n 7. The TLS Handshaking Protocols ..................................26\n 7.1. Change Cipher Spec Protocol ...............................27\n 7.2. Alert Protocol ............................................28\n 7.2.1. Closure Alerts .....................................29\n 7.2.2. Error Alerts .......................................30\n 7.3. Handshake Protocol Overview ...............................33\n 7.4. Handshake Protocol ........................................37\n 7.4.1. Hello Messages .....................................38\n 7.4.1.1. Hello Request .............................38\n 7.4.1.2. Client Hello ..............................39\n 7.4.1.3. Server Hello ..............................42\n 7.4.1.4. Hello Extensions ..........................44\n 7.4.1.4.1. Signature Algorithms ...........45\n 7.4.2. Server Certificate .................................47\n 7.4.3. Server Key Exchange Message ........................50\n 7.4.4. Certificate Request ................................53\n 7.4.5. Server Hello Done ..................................55\n 7.4.6. Client Certificate .................................55\n 7.4.7. Client Key Exchange Message ........................57\n 7.4.7.1. RSA-Encrypted Premaster Secret Message ....58\n 7.4.7.2. Client Diffie-Hellman Public Value ........61\n 7.4.8. Certificate Verify .................................62\n 7.4.9. Finished ...........................................63\n 8. Cryptographic Computations .....................................64\n 8.1. Computing the Master Secret ...............................64\n 8.1.1. RSA ................................................65\n 8.1.2. Diffie-Hellman .....................................65\n 9. Mandatory Cipher Suites ........................................65\n 10. Application Data Protocol .....................................65\n 11. Security Considerations .......................................65\n 12. IANA Considerations ...........................................65\n Appendix A. Protocol Data Structures and Constant Values ..........68\n A.1. Record Layer ..............................................68\n A.2. Change Cipher Specs Message ...............................69\n A.3. Alert Messages ............................................69\n A.4. Handshake Protocol ........................................70\n A.4.1. Hello Messages .....................................71\n A.4.2. Server Authentication and Key Exchange Messages ....72\n A.4.3. Client Authentication and Key Exchange Messages ....74\n A.4.4. Handshake Finalization Message .....................74\n A.5. The Cipher Suite ..........................................75\n A.6. The Security Parameters ...................................77\n\n\n\nDierks & Rescorla Standards Track [Page 2]\n\f\nRFC 5246 TLS August 2008\n\n\n A.7. Changes to RFC 4492 .......................................78\n Appendix B. Glossary ..............................................78\n Appendix C. Cipher Suite Definitions ..............................83\n Appendix D. Implementation Notes ..................................85\n D.1. Random Number Generation and Seeding ......................85\n D.2. Certificates and Authentication ...........................85\n D.3. Cipher Suites .............................................85\n D.4. Implementation Pitfalls ...................................85\n Appendix E. Backward Compatibility ................................87\n E.1. Compatibility with TLS 1.0/1.1 and SSL 3.0 ................87\n E.2. Compatibility with SSL 2.0 ................................88\n E.3. Avoiding Man-in-the-Middle Version Rollback ...............90\n Appendix F. Security Analysis .....................................91\n F.1. Handshake Protocol ........................................91\n F.1.1. Authentication and Key Exchange ....................91\n F.1.1.1. Anonymous Key Exchange ....................91\n F.1.1.2. RSA Key Exchange and Authentication .......92\n F.1.1.3. Diffie-Hellman Key Exchange with\n Authentication ............................92\n F.1.2. Version Rollback Attacks ...........................93\n F.1.3. Detecting Attacks Against the Handshake Protocol ...94\n F.1.4. Resuming Sessions ..................................94\n F.2. Protecting Application Data ...............................94\n F.3. Explicit IVs ..............................................95\n F.4. Security of Composite Cipher Modes ........................95\n F.5. Denial of Service .........................................96\n F.6. Final Notes ...............................................96\n Normative References ..............................................97\n Informative References ............................................98\n Working Group Information ........................................101\n Contributors .....................................................101\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 3]\n\f\nRFC 5246 TLS August 2008\n\n\n1. Introduction\n\n The primary goal of the TLS protocol is to provide privacy and data\n integrity between two communicating applications. The protocol is\n composed of two layers: the TLS Record Protocol and the TLS Handshake\n Protocol. At the lowest level, layered on top of some reliable\n transport protocol (e.g., TCP [TCP]), is the TLS Record Protocol.\n The TLS Record Protocol provides connection security that has two\n basic properties:\n\n - The connection is private. Symmetric cryptography is used for\n data encryption (e.g., AES [AES], RC4 [SCH], etc.). The keys for\n this symmetric encryption are generated uniquely for each\n connection and are based on a secret negotiated by another\n protocol (such as the TLS Handshake Protocol). The Record\n Protocol can also be used without encryption.\n\n - The connection is reliable. Message transport includes a message\n integrity check using a keyed MAC. Secure hash functions (e.g.,\n SHA-1, etc.) are used for MAC computations. The Record Protocol\n can operate without a MAC, but is generally only used in this mode\n while another protocol is using the Record Protocol as a transport\n for negotiating security parameters.\n\n The TLS Record Protocol is used for encapsulation of various higher-\n level protocols. One such encapsulated protocol, the TLS Handshake\n Protocol, allows the server and client to authenticate each other and\n to negotiate an encryption algorithm and cryptographic keys before\n the application protocol transmits or receives its first byte of\n data. The TLS Handshake Protocol provides connection security that\n has three basic properties:\n\n - The peer's identity can be authenticated using asymmetric, or\n public key, cryptography (e.g., RSA [RSA], DSA [DSS], etc.). This\n authentication can be made optional, but is generally required for\n at least one of the peers.\n\n - The negotiation of a shared secret is secure: the negotiated\n secret is unavailable to eavesdroppers, and for any authenticated\n connection the secret cannot be obtained, even by an attacker who\n can place himself in the middle of the connection.\n\n - The negotiation is reliable: no attacker can modify the\n negotiation communication without being detected by the parties to\n the communication.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 4]\n\f\nRFC 5246 TLS August 2008\n\n\n One advantage of TLS is that it is application protocol independent.\n Higher-level protocols can layer on top of the TLS protocol\n transparently. The TLS standard, however, does not specify how\n protocols add security with TLS; the decisions on how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n1.1. Requirements Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [REQ].\n\n1.2. Major Differences from TLS 1.1\n\n This document is a revision of the TLS 1.1 [TLS1.1] protocol which\n contains improved flexibility, particularly for negotiation of\n cryptographic algorithms. The major changes are:\n\n - The MD5/SHA-1 combination in the pseudorandom function (PRF) has\n been replaced with cipher-suite-specified PRFs. All cipher suites\n in this document use P_SHA256.\n\n - The MD5/SHA-1 combination in the digitally-signed element has been\n replaced with a single hash. Signed elements now include a field\n that explicitly specifies the hash algorithm used.\n\n - Substantial cleanup to the client's and server's ability to\n specify which hash and signature algorithms they will accept.\n Note that this also relaxes some of the constraints on signature\n and hash algorithms from previous versions of TLS.\n\n - Addition of support for authenticated encryption with additional\n data modes.\n\n - TLS Extensions definition and AES Cipher Suites were merged in\n from external [TLSEXT] and [TLSAES].\n\n - Tighter checking of EncryptedPreMasterSecret version numbers.\n\n - Tightened up a number of requirements.\n\n - Verify_data length now depends on the cipher suite (default is\n still 12).\n\n - Cleaned up description of Bleichenbacher/Klima attack defenses.\n\n\n\n\nDierks & Rescorla Standards Track [Page 5]\n\f\nRFC 5246 TLS August 2008\n\n\n - Alerts MUST now be sent in many cases.\n\n - After a certificate_request, if no certificates are available,\n clients now MUST send an empty certificate list.\n\n - TLS_RSA_WITH_AES_128_CBC_SHA is now the mandatory to implement\n cipher suite.\n\n - Added HMAC-SHA256 cipher suites.\n\n - Removed IDEA and DES cipher suites. They are now deprecated and\n will be documented in a separate document.\n\n - Support for the SSLv2 backward-compatible hello is now a MAY, not\n a SHOULD, with sending it a SHOULD NOT. Support will probably\n become a SHOULD NOT in the future.\n\n - Added limited \"fall-through\" to the presentation language to allow\n multiple case arms to have the same encoding.\n\n - Added an Implementation Pitfalls sections\n\n - The usual clarifications and editorial work.\n\n2. Goals\n\n The goals of the TLS protocol, in order of priority, are as follows:\n\n 1. Cryptographic security: TLS should be used to establish a secure\n connection between two parties.\n\n 2. Interoperability: Independent programmers should be able to\n develop applications utilizing TLS that can successfully exchange\n cryptographic parameters without knowledge of one another's code.\n\n 3. Extensibility: TLS seeks to provide a framework into which new\n public key and bulk encryption methods can be incorporated as\n necessary. This will also accomplish two sub-goals: preventing\n the need to create a new protocol (and risking the introduction of\n possible new weaknesses) and avoiding the need to implement an\n entire new security library.\n\n 4. Relative efficiency: Cryptographic operations tend to be highly\n CPU intensive, particularly public key operations. For this\n reason, the TLS protocol has incorporated an optional session\n caching scheme to reduce the number of connections that need to be\n established from scratch. Additionally, care has been taken to\n reduce network activity.\n\n\n\nDierks & Rescorla Standards Track [Page 6]\n\f\nRFC 5246 TLS August 2008\n\n\n3. Goals of This Document\n\n This document and the TLS protocol itself are based on the SSL 3.0\n Protocol Specification as published by Netscape. The differences\n between this protocol and SSL 3.0 are not dramatic, but they are\n significant enough that the various versions of TLS and SSL 3.0 do\n not interoperate (although each protocol incorporates a mechanism by\n which an implementation can back down to prior versions). This\n document is intended primarily for readers who will be implementing\n the protocol and for those doing cryptographic analysis of it. The\n specification has been written with this in mind, and it is intended\n to reflect the needs of those two groups. For that reason, many of\n the algorithm-dependent data structures and rules are included in the\n body of the text (as opposed to in an appendix), providing easier\n access to them.\n\n This document is not intended to supply any details of service\n definition or of interface definition, although it does cover select\n areas of policy as they are required for the maintenance of solid\n security.\n\n4. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used. The syntax draws from\n several sources in its structure. Although it resembles the\n programming language \"C\" in its syntax and XDR [XDR] in both its\n syntax and intent, it would be risky to draw too many parallels. The\n purpose of this presentation language is to document TLS only; it has\n no general application beyond that particular goal.\n\n4.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 7]\n\f\nRFC 5246 TLS August 2008\n\n\n4.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" double\n brackets.\n\n Single-byte entities containing uninterpreted data are of type\n opaque.\n\n4.3. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* 3 consecutive 3 byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n In the following example, mandatory is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 4.4). On the\n other hand, longer can represent up to 800 bytes of data, or 400\n uint16 elements, and it may be empty. Its encoding will include a\n\n\n\nDierks & Rescorla Standards Track [Page 8]\n\f\nRFC 5246 TLS August 2008\n\n\n two-byte actual length field prepended to the vector. The length of\n an encoded vector must be an even multiple of the length of a single\n element (for example, a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is 2 bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n4.4. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are formed from fixed-length series of bytes\n concatenated as described in Section 4.1 and are also unsigned. The\n following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are stored in\n network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n Note that in some cases (e.g., DH parameters) it is necessary to\n represent integers as opaque vectors. In such cases, they are\n represented as unsigned integers (i.e., leading zero octets are not\n required even if the most significant bit is set).\n\n4.5. Enumerateds\n\n An additional sparse data type is available called enum. A field of\n type enum can only assume the values declared in the definition.\n Each definition is a different type. Only enumerateds of the same\n type may be assigned or compared. Every element of an enumerated\n must be assigned a value, as demonstrated in the following example.\n Since the elements of the enumerated are not ordered, they can be\n assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n\n\n\nDierks & Rescorla Standards Track [Page 9]\n\f\nRFC 5246 TLS August 2008\n\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n For enumerateds that are never converted to external representation,\n the numerical information may be omitted.\n\n enum { low, medium, high } Amount;\n\n4.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax for definition is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } [[T]];\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n Structure definitions may be embedded.\n\n4.6.1. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. There\n must be a case arm for every element of the enumeration declared in\n the select. Case arms have limited fall-through: if two case arms\n follow in immediate succession with no fields in between, then they\n\n\n\nDierks & Rescorla Standards Track [Page 10]\n\f\nRFC 5246 TLS August 2008\n\n\n both contain the same fields. Thus, in the example below, \"orange\"\n and \"banana\" both contain V2. Note that this is a new piece of\n syntax in TLS 1.2.\n\n The body of the variant structure may be given a label for reference.\n The mechanism by which the variant is selected at runtime is not\n prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1;\n case e2: Te2;\n case e3: case e4: Te3;\n ....\n case en: Ten;\n } [[fv]];\n } [[Tv]];\n\n For example:\n\n enum { apple, orange, banana } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n select (VariantTag) { /* value of selector is implicit */\n case apple:\n V1; /* VariantBody, tag = apple */\n case orange:\n case banana:\n V2; /* VariantBody, tag = orange or banana */\n } variant_body; /* optional label on variant */\n } VariantRecord;\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 11]\n\f\nRFC 5246 TLS August 2008\n\n\n4.7. Cryptographic Attributes\n\n The five cryptographic operations -- digital signing, stream cipher\n encryption, block cipher encryption, authenticated encryption with\n additional data (AEAD) encryption, and public key encryption -- are\n designated digitally-signed, stream-ciphered, block-ciphered, aead-\n ciphered, and public-key-encrypted, respectively. A field's\n cryptographic processing is specified by prepending an appropriate\n key word designation before the field's type specification.\n Cryptographic keys are implied by the current session state (see\n Section 6.1).\n\n A digitally-signed element is encoded as a struct DigitallySigned:\n\n struct {\n SignatureAndHashAlgorithm algorithm;\n opaque signature<0..2^16-1>;\n } DigitallySigned;\n\n The algorithm field specifies the algorithm used (see Section\n 7.4.1.4.1 for the definition of this field). Note that the\n introduction of the algorithm field is a change from previous\n versions. The signature is a digital signature using those\n algorithms over the contents of the element. The contents themselves\n do not appear on the wire but are simply calculated. The length of\n the signature is specified by the signing algorithm and key.\n\n In RSA signing, the opaque vector contains the signature generated\n using the RSASSA-PKCS1-v1_5 signature scheme defined in [PKCS1]. As\n discussed in [PKCS1], the DigestInfo MUST be DER-encoded [X680]\n [X690]. For hash algorithms without parameters (which includes\n SHA-1), the DigestInfo.AlgorithmIdentifier.parameters field MUST be\n NULL, but implementations MUST accept both without parameters and\n with NULL parameters. Note that earlier versions of TLS used a\n different RSA signature scheme that did not include a DigestInfo\n encoding.\n\n In DSA, the 20 bytes of the SHA-1 hash are run directly through the\n Digital Signing Algorithm with no additional hashing. This produces\n two values, r and s. The DSA signature is an opaque vector, as\n above, the contents of which are the DER encoding of:\n\n Dss-Sig-Value ::= SEQUENCE {\n r INTEGER,\n s INTEGER\n }\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 12]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: In current terminology, DSA refers to the Digital Signature\n Algorithm and DSS refers to the NIST standard. In the original SSL\n and TLS specs, \"DSS\" was used universally. This document uses \"DSA\"\n to refer to the algorithm, \"DSS\" to refer to the standard, and it\n uses \"DSS\" in the code point definitions for historical continuity.\n\n In stream cipher encryption, the plaintext is exclusive-ORed with an\n identical amount of output generated from a cryptographically secure\n keyed pseudorandom number generator.\n\n In block cipher encryption, every block of plaintext encrypts to a\n block of ciphertext. All block cipher encryption is done in CBC\n (Cipher Block Chaining) mode, and all items that are block-ciphered\n will be an exact multiple of the cipher block length.\n\n In AEAD encryption, the plaintext is simultaneously encrypted and\n integrity protected. The input may be of any length, and aead-\n ciphered output is generally larger than the input in order to\n accommodate the integrity check value.\n\n In public key encryption, a public key algorithm is used to encrypt\n data in such a way that it can be decrypted only with the matching\n private key. A public-key-encrypted element is encoded as an opaque\n vector <0..2^16-1>, where the length is specified by the encryption\n algorithm and key.\n\n RSA encryption is done using the RSAES-PKCS1-v1_5 encryption scheme\n defined in [PKCS1].\n\n In the following example\n\n stream-ciphered struct {\n uint8 field1;\n uint8 field2;\n digitally-signed opaque {\n uint8 field3<0..255>;\n uint8 field4;\n };\n } UserType;\n\n The contents of the inner struct (field3 and field4) are used as\n input for the signature/hash algorithm, and then the entire structure\n is encrypted with a stream cipher. The length of this structure, in\n bytes, would be equal to two bytes for field1 and field2, plus two\n bytes for the signature and hash algorithm, plus two bytes for the\n length of the signature, plus the length of the output of the signing\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 13]\n\f\nRFC 5246 TLS August 2008\n\n\n algorithm. The length of the signature is known because the\n algorithm and key used for the signing are known prior to encoding or\n decoding this structure.\n\n4.8. Constants\n\n Typed constants can be defined for purposes of specification by\n declaring a symbol of the desired type and assigning values to it.\n\n Under-specified types (opaque, variable-length vectors, and\n structures that contain opaque) cannot be assigned values. No fields\n of a multi-element structure or vector may be elided.\n\n For example:\n\n struct {\n uint8 f1;\n uint8 f2;\n } Example1;\n\n Example1 ex1 = {1, 4}; /* assigns f1 = 1, f2 = 4 */\n\n5. HMAC and the Pseudorandom Function\n\n The TLS record layer uses a keyed Message Authentication Code (MAC)\n to protect message integrity. The cipher suites defined in this\n document use a construction known as HMAC, described in [HMAC], which\n is based on a hash function. Other cipher suites MAY define their\n own MAC constructions, if needed.\n\n In addition, a construction is required to do expansion of secrets\n into blocks of data for the purposes of key generation or validation.\n This pseudorandom function (PRF) takes as input a secret, a seed, and\n an identifying label and produces an output of arbitrary length.\n\n In this section, we define one PRF, based on HMAC. This PRF with the\n SHA-256 hash function is used for all cipher suites defined in this\n document and in TLS documents published prior to this document when\n TLS 1.2 is negotiated. New cipher suites MUST explicitly specify a\n PRF and, in general, SHOULD use the TLS PRF with SHA-256 or a\n stronger standard hash function.\n\n First, we define a data expansion function, P_hash(secret, data),\n that uses a single hash function to expand a secret and seed into an\n arbitrary quantity of output:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 14]\n\f\nRFC 5246 TLS August 2008\n\n\n P_hash(secret, seed) = HMAC_hash(secret, A(1) + seed) +\n HMAC_hash(secret, A(2) + seed) +\n HMAC_hash(secret, A(3) + seed) + ...\n\n where + indicates concatenation.\n\n A() is defined as:\n\n A(0) = seed\n A(i) = HMAC_hash(secret, A(i-1))\n\n P_hash can be iterated as many times as necessary to produce the\n required quantity of data. For example, if P_SHA256 is being used to\n create 80 bytes of data, it will have to be iterated three times\n (through A(3)), creating 96 bytes of output data; the last 16 bytes\n of the final iteration will then be discarded, leaving 80 bytes of\n output data.\n\n TLS's PRF is created by applying P_hash to the secret as:\n\n PRF(secret, label, seed) = P_(secret, label + seed)\n\n The label is an ASCII string. It should be included in the exact\n form it is given without a length byte or trailing null character.\n For example, the label \"slithy toves\" would be processed by hashing\n the following bytes:\n\n 73 6C 69 74 68 79 20 74 6F 76 65 73\n\n6. The TLS Record Protocol\n\n The TLS Record Protocol is a layered protocol. At each layer,\n messages may include fields for length, description, and content.\n The Record Protocol takes messages to be transmitted, fragments the\n data into manageable blocks, optionally compresses the data, applies\n a MAC, encrypts, and transmits the result. Received data is\n decrypted, verified, decompressed, reassembled, and then delivered to\n higher-level clients.\n\n Four protocols that use the record protocol are described in this\n document: the handshake protocol, the alert protocol, the change\n cipher spec protocol, and the application data protocol. In order to\n allow extension of the TLS protocol, additional record content types\n can be supported by the record protocol. New record content type\n values are assigned by IANA in the TLS Content Type Registry as\n described in Section 12.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 15]\n\f\nRFC 5246 TLS August 2008\n\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST send an\n unexpected_message alert.\n\n Any protocol designed for use over TLS must be carefully designed to\n deal with all possible attacks against it. As a practical matter,\n this means that the protocol designer must be aware of what security\n properties TLS does and does not provide and cannot safely rely on\n the latter.\n\n Note in particular that type and length of a record are not protected\n by encryption. If this information is itself sensitive, application\n designers may wish to take steps (padding, cover traffic) to minimize\n information leakage.\n\n6.1. Connection States\n\n\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 8446 Mozilla\nObsoletes: 5077, 5246, 6961 August 2018\nUpdates: 5705, 6066\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS) Protocol Version 1.3\n\nAbstract\n\n This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol. TLS allows client/server applications to communicate\n over the Internet in a way that is designed to prevent eavesdropping,\n tampering, and message forgery.\n\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077,\n 5246, and 6961. This document also specifies new requirements for\n TLS 1.2 implementations.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 7841.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n https://www.rfc-editor.org/info/rfc8446.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 8446 TLS August 2018\n\n\nCopyright Notice\n\n Copyright (c) 2018 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (https://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 8446 TLS August 2018\n\n\nTable of Contents\n\n 1. Introduction ....................................................6\n 1.1. Conventions and Terminology ................................7\n 1.2. Major Differences from TLS 1.2 .............................8\n 1.3. Updates Affecting TLS 1.2 ..................................9\n 2. Protocol Overview ..............................................10\n 2.1. Incorrect DHE Share .......................................14\n 2.2. Resumption and Pre-Shared Key (PSK) .......................15\n 2.3. 0-RTT Data ................................................17\n 3. Presentation Language ..........................................19\n 3.1. Basic Block Size ..........................................19\n 3.2. Miscellaneous .............................................20\n 3.3. Numbers ...................................................20\n 3.4. Vectors ...................................................20\n 3.5. Enumerateds ...............................................21\n 3.6. Constructed Types .........................................22\n 3.7. Constants .................................................23\n 3.8. Variants ..................................................23\n 4. Handshake Protocol .............................................24\n 4.1. Key Exchange Messages .....................................25\n 4.1.1. Cryptographic Negotiation ..........................26\n 4.1.2. Client Hello .......................................27\n 4.1.3. Server Hello .......................................31\n 4.1.4. Hello Retry Request ................................33\n 4.2. Extensions ................................................35\n 4.2.1. Supported Versions .................................39\n 4.2.2. Cookie .............................................40\n 4.2.3. Signature Algorithms ...............................41\n 4.2.4. Certificate Authorities ............................45\n 4.2.5. OID Filters ........................................45\n 4.2.6. Post-Handshake Client Authentication ...............47\n 4.2.7. Supported Groups ...................................47\n 4.2.8. Key Share ..........................................48\n 4.2.9. Pre-Shared Key Exchange Modes ......................51\n 4.2.10. Early Data Indication .............................52\n 4.2.11. Pre-Shared Key Extension ..........................55\n 4.3. Server Parameters .........................................59\n 4.3.1. Encrypted Extensions ...............................60\n 4.3.2. Certificate Request ................................60\n 4.4. Authentication Messages ...................................61\n 4.4.1. The Transcript Hash ................................63\n 4.4.2. Certificate ........................................64\n 4.4.3. Certificate Verify .................................69\n 4.4.4. Finished ...........................................71\n 4.5. End of Early Data .........................................72\n\n\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 8446 TLS August 2018\n\n\n 4.6. Post-Handshake Messages ...................................73\n 4.6.1. New Session Ticket Message .........................73\n 4.6.2. Post-Handshake Authentication ......................75\n 4.6.3. Key and Initialization Vector Update ...............76\n 5. Record Protocol ................................................77\n 5.1. Record Layer ..............................................78\n 5.2. Record Payload Protection .................................80\n 5.3. Per-Record Nonce ..........................................82\n 5.4. Record Padding ............................................83\n 5.5. Limits on Key Usage .......................................84\n 6. Alert Protocol .................................................85\n 6.1. Closure Alerts ............................................87\n 6.2. Error Alerts ..............................................88\n 7. Cryptographic Computations .....................................90\n 7.1. Key Schedule ..............................................91\n 7.2. Updating Traffic Secrets ..................................94\n 7.3. Traffic Key Calculation ...................................95\n 7.4. (EC)DHE Shared Secret Calculation .........................95\n 7.4.1. Finite Field Diffie-Hellman ........................95\n 7.4.2. Elliptic Curve Diffie-Hellman ......................96\n 7.5. Exporters .................................................97\n 8. 0-RTT and Anti-Replay ..........................................98\n 8.1. Single-Use Tickets ........................................99\n 8.2. Client Hello Recording ....................................99\n 8.3. Freshness Checks .........................................101\n 9. Compliance Requirements .......................................102\n 9.1. Mandatory-to-Implement Cipher Suites .....................102\n 9.2. Mandatory-to-Implement Extensions ........................103\n 9.3. Protocol Invariants ......................................104\n 10. Security Considerations ......................................106\n 11. IANA Considerations ..........................................106\n 12. References ...................................................109\n 12.1. Normative References ....................................109\n 12.2. Informative References ..................................112\n Appendix A. State Machine ........................................120\n A.1. Client ....................................................120\n A.2. Server ....................................................121\n Appendix B. Protocol Data Structures and Constant Values .........122\n B.1. Record Layer ..............................................122\n B.2. Alert Messages ............................................123\n B.3. Handshake Protocol ........................................124\n B.3.1. Key Exchange Messages .................................125\n B.3.2. Server Parameters Messages ............................131\n B.3.3. Authentication Messages ...............................132\n B.3.4. Ticket Establishment ..................................132\n B.3.5. Updating Keys .........................................133\n B.4. Cipher Suites .............................................133\n\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 8446 TLS August 2018\n\n\n Appendix C. Implementation Notes .................................134\n C.1. Random Number Generation and Seeding ......................134\n C.2. Certificates and Authentication ...........................135\n C.3. Implementation Pitfalls ...................................135\n C.4. Client Tracking Prevention ................................137\n C.5. Unauthenticated Operation .................................137\n Appendix D. Backward Compatibility ...............................138\n D.1. Negotiating with an Older Server ..........................139\n D.2. Negotiating with an Older Client ..........................139\n D.3. 0-RTT Backward Compatibility ..............................140\n D.4. Middlebox Compatibility Mode ..............................140\n D.5. Security Restrictions Related to Backward Compatibility ...141\n Appendix E. Overview of Security Properties ......................142\n E.1. Handshake .................................................142\n E.1.1. Key Derivation and HKDF ...............................145\n E.1.2. Client Authentication .................................146\n E.1.3. 0-RTT .................................................146\n E.1.4. Exporter Independence .................................146\n E.1.5. Post-Compromise Security ..............................146\n E.1.6. External References ...................................147\n E.2. Record Layer ..............................................147\n E.2.1. External References ...................................148\n E.3. Traffic Analysis ..........................................148\n E.4. Side-Channel Attacks ......................................149\n E.5. Replay Attacks on 0-RTT ...................................150\n E.5.1. Replay and Exporters ..................................151\n E.6. PSK Identity Exposure .....................................152\n E.7. Sharing PSKs ..............................................152\n E.8. Attacks on Static RSA .....................................152\n Contributors .....................................................153\n Author's Address .................................................160\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 8446 TLS August 2018\n\n\n1. Introduction\n\n The primary goal of TLS is to provide a secure channel between two\n communicating peers; the only requirement from the underlying\n transport is a reliable, in-order data stream. Specifically, the\n secure channel should provide the following properties:\n\n - Authentication: The server side of the channel is always\n authenticated; the client side is optionally authenticated.\n Authentication can happen via asymmetric cryptography (e.g., RSA\n [RSA], the Elliptic Curve Digital Signature Algorithm (ECDSA)\n [ECDSA], or the Edwards-Curve Digital Signature Algorithm (EdDSA)\n [RFC8032]) or a symmetric pre-shared key (PSK).\n\n - Confidentiality: Data sent over the channel after establishment is\n only visible to the endpoints. TLS does not hide the length of\n the data it transmits, though endpoints are able to pad TLS\n records in order to obscure lengths and improve protection against\n traffic analysis techniques.\n\n - Integrity: Data sent over the channel after establishment cannot\n be modified by attackers without detection.\n\n These properties should be true even in the face of an attacker who\n has complete control of the network, as described in [RFC3552]. See\n Appendix E for a more complete statement of the relevant security\n properties.\n\n TLS consists of two primary components:\n\n - A handshake protocol (Section 4) that authenticates the\n communicating parties, negotiates cryptographic modes and\n parameters, and establishes shared keying material. The handshake\n protocol is designed to resist tampering; an active attacker\n should not be able to force the peers to negotiate different\n parameters than they would if the connection were not under\n attack.\n\n - A record protocol (Section 5) that uses the parameters established\n by the handshake protocol to protect traffic between the\n communicating peers. The record protocol divides traffic up into\n a series of records, each of which is independently protected\n using the traffic keys.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 8446 TLS August 2018\n\n\n TLS is application protocol independent; higher-level protocols can\n layer on top of TLS transparently. The TLS standard, however, does\n not specify how protocols add security with TLS; how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n This document defines TLS version 1.3. While TLS 1.3 is not directly\n compatible with previous versions, all versions of TLS incorporate a\n versioning mechanism which allows clients and servers to\n interoperably negotiate a common version if one is supported by both\n peers.\n\n This document supersedes and obsoletes previous versions of TLS,\n including version 1.2 [RFC5246]. It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2. Because TLS 1.3 changes the way keys are\n derived, it updates [RFC5705] as described in Section 7.5. It also\n changes how Online Certificate Status Protocol (OCSP) messages are\n carried and therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in Section 4.4.2.1.\n\n1.1. Conventions and Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all\n capitals, as shown here.\n\n The following terms are used:\n\n client: The endpoint initiating the TLS connection.\n\n connection: A transport-layer connection between two endpoints.\n\n endpoint: Either the client or server of the connection.\n\n handshake: An initial negotiation between client and server that\n establishes the parameters of their subsequent interactions\n within TLS.\n\n peer: An endpoint. When discussing a particular endpoint, \"peer\"\n refers to the endpoint that is not the primary subject of\n discussion.\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\f\nRFC 8446 TLS August 2018\n\n\n receiver: An endpoint that is receiving records.\n\n sender: An endpoint that is transmitting records.\n\n server: The endpoint that did not initiate the TLS connection.\n\n1.2. Major Differences from TLS 1.2\n\n The following is a list of the major functional differences between\n TLS 1.2 and TLS 1.3. It is not intended to be exhaustive, and there\n are many minor differences.\n\n - The list of supported symmetric encryption algorithms has been\n pruned of all algorithms that are considered legacy. Those that\n remain are all Authenticated Encryption with Associated Data\n (AEAD) algorithms. The cipher suite concept has been changed to\n separate the authentication and key exchange mechanisms from the\n record protection algorithm (including secret key length) and a\n hash to be used with both the key derivation function and\n handshake message authentication code (MAC).\n\n - A zero round-trip time (0-RTT) mode was added, saving a round trip\n at connection setup for some application data, at the cost of\n certain security properties.\n\n - Static RSA and Diffie-Hellman cipher suites have been removed; all\n public-key based key exchange mechanisms now provide forward\n secrecy.\n\n - All handshake messages after the ServerHello are now encrypted.\n The newly introduced EncryptedExtensions message allows various\n extensions previously sent in the clear in the ServerHello to also\n enjoy confidentiality protection.\n\n - The key derivation functions have been redesigned. The new design\n allows easier analysis by cryptographers due to their improved key\n separation properties. The HMAC-based Extract-and-Expand Key\n Derivation Function (HKDF) is used as an underlying primitive.\n\n - The handshake state machine has been significantly restructured to\n be more consistent and to remove superfluous messages such as\n ChangeCipherSpec (except when needed for middlebox compatibility).\n\n - Elliptic curve algorithms are now in the base spec, and new\n signature algorithms, such as EdDSA, are included. TLS 1.3\n removed point format negotiation in favor of a single point format\n for each curve.\n\n\n\n\nRescorla Standards Track [Page 8]\n\f\nRFC 8446 TLS August 2018\n\n\n - Other cryptographic improvements were made, including changing the\n RSA padding to use the RSA Probabilistic Signature Scheme\n (RSASSA-PSS), and the removal of compression, the Digital\n Signature Algorithm (DSA), and custom Ephemeral Diffie-Hellman\n (DHE) groups.\n\n - The TLS 1.2 version negotiation mechanism has been deprecated in\n favor of a version list in an extension. This increases\n compatibility with existing servers that incorrectly implemented\n version negotiation.\n\n - Session resumption with and without server-side state as well as\n the PSK-based cipher suites of earlier TLS versions have been\n replaced by a single new PSK exchange.\n\n - References have been updated to point to the updated versions of\n RFCs, as appropriate (e.g., RFC 5280 rather than RFC 3280).\n\n1.3. Updates Affecting TLS 1.2\n\n This document defines several changes that optionally affect\n implementations of TLS 1.2, including those which do not also support\n TLS 1.3:\n\n - A version downgrade protection mechanism is described in\n Section 4.1.3.\n\n - RSASSA-PSS signature schemes are defined in Section 4.2.3.\n\n - The \"supported_versions\" ClientHello extension can be used to\n negotiate the version of TLS to use, in preference to the\n legacy_version field of the ClientHello.\n\n - The \"signature_algorithms_cert\" extension allows a client to\n indicate which signature algorithms it can validate in X.509\n certificates.\n\n Additionally, this document clarifies some compliance requirements\n for earlier versions of TLS; see Section 9.3.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 9]\n\f\nRFC 8446 TLS August 2018\n\n\n2. Protocol Overview\n\n The cryptographic parameters used by the secure channel are produced\n by the TLS handshake protocol. This sub-protocol of TLS is used by\n the client and server when first communicating with each other. The\n handshake protocol allows peers to negotiate a protocol version,\n select cryptographic algorithms, optionally authenticate each other,\n and establish shared secret keying material. Once the handshake is\n complete, the peers use the established keys to protect the\n application-layer traffic.\n\n A failure of the handshake or other protocol error triggers the\n termination of the connection, optionally preceded by an alert\n message (Section 6).\n\n TLS supports three basic key exchange modes:\n\n - (EC)DHE (Diffie-Hellman over either finite fields or elliptic\n curves)\n\n - PSK-only\n\n - PSK with (EC)DHE\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 10]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 1 below shows the basic full TLS handshake:\n\n Client Server\n\nKey ^ ClientHello\nExch | + key_share*\n | + signature_algorithms*\n | + psk_key_exchange_modes*\n v + pre_shared_key* -------->\n ServerHello ^ Key\n + key_share* | Exch\n + pre_shared_key* v\n {EncryptedExtensions} ^ Server\n {CertificateRequest*} v Params\n {Certificate*} ^\n {CertificateVerify*} | Auth\n {Finished} v\n <-------- [Application Data*]\n ^ {Certificate*}\nAuth | {CertificateVerify*}\n v {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 1: Message Flow for Full TLS Handshake\n\n The handshake can be thought of as having three phases (indicated in\n the diagram above):\n\n - Key Exchange: Establish shared keying material and select the\n cryptographic parameters. Everything after this phase is\n encrypted.\n\n - Server Parameters: Establish other handshake parameters\n (whether the client is authenticated, application-layer protocol\n support, etc.).\n\n\n\n\nRescorla Standards Track [Page 11]\n\f\nRFC 8446 TLS August 2018\n\n\n - Authentication: Authenticate the server (and, optionally, the\n client) and provide key confirmation and handshake integrity.\n\n In the Key Exchange phase, the client sends the ClientHello\n (Section 4.1.2) message, which contains a random nonce\n (ClientHello.random); its offered protocol versions; a list of\n symmetric cipher/HKDF hash pairs; either a set of Diffie-Hellman key\n shares (in the \"key_share\" (Section 4.2.8) extension), a set of\n pre-shared key labels (in the \"pre_shared_key\" (Section 4.2.11)\n extension), or both; and potentially additional extensions.\n Additional fields and/or messages may also be present for middlebox\n compatibility.\n\n The server processes the ClientHello and determines the appropriate\n cryptographic parameters for the connection. It then responds with\n its own ServerHello (Section 4.1.3), which indicates the negotiated\n connection parameters. The combination of the ClientHello and the\n ServerHello determines the shared keys. If (EC)DHE key establishment\n is in use, then the ServerHello contains a \"key_share\" extension with\n the server's ephemeral Diffie-Hellman share; the server's share MUST\n be in the same group as one of the client's shares. If PSK key\n establishment is in use, then the ServerHello contains a\n \"pre_shared_key\" extension indicating which of the client's offered\n PSKs was selected. Note that implementations can use (EC)DHE and PSK\n together, in which case both extensions will be supplied.\n\n The server then sends two messages to establish the Server\n Parameters:\n\n EncryptedExtensions: responses to ClientHello extensions that are\n not required to determine the cryptographic parameters, other than\n those that are specific to individual certificates.\n [Section 4.3.1]\n\n CertificateRequest: if certificate-based client authentication is\n desired, the desired parameters for that certificate. This\n message is omitted if client authentication is not desired.\n [Section 4.3.2]\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 12]\n\f\nRFC 8446 TLS August 2018\n\n\n Finally, the client and server exchange Authentication messages. TLS\n uses the same set of messages every time that certificate-based\n authentication is needed. (PSK-based authentication happens as a\n side effect of key exchange.) Specifically:\n\n Certificate: The certificate of the endpoint and any per-certificate\n extensions. This message is omitted by the server if not\n authenticating with a certificate and by the client if the server\n did not send CertificateRequest (thus indicating that the client\n should not authenticate with a certificate). Note that if raw\n public keys [RFC7250] or the cached information extension\n [RFC7924] are in use, then this message will not contain a\n certificate but rather some other value corresponding to the\n server's long-term key. [Section 4.4.2]\n\n CertificateVerify: A signature over the entire handshake using the\n private key corresponding to the public key in the Certificate\n message. This message is omitted if the endpoint is not\n authenticating via a certificate. [Section 4.4.3]\n\n Finished: A MAC (Message Authentication Code) over the entire\n handshake. This message provides key confirmation, binds the\n endpoint's identity to the exchanged keys, and in PSK mode also\n authenticates the handshake. [Section 4.4.4]\n\n Upon receiving the server's messages, the client responds with its\n Authentication messages, namely Certificate and CertificateVerify (if\n requested), and Finished.\n\n At this point, the handshake is complete, and the client and server\n derive the keying material required by the record layer to exchange\n application-layer data protected through authenticated encryption.\n Application Data MUST NOT be sent prior to sending the Finished\n message, except as specified in Section 2.3. Note that while the\n server may send Application Data prior to receiving the client's\n Authentication messages, any data sent at that point is, of course,\n being sent to an unauthenticated peer.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 13]\n\f\nRFC 8446 TLS August 2018\n\n\n2.1. Incorrect DHE Share\n\n If the client has not provided a sufficient \"key_share\" extension\n (e.g., it includes only DHE or ECDHE groups unacceptable to or\n unsupported by the server), the server corrects the mismatch with a\n HelloRetryRequest and the client needs to restart the handshake with\n an appropriate \"key_share\" extension, as shown in Figure 2. If no\n common cryptographic parameters can be negotiated, the server MUST\n abort the handshake with an appropriate alert.\n\n Client Server\n\n ClientHello\n + key_share -------->\n HelloRetryRequest\n <-------- + key_share\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 2: Message Flow for a Full Handshake with\n Mismatched Parameters\n\n Note: The handshake transcript incorporates the initial\n ClientHello/HelloRetryRequest exchange; it is not reset with the\n new ClientHello.\n\n TLS also allows several optimized variants of the basic handshake, as\n described in the following sections.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 14]\n\f\nRFC 8446 TLS August 2018\n\n\n2.2. Resumption and Pre-Shared Key (PSK)\n\n Although TLS PSKs can be established out of band, PSKs can also be\n established in a previous connection and then used to establish a new\n connection (\"session resumption\" or \"resuming\" with a PSK). Once a\n handshake has completed, the server can send the client a PSK\n identity that corresponds to a unique key derived from the initial\n handshake (see Section 4.6.1). The client can then use that PSK\n identity in future handshakes to negotiate the use of the associated\n PSK. If the server accepts the PSK, then the security context of the\n new connection is cryptographically tied to the original connection\n and the key derived from the initial handshake is used to bootstrap\n the cryptographic state instead of a full handshake. In TLS 1.2 and\n below, this functionality was provided by \"session IDs\" and \"session\n tickets\" [RFC5077]. Both mechanisms are obsoleted in TLS 1.3.\n\n PSKs can be used with (EC)DHE key exchange in order to provide\n forward secrecy in combination with shared keys, or can be used\n alone, at the cost of losing forward secrecy for the application\n data.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 15]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 3 shows a pair of handshakes in which the first handshake\n establishes a PSK and the second handshake uses it:\n\n Client Server\n\n Initial Handshake:\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n <-------- [NewSessionTicket]\n [Application Data] <-------> [Application Data]\n\n\n Subsequent Handshake:\n ClientHello\n + key_share*\n + pre_shared_key -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n {Finished}\n <-------- [Application Data*]\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 3: Message Flow for Resumption and PSK\n\n As the server is authenticating via a PSK, it does not send a\n Certificate or a CertificateVerify message. When a client offers\n resumption via a PSK, it SHOULD also supply a \"key_share\" extension\n to the server to allow the server to decline resumption and fall back\n to a full handshake, if needed. The server responds with a\n \"pre_shared_key\" extension to negotiate the use of PSK key\n establishment and can (as shown here) respond with a \"key_share\"\n extension to do (EC)DHE key establishment, thus providing forward\n secrecy.\n\n\n\n\nRescorla Standards Track [Page 16]\n\f\nRFC 8446 TLS August 2018\n\n\n When PSKs are provisioned out of band, the PSK identity and the KDF\n hash algorithm to be used with the PSK MUST also be provisioned.\n\n Note: When using an out-of-band provisioned pre-shared secret, a\n critical consideration is using sufficient entropy during the key\n generation, as discussed in [RFC4086]. Deriving a shared secret\n from a password or other low-entropy sources is not secure. A\n low-entropy secret, or password, is subject to dictionary attacks\n based on the PSK binder. The specified PSK authentication is not\n a strong password-based authenticated key exchange even when used\n with Diffie-Hellman key establishment. Specifically, it does not\n prevent an attacker that can observe the handshake from performing\n a brute-force attack on the password/pre-shared key.\n\n2.3. 0-RTT Data\n\n When clients and servers share a PSK (either obtained externally or\n via a previous handshake), TLS 1.3 allows clients to send data on the\n first flight (\"early data\"). The client uses the PSK to authenticate\n the server and to encrypt the early data.\n\n As shown in Figure 4, the 0-RTT data is just added to the 1-RTT\n handshake in the first flight. The rest of the handshake uses the\n same messages as for a 1-RTT handshake with PSK resumption.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 17]\n\f\nRFC 8446 TLS August 2018\n\n\n Client Server\n\n ClientHello\n + early_data\n + key_share*\n + psk_key_exchange_modes\n + pre_shared_key\n (Application Data*) -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n + early_data*\n {Finished}\n <-------- [Application Data*]\n (EndOfEarlyData)\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n () Indicates messages protected using keys\n derived from a client_early_traffic_secret.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 4: Message Flow for a 0-RTT Handshake\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 18]\n\f\nRFC 8446 TLS August 2018\n\n\n IMPORTANT NOTE: The security properties for 0-RTT data are weaker\n than those for other kinds of TLS data. Specifically:\n\n 1. This data is not forward secret, as it is encrypted solely under\n keys derived using the offered PSK.\n\n 2. There are no guarantees of non-replay between connections.\n Protection against replay for ordinary TLS 1.3 1-RTT data is\n provided via the server's Random value, but 0-RTT data does not\n depend on the ServerHello and therefore has weaker guarantees.\n This is especially relevant if the data is authenticated either\n with TLS client authentication or inside the application\n protocol. The same warnings apply to any use of the\n early_exporter_master_secret.\n\n\n\n===== DOCUMENT =====\n\n A TLS connection state is the operating environment of the TLS Record\n Protocol. It specifies a compression algorithm, an encryption\n algorithm, and a MAC algorithm. In addition, the parameters for\n these algorithms are known: the MAC key and the bulk encryption keys\n for the connection in both the read and the write directions.\n Logically, there are always four connection states outstanding: the\n current read and write states, and the pending read and write states.\n All records are processed under the current read and write states.\n The security parameters for the pending states can be set by the TLS\n Handshake Protocol, and the ChangeCipherSpec can selectively make\n either of the pending states current, in which case the appropriate\n current state is disposed of and replaced with the pending state; the\n pending state is then reinitialized to an empty state. It is illegal\n to make a state that has not been initialized with security\n parameters a current state. The initial current state always\n specifies that no encryption, compression, or MAC will be used.\n\n The security parameters for a TLS Connection read and write state are\n set by providing the following values:\n\n connection end\n Whether this entity is considered the \"client\" or the \"server\" in\n this connection.\n\n PRF algorithm\n An algorithm used to generate keys from the master secret (see\n Sections 5 and 6.3).\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 16]\n\f\nRFC 5246 TLS August 2008\n\n\n bulk encryption algorithm\n An algorithm to be used for bulk encryption. This specification\n includes the key size of this algorithm, whether it is a block,\n stream, or AEAD cipher, the block size of the cipher (if\n appropriate), and the lengths of explicit and implicit\n initialization vectors (or nonces).\n\n MAC algorithm\n An algorithm to be used for message authentication. This\n specification includes the size of the value returned by the MAC\n algorithm.\n\n compression algorithm\n An algorithm to be used for data compression. This specification\n must include all information the algorithm requires to do\n compression.\n\n master secret\n A 48-byte secret shared between the two peers in the connection.\n\n client random\n A 32-byte value provided by the client.\n\n server random\n A 32-byte value provided by the server.\n\n These parameters are defined in the presentation language as:\n\n enum { server, client } ConnectionEnd;\n\n enum { tls_prf_sha256 } PRFAlgorithm;\n\n enum { null, rc4, 3des, aes }\n BulkCipherAlgorithm;\n\n enum { stream, block, aead } CipherType;\n\n enum { null, hmac_md5, hmac_sha1, hmac_sha256,\n hmac_sha384, hmac_sha512} MACAlgorithm;\n\n enum { null(0), (255) } CompressionMethod;\n\n /* The algorithms specified in CompressionMethod, PRFAlgorithm,\n BulkCipherAlgorithm, and MACAlgorithm may be added to. */\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 17]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ConnectionEnd entity;\n PRFAlgorithm prf_algorithm;\n BulkCipherAlgorithm bulk_cipher_algorithm;\n CipherType cipher_type;\n uint8 enc_key_length;\n uint8 block_length;\n uint8 fixed_iv_length;\n uint8 record_iv_length;\n MACAlgorithm mac_algorithm;\n uint8 mac_length;\n uint8 mac_key_length;\n CompressionMethod compression_algorithm;\n opaque master_secret[48];\n opaque client_random[32];\n opaque server_random[32];\n } SecurityParameters;\n\n The record layer will use the security parameters to generate the\n following six items (some of which are not required by all ciphers,\n and are thus empty):\n\n client write MAC key\n server write MAC key\n client write encryption key\n server write encryption key\n client write IV\n server write IV\n\n The client write parameters are used by the server when receiving and\n processing records and vice versa. The algorithm used for generating\n these items from the security parameters is described in Section 6.3.\n\n Once the security parameters have been set and the keys have been\n generated, the connection states can be instantiated by making them\n the current states. These current states MUST be updated for each\n record processed. Each connection state includes the following\n elements:\n\n compression state\n The current state of the compression algorithm.\n\n cipher state\n The current state of the encryption algorithm. This will consist\n of the scheduled key for that connection. For stream ciphers,\n this will also contain whatever state information is necessary to\n allow the stream to continue to encrypt or decrypt data.\n\n\n\n\nDierks & Rescorla Standards Track [Page 18]\n\f\nRFC 5246 TLS August 2008\n\n\n MAC key\n The MAC key for this connection, as generated above.\n\n sequence number\n Each connection state contains a sequence number, which is\n maintained separately for read and write states. The sequence\n number MUST be set to zero whenever a connection state is made the\n active state. Sequence numbers are of type uint64 and may not\n exceed 2^64-1. Sequence numbers do not wrap. If a TLS\n implementation would need to wrap a sequence number, it must\n renegotiate instead. A sequence number is incremented after each\n record: specifically, the first record transmitted under a\n particular connection state MUST use sequence number 0.\n\n6.2. Record Layer\n\n The TLS record layer receives uninterpreted data from higher layers\n in non-empty blocks of arbitrary size.\n\n6.2.1. Fragmentation\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Client\n message boundaries are not preserved in the record layer (i.e.,\n multiple client messages of the same ContentType MAY be coalesced\n into a single TLSPlaintext record, or a single message MAY be\n fragmented across several records).\n\n struct {\n uint8 major;\n uint8 minor;\n } ProtocolVersion;\n\n enum {\n change_cipher_spec(20), alert(21), handshake(22),\n application_data(23), (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type\n The higher-level protocol used to process the enclosed fragment.\n\n\n\n\nDierks & Rescorla Standards Track [Page 19]\n\f\nRFC 5246 TLS August 2008\n\n\n version\n The version of the protocol being employed. This document\n describes TLS Version 1.2, which uses the version { 3, 3 }. The\n version value 3.3 is historical, deriving from the use of {3, 1}\n for TLS 1.0. (See Appendix A.1.) Note that a client that\n supports multiple versions of TLS may not know what version will\n be employed before it receives the ServerHello. See Appendix E\n for discussion about what record layer version number should be\n employed for ClientHello.\n\n length\n The length (in bytes) of the following TLSPlaintext.fragment. The\n length MUST NOT exceed 2^14.\n\n fragment\n The application data. This data is transparent and treated as an\n independent block to be dealt with by the higher-level protocol\n specified by the type field.\n\n Implementations MUST NOT send zero-length fragments of Handshake,\n Alert, or ChangeCipherSpec content types. Zero-length fragments of\n Application data MAY be sent as they are potentially useful as a\n traffic analysis countermeasure.\n\n Note: Data of different TLS record layer content types MAY be\n interleaved. Application data is generally of lower precedence for\n transmission than other content types. However, records MUST be\n delivered to the network in the same order as they are protected by\n the record layer. Recipients MUST receive and process interleaved\n application layer traffic during handshakes subsequent to the first\n one on a connection.\n\n6.2.2. Record Compression and Decompression\n\n All records are compressed using the compression algorithm defined in\n the current session state. There is always an active compression\n algorithm; however, initially it is defined as\n CompressionMethod.null. The compression algorithm translates a\n TLSPlaintext structure into a TLSCompressed structure. Compression\n functions are initialized with default state information whenever a\n connection state is made active. [RFC3749] describes compression\n algorithms for TLS.\n\n Compression must be lossless and may not increase the content length\n by more than 1024 bytes. If the decompression function encounters a\n TLSCompressed.fragment that would decompress to a length in excess of\n 2^14 bytes, it MUST report a fatal decompression failure error.\n\n\n\n\nDierks & Rescorla Standards Track [Page 20]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ContentType type; /* same as TLSPlaintext.type */\n ProtocolVersion version;/* same as TLSPlaintext.version */\n uint16 length;\n opaque fragment[TLSCompressed.length];\n } TLSCompressed;\n\n length\n The length (in bytes) of the following TLSCompressed.fragment.\n The length MUST NOT exceed 2^14 + 1024.\n\n fragment\n The compressed form of TLSPlaintext.fragment.\n\n Note: A CompressionMethod.null operation is an identity operation;\n no fields are altered.\n\n Implementation note: Decompression functions are responsible for\n ensuring that messages cannot cause internal buffer overflows.\n\n6.2.3. Record Payload Protection\n\n The encryption and MAC functions translate a TLSCompressed\n structure into a TLSCiphertext. The decryption functions reverse\n the process. The MAC of the record also includes a sequence\n number so that missing, extra, or repeated messages are\n detectable.\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n select (SecurityParameters.cipher_type) {\n case stream: GenericStreamCipher;\n case block: GenericBlockCipher;\n case aead: GenericAEADCipher;\n } fragment;\n } TLSCiphertext;\n\n type\n The type field is identical to TLSCompressed.type.\n\n version\n The version field is identical to TLSCompressed.version.\n\n length\n The length (in bytes) of the following TLSCiphertext.fragment.\n The length MUST NOT exceed 2^14 + 2048.\n\n\n\nDierks & Rescorla Standards Track [Page 21]\n\f\nRFC 5246 TLS August 2008\n\n\n fragment\n The encrypted form of TLSCompressed.fragment, with the MAC.\n\n6.2.3.1. Null or Standard Stream Cipher\n\n Stream ciphers (including BulkCipherAlgorithm.null; see Appendix A.6)\n convert TLSCompressed.fragment structures to and from stream\n TLSCiphertext.fragment structures.\n\n stream-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n } GenericStreamCipher;\n\n The MAC is generated as:\n\n MAC(MAC_write_key, seq_num +\n TLSCompressed.type +\n TLSCompressed.version +\n TLSCompressed.length +\n TLSCompressed.fragment);\n\n where \"+\" denotes concatenation.\n\n seq_num\n The sequence number for this record.\n\n MAC\n The MAC algorithm specified by SecurityParameters.mac_algorithm.\n\n Note that the MAC is computed before encryption. The stream cipher\n encrypts the entire block, including the MAC. For stream ciphers\n that do not use a synchronization vector (such as RC4), the stream\n cipher state from the end of one record is simply used on the\n subsequent packet. If the cipher suite is TLS_NULL_WITH_NULL_NULL,\n encryption consists of the identity operation (i.e., the data is not\n encrypted, and the MAC size is zero, implying that no MAC is used).\n For both null and stream ciphers, TLSCiphertext.length is\n TLSCompressed.length plus SecurityParameters.mac_length.\n\n6.2.3.2. CBC Block Cipher\n\n For block ciphers (such as 3DES or AES), the encryption and MAC\n functions convert TLSCompressed.fragment structures to and from block\n TLSCiphertext.fragment structures.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 22]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n opaque IV[SecurityParameters.record_iv_length];\n block-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n uint8 padding[GenericBlockCipher.padding_length];\n uint8 padding_length;\n };\n } GenericBlockCipher;\n\n The MAC is generated as described in Section 6.2.3.1.\n\n IV\n The Initialization Vector (IV) SHOULD be chosen at random, and\n MUST be unpredictable. Note that in versions of TLS prior to 1.1,\n there was no IV field, and the last ciphertext block of the\n previous record (the \"CBC residue\") was used as the IV. This was\n changed to prevent the attacks described in [CBCATT]. For block\n ciphers, the IV length is of length\n SecurityParameters.record_iv_length, which is equal to the\n SecurityParameters.block_size.\n\n padding\n Padding that is added to force the length of the plaintext to be\n an integral multiple of the block cipher's block length. The\n padding MAY be any length up to 255 bytes, as long as it results\n in the TLSCiphertext.length being an integral multiple of the\n block length. Lengths longer than necessary might be desirable to\n frustrate attacks on a protocol that are based on analysis of the\n lengths of exchanged messages. Each uint8 in the padding data\n vector MUST be filled with the padding length value. The receiver\n MUST check this padding and MUST use the bad_record_mac alert to\n indicate padding errors.\n\n padding_length\n The padding length MUST be such that the total size of the\n GenericBlockCipher structure is a multiple of the cipher's block\n length. Legal values range from zero to 255, inclusive. This\n length specifies the length of the padding field exclusive of the\n padding_length field itself.\n\n The encrypted data length (TLSCiphertext.length) is one more than the\n sum of SecurityParameters.block_length, TLSCompressed.length,\n SecurityParameters.mac_length, and padding_length.\n\n Example: If the block length is 8 bytes, the content length\n (TLSCompressed.length) is 61 bytes, and the MAC length is 20 bytes,\n then the length before padding is 82 bytes (this does not include the\n\n\n\nDierks & Rescorla Standards Track [Page 23]\n\f\nRFC 5246 TLS August 2008\n\n\n IV. Thus, the padding length modulo 8 must be equal to 6 in order to\n make the total length an even multiple of 8 bytes (the block length).\n The padding length can be 6, 14, 22, and so on, through 254. If the\n padding length were the minimum necessary, 6, the padding would be 6\n bytes, each containing the value 6. Thus, the last 8 octets of the\n GenericBlockCipher before block encryption would be xx 06 06 06 06 06\n 06 06, where xx is the last octet of the MAC.\n\n Note: With block ciphers in CBC mode (Cipher Block Chaining), it is\n critical that the entire plaintext of the record be known before any\n ciphertext is transmitted. Otherwise, it is possible for the\n attacker to mount the attack described in [CBCATT].\n\n Implementation note: Canvel et al. [CBCTIME] have demonstrated a\n timing attack on CBC padding based on the time required to compute\n the MAC. In order to defend against this attack, implementations\n MUST ensure that record processing time is essentially the same\n whether or not the padding is correct. In general, the best way to\n do this is to compute the MAC even if the padding is incorrect, and\n only then reject the packet. For instance, if the pad appears to be\n incorrect, the implementation might assume a zero-length pad and then\n compute the MAC. This leaves a small timing channel, since MAC\n performance depends to some extent on the size of the data fragment,\n but it is not believed to be large enough to be exploitable, due to\n the large block size of existing MACs and the small size of the\n timing signal.\n\n6.2.3.3. AEAD Ciphers\n\n For AEAD [AEAD] ciphers (such as [CCM] or [GCM]), the AEAD function\n converts TLSCompressed.fragment structures to and from AEAD\n TLSCiphertext.fragment structures.\n\n struct {\n opaque nonce_explicit[SecurityParameters.record_iv_length];\n aead-ciphered struct {\n opaque content[TLSCompressed.length];\n };\n } GenericAEADCipher;\n\n AEAD ciphers take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [AEAD]. The key is either the\n client_write_key or the server_write_key. No MAC key is used.\n\n Each AEAD cipher suite MUST specify how the nonce supplied to the\n AEAD operation is constructed, and what is the length of the\n GenericAEADCipher.nonce_explicit part. In many cases, it is\n\n\n\nDierks & Rescorla Standards Track [Page 24]\n\f\nRFC 5246 TLS August 2008\n\n\n appropriate to use the partially implicit nonce technique described\n in Section 3.2.1 of [AEAD]; with record_iv_length being the length of\n the explicit part. In this case, the implicit part SHOULD be derived\n from key_block as client_write_iv and server_write_iv (as described\n in Section 6.3), and the explicit part is included in\n GenericAEAEDCipher.nonce_explicit.\n\n The plaintext is the TLSCompressed.fragment.\n\n The additional authenticated data, which we denote as\n additional_data, is defined as follows:\n\n additional_data = seq_num + TLSCompressed.type +\n TLSCompressed.version + TLSCompressed.length;\n\n where \"+\" denotes concatenation.\n\n The aead_output consists of the ciphertext output by the AEAD\n encryption operation. The length will generally be larger than\n TLSCompressed.length, but by an amount that varies with the AEAD\n cipher. Since the ciphers might incorporate padding, the amount of\n overhead could vary with different TLSCompressed.length values. Each\n AEAD cipher MUST NOT produce an expansion of greater than 1024 bytes.\n Symbolically,\n\n AEADEncrypted = AEAD-Encrypt(write_key, nonce, plaintext,\n additional_data)\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, the \"additional_data\", and the AEADEncrypted value. The\n output is either the plaintext or an error indicating that the\n decryption failed. There is no separate integrity check. That is:\n\n TLSCompressed.fragment = AEAD-Decrypt(write_key, nonce,\n AEADEncrypted,\n additional_data)\n\n If the decryption fails, a fatal bad_record_mac alert MUST be\n generated.\n\n6.3. Key Calculation\n\n The Record Protocol requires an algorithm to generate keys required\n by the current connection state (see Appendix A.6) from the security\n parameters provided by the handshake protocol.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 25]\n\f\nRFC 5246 TLS August 2008\n\n\n The master secret is expanded into a sequence of secure bytes, which\n is then split to a client write MAC key, a server write MAC key, a\n client write encryption key, and a server write encryption key. Each\n of these is generated from the byte sequence in that order. Unused\n values are empty. Some AEAD ciphers may additionally require a\n client write IV and a server write IV (see Section 6.2.3.3).\n\n When keys and MAC keys are generated, the master secret is used as an\n entropy source.\n\n To generate the key material, compute\n\n key_block = PRF(SecurityParameters.master_secret,\n \"key expansion\",\n SecurityParameters.server_random +\n SecurityParameters.client_random);\n\n until enough output has been generated. Then, the key_block is\n partitioned as follows:\n\n client_write_MAC_key[SecurityParameters.mac_key_length]\n server_write_MAC_key[SecurityParameters.mac_key_length]\n client_write_key[SecurityParameters.enc_key_length]\n server_write_key[SecurityParameters.enc_key_length]\n client_write_IV[SecurityParameters.fixed_iv_length]\n server_write_IV[SecurityParameters.fixed_iv_length]\n\n Currently, the client_write_IV and server_write_IV are only generated\n for implicit nonce techniques as described in Section 3.2.1 of\n [AEAD].\n\n Implementation note: The currently defined cipher suite which\n requires the most material is AES_256_CBC_SHA256. It requires 2 x 32\n byte keys and 2 x 32 byte MAC keys, for a total 128 bytes of key\n material.\n\n7. The TLS Handshaking Protocols\n\n TLS has three subprotocols that are used to allow peers to agree upon\n security parameters for the record layer, to authenticate themselves,\n to instantiate negotiated security parameters, and to report error\n conditions to each other.\n\n The Handshake Protocol is responsible for negotiating a session,\n which consists of the following items:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 26]\n\f\nRFC 5246 TLS August 2008\n\n\n session identifier\n An arbitrary byte sequence chosen by the server to identify an\n active or resumable session state.\n\n peer certificate\n X509v3 [PKIX] certificate of the peer. This element of the state\n may be null.\n\n compression method\n The algorithm used to compress data prior to encryption.\n\n cipher spec\n Specifies the pseudorandom function (PRF) used to generate keying\n material, the bulk data encryption algorithm (such as null, AES,\n etc.) and the MAC algorithm (such as HMAC-SHA1). It also defines\n cryptographic attributes such as the mac_length. (See Appendix\n A.6 for formal definition.)\n\n master secret\n 48-byte secret shared between the client and server.\n\n is resumable\n A flag indicating whether the session can be used to initiate new\n connections.\n\n These items are then used to create security parameters for use by\n the record layer when protecting application data. Many connections\n can be instantiated using the same session through the resumption\n feature of the TLS Handshake Protocol.\n\n7.1. Change Cipher Spec Protocol\n\n The change cipher spec protocol exists to signal transitions in\n ciphering strategies. The protocol consists of a single message,\n which is encrypted and compressed under the current (not the pending)\n connection state. The message consists of a single byte of value 1.\n\n struct {\n enum { change_cipher_spec(1), (255) } type;\n } ChangeCipherSpec;\n\n The ChangeCipherSpec message is sent by both the client and the\n server to notify the receiving party that subsequent records will be\n protected under the newly negotiated CipherSpec and keys. Reception\n of this message causes the receiver to instruct the record layer to\n immediately copy the read pending state into the read current state.\n Immediately after sending this message, the sender MUST instruct the\n record layer to make the write pending state the write active state.\n\n\n\nDierks & Rescorla Standards Track [Page 27]\n\f\nRFC 5246 TLS August 2008\n\n\n (See Section 6.1.) The ChangeCipherSpec message is sent during the\n handshake after the security parameters have been agreed upon, but\n before the verifying Finished message is sent.\n\n Note: If a rehandshake occurs while data is flowing on a connection,\n the communicating parties may continue to send data using the old\n CipherSpec. However, once the ChangeCipherSpec has been sent, the\n new CipherSpec MUST be used. The first side to send the\n ChangeCipherSpec does not know that the other side has finished\n computing the new keying material (e.g., if it has to perform a\n time-consuming public key operation). Thus, a small window of time,\n during which the recipient must buffer the data, MAY exist. In\n practice, with modern machines this interval is likely to be fairly\n short.\n\n7.2. Alert Protocol\n\n One of the content types supported by the TLS record layer is the\n alert type. Alert messages convey the severity of the message\n (warning or fatal) and a description of the alert. Alert messages\n with a level of fatal result in the immediate termination of the\n connection. In this case, other connections corresponding to the\n session may continue, but the session identifier MUST be invalidated,\n preventing the failed session from being used to establish new\n connections. Like other messages, alert messages are encrypted and\n compressed, as specified by the current connection state.\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n\n\n\nDierks & Rescorla Standards Track [Page 28]\n\f\nRFC 5246 TLS August 2008\n\n\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n7.2.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack. Either party may\n initiate the exchange of closing messages.\n\n close_notify\n This message notifies the recipient that the sender will not send\n any more messages on this connection. Note that as of TLS 1.1,\n failure to properly close a connection no longer requires that a\n session not be resumed. This is a change from TLS 1.0 to conform\n with widespread implementation practice.\n\n Either party may initiate a close by sending a close_notify alert.\n Any data received after a closure alert is ignored.\n\n Unless some other fatal alert has been transmitted, each party is\n required to send a close_notify alert before closing the write side\n of the connection. The other party MUST respond with a close_notify\n alert of its own and close down the connection immediately,\n discarding any pending writes. It is not required for the initiator\n of the close to wait for the responding close_notify alert before\n closing the read side of the connection.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation must receive the responding\n close_notify alert before indicating to the application layer that\n the TLS connection has ended. If the application protocol will not\n transfer any additional data, but will only close the underlying\n transport connection, then the implementation MAY choose to close the\n transport without waiting for the responding close_notify. No part\n\n\n\n\nDierks & Rescorla Standards Track [Page 29]\n\f\nRFC 5246 TLS August 2008\n\n\n of this standard should be taken to dictate the manner in which a\n usage profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing a connection reliably delivers\n pending data before destroying the transport.\n\n7.2.2. Error Alerts\n\n Error handling in the TLS Handshake protocol is very simple. When an\n error is detected, the detecting party sends a message to the other\n party. Upon transmission or receipt of a fatal alert message, both\n parties immediately close the connection. Servers and clients MUST\n forget any session-identifiers, keys, and secrets associated with a\n failed connection. Thus, any connection terminated with a fatal\n alert MUST NOT be resumed.\n\n Whenever an implementation encounters a condition which is defined as\n a fatal alert, it MUST send the appropriate alert prior to closing\n the connection. For all errors where an alert level is not\n explicitly specified, the sending party MAY determine at its\n discretion whether to treat this as a fatal error or not. If the\n implementation chooses to send an alert but intends to close the\n connection immediately afterwards, it MUST send that alert at the\n fatal alert level.\n\n If an alert with a level of warning is sent and received, generally\n the connection can continue normally. If the receiving party decides\n not to proceed with the connection (e.g., after having received a\n no_renegotiation alert that it is not willing to accept), it SHOULD\n send a fatal alert to terminate the connection. Given this, the\n sending party cannot, in general, know how the receiving party will\n behave. Therefore, warning alerts are not very useful when the\n sending party wants to continue the connection, and thus are\n sometimes omitted. For example, if a peer decides to accept an\n expired certificate (perhaps after confirming this with the user) and\n wants to continue the connection, it would not generally send a\n certificate_expired alert.\n\n The following error alerts are defined:\n\n unexpected_message\n An inappropriate message was received. This alert is always fatal\n and should never be observed in communication between proper\n implementations.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 30]\n\f\nRFC 5246 TLS August 2008\n\n\n bad_record_mac\n This alert is returned if a record is received with an incorrect\n MAC. This alert also MUST be returned if an alert is sent because\n a TLSCiphertext decrypted in an invalid way: either it wasn't an\n even multiple of the block length, or its padding values, when\n checked, weren't correct. This message is always fatal and should\n never be observed in communication between proper implementations\n (except when messages were corrupted in the network).\n\n decryption_failed_RESERVED\n This alert was used in some earlier versions of TLS, and may have\n permitted certain attacks against the CBC mode [CBCATT]. It MUST\n NOT be sent by compliant implementations.\n\n record_overflow\n A TLSCiphertext record was received that had a length more than\n 2^14+2048 bytes, or a record decrypted to a TLSCompressed record\n with more than 2^14+1024 bytes. This message is always fatal and\n should never be observed in communication between proper\n implementations (except when messages were corrupted in the\n network).\n\n decompression_failure\n The decompression function received improper input (e.g., data\n that would expand to excessive length). This message is always\n fatal and should never be observed in communication between proper\n implementations.\n\n handshake_failure\n Reception of a handshake_failure alert message indicates that the\n sender was unable to negotiate an acceptable set of security\n parameters given the options available. This is a fatal error.\n\n no_certificate_RESERVED\n This alert was used in SSLv3 but not any version of TLS. It MUST\n NOT be sent by compliant implementations.\n\n bad_certificate\n A certificate was corrupt, contained signatures that did not\n verify correctly, etc.\n\n unsupported_certificate\n A certificate was of an unsupported type.\n\n certificate_revoked\n A certificate was revoked by its signer.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 31]\n\f\nRFC 5246 TLS August 2008\n\n\n certificate_expired\n A certificate has expired or is not currently valid.\n\n certificate_unknown\n Some other (unspecified) issue arose in processing the\n certificate, rendering it unacceptable.\n\n illegal_parameter\n A field in the handshake was out of range or inconsistent with\n other fields. This message is always fatal.\n\n unknown_ca\n A valid certificate chain or partial chain was received, but the\n certificate was not accepted because the CA certificate could not\n be located or couldn't be matched with a known, trusted CA. This\n message is always fatal.\n\n access_denied\n A valid certificate was received, but when access control was\n applied, the sender decided not to proceed with negotiation. This\n message is always fatal.\n\n decode_error\n A message could not be decoded because some field was out of the\n specified range or the length of the message was incorrect. This\n message is always fatal and should never be observed in\n communication between proper implementations (except when messages\n were corrupted in the network).\n\n decrypt_error\n A handshake cryptographic operation failed, including being unable\n to correctly verify a signature or validate a Finished message.\n This message is always fatal.\n\n export_restriction_RESERVED\n This alert was used in some earlier versions of TLS. It MUST NOT\n be sent by compliant implementations.\n\n protocol_version\n The protocol version the client has attempted to negotiate is\n recognized but not supported. (For example, old protocol versions\n might be avoided for security reasons.) This message is always\n fatal.\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 32]\n\f\nRFC 5246 TLS August 2008\n\n\n insufficient_security\n Returned instead of handshake_failure when a negotiation has\n failed specifically because the server requires ciphers more\n secure than those supported by the client. This message is always\n fatal.\n\n internal_error\n An internal error unrelated to the peer or the correctness of the\n protocol (such as a memory allocation failure) makes it impossible\n to continue. This message is always fatal.\n\n\n\n===== DOCUMENT =====\n\nNetwork Working Group E. Rescorla\nInternet-Draft RTFM, Inc.\nObsoletes: 5077, 5246, 6961 (if March 20, 2018\n approved)\nUpdates: 4492, 5705, 6066 (if approved)\nIntended status: Standards Track\nExpires: September 21, 2018\n\n\n\n===== DOCUMENT =====\n\n user_canceled\n This handshake is being canceled for some reason unrelated to a\n protocol failure. If the user cancels an operation after the\n handshake is complete, just closing the connection by sending a\n close_notify is more appropriate. This alert should be followed\n by a close_notify. This message is generally a warning.\n\n no_renegotiation\n Sent by the client in response to a hello request or by the server\n in response to a client hello after initial handshaking. Either\n of these would normally lead to renegotiation; when that is not\n appropriate, the recipient should respond with this alert. At\n that point, the original requester can decide whether to proceed\n with the connection. One case where this would be appropriate is\n where a server has spawned a process to satisfy a request; the\n process might receive security parameters (key length,\n authentication, etc.) at startup, and it might be difficult to\n communicate changes to these parameters after that point. This\n message is always a warning.\n\n unsupported_extension\n sent by clients that receive an extended server hello containing\n an extension that they did not put in the corresponding client\n hello. This message is always fatal.\n\n New Alert values are assigned by IANA as described in Section 12.\n\n7.3. Handshake Protocol Overview\n\n The cryptographic parameters of the session state are produced by the\n TLS Handshake Protocol, which operates on top of the TLS record\n layer. When a TLS client and server first start communicating, they\n agree on a protocol version, select cryptographic algorithms,\n optionally authenticate each other, and use public-key encryption\n techniques to generate shared secrets.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 33]\n\f\nRFC 5246 TLS August 2008\n\n\n The TLS Handshake Protocol involves the following steps:\n\n - Exchange hello messages to agree on algorithms, exchange random\n values, and check for session resumption.\n\n - Exchange the necessary cryptographic parameters to allow the\n client and server to agree on a premaster secret.\n\n - Exchange certificates and cryptographic information to allow the\n client and server to authenticate themselves.\n\n - Generate a master secret from the premaster secret and exchanged\n random values.\n\n - Provide security parameters to the record layer.\n\n - Allow the client and server to verify that their peer has\n calculated the same security parameters and that the handshake\n occurred without tampering by an attacker.\n\n Note that higher layers should not be overly reliant on whether TLS\n always negotiates the strongest possible connection between two\n peers. There are a number of ways in which a man-in-the-middle\n attacker can attempt to make two entities drop down to the least\n secure method they support. The protocol has been designed to\n minimize this risk, but there are still attacks available: for\n example, an attacker could block access to the port a secure service\n runs on, or attempt to get the peers to negotiate an unauthenticated\n connection. The fundamental rule is that higher levels must be\n cognizant of what their security requirements are and never transmit\n information over a channel less secure than what they require. The\n TLS protocol is secure in that any cipher suite offers its promised\n level of security: if you negotiate 3DES with a 1024-bit RSA key\n exchange with a host whose certificate you have verified, you can\n expect to be that secure.\n\n These goals are achieved by the handshake protocol, which can be\n summarized as follows: The client sends a ClientHello message to\n which the server must respond with a ServerHello message, or else a\n fatal error will occur and the connection will fail. The ClientHello\n and ServerHello are used to establish security enhancement\n capabilities between client and server. The ClientHello and\n ServerHello establish the following attributes: Protocol Version,\n Session ID, Cipher Suite, and Compression Method. Additionally, two\n random values are generated and exchanged: ClientHello.random and\n ServerHello.random.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 34]\n\f\nRFC 5246 TLS August 2008\n\n\n The actual key exchange uses up to four messages: the server\n Certificate, the ServerKeyExchange, the client Certificate, and the\n ClientKeyExchange. New key exchange methods can be created by\n specifying a format for these messages and by defining the use of the\n messages to allow the client and server to agree upon a shared\n secret. This secret MUST be quite long; currently defined key\n exchange methods exchange secrets that range from 46 bytes upwards.\n\n Following the hello messages, the server will send its certificate in\n a Certificate message if it is to be authenticated. Additionally, a\n ServerKeyExchange message may be sent, if it is required (e.g., if\n the server has no certificate, or if its certificate is for signing\n only). If the server is authenticated, it may request a certificate\n from the client, if that is appropriate to the cipher suite selected.\n Next, the server will send the ServerHelloDone message, indicating\n that the hello-message phase of the handshake is complete. The\n server will then wait for a client response. If the server has sent\n a CertificateRequest message, the client MUST send the Certificate\n message. The ClientKeyExchange message is now sent, and the content\n of that message will depend on the public key algorithm selected\n between the ClientHello and the ServerHello. If the client has sent\n a certificate with signing ability, a digitally-signed\n CertificateVerify message is sent to explicitly verify possession of\n the private key in the certificate.\n\n At this point, a ChangeCipherSpec message is sent by the client, and\n the client copies the pending Cipher Spec into the current Cipher\n Spec. The client then immediately sends the Finished message under\n the new algorithms, keys, and secrets. In response, the server will\n send its own ChangeCipherSpec message, transfer the pending to the\n current Cipher Spec, and send its Finished message under the new\n Cipher Spec. At this point, the handshake is complete, and the\n client and server may begin to exchange application layer data. (See\n flow chart below.) Application data MUST NOT be sent prior to the\n completion of the first handshake (before a cipher suite other than\n TLS_NULL_WITH_NULL_NULL is established).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 35]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1. Message flow for a full handshake\n\n * Indicates optional or situation-dependent messages that are not\n always sent.\n\n Note: To help avoid pipeline stalls, ChangeCipherSpec is an\n independent TLS protocol content type, and is not actually a TLS\n handshake message.\n\n When the client and server decide to resume a previous session or\n duplicate an existing session (instead of negotiating new security\n parameters), the message flow is as follows:\n\n The client sends a ClientHello using the Session ID of the session to\n be resumed. The server then checks its session cache for a match.\n If a match is found, and the server is willing to re-establish the\n connection under the specified session state, it will send a\n ServerHello with the same Session ID value. At this point, both\n client and server MUST send ChangeCipherSpec messages and proceed\n directly to Finished messages. Once the re-establishment is\n complete, the client and server MAY begin to exchange application\n layer data. (See flow chart below.) If a Session ID match is not\n found, the server generates a new session ID, and the TLS client and\n server perform a full handshake.\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 36]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2. Message flow for an abbreviated handshake\n\n The contents and significance of each message will be presented in\n detail in the following sections.\n\n7.4. Handshake Protocol\n\n The TLS Handshake Protocol is one of the defined higher-level clients\n of the TLS Record Protocol. This protocol is used to negotiate the\n secure attributes of a session. Handshake messages are supplied to\n the TLS record layer, where they are encapsulated within one or more\n TLSPlaintext structures, which are processed and transmitted as\n specified by the current active session state.\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n } body;\n } Handshake;\n\n\n\nDierks & Rescorla Standards Track [Page 37]\n\f\nRFC 5246 TLS August 2008\n\n\n The handshake protocol messages are presented below in the order they\n MUST be sent; sending handshake messages in an unexpected order\n results in a fatal error. Unneeded handshake messages can be\n omitted, however. Note one exception to the ordering: the\n Certificate message is used twice in the handshake (from server to\n client, then from client to server), but described only in its first\n position. The one message that is not bound by these ordering rules\n is the HelloRequest message, which can be sent at any time, but which\n SHOULD be ignored by the client if it arrives in the middle of a\n handshake.\n\n New handshake message types are assigned by IANA as described in\n Section 12.\n\n7.4.1. Hello Messages\n\n The hello phase messages are used to exchange security enhancement\n capabilities between the client and server. When a new session\n begins, the record layer's connection state encryption, hash, and\n compression algorithms are initialized to null. The current\n connection state is used for renegotiation messages.\n\n7.4.1.1. Hello Request\n\n When this message will be sent:\n\n The HelloRequest message MAY be sent by the server at any time.\n\n Meaning of this message:\n\n HelloRequest is a simple notification that the client should begin\n the negotiation process anew. In response, the client should send\n a ClientHello message when convenient. This message is not\n intended to establish which side is the client or server but\n merely to initiate a new negotiation. Servers SHOULD NOT send a\n HelloRequest immediately upon the client's initial connection. It\n is the client's job to send a ClientHello at that time.\n\n This message will be ignored by the client if the client is\n currently negotiating a session. This message MAY be ignored by\n the client if it does not wish to renegotiate a session, or the\n client may, if it wishes, respond with a no_renegotiation alert.\n Since handshake messages are intended to have transmission\n precedence over application data, it is expected that the\n negotiation will begin before no more than a few records are\n received from the client. If the server sends a HelloRequest but\n does not receive a ClientHello in response, it may close the\n connection with a fatal alert.\n\n\n\nDierks & Rescorla Standards Track [Page 38]\n\f\nRFC 5246 TLS August 2008\n\n\n After sending a HelloRequest, servers SHOULD NOT repeat the\n request until the subsequent handshake negotiation is complete.\n\n Structure of this message:\n\n struct { } HelloRequest;\n\n This message MUST NOT be included in the message hashes that are\n maintained throughout the handshake and used in the Finished messages\n and the certificate verify message.\n\n7.4.1.2. Client Hello\n\n When this message will be sent:\n\n When a client first connects to a server, it is required to send\n the ClientHello as its first message. The client can also send a\n ClientHello in response to a HelloRequest or on its own initiative\n in order to renegotiate the security parameters in an existing\n connection.\n\n Structure of this message:\n\n The ClientHello message includes a random structure, which is used\n later in the protocol.\n\n struct {\n uint32 gmt_unix_time;\n opaque random_bytes[28];\n } Random;\n\n gmt_unix_time\n The current time and date in standard UNIX 32-bit format\n (seconds since the midnight starting Jan 1, 1970, UTC, ignoring\n leap seconds) according to the sender's internal clock. Clocks\n are not required to be set correctly by the basic TLS protocol;\n higher-level or application protocols may define additional\n requirements. Note that, for historical reasons, the data\n element is named using GMT, the predecessor of the current\n worldwide time base, UTC.\n\n random_bytes\n 28 bytes generated by a secure random number generator.\n\n The ClientHello message includes a variable-length session\n identifier. If not empty, the value identifies a session between the\n same client and server whose security parameters the client wishes to\n reuse. The session identifier MAY be from an earlier connection,\n\n\n\nDierks & Rescorla Standards Track [Page 39]\n\f\nRFC 5246 TLS August 2008\n\n\n this connection, or from another currently active connection. The\n second option is useful if the client only wishes to update the\n random structures and derived values of a connection, and the third\n option makes it possible to establish several independent secure\n connections without repeating the full handshake protocol. These\n independent connections may occur sequentially or simultaneously; a\n SessionID becomes valid when the handshake negotiating it completes\n with the exchange of Finished messages and persists until it is\n removed due to aging or because a fatal error was encountered on a\n connection associated with the session. The actual contents of the\n SessionID are defined by the server.\n\n opaque SessionID<0..32>;\n\n Warning: Because the SessionID is transmitted without encryption or\n immediate MAC protection, servers MUST NOT place confidential\n information in session identifiers or let the contents of fake\n session identifiers cause any breach of security. (Note that the\n content of the handshake as a whole, including the SessionID, is\n protected by the Finished messages exchanged at the end of the\n handshake.)\n\n The cipher suite list, passed from the client to the server in the\n ClientHello message, contains the combinations of cryptographic\n algorithms supported by the client in order of the client's\n preference (favorite choice first). Each cipher suite defines a key\n exchange algorithm, a bulk encryption algorithm (including secret key\n length), a MAC algorithm, and a PRF. The server will select a cipher\n suite or, if no acceptable choices are presented, return a handshake\n failure alert and close the connection. If the list contains cipher\n suites the server does not recognize, support, or wish to use, the\n server MUST ignore those cipher suites, and process the remaining\n ones as usual.\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n The ClientHello includes a list of compression algorithms supported\n by the client, ordered according to the client's preference.\n\n enum { null(0), (255) } CompressionMethod;\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 40]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ProtocolVersion client_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suites<2..2^16-2>;\n CompressionMethod compression_methods<1..2^8-1>;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ClientHello;\n\n TLS allows extensions to follow the compression_methods field in an\n extensions block. The presence of extensions can be detected by\n determining whether there are bytes following the compression_methods\n at the end of the ClientHello. Note that this method of detecting\n optional data differs from the normal TLS method of having a\n variable-length field, but it is used for compatibility with TLS\n before extensions were defined.\n\n client_version\n The version of the TLS protocol by which the client wishes to\n communicate during this session. This SHOULD be the latest\n (highest valued) version supported by the client. For this\n version of the specification, the version will be 3.3 (see\n Appendix E for details about backward compatibility).\n\n random\n A client-generated random structure.\n\n session_id\n The ID of a session the client wishes to use for this connection.\n This field is empty if no session_id is available, or if the\n client wishes to generate new security parameters.\n\n cipher_suites\n This is a list of the cryptographic options supported by the\n client, with the client's first preference first. If the\n session_id field is not empty (implying a session resumption\n request), this vector MUST include at least the cipher_suite from\n that session. Values are defined in Appendix A.5.\n\n compression_methods\n This is a list of the compression methods supported by the client,\n sorted by client preference. If the session_id field is not empty\n (implying a session resumption request), it MUST include the\n\n\n\nDierks & Rescorla Standards Track [Page 41]\n\f\nRFC 5246 TLS August 2008\n\n\n compression_method from that session. This vector MUST contain,\n and all implementations MUST support, CompressionMethod.null.\n Thus, a client and server will always be able to agree on a\n compression method.\n\n extensions\n Clients MAY request extended functionality from servers by sending\n data in the extensions field. The actual \"Extension\" format is\n defined in Section 7.4.1.4.\n\n In the event that a client requests additional functionality using\n extensions, and this functionality is not supplied by the server, the\n client MAY abort the handshake. A server MUST accept ClientHello\n messages both with and without the extensions field, and (as for all\n other messages) it MUST check that the amount of data in the message\n precisely matches one of these formats; if not, then it MUST send a\n fatal \"decode_error\" alert.\n\n After sending the ClientHello message, the client waits for a\n ServerHello message. Any handshake message returned by the server,\n except for a HelloRequest, is treated as a fatal error.\n\n7.4.1.3. Server Hello\n\n When this message will be sent:\n\n The server will send this message in response to a ClientHello\n message when it was able to find an acceptable set of algorithms.\n If it cannot find such a match, it will respond with a handshake\n failure alert.\n\n Structure of this message:\n\n struct {\n ProtocolVersion server_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ServerHello;\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 42]\n\f\nRFC 5246 TLS August 2008\n\n\n The presence of extensions can be detected by determining whether\n there are bytes following the compression_method field at the end of\n the ServerHello.\n\n server_version\n This field will contain the lower of that suggested by the client\n in the client hello and the highest supported by the server. For\n this version of the specification, the version is 3.3. (See\n Appendix E for details about backward compatibility.)\n\n random\n This structure is generated by the server and MUST be\n independently generated from the ClientHello.random.\n\n session_id\n This is the identity of the session corresponding to this\n connection. If the ClientHello.session_id was non-empty, the\n server will look in its session cache for a match. If a match is\n found and the server is willing to establish the new connection\n using the specified session state, the server will respond with\n the same value as was supplied by the client. This indicates a\n resumed session and dictates that the parties must proceed\n directly to the Finished messages. Otherwise, this field will\n contain a different value identifying the new session. The server\n may return an empty session_id to indicate that the session will\n not be cached and therefore cannot be resumed. If a session is\n resumed, it must be resumed using the same cipher suite it was\n originally negotiated with. Note that there is no requirement\n that the server resume any session even if it had formerly\n provided a session_id. Clients MUST be prepared to do a full\n negotiation -- including negotiating new cipher suites -- during\n any handshake.\n\n cipher_suite\n The single cipher suite selected by the server from the list in\n ClientHello.cipher_suites. For resumed sessions, this field is\n the value from the state of the session being resumed.\n\n compression_method\n The single compression algorithm selected by the server from the\n list in ClientHello.compression_methods. For resumed sessions,\n this field is the value from the resumed session state.\n\n extensions\n A list of extensions. Note that only extensions offered by the\n client can appear in the server's list.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 43]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.1.4. Hello Extensions\n\n The extension format is:\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n signature_algorithms(13), (65535)\n } ExtensionType;\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The initial set of extensions is defined in a companion document\n [TLSEXT]. The list of extension types is maintained by IANA as\n described in Section 12.\n\n An extension type MUST NOT appear in the ServerHello unless the same\n extension type appeared in the corresponding ClientHello. If a\n client receives an extension type in ServerHello that it did not\n request in the associated ClientHello, it MUST abort the handshake\n with an unsupported_extension fatal alert.\n\n Nonetheless, \"server-oriented\" extensions may be provided in the\n future within this framework. Such an extension (say, of type x)\n would require the client to first send an extension of type x in a\n ClientHello with empty extension_data to indicate that it supports\n the extension type. In this case, the client is offering the\n capability to understand the extension type, and the server is taking\n the client up on its offer.\n\n When multiple extensions of different types are present in the\n ClientHello or ServerHello messages, the extensions MAY appear in any\n order. There MUST NOT be more than one extension of the same type.\n\n Finally, note that extensions can be sent both when starting a new\n session and when requesting session resumption. Indeed, a client\n that requests session resumption does not in general know whether the\n server will accept this request, and therefore it SHOULD send the\n same extensions as it would send if it were not attempting\n resumption.\n\n\n\nDierks & Rescorla Standards Track [Page 44]\n\f\nRFC 5246 TLS August 2008\n\n\n In general, the specification of each extension type needs to\n describe the effect of the extension both during full handshake and\n session resumption. Most current TLS extensions are relevant only\n when a session is initiated: when an older session is resumed, the\n server does not process these extensions in Client Hello, and does\n not include them in Server Hello. However, some extensions may\n specify different behavior during session resumption.\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions, and some are simply refusals to support particular\n features. In general, error alerts should be used for the former,\n and a field in the server extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem.\n\n Often the fact that the extension fields are included in the\n inputs to the Finished message hashes will be sufficient, but\n extreme care is needed when the extension changes the meaning of\n messages sent in the handshake phase. Designers and implementors\n should be aware of the fact that until the handshake has been\n authenticated, active attackers can modify messages and insert,\n remove, or replace extensions.\n\n - It would be technically possible to use extensions to change major\n aspects of the design of TLS; for example the design of cipher\n suite negotiation. This is not recommended; it would be more\n appropriate to define a new version of TLS -- particularly since\n the TLS handshake algorithms have specific protection against\n version rollback attacks based on the version number, and the\n possibility of version rollback should be a significant\n consideration in any major design change.\n\n7.4.1.4.1. Signature Algorithms\n\n The client uses the \"signature_algorithms\" extension to indicate to\n the server which signature/hash algorithm pairs may be used in\n digital signatures. The \"extension_data\" field of this extension\n contains a \"supported_signature_algorithms\" value.\n\n\n\nDierks & Rescorla Standards Track [Page 45]\n\f\nRFC 5246 TLS August 2008\n\n\n enum {\n none(0), md5(1), sha1(2), sha224(3), sha256(4), sha384(5),\n sha512(6), (255)\n } HashAlgorithm;\n\n enum { anonymous(0), rsa(1), dsa(2), ecdsa(3), (255) }\n SignatureAlgorithm;\n\n struct {\n HashAlgorithm hash;\n SignatureAlgorithm signature;\n } SignatureAndHashAlgorithm;\n\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2..2^16-2>;\n\n Each SignatureAndHashAlgorithm value lists a single hash/signature\n pair that the client is willing to verify. The values are indicated\n in descending order of preference.\n\n Note: Because not all signature algorithms and hash algorithms may be\n accepted by an implementation (e.g., DSA with SHA-1, but not\n SHA-256), algorithms here are listed in pairs.\n\n hash\n This field indicates the hash algorithm which may be used. The\n values indicate support for unhashed data, MD5 [MD5], SHA-1,\n SHA-224, SHA-256, SHA-384, and SHA-512 [SHS], respectively. The\n \"none\" value is provided for future extensibility, in case of a\n signature algorithm which does not require hashing before signing.\n\n signature\n This field indicates the signature algorithm that may be used.\n The values indicate anonymous signatures, RSASSA-PKCS1-v1_5\n [PKCS1] and DSA [DSS], and ECDSA [ECDSA], respectively. The\n \"anonymous\" value is meaningless in this context but used in\n Section 7.4.3. It MUST NOT appear in this extension.\n\n The semantics of this extension are somewhat complicated because the\n cipher suite indicates permissible signature algorithms but not hash\n algorithms. Sections 7.4.2 and 7.4.3 describe the appropriate rules.\n\n If the client supports only the default hash and signature algorithms\n (listed in this section), it MAY omit the signature_algorithms\n extension. If the client does not support the default algorithms, or\n supports other hash and signature algorithms (and it is willing to\n use them for verifying messages sent by the server, i.e., server\n certificates and server key exchange), it MUST send the\n\n\n\nDierks & Rescorla Standards Track [Page 46]\n\f\nRFC 5246 TLS August 2008\n\n\n signature_algorithms extension, listing the algorithms it is willing\n to accept.\n\n If the client does not send the signature_algorithms extension, the\n server MUST do the following:\n\n - If the negotiated key exchange algorithm is one of (RSA, DHE_RSA,\n DH_RSA, RSA_PSK, ECDH_RSA, ECDHE_RSA), behave as if client had\n sent the value {sha1,rsa}.\n\n - If the negotiated key exchange algorithm is one of (DHE_DSS,\n DH_DSS), behave as if the client had sent the value {sha1,dsa}.\n\n - If the negotiated key exchange algorithm is one of (ECDH_ECDSA,\n ECDHE_ECDSA), behave as if the client had sent value {sha1,ecdsa}.\n\n Note: this is a change from TLS 1.1 where there are no explicit\n rules, but as a practical matter one can assume that the peer\n supports MD5 and SHA-1.\n\n Note: this extension is not meaningful for TLS versions prior to 1.2.\n Clients MUST NOT offer it if they are offering prior versions.\n However, even if clients do offer it, the rules specified in [TLSEXT]\n require servers to ignore extensions they do not understand.\n\n Servers MUST NOT send this extension. TLS servers MUST support\n receiving this extension.\n\n When performing session resumption, this extension is not included in\n Server Hello, and the server ignores the extension in Client Hello\n (if present).\n\n7.4.2. Server Certificate\n\n When this message will be sent:\n\n The server MUST send a Certificate message whenever the agreed-\n upon key exchange method uses certificates for authentication\n (this includes all key exchange methods defined in this document\n except DH_anon). This message will always immediately follow the\n ServerHello message.\n\n Meaning of this message:\n\n This message conveys the server's certificate chain to the client.\n\n The certificate MUST be appropriate for the negotiated cipher\n suite's key exchange algorithm and any negotiated extensions.\n\n\n\nDierks & Rescorla Standards Track [Page 47]\n\f\nRFC 5246 TLS August 2008\n\n\n Structure of this message:\n\n opaque ASN.1Cert<1..2^24-1>;\n\n struct {\n ASN.1Cert certificate_list<0..2^24-1>;\n } Certificate;\n\n certificate_list\n This is a sequence (chain) of certificates. The sender's\n certificate MUST come first in the list. Each following\n certificate MUST directly certify the one preceding it. Because\n certificate validation requires that root keys be distributed\n independently, the self-signed certificate that specifies the root\n certificate authority MAY be omitted from the chain, under the\n assumption that the remote end must already possess it in order to\n validate it in any case.\n\n The same message type and structure will be used for the client's\n response to a certificate request message. Note that a client MAY\n send no certificates if it does not have an appropriate certificate\n to send in response to the server's authentication request.\n\n Note: PKCS #7 [PKCS7] is not used as the format for the certificate\n vector because PKCS #6 [PKCS6] extended certificates are not used.\n Also, PKCS #7 defines a SET rather than a SEQUENCE, making the task\n of parsing the list more difficult.\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3, unless explicitly negotiated\n otherwise (e.g., [TLSPGP]).\n\n - The end entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected key exchange\n algorithm.\n\n Key Exchange Alg. Certificate Key Type\n\n RSA RSA public key; the certificate MUST allow the\n RSA_PSK key to be used for encryption (the\n keyEncipherment bit MUST be set if the key\n usage extension is present).\n Note: RSA_PSK is defined in [TLSPSK].\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 48]\n\f\nRFC 5246 TLS August 2008\n\n\n DHE_RSA RSA public key; the certificate MUST allow the\n ECDHE_RSA key to be used for signing (the\n digitalSignature bit MUST be set if the key\n usage extension is present) with the signature\n scheme and hash algorithm that will be employed\n in the server key exchange message.\n Note: ECDHE_RSA is defined in [TLSECC].\n\n DHE_DSS DSA public key; the certificate MUST allow the\n key to be used for signing with the hash\n algorithm that will be employed in the server\n key exchange message.\n\n DH_DSS Diffie-Hellman public key; the keyAgreement bit\n DH_RSA MUST be set if the key usage extension is\n present.\n\n ECDH_ECDSA ECDH-capable public key; the public key MUST\n ECDH_RSA use a curve and point format supported by the\n client, as described in [TLSECC].\n\n ECDHE_ECDSA ECDSA-capable public key; the certificate MUST\n allow the key to be used for signing with the\n hash algorithm that will be employed in the\n server key exchange message. The public key\n MUST use a curve and point format supported by\n the client, as described in [TLSECC].\n\n - The \"server_name\" and \"trusted_ca_keys\" extensions [TLSEXT] are\n used to guide certificate selection.\n\n\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) Y. Pettersen\nRequest for Comments: 6961 June 2013\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\n\nAbstract\n\n This document defines the Transport Layer Security (TLS) Certificate\n Status Version 2 Extension to allow clients to specify and support\n several certificate status methods. (The use of the Certificate\n Status extension is commonly referred to as \"OCSP stapling\".) Also\n defined is a new method based on the Online Certificate Status\n Protocol (OCSP) that servers can use to provide status information\n about not only the server's own certificate but also the status of\n intermediate certificates in the chain.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6961.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 1]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nCopyright Notice\n\n Copyright (c) 2013 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n1. Introduction\n\n The Transport Layer Security (TLS) Extension [RFC6066] framework\n defines, among other extensions, the Certificate Status extension\n (also referred to as \"OCSP stapling\") that clients can use to request\n the server's copy of the current status of its certificate. The\n benefits of this extension include a reduced number of roundtrips and\n network delays for the client to verify the status of the server's\n certificate and a reduced load on the certificate issuer's status\n response servers, thus solving a problem that can become significant\n when the issued certificate is presented by a frequently visited\n server.\n\n There are two problems with the existing Certificate Status\n extension. First, it does not provide functionality to request the\n status information about intermediate Certification Authority (CA)\n certificates, which means the client has to request status\n information through other methods, such as Certificate Revocation\n Lists (CRLs), introducing further delays. Second, the current format\n of the extension and requirements in the TLS protocol prevent a\n client from offering the server multiple status methods.\n\n\n\nPettersen Standards Track [Page 2]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n Many CAs are now issuing intermediate CA certificates that not only\n specify the publication point for their CRLs in a CRL Distribution\n Point [RFC5280] but also specify a URL for their OCSP [RFC6960]\n server in Authority Information Access [RFC5280]. Given that\n client-cached CRLs are frequently out of date, clients would benefit\n from using OCSP to access up-to-date status information about\n intermediate CA certificates. The benefit to the issuing CA is less\n clear, as providing the bandwidth for the OCSP responder can be\n costly, especially for CAs with many high-traffic subscriber sites,\n and this cost is a concern for many CAs. There are cases where OCSP\n requests for a single high-traffic site caused significant network\n problems for the issuing CA.\n\n Clients will benefit from the TLS server providing certificate status\n information regardless of type, not just for the server certificate\n but also for the intermediate CA certificates. Combining the status\n checks into one extension will reduce the roundtrips needed to\n complete the handshake by the client to just those needed for\n negotiating the TLS connection. Also, for the Certification\n Authorities, the load on their servers will depend on the number of\n certificates they have issued, not on the number of visitors to those\n sites. Additionally, using this extension significantly reduces\n privacy concerns around the clients informing the certificate issuer\n about which sites they are visiting.\n\n For such a new system to be introduced seamlessly, clients need to be\n able to indicate support for the existing OCSP Certificate Status\n method and a new multiple-OCSP mode.\n\n Unfortunately, the definition of the Certificate Status extension\n only allows a single Certificate Status extension to be defined in a\n single extension record in the handshake, and the TLS protocol\n [RFC5246] only allows a single record in the extension list for any\n given extension. This means that it is not possible for clients to\n indicate support for new methods while still supporting older\n methods, which would cause problems for interoperability between\n newer clients and older servers. This will not just be an issue for\n the multiple status request mode proposed above but also for any\n other future status methods that might be introduced. This will be\n true not just for the current PKIX infrastructure [RFC5280] but also\n for alternative PKI structures.\n\n The solution to this problem is to define a new extension,\n \"status_request_v2\", with an extended format that allows the client\n to indicate support for multiple status request methods. This is\n implemented using a list of CertificateStatusRequestItemV2 records in\n the extension record. As the server will select the single status\n\n\n\n\nPettersen Standards Track [Page 3]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n method based on the selected cipher suite and the certificate\n presented, no significant changes are needed in the server's\n extension format.\n\n1.1. Requirements Language\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [RFC2119].\n\n1.2. Presentation Language\n\n This document defines protocol structures using the same conventions\n and presentation language as defined in Section 4 of [RFC5246].\n\n2. Multiple Certificate Status Extension\n\n2.1. New Extension\n\n The extension defined by this document is indicated by\n \"status_request_v2\" in the ExtensionType enum (originally defined by\n [RFC6066]), which uses the following value:\n\n enum {\n status_request_v2(17), (65535)\n } ExtensionType;\n\n2.2. Multiple Certificate Status Request Record\n\n Clients that support a certificate status protocol like OCSP may send\n the \"status_request_v2\" extension to the server in order to use the\n TLS handshake to transfer such data instead of downloading it through\n separate connections. When using this extension, the\n \"extension_data\" field (defined in Section 7.4.1.4 of [RFC5246]) of\n the extension SHALL contain a CertificateStatusRequestListV2 where:\n\n struct {\n CertificateStatusType status_type;\n uint16 request_length; /* Length of request field in bytes */\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n case ocsp_multi: OCSPStatusRequest;\n } request;\n } CertificateStatusRequestItemV2;\n\n enum { ocsp(1), ocsp_multi(2), (255) } CertificateStatusType;\n\n\n\n\n\nPettersen Standards Track [Page 4]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n struct {\n CertificateStatusRequestItemV2\n certificate_status_req_list<1..2^16-1>;\n } CertificateStatusRequestListV2;\n\n In the OCSPStatusRequest (originally defined by [RFC6066]), the\n \"ResponderID\" provides a list of OCSP responders that the client\n trusts. A zero-length \"responder_id_list\" sequence has the special\n meaning that the responders are implicitly known to the server, e.g.,\n by prior arrangement, or are identified by the certificates used by\n the server. \"Extensions\" is a DER encoding [X.690] of the OCSP\n request extensions, and if the server supports the forwarding of OCSP\n request extensions, this value MUST be forwarded without\n modification.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC6960]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a DER-encoded zero-length ASN.1 SEQUENCE,\n which is not valid for the \"Extensions\" type).\n\n Servers that support a client's selection of responders using the\n ResponderID field could implement this selection by matching the\n responder ID values from the client's list with the ResponderIDs of\n known OCSP responders, either by using a binary compare of the values\n or a hash calculation and compare method.\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement). This\n has been clarified in [RFC6960].\n\n The items in the list of CertificateStatusRequestItemV2 entries are\n ordered according to the client's preference (favorite choice first).\n\n A server that receives a client hello containing the\n \"status_request_v2\" extension MAY return a suitable certificate\n status response message to the client along with the server's\n\n\n\nPettersen Standards Track [Page 5]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n certificate message. If OCSP is requested, it SHOULD use the\n information contained in the extension when selecting an OCSP\n responder and SHOULD include request_extensions in the OCSP request.\n\n The server returns a certificate status response along with its\n certificate by sending a \"CertificateStatus\" message (originally\n defined by [RFC6066]) immediately after the \"Certificate\" message\n (Section 7.4.2 of [RFC5246]) (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message in response to a \"status_request_v2\"\n request, then the server MUST have included an extension of type\n \"status_request_v2\" with empty \"extension_data\" in the extended\n server hello.\n\n The \"CertificateStatus\" message is conveyed using the handshake\n message type \"certificate_status\" (defined in [RFC6066]) as follows\n (updated from the definition in [RFC6066]):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n case ocsp_multi: OCSPResponseList;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<0..2^24-1>;\n\n struct {\n OCSPResponse ocsp_response_list<1..2^24-1>;\n } OCSPResponseList;\n\n An \"OCSPResponse\" element (originally defined by [RFC6066]) contains\n a complete, DER-encoded OCSP response (using the ASN.1 [X.680] type\n OCSPResponse defined in [RFC6960]). Only one OCSP response, with a\n length of at least one byte, may be sent for status_type \"ocsp\".\n\n An \"ocsp_response_list\" contains a list of \"OCSPResponse\" elements,\n as specified above, each containing the OCSP response for the\n matching corresponding certificate in the server's Certificate TLS\n handshake message. That is, the first entry is the OCSP response for\n the first certificate in the Certificate list, the second entry is\n the response for the second certificate, and so on. The list MAY\n contain fewer OCSP responses than there were certificates in the\n Certificate handshake message, but there MUST NOT be more responses\n than there were certificates in the list. Individual elements of the\n list MAY have a length of 0 (zero) bytes if the server does not have\n the OCSP response for that particular certificate stored, in which\n\n\n\nPettersen Standards Track [Page 6]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n case the client MUST act as if a response was not received for that\n particular certificate. If the client receives a\n \"ocsp_response_list\" that does not contain a response for one or more\n of the certificates in the completed certificate chain, the client\n SHOULD attempt to validate the certificate using an alternative\n retrieval method, such as downloading the relevant CRL; OCSP SHOULD\n in this situation only be used for the end-entity certificate, not\n intermediate CA certificates, for reasons stated above.\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if it has received a \"status_request_v2\" extension in\n the client hello message and has sent a \"status_request_v2\" extension\n in the server hello message. Additionally, note that a server MUST\n NOT send the \"CertificateStatus\" message unless it received either a\n \"status_request\" or \"status_request_v2\" extension in the client hello\n message and sent a corresponding \"status_request\" or\n \"status_request_v2\" extension in the server hello message.\n\n Clients requesting an OCSP response and receiving one or more OCSP\n responses in a \"CertificateStatus\" message MUST check the OCSP\n response(s) and abort the handshake if the response is a \"revoked\"\n status or other unacceptable responses (as determined by client\n policy) with a bad_certificate_status_response(113) alert. This\n alert is always fatal.\n\n If the OCSP response received from the server does not result in a\n definite \"good\" or \"revoked\" status, it is inconclusive. A TLS\n client in such a case MAY check the validity of the server\n certificate through other means, e.g., by directly querying the\n certificate issuer. If such processing still results in an\n inconclusive response, then the application using the TLS connection\n will have to decide whether to close the connection or not. Note\n that this problem cannot be decided by the generic TLS client code\n without information from the application. If the application doesn't\n provide any such information, then the client MUST abort the\n connection, since the server certificate has not been sufficiently\n validated.\n\n An example of where the application might wish to continue is with\n EAP-TLS (Extensible Authentication Protocol - TLS), where the\n application can use another mechanism to check the status of a\n certificate once it obtains network access. In this case, the\n application could have the client continue with the handshake, but it\n MUST NOT disclose a username and password until it has fully\n validated the server certificate.\n\n\n\n\n\n\nPettersen Standards Track [Page 7]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n3. IANA Considerations\n\n Section 2.1 defines the new TLS extension status_request_v2 (17)\n enum, which has been added to the \"ExtensionType Values\" list in the\n IANA \"Transport Layer Security (TLS) Extensions\" registry.\n\n Section 2.2 describes a TLS CertificateStatusType registry that is\n now maintained by IANA. The \"TLS Certificate Status Types\" registry\n has been created under the \"Transport Layer Security (TLS)\n Extensions\" registry. CertificateStatusType values are to be\n assigned via IETF Review as defined in [RFC5226]. The initial\n registry corresponds to the definition of \"CertificateStatusType\" in\n Section 2.2.\n\n Value Description Reference\n -----------------------------------------\n 0 Reserved [RFC6961]\n 1 ocsp [RFC6066] [RFC6961]\n 2 ocsp_multi [RFC6961]\n 3-255 Unassigned\n\n4. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security considerations for the particular extension\n specified in this document are given below. In general, implementers\n should continue to monitor the state of the art and address any\n weaknesses identified.\n\n4.1. Security Considerations for status_request_v2\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC6960] for further details.\n\n This extension allows the client to send arbitrary data to the\n server. The server implementers need to handle such data carefully\n to avoid introducing security vulnerabilities.\n\n The security considerations of [RFC6960] apply to OCSP requests and\n responses.\n\n\n\n\nPettersen Standards Track [Page 8]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n5. Acknowledgements\n\n This document is based on [RFC6066], authored by Donald Eastlake 3rd.\n\n6. References\n\n6.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246, August 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC6066] Eastlake, D., \"Transport Layer Security (TLS) Extensions:\n Extension Definitions\", RFC 6066, January 2011.\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, June 2013.\n\n [X.680] ITU-T Recommendation X.680 (2008) | ISO/IEC 8824-1:2008,\n \"Abstract Syntax Notation One (ASN.1): Specification of\n basic notation\", November 2008.\n\n [X.690] ITU-T Recommendation X.690 (2008) | ISO/IEC 8825-1:2008,\n \"ASN.1 encoding rules: Specification of Basic Encoding\n Rules (BER), Canonical Encoding Rules (CER) and\n Distinguished Encoding Rules (DER)\", November 2008.\n\n6.2. Informative References\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and C.\n Adams, \"X.509 Internet Public Key Infrastructure Online\n Certificate Status Protocol - OCSP\", RFC 2560, June 1999.\n\n\n\n\n\n\nPettersen Standards Track [Page 9]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nAuthor's Address\n\n Yngve N. Pettersen\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 10]\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 5705 RTFM, Inc.\nCategory: Standards Track March 2010\nISSN: 2070-1721\n\n\n Keying Material Exporters for Transport Layer Security (TLS)\n\nAbstract\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n to perform key establishment but then use some of the keying material\n for their own purposes. This document describes a general mechanism\n for allowing that.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/5705.\n\nCopyright Notice\n\n Copyright (c) 2010 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2\n 2. Conventions Used In This Document . . . . . . . . . . . . . . . 3\n 3. Binding to Application Contexts . . . . . . . . . . . . . . . . 3\n 4. Exporter Definition . . . . . . . . . . . . . . . . . . . . . . 4\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . . 5\n 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 6\n 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . 6\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 7\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 7\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 7\n\n1. Introduction\n\n Note: The mechanism described in this document was previously known\n as \"TLS Extractors\" but was changed to avoid a name conflict\n with the use of the term \"Extractor\" in the cryptographic\n community.\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n [RFC5246] or Datagram TLS (DTLS) [RFC4347] to perform key\n establishment but then use some of the keying material for their own\n purposes. A typical example is DTLS-SRTP [DTLS-SRTP], a key\n management scheme for the Secure Real-time Transport Protocol (SRTP)\n that uses DTLS to perform a key exchange and negotiate the SRTP\n [RFC3711] protection suite and then uses the DTLS master_secret to\n generate the SRTP keys.\n\n These applications imply a need to be able to export keying material\n (later called Exported Keying Material or EKM) from TLS/DTLS to an\n application or protocol residing at an upper layer, and to securely\n agree on the upper-layer context where the keying material will be\n used. The mechanism for exporting the keying material has the\n following requirements:\n\n o Both client and server need to be able to export the same EKM\n value.\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n o EKM values should be indistinguishable from random data to\n attackers who don't know the master_secret.\n\n o It should be possible to export multiple EKM values from the same\n TLS/DTLS association.\n\n o Knowing one EKM value should not reveal any useful information\n about the master_secret or about other EKM values.\n\n The mechanism described in this document is intended to fulfill these\n requirements. This mechanism is compatible with all versions of TLS.\n\n2. Conventions Used In This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Binding to Application Contexts\n\n In addition to using an exporter to obtain keying material, an\n application using the keying material has to securely establish the\n upper-layer context where the keying material will be used. The\n details of this context depend on the application, but it could\n include things such as algorithms and parameters that will be used\n with the keys, identifier(s) for the endpoint(s) who will use the\n keys, identifier(s) for the session(s) where the keys will be used,\n and the lifetime(s) for the context and/or keys. At a minimum, there\n should be some mechanism for signaling that an exporter will be used.\n\n This specification does not mandate a single mechanism for agreeing\n on such context; instead, there are several possibilities that can be\n used (and can complement each other). For example:\n\n o Information about the upper-layer context can be included in the\n optional data after the exporter label (see Section 4).\n\n o Information about the upper-layer context can be exchanged in TLS\n extensions included in the ClientHello and ServerHello messages.\n This approach is used in [DTLS-SRTP]. The handshake messages are\n protected by the Finished messages, so once the handshake\n completes, the peers will have the same view of the information.\n Extensions also allow a limited form of negotiation: for example,\n the TLS client could propose several alternatives for some context\n parameters, and the TLS server could select one of them.\n\n o The upper-layer protocol can include its own handshake, which can\n be protected using the keys exported by TLS.\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n No matter how the context is agreed, it is required that it has one\n part that indicates which application will use the exported keys.\n This part is the disambiguating label string (see Section 4).\n\n It is important to note that just embedding TLS messages in the\n upper-layer protocol may not automatically secure all the important\n context information, since the upper-layer messages are not covered\n by TLS Finished messages.\n\n4. Exporter Definition\n\n The output of the exporter is intended to be used in a single scope,\n which is associated with the TLS session, the label, and the context\n value.\n\n The exporter takes three input values:\n\n o a disambiguating label string,\n\n o a per-association context value provided by the application using\n the exporter, and\n\n o a length value.\n\n If no context is provided, it then computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random\n )[length]\n\n If context is provided, it computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random +\n context_value_length + context_value\n )[length]\n\n Where PRF is the TLS Pseudorandom Function in use for the session.\n The output is a pseudorandom bit string of length bytes generated\n from the master_secret. (This construction allows for\n interoperability with older exporter-type constructions which do not\n use context values, e.g., [RFC5281]).\n\n Labels here have the same definition as in TLS, i.e., an ASCII string\n with no terminating NULL. Label values beginning with \"EXPERIMENTAL\"\n MAY be used for private use without registration. All other label\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n values MUST be registered via Specification Required as described by\n RFC 5226 [RFC5226]. Note that exporter labels have the potential to\n collide with existing PRF labels. In order to prevent this, labels\n SHOULD begin with \"EXPORTER\". This is not a MUST because there are\n existing uses that have labels which do not begin with this prefix.\n\n The context value allows the application using the exporter to mix\n its own data with the TLS PRF for the exporter output. One example\n of where this might be useful is an authentication setting where the\n client credentials are valid for more than one identity; the context\n value could then be used to mix the expected identity into the keying\n material, thus preventing substitution attacks. The context value\n length is encoded as an unsigned, 16-bit quantity (uint16; see\n [RFC5246], Section 4.4) representing the length of the context value.\n The context MAY be zero length. Because the context value is mixed\n with the master_secret via the PRF, it is safe to mix confidential\n information into the exporter, provided that the master_secret will\n not be known to the attacker.\n\n5. Security Considerations\n\n The prime security requirement for exporter outputs is that they be\n independent. More formally, after a particular TLS session, if an\n adversary is allowed to choose multiple (label, context value) pairs\n and is given the output of the PRF for those values, the attacker is\n still unable to distinguish between the output of the PRF for a\n (label, context value) pair (different from the ones that it\n submitted) and a random value of the same length. In particular,\n there may be settings, such as the one described in Section 4, where\n the attacker can control the context value; such an attacker MUST NOT\n be able to predict the output of the exporter. Similarly, an\n attacker who does not know the master secret should not be able to\n distinguish valid exporter outputs from random values. The current\n set of TLS PRFs is believed to meet this objective, provided the\n master secret is randomly generated.\n\n Because an exporter produces the same value if applied twice with the\n same label to the same master_secret, it is critical that two EKM\n values generated with the same label not be used for two different\n purposes -- hence, the requirement for IANA registration. However,\n because exporters depend on the TLS PRF, it is not a threat to the\n use of an EKM value generated from one label to reveal an EKM value\n generated from another label.\n\n With certain TLS cipher suites, the TLS master secret is not\n necessarily unique to a single TLS session. In particular, with RSA\n key exchange, a malicious party acting as TLS server in one session\n and as TLS client in another session can cause those two sessions to\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n have the same TLS master secret (though the sessions must be\n established simultaneously to get adequate control of the Random\n values). Applications using the EKM need to consider this in how\n they use the EKM; in some cases, requiring the use of other cipher\n suites (such as those using a Diffie-Hellman key exchange) may be\n advisable.\n\n Designing a secure mechanism that uses exporters is not necessarily\n straightforward. This document only provides the exporter mechanism,\n but the problem of agreeing on the surrounding context and the\n meaning of the information passed to and from the exporter remains.\n Any new uses of the exporter mechanism should be subject to careful\n review.\n\n6. IANA Considerations\n\n IANA has created a TLS Exporter Label registry for this purpose. The\n initial contents of the registry are given below:\n\n Value Reference Note\n ----------------------------- --------- ----\n client finished [RFC5246] (1)\n server finished [RFC5246] (1)\n master secret [RFC5246] (1)\n key expansion [RFC5246] (1)\n client EAP encryption [RFC5216]\n ttls keying material [RFC5281]\n ttls challenge [RFC5281]\n\n Note: (1) These entries are reserved and MUST NOT be used for the\n purpose described in RFC 5705, in order to avoid confusion with\n similar, but distinct, use in RFC 5246.\n\n Future values are allocated via the RFC 5226 Specification Required\n policy. The label is a string consisting of printable ASCII\n characters. IANA MUST also verify that one label is not a prefix of\n any other label. For example, labels \"key\" or \"master secretary\" are\n forbidden.\n\n7. Acknowledgments\n\n Thanks to Pasi Eronen for valuable comments and for the contents of\n the IANA section and Section 3. Thanks to David McGrew for helpful\n discussion of the security considerations and to Vijay Gurbani and\n Alfred Hoenes for editorial comments.\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246,\n August 2008.\n\n8.2. Informative References\n\n [DTLS-SRTP] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for Secure\n Real-time Transport Protocol (SRTP)\", Work in Progress,\n February 2009.\n\n [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and\n K. Norrman, \"The Secure Real-time Transport Protocol\n (SRTP)\", RFC 3711, March 2004.\n\n [RFC4347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security\", RFC 4347, April 2006.\n\n [RFC5216] Simon, D., Aboba, B., and R. Hurst, \"The EAP-TLS\n Authentication Protocol\", RFC 5216, March 2008.\n\n [RFC5281] Funk, P. and S. Blake-Wilson, \"Extensible Authentication\n Protocol Tunneled Transport Layer Security Authenticated\n Protocol Version 0 (EAP-TTLSv0)\", RFC 5281, August 2008.\n\nAuthor's Address\n\n Eric Rescorla\n RTFM, Inc.\n 2064 Edgewood Drive\n Palo Alto, CA 94303\n USA\n\n EMail: [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) D. Eastlake 3rd\nRequest for Comments: 6066 Huawei\nObsoletes: 4366 January 2011\nCategory: Standards Track\nISSN: 2070-1721\n\n\n Transport Layer Security (TLS) Extensions: Extension Definitions\n\nAbstract\n\n This document provides specifications for existing TLS extensions.\n It is a companion document for RFC 5246, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\". The extensions specified are\n server_name, max_fragment_length, client_certificate_url,\n trusted_ca_keys, truncated_hmac, and status_request.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6066.\n\nCopyright Notice\n\n Copyright (c) 2011 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n\n\n\n\n\nEastlake Standards Track [Page 1]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction ....................................................3\n 1.1. Specific Extensions Covered ................................3\n 1.2. Conventions Used in This Document ..........................5\n 2. Extensions to the Handshake Protocol ............................5\n 3. Server Name Indication ..........................................6\n 4. Maximum Fragment Length Negotiation .............................8\n 5. Client Certificate URLs .........................................9\n 6. Trusted CA Indication ..........................................12\n 7. Truncated HMAC .................................................13\n 8. Certificate Status Request .....................................14\n 9. Error Alerts ...................................................16\n 10. IANA Considerations ...........................................17\n 10.1. pkipath MIME Type Registration ...........................17\n 10.2. Reference for TLS Alerts, TLS HandshakeTypes, and\n ExtensionTypes ...........................................19\n 11. Security Considerations .......................................19\n 11.1. Security Considerations for server_name ..................19\n 11.2. Security Considerations for max_fragment_length ..........20\n 11.3. Security Considerations for client_certificate_url .......20\n 11.4. Security Considerations for trusted_ca_keys ..............21\n 11.5. Security Considerations for truncated_hmac ...............21\n 11.6. Security Considerations for status_request ...............22\n 12. Normative References ..........................................22\n 13. Informative References ........................................23\n Appendix A. Changes from RFC 4366 .................................24\n Appendix B. Acknowledgements ......................................25\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 2]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1. Introduction\n\n The Transport Layer Security (TLS) Protocol Version 1.2 is specified\n in [RFC5246]. That specification includes the framework for\n extensions to TLS, considerations in designing such extensions (see\n Section 7.4.1.4 of [RFC5246]), and IANA Considerations for the\n allocation of new extension code points; however, it does not specify\n any particular extensions other than Signature Algorithms (see\n Section 7.4.1.4.1 of [RFC5246]).\n\n This document provides the specifications for existing TLS\n extensions. It is, for the most part, the adaptation and editing of\n material from RFC 4366, which covered TLS extensions for TLS 1.0 (RFC\n 2246) and TLS 1.1 (RFC 4346).\n\n1.1. Specific Extensions Covered\n\n The extensions described here focus on extending the functionality\n provided by the TLS protocol message formats. Other issues, such as\n the addition of new cipher suites, are deferred.\n\n The extension types defined in this document are:\n\n enum {\n server_name(0), max_fragment_length(1),\n client_certificate_url(2), trusted_ca_keys(3),\n truncated_hmac(4), status_request(5), (65535)\n } ExtensionType;\n\n Specifically, the extensions described in this document:\n\n - Allow TLS clients to provide to the TLS server the name of the\n server they are contacting. This functionality is desirable in\n order to facilitate secure connections to servers that host\n multiple 'virtual' servers at a single underlying network address.\n\n - Allow TLS clients and servers to negotiate the maximum fragment\n length to be sent. This functionality is desirable as a result of\n memory constraints among some clients, and bandwidth constraints\n among some access networks.\n\n - Allow TLS clients and servers to negotiate the use of client\n certificate URLs. This functionality is desirable in order to\n conserve memory on constrained clients.\n\n\n\n\n\n\n\nEastlake Standards Track [Page 3]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - Allow TLS clients to indicate to TLS servers which certification\n authority (CA) root keys they possess. This functionality is\n desirable in order to prevent multiple handshake failures\n involving TLS clients that are only able to store a small number\n of CA root keys due to memory limitations.\n\n - Allow TLS clients and servers to negotiate the use of truncated\n Message Authentication Codes (MACs). This functionality is\n desirable in order to conserve bandwidth in constrained access\n networks.\n\n - Allow TLS clients and servers to negotiate that the server sends\n the client certificate status information (e.g., an Online\n Certificate Status Protocol (OCSP) [RFC2560] response) during a\n TLS handshake. This functionality is desirable in order to avoid\n sending a Certificate Revocation List (CRL) over a constrained\n access network and therefore saving bandwidth.\n\n TLS clients and servers may use the extensions described in this\n document. The extensions are designed to be backwards compatible,\n meaning that TLS clients that support the extensions can talk to TLS\n servers that do not support the extensions, and vice versa.\n\n Note that any messages associated with these extensions that are sent\n during the TLS handshake MUST be included in the hash calculations\n involved in \"Finished\" messages.\n\n Note also that all the extensions defined in this document are\n relevant only when a session is initiated. A client that requests\n session resumption does not in general know whether the server will\n accept this request, and therefore it SHOULD send the same extensions\n as it would send if it were not attempting resumption. When a client\n includes one or more of the defined extension types in an extended\n client hello while requesting session resumption:\n\n - The server name indication extension MAY be used by the server\n when deciding whether or not to resume a session as described in\n Section 3.\n\n - If the resumption request is denied, the use of the extensions is\n negotiated as normal.\n\n - If, on the other hand, the older session is resumed, then the\n server MUST ignore the extensions and send a server hello\n containing none of the extension types. In this case, the\n functionality of these extensions negotiated during the original\n session initiation is applied to the resumed session.\n\n\n\n\nEastlake Standards Track [Page 4]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1.2. Conventions Used in This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n [RFC2119].\n\n2. Extensions to the Handshake Protocol\n\n This document specifies the use of two new handshake messages,\n \"CertificateURL\" and \"CertificateStatus\". These messages are\n described in Sections 5 and 8, respectively. The new handshake\n message structure therefore becomes:\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), certificate_url(21), certificate_status(22),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case certificate_url: CertificateURL;\n case certificate_status: CertificateStatus;\n } body;\n } Handshake;\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 5]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n3. Server Name Indication\n\n TLS does not provide a mechanism for a client to tell a server the\n name of the server it is contacting. It may be desirable for clients\n to provide this information to facilitate secure connections to\n servers that host multiple 'virtual' servers at a single underlying\n network address.\n\n In order to provide any of the server names, clients MAY include an\n extension of type \"server_name\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"ServerNameList\" where:\n\n struct {\n NameType name_type;\n select (name_type) {\n case host_name: HostName;\n } name;\n } ServerName;\n\n enum {\n host_name(0), (255)\n } NameType;\n\n opaque HostName<1..2^16-1>;\n\n struct {\n ServerName server_name_list<1..2^16-1>\n } ServerNameList;\n\n The ServerNameList MUST NOT contain more than one name of the same\n name_type. If the server understood the ClientHello extension but\n does not recognize the server name, the server SHOULD take one of two\n actions: either abort the handshake by sending a fatal-level\n unrecognized_name(112) alert or continue the handshake. It is NOT\n RECOMMENDED to send a warning-level unrecognized_name(112) alert,\n because the client's behavior in response to warning-level alerts is\n unpredictable. If there is a mismatch between the server name used\n by the client application and the server name of the credential\n chosen by the server, this mismatch will become apparent when the\n client application performs the server endpoint identification, at\n which point the client application will have to decide whether to\n proceed with the communication. TLS implementations are encouraged\n to make information available to application callers about warning-\n level alerts that were received or sent during a TLS handshake. Such\n information can be useful for diagnostic purposes.\n\n\n\n\n\nEastlake Standards Track [Page 6]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note: Earlier versions of this specification permitted multiple\n names of the same name_type. In practice, current client\n implementations only send one name, and the client cannot\n necessarily find out which name the server selected. Multiple\n names of the same name_type are therefore now prohibited.\n\n Currently, the only server names supported are DNS hostnames;\n however, this does not imply any dependency of TLS on DNS, and other\n name types may be added in the future (by an RFC that updates this\n document). The data structure associated with the host_name NameType\n is a variable-length vector that begins with a 16-bit length. For\n backward compatibility, all future data structures associated with\n new NameTypes MUST begin with a 16-bit length field. TLS MAY treat\n provided server names as opaque data and pass the names and types to\n the application.\n\n \"HostName\" contains the fully qualified DNS hostname of the server,\n as understood by the client. The hostname is represented as a byte\n string using ASCII encoding without a trailing dot. This allows the\n support of internationalized domain names through the use of A-labels\n defined in [RFC5890]. DNS hostnames are case-insensitive. The\n algorithm to compare hostnames is described in [RFC5890], Section\n 2.3.2.4.\n\n Literal IPv4 and IPv6 addresses are not permitted in \"HostName\".\n\n It is RECOMMENDED that clients include an extension of type\n \"server_name\" in the client hello whenever they locate a server by a\n supported name type.\n\n A server that receives a client hello containing the \"server_name\"\n extension MAY use the information contained in the extension to guide\n its selection of an appropriate certificate to return to the client,\n and/or other aspects of security policy. In this event, the server\n SHALL include an extension of type \"server_name\" in the (extended)\n server hello. The \"extension_data\" field of this extension SHALL be\n empty.\n\n When the server is deciding whether or not to accept a request to\n resume a session, the contents of a server_name extension MAY be used\n in the lookup of the session in the session cache. The client SHOULD\n include the same server_name extension in the session resumption\n request as it did in the full handshake that established the session.\n A server that implements this extension MUST NOT accept the request\n to resume the session if the server_name extension contains a\n different name. Instead, it proceeds with a full handshake to\n establish a new session. When resuming a session, the server MUST\n NOT include a server_name extension in the server hello.\n\n\n\nEastlake Standards Track [Page 7]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n If an application negotiates a server name using an application\n protocol and then upgrades to TLS, and if a server_name extension is\n sent, then the extension SHOULD contain the same name that was\n negotiated in the application protocol. If the server_name is\n established in the TLS session handshake, the client SHOULD NOT\n attempt to request a different server name at the application layer.\n\n4. Maximum Fragment Length Negotiation\n\n Without this extension, TLS specifies a fixed maximum plaintext\n fragment length of 2^14 bytes. It may be desirable for constrained\n clients to negotiate a smaller maximum fragment length due to memory\n limitations or bandwidth limitations.\n\n In order to negotiate smaller maximum fragment lengths, clients MAY\n include an extension of type \"max_fragment_length\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain:\n\n enum{\n 2^9(1), 2^10(2), 2^11(3), 2^12(4), (255)\n } MaxFragmentLength;\n\n whose value is the desired maximum fragment length. The allowed\n values for this field are: 2^9, 2^10, 2^11, and 2^12.\n\n Servers that receive an extended client hello containing a\n \"max_fragment_length\" extension MAY accept the requested maximum\n fragment length by including an extension of type\n \"max_fragment_length\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL contain a\n \"MaxFragmentLength\" whose value is the same as the requested maximum\n fragment length.\n\n If a server receives a maximum fragment length negotiation request\n for a value other than the allowed values, it MUST abort the\n handshake with an \"illegal_parameter\" alert. Similarly, if a client\n receives a maximum fragment length negotiation response that differs\n from the length it requested, it MUST also abort the handshake with\n an \"illegal_parameter\" alert.\n\n Once a maximum fragment length other than 2^14 has been successfully\n negotiated, the client and server MUST immediately begin fragmenting\n messages (including handshake messages) to ensure that no fragment\n larger than the negotiated length is sent. Note that TLS already\n requires clients and servers to support fragmentation of handshake\n messages.\n\n\n\n\nEastlake Standards Track [Page 8]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n The negotiated length applies for the duration of the session\n including session resumptions.\n\n The negotiated length limits the input that the record layer may\n process without fragmentation (that is, the maximum value of\n TLSPlaintext.length; see [RFC5246], Section 6.2.1). Note that the\n output of the record layer may be larger. For example, if the\n negotiated length is 2^9=512, then, when using currently defined\n cipher suites (those defined in [RFC5246] and [RFC2712]) and null\n compression, the record-layer output can be at most 805 bytes: 5\n bytes of headers, 512 bytes of application data, 256 bytes of\n padding, and 32 bytes of MAC. This means that in this event a TLS\n record-layer peer receiving a TLS record-layer message larger than\n 805 bytes MUST discard the message and send a \"record_overflow\"\n alert, without decrypting the message. When this extension is used\n with Datagram Transport Layer Security (DTLS), implementations SHOULD\n NOT generate record_overflow alerts unless the packet passes message\n authentication.\n\n5. Client Certificate URLs\n\n Without this extension, TLS specifies that when client authentication\n is performed, client certificates are sent by clients to servers\n during the TLS handshake. It may be desirable for constrained\n clients to send certificate URLs in place of certificates, so that\n they do not need to store their certificates and can therefore save\n memory.\n\n In order to negotiate sending certificate URLs to a server, clients\n MAY include an extension of type \"client_certificate_url\" in the\n (extended) client hello. The \"extension_data\" field of this\n extension SHALL be empty.\n\n (Note that it is necessary to negotiate the use of client certificate\n URLs in order to avoid \"breaking\" existing TLS servers.)\n\n Servers that receive an extended client hello containing a\n \"client_certificate_url\" extension MAY indicate that they are willing\n to accept certificate URLs by including an extension of type\n \"client_certificate_url\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n After negotiation of the use of client certificate URLs has been\n successfully completed (by exchanging hellos including\n \"client_certificate_url\" extensions), clients MAY send a\n \"CertificateURL\" message in place of a \"Certificate\" message as\n follows (see also Section 2):\n\n\n\n\nEastlake Standards Track [Page 9]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n enum {\n individual_certs(0), pkipath(1), (255)\n } CertChainType;\n\n struct {\n CertChainType type;\n URLAndHash url_and_hash_list<1..2^16-1>;\n } CertificateURL;\n\n struct {\n opaque url<1..2^16-1>;\n unint8 padding;\n opaque SHA1Hash[20];\n } URLAndHash;\n\n Here, \"url_and_hash_list\" contains a sequence of URLs and hashes.\n Each \"url\" MUST be an absolute URI reference according to [RFC3986]\n that can be immediately used to fetch the certificate(s).\n\n When X.509 certificates are used, there are two possibilities:\n\n - If CertificateURL.type is \"individual_certs\", each URL refers to a\n single DER-encoded X.509v3 certificate, with the URL for the\n client's certificate first.\n\n - If CertificateURL.type is \"pkipath\", the list contains a single\n URL referring to a DER-encoded certificate chain, using the type\n PkiPath described in Section 10.1.\n\n When any other certificate format is used, the specification that\n describes use of that format in TLS should define the encoding format\n of certificates or certificate chains, and any constraint on their\n ordering.\n\n The \"padding\" byte MUST be 0x01. It is present to make the structure\n backwards compatible.\n\n The hash corresponding to each URL is the SHA-1 hash of the\n certificate or certificate chain (in the case of X.509 certificates,\n the DER-encoded certificate or the DER-encoded PkiPath).\n\n Note that when a list of URLs for X.509 certificates is used, the\n ordering of URLs is the same as that used in the TLS Certificate\n message (see [RFC5246], Section 7.4.2), but opposite to the order in\n which certificates are encoded in PkiPath. In either case, the self-\n signed root certificate MAY be omitted from the chain, under the\n assumption that the server must already possess it in order to\n validate it.\n\n\n\nEastlake Standards Track [Page 10]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Servers receiving \"CertificateURL\" SHALL attempt to retrieve the\n client's certificate chain from the URLs and then process the\n certificate chain as usual. A cached copy of the content of any URL\n in the chain MAY be used, provided that the SHA-1 hash matches the\n hash of the cached copy.\n\n Servers that support this extension MUST support the 'http' URI\n scheme for certificate URLs and MAY support other schemes. Use of\n other schemes than 'http', 'https', or 'ftp' may create unexpected\n problems.\n\n If the protocol used is HTTP, then the HTTP server can be configured\n to use the Cache-Control and Expires directives described in\n [RFC2616] to specify whether and for how long certificates or\n certificate chains should be cached.\n\n The TLS server MUST NOT follow HTTP redirects when retrieving the\n certificates or certificate chain. The URLs used in this extension\n MUST NOT be chosen to depend on such redirects.\n\n If the protocol used to retrieve certificates or certificate chains\n returns a MIME-formatted response (as HTTP does), then the following\n MIME Content-Types SHALL be used: when a single X.509v3 certificate\n is returned, the Content-Type is \"application/pkix-cert\" [RFC2585],\n and when a chain of X.509v3 certificates is returned, the Content-\n Type is \"application/pkix-pkipath\" (Section 10.1).\n\n The server MUST check that the SHA-1 hash of the contents of the\n object retrieved from that URL (after decoding any MIME Content-\n Transfer-Encoding) matches the given hash. If any retrieved object\n does not have the correct SHA-1 hash, the server MUST abort the\n handshake with a bad_certificate_hash_value(114) alert. This alert\n is always fatal.\n\n Clients may choose to send either \"Certificate\" or \"CertificateURL\"\n after successfully negotiating the option to send certificate URLs.\n The option to send a certificate is included to provide flexibility\n to clients possessing multiple certificates.\n\n If a server is unable to obtain certificates in a given\n CertificateURL, it MUST send a fatal certificate_unobtainable(111)\n alert if it requires the certificates to complete the handshake. If\n the server does not require the certificates, then the server\n continues the handshake. The server MAY send a warning-level alert\n in this case. Clients receiving such an alert SHOULD log the alert\n and continue with the handshake if possible.\n\n\n\n\n\nEastlake Standards Track [Page 11]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n6. Trusted CA Indication\n\n Constrained clients that, due to memory limitations, possess only a\n small number of CA root keys may wish to indicate to servers which\n root keys they possess, in order to avoid repeated handshake\n failures.\n\n In order to indicate which CA root keys they possess, clients MAY\n include an extension of type \"trusted_ca_keys\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain \"TrustedAuthorities\" where:\n\n struct {\n TrustedAuthority trusted_authorities_list<0..2^16-1>;\n } TrustedAuthorities;\n\n struct {\n IdentifierType identifier_type;\n select (identifier_type) {\n case pre_agreed: struct {};\n case key_sha1_hash: SHA1Hash;\n case x509_name: DistinguishedName;\n case cert_sha1_hash: SHA1Hash;\n } identifier;\n } TrustedAuthority;\n\n enum {\n pre_agreed(0), key_sha1_hash(1), x509_name(2),\n cert_sha1_hash(3), (255)\n } IdentifierType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n Here, \"TrustedAuthorities\" provides a list of CA root key identifiers\n that the client possesses. Each CA root key is identified via\n either:\n\n - \"pre_agreed\": no CA root key identity supplied.\n\n - \"key_sha1_hash\": contains the SHA-1 hash of the CA root key. For\n Digital Signature Algorithm (DSA) and Elliptic Curve Digital\n Signature Algorithm (ECDSA) keys, this is the hash of the\n \"subjectPublicKey\" value. For RSA keys, the hash is of the big-\n endian byte string representation of the modulus without any\n initial zero-valued bytes. (This copies the key hash formats\n deployed in other environments.)\n\n\n\n\n\nEastlake Standards Track [Page 12]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - \"x509_name\": contains the DER-encoded X.509 DistinguishedName of\n the CA.\n\n - \"cert_sha1_hash\": contains the SHA-1 hash of a DER-encoded\n Certificate containing the CA root key.\n\n Note that clients may include none, some, or all of the CA root keys\n they possess in this extension.\n\n Note also that it is possible that a key hash or a Distinguished Name\n alone may not uniquely identify a certificate issuer (for example, if\n a particular CA has multiple key pairs). However, here we assume\n this is the case following the use of Distinguished Names to identify\n certificate issuers in TLS.\n\n The option to include no CA root keys is included to allow the client\n to indicate possession of some pre-defined set of CA root keys.\n\n Servers that receive a client hello containing the \"trusted_ca_keys\"\n extension MAY use the information contained in the extension to guide\n their selection of an appropriate certificate chain to return to the\n client. In this event, the server SHALL include an extension of type\n \"trusted_ca_keys\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n7. Truncated HMAC\n\n Currently defined TLS cipher suites use the MAC construction HMAC\n [RFC2104] to authenticate record-layer communications. In TLS, the\n entire output of the hash function is used as the MAC tag. However,\n it may be desirable in constrained environments to save bandwidth by\n truncating the output of the hash function to 80 bits when forming\n MAC tags.\n\n In order to negotiate the use of 80-bit truncated HMAC, clients MAY\n include an extension of type \"truncated_hmac\" in the extended client\n hello. The \"extension_data\" field of this extension SHALL be empty.\n\n Servers that receive an extended hello containing a \"truncated_hmac\"\n extension MAY agree to use a truncated HMAC by including an extension\n of type \"truncated_hmac\", with empty \"extension_data\", in the\n extended server hello.\n\n Note that if new cipher suites are added that do not use HMAC, and\n the session negotiates one of these cipher suites, this extension\n will have no effect. It is strongly recommended that any new cipher\n suites using other MACs consider the MAC size an integral part of the\n\n\n\n\nEastlake Standards Track [Page 13]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n cipher suite definition, taking into account both security and\n bandwidth considerations.\n\n If HMAC truncation has been successfully negotiated during a TLS\n handshake, and the negotiated cipher suite uses HMAC, both the client\n and the server pass this fact to the TLS record layer along with the\n other negotiated security parameters. Subsequently during the\n session, clients and servers MUST use truncated HMACs, calculated as\n specified in [RFC2104]. That is, SecurityParameters.mac_length is 10\n bytes, and only the first 10 bytes of the HMAC output are transmitted\n and checked. Note that this extension does not affect the\n calculation of the pseudo-random function (PRF) as part of\n handshaking or key derivation.\n\n The negotiated HMAC truncation size applies for the duration of the\n session including session resumptions.\n\n8. Certificate Status Request\n\n Constrained clients may wish to use a certificate-status protocol\n such as OCSP [RFC2560] to check the validity of server certificates,\n in order to avoid transmission of CRLs and therefore save bandwidth\n on constrained networks. This extension allows for such information\n to be sent in the TLS handshake, saving roundtrips and resources.\n\n In order to indicate their desire to receive certificate status\n information, clients MAY include an extension of type\n \"status_request\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"CertificateStatusRequest\" where:\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n } request;\n } CertificateStatusRequest;\n\n enum { ocsp(1), (255) } CertificateStatusType;\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n\n\n\nEastlake Standards Track [Page 14]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n In the OCSPStatusRequest, the \"ResponderIDs\" provides a list of OCSP\n responders that the client trusts. A zero-length \"responder_id_list\"\n sequence has the special meaning that the responders are implicitly\n known to the server, e.g., by prior arrangement. \"Extensions\" is a\n DER encoding of OCSP request extensions.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC2560]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a zero-length ASN.1 SEQUENCE, which is not\n valid for the \"Extensions\" type).\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement).\n\n Servers that receive a client hello containing the \"status_request\"\n extension MAY return a suitable certificate status response to the\n client along with their certificate. If OCSP is requested, they\n SHOULD use the information contained in the extension when selecting\n an OCSP responder and SHOULD include request_extensions in the OCSP\n request.\n\n Servers return a certificate response along with their certificate by\n sending a \"CertificateStatus\" message immediately after the\n \"Certificate\" message (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message, then the server MUST have included an\n extension of type \"status_request\" with empty \"extension_data\" in the\n extended server hello. The \"CertificateStatus\" message is conveyed\n using the handshake message type \"certificate_status\" as follows (see\n also Section 2):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<1..2^24-1>;\n\n An \"ocsp_response\" contains a complete, DER-encoded OCSP response\n (using the ASN.1 type OCSPResponse defined in [RFC2560]). Only one\n OCSP response may be sent.\n\n\n\n\nEastlake Standards Track [Page 15]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if has received a \"status_request\" extension in the\n client hello message and has sent a \"status_request\" extension in the\n server hello message.\n\n Note in addition that a server MUST NOT send the \"CertificateStatus\"\n message unless it received a \"status_request\" extension in the client\n hello message and sent a \"status_request\" extension in the server\n hello message.\n\n Clients requesting an OCSP response and receiving an OCSP response in\n a \"CertificateStatus\" message MUST check the OCSP response and abort\n the handshake if the response is not satisfactory with\n bad_certificate_status_response(113) alert. This alert is always\n fatal.\n\n9. Error Alerts\n\n Four new error alerts are defined for use with the TLS extensions\n defined in this document. To avoid \"breaking\" existing clients and\n servers, these alerts MUST NOT be sent unless the sending party has\n received an extended hello message from the party they are\n communicating with. These error alerts are conveyed using the\n following syntax. The new alerts are the last four, as indicated by\n the comments on the same line as the error alert number.\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n /* 41 is not defined, for historical reasons */\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction(60),\n protocol_version(70),\n insufficient_security(71),\n\n\n\nEastlake Standards Track [Page 16]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n certificate_unobtainable(111), /* new */\n unrecognized_name(112), /* new */\n bad_certificate_status_response(113), /* new */\n bad_certificate_hash_value(114), /* new */\n (255)\n } AlertDescription;\n\n \"certificate_unobtainable\" is described in Section 5.\n \"unrecognized_name\" is described in Section 3.\n \"bad_certificate_status_response\" is described in Section 8.\n \"bad_certificate_hash_value\" is described in Section 5.\n\n10. IANA Considerations\n\n IANA Considerations for TLS extensions and the creation of a registry\n are covered in Section 12 of [RFC5246] except for the registration of\n MIME type application/pkix-pkipath, which appears below.\n\n The IANA TLS extensions and MIME type application/pkix-pkipath\n registry entries that reference RFC 4366 have been updated to\n reference this document.\n\n10.1. pkipath MIME Type Registration\n\n MIME media type name: application\n MIME subtype name: pkix-pkipath\n Required parameters: none\n\n Optional parameters: version (default value is \"1\")\n\n Encoding considerations:\n Binary; this MIME type is a DER encoding of the ASN.1 type\n PkiPath, defined as follows:\n PkiPath ::= SEQUENCE OF Certificate\n PkiPath is used to represent a certification path. Within the\n sequence, the order of certificates is such that the subject of\n the first certificate is the issuer of the second certificate,\n etc.\n This is identical to the definition published in [X509-4th-TC1];\n note that it is different from that in [X509-4th].\n\n All Certificates MUST conform to [RFC5280]. (This should be\n interpreted as a requirement to encode only PKIX-conformant\n certificates using this type. It does not necessarily require\n\n\n\nEastlake Standards Track [Page 17]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n that all certificates that are not strictly PKIX-conformant must\n be rejected by relying parties, although the security consequences\n of accepting any such certificates should be considered\n carefully.)\n\n DER (as opposed to BER) encoding MUST be used. If this type is\n sent over a 7-bit transport, base64 encoding SHOULD be used.\n\n Security considerations:\n The security considerations of [X509-4th] and [RFC5280] (or any\n updates to them) apply, as well as those of any protocol that uses\n this type (e.g., TLS).\n\n Note that this type only specifies a certificate chain that can be\n assessed for validity according to the relying party's existing\n configuration of trusted CAs; it is not intended to be used to\n specify any change to that configuration.\n\n Interoperability considerations:\n No specific interoperability problems are known with this type,\n but for recommendations relating to X.509 certificates in general,\n see [RFC5280].\n\n Published specification: This document and [RFC5280].\n\n Applications that use this media type:\n TLS. It may also be used by other protocols or for general\n interchange of PKIX certificate chains.\n\n Additional information:\n Magic number(s): DER-encoded ASN.1 can be easily recognized.\n Further parsing is required to distinguish it from other ASN.1\n types.\n File extension(s): .pkipath\n Macintosh File Type Code(s): not specified\n\n Person & email address to contact for further information:\n Magnus Nystrom <[redacted-email]>\n\n Intended usage: COMMON\n\n Change controller: IESG <[redacted-email]>\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 18]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n10.2. Reference for TLS Alerts, TLS HandshakeTypes, and ExtensionTypes\n\n The following values in the TLS Alert Registry have been updated to\n reference this document:\n\n 111 certificate_unobtainable\n 112 unrecognized_name\n 113 bad_certificate_status_response\n 114 bad_certificate_hash_value\n\n The following values in the TLS HandshakeType Registry have been\n updated to reference this document:\n\n 21 certificate_url\n 22 certificate_status\n\n The following ExtensionType values have been updated to reference\n this document:\n\n 0 server_name\n 1 max_fragment_length\n 2 client_certificate_url\n 3 trusted_ca_keys\n 4 truncated_hmac\n 5 status_request\n\n11. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security Considerations for particular extensions\n specified in this document are given below.\n\n In general, implementers should continue to monitor the state of the\n art and address any weaknesses identified.\n\n11.1. Security Considerations for server_name\n\n If a single server hosts several domains, then clearly it is\n necessary for the owners of each domain to ensure that this satisfies\n their security needs. Apart from this, server_name does not appear\n to introduce significant security issues.\n\n Since it is possible for a client to present a different server_name\n in the application protocol, application server implementations that\n rely upon these names being the same MUST check to make sure the\n client did not present a different name in the application protocol.\n\n\n\n\n\nEastlake Standards Track [Page 19]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Implementations MUST ensure that a buffer overflow does not occur,\n whatever the values of the length fields in server_name.\n\n11.2. Security Considerations for max_fragment_length\n\n The maximum fragment length takes effect immediately, including for\n handshake messages. However, that does not introduce any security\n complications that are not already present in TLS, since TLS requires\n implementations to be able to handle fragmented handshake messages.\n\n Note that, as described in Section 4, once a non-null cipher suite\n has been activated, the effective maximum fragment length depends on\n the cipher suite and compression method, as well as on the negotiated\n max_fragment_length. This must be taken into account when sizing\n buffers and checking for buffer overflow.\n\n11.3. Security Considerations for client_certificate_url\n\n Support for client_certificate_url involves the server's acting as a\n client in another URI-scheme-dependent protocol. The server\n therefore becomes subject to many of the same security concerns that\n clients of the URI scheme are subject to, with the added concern that\n the client can attempt to prompt the server to connect to some\n (possibly weird-looking) URL.\n\n In general, this issue means that an attacker might use the server to\n indirectly attack another host that is vulnerable to some security\n flaw. It also introduces the possibility of denial-of-service\n attacks in which an attacker makes many connections to the server,\n each of which results in the server's attempting a connection to the\n target of the attack.\n\n Note that the server may be behind a firewall or otherwise able to\n access hosts that would not be directly accessible from the public\n Internet. This could exacerbate the potential security and denial-\n of-service problems described above, as well as allow the existence\n of internal hosts to be confirmed when they would otherwise be\n hidden.\n\n The detailed security concerns involved will depend on the URI\n schemes supported by the server. In the case of HTTP, the concerns\n are similar to those that apply to a publicly accessible HTTP proxy\n server. In the case of HTTPS, loops and deadlocks may be created,\n and this should be addressed. In the case of FTP, attacks arise that\n are similar to FTP bounce attacks.\n\n\n\n\n\n\nEastlake Standards Track [Page 20]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n As a result of this issue, it is RECOMMENDED that the\n client_certificate_url extension should have to be specifically\n enabled by a server administrator, rather than be enabled by default.\n It is also RECOMMENDED that URI schemes be enabled by the\n administrator individually, and only a minimal set of schemes be\n enabled. Unusual protocols that offer limited security or whose\n security is not well understood SHOULD be avoided.\n\n As discussed in [RFC3986], URLs that specify ports other than the\n default may cause problems, as may very long URLs (which are more\n likely to be useful in exploiting buffer overflow bugs).\n\n This extension continues to use SHA-1 (as in RFC 4366) and does not\n provide algorithm agility. The property required of SHA-1 in this\n case is second pre-image resistance, not collision resistance.\n Furthermore, even if second pre-image attacks against SHA-1 are found\n in the future, an attack against client_certificate_url would require\n a second pre-image that is accepted as a valid certificate by the\n server and contains the same public key.\n\n Also note that HTTP caching proxies are common on the Internet, and\n some proxies do not check for the latest version of an object\n correctly. If a request using HTTP (or another caching protocol)\n goes through a misconfigured or otherwise broken proxy, the proxy may\n return an out-of-date response.\n\n11.4. Security Considerations for trusted_ca_keys\n\n Potentially, the CA root keys a client possesses could be regarded as\n confidential information. As a result, the CA root key indication\n extension should be used with care.\n\n The use of the SHA-1 certificate hash alternative ensures that each\n certificate is specified unambiguously. This context does not\n require a cryptographic hash function, so the use of SHA-1 is\n considered acceptable, and no algorithm agility is provided.\n\n11.5. Security Considerations for truncated_hmac\n\n It is possible that truncated MACs are weaker than \"un-truncated\"\n MACs. However, no significant weaknesses are currently known or\n expected to exist for HMAC with MD5 or SHA-1, truncated to 80 bits.\n\n Note that the output length of a MAC need not be as long as the\n length of a symmetric cipher key, since forging of MAC values cannot\n be done off-line: in TLS, a single failed MAC guess will cause the\n immediate termination of the TLS session.\n\n\n\n\nEastlake Standards Track [Page 21]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Since the MAC algorithm only takes effect after all handshake\n messages that affect extension parameters have been authenticated by\n the hashes in the Finished messages, it is not possible for an active\n attacker to force negotiation of the truncated HMAC extension where\n it would not otherwise be used (to the extent that the handshake\n authentication is secure). Therefore, in the event that any security\n problems were found with truncated HMAC in the future, if either the\n client or the server for a given session were updated to take the\n problem into account, it would be able to veto use of this extension.\n\n11.6. Security Considerations for status_request\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC2560] for further details.\n\n12. Normative References\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC:\n Keyed-Hashing for Message Authentication\", RFC 2104,\n February 1997.\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and\n C. Adams, \"X.509 Internet Public Key Infrastructure\n Online Certificate Status Protocol - OCSP\", RFC 2560,\n June 1999.\n\n [RFC2585] Housley, R. and P. Hoffman, \"Internet X.509 Public Key\n Infrastructure Operational Protocols: FTP and HTTP\",\n RFC 2585, May 1999.\n\n [RFC2616] Fielding, R., Gettys, J., Mogul, J., Frystyk, H.,\n Masinter, L., Leach, P., and T. Berners-Lee,\n \"Hypertext Transfer Protocol -- HTTP/1.1\", RFC 2616,\n June 1999.\n\n [RFC3986] Berners-Lee, T., Fielding, R., and L. Masinter,\n \"Uniform Resource Identifier (URI): Generic Syntax\",\n STD 66, RFC 3986, January 2005.\n\n\n\nEastlake Standards Track [Page 22]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246, August\n 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation\n List (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC5890] Klensin, J., \"Internationalized Domain Names for\n Applications (IDNA): Definitions and Document\n Framework\", RFC 5890, August 2010.\n\n13. Informative References\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [X509-4th] ITU-T Recommendation X.509 (2000) | ISO/IEC\n 9594-8:2001, \"Information Systems - Open Systems\n Interconnection - The Directory: Public key and\n attribute certificate frameworks\".\n\n [X509-4th-TC1] ITU-T Recommendation X.509(2000) Corrigendum 1(2001) |\n ISO/IEC 9594-8:2001/Cor.1:2002, Technical Corrigendum\n 1 to ISO/IEC 9594:8:2001.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 23]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix A. Changes from RFC 4366\n\n The significant changes between RFC 4366 and this document are\n described below.\n\n RFC 4366 described both general extension mechanisms (for the TLS\n handshake and client and server hellos) as well as specific\n extensions. RFC 4366 was associated with RFC 4346, TLS 1.1. The\n client and server hello extension mechanisms have been moved into RFC\n 5246, TLS 1.2, so this document, which is associated with RFC 5246,\n includes only the handshake extension mechanisms and the specific\n extensions from RFC 4366. RFC 5246 also specifies the unknown\n extension error and new extension specification considerations, so\n that material has been removed from this document.\n\n The Server Name extension now specifies only ASCII representation,\n eliminating UTF-8. It is provided that the ServerNameList can\n contain more than only one name of any particular name_type. If a\n server name is provided but not recognized, the server should either\n continue the handshake without an error or send a fatal error.\n Sending a warning-level message is not recommended because client\n behavior will be unpredictable. Provision was added for the user\n using the server_name extension in deciding whether or not to resume\n a session. Furthermore, this extension should be the same in a\n session resumption request as it was in the full handshake that\n established the session. Such a resumption request must not be\n accepted if the server_name extension is different, but instead a\n full handshake must be done to possibly establish a new session.\n\n The Client Certificate URLs extension has been changed to make the\n presence of a hash mandatory.\n\n For the case of DTLS, the requirement to report an overflow of the\n negotiated maximum fragment length is made conditional on passing\n authentication.\n\n TLS servers are now prohibited from following HTTP redirects when\n retrieving certificates.\n\n The material was also re-organized in minor ways. For example,\n information as to which errors are fatal is moved from the \"Error\n Alerts\" section to the individual extension specifications.\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 24]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix B. Acknowledgements\n\n This document is based on material from RFC 4366 for which the\n authors were S. Blake-Wilson, M. Nystrom, D. Hopwood, J. Mikkelsen,\n and T. Wright. Other contributors include Joseph Salowey, Alexey\n Melnikov, Peter Saint-Andre, and Adrian Farrel.\n\nAuthor's Address\n\n Donald Eastlake 3rd\n Huawei\n 155 Beaver Street\n Milford, MA 01757 USA\n\n Phone: +1-508-333-2270\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 25]\n\nAnswer using only the documents and the scenario in the question. Return exactly the requested JSON object; use \"UNKNOWN\" only per unresolved field.", "context_source_relation_count": 22, "counterfactual_twin": {"byte_end": 447, "byte_start": 361, "char_end": 447, "char_start": 361, "evidence_id": "current_protocol", "parent_source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd", "parent_text_sha256": "d596886622bceee8500b3f3c9d869d947b4e70a2f05ea65fdd75666b46debd81", "parent_value": "This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol.", "provenance_operation": "exclude_exact_source_span", "record_id": "ietf:rfc:8446", "source_manifest_sha256": "287ca4031b223ad818d21781c575905d5ed30eb7e4e0e38fef196cf1ca1865b4", "source_origin": "synthetic_counterfactual", "text_sha256": "98ef2decaa389708a963efbd12018cad07c13f824ea89269ccae34e25ced38b7", "value": " "}, "data_stage": "train_ready", "dependency_class": "long_range_retrieval", "diagnostic_only": true, "difficulty": {"context_tokens": 65513, "evidence_span_tokens": 65134, "max_evidence_distance": 65141, "proof_depth": 2, "state_updates": 10}, "document_context": "Network Working Group J. Salowey\nRequest for Comments: 5077 H. Zhou\nObsoletes: 4507 Cisco Systems\nCategory: Standards Track P. Eronen\n Nokia\n H. Tschofenig\n Nokia Siemens Networks\n January 2008\n\n\n Transport Layer Security (TLS) Session Resumption without\n Server-Side State\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document describes a mechanism that enables the Transport Layer\n Security (TLS) server to resume sessions and avoid keeping per-client\n session state. The TLS server encapsulates the session state into a\n ticket and forwards it to the client. The client can subsequently\n resume a session using the obtained ticket. This document obsoletes\n RFC 4507.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 1]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3. Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . 4\n 3.2. SessionTicket TLS Extension . . . . . . . . . . . . . . . 7\n 3.3. NewSessionTicket Handshake Message . . . . . . . . . . . . 8\n 3.4. Interaction with TLS Session ID . . . . . . . . . . . . . 9\n 4. Recommended Ticket Construction . . . . . . . . . . . . . . . 10\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . 12\n 5.1. Invalidating Sessions . . . . . . . . . . . . . . . . . . 12\n 5.2. Stolen Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.3. Forged Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.4. Denial of Service Attacks . . . . . . . . . . . . . . . . 12\n 5.5. Ticket Protection Key Management . . . . . . . . . . . . . 13\n 5.6. Ticket Lifetime . . . . . . . . . . . . . . . . . . . . . 13\n 5.7. Alternate Ticket Formats and Distribution Schemes . . . . 13\n 5.8. Identity Privacy, Anonymity, and Unlinkability . . . . . . 14\n 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 14\n 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 15\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 15\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 15\n Appendix A. Discussion of Changes to RFC 4507 . . . . . . . . . . 17\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 2]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n1. Introduction\n\n This document defines a way to resume a Transport Layer Security\n (TLS) session without requiring session-specific state at the TLS\n server. This mechanism may be used with any TLS ciphersuite. This\n document applies to both TLS 1.0 defined in [RFC2246], and TLS 1.1\n defined in [RFC4346]. The mechanism makes use of TLS extensions\n defined in [RFC4366] and defines a new TLS message type.\n\n This mechanism is useful in the following situations:\n\n 1. servers that handle a large number of transactions from different\n users\n\n 2. servers that desire to cache sessions for a long time\n\n 3. ability to load balance requests across servers\n\n 4. embedded servers with little memory\n\n This document obsoletes RFC 4507 [RFC4507] to correct an error in the\n encoding that caused the specification to differ from deployed\n implementations. At the time of this writing, there are no known\n implementations that follow the encoding specified in RFC 4507. This\n update to RFC 4507 aligns the document with currently deployed\n implementations. More details of the change are given in Appendix A.\n\n2. Terminology\n\n Within this document, the term 'ticket' refers to a cryptographically\n protected data structure that is created and consumed by the server\n to rebuild session-specific state.\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Protocol\n\n This specification describes a mechanism to distribute encrypted\n session-state information to the client in the form of a ticket and a\n mechanism to present the ticket back to the server. The ticket is\n created by a TLS server and sent to a TLS client. The TLS client\n presents the ticket to the TLS server to resume a session.\n Implementations of this specification are expected to support both\n mechanisms. Other specifications can take advantage of the session\n tickets, perhaps specifying alternative means for distribution or\n selection. For example, a separate specification may describe an\n\n\n\nSalowey, et al. Standards Track [Page 3]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n alternate way to distribute a ticket and use the TLS extension in\n this document to resume the session. This behavior is beyond the\n scope of the document and would need to be described in a separate\n specification.\n\n3.1. Overview\n\n The client indicates that it supports this mechanism by including a\n SessionTicket TLS extension in the ClientHello message. The\n extension will be empty if the client does not already possess a\n ticket for the server. The server sends an empty SessionTicket\n extension to indicate that it will send a new session ticket using\n the NewSessionTicket handshake message. The extension is described\n in Section 3.2.\n\n If the server wants to use this mechanism, it stores its session\n state (such as ciphersuite and master secret) to a ticket that is\n encrypted and integrity-protected by a key known only to the server.\n The ticket is distributed to the client using the NewSessionTicket\n TLS handshake message described in Section 3.3. This message is sent\n during the TLS handshake before the ChangeCipherSpec message, after\n the server has successfully verified the client's Finished message.\n\n Client Server\n\n ClientHello\n (empty SessionTicket extension)-------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1: Message Flow for Full Handshake Issuing New Session Ticket\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 4]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n The client caches this ticket along with the master secret and other\n parameters associated with the current session. When the client\n wishes to resume the session, it includes the ticket in the\n SessionTicket extension within the ClientHello message. Appendix A\n provides a detailed description of the encoding of the extension and\n changes from RFC 4507. The server then decrypts the received ticket,\n verifies the ticket's validity, retrieves the session state from the\n contents of the ticket, and uses this state to resume the session.\n The interaction with the TLS Session ID is described in Section 3.4.\n If the server successfully verifies the client's ticket, then it may\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello.\n\n Client Server\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2: Message Flow for Abbreviated Handshake Using New Session\n Ticket\n\n A recommended ticket format is given in Section 4.\n\n If the server cannot or does not want to honor the ticket, then it\n can initiate a full handshake with the client.\n\n In the case that the server does not wish to issue a new ticket at\n this time, it just completes the handshake without including a\n SessionTicket extension or NewSessionTicket handshake message. This\n is shown below (this flow is identical to Figure 1 in RFC 4346,\n except for the SessionTicket extension in the first message):\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 5]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 3: Message Flow for Server Completing Full Handshake Without\n Issuing New Session Ticket\n\n It is also permissible to have an exchange similar to Figure 3 using\n the abbreviated handshake defined in Figure 2 of RFC 4346, where the\n client uses the SessionTicket extension to resume the session, but\n the server does not wish to issue a new ticket, and therefore does\n not send a SessionTicket extension.\n\n If the server rejects the ticket, it may still wish to issue a new\n ticket after performing the full handshake as shown below (this flow\n is identical to Figure 1, except the SessionTicket extension in the\n ClientHello is not empty):\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 6]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 4: Message Flow for Server Rejecting Ticket, Performing Full\n Handshake, and Issuing New Session Ticket\n\n3.2. SessionTicket TLS Extension\n\n The SessionTicket TLS extension is based on [RFC4366]. The format of\n the ticket is an opaque structure used to carry session-specific\n state information. This extension may be sent in the ClientHello and\n ServerHello.\n\n If the client possesses a ticket that it wants to use to resume a\n session, then it includes the ticket in the SessionTicket extension\n in the ClientHello. If the client does not have a ticket and is\n prepared to receive one in the NewSessionTicket handshake message,\n then it MUST include a zero-length ticket in the SessionTicket\n extension. If the client is not prepared to receive a ticket in the\n NewSessionTicket handshake message, then it MUST NOT include a\n SessionTicket extension unless it is sending a non-empty ticket it\n received through some other means from the server.\n\n The server uses a zero-length SessionTicket extension to indicate to\n the client that it will send a new session ticket using the\n NewSessionTicket handshake message described in Section 3.3. The\n server MUST send this extension in the ServerHello if it wishes to\n issue a new ticket to the client using the NewSessionTicket handshake\n message. The server MUST NOT send this extension if it does not\n receive one in the ClientHello.\n\n\n\n\nSalowey, et al. Standards Track [Page 7]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n If the server fails to verify the ticket, then it falls back to\n performing a full handshake. If the ticket is accepted by the server\n but the handshake fails, the client SHOULD delete the ticket.\n\n The SessionTicket extension has been assigned the number 35. The\n extension_data field of SessionTicket extension contains the ticket.\n\n3.3. NewSessionTicket Handshake Message\n\n This message is sent by the server during the TLS handshake before\n the ChangeCipherSpec message. This message MUST be sent if the\n server included a SessionTicket extension in the ServerHello. This\n message MUST NOT be sent if the server did not include a\n SessionTicket extension in the ServerHello. This message is included\n in the hash used to create and verify the Finished message. In the\n case of a full handshake, the server MUST verify the client's\n Finished message before sending the ticket. The client MUST NOT\n treat the ticket as valid until it has verified the server's Finished\n message. If the server determines that it does not want to include a\n ticket after it has included the SessionTicket extension in the\n ServerHello, then it sends a zero-length ticket in the\n NewSessionTicket handshake message.\n\n If the server successfully verifies the client's ticket, then it MAY\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello in the abbreviated handshake. The client\n should start using the new ticket as soon as possible after it\n verifies the server's Finished message for new connections. Note\n that since the updated ticket is issued before the handshake\n completes, it is possible that the client may not put the new ticket\n into use before it initiates new connections. The server MUST NOT\n assume that the client actually received the updated ticket until it\n successfully verifies the client's Finished message.\n\n The NewSessionTicket handshake message has been assigned the number 4\n and its definition is given at the end of this section. The\n ticket_lifetime_hint field contains a hint from the server about how\n long the ticket should be stored. The value indicates the lifetime\n in seconds as a 32-bit unsigned integer in network byte order\n relative to when the ticket is received. A value of zero is reserved\n to indicate that the lifetime of the ticket is unspecified. A client\n SHOULD delete the ticket and associated state when the time expires.\n It MAY delete the ticket earlier based on local policy. A server MAY\n treat a ticket as valid for a shorter or longer period of time than\n what is stated in the ticket_lifetime_hint.\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 8]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n HandshakeType msg_type;\n uint24 length;\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case session_ticket: NewSessionTicket; /* NEW */\n } body;\n } Handshake;\n\n\n struct {\n uint32 ticket_lifetime_hint;\n opaque ticket<0..2^16-1>;\n } NewSessionTicket;\n\n3.4. Interaction with TLS Session ID\n\n If a server is planning on issuing a session ticket to a client that\n does not present one, it SHOULD include an empty Session ID in the\n ServerHello. If the server rejects the ticket and falls back to the\n full handshake then it may include a non-empty Session ID to indicate\n its support for stateful session resumption. If the client receives\n a session ticket from the server, then it discards any Session ID\n that was sent in the ServerHello.\n\n When presenting a ticket, the client MAY generate and include a\n Session ID in the TLS ClientHello. If the server accepts the ticket\n and the Session ID is not empty, then it MUST respond with the same\n Session ID present in the ClientHello. This allows the client to\n easily differentiate when the server is resuming a session from when\n it is falling back to a full handshake. Since the client generates a\n Session ID, the server MUST NOT rely upon the Session ID having a\n particular value when validating the ticket. If a ticket is\n presented by the client, the server MUST NOT attempt to use the\n Session ID in the ClientHello for stateful session resumption.\n Alternatively, the client MAY include an empty Session ID in the\n ClientHello. In this case, the client ignores the Session ID sent in\n the ServerHello and determines if the server is resuming a session by\n the subsequent handshake messages.\n\n\n\nSalowey, et al. Standards Track [Page 9]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n4. Recommended Ticket Construction\n\n This section describes a recommended format and protection for the\n ticket. Note that the ticket is opaque to the client, so the\n structure is not subject to interoperability concerns, and\n implementations may diverge from this format. If implementations do\n diverge from this format, they must take security concerns seriously.\n Clients MUST NOT examine the ticket under the assumption that it\n complies with this document.\n\n The server uses two different keys: one 128-bit key for Advanced\n Encryption Standard (AES) [AES] in Cipher Block Chaining (CBC) mode\n [CBC] encryption and one 256-bit key for HMAC-SHA-256 [RFC4634].\n\n The ticket is structured as follows:\n\n struct {\n opaque key_name[16];\n opaque iv[16];\n opaque encrypted_state<0..2^16-1>;\n opaque mac[32];\n } ticket;\n\n Here, key_name serves to identify a particular set of keys used to\n protect the ticket. It enables the server to easily recognize\n tickets it has issued. The key_name should be randomly generated to\n avoid collisions between servers. One possibility is to generate new\n random keys and key_name every time the server is started.\n\n The actual state information in encrypted_state is encrypted using\n 128-bit AES in CBC mode with the given IV. The Message\n Authentication Code (MAC) is calculated using HMAC-SHA-256 over\n key_name (16 octets) and IV (16 octets), followed by the length of\n the encrypted_state field (2 octets) and its contents (variable\n length).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 10]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n ProtocolVersion protocol_version;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n opaque master_secret[48];\n ClientIdentity client_identity;\n uint32 timestamp;\n } StatePlaintext;\n\n enum {\n anonymous(0),\n certificate_based(1),\n psk(2)\n } ClientAuthenticationType;\n\n struct {\n ClientAuthenticationType client_authentication_type;\n select (ClientAuthenticationType) {\n case anonymous: struct {};\n case certificate_based:\n ASN.1Cert certificate_list<0..2^24-1>;\n case psk:\n opaque psk_identity<0..2^16-1>; /* from [RFC4279] */\n };\n } ClientIdentity;\n\n The structure StatePlaintext stores the TLS session state including\n the master_secret. The timestamp within this structure allows the\n TLS server to expire tickets. To cover the authentication and key\n exchange protocols provided by TLS, the ClientIdentity structure\n contains the authentication type of the client used in the initial\n exchange (see ClientAuthenticationType). To offer the TLS server\n with the same capabilities for authentication and authorization, a\n certificate list is included in case of public-key-based\n authentication. The TLS server is therefore able to inspect a number\n of different attributes within these certificates. A specific\n implementation might choose to store a subset of this information or\n additional information. Other authentication mechanisms, such as\n Kerberos [RFC2712], would require different client identity data.\n Other TLS extensions may require the inclusion of additional data in\n the StatePlaintext structure.\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 11]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5. Security Considerations\n\n This section addresses security issues related to the usage of a\n ticket. Tickets must be authenticated and encrypted to prevent\n modification or eavesdropping by an attacker. Several attacks\n described below will be possible if this is not carefully done.\n\n Implementations should take care to ensure that the processing of\n tickets does not increase the chance of denial of service as\n described below.\n\n5.1. Invalidating Sessions\n\n The TLS specification requires that TLS sessions be invalidated when\n errors occur. [CSSC] discusses the security implications of this in\n detail. In the analysis within this paper, failure to invalidate\n sessions does not pose a security risk. This is because the TLS\n handshake uses a non-reversible function to derive keys for a session\n so information about one session does not provide an advantage to\n attack the master secret or a different session. If a session\n invalidation scheme is used, the implementation should verify the\n integrity of the ticket before using the contents to invalidate a\n session to ensure that an attacker cannot invalidate a chosen\n session.\n\n5.2. Stolen Tickets\n\n An eavesdropper or man-in-the-middle may obtain the ticket and\n attempt to use it to establish a session with the server; however,\n since the ticket is encrypted and the attacker does not know the\n secret key, a stolen ticket does not help an attacker resume a\n session. A TLS server MUST use strong encryption and integrity\n protection for the ticket to prevent an attacker from using a brute\n force mechanism to obtain the ticket's contents.\n\n5.3. Forged Tickets\n\n A malicious user could forge or alter a ticket in order to resume a\n session, to extend its lifetime, to impersonate another user, or to\n gain additional privileges. This attack is not possible if the\n ticket is protected using a strong integrity protection algorithm\n such as a keyed HMAC-SHA-256.\n\n5.4. Denial of Service Attacks\n\n The key_name field defined in the recommended ticket format helps the\n server efficiently reject tickets that it did not issue. However, an\n adversary could store or generate a large number of tickets to send\n\n\n\nSalowey, et al. Standards Track [Page 12]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n to the TLS server for verification. To minimize the possibility of a\n denial of service, the verification of the ticket should be\n lightweight (e.g., using efficient symmetric key cryptographic\n algorithms).\n\n5.5. Ticket Protection Key Management\n\n A full description of the management of the keys used to protect the\n ticket is beyond the scope of this document. A list of RECOMMENDED\n practices is given below.\n\n o The keys should be generated securely following the randomness\n recommendations in [RFC4086].\n\n o The keys and cryptographic protection algorithms should be at\n least 128 bits in strength. Some ciphersuites and applications\n may require cryptographic protection greater than 128 bits in\n strength.\n\n o The keys should not be used for any purpose other than generating\n and verifying tickets.\n\n o The keys should be changed regularly.\n\n o The keys should be changed if the ticket format or cryptographic\n protection algorithms change.\n\n5.6. Ticket Lifetime\n\n The TLS server controls the lifetime of the ticket. Servers\n determine the acceptable lifetime based on the operational and\n security requirements of the environments in which they are deployed.\n The ticket lifetime may be longer than the 24-hour lifetime\n recommended in [RFC4346]. TLS clients may be given a hint of the\n lifetime of the ticket. Since the lifetime of a ticket may be\n unspecified, a client has its own local policy that determines when\n it discards tickets.\n\n5.7. Alternate Ticket Formats and Distribution Schemes\n\n If the ticket format or distribution scheme defined in this document\n is not used, then great care must be taken in analyzing the security\n of the solution. In particular, if confidential information, such as\n a secret key, is transferred to the client, it MUST be done using\n secure communication so as to prevent attackers from obtaining or\n modifying the key. Also, the ticket MUST have its integrity and\n confidentiality protected with strong cryptographic techniques to\n prevent a breach in the security of the system.\n\n\n\nSalowey, et al. Standards Track [Page 13]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5.8. Identity Privacy, Anonymity, and Unlinkability\n\n This document mandates that the content of the ticket is\n confidentiality protected in order to avoid leakage of its content,\n such as user-relevant information. As such, it prevents disclosure\n of potentially sensitive information carried within the ticket.\n\n The initial handshake exchange, which was used to obtain the ticket,\n might not provide identity confidentiality of the client based on the\n properties of TLS. Another relevant security threat is the ability\n for an on-path adversary to observe multiple TLS handshakes where the\n same ticket is used, therefore concluding they belong to the same\n communication endpoints. Application designers that use the ticket\n mechanism described in this document should consider that\n unlinkability [ANON] is not necessarily provided.\n\n While a full discussion of these topics is beyond the scope of this\n document, it should be noted that it is possible to issue a ticket\n using a TLS renegotiation handshake that occurs after a secure tunnel\n has been established by a previous handshake. This may help address\n some privacy and unlinkability issues in some environments.\n\n6. Acknowledgements\n\n The authors would like to thank the following people for their help\n with preparing and reviewing this document: Eric Rescorla, Mohamad\n Badra, Tim Dierks, Nelson Bolyard, Nancy Cam-Winget, David McGrew,\n Rob Dugal, Russ Housley, Amir Herzberg, Bernard Aboba, and members of\n the TLS working group.\n\n [CSSC] describes a solution that is very similar to the one described\n in this document and gives a detailed analysis of the security\n considerations involved. [RFC2712] describes a mechanism for using\n Kerberos [RFC4120] in TLS ciphersuites, which helped inspire the use\n of tickets to avoid server state. [RFC4851] makes use of a similar\n mechanism to avoid maintaining server state for the cryptographic\n tunnel. [SC97] also investigates the concept of stateless sessions.\n\n The authors would also like to thank Jan Nordqvist, who found the\n encoding error in RFC 4507, corrected by this document. In addition\n Nagendra Modadugu, Wan-Teh Chang, and Michael D'Errico provided\n useful feedback during the review of this document.\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 14]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n7. IANA Considerations\n\n IANA has assigned a TLS extension number of 35 to the SessionTicket\n TLS extension from the TLS registry of ExtensionType values defined\n in [RFC4366].\n\n IANA has assigned a TLS HandshakeType number 4 to the\n NewSessionTicket handshake type from the TLS registry of\n HandshakeType values defined in [RFC4346].\n\n This document does not require any actions or assignments from IANA.\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2246] Dierks, T. and C. Allen, \"The TLS Protocol Version 1.0\",\n RFC 2246, January 1999.\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346, April 2006.\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, April 2006.\n\n [RFC4507] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 4507, May 2006.\n\n8.2. Informative References\n\n [AES] National Institute of Standards and Technology, \"Advanced\n Encryption Standard (AES)\", Federal Information Processing\n Standards (FIPS) Publication 197, November 2001.\n\n [ANON] Pfitzmann, A. and M. Hansen, \"Anonymity, Unlinkability,\n Unobservability, Pseudonymity, and Identity Management - A\n Consolidated Proposal for Terminology\", http://\n dud.inf.tu-dresden.de/literatur/\n Anon_Terminology_v0.26-1.pdf Version 0.26, December 2005.\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 15]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n [CBC] National Institute of Standards and Technology,\n \"Recommendation for Block Cipher Modes of Operation -\n Methods and Techniques\", NIST Special Publication 800-38A,\n December 2001.\n\n [CSSC] Shacham, H., Boneh, D., and E. Rescorla, \"Client-side\n caching for TLS\", Transactions on Information and System\n Security (TISSEC) , Volume 7, Issue 4, November 2004.\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [RFC4086] Eastlake, D., Schiller, J., and S. Crocker, \"Randomness\n Requirements for Security\", BCP 106, RFC 4086, June 2005.\n\n [RFC4120] Neuman, C., Yu, T., Hartman, S., and K. Raeburn, \"The\n Kerberos Network Authentication Service (V5)\", RFC 4120,\n July 2005.\n\n [RFC4279] Eronen, P. and H. Tschofenig, \"Pre-Shared Key Ciphersuites\n for Transport Layer Security (TLS)\", RFC 4279,\n December 2005.\n\n [RFC4634] Eastlake, D. and T. Hansen, \"US Secure Hash Algorithms\n (SHA and HMAC-SHA)\", RFC 4634, July 2006.\n\n [RFC4851] Cam-Winget, N., McGrew, D., Salowey, J., and H. Zhou, \"The\n Flexible Authentication via Secure Tunneling Extensible\n Authentication Protocol Method (EAP-FAST)\", RFC 4851,\n May 2007.\n\n [SC97] Aura, T. and P. Nikander, \"Stateless Connections\",\n Proceedings of the First International Conference on\n Information and Communication Security (ICICS '97) , 1997.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 16]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAppendix A. Discussion of Changes to RFC 4507\n\n RFC 4507 [RFC4507] defines a mechanism to resume a TLS session\n without maintaining server side state by specifying an encrypted\n ticket that is maintained on the client. The client presents this\n ticket to the server in a SessionTicket hello extension. The\n encoding in RFC 4507 used the XDR style encoding specified in TLS\n [RFC4346].\n\n An error in the encoding caused the specification to differ from\n deployed implementations. At the time of this writing there are no\n known implementations that follow the encoding specified in RFC 4507.\n This update to RFC 4507 aligns the document with these currently\n deployed implementations.\n\n Erroneous encoding in RFC 4507 resulted in two length fields; one for\n the extension contents and one for the ticket itself. Hence, for a\n ticket that is 256 bytes long and begins with the hex value FF FF,\n the encoding of the extension would be as follows according to RFC\n 4507:\n\n 00 23 Ticket Extension type 35\n 01 02 Length of extension contents\n 01 00 Length of ticket\n FF FF .. .. Actual ticket\n\n The update proposed in this document reflects what implementations\n actually encode, namely it removes the redundant length field. So,\n for a ticket that is 256 bytes long and begins with the hex value FF\n FF, the encoding of the extension would be as follows according to\n this update:\n\n 00 23 Extension type 35\n 01 00 Length of extension contents (ticket)\n FF FF .. .. Actual ticket\n\n A server implemented according to RFC 4507 receiving a ticket\n extension from a client conforming to this document would interpret\n the first two bytes of the ticket as the length of this ticket. This\n will result in either an inconsistent length field or in the\n processing of a ticket missing the first two bytes. In the first\n case, the server should reject the request based on a malformed\n length. In the second case, the server should reject the ticket\n based on a malformed ticket, incorrect key version, or failed\n decryption. A server implementation based on this update receiving\n an RFC 4507 extension would interpret the first length field as the\n\n\n\n\n\nSalowey, et al. Standards Track [Page 17]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n length of the ticket and include the second two length bytes as the\n first bytes in the ticket, resulting in the ticket being rejected\n based on a malformed ticket, incorrect key version, or failed\n decryption.\n\n Note that the encoding of an empty SessionTicket extension was\n ambiguous in RFC 4507. An RFC 4507 implementation may have encoded\n it as:\n\n 00 23 Extension type 35\n 00 02 Length of extension contents\n 00 00 Length of ticket\n\n or it may have encoded it the same way as this update:\n\n 00 23 Extension type 35\n 00 00 Length of extension contents\n\n A server wishing to support RFC 4507 clients should respond to an\n empty SessionTicket extension encoded the same way as it received it.\n\n A server implementation can construct tickets such that it can detect\n an RFC 4507 implementation, if one existed, by including a cookie at\n the beginning of the tickets that can be differentiated from a valid\n length. For example, if an implementation constructed tickets to\n start with the hex values FF FF, then it could determine where the\n ticket begins and determine the length correctly from the type of\n length fields present.\n\n This document makes a few additional changes to RFC 4507 listed\n below.\n\n o Clarifying that the server can allow session resumption using a\n ticket without issuing a new ticket in Section 3.1.\n\n o Clarifying that the lifetime is relative to when the ticket is\n received in section 3.3.\n\n o Clarifying that the NewSessionTicket handshake message is included\n in the hash generated for the Finished messages in Section 3.3.\n\n o Clarifying the interaction with TLS Session ID in Section 3.4.\n\n o Recommending the use of SHA-256 for the integrity protection of\n the ticket in Section 4.\n\n o Clarifying that additional data can be included in the\n StatePlaintext structure in Section 4.\n\n\n\nSalowey, et al. Standards Track [Page 18]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAuthors' Addresses\n\n Joseph Salowey\n Cisco Systems\n 2901 3rd Ave\n Seattle, WA 98121\n US\n\n EMail: [redacted-email]\n\n\n Hao Zhou\n Cisco Systems\n 4125 Highlander Parkway\n Richfield, OH 44286\n US\n\n EMail: [redacted-email]\n\n\n Pasi Eronen\n Nokia Research Center\n P.O. Box 407\n FIN-00045 Nokia Group\n Finland\n\n EMail: [redacted-email]\n\n\n Hannes Tschofenig\n Nokia Siemens Networks\n Otto-Hahn-Ring 6\n Munich, Bayern 81739\n Germany\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 19]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nFull Copyright Statement\n\n Copyright (C) The IETF Trust (2008).\n\n This document is subject to the rights, licenses and restrictions\n contained in BCP 78, and except as set forth therein, the authors\n retain all their rights.\n\n This document and the information contained herein are provided on an\n \"AS IS\" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\n OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\n THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\n OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\n THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\n WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\n\nIntellectual Property\n\n The IETF takes no position regarding the validity or scope of any\n Intellectual Property Rights or other rights that might be claimed to\n pertain to the implementation or use of the technology described in\n this document or the extent to which any license under such rights\n might or might not be available; nor does it represent that it has\n made any independent effort to identify any such rights. Information\n on the procedures with respect to rights in RFC documents can be\n found in BCP 78 and BCP 79.\n\n Copies of IPR disclosures made to the IETF Secretariat and any\n assurances of licenses to be made available, or the result of an\n attempt made to obtain a general license or permission for the use of\n such proprietary rights by implementers or users of this\n specification can be obtained from the IETF on-line IPR repository at\n http://www.ietf.org/ipr.\n\n The IETF invites any interested party to bring to its attention any\n copyrights, patents or patent applications, or other proprietary\n rights that may cover technology that may be required to implement\n this standard. Please address the information to the IETF at\n ietf-ipr@ietf.org.\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 20]\n\n===== DOCUMENT =====\n\n\nRFC 8446 TLS August 2018\n\n\n Sam Scott\n Royal Holloway, University of London\n [redacted-email]\n\n Thomas Shrimpton\n University of Florida\n [redacted-email]\n\n Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n Brian Smith\n Independent\n [redacted-email]\n\n Brian Sniffen\n Akamai Technologies\n [redacted-email]\n\n Nick Sullivan\n Cloudflare Inc.\n [redacted-email]\n\n Bjoern Tackmann\n University of California, San Diego\n [redacted-email]\n\n Tim Taubert\n Mozilla\n [redacted-email]\n\n Martin Thomson\n Mozilla\n [redacted-email]\n\n Hannes Tschofenig\n Arm Limited\n [redacted-email]\n\n Sean Turner\n sn3rd\n [redacted-email]\n\n Steven Valdez\n Google\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 159]\n\f\nRFC 8446 TLS August 2018\n\n\n Filippo Valsorda\n Cloudflare Inc.\n [redacted-email]\n\n Thyla van der Merwe\n Royal Holloway, University of London\n [redacted-email]\n\n Victor Vasiliev\n Google\n [redacted-email]\n\n Hoeteck Wee\n Ecole Normale Superieure, Paris\n [redacted-email]\n\n Tom Weinstein\n\n David Wong\n NCC Group\n [redacted-email]\n\n Christopher A. Wood\n Apple Inc.\n [redacted-email]\n\n Tim Wright\n Vodafone\n [redacted-email]\n\n Peter Wu\n Independent\n [redacted-email]\n\n Kazu Yamamoto\n Internet Initiative Japan Inc.\n [redacted-email]\n\nAuthor's Address\n\n Eric Rescorla\n Mozilla\n\n Email: [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 160]\n\n===== DOCUMENT =====\n\nNetwork Working Group T. Dierks\nRequest for Comments: 5246 Independent\nObsoletes: 3268, 4346, 4366 E. Rescorla\nUpdates: 4492 RTFM, Inc.\nCategory: Standards Track August 2008\n\n\n The Transport Layer Security (TLS) Protocol\n Version 1.2\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document specifies Version 1.2 of the Transport Layer Security\n (TLS) protocol. The TLS protocol provides communications security\n over the Internet. The protocol allows client/server applications to\n communicate in a way that is designed to prevent eavesdropping,\n tampering, or message forgery.\n\nTable of Contents\n\n 1. Introduction ....................................................4\n 1.1. Requirements Terminology ...................................5\n 1.2. Major Differences from TLS 1.1 .............................5\n 2. Goals ...........................................................6\n 3. Goals of This Document ..........................................7\n 4. Presentation Language ...........................................7\n 4.1. Basic Block Size ...........................................7\n 4.2. Miscellaneous ..............................................8\n 4.3. Vectors ....................................................8\n 4.4. Numbers ....................................................9\n 4.5. Enumerateds ................................................9\n 4.6. Constructed Types .........................................10\n 4.6.1. Variants ...........................................10\n 4.7. Cryptographic Attributes ..................................12\n 4.8. Constants .................................................14\n 5. HMAC and the Pseudorandom Function .............................14\n 6. The TLS Record Protocol ........................................15\n 6.1. Connection States .........................................16\n 6.2. Record Layer ..............................................19\n 6.2.1. Fragmentation ......................................19\n\n\n\nDierks & Rescorla Standards Track [Page 1]\n\f\nRFC 5246 TLS August 2008\n\n\n 6.2.2. Record Compression and Decompression ...............20\n 6.2.3. Record Payload Protection ..........................21\n 6.2.3.1. Null or Standard Stream Cipher ............22\n 6.2.3.2. CBC Block Cipher ..........................22\n 6.2.3.3. AEAD Ciphers ..............................24\n 6.3. Key Calculation ...........................................25\n 7. The TLS Handshaking Protocols ..................................26\n 7.1. Change Cipher Spec Protocol ...............................27\n 7.2. Alert Protocol ............................................28\n 7.2.1. Closure Alerts .....................................29\n 7.2.2. Error Alerts .......................................30\n 7.3. Handshake Protocol Overview ...............................33\n 7.4. Handshake Protocol ........................................37\n 7.4.1. Hello Messages .....................................38\n 7.4.1.1. Hello Request .............................38\n 7.4.1.2. Client Hello ..............................39\n 7.4.1.3. Server Hello ..............................42\n 7.4.1.4. Hello Extensions ..........................44\n 7.4.1.4.1. Signature Algorithms ...........45\n 7.4.2. Server Certificate .................................47\n 7.4.3. Server Key Exchange Message ........................50\n 7.4.4. Certificate Request ................................53\n 7.4.5. Server Hello Done ..................................55\n 7.4.6. Client Certificate .................................55\n 7.4.7. Client Key Exchange Message ........................57\n 7.4.7.1. RSA-Encrypted Premaster Secret Message ....58\n 7.4.7.2. Client Diffie-Hellman Public Value ........61\n 7.4.8. Certificate Verify .................................62\n 7.4.9. Finished ...........................................63\n 8. Cryptographic Computations .....................................64\n 8.1. Computing the Master Secret ...............................64\n 8.1.1. RSA ................................................65\n 8.1.2. Diffie-Hellman .....................................65\n 9. Mandatory Cipher Suites ........................................65\n 10. Application Data Protocol .....................................65\n 11. Security Considerations .......................................65\n 12. IANA Considerations ...........................................65\n Appendix A. Protocol Data Structures and Constant Values ..........68\n A.1. Record Layer ..............................................68\n A.2. Change Cipher Specs Message ...............................69\n A.3. Alert Messages ............................................69\n A.4. Handshake Protocol ........................................70\n A.4.1. Hello Messages .....................................71\n A.4.2. Server Authentication and Key Exchange Messages ....72\n A.4.3. Client Authentication and Key Exchange Messages ....74\n A.4.4. Handshake Finalization Message .....................74\n A.5. The Cipher Suite ..........................................75\n A.6. The Security Parameters ...................................77\n\n\n\nDierks & Rescorla Standards Track [Page 2]\n\f\nRFC 5246 TLS August 2008\n\n\n A.7. Changes to RFC 4492 .......................................78\n Appendix B. Glossary ..............................................78\n Appendix C. Cipher Suite Definitions ..............................83\n Appendix D. Implementation Notes ..................................85\n D.1. Random Number Generation and Seeding ......................85\n D.2. Certificates and Authentication ...........................85\n D.3. Cipher Suites .............................................85\n D.4. Implementation Pitfalls ...................................85\n Appendix E. Backward Compatibility ................................87\n E.1. Compatibility with TLS 1.0/1.1 and SSL 3.0 ................87\n E.2. Compatibility with SSL 2.0 ................................88\n E.3. Avoiding Man-in-the-Middle Version Rollback ...............90\n Appendix F. Security Analysis .....................................91\n F.1. Handshake Protocol ........................................91\n F.1.1. Authentication and Key Exchange ....................91\n F.1.1.1. Anonymous Key Exchange ....................91\n F.1.1.2. RSA Key Exchange and Authentication .......92\n F.1.1.3. Diffie-Hellman Key Exchange with\n Authentication ............................92\n F.1.2. Version Rollback Attacks ...........................93\n F.1.3. Detecting Attacks Against the Handshake Protocol ...94\n F.1.4. Resuming Sessions ..................................94\n F.2. Protecting Application Data ...............................94\n F.3. Explicit IVs ..............................................95\n F.4. Security of Composite Cipher Modes ........................95\n F.5. Denial of Service .........................................96\n F.6. Final Notes ...............................................96\n Normative References ..............................................97\n Informative References ............................................98\n Working Group Information ........................................101\n Contributors .....................................................101\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 3]\n\f\nRFC 5246 TLS August 2008\n\n\n1. Introduction\n\n The primary goal of the TLS protocol is to provide privacy and data\n integrity between two communicating applications. The protocol is\n composed of two layers: the TLS Record Protocol and the TLS Handshake\n Protocol. At the lowest level, layered on top of some reliable\n transport protocol (e.g., TCP [TCP]), is the TLS Record Protocol.\n The TLS Record Protocol provides connection security that has two\n basic properties:\n\n - The connection is private. Symmetric cryptography is used for\n data encryption (e.g., AES [AES], RC4 [SCH], etc.). The keys for\n this symmetric encryption are generated uniquely for each\n connection and are based on a secret negotiated by another\n protocol (such as the TLS Handshake Protocol). The Record\n Protocol can also be used without encryption.\n\n - The connection is reliable. Message transport includes a message\n integrity check using a keyed MAC. Secure hash functions (e.g.,\n SHA-1, etc.) are used for MAC computations. The Record Protocol\n can operate without a MAC, but is generally only used in this mode\n while another protocol is using the Record Protocol as a transport\n for negotiating security parameters.\n\n The TLS Record Protocol is used for encapsulation of various higher-\n level protocols. One such encapsulated protocol, the TLS Handshake\n Protocol, allows the server and client to authenticate each other and\n to negotiate an encryption algorithm and cryptographic keys before\n the application protocol transmits or receives its first byte of\n data. The TLS Handshake Protocol provides connection security that\n has three basic properties:\n\n - The peer's identity can be authenticated using asymmetric, or\n public key, cryptography (e.g., RSA [RSA], DSA [DSS], etc.). This\n authentication can be made optional, but is generally required for\n at least one of the peers.\n\n - The negotiation of a shared secret is secure: the negotiated\n secret is unavailable to eavesdroppers, and for any authenticated\n connection the secret cannot be obtained, even by an attacker who\n can place himself in the middle of the connection.\n\n - The negotiation is reliable: no attacker can modify the\n negotiation communication without being detected by the parties to\n the communication.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 4]\n\f\nRFC 5246 TLS August 2008\n\n\n One advantage of TLS is that it is application protocol independent.\n Higher-level protocols can layer on top of the TLS protocol\n transparently. The TLS standard, however, does not specify how\n protocols add security with TLS; the decisions on how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n1.1. Requirements Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [REQ].\n\n1.2. Major Differences from TLS 1.1\n\n This document is a revision of the TLS 1.1 [TLS1.1] protocol which\n contains improved flexibility, particularly for negotiation of\n cryptographic algorithms. The major changes are:\n\n - The MD5/SHA-1 combination in the pseudorandom function (PRF) has\n been replaced with cipher-suite-specified PRFs. All cipher suites\n in this document use P_SHA256.\n\n - The MD5/SHA-1 combination in the digitally-signed element has been\n replaced with a single hash. Signed elements now include a field\n that explicitly specifies the hash algorithm used.\n\n - Substantial cleanup to the client's and server's ability to\n specify which hash and signature algorithms they will accept.\n Note that this also relaxes some of the constraints on signature\n and hash algorithms from previous versions of TLS.\n\n - Addition of support for authenticated encryption with additional\n data modes.\n\n - TLS Extensions definition and AES Cipher Suites were merged in\n from external [TLSEXT] and [TLSAES].\n\n - Tighter checking of EncryptedPreMasterSecret version numbers.\n\n - Tightened up a number of requirements.\n\n - Verify_data length now depends on the cipher suite (default is\n still 12).\n\n - Cleaned up description of Bleichenbacher/Klima attack defenses.\n\n\n\n\nDierks & Rescorla Standards Track [Page 5]\n\f\nRFC 5246 TLS August 2008\n\n\n - Alerts MUST now be sent in many cases.\n\n - After a certificate_request, if no certificates are available,\n clients now MUST send an empty certificate list.\n\n - TLS_RSA_WITH_AES_128_CBC_SHA is now the mandatory to implement\n cipher suite.\n\n - Added HMAC-SHA256 cipher suites.\n\n - Removed IDEA and DES cipher suites. They are now deprecated and\n will be documented in a separate document.\n\n - Support for the SSLv2 backward-compatible hello is now a MAY, not\n a SHOULD, with sending it a SHOULD NOT. Support will probably\n become a SHOULD NOT in the future.\n\n - Added limited \"fall-through\" to the presentation language to allow\n multiple case arms to have the same encoding.\n\n - Added an Implementation Pitfalls sections\n\n - The usual clarifications and editorial work.\n\n2. Goals\n\n The goals of the TLS protocol, in order of priority, are as follows:\n\n 1. Cryptographic security: TLS should be used to establish a secure\n connection between two parties.\n\n 2. Interoperability: Independent programmers should be able to\n develop applications utilizing TLS that can successfully exchange\n cryptographic parameters without knowledge of one another's code.\n\n 3. Extensibility: TLS seeks to provide a framework into which new\n public key and bulk encryption methods can be incorporated as\n necessary. This will also accomplish two sub-goals: preventing\n the need to create a new protocol (and risking the introduction of\n possible new weaknesses) and avoiding the need to implement an\n entire new security library.\n\n 4. Relative efficiency: Cryptographic operations tend to be highly\n CPU intensive, particularly public key operations. For this\n reason, the TLS protocol has incorporated an optional session\n caching scheme to reduce the number of connections that need to be\n established from scratch. Additionally, care has been taken to\n reduce network activity.\n\n\n\nDierks & Rescorla Standards Track [Page 6]\n\f\nRFC 5246 TLS August 2008\n\n\n3. Goals of This Document\n\n This document and the TLS protocol itself are based on the SSL 3.0\n Protocol Specification as published by Netscape. The differences\n between this protocol and SSL 3.0 are not dramatic, but they are\n significant enough that the various versions of TLS and SSL 3.0 do\n not interoperate (although each protocol incorporates a mechanism by\n which an implementation can back down to prior versions). This\n document is intended primarily for readers who will be implementing\n the protocol and for those doing cryptographic analysis of it. The\n specification has been written with this in mind, and it is intended\n to reflect the needs of those two groups. For that reason, many of\n the algorithm-dependent data structures and rules are included in the\n body of the text (as opposed to in an appendix), providing easier\n access to them.\n\n This document is not intended to supply any details of service\n definition or of interface definition, although it does cover select\n areas of policy as they are required for the maintenance of solid\n security.\n\n4. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used. The syntax draws from\n several sources in its structure. Although it resembles the\n programming language \"C\" in its syntax and XDR [XDR] in both its\n syntax and intent, it would be risky to draw too many parallels. The\n purpose of this presentation language is to document TLS only; it has\n no general application beyond that particular goal.\n\n4.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 7]\n\f\nRFC 5246 TLS August 2008\n\n\n4.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" double\n brackets.\n\n Single-byte entities containing uninterpreted data are of type\n opaque.\n\n4.3. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* 3 consecutive 3 byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n In the following example, mandatory is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 4.4). On the\n other hand, longer can represent up to 800 bytes of data, or 400\n uint16 elements, and it may be empty. Its encoding will include a\n\n\n\nDierks & Rescorla Standards Track [Page 8]\n\f\nRFC 5246 TLS August 2008\n\n\n two-byte actual length field prepended to the vector. The length of\n an encoded vector must be an even multiple of the length of a single\n element (for example, a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is 2 bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n4.4. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are formed from fixed-length series of bytes\n concatenated as described in Section 4.1 and are also unsigned. The\n following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are stored in\n network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n Note that in some cases (e.g., DH parameters) it is necessary to\n represent integers as opaque vectors. In such cases, they are\n represented as unsigned integers (i.e., leading zero octets are not\n required even if the most significant bit is set).\n\n4.5. Enumerateds\n\n An additional sparse data type is available called enum. A field of\n type enum can only assume the values declared in the definition.\n Each definition is a different type. Only enumerateds of the same\n type may be assigned or compared. Every element of an enumerated\n must be assigned a value, as demonstrated in the following example.\n Since the elements of the enumerated are not ordered, they can be\n assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n\n\n\nDierks & Rescorla Standards Track [Page 9]\n\f\nRFC 5246 TLS August 2008\n\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n For enumerateds that are never converted to external representation,\n the numerical information may be omitted.\n\n enum { low, medium, high } Amount;\n\n4.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax for definition is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } [[T]];\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n Structure definitions may be embedded.\n\n4.6.1. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. There\n must be a case arm for every element of the enumeration declared in\n the select. Case arms have limited fall-through: if two case arms\n follow in immediate succession with no fields in between, then they\n\n\n\nDierks & Rescorla Standards Track [Page 10]\n\f\nRFC 5246 TLS August 2008\n\n\n both contain the same fields. Thus, in the example below, \"orange\"\n and \"banana\" both contain V2. Note that this is a new piece of\n syntax in TLS 1.2.\n\n The body of the variant structure may be given a label for reference.\n The mechanism by which the variant is selected at runtime is not\n prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1;\n case e2: Te2;\n case e3: case e4: Te3;\n ....\n case en: Ten;\n } [[fv]];\n } [[Tv]];\n\n For example:\n\n enum { apple, orange, banana } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n select (VariantTag) { /* value of selector is implicit */\n case apple:\n V1; /* VariantBody, tag = apple */\n case orange:\n case banana:\n V2; /* VariantBody, tag = orange or banana */\n } variant_body; /* optional label on variant */\n } VariantRecord;\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 11]\n\f\nRFC 5246 TLS August 2008\n\n\n4.7. Cryptographic Attributes\n\n The five cryptographic operations -- digital signing, stream cipher\n encryption, block cipher encryption, authenticated encryption with\n additional data (AEAD) encryption, and public key encryption -- are\n designated digitally-signed, stream-ciphered, block-ciphered, aead-\n ciphered, and public-key-encrypted, respectively. A field's\n cryptographic processing is specified by prepending an appropriate\n key word designation before the field's type specification.\n Cryptographic keys are implied by the current session state (see\n Section 6.1).\n\n A digitally-signed element is encoded as a struct DigitallySigned:\n\n struct {\n SignatureAndHashAlgorithm algorithm;\n opaque signature<0..2^16-1>;\n } DigitallySigned;\n\n The algorithm field specifies the algorithm used (see Section\n 7.4.1.4.1 for the definition of this field). Note that the\n introduction of the algorithm field is a change from previous\n versions. The signature is a digital signature using those\n algorithms over the contents of the element. The contents themselves\n do not appear on the wire but are simply calculated. The length of\n the signature is specified by the signing algorithm and key.\n\n In RSA signing, the opaque vector contains the signature generated\n using the RSASSA-PKCS1-v1_5 signature scheme defined in [PKCS1]. As\n discussed in [PKCS1], the DigestInfo MUST be DER-encoded [X680]\n [X690]. For hash algorithms without parameters (which includes\n SHA-1), the DigestInfo.AlgorithmIdentifier.parameters field MUST be\n NULL, but implementations MUST accept both without parameters and\n with NULL parameters. Note that earlier versions of TLS used a\n different RSA signature scheme that did not include a DigestInfo\n encoding.\n\n In DSA, the 20 bytes of the SHA-1 hash are run directly through the\n Digital Signing Algorithm with no additional hashing. This produces\n two values, r and s. The DSA signature is an opaque vector, as\n above, the contents of which are the DER encoding of:\n\n Dss-Sig-Value ::= SEQUENCE {\n r INTEGER,\n s INTEGER\n }\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 12]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: In current terminology, DSA refers to the Digital Signature\n Algorithm and DSS refers to the NIST standard. In the original SSL\n and TLS specs, \"DSS\" was used universally. This document uses \"DSA\"\n to refer to the algorithm, \"DSS\" to refer to the standard, and it\n uses \"DSS\" in the code point definitions for historical continuity.\n\n In stream cipher encryption, the plaintext is exclusive-ORed with an\n identical amount of output generated from a cryptographically secure\n keyed pseudorandom number generator.\n\n In block cipher encryption, every block of plaintext encrypts to a\n block of ciphertext. All block cipher encryption is done in CBC\n (Cipher Block Chaining) mode, and all items that are block-ciphered\n will be an exact multiple of the cipher block length.\n\n In AEAD encryption, the plaintext is simultaneously encrypted and\n integrity protected. The input may be of any length, and aead-\n ciphered output is generally larger than the input in order to\n accommodate the integrity check value.\n\n In public key encryption, a public key algorithm is used to encrypt\n data in such a way that it can be decrypted only with the matching\n private key. A public-key-encrypted element is encoded as an opaque\n vector <0..2^16-1>, where the length is specified by the encryption\n algorithm and key.\n\n RSA encryption is done using the RSAES-PKCS1-v1_5 encryption scheme\n defined in [PKCS1].\n\n In the following example\n\n stream-ciphered struct {\n uint8 field1;\n uint8 field2;\n digitally-signed opaque {\n uint8 field3<0..255>;\n uint8 field4;\n };\n } UserType;\n\n The contents of the inner struct (field3 and field4) are used as\n input for the signature/hash algorithm, and then the entire structure\n is encrypted with a stream cipher. The length of this structure, in\n bytes, would be equal to two bytes for field1 and field2, plus two\n bytes for the signature and hash algorithm, plus two bytes for the\n length of the signature, plus the length of the output of the signing\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 13]\n\f\nRFC 5246 TLS August 2008\n\n\n algorithm. The length of the signature is known because the\n algorithm and key used for the signing are known prior to encoding or\n decoding this structure.\n\n4.8. Constants\n\n Typed constants can be defined for purposes of specification by\n declaring a symbol of the desired type and assigning values to it.\n\n Under-specified types (opaque, variable-length vectors, and\n structures that contain opaque) cannot be assigned values. No fields\n of a multi-element structure or vector may be elided.\n\n For example:\n\n struct {\n uint8 f1;\n uint8 f2;\n } Example1;\n\n Example1 ex1 = {1, 4}; /* assigns f1 = 1, f2 = 4 */\n\n5. HMAC and the Pseudorandom Function\n\n The TLS record layer uses a keyed Message Authentication Code (MAC)\n to protect message integrity. The cipher suites defined in this\n document use a construction known as HMAC, described in [HMAC], which\n is based on a hash function. Other cipher suites MAY define their\n own MAC constructions, if needed.\n\n In addition, a construction is required to do expansion of secrets\n into blocks of data for the purposes of key generation or validation.\n This pseudorandom function (PRF) takes as input a secret, a seed, and\n an identifying label and produces an output of arbitrary length.\n\n In this section, we define one PRF, based on HMAC. This PRF with the\n SHA-256 hash function is used for all cipher suites defined in this\n document and in TLS documents published prior to this document when\n TLS 1.2 is negotiated. New cipher suites MUST explicitly specify a\n PRF and, in general, SHOULD use the TLS PRF with SHA-256 or a\n stronger standard hash function.\n\n First, we define a data expansion function, P_hash(secret, data),\n that uses a single hash function to expand a secret and seed into an\n arbitrary quantity of output:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 14]\n\f\nRFC 5246 TLS August 2008\n\n\n P_hash(secret, seed) = HMAC_hash(secret, A(1) + seed) +\n HMAC_hash(secret, A(2) + seed) +\n HMAC_hash(secret, A(3) + seed) + ...\n\n where + indicates concatenation.\n\n A() is defined as:\n\n A(0) = seed\n A(i) = HMAC_hash(secret, A(i-1))\n\n P_hash can be iterated as many times as necessary to produce the\n required quantity of data. For example, if P_SHA256 is being used to\n create 80 bytes of data, it will have to be iterated three times\n (through A(3)), creating 96 bytes of output data; the last 16 bytes\n of the final iteration will then be discarded, leaving 80 bytes of\n output data.\n\n TLS's PRF is created by applying P_hash to the secret as:\n\n PRF(secret, label, seed) = P_(secret, label + seed)\n\n The label is an ASCII string. It should be included in the exact\n form it is given without a length byte or trailing null character.\n For example, the label \"slithy toves\" would be processed by hashing\n the following bytes:\n\n 73 6C 69 74 68 79 20 74 6F 76 65 73\n\n6. The TLS Record Protocol\n\n The TLS Record Protocol is a layered protocol. At each layer,\n messages may include fields for length, description, and content.\n The Record Protocol takes messages to be transmitted, fragments the\n data into manageable blocks, optionally compresses the data, applies\n a MAC, encrypts, and transmits the result. Received data is\n decrypted, verified, decompressed, reassembled, and then delivered to\n higher-level clients.\n\n Four protocols that use the record protocol are described in this\n document: the handshake protocol, the alert protocol, the change\n cipher spec protocol, and the application data protocol. In order to\n allow extension of the TLS protocol, additional record content types\n can be supported by the record protocol. New record content type\n values are assigned by IANA in the TLS Content Type Registry as\n described in Section 12.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 15]\n\f\nRFC 5246 TLS August 2008\n\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST send an\n unexpected_message alert.\n\n Any protocol designed for use over TLS must be carefully designed to\n deal with all possible attacks against it. As a practical matter,\n this means that the protocol designer must be aware of what security\n properties TLS does and does not provide and cannot safely rely on\n the latter.\n\n Note in particular that type and length of a record are not protected\n by encryption. If this information is itself sensitive, application\n designers may wish to take steps (padding, cover traffic) to minimize\n information leakage.\n\n6.1. Connection States\n\n\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 8446 Mozilla\nObsoletes: 5077, 5246, 6961 August 2018\nUpdates: 5705, 6066\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS) Protocol Version 1.3\n\nAbstract\n\n This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol. TLS allows client/server applications to communicate\n over the Internet in a way that is designed to prevent eavesdropping,\n tampering, and message forgery.\n\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077,\n 5246, and 6961. This document also specifies new requirements for\n TLS 1.2 implementations.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 7841.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n https://www.rfc-editor.org/info/rfc8446.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 8446 TLS August 2018\n\n\nCopyright Notice\n\n Copyright (c) 2018 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (https://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 8446 TLS August 2018\n\n\nTable of Contents\n\n 1. Introduction ....................................................6\n 1.1. Conventions and Terminology ................................7\n 1.2. Major Differences from TLS 1.2 .............................8\n 1.3. Updates Affecting TLS 1.2 ..................................9\n 2. Protocol Overview ..............................................10\n 2.1. Incorrect DHE Share .......................................14\n 2.2. Resumption and Pre-Shared Key (PSK) .......................15\n 2.3. 0-RTT Data ................................................17\n 3. Presentation Language ..........................................19\n 3.1. Basic Block Size ..........................................19\n 3.2. Miscellaneous .............................................20\n 3.3. Numbers ...................................................20\n 3.4. Vectors ...................................................20\n 3.5. Enumerateds ...............................................21\n 3.6. Constructed Types .........................................22\n 3.7. Constants .................................................23\n 3.8. Variants ..................................................23\n 4. Handshake Protocol .............................................24\n 4.1. Key Exchange Messages .....................................25\n 4.1.1. Cryptographic Negotiation ..........................26\n 4.1.2. Client Hello .......................................27\n 4.1.3. Server Hello .......................................31\n 4.1.4. Hello Retry Request ................................33\n 4.2. Extensions ................................................35\n 4.2.1. Supported Versions .................................39\n 4.2.2. Cookie .............................................40\n 4.2.3. Signature Algorithms ...............................41\n 4.2.4. Certificate Authorities ............................45\n 4.2.5. OID Filters ........................................45\n 4.2.6. Post-Handshake Client Authentication ...............47\n 4.2.7. Supported Groups ...................................47\n 4.2.8. Key Share ..........................................48\n 4.2.9. Pre-Shared Key Exchange Modes ......................51\n 4.2.10. Early Data Indication .............................52\n 4.2.11. Pre-Shared Key Extension ..........................55\n 4.3. Server Parameters .........................................59\n 4.3.1. Encrypted Extensions ...............................60\n 4.3.2. Certificate Request ................................60\n 4.4. Authentication Messages ...................................61\n 4.4.1. The Transcript Hash ................................63\n 4.4.2. Certificate ........................................64\n 4.4.3. Certificate Verify .................................69\n 4.4.4. Finished ...........................................71\n 4.5. End of Early Data .........................................72\n\n\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 8446 TLS August 2018\n\n\n 4.6. Post-Handshake Messages ...................................73\n 4.6.1. New Session Ticket Message .........................73\n 4.6.2. Post-Handshake Authentication ......................75\n 4.6.3. Key and Initialization Vector Update ...............76\n 5. Record Protocol ................................................77\n 5.1. Record Layer ..............................................78\n 5.2. Record Payload Protection .................................80\n 5.3. Per-Record Nonce ..........................................82\n 5.4. Record Padding ............................................83\n 5.5. Limits on Key Usage .......................................84\n 6. Alert Protocol .................................................85\n 6.1. Closure Alerts ............................................87\n 6.2. Error Alerts ..............................................88\n 7. Cryptographic Computations .....................................90\n 7.1. Key Schedule ..............................................91\n 7.2. Updating Traffic Secrets ..................................94\n 7.3. Traffic Key Calculation ...................................95\n 7.4. (EC)DHE Shared Secret Calculation .........................95\n 7.4.1. Finite Field Diffie-Hellman ........................95\n 7.4.2. Elliptic Curve Diffie-Hellman ......................96\n 7.5. Exporters .................................................97\n 8. 0-RTT and Anti-Replay ..........................................98\n 8.1. Single-Use Tickets ........................................99\n 8.2. Client Hello Recording ....................................99\n 8.3. Freshness Checks .........................................101\n 9. Compliance Requirements .......................................102\n 9.1. Mandatory-to-Implement Cipher Suites .....................102\n 9.2. Mandatory-to-Implement Extensions ........................103\n 9.3. Protocol Invariants ......................................104\n 10. Security Considerations ......................................106\n 11. IANA Considerations ..........................................106\n 12. References ...................................................109\n 12.1. Normative References ....................................109\n 12.2. Informative References ..................................112\n Appendix A. State Machine ........................................120\n A.1. Client ....................................................120\n A.2. Server ....................................................121\n Appendix B. Protocol Data Structures and Constant Values .........122\n B.1. Record Layer ..............................................122\n B.2. Alert Messages ............................................123\n B.3. Handshake Protocol ........................................124\n B.3.1. Key Exchange Messages .................................125\n B.3.2. Server Parameters Messages ............................131\n B.3.3. Authentication Messages ...............................132\n B.3.4. Ticket Establishment ..................................132\n B.3.5. Updating Keys .........................................133\n B.4. Cipher Suites .............................................133\n\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 8446 TLS August 2018\n\n\n Appendix C. Implementation Notes .................................134\n C.1. Random Number Generation and Seeding ......................134\n C.2. Certificates and Authentication ...........................135\n C.3. Implementation Pitfalls ...................................135\n C.4. Client Tracking Prevention ................................137\n C.5. Unauthenticated Operation .................................137\n Appendix D. Backward Compatibility ...............................138\n D.1. Negotiating with an Older Server ..........................139\n D.2. Negotiating with an Older Client ..........................139\n D.3. 0-RTT Backward Compatibility ..............................140\n D.4. Middlebox Compatibility Mode ..............................140\n D.5. Security Restrictions Related to Backward Compatibility ...141\n Appendix E. Overview of Security Properties ......................142\n E.1. Handshake .................................................142\n E.1.1. Key Derivation and HKDF ...............................145\n E.1.2. Client Authentication .................................146\n E.1.3. 0-RTT .................................................146\n E.1.4. Exporter Independence .................................146\n E.1.5. Post-Compromise Security ..............................146\n E.1.6. External References ...................................147\n E.2. Record Layer ..............................................147\n E.2.1. External References ...................................148\n E.3. Traffic Analysis ..........................................148\n E.4. Side-Channel Attacks ......................................149\n E.5. Replay Attacks on 0-RTT ...................................150\n E.5.1. Replay and Exporters ..................................151\n E.6. PSK Identity Exposure .....................................152\n E.7. Sharing PSKs ..............................................152\n E.8. Attacks on Static RSA .....................................152\n Contributors .....................................................153\n Author's Address .................................................160\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 8446 TLS August 2018\n\n\n1. Introduction\n\n The primary goal of TLS is to provide a secure channel between two\n communicating peers; the only requirement from the underlying\n transport is a reliable, in-order data stream. Specifically, the\n secure channel should provide the following properties:\n\n - Authentication: The server side of the channel is always\n authenticated; the client side is optionally authenticated.\n Authentication can happen via asymmetric cryptography (e.g., RSA\n [RSA], the Elliptic Curve Digital Signature Algorithm (ECDSA)\n [ECDSA], or the Edwards-Curve Digital Signature Algorithm (EdDSA)\n [RFC8032]) or a symmetric pre-shared key (PSK).\n\n - Confidentiality: Data sent over the channel after establishment is\n only visible to the endpoints. TLS does not hide the length of\n the data it transmits, though endpoints are able to pad TLS\n records in order to obscure lengths and improve protection against\n traffic analysis techniques.\n\n - Integrity: Data sent over the channel after establishment cannot\n be modified by attackers without detection.\n\n These properties should be true even in the face of an attacker who\n has complete control of the network, as described in [RFC3552]. See\n Appendix E for a more complete statement of the relevant security\n properties.\n\n TLS consists of two primary components:\n\n - A handshake protocol (Section 4) that authenticates the\n communicating parties, negotiates cryptographic modes and\n parameters, and establishes shared keying material. The handshake\n protocol is designed to resist tampering; an active attacker\n should not be able to force the peers to negotiate different\n parameters than they would if the connection were not under\n attack.\n\n - A record protocol (Section 5) that uses the parameters established\n by the handshake protocol to protect traffic between the\n communicating peers. The record protocol divides traffic up into\n a series of records, each of which is independently protected\n using the traffic keys.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 8446 TLS August 2018\n\n\n TLS is application protocol independent; higher-level protocols can\n layer on top of TLS transparently. The TLS standard, however, does\n not specify how protocols add security with TLS; how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n This document defines TLS version 1.3. While TLS 1.3 is not directly\n compatible with previous versions, all versions of TLS incorporate a\n versioning mechanism which allows clients and servers to\n interoperably negotiate a common version if one is supported by both\n peers.\n\n This document supersedes and obsoletes previous versions of TLS,\n including version 1.2 [RFC5246]. It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2. Because TLS 1.3 changes the way keys are\n derived, it updates [RFC5705] as described in Section 7.5. It also\n changes how Online Certificate Status Protocol (OCSP) messages are\n carried and therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in Section 4.4.2.1.\n\n1.1. Conventions and Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all\n capitals, as shown here.\n\n The following terms are used:\n\n client: The endpoint initiating the TLS connection.\n\n connection: A transport-layer connection between two endpoints.\n\n endpoint: Either the client or server of the connection.\n\n handshake: An initial negotiation between client and server that\n establishes the parameters of their subsequent interactions\n within TLS.\n\n peer: An endpoint. When discussing a particular endpoint, \"peer\"\n refers to the endpoint that is not the primary subject of\n discussion.\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\f\nRFC 8446 TLS August 2018\n\n\n receiver: An endpoint that is receiving records.\n\n sender: An endpoint that is transmitting records.\n\n server: The endpoint that did not initiate the TLS connection.\n\n1.2. Major Differences from TLS 1.2\n\n The following is a list of the major functional differences between\n TLS 1.2 and TLS 1.3. It is not intended to be exhaustive, and there\n are many minor differences.\n\n - The list of supported symmetric encryption algorithms has been\n pruned of all algorithms that are considered legacy. Those that\n remain are all Authenticated Encryption with Associated Data\n (AEAD) algorithms. The cipher suite concept has been changed to\n separate the authentication and key exchange mechanisms from the\n record protection algorithm (including secret key length) and a\n hash to be used with both the key derivation function and\n handshake message authentication code (MAC).\n\n - A zero round-trip time (0-RTT) mode was added, saving a round trip\n at connection setup for some application data, at the cost of\n certain security properties.\n\n - Static RSA and Diffie-Hellman cipher suites have been removed; all\n public-key based key exchange mechanisms now provide forward\n secrecy.\n\n - All handshake messages after the ServerHello are now encrypted.\n The newly introduced EncryptedExtensions message allows various\n extensions previously sent in the clear in the ServerHello to also\n enjoy confidentiality protection.\n\n - The key derivation functions have been redesigned. The new design\n allows easier analysis by cryptographers due to their improved key\n separation properties. The HMAC-based Extract-and-Expand Key\n Derivation Function (HKDF) is used as an underlying primitive.\n\n - The handshake state machine has been significantly restructured to\n be more consistent and to remove superfluous messages such as\n ChangeCipherSpec (except when needed for middlebox compatibility).\n\n - Elliptic curve algorithms are now in the base spec, and new\n signature algorithms, such as EdDSA, are included. TLS 1.3\n removed point format negotiation in favor of a single point format\n for each curve.\n\n\n\n\nRescorla Standards Track [Page 8]\n\f\nRFC 8446 TLS August 2018\n\n\n - Other cryptographic improvements were made, including changing the\n RSA padding to use the RSA Probabilistic Signature Scheme\n (RSASSA-PSS), and the removal of compression, the Digital\n Signature Algorithm (DSA), and custom Ephemeral Diffie-Hellman\n (DHE) groups.\n\n - The TLS 1.2 version negotiation mechanism has been deprecated in\n favor of a version list in an extension. This increases\n compatibility with existing servers that incorrectly implemented\n version negotiation.\n\n - Session resumption with and without server-side state as well as\n the PSK-based cipher suites of earlier TLS versions have been\n replaced by a single new PSK exchange.\n\n - References have been updated to point to the updated versions of\n RFCs, as appropriate (e.g., RFC 5280 rather than RFC 3280).\n\n1.3. Updates Affecting TLS 1.2\n\n This document defines several changes that optionally affect\n implementations of TLS 1.2, including those which do not also support\n TLS 1.3:\n\n - A version downgrade protection mechanism is described in\n Section 4.1.3.\n\n - RSASSA-PSS signature schemes are defined in Section 4.2.3.\n\n - The \"supported_versions\" ClientHello extension can be used to\n negotiate the version of TLS to use, in preference to the\n legacy_version field of the ClientHello.\n\n - The \"signature_algorithms_cert\" extension allows a client to\n indicate which signature algorithms it can validate in X.509\n certificates.\n\n Additionally, this document clarifies some compliance requirements\n for earlier versions of TLS; see Section 9.3.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 9]\n\f\nRFC 8446 TLS August 2018\n\n\n2. Protocol Overview\n\n The cryptographic parameters used by the secure channel are produced\n by the TLS handshake protocol. This sub-protocol of TLS is used by\n the client and server when first communicating with each other. The\n handshake protocol allows peers to negotiate a protocol version,\n select cryptographic algorithms, optionally authenticate each other,\n and establish shared secret keying material. Once the handshake is\n complete, the peers use the established keys to protect the\n application-layer traffic.\n\n A failure of the handshake or other protocol error triggers the\n termination of the connection, optionally preceded by an alert\n message (Section 6).\n\n TLS supports three basic key exchange modes:\n\n - (EC)DHE (Diffie-Hellman over either finite fields or elliptic\n curves)\n\n - PSK-only\n\n - PSK with (EC)DHE\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 10]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 1 below shows the basic full TLS handshake:\n\n Client Server\n\nKey ^ ClientHello\nExch | + key_share*\n | + signature_algorithms*\n | + psk_key_exchange_modes*\n v + pre_shared_key* -------->\n ServerHello ^ Key\n + key_share* | Exch\n + pre_shared_key* v\n {EncryptedExtensions} ^ Server\n {CertificateRequest*} v Params\n {Certificate*} ^\n {CertificateVerify*} | Auth\n {Finished} v\n <-------- [Application Data*]\n ^ {Certificate*}\nAuth | {CertificateVerify*}\n v {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 1: Message Flow for Full TLS Handshake\n\n The handshake can be thought of as having three phases (indicated in\n the diagram above):\n\n - Key Exchange: Establish shared keying material and select the\n cryptographic parameters. Everything after this phase is\n encrypted.\n\n - Server Parameters: Establish other handshake parameters\n (whether the client is authenticated, application-layer protocol\n support, etc.).\n\n\n\n\nRescorla Standards Track [Page 11]\n\f\nRFC 8446 TLS August 2018\n\n\n - Authentication: Authenticate the server (and, optionally, the\n client) and provide key confirmation and handshake integrity.\n\n In the Key Exchange phase, the client sends the ClientHello\n (Section 4.1.2) message, which contains a random nonce\n (ClientHello.random); its offered protocol versions; a list of\n symmetric cipher/HKDF hash pairs; either a set of Diffie-Hellman key\n shares (in the \"key_share\" (Section 4.2.8) extension), a set of\n pre-shared key labels (in the \"pre_shared_key\" (Section 4.2.11)\n extension), or both; and potentially additional extensions.\n Additional fields and/or messages may also be present for middlebox\n compatibility.\n\n The server processes the ClientHello and determines the appropriate\n cryptographic parameters for the connection. It then responds with\n its own ServerHello (Section 4.1.3), which indicates the negotiated\n connection parameters. The combination of the ClientHello and the\n ServerHello determines the shared keys. If (EC)DHE key establishment\n is in use, then the ServerHello contains a \"key_share\" extension with\n the server's ephemeral Diffie-Hellman share; the server's share MUST\n be in the same group as one of the client's shares. If PSK key\n establishment is in use, then the ServerHello contains a\n \"pre_shared_key\" extension indicating which of the client's offered\n PSKs was selected. Note that implementations can use (EC)DHE and PSK\n together, in which case both extensions will be supplied.\n\n The server then sends two messages to establish the Server\n Parameters:\n\n EncryptedExtensions: responses to ClientHello extensions that are\n not required to determine the cryptographic parameters, other than\n those that are specific to individual certificates.\n [Section 4.3.1]\n\n CertificateRequest: if certificate-based client authentication is\n desired, the desired parameters for that certificate. This\n message is omitted if client authentication is not desired.\n [Section 4.3.2]\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 12]\n\f\nRFC 8446 TLS August 2018\n\n\n Finally, the client and server exchange Authentication messages. TLS\n uses the same set of messages every time that certificate-based\n authentication is needed. (PSK-based authentication happens as a\n side effect of key exchange.) Specifically:\n\n Certificate: The certificate of the endpoint and any per-certificate\n extensions. This message is omitted by the server if not\n authenticating with a certificate and by the client if the server\n did not send CertificateRequest (thus indicating that the client\n should not authenticate with a certificate). Note that if raw\n public keys [RFC7250] or the cached information extension\n [RFC7924] are in use, then this message will not contain a\n certificate but rather some other value corresponding to the\n server's long-term key. [Section 4.4.2]\n\n CertificateVerify: A signature over the entire handshake using the\n private key corresponding to the public key in the Certificate\n message. This message is omitted if the endpoint is not\n authenticating via a certificate. [Section 4.4.3]\n\n Finished: A MAC (Message Authentication Code) over the entire\n handshake. This message provides key confirmation, binds the\n endpoint's identity to the exchanged keys, and in PSK mode also\n authenticates the handshake. [Section 4.4.4]\n\n Upon receiving the server's messages, the client responds with its\n Authentication messages, namely Certificate and CertificateVerify (if\n requested), and Finished.\n\n At this point, the handshake is complete, and the client and server\n derive the keying material required by the record layer to exchange\n application-layer data protected through authenticated encryption.\n Application Data MUST NOT be sent prior to sending the Finished\n message, except as specified in Section 2.3. Note that while the\n server may send Application Data prior to receiving the client's\n Authentication messages, any data sent at that point is, of course,\n being sent to an unauthenticated peer.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 13]\n\f\nRFC 8446 TLS August 2018\n\n\n2.1. Incorrect DHE Share\n\n If the client has not provided a sufficient \"key_share\" extension\n (e.g., it includes only DHE or ECDHE groups unacceptable to or\n unsupported by the server), the server corrects the mismatch with a\n HelloRetryRequest and the client needs to restart the handshake with\n an appropriate \"key_share\" extension, as shown in Figure 2. If no\n common cryptographic parameters can be negotiated, the server MUST\n abort the handshake with an appropriate alert.\n\n Client Server\n\n ClientHello\n + key_share -------->\n HelloRetryRequest\n <-------- + key_share\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 2: Message Flow for a Full Handshake with\n Mismatched Parameters\n\n Note: The handshake transcript incorporates the initial\n ClientHello/HelloRetryRequest exchange; it is not reset with the\n new ClientHello.\n\n TLS also allows several optimized variants of the basic handshake, as\n described in the following sections.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 14]\n\f\nRFC 8446 TLS August 2018\n\n\n2.2. Resumption and Pre-Shared Key (PSK)\n\n Although TLS PSKs can be established out of band, PSKs can also be\n established in a previous connection and then used to establish a new\n connection (\"session resumption\" or \"resuming\" with a PSK). Once a\n handshake has completed, the server can send the client a PSK\n identity that corresponds to a unique key derived from the initial\n handshake (see Section 4.6.1). The client can then use that PSK\n identity in future handshakes to negotiate the use of the associated\n PSK. If the server accepts the PSK, then the security context of the\n new connection is cryptographically tied to the original connection\n and the key derived from the initial handshake is used to bootstrap\n the cryptographic state instead of a full handshake. In TLS 1.2 and\n below, this functionality was provided by \"session IDs\" and \"session\n tickets\" [RFC5077]. Both mechanisms are obsoleted in TLS 1.3.\n\n PSKs can be used with (EC)DHE key exchange in order to provide\n forward secrecy in combination with shared keys, or can be used\n alone, at the cost of losing forward secrecy for the application\n data.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 15]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 3 shows a pair of handshakes in which the first handshake\n establishes a PSK and the second handshake uses it:\n\n Client Server\n\n Initial Handshake:\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n <-------- [NewSessionTicket]\n [Application Data] <-------> [Application Data]\n\n\n Subsequent Handshake:\n ClientHello\n + key_share*\n + pre_shared_key -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n {Finished}\n <-------- [Application Data*]\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 3: Message Flow for Resumption and PSK\n\n As the server is authenticating via a PSK, it does not send a\n Certificate or a CertificateVerify message. When a client offers\n resumption via a PSK, it SHOULD also supply a \"key_share\" extension\n to the server to allow the server to decline resumption and fall back\n to a full handshake, if needed. The server responds with a\n \"pre_shared_key\" extension to negotiate the use of PSK key\n establishment and can (as shown here) respond with a \"key_share\"\n extension to do (EC)DHE key establishment, thus providing forward\n secrecy.\n\n\n\n\nRescorla Standards Track [Page 16]\n\f\nRFC 8446 TLS August 2018\n\n\n When PSKs are provisioned out of band, the PSK identity and the KDF\n hash algorithm to be used with the PSK MUST also be provisioned.\n\n Note: When using an out-of-band provisioned pre-shared secret, a\n critical consideration is using sufficient entropy during the key\n generation, as discussed in [RFC4086]. Deriving a shared secret\n from a password or other low-entropy sources is not secure. A\n low-entropy secret, or password, is subject to dictionary attacks\n based on the PSK binder. The specified PSK authentication is not\n a strong password-based authenticated key exchange even when used\n with Diffie-Hellman key establishment. Specifically, it does not\n prevent an attacker that can observe the handshake from performing\n a brute-force attack on the password/pre-shared key.\n\n2.3. 0-RTT Data\n\n When clients and servers share a PSK (either obtained externally or\n via a previous handshake), TLS 1.3 allows clients to send data on the\n first flight (\"early data\"). The client uses the PSK to authenticate\n the server and to encrypt the early data.\n\n As shown in Figure 4, the 0-RTT data is just added to the 1-RTT\n handshake in the first flight. The rest of the handshake uses the\n same messages as for a 1-RTT handshake with PSK resumption.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 17]\n\f\nRFC 8446 TLS August 2018\n\n\n Client Server\n\n ClientHello\n + early_data\n + key_share*\n + psk_key_exchange_modes\n + pre_shared_key\n (Application Data*) -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n + early_data*\n {Finished}\n <-------- [Application Data*]\n (EndOfEarlyData)\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n () Indicates messages protected using keys\n derived from a client_early_traffic_secret.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 4: Message Flow for a 0-RTT Handshake\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 18]\n\f\nRFC 8446 TLS August 2018\n\n\n IMPORTANT NOTE: The security properties for 0-RTT data are weaker\n than those for other kinds of TLS data. Specifically:\n\n 1. This data is not forward secret, as it is encrypted solely under\n keys derived using the offered PSK.\n\n 2. There are no guarantees of non-replay between connections.\n Protection against replay for ordinary TLS 1.3 1-RTT data is\n provided via the server's Random value, but 0-RTT data does not\n depend on the ServerHello and therefore has weaker guarantees.\n This is especially relevant if the data is authenticated either\n with TLS client authentication or inside the application\n protocol. The same warnings apply to any use of the\n early_exporter_master_secret.\n\n\n\n===== DOCUMENT =====\n\n A TLS connection state is the operating environment of the TLS Record\n Protocol. It specifies a compression algorithm, an encryption\n algorithm, and a MAC algorithm. In addition, the parameters for\n these algorithms are known: the MAC key and the bulk encryption keys\n for the connection in both the read and the write directions.\n Logically, there are always four connection states outstanding: the\n current read and write states, and the pending read and write states.\n All records are processed under the current read and write states.\n The security parameters for the pending states can be set by the TLS\n Handshake Protocol, and the ChangeCipherSpec can selectively make\n either of the pending states current, in which case the appropriate\n current state is disposed of and replaced with the pending state; the\n pending state is then reinitialized to an empty state. It is illegal\n to make a state that has not been initialized with security\n parameters a current state. The initial current state always\n specifies that no encryption, compression, or MAC will be used.\n\n The security parameters for a TLS Connection read and write state are\n set by providing the following values:\n\n connection end\n Whether this entity is considered the \"client\" or the \"server\" in\n this connection.\n\n PRF algorithm\n An algorithm used to generate keys from the master secret (see\n Sections 5 and 6.3).\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 16]\n\f\nRFC 5246 TLS August 2008\n\n\n bulk encryption algorithm\n An algorithm to be used for bulk encryption. This specification\n includes the key size of this algorithm, whether it is a block,\n stream, or AEAD cipher, the block size of the cipher (if\n appropriate), and the lengths of explicit and implicit\n initialization vectors (or nonces).\n\n MAC algorithm\n An algorithm to be used for message authentication. This\n specification includes the size of the value returned by the MAC\n algorithm.\n\n compression algorithm\n An algorithm to be used for data compression. This specification\n must include all information the algorithm requires to do\n compression.\n\n master secret\n A 48-byte secret shared between the two peers in the connection.\n\n client random\n A 32-byte value provided by the client.\n\n server random\n A 32-byte value provided by the server.\n\n These parameters are defined in the presentation language as:\n\n enum { server, client } ConnectionEnd;\n\n enum { tls_prf_sha256 } PRFAlgorithm;\n\n enum { null, rc4, 3des, aes }\n BulkCipherAlgorithm;\n\n enum { stream, block, aead } CipherType;\n\n enum { null, hmac_md5, hmac_sha1, hmac_sha256,\n hmac_sha384, hmac_sha512} MACAlgorithm;\n\n enum { null(0), (255) } CompressionMethod;\n\n /* The algorithms specified in CompressionMethod, PRFAlgorithm,\n BulkCipherAlgorithm, and MACAlgorithm may be added to. */\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 17]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ConnectionEnd entity;\n PRFAlgorithm prf_algorithm;\n BulkCipherAlgorithm bulk_cipher_algorithm;\n CipherType cipher_type;\n uint8 enc_key_length;\n uint8 block_length;\n uint8 fixed_iv_length;\n uint8 record_iv_length;\n MACAlgorithm mac_algorithm;\n uint8 mac_length;\n uint8 mac_key_length;\n CompressionMethod compression_algorithm;\n opaque master_secret[48];\n opaque client_random[32];\n opaque server_random[32];\n } SecurityParameters;\n\n The record layer will use the security parameters to generate the\n following six items (some of which are not required by all ciphers,\n and are thus empty):\n\n client write MAC key\n server write MAC key\n client write encryption key\n server write encryption key\n client write IV\n server write IV\n\n The client write parameters are used by the server when receiving and\n processing records and vice versa. The algorithm used for generating\n these items from the security parameters is described in Section 6.3.\n\n Once the security parameters have been set and the keys have been\n generated, the connection states can be instantiated by making them\n the current states. These current states MUST be updated for each\n record processed. Each connection state includes the following\n elements:\n\n compression state\n The current state of the compression algorithm.\n\n cipher state\n The current state of the encryption algorithm. This will consist\n of the scheduled key for that connection. For stream ciphers,\n this will also contain whatever state information is necessary to\n allow the stream to continue to encrypt or decrypt data.\n\n\n\n\nDierks & Rescorla Standards Track [Page 18]\n\f\nRFC 5246 TLS August 2008\n\n\n MAC key\n The MAC key for this connection, as generated above.\n\n sequence number\n Each connection state contains a sequence number, which is\n maintained separately for read and write states. The sequence\n number MUST be set to zero whenever a connection state is made the\n active state. Sequence numbers are of type uint64 and may not\n exceed 2^64-1. Sequence numbers do not wrap. If a TLS\n implementation would need to wrap a sequence number, it must\n renegotiate instead. A sequence number is incremented after each\n record: specifically, the first record transmitted under a\n particular connection state MUST use sequence number 0.\n\n6.2. Record Layer\n\n The TLS record layer receives uninterpreted data from higher layers\n in non-empty blocks of arbitrary size.\n\n6.2.1. Fragmentation\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Client\n message boundaries are not preserved in the record layer (i.e.,\n multiple client messages of the same ContentType MAY be coalesced\n into a single TLSPlaintext record, or a single message MAY be\n fragmented across several records).\n\n struct {\n uint8 major;\n uint8 minor;\n } ProtocolVersion;\n\n enum {\n change_cipher_spec(20), alert(21), handshake(22),\n application_data(23), (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type\n The higher-level protocol used to process the enclosed fragment.\n\n\n\n\nDierks & Rescorla Standards Track [Page 19]\n\f\nRFC 5246 TLS August 2008\n\n\n version\n The version of the protocol being employed. This document\n describes TLS Version 1.2, which uses the version { 3, 3 }. The\n version value 3.3 is historical, deriving from the use of {3, 1}\n for TLS 1.0. (See Appendix A.1.) Note that a client that\n supports multiple versions of TLS may not know what version will\n be employed before it receives the ServerHello. See Appendix E\n for discussion about what record layer version number should be\n employed for ClientHello.\n\n length\n The length (in bytes) of the following TLSPlaintext.fragment. The\n length MUST NOT exceed 2^14.\n\n fragment\n The application data. This data is transparent and treated as an\n independent block to be dealt with by the higher-level protocol\n specified by the type field.\n\n Implementations MUST NOT send zero-length fragments of Handshake,\n Alert, or ChangeCipherSpec content types. Zero-length fragments of\n Application data MAY be sent as they are potentially useful as a\n traffic analysis countermeasure.\n\n Note: Data of different TLS record layer content types MAY be\n interleaved. Application data is generally of lower precedence for\n transmission than other content types. However, records MUST be\n delivered to the network in the same order as they are protected by\n the record layer. Recipients MUST receive and process interleaved\n application layer traffic during handshakes subsequent to the first\n one on a connection.\n\n6.2.2. Record Compression and Decompression\n\n All records are compressed using the compression algorithm defined in\n the current session state. There is always an active compression\n algorithm; however, initially it is defined as\n CompressionMethod.null. The compression algorithm translates a\n TLSPlaintext structure into a TLSCompressed structure. Compression\n functions are initialized with default state information whenever a\n connection state is made active. [RFC3749] describes compression\n algorithms for TLS.\n\n Compression must be lossless and may not increase the content length\n by more than 1024 bytes. If the decompression function encounters a\n TLSCompressed.fragment that would decompress to a length in excess of\n 2^14 bytes, it MUST report a fatal decompression failure error.\n\n\n\n\nDierks & Rescorla Standards Track [Page 20]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ContentType type; /* same as TLSPlaintext.type */\n ProtocolVersion version;/* same as TLSPlaintext.version */\n uint16 length;\n opaque fragment[TLSCompressed.length];\n } TLSCompressed;\n\n length\n The length (in bytes) of the following TLSCompressed.fragment.\n The length MUST NOT exceed 2^14 + 1024.\n\n fragment\n The compressed form of TLSPlaintext.fragment.\n\n Note: A CompressionMethod.null operation is an identity operation;\n no fields are altered.\n\n Implementation note: Decompression functions are responsible for\n ensuring that messages cannot cause internal buffer overflows.\n\n6.2.3. Record Payload Protection\n\n The encryption and MAC functions translate a TLSCompressed\n structure into a TLSCiphertext. The decryption functions reverse\n the process. The MAC of the record also includes a sequence\n number so that missing, extra, or repeated messages are\n detectable.\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n select (SecurityParameters.cipher_type) {\n case stream: GenericStreamCipher;\n case block: GenericBlockCipher;\n case aead: GenericAEADCipher;\n } fragment;\n } TLSCiphertext;\n\n type\n The type field is identical to TLSCompressed.type.\n\n version\n The version field is identical to TLSCompressed.version.\n\n length\n The length (in bytes) of the following TLSCiphertext.fragment.\n The length MUST NOT exceed 2^14 + 2048.\n\n\n\nDierks & Rescorla Standards Track [Page 21]\n\f\nRFC 5246 TLS August 2008\n\n\n fragment\n The encrypted form of TLSCompressed.fragment, with the MAC.\n\n6.2.3.1. Null or Standard Stream Cipher\n\n Stream ciphers (including BulkCipherAlgorithm.null; see Appendix A.6)\n convert TLSCompressed.fragment structures to and from stream\n TLSCiphertext.fragment structures.\n\n stream-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n } GenericStreamCipher;\n\n The MAC is generated as:\n\n MAC(MAC_write_key, seq_num +\n TLSCompressed.type +\n TLSCompressed.version +\n TLSCompressed.length +\n TLSCompressed.fragment);\n\n where \"+\" denotes concatenation.\n\n seq_num\n The sequence number for this record.\n\n MAC\n The MAC algorithm specified by SecurityParameters.mac_algorithm.\n\n Note that the MAC is computed before encryption. The stream cipher\n encrypts the entire block, including the MAC. For stream ciphers\n that do not use a synchronization vector (such as RC4), the stream\n cipher state from the end of one record is simply used on the\n subsequent packet. If the cipher suite is TLS_NULL_WITH_NULL_NULL,\n encryption consists of the identity operation (i.e., the data is not\n encrypted, and the MAC size is zero, implying that no MAC is used).\n For both null and stream ciphers, TLSCiphertext.length is\n TLSCompressed.length plus SecurityParameters.mac_length.\n\n6.2.3.2. CBC Block Cipher\n\n For block ciphers (such as 3DES or AES), the encryption and MAC\n functions convert TLSCompressed.fragment structures to and from block\n TLSCiphertext.fragment structures.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 22]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n opaque IV[SecurityParameters.record_iv_length];\n block-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n uint8 padding[GenericBlockCipher.padding_length];\n uint8 padding_length;\n };\n } GenericBlockCipher;\n\n The MAC is generated as described in Section 6.2.3.1.\n\n IV\n The Initialization Vector (IV) SHOULD be chosen at random, and\n MUST be unpredictable. Note that in versions of TLS prior to 1.1,\n there was no IV field, and the last ciphertext block of the\n previous record (the \"CBC residue\") was used as the IV. This was\n changed to prevent the attacks described in [CBCATT]. For block\n ciphers, the IV length is of length\n SecurityParameters.record_iv_length, which is equal to the\n SecurityParameters.block_size.\n\n padding\n Padding that is added to force the length of the plaintext to be\n an integral multiple of the block cipher's block length. The\n padding MAY be any length up to 255 bytes, as long as it results\n in the TLSCiphertext.length being an integral multiple of the\n block length. Lengths longer than necessary might be desirable to\n frustrate attacks on a protocol that are based on analysis of the\n lengths of exchanged messages. Each uint8 in the padding data\n vector MUST be filled with the padding length value. The receiver\n MUST check this padding and MUST use the bad_record_mac alert to\n indicate padding errors.\n\n padding_length\n The padding length MUST be such that the total size of the\n GenericBlockCipher structure is a multiple of the cipher's block\n length. Legal values range from zero to 255, inclusive. This\n length specifies the length of the padding field exclusive of the\n padding_length field itself.\n\n The encrypted data length (TLSCiphertext.length) is one more than the\n sum of SecurityParameters.block_length, TLSCompressed.length,\n SecurityParameters.mac_length, and padding_length.\n\n Example: If the block length is 8 bytes, the content length\n (TLSCompressed.length) is 61 bytes, and the MAC length is 20 bytes,\n then the length before padding is 82 bytes (this does not include the\n\n\n\nDierks & Rescorla Standards Track [Page 23]\n\f\nRFC 5246 TLS August 2008\n\n\n IV. Thus, the padding length modulo 8 must be equal to 6 in order to\n make the total length an even multiple of 8 bytes (the block length).\n The padding length can be 6, 14, 22, and so on, through 254. If the\n padding length were the minimum necessary, 6, the padding would be 6\n bytes, each containing the value 6. Thus, the last 8 octets of the\n GenericBlockCipher before block encryption would be xx 06 06 06 06 06\n 06 06, where xx is the last octet of the MAC.\n\n Note: With block ciphers in CBC mode (Cipher Block Chaining), it is\n critical that the entire plaintext of the record be known before any\n ciphertext is transmitted. Otherwise, it is possible for the\n attacker to mount the attack described in [CBCATT].\n\n Implementation note: Canvel et al. [CBCTIME] have demonstrated a\n timing attack on CBC padding based on the time required to compute\n the MAC. In order to defend against this attack, implementations\n MUST ensure that record processing time is essentially the same\n whether or not the padding is correct. In general, the best way to\n do this is to compute the MAC even if the padding is incorrect, and\n only then reject the packet. For instance, if the pad appears to be\n incorrect, the implementation might assume a zero-length pad and then\n compute the MAC. This leaves a small timing channel, since MAC\n performance depends to some extent on the size of the data fragment,\n but it is not believed to be large enough to be exploitable, due to\n the large block size of existing MACs and the small size of the\n timing signal.\n\n6.2.3.3. AEAD Ciphers\n\n For AEAD [AEAD] ciphers (such as [CCM] or [GCM]), the AEAD function\n converts TLSCompressed.fragment structures to and from AEAD\n TLSCiphertext.fragment structures.\n\n struct {\n opaque nonce_explicit[SecurityParameters.record_iv_length];\n aead-ciphered struct {\n opaque content[TLSCompressed.length];\n };\n } GenericAEADCipher;\n\n AEAD ciphers take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [AEAD]. The key is either the\n client_write_key or the server_write_key. No MAC key is used.\n\n Each AEAD cipher suite MUST specify how the nonce supplied to the\n AEAD operation is constructed, and what is the length of the\n GenericAEADCipher.nonce_explicit part. In many cases, it is\n\n\n\nDierks & Rescorla Standards Track [Page 24]\n\f\nRFC 5246 TLS August 2008\n\n\n appropriate to use the partially implicit nonce technique described\n in Section 3.2.1 of [AEAD]; with record_iv_length being the length of\n the explicit part. In this case, the implicit part SHOULD be derived\n from key_block as client_write_iv and server_write_iv (as described\n in Section 6.3), and the explicit part is included in\n GenericAEAEDCipher.nonce_explicit.\n\n The plaintext is the TLSCompressed.fragment.\n\n The additional authenticated data, which we denote as\n additional_data, is defined as follows:\n\n additional_data = seq_num + TLSCompressed.type +\n TLSCompressed.version + TLSCompressed.length;\n\n where \"+\" denotes concatenation.\n\n The aead_output consists of the ciphertext output by the AEAD\n encryption operation. The length will generally be larger than\n TLSCompressed.length, but by an amount that varies with the AEAD\n cipher. Since the ciphers might incorporate padding, the amount of\n overhead could vary with different TLSCompressed.length values. Each\n AEAD cipher MUST NOT produce an expansion of greater than 1024 bytes.\n Symbolically,\n\n AEADEncrypted = AEAD-Encrypt(write_key, nonce, plaintext,\n additional_data)\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, the \"additional_data\", and the AEADEncrypted value. The\n output is either the plaintext or an error indicating that the\n decryption failed. There is no separate integrity check. That is:\n\n TLSCompressed.fragment = AEAD-Decrypt(write_key, nonce,\n AEADEncrypted,\n additional_data)\n\n If the decryption fails, a fatal bad_record_mac alert MUST be\n generated.\n\n6.3. Key Calculation\n\n The Record Protocol requires an algorithm to generate keys required\n by the current connection state (see Appendix A.6) from the security\n parameters provided by the handshake protocol.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 25]\n\f\nRFC 5246 TLS August 2008\n\n\n The master secret is expanded into a sequence of secure bytes, which\n is then split to a client write MAC key, a server write MAC key, a\n client write encryption key, and a server write encryption key. Each\n of these is generated from the byte sequence in that order. Unused\n values are empty. Some AEAD ciphers may additionally require a\n client write IV and a server write IV (see Section 6.2.3.3).\n\n When keys and MAC keys are generated, the master secret is used as an\n entropy source.\n\n To generate the key material, compute\n\n key_block = PRF(SecurityParameters.master_secret,\n \"key expansion\",\n SecurityParameters.server_random +\n SecurityParameters.client_random);\n\n until enough output has been generated. Then, the key_block is\n partitioned as follows:\n\n client_write_MAC_key[SecurityParameters.mac_key_length]\n server_write_MAC_key[SecurityParameters.mac_key_length]\n client_write_key[SecurityParameters.enc_key_length]\n server_write_key[SecurityParameters.enc_key_length]\n client_write_IV[SecurityParameters.fixed_iv_length]\n server_write_IV[SecurityParameters.fixed_iv_length]\n\n Currently, the client_write_IV and server_write_IV are only generated\n for implicit nonce techniques as described in Section 3.2.1 of\n [AEAD].\n\n Implementation note: The currently defined cipher suite which\n requires the most material is AES_256_CBC_SHA256. It requires 2 x 32\n byte keys and 2 x 32 byte MAC keys, for a total 128 bytes of key\n material.\n\n7. The TLS Handshaking Protocols\n\n TLS has three subprotocols that are used to allow peers to agree upon\n security parameters for the record layer, to authenticate themselves,\n to instantiate negotiated security parameters, and to report error\n conditions to each other.\n\n The Handshake Protocol is responsible for negotiating a session,\n which consists of the following items:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 26]\n\f\nRFC 5246 TLS August 2008\n\n\n session identifier\n An arbitrary byte sequence chosen by the server to identify an\n active or resumable session state.\n\n peer certificate\n X509v3 [PKIX] certificate of the peer. This element of the state\n may be null.\n\n compression method\n The algorithm used to compress data prior to encryption.\n\n cipher spec\n Specifies the pseudorandom function (PRF) used to generate keying\n material, the bulk data encryption algorithm (such as null, AES,\n etc.) and the MAC algorithm (such as HMAC-SHA1). It also defines\n cryptographic attributes such as the mac_length. (See Appendix\n A.6 for formal definition.)\n\n master secret\n 48-byte secret shared between the client and server.\n\n is resumable\n A flag indicating whether the session can be used to initiate new\n connections.\n\n These items are then used to create security parameters for use by\n the record layer when protecting application data. Many connections\n can be instantiated using the same session through the resumption\n feature of the TLS Handshake Protocol.\n\n7.1. Change Cipher Spec Protocol\n\n The change cipher spec protocol exists to signal transitions in\n ciphering strategies. The protocol consists of a single message,\n which is encrypted and compressed under the current (not the pending)\n connection state. The message consists of a single byte of value 1.\n\n struct {\n enum { change_cipher_spec(1), (255) } type;\n } ChangeCipherSpec;\n\n The ChangeCipherSpec message is sent by both the client and the\n server to notify the receiving party that subsequent records will be\n protected under the newly negotiated CipherSpec and keys. Reception\n of this message causes the receiver to instruct the record layer to\n immediately copy the read pending state into the read current state.\n Immediately after sending this message, the sender MUST instruct the\n record layer to make the write pending state the write active state.\n\n\n\nDierks & Rescorla Standards Track [Page 27]\n\f\nRFC 5246 TLS August 2008\n\n\n (See Section 6.1.) The ChangeCipherSpec message is sent during the\n handshake after the security parameters have been agreed upon, but\n before the verifying Finished message is sent.\n\n Note: If a rehandshake occurs while data is flowing on a connection,\n the communicating parties may continue to send data using the old\n CipherSpec. However, once the ChangeCipherSpec has been sent, the\n new CipherSpec MUST be used. The first side to send the\n ChangeCipherSpec does not know that the other side has finished\n computing the new keying material (e.g., if it has to perform a\n time-consuming public key operation). Thus, a small window of time,\n during which the recipient must buffer the data, MAY exist. In\n practice, with modern machines this interval is likely to be fairly\n short.\n\n7.2. Alert Protocol\n\n One of the content types supported by the TLS record layer is the\n alert type. Alert messages convey the severity of the message\n (warning or fatal) and a description of the alert. Alert messages\n with a level of fatal result in the immediate termination of the\n connection. In this case, other connections corresponding to the\n session may continue, but the session identifier MUST be invalidated,\n preventing the failed session from being used to establish new\n connections. Like other messages, alert messages are encrypted and\n compressed, as specified by the current connection state.\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n\n\n\nDierks & Rescorla Standards Track [Page 28]\n\f\nRFC 5246 TLS August 2008\n\n\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n7.2.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack. Either party may\n initiate the exchange of closing messages.\n\n close_notify\n This message notifies the recipient that the sender will not send\n any more messages on this connection. Note that as of TLS 1.1,\n failure to properly close a connection no longer requires that a\n session not be resumed. This is a change from TLS 1.0 to conform\n with widespread implementation practice.\n\n Either party may initiate a close by sending a close_notify alert.\n Any data received after a closure alert is ignored.\n\n Unless some other fatal alert has been transmitted, each party is\n required to send a close_notify alert before closing the write side\n of the connection. The other party MUST respond with a close_notify\n alert of its own and close down the connection immediately,\n discarding any pending writes. It is not required for the initiator\n of the close to wait for the responding close_notify alert before\n closing the read side of the connection.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation must receive the responding\n close_notify alert before indicating to the application layer that\n the TLS connection has ended. If the application protocol will not\n transfer any additional data, but will only close the underlying\n transport connection, then the implementation MAY choose to close the\n transport without waiting for the responding close_notify. No part\n\n\n\n\nDierks & Rescorla Standards Track [Page 29]\n\f\nRFC 5246 TLS August 2008\n\n\n of this standard should be taken to dictate the manner in which a\n usage profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing a connection reliably delivers\n pending data before destroying the transport.\n\n7.2.2. Error Alerts\n\n Error handling in the TLS Handshake protocol is very simple. When an\n error is detected, the detecting party sends a message to the other\n party. Upon transmission or receipt of a fatal alert message, both\n parties immediately close the connection. Servers and clients MUST\n forget any session-identifiers, keys, and secrets associated with a\n failed connection. Thus, any connection terminated with a fatal\n alert MUST NOT be resumed.\n\n Whenever an implementation encounters a condition which is defined as\n a fatal alert, it MUST send the appropriate alert prior to closing\n the connection. For all errors where an alert level is not\n explicitly specified, the sending party MAY determine at its\n discretion whether to treat this as a fatal error or not. If the\n implementation chooses to send an alert but intends to close the\n connection immediately afterwards, it MUST send that alert at the\n fatal alert level.\n\n If an alert with a level of warning is sent and received, generally\n the connection can continue normally. If the receiving party decides\n not to proceed with the connection (e.g., after having received a\n no_renegotiation alert that it is not willing to accept), it SHOULD\n send a fatal alert to terminate the connection. Given this, the\n sending party cannot, in general, know how the receiving party will\n behave. Therefore, warning alerts are not very useful when the\n sending party wants to continue the connection, and thus are\n sometimes omitted. For example, if a peer decides to accept an\n expired certificate (perhaps after confirming this with the user) and\n wants to continue the connection, it would not generally send a\n certificate_expired alert.\n\n The following error alerts are defined:\n\n unexpected_message\n An inappropriate message was received. This alert is always fatal\n and should never be observed in communication between proper\n implementations.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 30]\n\f\nRFC 5246 TLS August 2008\n\n\n bad_record_mac\n This alert is returned if a record is received with an incorrect\n MAC. This alert also MUST be returned if an alert is sent because\n a TLSCiphertext decrypted in an invalid way: either it wasn't an\n even multiple of the block length, or its padding values, when\n checked, weren't correct. This message is always fatal and should\n never be observed in communication between proper implementations\n (except when messages were corrupted in the network).\n\n decryption_failed_RESERVED\n This alert was used in some earlier versions of TLS, and may have\n permitted certain attacks against the CBC mode [CBCATT]. It MUST\n NOT be sent by compliant implementations.\n\n record_overflow\n A TLSCiphertext record was received that had a length more than\n 2^14+2048 bytes, or a record decrypted to a TLSCompressed record\n with more than 2^14+1024 bytes. This message is always fatal and\n should never be observed in communication between proper\n implementations (except when messages were corrupted in the\n network).\n\n decompression_failure\n The decompression function received improper input (e.g., data\n that would expand to excessive length). This message is always\n fatal and should never be observed in communication between proper\n implementations.\n\n handshake_failure\n Reception of a handshake_failure alert message indicates that the\n sender was unable to negotiate an acceptable set of security\n parameters given the options available. This is a fatal error.\n\n no_certificate_RESERVED\n This alert was used in SSLv3 but not any version of TLS. It MUST\n NOT be sent by compliant implementations.\n\n bad_certificate\n A certificate was corrupt, contained signatures that did not\n verify correctly, etc.\n\n unsupported_certificate\n A certificate was of an unsupported type.\n\n certificate_revoked\n A certificate was revoked by its signer.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 31]\n\f\nRFC 5246 TLS August 2008\n\n\n certificate_expired\n A certificate has expired or is not currently valid.\n\n certificate_unknown\n Some other (unspecified) issue arose in processing the\n certificate, rendering it unacceptable.\n\n illegal_parameter\n A field in the handshake was out of range or inconsistent with\n other fields. This message is always fatal.\n\n unknown_ca\n A valid certificate chain or partial chain was received, but the\n certificate was not accepted because the CA certificate could not\n be located or couldn't be matched with a known, trusted CA. This\n message is always fatal.\n\n access_denied\n A valid certificate was received, but when access control was\n applied, the sender decided not to proceed with negotiation. This\n message is always fatal.\n\n decode_error\n A message could not be decoded because some field was out of the\n specified range or the length of the message was incorrect. This\n message is always fatal and should never be observed in\n communication between proper implementations (except when messages\n were corrupted in the network).\n\n decrypt_error\n A handshake cryptographic operation failed, including being unable\n to correctly verify a signature or validate a Finished message.\n This message is always fatal.\n\n export_restriction_RESERVED\n This alert was used in some earlier versions of TLS. It MUST NOT\n be sent by compliant implementations.\n\n protocol_version\n The protocol version the client has attempted to negotiate is\n recognized but not supported. (For example, old protocol versions\n might be avoided for security reasons.) This message is always\n fatal.\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 32]\n\f\nRFC 5246 TLS August 2008\n\n\n insufficient_security\n Returned instead of handshake_failure when a negotiation has\n failed specifically because the server requires ciphers more\n secure than those supported by the client. This message is always\n fatal.\n\n internal_error\n An internal error unrelated to the peer or the correctness of the\n protocol (such as a memory allocation failure) makes it impossible\n to continue. This message is always fatal.\n\n\n\n===== DOCUMENT =====\n\nNetwork Working Group E. Rescorla\nInternet-Draft RTFM, Inc.\nObsoletes: 5077, 5246, 6961 (if March 20, 2018\n approved)\nUpdates: 4492, 5705, 6066 (if approved)\nIntended status: Standards Track\nExpires: September 21, 2018\n\n\n\n===== DOCUMENT =====\n\n user_canceled\n This handshake is being canceled for some reason unrelated to a\n protocol failure. If the user cancels an operation after the\n handshake is complete, just closing the connection by sending a\n close_notify is more appropriate. This alert should be followed\n by a close_notify. This message is generally a warning.\n\n no_renegotiation\n Sent by the client in response to a hello request or by the server\n in response to a client hello after initial handshaking. Either\n of these would normally lead to renegotiation; when that is not\n appropriate, the recipient should respond with this alert. At\n that point, the original requester can decide whether to proceed\n with the connection. One case where this would be appropriate is\n where a server has spawned a process to satisfy a request; the\n process might receive security parameters (key length,\n authentication, etc.) at startup, and it might be difficult to\n communicate changes to these parameters after that point. This\n message is always a warning.\n\n unsupported_extension\n sent by clients that receive an extended server hello containing\n an extension that they did not put in the corresponding client\n hello. This message is always fatal.\n\n New Alert values are assigned by IANA as described in Section 12.\n\n7.3. Handshake Protocol Overview\n\n The cryptographic parameters of the session state are produced by the\n TLS Handshake Protocol, which operates on top of the TLS record\n layer. When a TLS client and server first start communicating, they\n agree on a protocol version, select cryptographic algorithms,\n optionally authenticate each other, and use public-key encryption\n techniques to generate shared secrets.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 33]\n\f\nRFC 5246 TLS August 2008\n\n\n The TLS Handshake Protocol involves the following steps:\n\n - Exchange hello messages to agree on algorithms, exchange random\n values, and check for session resumption.\n\n - Exchange the necessary cryptographic parameters to allow the\n client and server to agree on a premaster secret.\n\n - Exchange certificates and cryptographic information to allow the\n client and server to authenticate themselves.\n\n - Generate a master secret from the premaster secret and exchanged\n random values.\n\n - Provide security parameters to the record layer.\n\n - Allow the client and server to verify that their peer has\n calculated the same security parameters and that the handshake\n occurred without tampering by an attacker.\n\n Note that higher layers should not be overly reliant on whether TLS\n always negotiates the strongest possible connection between two\n peers. There are a number of ways in which a man-in-the-middle\n attacker can attempt to make two entities drop down to the least\n secure method they support. The protocol has been designed to\n minimize this risk, but there are still attacks available: for\n example, an attacker could block access to the port a secure service\n runs on, or attempt to get the peers to negotiate an unauthenticated\n connection. The fundamental rule is that higher levels must be\n cognizant of what their security requirements are and never transmit\n information over a channel less secure than what they require. The\n TLS protocol is secure in that any cipher suite offers its promised\n level of security: if you negotiate 3DES with a 1024-bit RSA key\n exchange with a host whose certificate you have verified, you can\n expect to be that secure.\n\n These goals are achieved by the handshake protocol, which can be\n summarized as follows: The client sends a ClientHello message to\n which the server must respond with a ServerHello message, or else a\n fatal error will occur and the connection will fail. The ClientHello\n and ServerHello are used to establish security enhancement\n capabilities between client and server. The ClientHello and\n ServerHello establish the following attributes: Protocol Version,\n Session ID, Cipher Suite, and Compression Method. Additionally, two\n random values are generated and exchanged: ClientHello.random and\n ServerHello.random.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 34]\n\f\nRFC 5246 TLS August 2008\n\n\n The actual key exchange uses up to four messages: the server\n Certificate, the ServerKeyExchange, the client Certificate, and the\n ClientKeyExchange. New key exchange methods can be created by\n specifying a format for these messages and by defining the use of the\n messages to allow the client and server to agree upon a shared\n secret. This secret MUST be quite long; currently defined key\n exchange methods exchange secrets that range from 46 bytes upwards.\n\n Following the hello messages, the server will send its certificate in\n a Certificate message if it is to be authenticated. Additionally, a\n ServerKeyExchange message may be sent, if it is required (e.g., if\n the server has no certificate, or if its certificate is for signing\n only). If the server is authenticated, it may request a certificate\n from the client, if that is appropriate to the cipher suite selected.\n Next, the server will send the ServerHelloDone message, indicating\n that the hello-message phase of the handshake is complete. The\n server will then wait for a client response. If the server has sent\n a CertificateRequest message, the client MUST send the Certificate\n message. The ClientKeyExchange message is now sent, and the content\n of that message will depend on the public key algorithm selected\n between the ClientHello and the ServerHello. If the client has sent\n a certificate with signing ability, a digitally-signed\n CertificateVerify message is sent to explicitly verify possession of\n the private key in the certificate.\n\n At this point, a ChangeCipherSpec message is sent by the client, and\n the client copies the pending Cipher Spec into the current Cipher\n Spec. The client then immediately sends the Finished message under\n the new algorithms, keys, and secrets. In response, the server will\n send its own ChangeCipherSpec message, transfer the pending to the\n current Cipher Spec, and send its Finished message under the new\n Cipher Spec. At this point, the handshake is complete, and the\n client and server may begin to exchange application layer data. (See\n flow chart below.) Application data MUST NOT be sent prior to the\n completion of the first handshake (before a cipher suite other than\n TLS_NULL_WITH_NULL_NULL is established).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 35]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1. Message flow for a full handshake\n\n * Indicates optional or situation-dependent messages that are not\n always sent.\n\n Note: To help avoid pipeline stalls, ChangeCipherSpec is an\n independent TLS protocol content type, and is not actually a TLS\n handshake message.\n\n When the client and server decide to resume a previous session or\n duplicate an existing session (instead of negotiating new security\n parameters), the message flow is as follows:\n\n The client sends a ClientHello using the Session ID of the session to\n be resumed. The server then checks its session cache for a match.\n If a match is found, and the server is willing to re-establish the\n connection under the specified session state, it will send a\n ServerHello with the same Session ID value. At this point, both\n client and server MUST send ChangeCipherSpec messages and proceed\n directly to Finished messages. Once the re-establishment is\n complete, the client and server MAY begin to exchange application\n layer data. (See flow chart below.) If a Session ID match is not\n found, the server generates a new session ID, and the TLS client and\n server perform a full handshake.\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 36]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2. Message flow for an abbreviated handshake\n\n The contents and significance of each message will be presented in\n detail in the following sections.\n\n7.4. Handshake Protocol\n\n The TLS Handshake Protocol is one of the defined higher-level clients\n of the TLS Record Protocol. This protocol is used to negotiate the\n secure attributes of a session. Handshake messages are supplied to\n the TLS record layer, where they are encapsulated within one or more\n TLSPlaintext structures, which are processed and transmitted as\n specified by the current active session state.\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n } body;\n } Handshake;\n\n\n\nDierks & Rescorla Standards Track [Page 37]\n\f\nRFC 5246 TLS August 2008\n\n\n The handshake protocol messages are presented below in the order they\n MUST be sent; sending handshake messages in an unexpected order\n results in a fatal error. Unneeded handshake messages can be\n omitted, however. Note one exception to the ordering: the\n Certificate message is used twice in the handshake (from server to\n client, then from client to server), but described only in its first\n position. The one message that is not bound by these ordering rules\n is the HelloRequest message, which can be sent at any time, but which\n SHOULD be ignored by the client if it arrives in the middle of a\n handshake.\n\n New handshake message types are assigned by IANA as described in\n Section 12.\n\n7.4.1. Hello Messages\n\n The hello phase messages are used to exchange security enhancement\n capabilities between the client and server. When a new session\n begins, the record layer's connection state encryption, hash, and\n compression algorithms are initialized to null. The current\n connection state is used for renegotiation messages.\n\n7.4.1.1. Hello Request\n\n When this message will be sent:\n\n The HelloRequest message MAY be sent by the server at any time.\n\n Meaning of this message:\n\n HelloRequest is a simple notification that the client should begin\n the negotiation process anew. In response, the client should send\n a ClientHello message when convenient. This message is not\n intended to establish which side is the client or server but\n merely to initiate a new negotiation. Servers SHOULD NOT send a\n HelloRequest immediately upon the client's initial connection. It\n is the client's job to send a ClientHello at that time.\n\n This message will be ignored by the client if the client is\n currently negotiating a session. This message MAY be ignored by\n the client if it does not wish to renegotiate a session, or the\n client may, if it wishes, respond with a no_renegotiation alert.\n Since handshake messages are intended to have transmission\n precedence over application data, it is expected that the\n negotiation will begin before no more than a few records are\n received from the client. If the server sends a HelloRequest but\n does not receive a ClientHello in response, it may close the\n connection with a fatal alert.\n\n\n\nDierks & Rescorla Standards Track [Page 38]\n\f\nRFC 5246 TLS August 2008\n\n\n After sending a HelloRequest, servers SHOULD NOT repeat the\n request until the subsequent handshake negotiation is complete.\n\n Structure of this message:\n\n struct { } HelloRequest;\n\n This message MUST NOT be included in the message hashes that are\n maintained throughout the handshake and used in the Finished messages\n and the certificate verify message.\n\n7.4.1.2. Client Hello\n\n When this message will be sent:\n\n When a client first connects to a server, it is required to send\n the ClientHello as its first message. The client can also send a\n ClientHello in response to a HelloRequest or on its own initiative\n in order to renegotiate the security parameters in an existing\n connection.\n\n Structure of this message:\n\n The ClientHello message includes a random structure, which is used\n later in the protocol.\n\n struct {\n uint32 gmt_unix_time;\n opaque random_bytes[28];\n } Random;\n\n gmt_unix_time\n The current time and date in standard UNIX 32-bit format\n (seconds since the midnight starting Jan 1, 1970, UTC, ignoring\n leap seconds) according to the sender's internal clock. Clocks\n are not required to be set correctly by the basic TLS protocol;\n higher-level or application protocols may define additional\n requirements. Note that, for historical reasons, the data\n element is named using GMT, the predecessor of the current\n worldwide time base, UTC.\n\n random_bytes\n 28 bytes generated by a secure random number generator.\n\n The ClientHello message includes a variable-length session\n identifier. If not empty, the value identifies a session between the\n same client and server whose security parameters the client wishes to\n reuse. The session identifier MAY be from an earlier connection,\n\n\n\nDierks & Rescorla Standards Track [Page 39]\n\f\nRFC 5246 TLS August 2008\n\n\n this connection, or from another currently active connection. The\n second option is useful if the client only wishes to update the\n random structures and derived values of a connection, and the third\n option makes it possible to establish several independent secure\n connections without repeating the full handshake protocol. These\n independent connections may occur sequentially or simultaneously; a\n SessionID becomes valid when the handshake negotiating it completes\n with the exchange of Finished messages and persists until it is\n removed due to aging or because a fatal error was encountered on a\n connection associated with the session. The actual contents of the\n SessionID are defined by the server.\n\n opaque SessionID<0..32>;\n\n Warning: Because the SessionID is transmitted without encryption or\n immediate MAC protection, servers MUST NOT place confidential\n information in session identifiers or let the contents of fake\n session identifiers cause any breach of security. (Note that the\n content of the handshake as a whole, including the SessionID, is\n protected by the Finished messages exchanged at the end of the\n handshake.)\n\n The cipher suite list, passed from the client to the server in the\n ClientHello message, contains the combinations of cryptographic\n algorithms supported by the client in order of the client's\n preference (favorite choice first). Each cipher suite defines a key\n exchange algorithm, a bulk encryption algorithm (including secret key\n length), a MAC algorithm, and a PRF. The server will select a cipher\n suite or, if no acceptable choices are presented, return a handshake\n failure alert and close the connection. If the list contains cipher\n suites the server does not recognize, support, or wish to use, the\n server MUST ignore those cipher suites, and process the remaining\n ones as usual.\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n The ClientHello includes a list of compression algorithms supported\n by the client, ordered according to the client's preference.\n\n enum { null(0), (255) } CompressionMethod;\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 40]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ProtocolVersion client_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suites<2..2^16-2>;\n CompressionMethod compression_methods<1..2^8-1>;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ClientHello;\n\n TLS allows extensions to follow the compression_methods field in an\n extensions block. The presence of extensions can be detected by\n determining whether there are bytes following the compression_methods\n at the end of the ClientHello. Note that this method of detecting\n optional data differs from the normal TLS method of having a\n variable-length field, but it is used for compatibility with TLS\n before extensions were defined.\n\n client_version\n The version of the TLS protocol by which the client wishes to\n communicate during this session. This SHOULD be the latest\n (highest valued) version supported by the client. For this\n version of the specification, the version will be 3.3 (see\n Appendix E for details about backward compatibility).\n\n random\n A client-generated random structure.\n\n session_id\n The ID of a session the client wishes to use for this connection.\n This field is empty if no session_id is available, or if the\n client wishes to generate new security parameters.\n\n cipher_suites\n This is a list of the cryptographic options supported by the\n client, with the client's first preference first. If the\n session_id field is not empty (implying a session resumption\n request), this vector MUST include at least the cipher_suite from\n that session. Values are defined in Appendix A.5.\n\n compression_methods\n This is a list of the compression methods supported by the client,\n sorted by client preference. If the session_id field is not empty\n (implying a session resumption request), it MUST include the\n\n\n\nDierks & Rescorla Standards Track [Page 41]\n\f\nRFC 5246 TLS August 2008\n\n\n compression_method from that session. This vector MUST contain,\n and all implementations MUST support, CompressionMethod.null.\n Thus, a client and server will always be able to agree on a\n compression method.\n\n extensions\n Clients MAY request extended functionality from servers by sending\n data in the extensions field. The actual \"Extension\" format is\n defined in Section 7.4.1.4.\n\n In the event that a client requests additional functionality using\n extensions, and this functionality is not supplied by the server, the\n client MAY abort the handshake. A server MUST accept ClientHello\n messages both with and without the extensions field, and (as for all\n other messages) it MUST check that the amount of data in the message\n precisely matches one of these formats; if not, then it MUST send a\n fatal \"decode_error\" alert.\n\n After sending the ClientHello message, the client waits for a\n ServerHello message. Any handshake message returned by the server,\n except for a HelloRequest, is treated as a fatal error.\n\n7.4.1.3. Server Hello\n\n When this message will be sent:\n\n The server will send this message in response to a ClientHello\n message when it was able to find an acceptable set of algorithms.\n If it cannot find such a match, it will respond with a handshake\n failure alert.\n\n Structure of this message:\n\n struct {\n ProtocolVersion server_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ServerHello;\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 42]\n\f\nRFC 5246 TLS August 2008\n\n\n The presence of extensions can be detected by determining whether\n there are bytes following the compression_method field at the end of\n the ServerHello.\n\n server_version\n This field will contain the lower of that suggested by the client\n in the client hello and the highest supported by the server. For\n this version of the specification, the version is 3.3. (See\n Appendix E for details about backward compatibility.)\n\n random\n This structure is generated by the server and MUST be\n independently generated from the ClientHello.random.\n\n session_id\n This is the identity of the session corresponding to this\n connection. If the ClientHello.session_id was non-empty, the\n server will look in its session cache for a match. If a match is\n found and the server is willing to establish the new connection\n using the specified session state, the server will respond with\n the same value as was supplied by the client. This indicates a\n resumed session and dictates that the parties must proceed\n directly to the Finished messages. Otherwise, this field will\n contain a different value identifying the new session. The server\n may return an empty session_id to indicate that the session will\n not be cached and therefore cannot be resumed. If a session is\n resumed, it must be resumed using the same cipher suite it was\n originally negotiated with. Note that there is no requirement\n that the server resume any session even if it had formerly\n provided a session_id. Clients MUST be prepared to do a full\n negotiation -- including negotiating new cipher suites -- during\n any handshake.\n\n cipher_suite\n The single cipher suite selected by the server from the list in\n ClientHello.cipher_suites. For resumed sessions, this field is\n the value from the state of the session being resumed.\n\n compression_method\n The single compression algorithm selected by the server from the\n list in ClientHello.compression_methods. For resumed sessions,\n this field is the value from the resumed session state.\n\n extensions\n A list of extensions. Note that only extensions offered by the\n client can appear in the server's list.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 43]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.1.4. Hello Extensions\n\n The extension format is:\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n signature_algorithms(13), (65535)\n } ExtensionType;\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The initial set of extensions is defined in a companion document\n [TLSEXT]. The list of extension types is maintained by IANA as\n described in Section 12.\n\n An extension type MUST NOT appear in the ServerHello unless the same\n extension type appeared in the corresponding ClientHello. If a\n client receives an extension type in ServerHello that it did not\n request in the associated ClientHello, it MUST abort the handshake\n with an unsupported_extension fatal alert.\n\n Nonetheless, \"server-oriented\" extensions may be provided in the\n future within this framework. Such an extension (say, of type x)\n would require the client to first send an extension of type x in a\n ClientHello with empty extension_data to indicate that it supports\n the extension type. In this case, the client is offering the\n capability to understand the extension type, and the server is taking\n the client up on its offer.\n\n When multiple extensions of different types are present in the\n ClientHello or ServerHello messages, the extensions MAY appear in any\n order. There MUST NOT be more than one extension of the same type.\n\n Finally, note that extensions can be sent both when starting a new\n session and when requesting session resumption. Indeed, a client\n that requests session resumption does not in general know whether the\n server will accept this request, and therefore it SHOULD send the\n same extensions as it would send if it were not attempting\n resumption.\n\n\n\nDierks & Rescorla Standards Track [Page 44]\n\f\nRFC 5246 TLS August 2008\n\n\n In general, the specification of each extension type needs to\n describe the effect of the extension both during full handshake and\n session resumption. Most current TLS extensions are relevant only\n when a session is initiated: when an older session is resumed, the\n server does not process these extensions in Client Hello, and does\n not include them in Server Hello. However, some extensions may\n specify different behavior during session resumption.\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions, and some are simply refusals to support particular\n features. In general, error alerts should be used for the former,\n and a field in the server extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem.\n\n Often the fact that the extension fields are included in the\n inputs to the Finished message hashes will be sufficient, but\n extreme care is needed when the extension changes the meaning of\n messages sent in the handshake phase. Designers and implementors\n should be aware of the fact that until the handshake has been\n authenticated, active attackers can modify messages and insert,\n remove, or replace extensions.\n\n - It would be technically possible to use extensions to change major\n aspects of the design of TLS; for example the design of cipher\n suite negotiation. This is not recommended; it would be more\n appropriate to define a new version of TLS -- particularly since\n the TLS handshake algorithms have specific protection against\n version rollback attacks based on the version number, and the\n possibility of version rollback should be a significant\n consideration in any major design change.\n\n7.4.1.4.1. Signature Algorithms\n\n The client uses the \"signature_algorithms\" extension to indicate to\n the server which signature/hash algorithm pairs may be used in\n digital signatures. The \"extension_data\" field of this extension\n contains a \"supported_signature_algorithms\" value.\n\n\n\nDierks & Rescorla Standards Track [Page 45]\n\f\nRFC 5246 TLS August 2008\n\n\n enum {\n none(0), md5(1), sha1(2), sha224(3), sha256(4), sha384(5),\n sha512(6), (255)\n } HashAlgorithm;\n\n enum { anonymous(0), rsa(1), dsa(2), ecdsa(3), (255) }\n SignatureAlgorithm;\n\n struct {\n HashAlgorithm hash;\n SignatureAlgorithm signature;\n } SignatureAndHashAlgorithm;\n\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2..2^16-2>;\n\n Each SignatureAndHashAlgorithm value lists a single hash/signature\n pair that the client is willing to verify. The values are indicated\n in descending order of preference.\n\n Note: Because not all signature algorithms and hash algorithms may be\n accepted by an implementation (e.g., DSA with SHA-1, but not\n SHA-256), algorithms here are listed in pairs.\n\n hash\n This field indicates the hash algorithm which may be used. The\n values indicate support for unhashed data, MD5 [MD5], SHA-1,\n SHA-224, SHA-256, SHA-384, and SHA-512 [SHS], respectively. The\n \"none\" value is provided for future extensibility, in case of a\n signature algorithm which does not require hashing before signing.\n\n signature\n This field indicates the signature algorithm that may be used.\n The values indicate anonymous signatures, RSASSA-PKCS1-v1_5\n [PKCS1] and DSA [DSS], and ECDSA [ECDSA], respectively. The\n \"anonymous\" value is meaningless in this context but used in\n Section 7.4.3. It MUST NOT appear in this extension.\n\n The semantics of this extension are somewhat complicated because the\n cipher suite indicates permissible signature algorithms but not hash\n algorithms. Sections 7.4.2 and 7.4.3 describe the appropriate rules.\n\n If the client supports only the default hash and signature algorithms\n (listed in this section), it MAY omit the signature_algorithms\n extension. If the client does not support the default algorithms, or\n supports other hash and signature algorithms (and it is willing to\n use them for verifying messages sent by the server, i.e., server\n certificates and server key exchange), it MUST send the\n\n\n\nDierks & Rescorla Standards Track [Page 46]\n\f\nRFC 5246 TLS August 2008\n\n\n signature_algorithms extension, listing the algorithms it is willing\n to accept.\n\n If the client does not send the signature_algorithms extension, the\n server MUST do the following:\n\n - If the negotiated key exchange algorithm is one of (RSA, DHE_RSA,\n DH_RSA, RSA_PSK, ECDH_RSA, ECDHE_RSA), behave as if client had\n sent the value {sha1,rsa}.\n\n - If the negotiated key exchange algorithm is one of (DHE_DSS,\n DH_DSS), behave as if the client had sent the value {sha1,dsa}.\n\n - If the negotiated key exchange algorithm is one of (ECDH_ECDSA,\n ECDHE_ECDSA), behave as if the client had sent value {sha1,ecdsa}.\n\n Note: this is a change from TLS 1.1 where there are no explicit\n rules, but as a practical matter one can assume that the peer\n supports MD5 and SHA-1.\n\n Note: this extension is not meaningful for TLS versions prior to 1.2.\n Clients MUST NOT offer it if they are offering prior versions.\n However, even if clients do offer it, the rules specified in [TLSEXT]\n require servers to ignore extensions they do not understand.\n\n Servers MUST NOT send this extension. TLS servers MUST support\n receiving this extension.\n\n When performing session resumption, this extension is not included in\n Server Hello, and the server ignores the extension in Client Hello\n (if present).\n\n7.4.2. Server Certificate\n\n When this message will be sent:\n\n The server MUST send a Certificate message whenever the agreed-\n upon key exchange method uses certificates for authentication\n (this includes all key exchange methods defined in this document\n except DH_anon). This message will always immediately follow the\n ServerHello message.\n\n Meaning of this message:\n\n This message conveys the server's certificate chain to the client.\n\n The certificate MUST be appropriate for the negotiated cipher\n suite's key exchange algorithm and any negotiated extensions.\n\n\n\nDierks & Rescorla Standards Track [Page 47]\n\f\nRFC 5246 TLS August 2008\n\n\n Structure of this message:\n\n opaque ASN.1Cert<1..2^24-1>;\n\n struct {\n ASN.1Cert certificate_list<0..2^24-1>;\n } Certificate;\n\n certificate_list\n This is a sequence (chain) of certificates. The sender's\n certificate MUST come first in the list. Each following\n certificate MUST directly certify the one preceding it. Because\n certificate validation requires that root keys be distributed\n independently, the self-signed certificate that specifies the root\n certificate authority MAY be omitted from the chain, under the\n assumption that the remote end must already possess it in order to\n validate it in any case.\n\n The same message type and structure will be used for the client's\n response to a certificate request message. Note that a client MAY\n send no certificates if it does not have an appropriate certificate\n to send in response to the server's authentication request.\n\n Note: PKCS #7 [PKCS7] is not used as the format for the certificate\n vector because PKCS #6 [PKCS6] extended certificates are not used.\n Also, PKCS #7 defines a SET rather than a SEQUENCE, making the task\n of parsing the list more difficult.\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3, unless explicitly negotiated\n otherwise (e.g., [TLSPGP]).\n\n - The end entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected key exchange\n algorithm.\n\n Key Exchange Alg. Certificate Key Type\n\n RSA RSA public key; the certificate MUST allow the\n RSA_PSK key to be used for encryption (the\n keyEncipherment bit MUST be set if the key\n usage extension is present).\n Note: RSA_PSK is defined in [TLSPSK].\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 48]\n\f\nRFC 5246 TLS August 2008\n\n\n DHE_RSA RSA public key; the certificate MUST allow the\n ECDHE_RSA key to be used for signing (the\n digitalSignature bit MUST be set if the key\n usage extension is present) with the signature\n scheme and hash algorithm that will be employed\n in the server key exchange message.\n Note: ECDHE_RSA is defined in [TLSECC].\n\n DHE_DSS DSA public key; the certificate MUST allow the\n key to be used for signing with the hash\n algorithm that will be employed in the server\n key exchange message.\n\n DH_DSS Diffie-Hellman public key; the keyAgreement bit\n DH_RSA MUST be set if the key usage extension is\n present.\n\n ECDH_ECDSA ECDH-capable public key; the public key MUST\n ECDH_RSA use a curve and point format supported by the\n client, as described in [TLSECC].\n\n ECDHE_ECDSA ECDSA-capable public key; the certificate MUST\n allow the key to be used for signing with the\n hash algorithm that will be employed in the\n server key exchange message. The public key\n MUST use a curve and point format supported by\n the client, as described in [TLSECC].\n\n - The \"server_name\" and \"trusted_ca_keys\" extensions [TLSEXT] are\n used to guide certificate selection.\n\n\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) Y. Pettersen\nRequest for Comments: 6961 June 2013\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\n\nAbstract\n\n This document defines the Transport Layer Security (TLS) Certificate\n Status Version 2 Extension to allow clients to specify and support\n several certificate status methods. (The use of the Certificate\n Status extension is commonly referred to as \"OCSP stapling\".) Also\n defined is a new method based on the Online Certificate Status\n Protocol (OCSP) that servers can use to provide status information\n about not only the server's own certificate but also the status of\n intermediate certificates in the chain.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6961.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 1]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nCopyright Notice\n\n Copyright (c) 2013 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n1. Introduction\n\n The Transport Layer Security (TLS) Extension [RFC6066] framework\n defines, among other extensions, the Certificate Status extension\n (also referred to as \"OCSP stapling\") that clients can use to request\n the server's copy of the current status of its certificate. The\n benefits of this extension include a reduced number of roundtrips and\n network delays for the client to verify the status of the server's\n certificate and a reduced load on the certificate issuer's status\n response servers, thus solving a problem that can become significant\n when the issued certificate is presented by a frequently visited\n server.\n\n There are two problems with the existing Certificate Status\n extension. First, it does not provide functionality to request the\n status information about intermediate Certification Authority (CA)\n certificates, which means the client has to request status\n information through other methods, such as Certificate Revocation\n Lists (CRLs), introducing further delays. Second, the current format\n of the extension and requirements in the TLS protocol prevent a\n client from offering the server multiple status methods.\n\n\n\nPettersen Standards Track [Page 2]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n Many CAs are now issuing intermediate CA certificates that not only\n specify the publication point for their CRLs in a CRL Distribution\n Point [RFC5280] but also specify a URL for their OCSP [RFC6960]\n server in Authority Information Access [RFC5280]. Given that\n client-cached CRLs are frequently out of date, clients would benefit\n from using OCSP to access up-to-date status information about\n intermediate CA certificates. The benefit to the issuing CA is less\n clear, as providing the bandwidth for the OCSP responder can be\n costly, especially for CAs with many high-traffic subscriber sites,\n and this cost is a concern for many CAs. There are cases where OCSP\n requests for a single high-traffic site caused significant network\n problems for the issuing CA.\n\n Clients will benefit from the TLS server providing certificate status\n information regardless of type, not just for the server certificate\n but also for the intermediate CA certificates. Combining the status\n checks into one extension will reduce the roundtrips needed to\n complete the handshake by the client to just those needed for\n negotiating the TLS connection. Also, for the Certification\n Authorities, the load on their servers will depend on the number of\n certificates they have issued, not on the number of visitors to those\n sites. Additionally, using this extension significantly reduces\n privacy concerns around the clients informing the certificate issuer\n about which sites they are visiting.\n\n For such a new system to be introduced seamlessly, clients need to be\n able to indicate support for the existing OCSP Certificate Status\n method and a new multiple-OCSP mode.\n\n Unfortunately, the definition of the Certificate Status extension\n only allows a single Certificate Status extension to be defined in a\n single extension record in the handshake, and the TLS protocol\n [RFC5246] only allows a single record in the extension list for any\n given extension. This means that it is not possible for clients to\n indicate support for new methods while still supporting older\n methods, which would cause problems for interoperability between\n newer clients and older servers. This will not just be an issue for\n the multiple status request mode proposed above but also for any\n other future status methods that might be introduced. This will be\n true not just for the current PKIX infrastructure [RFC5280] but also\n for alternative PKI structures.\n\n The solution to this problem is to define a new extension,\n \"status_request_v2\", with an extended format that allows the client\n to indicate support for multiple status request methods. This is\n implemented using a list of CertificateStatusRequestItemV2 records in\n the extension record. As the server will select the single status\n\n\n\n\nPettersen Standards Track [Page 3]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n method based on the selected cipher suite and the certificate\n presented, no significant changes are needed in the server's\n extension format.\n\n1.1. Requirements Language\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [RFC2119].\n\n1.2. Presentation Language\n\n This document defines protocol structures using the same conventions\n and presentation language as defined in Section 4 of [RFC5246].\n\n2. Multiple Certificate Status Extension\n\n2.1. New Extension\n\n The extension defined by this document is indicated by\n \"status_request_v2\" in the ExtensionType enum (originally defined by\n [RFC6066]), which uses the following value:\n\n enum {\n status_request_v2(17), (65535)\n } ExtensionType;\n\n2.2. Multiple Certificate Status Request Record\n\n Clients that support a certificate status protocol like OCSP may send\n the \"status_request_v2\" extension to the server in order to use the\n TLS handshake to transfer such data instead of downloading it through\n separate connections. When using this extension, the\n \"extension_data\" field (defined in Section 7.4.1.4 of [RFC5246]) of\n the extension SHALL contain a CertificateStatusRequestListV2 where:\n\n struct {\n CertificateStatusType status_type;\n uint16 request_length; /* Length of request field in bytes */\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n case ocsp_multi: OCSPStatusRequest;\n } request;\n } CertificateStatusRequestItemV2;\n\n enum { ocsp(1), ocsp_multi(2), (255) } CertificateStatusType;\n\n\n\n\n\nPettersen Standards Track [Page 4]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n struct {\n CertificateStatusRequestItemV2\n certificate_status_req_list<1..2^16-1>;\n } CertificateStatusRequestListV2;\n\n In the OCSPStatusRequest (originally defined by [RFC6066]), the\n \"ResponderID\" provides a list of OCSP responders that the client\n trusts. A zero-length \"responder_id_list\" sequence has the special\n meaning that the responders are implicitly known to the server, e.g.,\n by prior arrangement, or are identified by the certificates used by\n the server. \"Extensions\" is a DER encoding [X.690] of the OCSP\n request extensions, and if the server supports the forwarding of OCSP\n request extensions, this value MUST be forwarded without\n modification.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC6960]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a DER-encoded zero-length ASN.1 SEQUENCE,\n which is not valid for the \"Extensions\" type).\n\n Servers that support a client's selection of responders using the\n ResponderID field could implement this selection by matching the\n responder ID values from the client's list with the ResponderIDs of\n known OCSP responders, either by using a binary compare of the values\n or a hash calculation and compare method.\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement). This\n has been clarified in [RFC6960].\n\n The items in the list of CertificateStatusRequestItemV2 entries are\n ordered according to the client's preference (favorite choice first).\n\n A server that receives a client hello containing the\n \"status_request_v2\" extension MAY return a suitable certificate\n status response message to the client along with the server's\n\n\n\nPettersen Standards Track [Page 5]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n certificate message. If OCSP is requested, it SHOULD use the\n information contained in the extension when selecting an OCSP\n responder and SHOULD include request_extensions in the OCSP request.\n\n The server returns a certificate status response along with its\n certificate by sending a \"CertificateStatus\" message (originally\n defined by [RFC6066]) immediately after the \"Certificate\" message\n (Section 7.4.2 of [RFC5246]) (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message in response to a \"status_request_v2\"\n request, then the server MUST have included an extension of type\n \"status_request_v2\" with empty \"extension_data\" in the extended\n server hello.\n\n The \"CertificateStatus\" message is conveyed using the handshake\n message type \"certificate_status\" (defined in [RFC6066]) as follows\n (updated from the definition in [RFC6066]):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n case ocsp_multi: OCSPResponseList;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<0..2^24-1>;\n\n struct {\n OCSPResponse ocsp_response_list<1..2^24-1>;\n } OCSPResponseList;\n\n An \"OCSPResponse\" element (originally defined by [RFC6066]) contains\n a complete, DER-encoded OCSP response (using the ASN.1 [X.680] type\n OCSPResponse defined in [RFC6960]). Only one OCSP response, with a\n length of at least one byte, may be sent for status_type \"ocsp\".\n\n An \"ocsp_response_list\" contains a list of \"OCSPResponse\" elements,\n as specified above, each containing the OCSP response for the\n matching corresponding certificate in the server's Certificate TLS\n handshake message. That is, the first entry is the OCSP response for\n the first certificate in the Certificate list, the second entry is\n the response for the second certificate, and so on. The list MAY\n contain fewer OCSP responses than there were certificates in the\n Certificate handshake message, but there MUST NOT be more responses\n than there were certificates in the list. Individual elements of the\n list MAY have a length of 0 (zero) bytes if the server does not have\n the OCSP response for that particular certificate stored, in which\n\n\n\nPettersen Standards Track [Page 6]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n case the client MUST act as if a response was not received for that\n particular certificate. If the client receives a\n \"ocsp_response_list\" that does not contain a response for one or more\n of the certificates in the completed certificate chain, the client\n SHOULD attempt to validate the certificate using an alternative\n retrieval method, such as downloading the relevant CRL; OCSP SHOULD\n in this situation only be used for the end-entity certificate, not\n intermediate CA certificates, for reasons stated above.\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if it has received a \"status_request_v2\" extension in\n the client hello message and has sent a \"status_request_v2\" extension\n in the server hello message. Additionally, note that a server MUST\n NOT send the \"CertificateStatus\" message unless it received either a\n \"status_request\" or \"status_request_v2\" extension in the client hello\n message and sent a corresponding \"status_request\" or\n \"status_request_v2\" extension in the server hello message.\n\n Clients requesting an OCSP response and receiving one or more OCSP\n responses in a \"CertificateStatus\" message MUST check the OCSP\n response(s) and abort the handshake if the response is a \"revoked\"\n status or other unacceptable responses (as determined by client\n policy) with a bad_certificate_status_response(113) alert. This\n alert is always fatal.\n\n If the OCSP response received from the server does not result in a\n definite \"good\" or \"revoked\" status, it is inconclusive. A TLS\n client in such a case MAY check the validity of the server\n certificate through other means, e.g., by directly querying the\n certificate issuer. If such processing still results in an\n inconclusive response, then the application using the TLS connection\n will have to decide whether to close the connection or not. Note\n that this problem cannot be decided by the generic TLS client code\n without information from the application. If the application doesn't\n provide any such information, then the client MUST abort the\n connection, since the server certificate has not been sufficiently\n validated.\n\n An example of where the application might wish to continue is with\n EAP-TLS (Extensible Authentication Protocol - TLS), where the\n application can use another mechanism to check the status of a\n certificate once it obtains network access. In this case, the\n application could have the client continue with the handshake, but it\n MUST NOT disclose a username and password until it has fully\n validated the server certificate.\n\n\n\n\n\n\nPettersen Standards Track [Page 7]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n3. IANA Considerations\n\n Section 2.1 defines the new TLS extension status_request_v2 (17)\n enum, which has been added to the \"ExtensionType Values\" list in the\n IANA \"Transport Layer Security (TLS) Extensions\" registry.\n\n Section 2.2 describes a TLS CertificateStatusType registry that is\n now maintained by IANA. The \"TLS Certificate Status Types\" registry\n has been created under the \"Transport Layer Security (TLS)\n Extensions\" registry. CertificateStatusType values are to be\n assigned via IETF Review as defined in [RFC5226]. The initial\n registry corresponds to the definition of \"CertificateStatusType\" in\n Section 2.2.\n\n Value Description Reference\n -----------------------------------------\n 0 Reserved [RFC6961]\n 1 ocsp [RFC6066] [RFC6961]\n 2 ocsp_multi [RFC6961]\n 3-255 Unassigned\n\n4. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security considerations for the particular extension\n specified in this document are given below. In general, implementers\n should continue to monitor the state of the art and address any\n weaknesses identified.\n\n4.1. Security Considerations for status_request_v2\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC6960] for further details.\n\n This extension allows the client to send arbitrary data to the\n server. The server implementers need to handle such data carefully\n to avoid introducing security vulnerabilities.\n\n The security considerations of [RFC6960] apply to OCSP requests and\n responses.\n\n\n\n\nPettersen Standards Track [Page 8]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n5. Acknowledgements\n\n This document is based on [RFC6066], authored by Donald Eastlake 3rd.\n\n6. References\n\n6.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246, August 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC6066] Eastlake, D., \"Transport Layer Security (TLS) Extensions:\n Extension Definitions\", RFC 6066, January 2011.\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, June 2013.\n\n [X.680] ITU-T Recommendation X.680 (2008) | ISO/IEC 8824-1:2008,\n \"Abstract Syntax Notation One (ASN.1): Specification of\n basic notation\", November 2008.\n\n [X.690] ITU-T Recommendation X.690 (2008) | ISO/IEC 8825-1:2008,\n \"ASN.1 encoding rules: Specification of Basic Encoding\n Rules (BER), Canonical Encoding Rules (CER) and\n Distinguished Encoding Rules (DER)\", November 2008.\n\n6.2. Informative References\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and C.\n Adams, \"X.509 Internet Public Key Infrastructure Online\n Certificate Status Protocol - OCSP\", RFC 2560, June 1999.\n\n\n\n\n\n\nPettersen Standards Track [Page 9]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nAuthor's Address\n\n Yngve N. Pettersen\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 10]\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 5705 RTFM, Inc.\nCategory: Standards Track March 2010\nISSN: 2070-1721\n\n\n Keying Material Exporters for Transport Layer Security (TLS)\n\nAbstract\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n to perform key establishment but then use some of the keying material\n for their own purposes. This document describes a general mechanism\n for allowing that.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/5705.\n\nCopyright Notice\n\n Copyright (c) 2010 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2\n 2. Conventions Used In This Document . . . . . . . . . . . . . . . 3\n 3. Binding to Application Contexts . . . . . . . . . . . . . . . . 3\n 4. Exporter Definition . . . . . . . . . . . . . . . . . . . . . . 4\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . . 5\n 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 6\n 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . 6\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 7\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 7\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 7\n\n1. Introduction\n\n Note: The mechanism described in this document was previously known\n as \"TLS Extractors\" but was changed to avoid a name conflict\n with the use of the term \"Extractor\" in the cryptographic\n community.\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n [RFC5246] or Datagram TLS (DTLS) [RFC4347] to perform key\n establishment but then use some of the keying material for their own\n purposes. A typical example is DTLS-SRTP [DTLS-SRTP], a key\n management scheme for the Secure Real-time Transport Protocol (SRTP)\n that uses DTLS to perform a key exchange and negotiate the SRTP\n [RFC3711] protection suite and then uses the DTLS master_secret to\n generate the SRTP keys.\n\n These applications imply a need to be able to export keying material\n (later called Exported Keying Material or EKM) from TLS/DTLS to an\n application or protocol residing at an upper layer, and to securely\n agree on the upper-layer context where the keying material will be\n used. The mechanism for exporting the keying material has the\n following requirements:\n\n o Both client and server need to be able to export the same EKM\n value.\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n o EKM values should be indistinguishable from random data to\n attackers who don't know the master_secret.\n\n o It should be possible to export multiple EKM values from the same\n TLS/DTLS association.\n\n o Knowing one EKM value should not reveal any useful information\n about the master_secret or about other EKM values.\n\n The mechanism described in this document is intended to fulfill these\n requirements. This mechanism is compatible with all versions of TLS.\n\n2. Conventions Used In This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Binding to Application Contexts\n\n In addition to using an exporter to obtain keying material, an\n application using the keying material has to securely establish the\n upper-layer context where the keying material will be used. The\n details of this context depend on the application, but it could\n include things such as algorithms and parameters that will be used\n with the keys, identifier(s) for the endpoint(s) who will use the\n keys, identifier(s) for the session(s) where the keys will be used,\n and the lifetime(s) for the context and/or keys. At a minimum, there\n should be some mechanism for signaling that an exporter will be used.\n\n This specification does not mandate a single mechanism for agreeing\n on such context; instead, there are several possibilities that can be\n used (and can complement each other). For example:\n\n o Information about the upper-layer context can be included in the\n optional data after the exporter label (see Section 4).\n\n o Information about the upper-layer context can be exchanged in TLS\n extensions included in the ClientHello and ServerHello messages.\n This approach is used in [DTLS-SRTP]. The handshake messages are\n protected by the Finished messages, so once the handshake\n completes, the peers will have the same view of the information.\n Extensions also allow a limited form of negotiation: for example,\n the TLS client could propose several alternatives for some context\n parameters, and the TLS server could select one of them.\n\n o The upper-layer protocol can include its own handshake, which can\n be protected using the keys exported by TLS.\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n No matter how the context is agreed, it is required that it has one\n part that indicates which application will use the exported keys.\n This part is the disambiguating label string (see Section 4).\n\n It is important to note that just embedding TLS messages in the\n upper-layer protocol may not automatically secure all the important\n context information, since the upper-layer messages are not covered\n by TLS Finished messages.\n\n4. Exporter Definition\n\n The output of the exporter is intended to be used in a single scope,\n which is associated with the TLS session, the label, and the context\n value.\n\n The exporter takes three input values:\n\n o a disambiguating label string,\n\n o a per-association context value provided by the application using\n the exporter, and\n\n o a length value.\n\n If no context is provided, it then computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random\n )[length]\n\n If context is provided, it computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random +\n context_value_length + context_value\n )[length]\n\n Where PRF is the TLS Pseudorandom Function in use for the session.\n The output is a pseudorandom bit string of length bytes generated\n from the master_secret. (This construction allows for\n interoperability with older exporter-type constructions which do not\n use context values, e.g., [RFC5281]).\n\n Labels here have the same definition as in TLS, i.e., an ASCII string\n with no terminating NULL. Label values beginning with \"EXPERIMENTAL\"\n MAY be used for private use without registration. All other label\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n values MUST be registered via Specification Required as described by\n RFC 5226 [RFC5226]. Note that exporter labels have the potential to\n collide with existing PRF labels. In order to prevent this, labels\n SHOULD begin with \"EXPORTER\". This is not a MUST because there are\n existing uses that have labels which do not begin with this prefix.\n\n The context value allows the application using the exporter to mix\n its own data with the TLS PRF for the exporter output. One example\n of where this might be useful is an authentication setting where the\n client credentials are valid for more than one identity; the context\n value could then be used to mix the expected identity into the keying\n material, thus preventing substitution attacks. The context value\n length is encoded as an unsigned, 16-bit quantity (uint16; see\n [RFC5246], Section 4.4) representing the length of the context value.\n The context MAY be zero length. Because the context value is mixed\n with the master_secret via the PRF, it is safe to mix confidential\n information into the exporter, provided that the master_secret will\n not be known to the attacker.\n\n5. Security Considerations\n\n The prime security requirement for exporter outputs is that they be\n independent. More formally, after a particular TLS session, if an\n adversary is allowed to choose multiple (label, context value) pairs\n and is given the output of the PRF for those values, the attacker is\n still unable to distinguish between the output of the PRF for a\n (label, context value) pair (different from the ones that it\n submitted) and a random value of the same length. In particular,\n there may be settings, such as the one described in Section 4, where\n the attacker can control the context value; such an attacker MUST NOT\n be able to predict the output of the exporter. Similarly, an\n attacker who does not know the master secret should not be able to\n distinguish valid exporter outputs from random values. The current\n set of TLS PRFs is believed to meet this objective, provided the\n master secret is randomly generated.\n\n Because an exporter produces the same value if applied twice with the\n same label to the same master_secret, it is critical that two EKM\n values generated with the same label not be used for two different\n purposes -- hence, the requirement for IANA registration. However,\n because exporters depend on the TLS PRF, it is not a threat to the\n use of an EKM value generated from one label to reveal an EKM value\n generated from another label.\n\n With certain TLS cipher suites, the TLS master secret is not\n necessarily unique to a single TLS session. In particular, with RSA\n key exchange, a malicious party acting as TLS server in one session\n and as TLS client in another session can cause those two sessions to\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n have the same TLS master secret (though the sessions must be\n established simultaneously to get adequate control of the Random\n values). Applications using the EKM need to consider this in how\n they use the EKM; in some cases, requiring the use of other cipher\n suites (such as those using a Diffie-Hellman key exchange) may be\n advisable.\n\n Designing a secure mechanism that uses exporters is not necessarily\n straightforward. This document only provides the exporter mechanism,\n but the problem of agreeing on the surrounding context and the\n meaning of the information passed to and from the exporter remains.\n Any new uses of the exporter mechanism should be subject to careful\n review.\n\n6. IANA Considerations\n\n IANA has created a TLS Exporter Label registry for this purpose. The\n initial contents of the registry are given below:\n\n Value Reference Note\n ----------------------------- --------- ----\n client finished [RFC5246] (1)\n server finished [RFC5246] (1)\n master secret [RFC5246] (1)\n key expansion [RFC5246] (1)\n client EAP encryption [RFC5216]\n ttls keying material [RFC5281]\n ttls challenge [RFC5281]\n\n Note: (1) These entries are reserved and MUST NOT be used for the\n purpose described in RFC 5705, in order to avoid confusion with\n similar, but distinct, use in RFC 5246.\n\n Future values are allocated via the RFC 5226 Specification Required\n policy. The label is a string consisting of printable ASCII\n characters. IANA MUST also verify that one label is not a prefix of\n any other label. For example, labels \"key\" or \"master secretary\" are\n forbidden.\n\n7. Acknowledgments\n\n Thanks to Pasi Eronen for valuable comments and for the contents of\n the IANA section and Section 3. Thanks to David McGrew for helpful\n discussion of the security considerations and to Vijay Gurbani and\n Alfred Hoenes for editorial comments.\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246,\n August 2008.\n\n8.2. Informative References\n\n [DTLS-SRTP] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for Secure\n Real-time Transport Protocol (SRTP)\", Work in Progress,\n February 2009.\n\n [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and\n K. Norrman, \"The Secure Real-time Transport Protocol\n (SRTP)\", RFC 3711, March 2004.\n\n [RFC4347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security\", RFC 4347, April 2006.\n\n [RFC5216] Simon, D., Aboba, B., and R. Hurst, \"The EAP-TLS\n Authentication Protocol\", RFC 5216, March 2008.\n\n [RFC5281] Funk, P. and S. Blake-Wilson, \"Extensible Authentication\n Protocol Tunneled Transport Layer Security Authenticated\n Protocol Version 0 (EAP-TTLSv0)\", RFC 5281, August 2008.\n\nAuthor's Address\n\n Eric Rescorla\n RTFM, Inc.\n 2064 Edgewood Drive\n Palo Alto, CA 94303\n USA\n\n EMail: [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) D. Eastlake 3rd\nRequest for Comments: 6066 Huawei\nObsoletes: 4366 January 2011\nCategory: Standards Track\nISSN: 2070-1721\n\n\n Transport Layer Security (TLS) Extensions: Extension Definitions\n\nAbstract\n\n This document provides specifications for existing TLS extensions.\n It is a companion document for RFC 5246, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\". The extensions specified are\n server_name, max_fragment_length, client_certificate_url,\n trusted_ca_keys, truncated_hmac, and status_request.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6066.\n\nCopyright Notice\n\n Copyright (c) 2011 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n\n\n\n\n\nEastlake Standards Track [Page 1]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction ....................................................3\n 1.1. Specific Extensions Covered ................................3\n 1.2. Conventions Used in This Document ..........................5\n 2. Extensions to the Handshake Protocol ............................5\n 3. Server Name Indication ..........................................6\n 4. Maximum Fragment Length Negotiation .............................8\n 5. Client Certificate URLs .........................................9\n 6. Trusted CA Indication ..........................................12\n 7. Truncated HMAC .................................................13\n 8. Certificate Status Request .....................................14\n 9. Error Alerts ...................................................16\n 10. IANA Considerations ...........................................17\n 10.1. pkipath MIME Type Registration ...........................17\n 10.2. Reference for TLS Alerts, TLS HandshakeTypes, and\n ExtensionTypes ...........................................19\n 11. Security Considerations .......................................19\n 11.1. Security Considerations for server_name ..................19\n 11.2. Security Considerations for max_fragment_length ..........20\n 11.3. Security Considerations for client_certificate_url .......20\n 11.4. Security Considerations for trusted_ca_keys ..............21\n 11.5. Security Considerations for truncated_hmac ...............21\n 11.6. Security Considerations for status_request ...............22\n 12. Normative References ..........................................22\n 13. Informative References ........................................23\n Appendix A. Changes from RFC 4366 .................................24\n Appendix B. Acknowledgements ......................................25\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 2]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1. Introduction\n\n The Transport Layer Security (TLS) Protocol Version 1.2 is specified\n in [RFC5246]. That specification includes the framework for\n extensions to TLS, considerations in designing such extensions (see\n Section 7.4.1.4 of [RFC5246]), and IANA Considerations for the\n allocation of new extension code points; however, it does not specify\n any particular extensions other than Signature Algorithms (see\n Section 7.4.1.4.1 of [RFC5246]).\n\n This document provides the specifications for existing TLS\n extensions. It is, for the most part, the adaptation and editing of\n material from RFC 4366, which covered TLS extensions for TLS 1.0 (RFC\n 2246) and TLS 1.1 (RFC 4346).\n\n1.1. Specific Extensions Covered\n\n The extensions described here focus on extending the functionality\n provided by the TLS protocol message formats. Other issues, such as\n the addition of new cipher suites, are deferred.\n\n The extension types defined in this document are:\n\n enum {\n server_name(0), max_fragment_length(1),\n client_certificate_url(2), trusted_ca_keys(3),\n truncated_hmac(4), status_request(5), (65535)\n } ExtensionType;\n\n Specifically, the extensions described in this document:\n\n - Allow TLS clients to provide to the TLS server the name of the\n server they are contacting. This functionality is desirable in\n order to facilitate secure connections to servers that host\n multiple 'virtual' servers at a single underlying network address.\n\n - Allow TLS clients and servers to negotiate the maximum fragment\n length to be sent. This functionality is desirable as a result of\n memory constraints among some clients, and bandwidth constraints\n among some access networks.\n\n - Allow TLS clients and servers to negotiate the use of client\n certificate URLs. This functionality is desirable in order to\n conserve memory on constrained clients.\n\n\n\n\n\n\n\nEastlake Standards Track [Page 3]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - Allow TLS clients to indicate to TLS servers which certification\n authority (CA) root keys they possess. This functionality is\n desirable in order to prevent multiple handshake failures\n involving TLS clients that are only able to store a small number\n of CA root keys due to memory limitations.\n\n - Allow TLS clients and servers to negotiate the use of truncated\n Message Authentication Codes (MACs). This functionality is\n desirable in order to conserve bandwidth in constrained access\n networks.\n\n - Allow TLS clients and servers to negotiate that the server sends\n the client certificate status information (e.g., an Online\n Certificate Status Protocol (OCSP) [RFC2560] response) during a\n TLS handshake. This functionality is desirable in order to avoid\n sending a Certificate Revocation List (CRL) over a constrained\n access network and therefore saving bandwidth.\n\n TLS clients and servers may use the extensions described in this\n document. The extensions are designed to be backwards compatible,\n meaning that TLS clients that support the extensions can talk to TLS\n servers that do not support the extensions, and vice versa.\n\n Note that any messages associated with these extensions that are sent\n during the TLS handshake MUST be included in the hash calculations\n involved in \"Finished\" messages.\n\n Note also that all the extensions defined in this document are\n relevant only when a session is initiated. A client that requests\n session resumption does not in general know whether the server will\n accept this request, and therefore it SHOULD send the same extensions\n as it would send if it were not attempting resumption. When a client\n includes one or more of the defined extension types in an extended\n client hello while requesting session resumption:\n\n - The server name indication extension MAY be used by the server\n when deciding whether or not to resume a session as described in\n Section 3.\n\n - If the resumption request is denied, the use of the extensions is\n negotiated as normal.\n\n - If, on the other hand, the older session is resumed, then the\n server MUST ignore the extensions and send a server hello\n containing none of the extension types. In this case, the\n functionality of these extensions negotiated during the original\n session initiation is applied to the resumed session.\n\n\n\n\nEastlake Standards Track [Page 4]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1.2. Conventions Used in This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n [RFC2119].\n\n2. Extensions to the Handshake Protocol\n\n This document specifies the use of two new handshake messages,\n \"CertificateURL\" and \"CertificateStatus\". These messages are\n described in Sections 5 and 8, respectively. The new handshake\n message structure therefore becomes:\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), certificate_url(21), certificate_status(22),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case certificate_url: CertificateURL;\n case certificate_status: CertificateStatus;\n } body;\n } Handshake;\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 5]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n3. Server Name Indication\n\n TLS does not provide a mechanism for a client to tell a server the\n name of the server it is contacting. It may be desirable for clients\n to provide this information to facilitate secure connections to\n servers that host multiple 'virtual' servers at a single underlying\n network address.\n\n In order to provide any of the server names, clients MAY include an\n extension of type \"server_name\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"ServerNameList\" where:\n\n struct {\n NameType name_type;\n select (name_type) {\n case host_name: HostName;\n } name;\n } ServerName;\n\n enum {\n host_name(0), (255)\n } NameType;\n\n opaque HostName<1..2^16-1>;\n\n struct {\n ServerName server_name_list<1..2^16-1>\n } ServerNameList;\n\n The ServerNameList MUST NOT contain more than one name of the same\n name_type. If the server understood the ClientHello extension but\n does not recognize the server name, the server SHOULD take one of two\n actions: either abort the handshake by sending a fatal-level\n unrecognized_name(112) alert or continue the handshake. It is NOT\n RECOMMENDED to send a warning-level unrecognized_name(112) alert,\n because the client's behavior in response to warning-level alerts is\n unpredictable. If there is a mismatch between the server name used\n by the client application and the server name of the credential\n chosen by the server, this mismatch will become apparent when the\n client application performs the server endpoint identification, at\n which point the client application will have to decide whether to\n proceed with the communication. TLS implementations are encouraged\n to make information available to application callers about warning-\n level alerts that were received or sent during a TLS handshake. Such\n information can be useful for diagnostic purposes.\n\n\n\n\n\nEastlake Standards Track [Page 6]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note: Earlier versions of this specification permitted multiple\n names of the same name_type. In practice, current client\n implementations only send one name, and the client cannot\n necessarily find out which name the server selected. Multiple\n names of the same name_type are therefore now prohibited.\n\n Currently, the only server names supported are DNS hostnames;\n however, this does not imply any dependency of TLS on DNS, and other\n name types may be added in the future (by an RFC that updates this\n document). The data structure associated with the host_name NameType\n is a variable-length vector that begins with a 16-bit length. For\n backward compatibility, all future data structures associated with\n new NameTypes MUST begin with a 16-bit length field. TLS MAY treat\n provided server names as opaque data and pass the names and types to\n the application.\n\n \"HostName\" contains the fully qualified DNS hostname of the server,\n as understood by the client. The hostname is represented as a byte\n string using ASCII encoding without a trailing dot. This allows the\n support of internationalized domain names through the use of A-labels\n defined in [RFC5890]. DNS hostnames are case-insensitive. The\n algorithm to compare hostnames is described in [RFC5890], Section\n 2.3.2.4.\n\n Literal IPv4 and IPv6 addresses are not permitted in \"HostName\".\n\n It is RECOMMENDED that clients include an extension of type\n \"server_name\" in the client hello whenever they locate a server by a\n supported name type.\n\n A server that receives a client hello containing the \"server_name\"\n extension MAY use the information contained in the extension to guide\n its selection of an appropriate certificate to return to the client,\n and/or other aspects of security policy. In this event, the server\n SHALL include an extension of type \"server_name\" in the (extended)\n server hello. The \"extension_data\" field of this extension SHALL be\n empty.\n\n When the server is deciding whether or not to accept a request to\n resume a session, the contents of a server_name extension MAY be used\n in the lookup of the session in the session cache. The client SHOULD\n include the same server_name extension in the session resumption\n request as it did in the full handshake that established the session.\n A server that implements this extension MUST NOT accept the request\n to resume the session if the server_name extension contains a\n different name. Instead, it proceeds with a full handshake to\n establish a new session. When resuming a session, the server MUST\n NOT include a server_name extension in the server hello.\n\n\n\nEastlake Standards Track [Page 7]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n If an application negotiates a server name using an application\n protocol and then upgrades to TLS, and if a server_name extension is\n sent, then the extension SHOULD contain the same name that was\n negotiated in the application protocol. If the server_name is\n established in the TLS session handshake, the client SHOULD NOT\n attempt to request a different server name at the application layer.\n\n4. Maximum Fragment Length Negotiation\n\n Without this extension, TLS specifies a fixed maximum plaintext\n fragment length of 2^14 bytes. It may be desirable for constrained\n clients to negotiate a smaller maximum fragment length due to memory\n limitations or bandwidth limitations.\n\n In order to negotiate smaller maximum fragment lengths, clients MAY\n include an extension of type \"max_fragment_length\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain:\n\n enum{\n 2^9(1), 2^10(2), 2^11(3), 2^12(4), (255)\n } MaxFragmentLength;\n\n whose value is the desired maximum fragment length. The allowed\n values for this field are: 2^9, 2^10, 2^11, and 2^12.\n\n Servers that receive an extended client hello containing a\n \"max_fragment_length\" extension MAY accept the requested maximum\n fragment length by including an extension of type\n \"max_fragment_length\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL contain a\n \"MaxFragmentLength\" whose value is the same as the requested maximum\n fragment length.\n\n If a server receives a maximum fragment length negotiation request\n for a value other than the allowed values, it MUST abort the\n handshake with an \"illegal_parameter\" alert. Similarly, if a client\n receives a maximum fragment length negotiation response that differs\n from the length it requested, it MUST also abort the handshake with\n an \"illegal_parameter\" alert.\n\n Once a maximum fragment length other than 2^14 has been successfully\n negotiated, the client and server MUST immediately begin fragmenting\n messages (including handshake messages) to ensure that no fragment\n larger than the negotiated length is sent. Note that TLS already\n requires clients and servers to support fragmentation of handshake\n messages.\n\n\n\n\nEastlake Standards Track [Page 8]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n The negotiated length applies for the duration of the session\n including session resumptions.\n\n The negotiated length limits the input that the record layer may\n process without fragmentation (that is, the maximum value of\n TLSPlaintext.length; see [RFC5246], Section 6.2.1). Note that the\n output of the record layer may be larger. For example, if the\n negotiated length is 2^9=512, then, when using currently defined\n cipher suites (those defined in [RFC5246] and [RFC2712]) and null\n compression, the record-layer output can be at most 805 bytes: 5\n bytes of headers, 512 bytes of application data, 256 bytes of\n padding, and 32 bytes of MAC. This means that in this event a TLS\n record-layer peer receiving a TLS record-layer message larger than\n 805 bytes MUST discard the message and send a \"record_overflow\"\n alert, without decrypting the message. When this extension is used\n with Datagram Transport Layer Security (DTLS), implementations SHOULD\n NOT generate record_overflow alerts unless the packet passes message\n authentication.\n\n5. Client Certificate URLs\n\n Without this extension, TLS specifies that when client authentication\n is performed, client certificates are sent by clients to servers\n during the TLS handshake. It may be desirable for constrained\n clients to send certificate URLs in place of certificates, so that\n they do not need to store their certificates and can therefore save\n memory.\n\n In order to negotiate sending certificate URLs to a server, clients\n MAY include an extension of type \"client_certificate_url\" in the\n (extended) client hello. The \"extension_data\" field of this\n extension SHALL be empty.\n\n (Note that it is necessary to negotiate the use of client certificate\n URLs in order to avoid \"breaking\" existing TLS servers.)\n\n Servers that receive an extended client hello containing a\n \"client_certificate_url\" extension MAY indicate that they are willing\n to accept certificate URLs by including an extension of type\n \"client_certificate_url\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n After negotiation of the use of client certificate URLs has been\n successfully completed (by exchanging hellos including\n \"client_certificate_url\" extensions), clients MAY send a\n \"CertificateURL\" message in place of a \"Certificate\" message as\n follows (see also Section 2):\n\n\n\n\nEastlake Standards Track [Page 9]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n enum {\n individual_certs(0), pkipath(1), (255)\n } CertChainType;\n\n struct {\n CertChainType type;\n URLAndHash url_and_hash_list<1..2^16-1>;\n } CertificateURL;\n\n struct {\n opaque url<1..2^16-1>;\n unint8 padding;\n opaque SHA1Hash[20];\n } URLAndHash;\n\n Here, \"url_and_hash_list\" contains a sequence of URLs and hashes.\n Each \"url\" MUST be an absolute URI reference according to [RFC3986]\n that can be immediately used to fetch the certificate(s).\n\n When X.509 certificates are used, there are two possibilities:\n\n - If CertificateURL.type is \"individual_certs\", each URL refers to a\n single DER-encoded X.509v3 certificate, with the URL for the\n client's certificate first.\n\n - If CertificateURL.type is \"pkipath\", the list contains a single\n URL referring to a DER-encoded certificate chain, using the type\n PkiPath described in Section 10.1.\n\n When any other certificate format is used, the specification that\n describes use of that format in TLS should define the encoding format\n of certificates or certificate chains, and any constraint on their\n ordering.\n\n The \"padding\" byte MUST be 0x01. It is present to make the structure\n backwards compatible.\n\n The hash corresponding to each URL is the SHA-1 hash of the\n certificate or certificate chain (in the case of X.509 certificates,\n the DER-encoded certificate or the DER-encoded PkiPath).\n\n Note that when a list of URLs for X.509 certificates is used, the\n ordering of URLs is the same as that used in the TLS Certificate\n message (see [RFC5246], Section 7.4.2), but opposite to the order in\n which certificates are encoded in PkiPath. In either case, the self-\n signed root certificate MAY be omitted from the chain, under the\n assumption that the server must already possess it in order to\n validate it.\n\n\n\nEastlake Standards Track [Page 10]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Servers receiving \"CertificateURL\" SHALL attempt to retrieve the\n client's certificate chain from the URLs and then process the\n certificate chain as usual. A cached copy of the content of any URL\n in the chain MAY be used, provided that the SHA-1 hash matches the\n hash of the cached copy.\n\n Servers that support this extension MUST support the 'http' URI\n scheme for certificate URLs and MAY support other schemes. Use of\n other schemes than 'http', 'https', or 'ftp' may create unexpected\n problems.\n\n If the protocol used is HTTP, then the HTTP server can be configured\n to use the Cache-Control and Expires directives described in\n [RFC2616] to specify whether and for how long certificates or\n certificate chains should be cached.\n\n The TLS server MUST NOT follow HTTP redirects when retrieving the\n certificates or certificate chain. The URLs used in this extension\n MUST NOT be chosen to depend on such redirects.\n\n If the protocol used to retrieve certificates or certificate chains\n returns a MIME-formatted response (as HTTP does), then the following\n MIME Content-Types SHALL be used: when a single X.509v3 certificate\n is returned, the Content-Type is \"application/pkix-cert\" [RFC2585],\n and when a chain of X.509v3 certificates is returned, the Content-\n Type is \"application/pkix-pkipath\" (Section 10.1).\n\n The server MUST check that the SHA-1 hash of the contents of the\n object retrieved from that URL (after decoding any MIME Content-\n Transfer-Encoding) matches the given hash. If any retrieved object\n does not have the correct SHA-1 hash, the server MUST abort the\n handshake with a bad_certificate_hash_value(114) alert. This alert\n is always fatal.\n\n Clients may choose to send either \"Certificate\" or \"CertificateURL\"\n after successfully negotiating the option to send certificate URLs.\n The option to send a certificate is included to provide flexibility\n to clients possessing multiple certificates.\n\n If a server is unable to obtain certificates in a given\n CertificateURL, it MUST send a fatal certificate_unobtainable(111)\n alert if it requires the certificates to complete the handshake. If\n the server does not require the certificates, then the server\n continues the handshake. The server MAY send a warning-level alert\n in this case. Clients receiving such an alert SHOULD log the alert\n and continue with the handshake if possible.\n\n\n\n\n\nEastlake Standards Track [Page 11]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n6. Trusted CA Indication\n\n Constrained clients that, due to memory limitations, possess only a\n small number of CA root keys may wish to indicate to servers which\n root keys they possess, in order to avoid repeated handshake\n failures.\n\n In order to indicate which CA root keys they possess, clients MAY\n include an extension of type \"trusted_ca_keys\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain \"TrustedAuthorities\" where:\n\n struct {\n TrustedAuthority trusted_authorities_list<0..2^16-1>;\n } TrustedAuthorities;\n\n struct {\n IdentifierType identifier_type;\n select (identifier_type) {\n case pre_agreed: struct {};\n case key_sha1_hash: SHA1Hash;\n case x509_name: DistinguishedName;\n case cert_sha1_hash: SHA1Hash;\n } identifier;\n } TrustedAuthority;\n\n enum {\n pre_agreed(0), key_sha1_hash(1), x509_name(2),\n cert_sha1_hash(3), (255)\n } IdentifierType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n Here, \"TrustedAuthorities\" provides a list of CA root key identifiers\n that the client possesses. Each CA root key is identified via\n either:\n\n - \"pre_agreed\": no CA root key identity supplied.\n\n - \"key_sha1_hash\": contains the SHA-1 hash of the CA root key. For\n Digital Signature Algorithm (DSA) and Elliptic Curve Digital\n Signature Algorithm (ECDSA) keys, this is the hash of the\n \"subjectPublicKey\" value. For RSA keys, the hash is of the big-\n endian byte string representation of the modulus without any\n initial zero-valued bytes. (This copies the key hash formats\n deployed in other environments.)\n\n\n\n\n\nEastlake Standards Track [Page 12]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - \"x509_name\": contains the DER-encoded X.509 DistinguishedName of\n the CA.\n\n - \"cert_sha1_hash\": contains the SHA-1 hash of a DER-encoded\n Certificate containing the CA root key.\n\n Note that clients may include none, some, or all of the CA root keys\n they possess in this extension.\n\n Note also that it is possible that a key hash or a Distinguished Name\n alone may not uniquely identify a certificate issuer (for example, if\n a particular CA has multiple key pairs). However, here we assume\n this is the case following the use of Distinguished Names to identify\n certificate issuers in TLS.\n\n The option to include no CA root keys is included to allow the client\n to indicate possession of some pre-defined set of CA root keys.\n\n Servers that receive a client hello containing the \"trusted_ca_keys\"\n extension MAY use the information contained in the extension to guide\n their selection of an appropriate certificate chain to return to the\n client. In this event, the server SHALL include an extension of type\n \"trusted_ca_keys\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n7. Truncated HMAC\n\n Currently defined TLS cipher suites use the MAC construction HMAC\n [RFC2104] to authenticate record-layer communications. In TLS, the\n entire output of the hash function is used as the MAC tag. However,\n it may be desirable in constrained environments to save bandwidth by\n truncating the output of the hash function to 80 bits when forming\n MAC tags.\n\n In order to negotiate the use of 80-bit truncated HMAC, clients MAY\n include an extension of type \"truncated_hmac\" in the extended client\n hello. The \"extension_data\" field of this extension SHALL be empty.\n\n Servers that receive an extended hello containing a \"truncated_hmac\"\n extension MAY agree to use a truncated HMAC by including an extension\n of type \"truncated_hmac\", with empty \"extension_data\", in the\n extended server hello.\n\n Note that if new cipher suites are added that do not use HMAC, and\n the session negotiates one of these cipher suites, this extension\n will have no effect. It is strongly recommended that any new cipher\n suites using other MACs consider the MAC size an integral part of the\n\n\n\n\nEastlake Standards Track [Page 13]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n cipher suite definition, taking into account both security and\n bandwidth considerations.\n\n If HMAC truncation has been successfully negotiated during a TLS\n handshake, and the negotiated cipher suite uses HMAC, both the client\n and the server pass this fact to the TLS record layer along with the\n other negotiated security parameters. Subsequently during the\n session, clients and servers MUST use truncated HMACs, calculated as\n specified in [RFC2104]. That is, SecurityParameters.mac_length is 10\n bytes, and only the first 10 bytes of the HMAC output are transmitted\n and checked. Note that this extension does not affect the\n calculation of the pseudo-random function (PRF) as part of\n handshaking or key derivation.\n\n The negotiated HMAC truncation size applies for the duration of the\n session including session resumptions.\n\n8. Certificate Status Request\n\n Constrained clients may wish to use a certificate-status protocol\n such as OCSP [RFC2560] to check the validity of server certificates,\n in order to avoid transmission of CRLs and therefore save bandwidth\n on constrained networks. This extension allows for such information\n to be sent in the TLS handshake, saving roundtrips and resources.\n\n In order to indicate their desire to receive certificate status\n information, clients MAY include an extension of type\n \"status_request\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"CertificateStatusRequest\" where:\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n } request;\n } CertificateStatusRequest;\n\n enum { ocsp(1), (255) } CertificateStatusType;\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n\n\n\nEastlake Standards Track [Page 14]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n In the OCSPStatusRequest, the \"ResponderIDs\" provides a list of OCSP\n responders that the client trusts. A zero-length \"responder_id_list\"\n sequence has the special meaning that the responders are implicitly\n known to the server, e.g., by prior arrangement. \"Extensions\" is a\n DER encoding of OCSP request extensions.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC2560]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a zero-length ASN.1 SEQUENCE, which is not\n valid for the \"Extensions\" type).\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement).\n\n Servers that receive a client hello containing the \"status_request\"\n extension MAY return a suitable certificate status response to the\n client along with their certificate. If OCSP is requested, they\n SHOULD use the information contained in the extension when selecting\n an OCSP responder and SHOULD include request_extensions in the OCSP\n request.\n\n Servers return a certificate response along with their certificate by\n sending a \"CertificateStatus\" message immediately after the\n \"Certificate\" message (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message, then the server MUST have included an\n extension of type \"status_request\" with empty \"extension_data\" in the\n extended server hello. The \"CertificateStatus\" message is conveyed\n using the handshake message type \"certificate_status\" as follows (see\n also Section 2):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<1..2^24-1>;\n\n An \"ocsp_response\" contains a complete, DER-encoded OCSP response\n (using the ASN.1 type OCSPResponse defined in [RFC2560]). Only one\n OCSP response may be sent.\n\n\n\n\nEastlake Standards Track [Page 15]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if has received a \"status_request\" extension in the\n client hello message and has sent a \"status_request\" extension in the\n server hello message.\n\n Note in addition that a server MUST NOT send the \"CertificateStatus\"\n message unless it received a \"status_request\" extension in the client\n hello message and sent a \"status_request\" extension in the server\n hello message.\n\n Clients requesting an OCSP response and receiving an OCSP response in\n a \"CertificateStatus\" message MUST check the OCSP response and abort\n the handshake if the response is not satisfactory with\n bad_certificate_status_response(113) alert. This alert is always\n fatal.\n\n9. Error Alerts\n\n Four new error alerts are defined for use with the TLS extensions\n defined in this document. To avoid \"breaking\" existing clients and\n servers, these alerts MUST NOT be sent unless the sending party has\n received an extended hello message from the party they are\n communicating with. These error alerts are conveyed using the\n following syntax. The new alerts are the last four, as indicated by\n the comments on the same line as the error alert number.\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n /* 41 is not defined, for historical reasons */\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction(60),\n protocol_version(70),\n insufficient_security(71),\n\n\n\nEastlake Standards Track [Page 16]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n certificate_unobtainable(111), /* new */\n unrecognized_name(112), /* new */\n bad_certificate_status_response(113), /* new */\n bad_certificate_hash_value(114), /* new */\n (255)\n } AlertDescription;\n\n \"certificate_unobtainable\" is described in Section 5.\n \"unrecognized_name\" is described in Section 3.\n \"bad_certificate_status_response\" is described in Section 8.\n \"bad_certificate_hash_value\" is described in Section 5.\n\n10. IANA Considerations\n\n IANA Considerations for TLS extensions and the creation of a registry\n are covered in Section 12 of [RFC5246] except for the registration of\n MIME type application/pkix-pkipath, which appears below.\n\n The IANA TLS extensions and MIME type application/pkix-pkipath\n registry entries that reference RFC 4366 have been updated to\n reference this document.\n\n10.1. pkipath MIME Type Registration\n\n MIME media type name: application\n MIME subtype name: pkix-pkipath\n Required parameters: none\n\n Optional parameters: version (default value is \"1\")\n\n Encoding considerations:\n Binary; this MIME type is a DER encoding of the ASN.1 type\n PkiPath, defined as follows:\n PkiPath ::= SEQUENCE OF Certificate\n PkiPath is used to represent a certification path. Within the\n sequence, the order of certificates is such that the subject of\n the first certificate is the issuer of the second certificate,\n etc.\n This is identical to the definition published in [X509-4th-TC1];\n note that it is different from that in [X509-4th].\n\n All Certificates MUST conform to [RFC5280]. (This should be\n interpreted as a requirement to encode only PKIX-conformant\n certificates using this type. It does not necessarily require\n\n\n\nEastlake Standards Track [Page 17]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n that all certificates that are not strictly PKIX-conformant must\n be rejected by relying parties, although the security consequences\n of accepting any such certificates should be considered\n carefully.)\n\n DER (as opposed to BER) encoding MUST be used. If this type is\n sent over a 7-bit transport, base64 encoding SHOULD be used.\n\n Security considerations:\n The security considerations of [X509-4th] and [RFC5280] (or any\n updates to them) apply, as well as those of any protocol that uses\n this type (e.g., TLS).\n\n Note that this type only specifies a certificate chain that can be\n assessed for validity according to the relying party's existing\n configuration of trusted CAs; it is not intended to be used to\n specify any change to that configuration.\n\n Interoperability considerations:\n No specific interoperability problems are known with this type,\n but for recommendations relating to X.509 certificates in general,\n see [RFC5280].\n\n Published specification: This document and [RFC5280].\n\n Applications that use this media type:\n TLS. It may also be used by other protocols or for general\n interchange of PKIX certificate chains.\n\n Additional information:\n Magic number(s): DER-encoded ASN.1 can be easily recognized.\n Further parsing is required to distinguish it from other ASN.1\n types.\n File extension(s): .pkipath\n Macintosh File Type Code(s): not specified\n\n Person & email address to contact for further information:\n Magnus Nystrom <[redacted-email]>\n\n Intended usage: COMMON\n\n Change controller: IESG <[redacted-email]>\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 18]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n10.2. Reference for TLS Alerts, TLS HandshakeTypes, and ExtensionTypes\n\n The following values in the TLS Alert Registry have been updated to\n reference this document:\n\n 111 certificate_unobtainable\n 112 unrecognized_name\n 113 bad_certificate_status_response\n 114 bad_certificate_hash_value\n\n The following values in the TLS HandshakeType Registry have been\n updated to reference this document:\n\n 21 certificate_url\n 22 certificate_status\n\n The following ExtensionType values have been updated to reference\n this document:\n\n 0 server_name\n 1 max_fragment_length\n 2 client_certificate_url\n 3 trusted_ca_keys\n 4 truncated_hmac\n 5 status_request\n\n11. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security Considerations for particular extensions\n specified in this document are given below.\n\n In general, implementers should continue to monitor the state of the\n art and address any weaknesses identified.\n\n11.1. Security Considerations for server_name\n\n If a single server hosts several domains, then clearly it is\n necessary for the owners of each domain to ensure that this satisfies\n their security needs. Apart from this, server_name does not appear\n to introduce significant security issues.\n\n Since it is possible for a client to present a different server_name\n in the application protocol, application server implementations that\n rely upon these names being the same MUST check to make sure the\n client did not present a different name in the application protocol.\n\n\n\n\n\nEastlake Standards Track [Page 19]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Implementations MUST ensure that a buffer overflow does not occur,\n whatever the values of the length fields in server_name.\n\n11.2. Security Considerations for max_fragment_length\n\n The maximum fragment length takes effect immediately, including for\n handshake messages. However, that does not introduce any security\n complications that are not already present in TLS, since TLS requires\n implementations to be able to handle fragmented handshake messages.\n\n Note that, as described in Section 4, once a non-null cipher suite\n has been activated, the effective maximum fragment length depends on\n the cipher suite and compression method, as well as on the negotiated\n max_fragment_length. This must be taken into account when sizing\n buffers and checking for buffer overflow.\n\n11.3. Security Considerations for client_certificate_url\n\n Support for client_certificate_url involves the server's acting as a\n client in another URI-scheme-dependent protocol. The server\n therefore becomes subject to many of the same security concerns that\n clients of the URI scheme are subject to, with the added concern that\n the client can attempt to prompt the server to connect to some\n (possibly weird-looking) URL.\n\n In general, this issue means that an attacker might use the server to\n indirectly attack another host that is vulnerable to some security\n flaw. It also introduces the possibility of denial-of-service\n attacks in which an attacker makes many connections to the server,\n each of which results in the server's attempting a connection to the\n target of the attack.\n\n Note that the server may be behind a firewall or otherwise able to\n access hosts that would not be directly accessible from the public\n Internet. This could exacerbate the potential security and denial-\n of-service problems described above, as well as allow the existence\n of internal hosts to be confirmed when they would otherwise be\n hidden.\n\n The detailed security concerns involved will depend on the URI\n schemes supported by the server. In the case of HTTP, the concerns\n are similar to those that apply to a publicly accessible HTTP proxy\n server. In the case of HTTPS, loops and deadlocks may be created,\n and this should be addressed. In the case of FTP, attacks arise that\n are similar to FTP bounce attacks.\n\n\n\n\n\n\nEastlake Standards Track [Page 20]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n As a result of this issue, it is RECOMMENDED that the\n client_certificate_url extension should have to be specifically\n enabled by a server administrator, rather than be enabled by default.\n It is also RECOMMENDED that URI schemes be enabled by the\n administrator individually, and only a minimal set of schemes be\n enabled. Unusual protocols that offer limited security or whose\n security is not well understood SHOULD be avoided.\n\n As discussed in [RFC3986], URLs that specify ports other than the\n default may cause problems, as may very long URLs (which are more\n likely to be useful in exploiting buffer overflow bugs).\n\n This extension continues to use SHA-1 (as in RFC 4366) and does not\n provide algorithm agility. The property required of SHA-1 in this\n case is second pre-image resistance, not collision resistance.\n Furthermore, even if second pre-image attacks against SHA-1 are found\n in the future, an attack against client_certificate_url would require\n a second pre-image that is accepted as a valid certificate by the\n server and contains the same public key.\n\n Also note that HTTP caching proxies are common on the Internet, and\n some proxies do not check for the latest version of an object\n correctly. If a request using HTTP (or another caching protocol)\n goes through a misconfigured or otherwise broken proxy, the proxy may\n return an out-of-date response.\n\n11.4. Security Considerations for trusted_ca_keys\n\n Potentially, the CA root keys a client possesses could be regarded as\n confidential information. As a result, the CA root key indication\n extension should be used with care.\n\n The use of the SHA-1 certificate hash alternative ensures that each\n certificate is specified unambiguously. This context does not\n require a cryptographic hash function, so the use of SHA-1 is\n considered acceptable, and no algorithm agility is provided.\n\n11.5. Security Considerations for truncated_hmac\n\n It is possible that truncated MACs are weaker than \"un-truncated\"\n MACs. However, no significant weaknesses are currently known or\n expected to exist for HMAC with MD5 or SHA-1, truncated to 80 bits.\n\n Note that the output length of a MAC need not be as long as the\n length of a symmetric cipher key, since forging of MAC values cannot\n be done off-line: in TLS, a single failed MAC guess will cause the\n immediate termination of the TLS session.\n\n\n\n\nEastlake Standards Track [Page 21]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Since the MAC algorithm only takes effect after all handshake\n messages that affect extension parameters have been authenticated by\n the hashes in the Finished messages, it is not possible for an active\n attacker to force negotiation of the truncated HMAC extension where\n it would not otherwise be used (to the extent that the handshake\n authentication is secure). Therefore, in the event that any security\n problems were found with truncated HMAC in the future, if either the\n client or the server for a given session were updated to take the\n problem into account, it would be able to veto use of this extension.\n\n11.6. Security Considerations for status_request\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC2560] for further details.\n\n12. Normative References\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC:\n Keyed-Hashing for Message Authentication\", RFC 2104,\n February 1997.\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and\n C. Adams, \"X.509 Internet Public Key Infrastructure\n Online Certificate Status Protocol - OCSP\", RFC 2560,\n June 1999.\n\n [RFC2585] Housley, R. and P. Hoffman, \"Internet X.509 Public Key\n Infrastructure Operational Protocols: FTP and HTTP\",\n RFC 2585, May 1999.\n\n [RFC2616] Fielding, R., Gettys, J., Mogul, J., Frystyk, H.,\n Masinter, L., Leach, P., and T. Berners-Lee,\n \"Hypertext Transfer Protocol -- HTTP/1.1\", RFC 2616,\n June 1999.\n\n [RFC3986] Berners-Lee, T., Fielding, R., and L. Masinter,\n \"Uniform Resource Identifier (URI): Generic Syntax\",\n STD 66, RFC 3986, January 2005.\n\n\n\nEastlake Standards Track [Page 22]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246, August\n 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation\n List (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC5890] Klensin, J., \"Internationalized Domain Names for\n Applications (IDNA): Definitions and Document\n Framework\", RFC 5890, August 2010.\n\n13. Informative References\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [X509-4th] ITU-T Recommendation X.509 (2000) | ISO/IEC\n 9594-8:2001, \"Information Systems - Open Systems\n Interconnection - The Directory: Public key and\n attribute certificate frameworks\".\n\n [X509-4th-TC1] ITU-T Recommendation X.509(2000) Corrigendum 1(2001) |\n ISO/IEC 9594-8:2001/Cor.1:2002, Technical Corrigendum\n 1 to ISO/IEC 9594:8:2001.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 23]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix A. Changes from RFC 4366\n\n The significant changes between RFC 4366 and this document are\n described below.\n\n RFC 4366 described both general extension mechanisms (for the TLS\n handshake and client and server hellos) as well as specific\n extensions. RFC 4366 was associated with RFC 4346, TLS 1.1. The\n client and server hello extension mechanisms have been moved into RFC\n 5246, TLS 1.2, so this document, which is associated with RFC 5246,\n includes only the handshake extension mechanisms and the specific\n extensions from RFC 4366. RFC 5246 also specifies the unknown\n extension error and new extension specification considerations, so\n that material has been removed from this document.\n\n The Server Name extension now specifies only ASCII representation,\n eliminating UTF-8. It is provided that the ServerNameList can\n contain more than only one name of any particular name_type. If a\n server name is provided but not recognized, the server should either\n continue the handshake without an error or send a fatal error.\n Sending a warning-level message is not recommended because client\n behavior will be unpredictable. Provision was added for the user\n using the server_name extension in deciding whether or not to resume\n a session. Furthermore, this extension should be the same in a\n session resumption request as it was in the full handshake that\n established the session. Such a resumption request must not be\n accepted if the server_name extension is different, but instead a\n full handshake must be done to possibly establish a new session.\n\n The Client Certificate URLs extension has been changed to make the\n presence of a hash mandatory.\n\n For the case of DTLS, the requirement to report an overflow of the\n negotiated maximum fragment length is made conditional on passing\n authentication.\n\n TLS servers are now prohibited from following HTTP redirects when\n retrieving certificates.\n\n The material was also re-organized in minor ways. For example,\n information as to which errors are fatal is moved from the \"Error\n Alerts\" section to the individual extension specifications.\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 24]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix B. Acknowledgements\n\n This document is based on material from RFC 4366 for which the\n authors were S. Blake-Wilson, M. Nystrom, D. Hopwood, J. Mikkelsen,\n and T. Wright. Other contributors include Joseph Salowey, Alexey\n Melnikov, Peter Saint-Andre, and Adrian Farrel.\n\nAuthor's Address\n\n Donald Eastlake 3rd\n Huawei\n 155 Beaver Street\n Milford, MA 01757 USA\n\n Phone: +1-508-333-2270\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 25]", "domain": "standards", "dossier_id": "a00388a585cd2f4d89b3", "essential_artifact_ids": ["ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"], "event_count": 6, "evidence_distance": 65141, "evidence_span_tokens": 65134, "executable_proof_id": "0c64de0f18afd43b8e63", "generation_integration": "task_replay_promotion_v1", "graph": {"hop_count": 2, "n_essential_artifacts": 7, "n_essential_events": 6, "proof_depth": 2}, "holdout": {"group_id": "af2707ad6cb50ab3", "strategy": "world"}, "hop_count": 2, "hybrid_causal_edges": [], "ietf_requirement_task": {"answer": {"current_protocol": "TLS_1_3_RFC8446", "obsoletes_ocsp_stapling": "RFC6961", "obsoletes_session_tickets": "RFC5077", "obsoletes_tls12": "RFC5246", "updates_exporters": "RFC5705", "updates_extensions": "RFC6066"}, "answer_program_id": "ietf.tls13_handshake_succession.v1", "cutoff": "2026-01-31T00:00:00Z", "essential_evidence_ids": ["current_protocol", "obsoletes_session_tickets", "obsoletes_tls12", "obsoletes_ocsp_stapling", "updates_exporters", "updates_extensions"], "essential_relation_ids": ["ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "ietf:updates:ietf:rfc:8446:ietf:rfc:6066"], "evidence_items": [{"char_end": 447, "char_start": 361, "evidence_id": "current_protocol", "evidence_quote": "This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol.", "quote_sha256": "213f297f8fd7940d76115d7115fd430d489bced81930ca9d9319c2eaddff65ea", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15703, "char_start": 15575, "evidence_id": "obsoletes_session_tickets", "evidence_quote": "It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2.", "quote_sha256": "957220687c3a075f9479d4799d2a2a6bc3065f4b5f968376a62cd724206781ff", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15573, "char_start": 15473, "evidence_id": "obsoletes_tls12", "evidence_quote": "This document supersedes and obsoletes previous versions of TLS,\n including version 1.2 [RFC5246].", "quote_sha256": "0ffd331074a612c946a7f65eb52848b1224f4497f7dad642bd1c34ac255fe340", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15989, "char_start": 15934, "evidence_id": "obsoletes_ocsp_stapling", "evidence_quote": "obsoletes [RFC6961] as\n described in Section 4.4.2.1.", "quote_sha256": "836a51c84442de8cd356808fe29a5bc51913ffc3248fa087461cea3fbe9f26cd", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15807, "char_start": 15705, "evidence_id": "updates_exporters", "evidence_quote": "Because TLS 1.3 changes the way keys are\n derived, it updates [RFC5705] as described in Section 7.5.", "quote_sha256": "e5c566bced7efdd9df1c7483ab71fa5ce896e674e1a35613667bd90d52341b0d", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15989, "char_start": 15902, "evidence_id": "updates_extensions", "evidence_quote": "therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in Section 4.4.2.1.", "quote_sha256": "46d4d3c9d3c46874270771eceb979e3b19a7f09528933d4e9071941d096f5b9f", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}], "query_type": "protocol_succession_resolution", "question": "Resolve the effective TLS 1.3 handshake protocol succession at 2026-01-31T00:00:00Z for this scenario (canonical JSON): {\"protocol\":\"tls\",\"publication\":\"rfc8446\",\"succession\":\"obsoletes_and_updates\",\"version\":\"1.3\"}. Return exactly one JSON object with these keys in this order: [\"current_protocol\",\"obsoletes_session_tickets\",\"obsoletes_tls12\",\"obsoletes_ocsp_stapling\",\"updates_exporters\",\"updates_extensions\"]. Use this exact per-field output codebook (canonical JSON): {\"current_protocol\":{\"code\":\"TLS_1_3_RFC8446\",\"meaning\":\"RFC 8446 is TLS protocol version 1.3\"},\"obsoletes_ocsp_stapling\":{\"code\":\"RFC6961\",\"meaning\":\"RFC 8446 obsoletes multiple certificate status requests\"},\"obsoletes_session_tickets\":{\"code\":\"RFC5077\",\"meaning\":\"RFC 8446 obsoletes TLS session resumption tickets\"},\"obsoletes_tls12\":{\"code\":\"RFC5246\",\"meaning\":\"RFC 8446 obsoletes TLS protocol version 1.2\"},\"updates_exporters\":{\"code\":\"RFC5705\",\"meaning\":\"RFC 8446 updates TLS keying material exporters\"},\"updates_extensions\":{\"code\":\"RFC6066\",\"meaning\":\"RFC 8446 updates TLS extension definitions\"}}. Resolve each field independently from the supplied RFC graph; use the string \"UNKNOWN\" only for a field whose required evidence or relation is absent.", "scenario": {"protocol": "tls", "publication": "rfc8446", "succession": "obsoletes_and_updates", "version": "1.3"}, "schema_version": "longworld.ietf-tls13-handshake-succession-task.v1", "source_manifest": {"authorization": {"allowed_actions": ["fetch_datatracker_document", "fetch_datatracker_relation", "fetch_draft_revision", "fetch_rfc"], "basis": "IETF open-records research export for a closed published_as plus Updates/Obsoletes graph. Distinct from HTTP/3/QUIC packing, OAuth, and RFC 9421 revision stacking. RFC 5077 and RFC 6961 are included because RFC 8446 authentically obsoletes them.", "record_id": "p57-ietf-tls13-handshake-graph-v1", "reviewed_at": "2026-09-07T04:00:00Z", "scope": "public IETF Datatracker, Internet-Draft archive, and RFC Editor records for the TLS 1.3 handshake requirement graph over RFC 8446 and its authentic Updates/Obsoletes targets"}, "data_stage": "source_inventory", "fetch_inventory_sha256": "c86bc411fe4e31382a729a63b451fff99251ebacc2b1323213b4fa8cba891c36", "fetch_receipt": {"allowed_actions": ["fetch_datatracker_document", "fetch_datatracker_relation", "fetch_draft_revision", "fetch_rfc"], "completed_at": "2026-09-07T04:14:41.225467Z", "max_retries": 2, "policy_url": "https://www.ietf.org/about/open-records/", "request_file": "ietf_fetch_request.json", "request_sha256": "577344d9c626e814c0e57a20755290a4e9bdaa66c1e7b545370366600ef34c5f", "requests_per_second": 1.0, "retrievals": [{"content_type": "application/json", "final_url": "https://datatracker.ietf.org/api/v1/doc/document/draft-ietf-tls-tls13/", "kind": "datatracker_document", "observed_at": "2026-09-07T04:14:30.932265Z", "redirect_chain": [], "requested_url": "https://datatracker.ietf.org/api/v1/doc/document/draft-ietf-tls-tls13/", "retrieval_file": "datatracker-draft-ietf-tls-tls13.json", "sha256": "a632cb66584c8f5d364d909b76eed121c1a9d00e341908463e019efad0faee0f", "status": 200}, {"content_type": "application/json", "final_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=draft-ietf-tls-tls13&limit=100", "kind": "datatracker_relation", "observed_at": "2026-09-07T04:14:32.295913Z", "redirect_chain": [], "requested_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=draft-ietf-tls-tls13&limit=100", "retrieval_file": "datatracker-draft-ietf-tls-tls13-relations.json", "sha256": "3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "kind": "draft_revision", "observed_at": "2026-09-07T04:14:32.641074Z", "redirect_chain": [], "requested_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "retrieval_file": "draft-ietf-tls-tls13-28.txt", "sha256": "9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:33.763234Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "retrieval_file": "rfc5077.txt", "sha256": "e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:34.735753Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "retrieval_file": "rfc5246.txt", "sha256": "2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:35.724001Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "retrieval_file": "rfc5705.txt", "sha256": "29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:36.718981Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "retrieval_file": "rfc6066.txt", "sha256": "6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:37.734605Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "retrieval_file": "rfc6961.txt", "sha256": "c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:38.737647Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "retrieval_file": "rfc8446.txt", "sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd", "status": 200}, {"content_type": "application/json", "final_url": "https://datatracker.ietf.org/api/v1/doc/document/rfc8446/", "kind": "datatracker_document", "observed_at": "2026-09-07T04:14:39.917810Z", "redirect_chain": [], "requested_url": "https://datatracker.ietf.org/api/v1/doc/document/rfc8446/", "retrieval_file": "datatracker-rfc8446.json", "sha256": "89b01010d1ff76dc260539d5a453dde0b116d857db43c9e09d4a3dc062c9ce31", "status": 200}, {"content_type": "application/json", "final_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=rfc8446&limit=100", "kind": "datatracker_relation", "observed_at": "2026-09-07T04:14:41.148823Z", "redirect_chain": [], "requested_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=rfc8446&limit=100", "retrieval_file": "datatracker-rfc8446-relations.json", "sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7", "status": 200}], "started_at": "2026-09-07T04:14:30.455103Z", "user_agent_sha256": "1c8887e16ad14ac078ad6e7a52c6cdf9f988eb709fa5671439746106b1265652"}, "generated_at": "2026-09-07T04:16:16.549092Z", "generation_integration": "disabled", "hybrid_train_ready": false, "n": 7, "n_relations": 11, "n_supporting_records": 4, "production_eligible": false, "records": [{"draft_references": [], "excluded_rfc_relation_targets": [{"kind": "obsoletes", "reason": "not_requested", "rfc_number": 4507}], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5077 H. Zhou", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "5077"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 4507 Cisco Systems", "fact_id": "obsoletes:4507", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "4507"}], "kind": "rfc", "occurred_at": "2008-01-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 4, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef", "record_id": "ietf:rfc:5077", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "rfc_number": 5077, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef", "source_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "temporal_semantics": "source_event_time", "text": "Network Working Group J. Salowey\nRequest for Comments: 5077 H. Zhou\nObsoletes: 4507 Cisco Systems\nCategory: Standards Track P. Eronen\n Nokia\n H. Tschofenig\n Nokia Siemens Networks\n January 2008\n\n\n Transport Layer Security (TLS) Session Resumption without\n Server-Side State\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document describes a mechanism that enables the Transport Layer\n Security (TLS) server to resume sessions and avoid keeping per-client\n session state. The TLS server encapsulates the session state into a\n ticket and forwards it to the client. The client can subsequently\n resume a session using the obtained ticket. This document obsoletes\n RFC 4507.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 1]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3. Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . 4\n 3.2. SessionTicket TLS Extension . . . . . . . . . . . . . . . 7\n 3.3. NewSessionTicket Handshake Message . . . . . . . . . . . . 8\n 3.4. Interaction with TLS Session ID . . . . . . . . . . . . . 9\n 4. Recommended Ticket Construction . . . . . . . . . . . . . . . 10\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . 12\n 5.1. Invalidating Sessions . . . . . . . . . . . . . . . . . . 12\n 5.2. Stolen Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.3. Forged Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.4. Denial of Service Attacks . . . . . . . . . . . . . . . . 12\n 5.5. Ticket Protection Key Management . . . . . . . . . . . . . 13\n 5.6. Ticket Lifetime . . . . . . . . . . . . . . . . . . . . . 13\n 5.7. Alternate Ticket Formats and Distribution Schemes . . . . 13\n 5.8. Identity Privacy, Anonymity, and Unlinkability . . . . . . 14\n 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 14\n 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 15\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 15\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 15\n Appendix A. Discussion of Changes to RFC 4507 . . . . . . . . . . 17\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 2]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n1. Introduction\n\n This document defines a way to resume a Transport Layer Security\n (TLS) session without requiring session-specific state at the TLS\n server. This mechanism may be used with any TLS ciphersuite. This\n document applies to both TLS 1.0 defined in [RFC2246], and TLS 1.1\n defined in [RFC4346]. The mechanism makes use of TLS extensions\n defined in [RFC4366] and defines a new TLS message type.\n\n This mechanism is useful in the following situations:\n\n 1. servers that handle a large number of transactions from different\n users\n\n 2. servers that desire to cache sessions for a long time\n\n 3. ability to load balance requests across servers\n\n 4. embedded servers with little memory\n\n This document obsoletes RFC 4507 [RFC4507] to correct an error in the\n encoding that caused the specification to differ from deployed\n implementations. At the time of this writing, there are no known\n implementations that follow the encoding specified in RFC 4507. This\n update to RFC 4507 aligns the document with currently deployed\n implementations. More details of the change are given in Appendix A.\n\n2. Terminology\n\n Within this document, the term 'ticket' refers to a cryptographically\n protected data structure that is created and consumed by the server\n to rebuild session-specific state.\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Protocol\n\n This specification describes a mechanism to distribute encrypted\n session-state information to the client in the form of a ticket and a\n mechanism to present the ticket back to the server. The ticket is\n created by a TLS server and sent to a TLS client. The TLS client\n presents the ticket to the TLS server to resume a session.\n Implementations of this specification are expected to support both\n mechanisms. Other specifications can take advantage of the session\n tickets, perhaps specifying alternative means for distribution or\n selection. For example, a separate specification may describe an\n\n\n\nSalowey, et al. Standards Track [Page 3]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n alternate way to distribute a ticket and use the TLS extension in\n this document to resume the session. This behavior is beyond the\n scope of the document and would need to be described in a separate\n specification.\n\n3.1. Overview\n\n The client indicates that it supports this mechanism by including a\n SessionTicket TLS extension in the ClientHello message. The\n extension will be empty if the client does not already possess a\n ticket for the server. The server sends an empty SessionTicket\n extension to indicate that it will send a new session ticket using\n the NewSessionTicket handshake message. The extension is described\n in Section 3.2.\n\n If the server wants to use this mechanism, it stores its session\n state (such as ciphersuite and master secret) to a ticket that is\n encrypted and integrity-protected by a key known only to the server.\n The ticket is distributed to the client using the NewSessionTicket\n TLS handshake message described in Section 3.3. This message is sent\n during the TLS handshake before the ChangeCipherSpec message, after\n the server has successfully verified the client's Finished message.\n\n Client Server\n\n ClientHello\n (empty SessionTicket extension)-------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1: Message Flow for Full Handshake Issuing New Session Ticket\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 4]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n The client caches this ticket along with the master secret and other\n parameters associated with the current session. When the client\n wishes to resume the session, it includes the ticket in the\n SessionTicket extension within the ClientHello message. Appendix A\n provides a detailed description of the encoding of the extension and\n changes from RFC 4507. The server then decrypts the received ticket,\n verifies the ticket's validity, retrieves the session state from the\n contents of the ticket, and uses this state to resume the session.\n The interaction with the TLS Session ID is described in Section 3.4.\n If the server successfully verifies the client's ticket, then it may\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello.\n\n Client Server\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2: Message Flow for Abbreviated Handshake Using New Session\n Ticket\n\n A recommended ticket format is given in Section 4.\n\n If the server cannot or does not want to honor the ticket, then it\n can initiate a full handshake with the client.\n\n In the case that the server does not wish to issue a new ticket at\n this time, it just completes the handshake without including a\n SessionTicket extension or NewSessionTicket handshake message. This\n is shown below (this flow is identical to Figure 1 in RFC 4346,\n except for the SessionTicket extension in the first message):\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 5]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 3: Message Flow for Server Completing Full Handshake Without\n Issuing New Session Ticket\n\n It is also permissible to have an exchange similar to Figure 3 using\n the abbreviated handshake defined in Figure 2 of RFC 4346, where the\n client uses the SessionTicket extension to resume the session, but\n the server does not wish to issue a new ticket, and therefore does\n not send a SessionTicket extension.\n\n If the server rejects the ticket, it may still wish to issue a new\n ticket after performing the full handshake as shown below (this flow\n is identical to Figure 1, except the SessionTicket extension in the\n ClientHello is not empty):\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 6]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 4: Message Flow for Server Rejecting Ticket, Performing Full\n Handshake, and Issuing New Session Ticket\n\n3.2. SessionTicket TLS Extension\n\n The SessionTicket TLS extension is based on [RFC4366]. The format of\n the ticket is an opaque structure used to carry session-specific\n state information. This extension may be sent in the ClientHello and\n ServerHello.\n\n If the client possesses a ticket that it wants to use to resume a\n session, then it includes the ticket in the SessionTicket extension\n in the ClientHello. If the client does not have a ticket and is\n prepared to receive one in the NewSessionTicket handshake message,\n then it MUST include a zero-length ticket in the SessionTicket\n extension. If the client is not prepared to receive a ticket in the\n NewSessionTicket handshake message, then it MUST NOT include a\n SessionTicket extension unless it is sending a non-empty ticket it\n received through some other means from the server.\n\n The server uses a zero-length SessionTicket extension to indicate to\n the client that it will send a new session ticket using the\n NewSessionTicket handshake message described in Section 3.3. The\n server MUST send this extension in the ServerHello if it wishes to\n issue a new ticket to the client using the NewSessionTicket handshake\n message. The server MUST NOT send this extension if it does not\n receive one in the ClientHello.\n\n\n\n\nSalowey, et al. Standards Track [Page 7]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n If the server fails to verify the ticket, then it falls back to\n performing a full handshake. If the ticket is accepted by the server\n but the handshake fails, the client SHOULD delete the ticket.\n\n The SessionTicket extension has been assigned the number 35. The\n extension_data field of SessionTicket extension contains the ticket.\n\n3.3. NewSessionTicket Handshake Message\n\n This message is sent by the server during the TLS handshake before\n the ChangeCipherSpec message. This message MUST be sent if the\n server included a SessionTicket extension in the ServerHello. This\n message MUST NOT be sent if the server did not include a\n SessionTicket extension in the ServerHello. This message is included\n in the hash used to create and verify the Finished message. In the\n case of a full handshake, the server MUST verify the client's\n Finished message before sending the ticket. The client MUST NOT\n treat the ticket as valid until it has verified the server's Finished\n message. If the server determines that it does not want to include a\n ticket after it has included the SessionTicket extension in the\n ServerHello, then it sends a zero-length ticket in the\n NewSessionTicket handshake message.\n\n If the server successfully verifies the client's ticket, then it MAY\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello in the abbreviated handshake. The client\n should start using the new ticket as soon as possible after it\n verifies the server's Finished message for new connections. Note\n that since the updated ticket is issued before the handshake\n completes, it is possible that the client may not put the new ticket\n into use before it initiates new connections. The server MUST NOT\n assume that the client actually received the updated ticket until it\n successfully verifies the client's Finished message.\n\n The NewSessionTicket handshake message has been assigned the number 4\n and its definition is given at the end of this section. The\n ticket_lifetime_hint field contains a hint from the server about how\n long the ticket should be stored. The value indicates the lifetime\n in seconds as a 32-bit unsigned integer in network byte order\n relative to when the ticket is received. A value of zero is reserved\n to indicate that the lifetime of the ticket is unspecified. A client\n SHOULD delete the ticket and associated state when the time expires.\n It MAY delete the ticket earlier based on local policy. A server MAY\n treat a ticket as valid for a shorter or longer period of time than\n what is stated in the ticket_lifetime_hint.\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 8]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n HandshakeType msg_type;\n uint24 length;\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case session_ticket: NewSessionTicket; /* NEW */\n } body;\n } Handshake;\n\n\n struct {\n uint32 ticket_lifetime_hint;\n opaque ticket<0..2^16-1>;\n } NewSessionTicket;\n\n3.4. Interaction with TLS Session ID\n\n If a server is planning on issuing a session ticket to a client that\n does not present one, it SHOULD include an empty Session ID in the\n ServerHello. If the server rejects the ticket and falls back to the\n full handshake then it may include a non-empty Session ID to indicate\n its support for stateful session resumption. If the client receives\n a session ticket from the server, then it discards any Session ID\n that was sent in the ServerHello.\n\n When presenting a ticket, the client MAY generate and include a\n Session ID in the TLS ClientHello. If the server accepts the ticket\n and the Session ID is not empty, then it MUST respond with the same\n Session ID present in the ClientHello. This allows the client to\n easily differentiate when the server is resuming a session from when\n it is falling back to a full handshake. Since the client generates a\n Session ID, the server MUST NOT rely upon the Session ID having a\n particular value when validating the ticket. If a ticket is\n presented by the client, the server MUST NOT attempt to use the\n Session ID in the ClientHello for stateful session resumption.\n Alternatively, the client MAY include an empty Session ID in the\n ClientHello. In this case, the client ignores the Session ID sent in\n the ServerHello and determines if the server is resuming a session by\n the subsequent handshake messages.\n\n\n\nSalowey, et al. Standards Track [Page 9]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n4. Recommended Ticket Construction\n\n This section describes a recommended format and protection for the\n ticket. Note that the ticket is opaque to the client, so the\n structure is not subject to interoperability concerns, and\n implementations may diverge from this format. If implementations do\n diverge from this format, they must take security concerns seriously.\n Clients MUST NOT examine the ticket under the assumption that it\n complies with this document.\n\n The server uses two different keys: one 128-bit key for Advanced\n Encryption Standard (AES) [AES] in Cipher Block Chaining (CBC) mode\n [CBC] encryption and one 256-bit key for HMAC-SHA-256 [RFC4634].\n\n The ticket is structured as follows:\n\n struct {\n opaque key_name[16];\n opaque iv[16];\n opaque encrypted_state<0..2^16-1>;\n opaque mac[32];\n } ticket;\n\n Here, key_name serves to identify a particular set of keys used to\n protect the ticket. It enables the server to easily recognize\n tickets it has issued. The key_name should be randomly generated to\n avoid collisions between servers. One possibility is to generate new\n random keys and key_name every time the server is started.\n\n The actual state information in encrypted_state is encrypted using\n 128-bit AES in CBC mode with the given IV. The Message\n Authentication Code (MAC) is calculated using HMAC-SHA-256 over\n key_name (16 octets) and IV (16 octets), followed by the length of\n the encrypted_state field (2 octets) and its contents (variable\n length).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 10]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n ProtocolVersion protocol_version;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n opaque master_secret[48];\n ClientIdentity client_identity;\n uint32 timestamp;\n } StatePlaintext;\n\n enum {\n anonymous(0),\n certificate_based(1),\n psk(2)\n } ClientAuthenticationType;\n\n struct {\n ClientAuthenticationType client_authentication_type;\n select (ClientAuthenticationType) {\n case anonymous: struct {};\n case certificate_based:\n ASN.1Cert certificate_list<0..2^24-1>;\n case psk:\n opaque psk_identity<0..2^16-1>; /* from [RFC4279] */\n };\n } ClientIdentity;\n\n The structure StatePlaintext stores the TLS session state including\n the master_secret. The timestamp within this structure allows the\n TLS server to expire tickets. To cover the authentication and key\n exchange protocols provided by TLS, the ClientIdentity structure\n contains the authentication type of the client used in the initial\n exchange (see ClientAuthenticationType). To offer the TLS server\n with the same capabilities for authentication and authorization, a\n certificate list is included in case of public-key-based\n authentication. The TLS server is therefore able to inspect a number\n of different attributes within these certificates. A specific\n implementation might choose to store a subset of this information or\n additional information. Other authentication mechanisms, such as\n Kerberos [RFC2712], would require different client identity data.\n Other TLS extensions may require the inclusion of additional data in\n the StatePlaintext structure.\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 11]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5. Security Considerations\n\n This section addresses security issues related to the usage of a\n ticket. Tickets must be authenticated and encrypted to prevent\n modification or eavesdropping by an attacker. Several attacks\n described below will be possible if this is not carefully done.\n\n Implementations should take care to ensure that the processing of\n tickets does not increase the chance of denial of service as\n described below.\n\n5.1. Invalidating Sessions\n\n The TLS specification requires that TLS sessions be invalidated when\n errors occur. [CSSC] discusses the security implications of this in\n detail. In the analysis within this paper, failure to invalidate\n sessions does not pose a security risk. This is because the TLS\n handshake uses a non-reversible function to derive keys for a session\n so information about one session does not provide an advantage to\n attack the master secret or a different session. If a session\n invalidation scheme is used, the implementation should verify the\n integrity of the ticket before using the contents to invalidate a\n session to ensure that an attacker cannot invalidate a chosen\n session.\n\n5.2. Stolen Tickets\n\n An eavesdropper or man-in-the-middle may obtain the ticket and\n attempt to use it to establish a session with the server; however,\n since the ticket is encrypted and the attacker does not know the\n secret key, a stolen ticket does not help an attacker resume a\n session. A TLS server MUST use strong encryption and integrity\n protection for the ticket to prevent an attacker from using a brute\n force mechanism to obtain the ticket's contents.\n\n5.3. Forged Tickets\n\n A malicious user could forge or alter a ticket in order to resume a\n session, to extend its lifetime, to impersonate another user, or to\n gain additional privileges. This attack is not possible if the\n ticket is protected using a strong integrity protection algorithm\n such as a keyed HMAC-SHA-256.\n\n5.4. Denial of Service Attacks\n\n The key_name field defined in the recommended ticket format helps the\n server efficiently reject tickets that it did not issue. However, an\n adversary could store or generate a large number of tickets to send\n\n\n\nSalowey, et al. Standards Track [Page 12]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n to the TLS server for verification. To minimize the possibility of a\n denial of service, the verification of the ticket should be\n lightweight (e.g., using efficient symmetric key cryptographic\n algorithms).\n\n5.5. Ticket Protection Key Management\n\n A full description of the management of the keys used to protect the\n ticket is beyond the scope of this document. A list of RECOMMENDED\n practices is given below.\n\n o The keys should be generated securely following the randomness\n recommendations in [RFC4086].\n\n o The keys and cryptographic protection algorithms should be at\n least 128 bits in strength. Some ciphersuites and applications\n may require cryptographic protection greater than 128 bits in\n strength.\n\n o The keys should not be used for any purpose other than generating\n and verifying tickets.\n\n o The keys should be changed regularly.\n\n o The keys should be changed if the ticket format or cryptographic\n protection algorithms change.\n\n5.6. Ticket Lifetime\n\n The TLS server controls the lifetime of the ticket. Servers\n determine the acceptable lifetime based on the operational and\n security requirements of the environments in which they are deployed.\n The ticket lifetime may be longer than the 24-hour lifetime\n recommended in [RFC4346]. TLS clients may be given a hint of the\n lifetime of the ticket. Since the lifetime of a ticket may be\n unspecified, a client has its own local policy that determines when\n it discards tickets.\n\n5.7. Alternate Ticket Formats and Distribution Schemes\n\n If the ticket format or distribution scheme defined in this document\n is not used, then great care must be taken in analyzing the security\n of the solution. In particular, if confidential information, such as\n a secret key, is transferred to the client, it MUST be done using\n secure communication so as to prevent attackers from obtaining or\n modifying the key. Also, the ticket MUST have its integrity and\n confidentiality protected with strong cryptographic techniques to\n prevent a breach in the security of the system.\n\n\n\nSalowey, et al. Standards Track [Page 13]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5.8. Identity Privacy, Anonymity, and Unlinkability\n\n This document mandates that the content of the ticket is\n confidentiality protected in order to avoid leakage of its content,\n such as user-relevant information. As such, it prevents disclosure\n of potentially sensitive information carried within the ticket.\n\n The initial handshake exchange, which was used to obtain the ticket,\n might not provide identity confidentiality of the client based on the\n properties of TLS. Another relevant security threat is the ability\n for an on-path adversary to observe multiple TLS handshakes where the\n same ticket is used, therefore concluding they belong to the same\n communication endpoints. Application designers that use the ticket\n mechanism described in this document should consider that\n unlinkability [ANON] is not necessarily provided.\n\n While a full discussion of these topics is beyond the scope of this\n document, it should be noted that it is possible to issue a ticket\n using a TLS renegotiation handshake that occurs after a secure tunnel\n has been established by a previous handshake. This may help address\n some privacy and unlinkability issues in some environments.\n\n6. Acknowledgements\n\n The authors would like to thank the following people for their help\n with preparing and reviewing this document: Eric Rescorla, Mohamad\n Badra, Tim Dierks, Nelson Bolyard, Nancy Cam-Winget, David McGrew,\n Rob Dugal, Russ Housley, Amir Herzberg, Bernard Aboba, and members of\n the TLS working group.\n\n [CSSC] describes a solution that is very similar to the one described\n in this document and gives a detailed analysis of the security\n considerations involved. [RFC2712] describes a mechanism for using\n Kerberos [RFC4120] in TLS ciphersuites, which helped inspire the use\n of tickets to avoid server state. [RFC4851] makes use of a similar\n mechanism to avoid maintaining server state for the cryptographic\n tunnel. [SC97] also investigates the concept of stateless sessions.\n\n The authors would also like to thank Jan Nordqvist, who found the\n encoding error in RFC 4507, corrected by this document. In addition\n Nagendra Modadugu, Wan-Teh Chang, and Michael D'Errico provided\n useful feedback during the review of this document.\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 14]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n7. IANA Considerations\n\n IANA has assigned a TLS extension number of 35 to the SessionTicket\n TLS extension from the TLS registry of ExtensionType values defined\n in [RFC4366].\n\n IANA has assigned a TLS HandshakeType number 4 to the\n NewSessionTicket handshake type from the TLS registry of\n HandshakeType values defined in [RFC4346].\n\n This document does not require any actions or assignments from IANA.\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2246] Dierks, T. and C. Allen, \"The TLS Protocol Version 1.0\",\n RFC 2246, January 1999.\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346, April 2006.\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, April 2006.\n\n [RFC4507] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 4507, May 2006.\n\n8.2. Informative References\n\n [AES] National Institute of Standards and Technology, \"Advanced\n Encryption Standard (AES)\", Federal Information Processing\n Standards (FIPS) Publication 197, November 2001.\n\n [ANON] Pfitzmann, A. and M. Hansen, \"Anonymity, Unlinkability,\n Unobservability, Pseudonymity, and Identity Management - A\n Consolidated Proposal for Terminology\", http://\n dud.inf.tu-dresden.de/literatur/\n Anon_Terminology_v0.26-1.pdf Version 0.26, December 2005.\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 15]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n [CBC] National Institute of Standards and Technology,\n \"Recommendation for Block Cipher Modes of Operation -\n Methods and Techniques\", NIST Special Publication 800-38A,\n December 2001.\n\n [CSSC] Shacham, H., Boneh, D., and E. Rescorla, \"Client-side\n caching for TLS\", Transactions on Information and System\n Security (TISSEC) , Volume 7, Issue 4, November 2004.\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [RFC4086] Eastlake, D., Schiller, J., and S. Crocker, \"Randomness\n Requirements for Security\", BCP 106, RFC 4086, June 2005.\n\n [RFC4120] Neuman, C., Yu, T., Hartman, S., and K. Raeburn, \"The\n Kerberos Network Authentication Service (V5)\", RFC 4120,\n July 2005.\n\n [RFC4279] Eronen, P. and H. Tschofenig, \"Pre-Shared Key Ciphersuites\n for Transport Layer Security (TLS)\", RFC 4279,\n December 2005.\n\n [RFC4634] Eastlake, D. and T. Hansen, \"US Secure Hash Algorithms\n (SHA and HMAC-SHA)\", RFC 4634, July 2006.\n\n [RFC4851] Cam-Winget, N., McGrew, D., Salowey, J., and H. Zhou, \"The\n Flexible Authentication via Secure Tunneling Extensible\n Authentication Protocol Method (EAP-FAST)\", RFC 4851,\n May 2007.\n\n [SC97] Aura, T. and P. Nikander, \"Stateless Connections\",\n Proceedings of the First International Conference on\n Information and Communication Security (ICICS '97) , 1997.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 16]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAppendix A. Discussion of Changes to RFC 4507\n\n RFC 4507 [RFC4507] defines a mechanism to resume a TLS session\n without maintaining server side state by specifying an encrypted\n ticket that is maintained on the client. The client presents this\n ticket to the server in a SessionTicket hello extension. The\n encoding in RFC 4507 used the XDR style encoding specified in TLS\n [RFC4346].\n\n An error in the encoding caused the specification to differ from\n deployed implementations. At the time of this writing there are no\n known implementations that follow the encoding specified in RFC 4507.\n This update to RFC 4507 aligns the document with these currently\n deployed implementations.\n\n Erroneous encoding in RFC 4507 resulted in two length fields; one for\n the extension contents and one for the ticket itself. Hence, for a\n ticket that is 256 bytes long and begins with the hex value FF FF,\n the encoding of the extension would be as follows according to RFC\n 4507:\n\n 00 23 Ticket Extension type 35\n 01 02 Length of extension contents\n 01 00 Length of ticket\n FF FF .. .. Actual ticket\n\n The update proposed in this document reflects what implementations\n actually encode, namely it removes the redundant length field. So,\n for a ticket that is 256 bytes long and begins with the hex value FF\n FF, the encoding of the extension would be as follows according to\n this update:\n\n 00 23 Extension type 35\n 01 00 Length of extension contents (ticket)\n FF FF .. .. Actual ticket\n\n A server implemented according to RFC 4507 receiving a ticket\n extension from a client conforming to this document would interpret\n the first two bytes of the ticket as the length of this ticket. This\n will result in either an inconsistent length field or in the\n processing of a ticket missing the first two bytes. In the first\n case, the server should reject the request based on a malformed\n length. In the second case, the server should reject the ticket\n based on a malformed ticket, incorrect key version, or failed\n decryption. A server implementation based on this update receiving\n an RFC 4507 extension would interpret the first length field as the\n\n\n\n\n\nSalowey, et al. Standards Track [Page 17]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n length of the ticket and include the second two length bytes as the\n first bytes in the ticket, resulting in the ticket being rejected\n based on a malformed ticket, incorrect key version, or failed\n decryption.\n\n Note that the encoding of an empty SessionTicket extension was\n ambiguous in RFC 4507. An RFC 4507 implementation may have encoded\n it as:\n\n 00 23 Extension type 35\n 00 02 Length of extension contents\n 00 00 Length of ticket\n\n or it may have encoded it the same way as this update:\n\n 00 23 Extension type 35\n 00 00 Length of extension contents\n\n A server wishing to support RFC 4507 clients should respond to an\n empty SessionTicket extension encoded the same way as it received it.\n\n A server implementation can construct tickets such that it can detect\n an RFC 4507 implementation, if one existed, by including a cookie at\n the beginning of the tickets that can be differentiated from a valid\n length. For example, if an implementation constructed tickets to\n start with the hex values FF FF, then it could determine where the\n ticket begins and determine the length correctly from the type of\n length fields present.\n\n This document makes a few additional changes to RFC 4507 listed\n below.\n\n o Clarifying that the server can allow session resumption using a\n ticket without issuing a new ticket in Section 3.1.\n\n o Clarifying that the lifetime is relative to when the ticket is\n received in section 3.3.\n\n o Clarifying that the NewSessionTicket handshake message is included\n in the hash generated for the Finished messages in Section 3.3.\n\n o Clarifying the interaction with TLS Session ID in Section 3.4.\n\n o Recommending the use of SHA-256 for the integrity protection of\n the ticket in Section 4.\n\n o Clarifying that additional data can be included in the\n StatePlaintext structure in Section 4.\n\n\n\nSalowey, et al. Standards Track [Page 18]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAuthors' Addresses\n\n Joseph Salowey\n Cisco Systems\n 2901 3rd Ave\n Seattle, WA 98121\n US\n\n EMail: [redacted-email]\n\n\n Hao Zhou\n Cisco Systems\n 4125 Highlander Parkway\n Richfield, OH 44286\n US\n\n EMail: [redacted-email]\n\n\n Pasi Eronen\n Nokia Research Center\n P.O. Box 407\n FIN-00045 Nokia Group\n Finland\n\n EMail: [redacted-email]\n\n\n Hannes Tschofenig\n Nokia Siemens Networks\n Otto-Hahn-Ring 6\n Munich, Bayern 81739\n Germany\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 19]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nFull Copyright Statement\n\n Copyright (C) The IETF Trust (2008).\n\n This document is subject to the rights, licenses and restrictions\n contained in BCP 78, and except as set forth therein, the authors\n retain all their rights.\n\n This document and the information contained herein are provided on an\n \"AS IS\" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\n OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\n THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\n OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\n THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\n WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\n\nIntellectual Property\n\n The IETF takes no position regarding the validity or scope of any\n Intellectual Property Rights or other rights that might be claimed to\n pertain to the implementation or use of the technology described in\n this document or the extent to which any license under such rights\n might or might not be available; nor does it represent that it has\n made any independent effort to identify any such rights. Information\n on the procedures with respect to rights in RFC documents can be\n found in BCP 78 and BCP 79.\n\n Copies of IPR disclosures made to the IETF Secretariat and any\n assurances of licenses to be made available, or the result of an\n attempt made to obtain a general license or permission for the use of\n such proprietary rights by implementers or users of this\n specification can be obtained from the IETF on-line IPR repository at\n http://www.ietf.org/ipr.\n\n The IETF invites any interested party to bring to its attention any\n copyrights, patents or patent applications, or other proprietary\n rights that may cover technology that may be required to implement\n this standard. Please address the information to the IETF at\n ietf-ipr@ietf.org.\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 20]", "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"draft_references": [], "excluded_rfc_relation_targets": [{"kind": "obsoletes", "reason": "not_requested", "rfc_number": 3268}, {"kind": "obsoletes", "reason": "not_requested", "rfc_number": 4346}, {"kind": "obsoletes", "reason": "not_requested", "rfc_number": 4366}, {"kind": "updates", "reason": "not_requested", "rfc_number": 4492}], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5246 Independent", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "5246"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 3268, 4346, 4366 E. Rescorla", "fact_id": "obsoletes:3268", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "3268"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 3268, 4346, 4366 E. Rescorla", "fact_id": "obsoletes:4346", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "4346"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 3268, 4346, 4366 E. Rescorla", "fact_id": "obsoletes:4366", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "4366"}, {"char_end": 291, "char_start": 219, "evidence_quote": "Updates: 4492 RTFM, Inc.", "fact_id": "updates:4492", "field": "updates", "raw_char_end": 297, "raw_char_start": 225, "value": "4492"}], "kind": "rfc", "occurred_at": "2008-08-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 22, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb", "record_id": "ietf:rfc:5246", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "rfc_number": 5246, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb", "source_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "temporal_semantics": "source_event_time", "text": "Network Working Group T. Dierks\nRequest for Comments: 5246 Independent\nObsoletes: 3268, 4346, 4366 E. Rescorla\nUpdates: 4492 RTFM, Inc.\nCategory: Standards Track August 2008\n\n\n The Transport Layer Security (TLS) Protocol\n Version 1.2\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document specifies Version 1.2 of the Transport Layer Security\n (TLS) protocol. The TLS protocol provides communications security\n over the Internet. The protocol allows client/server applications to\n communicate in a way that is designed to prevent eavesdropping,\n tampering, or message forgery.\n\nTable of Contents\n\n 1. Introduction ....................................................4\n 1.1. Requirements Terminology ...................................5\n 1.2. Major Differences from TLS 1.1 .............................5\n 2. Goals ...........................................................6\n 3. Goals of This Document ..........................................7\n 4. Presentation Language ...........................................7\n 4.1. Basic Block Size ...........................................7\n 4.2. Miscellaneous ..............................................8\n 4.3. Vectors ....................................................8\n 4.4. Numbers ....................................................9\n 4.5. Enumerateds ................................................9\n 4.6. Constructed Types .........................................10\n 4.6.1. Variants ...........................................10\n 4.7. Cryptographic Attributes ..................................12\n 4.8. Constants .................................................14\n 5. HMAC and the Pseudorandom Function .............................14\n 6. The TLS Record Protocol ........................................15\n 6.1. Connection States .........................................16\n 6.2. Record Layer ..............................................19\n 6.2.1. Fragmentation ......................................19\n\n\n\nDierks & Rescorla Standards Track [Page 1]\n\f\nRFC 5246 TLS August 2008\n\n\n 6.2.2. Record Compression and Decompression ...............20\n 6.2.3. Record Payload Protection ..........................21\n 6.2.3.1. Null or Standard Stream Cipher ............22\n 6.2.3.2. CBC Block Cipher ..........................22\n 6.2.3.3. AEAD Ciphers ..............................24\n 6.3. Key Calculation ...........................................25\n 7. The TLS Handshaking Protocols ..................................26\n 7.1. Change Cipher Spec Protocol ...............................27\n 7.2. Alert Protocol ............................................28\n 7.2.1. Closure Alerts .....................................29\n 7.2.2. Error Alerts .......................................30\n 7.3. Handshake Protocol Overview ...............................33\n 7.4. Handshake Protocol ........................................37\n 7.4.1. Hello Messages .....................................38\n 7.4.1.1. Hello Request .............................38\n 7.4.1.2. Client Hello ..............................39\n 7.4.1.3. Server Hello ..............................42\n 7.4.1.4. Hello Extensions ..........................44\n 7.4.1.4.1. Signature Algorithms ...........45\n 7.4.2. Server Certificate .................................47\n 7.4.3. Server Key Exchange Message ........................50\n 7.4.4. Certificate Request ................................53\n 7.4.5. Server Hello Done ..................................55\n 7.4.6. Client Certificate .................................55\n 7.4.7. Client Key Exchange Message ........................57\n 7.4.7.1. RSA-Encrypted Premaster Secret Message ....58\n 7.4.7.2. Client Diffie-Hellman Public Value ........61\n 7.4.8. Certificate Verify .................................62\n 7.4.9. Finished ...........................................63\n 8. Cryptographic Computations .....................................64\n 8.1. Computing the Master Secret ...............................64\n 8.1.1. RSA ................................................65\n 8.1.2. Diffie-Hellman .....................................65\n 9. Mandatory Cipher Suites ........................................65\n 10. Application Data Protocol .....................................65\n 11. Security Considerations .......................................65\n 12. IANA Considerations ...........................................65\n Appendix A. Protocol Data Structures and Constant Values ..........68\n A.1. Record Layer ..............................................68\n A.2. Change Cipher Specs Message ...............................69\n A.3. Alert Messages ............................................69\n A.4. Handshake Protocol ........................................70\n A.4.1. Hello Messages .....................................71\n A.4.2. Server Authentication and Key Exchange Messages ....72\n A.4.3. Client Authentication and Key Exchange Messages ....74\n A.4.4. Handshake Finalization Message .....................74\n A.5. The Cipher Suite ..........................................75\n A.6. The Security Parameters ...................................77\n\n\n\nDierks & Rescorla Standards Track [Page 2]\n\f\nRFC 5246 TLS August 2008\n\n\n A.7. Changes to RFC 4492 .......................................78\n Appendix B. Glossary ..............................................78\n Appendix C. Cipher Suite Definitions ..............................83\n Appendix D. Implementation Notes ..................................85\n D.1. Random Number Generation and Seeding ......................85\n D.2. Certificates and Authentication ...........................85\n D.3. Cipher Suites .............................................85\n D.4. Implementation Pitfalls ...................................85\n Appendix E. Backward Compatibility ................................87\n E.1. Compatibility with TLS 1.0/1.1 and SSL 3.0 ................87\n E.2. Compatibility with SSL 2.0 ................................88\n E.3. Avoiding Man-in-the-Middle Version Rollback ...............90\n Appendix F. Security Analysis .....................................91\n F.1. Handshake Protocol ........................................91\n F.1.1. Authentication and Key Exchange ....................91\n F.1.1.1. Anonymous Key Exchange ....................91\n F.1.1.2. RSA Key Exchange and Authentication .......92\n F.1.1.3. Diffie-Hellman Key Exchange with\n Authentication ............................92\n F.1.2. Version Rollback Attacks ...........................93\n F.1.3. Detecting Attacks Against the Handshake Protocol ...94\n F.1.4. Resuming Sessions ..................................94\n F.2. Protecting Application Data ...............................94\n F.3. Explicit IVs ..............................................95\n F.4. Security of Composite Cipher Modes ........................95\n F.5. Denial of Service .........................................96\n F.6. Final Notes ...............................................96\n Normative References ..............................................97\n Informative References ............................................98\n Working Group Information ........................................101\n Contributors .....................................................101\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 3]\n\f\nRFC 5246 TLS August 2008\n\n\n1. Introduction\n\n The primary goal of the TLS protocol is to provide privacy and data\n integrity between two communicating applications. The protocol is\n composed of two layers: the TLS Record Protocol and the TLS Handshake\n Protocol. At the lowest level, layered on top of some reliable\n transport protocol (e.g., TCP [TCP]), is the TLS Record Protocol.\n The TLS Record Protocol provides connection security that has two\n basic properties:\n\n - The connection is private. Symmetric cryptography is used for\n data encryption (e.g., AES [AES], RC4 [SCH], etc.). The keys for\n this symmetric encryption are generated uniquely for each\n connection and are based on a secret negotiated by another\n protocol (such as the TLS Handshake Protocol). The Record\n Protocol can also be used without encryption.\n\n - The connection is reliable. Message transport includes a message\n integrity check using a keyed MAC. Secure hash functions (e.g.,\n SHA-1, etc.) are used for MAC computations. The Record Protocol\n can operate without a MAC, but is generally only used in this mode\n while another protocol is using the Record Protocol as a transport\n for negotiating security parameters.\n\n The TLS Record Protocol is used for encapsulation of various higher-\n level protocols. One such encapsulated protocol, the TLS Handshake\n Protocol, allows the server and client to authenticate each other and\n to negotiate an encryption algorithm and cryptographic keys before\n the application protocol transmits or receives its first byte of\n data. The TLS Handshake Protocol provides connection security that\n has three basic properties:\n\n - The peer's identity can be authenticated using asymmetric, or\n public key, cryptography (e.g., RSA [RSA], DSA [DSS], etc.). This\n authentication can be made optional, but is generally required for\n at least one of the peers.\n\n - The negotiation of a shared secret is secure: the negotiated\n secret is unavailable to eavesdroppers, and for any authenticated\n connection the secret cannot be obtained, even by an attacker who\n can place himself in the middle of the connection.\n\n - The negotiation is reliable: no attacker can modify the\n negotiation communication without being detected by the parties to\n the communication.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 4]\n\f\nRFC 5246 TLS August 2008\n\n\n One advantage of TLS is that it is application protocol independent.\n Higher-level protocols can layer on top of the TLS protocol\n transparently. The TLS standard, however, does not specify how\n protocols add security with TLS; the decisions on how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n1.1. Requirements Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [REQ].\n\n1.2. Major Differences from TLS 1.1\n\n This document is a revision of the TLS 1.1 [TLS1.1] protocol which\n contains improved flexibility, particularly for negotiation of\n cryptographic algorithms. The major changes are:\n\n - The MD5/SHA-1 combination in the pseudorandom function (PRF) has\n been replaced with cipher-suite-specified PRFs. All cipher suites\n in this document use P_SHA256.\n\n - The MD5/SHA-1 combination in the digitally-signed element has been\n replaced with a single hash. Signed elements now include a field\n that explicitly specifies the hash algorithm used.\n\n - Substantial cleanup to the client's and server's ability to\n specify which hash and signature algorithms they will accept.\n Note that this also relaxes some of the constraints on signature\n and hash algorithms from previous versions of TLS.\n\n - Addition of support for authenticated encryption with additional\n data modes.\n\n - TLS Extensions definition and AES Cipher Suites were merged in\n from external [TLSEXT] and [TLSAES].\n\n - Tighter checking of EncryptedPreMasterSecret version numbers.\n\n - Tightened up a number of requirements.\n\n - Verify_data length now depends on the cipher suite (default is\n still 12).\n\n - Cleaned up description of Bleichenbacher/Klima attack defenses.\n\n\n\n\nDierks & Rescorla Standards Track [Page 5]\n\f\nRFC 5246 TLS August 2008\n\n\n - Alerts MUST now be sent in many cases.\n\n - After a certificate_request, if no certificates are available,\n clients now MUST send an empty certificate list.\n\n - TLS_RSA_WITH_AES_128_CBC_SHA is now the mandatory to implement\n cipher suite.\n\n - Added HMAC-SHA256 cipher suites.\n\n - Removed IDEA and DES cipher suites. They are now deprecated and\n will be documented in a separate document.\n\n - Support for the SSLv2 backward-compatible hello is now a MAY, not\n a SHOULD, with sending it a SHOULD NOT. Support will probably\n become a SHOULD NOT in the future.\n\n - Added limited \"fall-through\" to the presentation language to allow\n multiple case arms to have the same encoding.\n\n - Added an Implementation Pitfalls sections\n\n - The usual clarifications and editorial work.\n\n2. Goals\n\n The goals of the TLS protocol, in order of priority, are as follows:\n\n 1. Cryptographic security: TLS should be used to establish a secure\n connection between two parties.\n\n 2. Interoperability: Independent programmers should be able to\n develop applications utilizing TLS that can successfully exchange\n cryptographic parameters without knowledge of one another's code.\n\n 3. Extensibility: TLS seeks to provide a framework into which new\n public key and bulk encryption methods can be incorporated as\n necessary. This will also accomplish two sub-goals: preventing\n the need to create a new protocol (and risking the introduction of\n possible new weaknesses) and avoiding the need to implement an\n entire new security library.\n\n 4. Relative efficiency: Cryptographic operations tend to be highly\n CPU intensive, particularly public key operations. For this\n reason, the TLS protocol has incorporated an optional session\n caching scheme to reduce the number of connections that need to be\n established from scratch. Additionally, care has been taken to\n reduce network activity.\n\n\n\nDierks & Rescorla Standards Track [Page 6]\n\f\nRFC 5246 TLS August 2008\n\n\n3. Goals of This Document\n\n This document and the TLS protocol itself are based on the SSL 3.0\n Protocol Specification as published by Netscape. The differences\n between this protocol and SSL 3.0 are not dramatic, but they are\n significant enough that the various versions of TLS and SSL 3.0 do\n not interoperate (although each protocol incorporates a mechanism by\n which an implementation can back down to prior versions). This\n document is intended primarily for readers who will be implementing\n the protocol and for those doing cryptographic analysis of it. The\n specification has been written with this in mind, and it is intended\n to reflect the needs of those two groups. For that reason, many of\n the algorithm-dependent data structures and rules are included in the\n body of the text (as opposed to in an appendix), providing easier\n access to them.\n\n This document is not intended to supply any details of service\n definition or of interface definition, although it does cover select\n areas of policy as they are required for the maintenance of solid\n security.\n\n4. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used. The syntax draws from\n several sources in its structure. Although it resembles the\n programming language \"C\" in its syntax and XDR [XDR] in both its\n syntax and intent, it would be risky to draw too many parallels. The\n purpose of this presentation language is to document TLS only; it has\n no general application beyond that particular goal.\n\n4.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 7]\n\f\nRFC 5246 TLS August 2008\n\n\n4.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" double\n brackets.\n\n Single-byte entities containing uninterpreted data are of type\n opaque.\n\n4.3. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* 3 consecutive 3 byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n In the following example, mandatory is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 4.4). On the\n other hand, longer can represent up to 800 bytes of data, or 400\n uint16 elements, and it may be empty. Its encoding will include a\n\n\n\nDierks & Rescorla Standards Track [Page 8]\n\f\nRFC 5246 TLS August 2008\n\n\n two-byte actual length field prepended to the vector. The length of\n an encoded vector must be an even multiple of the length of a single\n element (for example, a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is 2 bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n4.4. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are formed from fixed-length series of bytes\n concatenated as described in Section 4.1 and are also unsigned. The\n following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are stored in\n network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n Note that in some cases (e.g., DH parameters) it is necessary to\n represent integers as opaque vectors. In such cases, they are\n represented as unsigned integers (i.e., leading zero octets are not\n required even if the most significant bit is set).\n\n4.5. Enumerateds\n\n An additional sparse data type is available called enum. A field of\n type enum can only assume the values declared in the definition.\n Each definition is a different type. Only enumerateds of the same\n type may be assigned or compared. Every element of an enumerated\n must be assigned a value, as demonstrated in the following example.\n Since the elements of the enumerated are not ordered, they can be\n assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n\n\n\nDierks & Rescorla Standards Track [Page 9]\n\f\nRFC 5246 TLS August 2008\n\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n For enumerateds that are never converted to external representation,\n the numerical information may be omitted.\n\n enum { low, medium, high } Amount;\n\n4.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax for definition is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } [[T]];\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n Structure definitions may be embedded.\n\n4.6.1. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. There\n must be a case arm for every element of the enumeration declared in\n the select. Case arms have limited fall-through: if two case arms\n follow in immediate succession with no fields in between, then they\n\n\n\nDierks & Rescorla Standards Track [Page 10]\n\f\nRFC 5246 TLS August 2008\n\n\n both contain the same fields. Thus, in the example below, \"orange\"\n and \"banana\" both contain V2. Note that this is a new piece of\n syntax in TLS 1.2.\n\n The body of the variant structure may be given a label for reference.\n The mechanism by which the variant is selected at runtime is not\n prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1;\n case e2: Te2;\n case e3: case e4: Te3;\n ....\n case en: Ten;\n } [[fv]];\n } [[Tv]];\n\n For example:\n\n enum { apple, orange, banana } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n select (VariantTag) { /* value of selector is implicit */\n case apple:\n V1; /* VariantBody, tag = apple */\n case orange:\n case banana:\n V2; /* VariantBody, tag = orange or banana */\n } variant_body; /* optional label on variant */\n } VariantRecord;\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 11]\n\f\nRFC 5246 TLS August 2008\n\n\n4.7. Cryptographic Attributes\n\n The five cryptographic operations -- digital signing, stream cipher\n encryption, block cipher encryption, authenticated encryption with\n additional data (AEAD) encryption, and public key encryption -- are\n designated digitally-signed, stream-ciphered, block-ciphered, aead-\n ciphered, and public-key-encrypted, respectively. A field's\n cryptographic processing is specified by prepending an appropriate\n key word designation before the field's type specification.\n Cryptographic keys are implied by the current session state (see\n Section 6.1).\n\n A digitally-signed element is encoded as a struct DigitallySigned:\n\n struct {\n SignatureAndHashAlgorithm algorithm;\n opaque signature<0..2^16-1>;\n } DigitallySigned;\n\n The algorithm field specifies the algorithm used (see Section\n 7.4.1.4.1 for the definition of this field). Note that the\n introduction of the algorithm field is a change from previous\n versions. The signature is a digital signature using those\n algorithms over the contents of the element. The contents themselves\n do not appear on the wire but are simply calculated. The length of\n the signature is specified by the signing algorithm and key.\n\n In RSA signing, the opaque vector contains the signature generated\n using the RSASSA-PKCS1-v1_5 signature scheme defined in [PKCS1]. As\n discussed in [PKCS1], the DigestInfo MUST be DER-encoded [X680]\n [X690]. For hash algorithms without parameters (which includes\n SHA-1), the DigestInfo.AlgorithmIdentifier.parameters field MUST be\n NULL, but implementations MUST accept both without parameters and\n with NULL parameters. Note that earlier versions of TLS used a\n different RSA signature scheme that did not include a DigestInfo\n encoding.\n\n In DSA, the 20 bytes of the SHA-1 hash are run directly through the\n Digital Signing Algorithm with no additional hashing. This produces\n two values, r and s. The DSA signature is an opaque vector, as\n above, the contents of which are the DER encoding of:\n\n Dss-Sig-Value ::= SEQUENCE {\n r INTEGER,\n s INTEGER\n }\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 12]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: In current terminology, DSA refers to the Digital Signature\n Algorithm and DSS refers to the NIST standard. In the original SSL\n and TLS specs, \"DSS\" was used universally. This document uses \"DSA\"\n to refer to the algorithm, \"DSS\" to refer to the standard, and it\n uses \"DSS\" in the code point definitions for historical continuity.\n\n In stream cipher encryption, the plaintext is exclusive-ORed with an\n identical amount of output generated from a cryptographically secure\n keyed pseudorandom number generator.\n\n In block cipher encryption, every block of plaintext encrypts to a\n block of ciphertext. All block cipher encryption is done in CBC\n (Cipher Block Chaining) mode, and all items that are block-ciphered\n will be an exact multiple of the cipher block length.\n\n In AEAD encryption, the plaintext is simultaneously encrypted and\n integrity protected. The input may be of any length, and aead-\n ciphered output is generally larger than the input in order to\n accommodate the integrity check value.\n\n In public key encryption, a public key algorithm is used to encrypt\n data in such a way that it can be decrypted only with the matching\n private key. A public-key-encrypted element is encoded as an opaque\n vector <0..2^16-1>, where the length is specified by the encryption\n algorithm and key.\n\n RSA encryption is done using the RSAES-PKCS1-v1_5 encryption scheme\n defined in [PKCS1].\n\n In the following example\n\n stream-ciphered struct {\n uint8 field1;\n uint8 field2;\n digitally-signed opaque {\n uint8 field3<0..255>;\n uint8 field4;\n };\n } UserType;\n\n The contents of the inner struct (field3 and field4) are used as\n input for the signature/hash algorithm, and then the entire structure\n is encrypted with a stream cipher. The length of this structure, in\n bytes, would be equal to two bytes for field1 and field2, plus two\n bytes for the signature and hash algorithm, plus two bytes for the\n length of the signature, plus the length of the output of the signing\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 13]\n\f\nRFC 5246 TLS August 2008\n\n\n algorithm. The length of the signature is known because the\n algorithm and key used for the signing are known prior to encoding or\n decoding this structure.\n\n4.8. Constants\n\n Typed constants can be defined for purposes of specification by\n declaring a symbol of the desired type and assigning values to it.\n\n Under-specified types (opaque, variable-length vectors, and\n structures that contain opaque) cannot be assigned values. No fields\n of a multi-element structure or vector may be elided.\n\n For example:\n\n struct {\n uint8 f1;\n uint8 f2;\n } Example1;\n\n Example1 ex1 = {1, 4}; /* assigns f1 = 1, f2 = 4 */\n\n5. HMAC and the Pseudorandom Function\n\n The TLS record layer uses a keyed Message Authentication Code (MAC)\n to protect message integrity. The cipher suites defined in this\n document use a construction known as HMAC, described in [HMAC], which\n is based on a hash function. Other cipher suites MAY define their\n own MAC constructions, if needed.\n\n In addition, a construction is required to do expansion of secrets\n into blocks of data for the purposes of key generation or validation.\n This pseudorandom function (PRF) takes as input a secret, a seed, and\n an identifying label and produces an output of arbitrary length.\n\n In this section, we define one PRF, based on HMAC. This PRF with the\n SHA-256 hash function is used for all cipher suites defined in this\n document and in TLS documents published prior to this document when\n TLS 1.2 is negotiated. New cipher suites MUST explicitly specify a\n PRF and, in general, SHOULD use the TLS PRF with SHA-256 or a\n stronger standard hash function.\n\n First, we define a data expansion function, P_hash(secret, data),\n that uses a single hash function to expand a secret and seed into an\n arbitrary quantity of output:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 14]\n\f\nRFC 5246 TLS August 2008\n\n\n P_hash(secret, seed) = HMAC_hash(secret, A(1) + seed) +\n HMAC_hash(secret, A(2) + seed) +\n HMAC_hash(secret, A(3) + seed) + ...\n\n where + indicates concatenation.\n\n A() is defined as:\n\n A(0) = seed\n A(i) = HMAC_hash(secret, A(i-1))\n\n P_hash can be iterated as many times as necessary to produce the\n required quantity of data. For example, if P_SHA256 is being used to\n create 80 bytes of data, it will have to be iterated three times\n (through A(3)), creating 96 bytes of output data; the last 16 bytes\n of the final iteration will then be discarded, leaving 80 bytes of\n output data.\n\n TLS's PRF is created by applying P_hash to the secret as:\n\n PRF(secret, label, seed) = P_(secret, label + seed)\n\n The label is an ASCII string. It should be included in the exact\n form it is given without a length byte or trailing null character.\n For example, the label \"slithy toves\" would be processed by hashing\n the following bytes:\n\n 73 6C 69 74 68 79 20 74 6F 76 65 73\n\n6. The TLS Record Protocol\n\n The TLS Record Protocol is a layered protocol. At each layer,\n messages may include fields for length, description, and content.\n The Record Protocol takes messages to be transmitted, fragments the\n data into manageable blocks, optionally compresses the data, applies\n a MAC, encrypts, and transmits the result. Received data is\n decrypted, verified, decompressed, reassembled, and then delivered to\n higher-level clients.\n\n Four protocols that use the record protocol are described in this\n document: the handshake protocol, the alert protocol, the change\n cipher spec protocol, and the application data protocol. In order to\n allow extension of the TLS protocol, additional record content types\n can be supported by the record protocol. New record content type\n values are assigned by IANA in the TLS Content Type Registry as\n described in Section 12.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 15]\n\f\nRFC 5246 TLS August 2008\n\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST send an\n unexpected_message alert.\n\n Any protocol designed for use over TLS must be carefully designed to\n deal with all possible attacks against it. As a practical matter,\n this means that the protocol designer must be aware of what security\n properties TLS does and does not provide and cannot safely rely on\n the latter.\n\n Note in particular that type and length of a record are not protected\n by encryption. If this information is itself sensitive, application\n designers may wish to take steps (padding, cover traffic) to minimize\n information leakage.\n\n6.1. Connection States\n\n A TLS connection state is the operating environment of the TLS Record\n Protocol. It specifies a compression algorithm, an encryption\n algorithm, and a MAC algorithm. In addition, the parameters for\n these algorithms are known: the MAC key and the bulk encryption keys\n for the connection in both the read and the write directions.\n Logically, there are always four connection states outstanding: the\n current read and write states, and the pending read and write states.\n All records are processed under the current read and write states.\n The security parameters for the pending states can be set by the TLS\n Handshake Protocol, and the ChangeCipherSpec can selectively make\n either of the pending states current, in which case the appropriate\n current state is disposed of and replaced with the pending state; the\n pending state is then reinitialized to an empty state. It is illegal\n to make a state that has not been initialized with security\n parameters a current state. The initial current state always\n specifies that no encryption, compression, or MAC will be used.\n\n The security parameters for a TLS Connection read and write state are\n set by providing the following values:\n\n connection end\n Whether this entity is considered the \"client\" or the \"server\" in\n this connection.\n\n PRF algorithm\n An algorithm used to generate keys from the master secret (see\n Sections 5 and 6.3).\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 16]\n\f\nRFC 5246 TLS August 2008\n\n\n bulk encryption algorithm\n An algorithm to be used for bulk encryption. This specification\n includes the key size of this algorithm, whether it is a block,\n stream, or AEAD cipher, the block size of the cipher (if\n appropriate), and the lengths of explicit and implicit\n initialization vectors (or nonces).\n\n MAC algorithm\n An algorithm to be used for message authentication. This\n specification includes the size of the value returned by the MAC\n algorithm.\n\n compression algorithm\n An algorithm to be used for data compression. This specification\n must include all information the algorithm requires to do\n compression.\n\n master secret\n A 48-byte secret shared between the two peers in the connection.\n\n client random\n A 32-byte value provided by the client.\n\n server random\n A 32-byte value provided by the server.\n\n These parameters are defined in the presentation language as:\n\n enum { server, client } ConnectionEnd;\n\n enum { tls_prf_sha256 } PRFAlgorithm;\n\n enum { null, rc4, 3des, aes }\n BulkCipherAlgorithm;\n\n enum { stream, block, aead } CipherType;\n\n enum { null, hmac_md5, hmac_sha1, hmac_sha256,\n hmac_sha384, hmac_sha512} MACAlgorithm;\n\n enum { null(0), (255) } CompressionMethod;\n\n /* The algorithms specified in CompressionMethod, PRFAlgorithm,\n BulkCipherAlgorithm, and MACAlgorithm may be added to. */\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 17]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ConnectionEnd entity;\n PRFAlgorithm prf_algorithm;\n BulkCipherAlgorithm bulk_cipher_algorithm;\n CipherType cipher_type;\n uint8 enc_key_length;\n uint8 block_length;\n uint8 fixed_iv_length;\n uint8 record_iv_length;\n MACAlgorithm mac_algorithm;\n uint8 mac_length;\n uint8 mac_key_length;\n CompressionMethod compression_algorithm;\n opaque master_secret[48];\n opaque client_random[32];\n opaque server_random[32];\n } SecurityParameters;\n\n The record layer will use the security parameters to generate the\n following six items (some of which are not required by all ciphers,\n and are thus empty):\n\n client write MAC key\n server write MAC key\n client write encryption key\n server write encryption key\n client write IV\n server write IV\n\n The client write parameters are used by the server when receiving and\n processing records and vice versa. The algorithm used for generating\n these items from the security parameters is described in Section 6.3.\n\n Once the security parameters have been set and the keys have been\n generated, the connection states can be instantiated by making them\n the current states. These current states MUST be updated for each\n record processed. Each connection state includes the following\n elements:\n\n compression state\n The current state of the compression algorithm.\n\n cipher state\n The current state of the encryption algorithm. This will consist\n of the scheduled key for that connection. For stream ciphers,\n this will also contain whatever state information is necessary to\n allow the stream to continue to encrypt or decrypt data.\n\n\n\n\nDierks & Rescorla Standards Track [Page 18]\n\f\nRFC 5246 TLS August 2008\n\n\n MAC key\n The MAC key for this connection, as generated above.\n\n sequence number\n Each connection state contains a sequence number, which is\n maintained separately for read and write states. The sequence\n number MUST be set to zero whenever a connection state is made the\n active state. Sequence numbers are of type uint64 and may not\n exceed 2^64-1. Sequence numbers do not wrap. If a TLS\n implementation would need to wrap a sequence number, it must\n renegotiate instead. A sequence number is incremented after each\n record: specifically, the first record transmitted under a\n particular connection state MUST use sequence number 0.\n\n6.2. Record Layer\n\n The TLS record layer receives uninterpreted data from higher layers\n in non-empty blocks of arbitrary size.\n\n6.2.1. Fragmentation\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Client\n message boundaries are not preserved in the record layer (i.e.,\n multiple client messages of the same ContentType MAY be coalesced\n into a single TLSPlaintext record, or a single message MAY be\n fragmented across several records).\n\n struct {\n uint8 major;\n uint8 minor;\n } ProtocolVersion;\n\n enum {\n change_cipher_spec(20), alert(21), handshake(22),\n application_data(23), (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type\n The higher-level protocol used to process the enclosed fragment.\n\n\n\n\nDierks & Rescorla Standards Track [Page 19]\n\f\nRFC 5246 TLS August 2008\n\n\n version\n The version of the protocol being employed. This document\n describes TLS Version 1.2, which uses the version { 3, 3 }. The\n version value 3.3 is historical, deriving from the use of {3, 1}\n for TLS 1.0. (See Appendix A.1.) Note that a client that\n supports multiple versions of TLS may not know what version will\n be employed before it receives the ServerHello. See Appendix E\n for discussion about what record layer version number should be\n employed for ClientHello.\n\n length\n The length (in bytes) of the following TLSPlaintext.fragment. The\n length MUST NOT exceed 2^14.\n\n fragment\n The application data. This data is transparent and treated as an\n independent block to be dealt with by the higher-level protocol\n specified by the type field.\n\n Implementations MUST NOT send zero-length fragments of Handshake,\n Alert, or ChangeCipherSpec content types. Zero-length fragments of\n Application data MAY be sent as they are potentially useful as a\n traffic analysis countermeasure.\n\n Note: Data of different TLS record layer content types MAY be\n interleaved. Application data is generally of lower precedence for\n transmission than other content types. However, records MUST be\n delivered to the network in the same order as they are protected by\n the record layer. Recipients MUST receive and process interleaved\n application layer traffic during handshakes subsequent to the first\n one on a connection.\n\n6.2.2. Record Compression and Decompression\n\n All records are compressed using the compression algorithm defined in\n the current session state. There is always an active compression\n algorithm; however, initially it is defined as\n CompressionMethod.null. The compression algorithm translates a\n TLSPlaintext structure into a TLSCompressed structure. Compression\n functions are initialized with default state information whenever a\n connection state is made active. [RFC3749] describes compression\n algorithms for TLS.\n\n Compression must be lossless and may not increase the content length\n by more than 1024 bytes. If the decompression function encounters a\n TLSCompressed.fragment that would decompress to a length in excess of\n 2^14 bytes, it MUST report a fatal decompression failure error.\n\n\n\n\nDierks & Rescorla Standards Track [Page 20]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ContentType type; /* same as TLSPlaintext.type */\n ProtocolVersion version;/* same as TLSPlaintext.version */\n uint16 length;\n opaque fragment[TLSCompressed.length];\n } TLSCompressed;\n\n length\n The length (in bytes) of the following TLSCompressed.fragment.\n The length MUST NOT exceed 2^14 + 1024.\n\n fragment\n The compressed form of TLSPlaintext.fragment.\n\n Note: A CompressionMethod.null operation is an identity operation;\n no fields are altered.\n\n Implementation note: Decompression functions are responsible for\n ensuring that messages cannot cause internal buffer overflows.\n\n6.2.3. Record Payload Protection\n\n The encryption and MAC functions translate a TLSCompressed\n structure into a TLSCiphertext. The decryption functions reverse\n the process. The MAC of the record also includes a sequence\n number so that missing, extra, or repeated messages are\n detectable.\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n select (SecurityParameters.cipher_type) {\n case stream: GenericStreamCipher;\n case block: GenericBlockCipher;\n case aead: GenericAEADCipher;\n } fragment;\n } TLSCiphertext;\n\n type\n The type field is identical to TLSCompressed.type.\n\n version\n The version field is identical to TLSCompressed.version.\n\n length\n The length (in bytes) of the following TLSCiphertext.fragment.\n The length MUST NOT exceed 2^14 + 2048.\n\n\n\nDierks & Rescorla Standards Track [Page 21]\n\f\nRFC 5246 TLS August 2008\n\n\n fragment\n The encrypted form of TLSCompressed.fragment, with the MAC.\n\n6.2.3.1. Null or Standard Stream Cipher\n\n Stream ciphers (including BulkCipherAlgorithm.null; see Appendix A.6)\n convert TLSCompressed.fragment structures to and from stream\n TLSCiphertext.fragment structures.\n\n stream-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n } GenericStreamCipher;\n\n The MAC is generated as:\n\n MAC(MAC_write_key, seq_num +\n TLSCompressed.type +\n TLSCompressed.version +\n TLSCompressed.length +\n TLSCompressed.fragment);\n\n where \"+\" denotes concatenation.\n\n seq_num\n The sequence number for this record.\n\n MAC\n The MAC algorithm specified by SecurityParameters.mac_algorithm.\n\n Note that the MAC is computed before encryption. The stream cipher\n encrypts the entire block, including the MAC. For stream ciphers\n that do not use a synchronization vector (such as RC4), the stream\n cipher state from the end of one record is simply used on the\n subsequent packet. If the cipher suite is TLS_NULL_WITH_NULL_NULL,\n encryption consists of the identity operation (i.e., the data is not\n encrypted, and the MAC size is zero, implying that no MAC is used).\n For both null and stream ciphers, TLSCiphertext.length is\n TLSCompressed.length plus SecurityParameters.mac_length.\n\n6.2.3.2. CBC Block Cipher\n\n For block ciphers (such as 3DES or AES), the encryption and MAC\n functions convert TLSCompressed.fragment structures to and from block\n TLSCiphertext.fragment structures.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 22]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n opaque IV[SecurityParameters.record_iv_length];\n block-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n uint8 padding[GenericBlockCipher.padding_length];\n uint8 padding_length;\n };\n } GenericBlockCipher;\n\n The MAC is generated as described in Section 6.2.3.1.\n\n IV\n The Initialization Vector (IV) SHOULD be chosen at random, and\n MUST be unpredictable. Note that in versions of TLS prior to 1.1,\n there was no IV field, and the last ciphertext block of the\n previous record (the \"CBC residue\") was used as the IV. This was\n changed to prevent the attacks described in [CBCATT]. For block\n ciphers, the IV length is of length\n SecurityParameters.record_iv_length, which is equal to the\n SecurityParameters.block_size.\n\n padding\n Padding that is added to force the length of the plaintext to be\n an integral multiple of the block cipher's block length. The\n padding MAY be any length up to 255 bytes, as long as it results\n in the TLSCiphertext.length being an integral multiple of the\n block length. Lengths longer than necessary might be desirable to\n frustrate attacks on a protocol that are based on analysis of the\n lengths of exchanged messages. Each uint8 in the padding data\n vector MUST be filled with the padding length value. The receiver\n MUST check this padding and MUST use the bad_record_mac alert to\n indicate padding errors.\n\n padding_length\n The padding length MUST be such that the total size of the\n GenericBlockCipher structure is a multiple of the cipher's block\n length. Legal values range from zero to 255, inclusive. This\n length specifies the length of the padding field exclusive of the\n padding_length field itself.\n\n The encrypted data length (TLSCiphertext.length) is one more than the\n sum of SecurityParameters.block_length, TLSCompressed.length,\n SecurityParameters.mac_length, and padding_length.\n\n Example: If the block length is 8 bytes, the content length\n (TLSCompressed.length) is 61 bytes, and the MAC length is 20 bytes,\n then the length before padding is 82 bytes (this does not include the\n\n\n\nDierks & Rescorla Standards Track [Page 23]\n\f\nRFC 5246 TLS August 2008\n\n\n IV. Thus, the padding length modulo 8 must be equal to 6 in order to\n make the total length an even multiple of 8 bytes (the block length).\n The padding length can be 6, 14, 22, and so on, through 254. If the\n padding length were the minimum necessary, 6, the padding would be 6\n bytes, each containing the value 6. Thus, the last 8 octets of the\n GenericBlockCipher before block encryption would be xx 06 06 06 06 06\n 06 06, where xx is the last octet of the MAC.\n\n Note: With block ciphers in CBC mode (Cipher Block Chaining), it is\n critical that the entire plaintext of the record be known before any\n ciphertext is transmitted. Otherwise, it is possible for the\n attacker to mount the attack described in [CBCATT].\n\n Implementation note: Canvel et al. [CBCTIME] have demonstrated a\n timing attack on CBC padding based on the time required to compute\n the MAC. In order to defend against this attack, implementations\n MUST ensure that record processing time is essentially the same\n whether or not the padding is correct. In general, the best way to\n do this is to compute the MAC even if the padding is incorrect, and\n only then reject the packet. For instance, if the pad appears to be\n incorrect, the implementation might assume a zero-length pad and then\n compute the MAC. This leaves a small timing channel, since MAC\n performance depends to some extent on the size of the data fragment,\n but it is not believed to be large enough to be exploitable, due to\n the large block size of existing MACs and the small size of the\n timing signal.\n\n6.2.3.3. AEAD Ciphers\n\n For AEAD [AEAD] ciphers (such as [CCM] or [GCM]), the AEAD function\n converts TLSCompressed.fragment structures to and from AEAD\n TLSCiphertext.fragment structures.\n\n struct {\n opaque nonce_explicit[SecurityParameters.record_iv_length];\n aead-ciphered struct {\n opaque content[TLSCompressed.length];\n };\n } GenericAEADCipher;\n\n AEAD ciphers take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [AEAD]. The key is either the\n client_write_key or the server_write_key. No MAC key is used.\n\n Each AEAD cipher suite MUST specify how the nonce supplied to the\n AEAD operation is constructed, and what is the length of the\n GenericAEADCipher.nonce_explicit part. In many cases, it is\n\n\n\nDierks & Rescorla Standards Track [Page 24]\n\f\nRFC 5246 TLS August 2008\n\n\n appropriate to use the partially implicit nonce technique described\n in Section 3.2.1 of [AEAD]; with record_iv_length being the length of\n the explicit part. In this case, the implicit part SHOULD be derived\n from key_block as client_write_iv and server_write_iv (as described\n in Section 6.3), and the explicit part is included in\n GenericAEAEDCipher.nonce_explicit.\n\n The plaintext is the TLSCompressed.fragment.\n\n The additional authenticated data, which we denote as\n additional_data, is defined as follows:\n\n additional_data = seq_num + TLSCompressed.type +\n TLSCompressed.version + TLSCompressed.length;\n\n where \"+\" denotes concatenation.\n\n The aead_output consists of the ciphertext output by the AEAD\n encryption operation. The length will generally be larger than\n TLSCompressed.length, but by an amount that varies with the AEAD\n cipher. Since the ciphers might incorporate padding, the amount of\n overhead could vary with different TLSCompressed.length values. Each\n AEAD cipher MUST NOT produce an expansion of greater than 1024 bytes.\n Symbolically,\n\n AEADEncrypted = AEAD-Encrypt(write_key, nonce, plaintext,\n additional_data)\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, the \"additional_data\", and the AEADEncrypted value. The\n output is either the plaintext or an error indicating that the\n decryption failed. There is no separate integrity check. That is:\n\n TLSCompressed.fragment = AEAD-Decrypt(write_key, nonce,\n AEADEncrypted,\n additional_data)\n\n If the decryption fails, a fatal bad_record_mac alert MUST be\n generated.\n\n6.3. Key Calculation\n\n The Record Protocol requires an algorithm to generate keys required\n by the current connection state (see Appendix A.6) from the security\n parameters provided by the handshake protocol.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 25]\n\f\nRFC 5246 TLS August 2008\n\n\n The master secret is expanded into a sequence of secure bytes, which\n is then split to a client write MAC key, a server write MAC key, a\n client write encryption key, and a server write encryption key. Each\n of these is generated from the byte sequence in that order. Unused\n values are empty. Some AEAD ciphers may additionally require a\n client write IV and a server write IV (see Section 6.2.3.3).\n\n When keys and MAC keys are generated, the master secret is used as an\n entropy source.\n\n To generate the key material, compute\n\n key_block = PRF(SecurityParameters.master_secret,\n \"key expansion\",\n SecurityParameters.server_random +\n SecurityParameters.client_random);\n\n until enough output has been generated. Then, the key_block is\n partitioned as follows:\n\n client_write_MAC_key[SecurityParameters.mac_key_length]\n server_write_MAC_key[SecurityParameters.mac_key_length]\n client_write_key[SecurityParameters.enc_key_length]\n server_write_key[SecurityParameters.enc_key_length]\n client_write_IV[SecurityParameters.fixed_iv_length]\n server_write_IV[SecurityParameters.fixed_iv_length]\n\n Currently, the client_write_IV and server_write_IV are only generated\n for implicit nonce techniques as described in Section 3.2.1 of\n [AEAD].\n\n Implementation note: The currently defined cipher suite which\n requires the most material is AES_256_CBC_SHA256. It requires 2 x 32\n byte keys and 2 x 32 byte MAC keys, for a total 128 bytes of key\n material.\n\n7. The TLS Handshaking Protocols\n\n TLS has three subprotocols that are used to allow peers to agree upon\n security parameters for the record layer, to authenticate themselves,\n to instantiate negotiated security parameters, and to report error\n conditions to each other.\n\n The Handshake Protocol is responsible for negotiating a session,\n which consists of the following items:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 26]\n\f\nRFC 5246 TLS August 2008\n\n\n session identifier\n An arbitrary byte sequence chosen by the server to identify an\n active or resumable session state.\n\n peer certificate\n X509v3 [PKIX] certificate of the peer. This element of the state\n may be null.\n\n compression method\n The algorithm used to compress data prior to encryption.\n\n cipher spec\n Specifies the pseudorandom function (PRF) used to generate keying\n material, the bulk data encryption algorithm (such as null, AES,\n etc.) and the MAC algorithm (such as HMAC-SHA1). It also defines\n cryptographic attributes such as the mac_length. (See Appendix\n A.6 for formal definition.)\n\n master secret\n 48-byte secret shared between the client and server.\n\n is resumable\n A flag indicating whether the session can be used to initiate new\n connections.\n\n These items are then used to create security parameters for use by\n the record layer when protecting application data. Many connections\n can be instantiated using the same session through the resumption\n feature of the TLS Handshake Protocol.\n\n7.1. Change Cipher Spec Protocol\n\n The change cipher spec protocol exists to signal transitions in\n ciphering strategies. The protocol consists of a single message,\n which is encrypted and compressed under the current (not the pending)\n connection state. The message consists of a single byte of value 1.\n\n struct {\n enum { change_cipher_spec(1), (255) } type;\n } ChangeCipherSpec;\n\n The ChangeCipherSpec message is sent by both the client and the\n server to notify the receiving party that subsequent records will be\n protected under the newly negotiated CipherSpec and keys. Reception\n of this message causes the receiver to instruct the record layer to\n immediately copy the read pending state into the read current state.\n Immediately after sending this message, the sender MUST instruct the\n record layer to make the write pending state the write active state.\n\n\n\nDierks & Rescorla Standards Track [Page 27]\n\f\nRFC 5246 TLS August 2008\n\n\n (See Section 6.1.) The ChangeCipherSpec message is sent during the\n handshake after the security parameters have been agreed upon, but\n before the verifying Finished message is sent.\n\n Note: If a rehandshake occurs while data is flowing on a connection,\n the communicating parties may continue to send data using the old\n CipherSpec. However, once the ChangeCipherSpec has been sent, the\n new CipherSpec MUST be used. The first side to send the\n ChangeCipherSpec does not know that the other side has finished\n computing the new keying material (e.g., if it has to perform a\n time-consuming public key operation). Thus, a small window of time,\n during which the recipient must buffer the data, MAY exist. In\n practice, with modern machines this interval is likely to be fairly\n short.\n\n7.2. Alert Protocol\n\n One of the content types supported by the TLS record layer is the\n alert type. Alert messages convey the severity of the message\n (warning or fatal) and a description of the alert. Alert messages\n with a level of fatal result in the immediate termination of the\n connection. In this case, other connections corresponding to the\n session may continue, but the session identifier MUST be invalidated,\n preventing the failed session from being used to establish new\n connections. Like other messages, alert messages are encrypted and\n compressed, as specified by the current connection state.\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n\n\n\nDierks & Rescorla Standards Track [Page 28]\n\f\nRFC 5246 TLS August 2008\n\n\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n7.2.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack. Either party may\n initiate the exchange of closing messages.\n\n close_notify\n This message notifies the recipient that the sender will not send\n any more messages on this connection. Note that as of TLS 1.1,\n failure to properly close a connection no longer requires that a\n session not be resumed. This is a change from TLS 1.0 to conform\n with widespread implementation practice.\n\n Either party may initiate a close by sending a close_notify alert.\n Any data received after a closure alert is ignored.\n\n Unless some other fatal alert has been transmitted, each party is\n required to send a close_notify alert before closing the write side\n of the connection. The other party MUST respond with a close_notify\n alert of its own and close down the connection immediately,\n discarding any pending writes. It is not required for the initiator\n of the close to wait for the responding close_notify alert before\n closing the read side of the connection.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation must receive the responding\n close_notify alert before indicating to the application layer that\n the TLS connection has ended. If the application protocol will not\n transfer any additional data, but will only close the underlying\n transport connection, then the implementation MAY choose to close the\n transport without waiting for the responding close_notify. No part\n\n\n\n\nDierks & Rescorla Standards Track [Page 29]\n\f\nRFC 5246 TLS August 2008\n\n\n of this standard should be taken to dictate the manner in which a\n usage profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing a connection reliably delivers\n pending data before destroying the transport.\n\n7.2.2. Error Alerts\n\n Error handling in the TLS Handshake protocol is very simple. When an\n error is detected, the detecting party sends a message to the other\n party. Upon transmission or receipt of a fatal alert message, both\n parties immediately close the connection. Servers and clients MUST\n forget any session-identifiers, keys, and secrets associated with a\n failed connection. Thus, any connection terminated with a fatal\n alert MUST NOT be resumed.\n\n Whenever an implementation encounters a condition which is defined as\n a fatal alert, it MUST send the appropriate alert prior to closing\n the connection. For all errors where an alert level is not\n explicitly specified, the sending party MAY determine at its\n discretion whether to treat this as a fatal error or not. If the\n implementation chooses to send an alert but intends to close the\n connection immediately afterwards, it MUST send that alert at the\n fatal alert level.\n\n If an alert with a level of warning is sent and received, generally\n the connection can continue normally. If the receiving party decides\n not to proceed with the connection (e.g., after having received a\n no_renegotiation alert that it is not willing to accept), it SHOULD\n send a fatal alert to terminate the connection. Given this, the\n sending party cannot, in general, know how the receiving party will\n behave. Therefore, warning alerts are not very useful when the\n sending party wants to continue the connection, and thus are\n sometimes omitted. For example, if a peer decides to accept an\n expired certificate (perhaps after confirming this with the user) and\n wants to continue the connection, it would not generally send a\n certificate_expired alert.\n\n The following error alerts are defined:\n\n unexpected_message\n An inappropriate message was received. This alert is always fatal\n and should never be observed in communication between proper\n implementations.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 30]\n\f\nRFC 5246 TLS August 2008\n\n\n bad_record_mac\n This alert is returned if a record is received with an incorrect\n MAC. This alert also MUST be returned if an alert is sent because\n a TLSCiphertext decrypted in an invalid way: either it wasn't an\n even multiple of the block length, or its padding values, when\n checked, weren't correct. This message is always fatal and should\n never be observed in communication between proper implementations\n (except when messages were corrupted in the network).\n\n decryption_failed_RESERVED\n This alert was used in some earlier versions of TLS, and may have\n permitted certain attacks against the CBC mode [CBCATT]. It MUST\n NOT be sent by compliant implementations.\n\n record_overflow\n A TLSCiphertext record was received that had a length more than\n 2^14+2048 bytes, or a record decrypted to a TLSCompressed record\n with more than 2^14+1024 bytes. This message is always fatal and\n should never be observed in communication between proper\n implementations (except when messages were corrupted in the\n network).\n\n decompression_failure\n The decompression function received improper input (e.g., data\n that would expand to excessive length). This message is always\n fatal and should never be observed in communication between proper\n implementations.\n\n handshake_failure\n Reception of a handshake_failure alert message indicates that the\n sender was unable to negotiate an acceptable set of security\n parameters given the options available. This is a fatal error.\n\n no_certificate_RESERVED\n This alert was used in SSLv3 but not any version of TLS. It MUST\n NOT be sent by compliant implementations.\n\n bad_certificate\n A certificate was corrupt, contained signatures that did not\n verify correctly, etc.\n\n unsupported_certificate\n A certificate was of an unsupported type.\n\n certificate_revoked\n A certificate was revoked by its signer.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 31]\n\f\nRFC 5246 TLS August 2008\n\n\n certificate_expired\n A certificate has expired or is not currently valid.\n\n certificate_unknown\n Some other (unspecified) issue arose in processing the\n certificate, rendering it unacceptable.\n\n illegal_parameter\n A field in the handshake was out of range or inconsistent with\n other fields. This message is always fatal.\n\n unknown_ca\n A valid certificate chain or partial chain was received, but the\n certificate was not accepted because the CA certificate could not\n be located or couldn't be matched with a known, trusted CA. This\n message is always fatal.\n\n access_denied\n A valid certificate was received, but when access control was\n applied, the sender decided not to proceed with negotiation. This\n message is always fatal.\n\n decode_error\n A message could not be decoded because some field was out of the\n specified range or the length of the message was incorrect. This\n message is always fatal and should never be observed in\n communication between proper implementations (except when messages\n were corrupted in the network).\n\n decrypt_error\n A handshake cryptographic operation failed, including being unable\n to correctly verify a signature or validate a Finished message.\n This message is always fatal.\n\n export_restriction_RESERVED\n This alert was used in some earlier versions of TLS. It MUST NOT\n be sent by compliant implementations.\n\n protocol_version\n The protocol version the client has attempted to negotiate is\n recognized but not supported. (For example, old protocol versions\n might be avoided for security reasons.) This message is always\n fatal.\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 32]\n\f\nRFC 5246 TLS August 2008\n\n\n insufficient_security\n Returned instead of handshake_failure when a negotiation has\n failed specifically because the server requires ciphers more\n secure than those supported by the client. This message is always\n fatal.\n\n internal_error\n An internal error unrelated to the peer or the correctness of the\n protocol (such as a memory allocation failure) makes it impossible\n to continue. This message is always fatal.\n\n user_canceled\n This handshake is being canceled for some reason unrelated to a\n protocol failure. If the user cancels an operation after the\n handshake is complete, just closing the connection by sending a\n close_notify is more appropriate. This alert should be followed\n by a close_notify. This message is generally a warning.\n\n no_renegotiation\n Sent by the client in response to a hello request or by the server\n in response to a client hello after initial handshaking. Either\n of these would normally lead to renegotiation; when that is not\n appropriate, the recipient should respond with this alert. At\n that point, the original requester can decide whether to proceed\n with the connection. One case where this would be appropriate is\n where a server has spawned a process to satisfy a request; the\n process might receive security parameters (key length,\n authentication, etc.) at startup, and it might be difficult to\n communicate changes to these parameters after that point. This\n message is always a warning.\n\n unsupported_extension\n sent by clients that receive an extended server hello containing\n an extension that they did not put in the corresponding client\n hello. This message is always fatal.\n\n New Alert values are assigned by IANA as described in Section 12.\n\n7.3. Handshake Protocol Overview\n\n The cryptographic parameters of the session state are produced by the\n TLS Handshake Protocol, which operates on top of the TLS record\n layer. When a TLS client and server first start communicating, they\n agree on a protocol version, select cryptographic algorithms,\n optionally authenticate each other, and use public-key encryption\n techniques to generate shared secrets.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 33]\n\f\nRFC 5246 TLS August 2008\n\n\n The TLS Handshake Protocol involves the following steps:\n\n - Exchange hello messages to agree on algorithms, exchange random\n values, and check for session resumption.\n\n - Exchange the necessary cryptographic parameters to allow the\n client and server to agree on a premaster secret.\n\n - Exchange certificates and cryptographic information to allow the\n client and server to authenticate themselves.\n\n - Generate a master secret from the premaster secret and exchanged\n random values.\n\n - Provide security parameters to the record layer.\n\n - Allow the client and server to verify that their peer has\n calculated the same security parameters and that the handshake\n occurred without tampering by an attacker.\n\n Note that higher layers should not be overly reliant on whether TLS\n always negotiates the strongest possible connection between two\n peers. There are a number of ways in which a man-in-the-middle\n attacker can attempt to make two entities drop down to the least\n secure method they support. The protocol has been designed to\n minimize this risk, but there are still attacks available: for\n example, an attacker could block access to the port a secure service\n runs on, or attempt to get the peers to negotiate an unauthenticated\n connection. The fundamental rule is that higher levels must be\n cognizant of what their security requirements are and never transmit\n information over a channel less secure than what they require. The\n TLS protocol is secure in that any cipher suite offers its promised\n level of security: if you negotiate 3DES with a 1024-bit RSA key\n exchange with a host whose certificate you have verified, you can\n expect to be that secure.\n\n These goals are achieved by the handshake protocol, which can be\n summarized as follows: The client sends a ClientHello message to\n which the server must respond with a ServerHello message, or else a\n fatal error will occur and the connection will fail. The ClientHello\n and ServerHello are used to establish security enhancement\n capabilities between client and server. The ClientHello and\n ServerHello establish the following attributes: Protocol Version,\n Session ID, Cipher Suite, and Compression Method. Additionally, two\n random values are generated and exchanged: ClientHello.random and\n ServerHello.random.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 34]\n\f\nRFC 5246 TLS August 2008\n\n\n The actual key exchange uses up to four messages: the server\n Certificate, the ServerKeyExchange, the client Certificate, and the\n ClientKeyExchange. New key exchange methods can be created by\n specifying a format for these messages and by defining the use of the\n messages to allow the client and server to agree upon a shared\n secret. This secret MUST be quite long; currently defined key\n exchange methods exchange secrets that range from 46 bytes upwards.\n\n Following the hello messages, the server will send its certificate in\n a Certificate message if it is to be authenticated. Additionally, a\n ServerKeyExchange message may be sent, if it is required (e.g., if\n the server has no certificate, or if its certificate is for signing\n only). If the server is authenticated, it may request a certificate\n from the client, if that is appropriate to the cipher suite selected.\n Next, the server will send the ServerHelloDone message, indicating\n that the hello-message phase of the handshake is complete. The\n server will then wait for a client response. If the server has sent\n a CertificateRequest message, the client MUST send the Certificate\n message. The ClientKeyExchange message is now sent, and the content\n of that message will depend on the public key algorithm selected\n between the ClientHello and the ServerHello. If the client has sent\n a certificate with signing ability, a digitally-signed\n CertificateVerify message is sent to explicitly verify possession of\n the private key in the certificate.\n\n At this point, a ChangeCipherSpec message is sent by the client, and\n the client copies the pending Cipher Spec into the current Cipher\n Spec. The client then immediately sends the Finished message under\n the new algorithms, keys, and secrets. In response, the server will\n send its own ChangeCipherSpec message, transfer the pending to the\n current Cipher Spec, and send its Finished message under the new\n Cipher Spec. At this point, the handshake is complete, and the\n client and server may begin to exchange application layer data. (See\n flow chart below.) Application data MUST NOT be sent prior to the\n completion of the first handshake (before a cipher suite other than\n TLS_NULL_WITH_NULL_NULL is established).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 35]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1. Message flow for a full handshake\n\n * Indicates optional or situation-dependent messages that are not\n always sent.\n\n Note: To help avoid pipeline stalls, ChangeCipherSpec is an\n independent TLS protocol content type, and is not actually a TLS\n handshake message.\n\n When the client and server decide to resume a previous session or\n duplicate an existing session (instead of negotiating new security\n parameters), the message flow is as follows:\n\n The client sends a ClientHello using the Session ID of the session to\n be resumed. The server then checks its session cache for a match.\n If a match is found, and the server is willing to re-establish the\n connection under the specified session state, it will send a\n ServerHello with the same Session ID value. At this point, both\n client and server MUST send ChangeCipherSpec messages and proceed\n directly to Finished messages. Once the re-establishment is\n complete, the client and server MAY begin to exchange application\n layer data. (See flow chart below.) If a Session ID match is not\n found, the server generates a new session ID, and the TLS client and\n server perform a full handshake.\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 36]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2. Message flow for an abbreviated handshake\n\n The contents and significance of each message will be presented in\n detail in the following sections.\n\n7.4. Handshake Protocol\n\n The TLS Handshake Protocol is one of the defined higher-level clients\n of the TLS Record Protocol. This protocol is used to negotiate the\n secure attributes of a session. Handshake messages are supplied to\n the TLS record layer, where they are encapsulated within one or more\n TLSPlaintext structures, which are processed and transmitted as\n specified by the current active session state.\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n } body;\n } Handshake;\n\n\n\nDierks & Rescorla Standards Track [Page 37]\n\f\nRFC 5246 TLS August 2008\n\n\n The handshake protocol messages are presented below in the order they\n MUST be sent; sending handshake messages in an unexpected order\n results in a fatal error. Unneeded handshake messages can be\n omitted, however. Note one exception to the ordering: the\n Certificate message is used twice in the handshake (from server to\n client, then from client to server), but described only in its first\n position. The one message that is not bound by these ordering rules\n is the HelloRequest message, which can be sent at any time, but which\n SHOULD be ignored by the client if it arrives in the middle of a\n handshake.\n\n New handshake message types are assigned by IANA as described in\n Section 12.\n\n7.4.1. Hello Messages\n\n The hello phase messages are used to exchange security enhancement\n capabilities between the client and server. When a new session\n begins, the record layer's connection state encryption, hash, and\n compression algorithms are initialized to null. The current\n connection state is used for renegotiation messages.\n\n7.4.1.1. Hello Request\n\n When this message will be sent:\n\n The HelloRequest message MAY be sent by the server at any time.\n\n Meaning of this message:\n\n HelloRequest is a simple notification that the client should begin\n the negotiation process anew. In response, the client should send\n a ClientHello message when convenient. This message is not\n intended to establish which side is the client or server but\n merely to initiate a new negotiation. Servers SHOULD NOT send a\n HelloRequest immediately upon the client's initial connection. It\n is the client's job to send a ClientHello at that time.\n\n This message will be ignored by the client if the client is\n currently negotiating a session. This message MAY be ignored by\n the client if it does not wish to renegotiate a session, or the\n client may, if it wishes, respond with a no_renegotiation alert.\n Since handshake messages are intended to have transmission\n precedence over application data, it is expected that the\n negotiation will begin before no more than a few records are\n received from the client. If the server sends a HelloRequest but\n does not receive a ClientHello in response, it may close the\n connection with a fatal alert.\n\n\n\nDierks & Rescorla Standards Track [Page 38]\n\f\nRFC 5246 TLS August 2008\n\n\n After sending a HelloRequest, servers SHOULD NOT repeat the\n request until the subsequent handshake negotiation is complete.\n\n Structure of this message:\n\n struct { } HelloRequest;\n\n This message MUST NOT be included in the message hashes that are\n maintained throughout the handshake and used in the Finished messages\n and the certificate verify message.\n\n7.4.1.2. Client Hello\n\n When this message will be sent:\n\n When a client first connects to a server, it is required to send\n the ClientHello as its first message. The client can also send a\n ClientHello in response to a HelloRequest or on its own initiative\n in order to renegotiate the security parameters in an existing\n connection.\n\n Structure of this message:\n\n The ClientHello message includes a random structure, which is used\n later in the protocol.\n\n struct {\n uint32 gmt_unix_time;\n opaque random_bytes[28];\n } Random;\n\n gmt_unix_time\n The current time and date in standard UNIX 32-bit format\n (seconds since the midnight starting Jan 1, 1970, UTC, ignoring\n leap seconds) according to the sender's internal clock. Clocks\n are not required to be set correctly by the basic TLS protocol;\n higher-level or application protocols may define additional\n requirements. Note that, for historical reasons, the data\n element is named using GMT, the predecessor of the current\n worldwide time base, UTC.\n\n random_bytes\n 28 bytes generated by a secure random number generator.\n\n The ClientHello message includes a variable-length session\n identifier. If not empty, the value identifies a session between the\n same client and server whose security parameters the client wishes to\n reuse. The session identifier MAY be from an earlier connection,\n\n\n\nDierks & Rescorla Standards Track [Page 39]\n\f\nRFC 5246 TLS August 2008\n\n\n this connection, or from another currently active connection. The\n second option is useful if the client only wishes to update the\n random structures and derived values of a connection, and the third\n option makes it possible to establish several independent secure\n connections without repeating the full handshake protocol. These\n independent connections may occur sequentially or simultaneously; a\n SessionID becomes valid when the handshake negotiating it completes\n with the exchange of Finished messages and persists until it is\n removed due to aging or because a fatal error was encountered on a\n connection associated with the session. The actual contents of the\n SessionID are defined by the server.\n\n opaque SessionID<0..32>;\n\n Warning: Because the SessionID is transmitted without encryption or\n immediate MAC protection, servers MUST NOT place confidential\n information in session identifiers or let the contents of fake\n session identifiers cause any breach of security. (Note that the\n content of the handshake as a whole, including the SessionID, is\n protected by the Finished messages exchanged at the end of the\n handshake.)\n\n The cipher suite list, passed from the client to the server in the\n ClientHello message, contains the combinations of cryptographic\n algorithms supported by the client in order of the client's\n preference (favorite choice first). Each cipher suite defines a key\n exchange algorithm, a bulk encryption algorithm (including secret key\n length), a MAC algorithm, and a PRF. The server will select a cipher\n suite or, if no acceptable choices are presented, return a handshake\n failure alert and close the connection. If the list contains cipher\n suites the server does not recognize, support, or wish to use, the\n server MUST ignore those cipher suites, and process the remaining\n ones as usual.\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n The ClientHello includes a list of compression algorithms supported\n by the client, ordered according to the client's preference.\n\n enum { null(0), (255) } CompressionMethod;\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 40]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ProtocolVersion client_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suites<2..2^16-2>;\n CompressionMethod compression_methods<1..2^8-1>;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ClientHello;\n\n TLS allows extensions to follow the compression_methods field in an\n extensions block. The presence of extensions can be detected by\n determining whether there are bytes following the compression_methods\n at the end of the ClientHello. Note that this method of detecting\n optional data differs from the normal TLS method of having a\n variable-length field, but it is used for compatibility with TLS\n before extensions were defined.\n\n client_version\n The version of the TLS protocol by which the client wishes to\n communicate during this session. This SHOULD be the latest\n (highest valued) version supported by the client. For this\n version of the specification, the version will be 3.3 (see\n Appendix E for details about backward compatibility).\n\n random\n A client-generated random structure.\n\n session_id\n The ID of a session the client wishes to use for this connection.\n This field is empty if no session_id is available, or if the\n client wishes to generate new security parameters.\n\n cipher_suites\n This is a list of the cryptographic options supported by the\n client, with the client's first preference first. If the\n session_id field is not empty (implying a session resumption\n request), this vector MUST include at least the cipher_suite from\n that session. Values are defined in Appendix A.5.\n\n compression_methods\n This is a list of the compression methods supported by the client,\n sorted by client preference. If the session_id field is not empty\n (implying a session resumption request), it MUST include the\n\n\n\nDierks & Rescorla Standards Track [Page 41]\n\f\nRFC 5246 TLS August 2008\n\n\n compression_method from that session. This vector MUST contain,\n and all implementations MUST support, CompressionMethod.null.\n Thus, a client and server will always be able to agree on a\n compression method.\n\n extensions\n Clients MAY request extended functionality from servers by sending\n data in the extensions field. The actual \"Extension\" format is\n defined in Section 7.4.1.4.\n\n In the event that a client requests additional functionality using\n extensions, and this functionality is not supplied by the server, the\n client MAY abort the handshake. A server MUST accept ClientHello\n messages both with and without the extensions field, and (as for all\n other messages) it MUST check that the amount of data in the message\n precisely matches one of these formats; if not, then it MUST send a\n fatal \"decode_error\" alert.\n\n After sending the ClientHello message, the client waits for a\n ServerHello message. Any handshake message returned by the server,\n except for a HelloRequest, is treated as a fatal error.\n\n7.4.1.3. Server Hello\n\n When this message will be sent:\n\n The server will send this message in response to a ClientHello\n message when it was able to find an acceptable set of algorithms.\n If it cannot find such a match, it will respond with a handshake\n failure alert.\n\n Structure of this message:\n\n struct {\n ProtocolVersion server_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ServerHello;\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 42]\n\f\nRFC 5246 TLS August 2008\n\n\n The presence of extensions can be detected by determining whether\n there are bytes following the compression_method field at the end of\n the ServerHello.\n\n server_version\n This field will contain the lower of that suggested by the client\n in the client hello and the highest supported by the server. For\n this version of the specification, the version is 3.3. (See\n Appendix E for details about backward compatibility.)\n\n random\n This structure is generated by the server and MUST be\n independently generated from the ClientHello.random.\n\n session_id\n This is the identity of the session corresponding to this\n connection. If the ClientHello.session_id was non-empty, the\n server will look in its session cache for a match. If a match is\n found and the server is willing to establish the new connection\n using the specified session state, the server will respond with\n the same value as was supplied by the client. This indicates a\n resumed session and dictates that the parties must proceed\n directly to the Finished messages. Otherwise, this field will\n contain a different value identifying the new session. The server\n may return an empty session_id to indicate that the session will\n not be cached and therefore cannot be resumed. If a session is\n resumed, it must be resumed using the same cipher suite it was\n originally negotiated with. Note that there is no requirement\n that the server resume any session even if it had formerly\n provided a session_id. Clients MUST be prepared to do a full\n negotiation -- including negotiating new cipher suites -- during\n any handshake.\n\n cipher_suite\n The single cipher suite selected by the server from the list in\n ClientHello.cipher_suites. For resumed sessions, this field is\n the value from the state of the session being resumed.\n\n compression_method\n The single compression algorithm selected by the server from the\n list in ClientHello.compression_methods. For resumed sessions,\n this field is the value from the resumed session state.\n\n extensions\n A list of extensions. Note that only extensions offered by the\n client can appear in the server's list.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 43]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.1.4. Hello Extensions\n\n The extension format is:\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n signature_algorithms(13), (65535)\n } ExtensionType;\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The initial set of extensions is defined in a companion document\n [TLSEXT]. The list of extension types is maintained by IANA as\n described in Section 12.\n\n An extension type MUST NOT appear in the ServerHello unless the same\n extension type appeared in the corresponding ClientHello. If a\n client receives an extension type in ServerHello that it did not\n request in the associated ClientHello, it MUST abort the handshake\n with an unsupported_extension fatal alert.\n\n Nonetheless, \"server-oriented\" extensions may be provided in the\n future within this framework. Such an extension (say, of type x)\n would require the client to first send an extension of type x in a\n ClientHello with empty extension_data to indicate that it supports\n the extension type. In this case, the client is offering the\n capability to understand the extension type, and the server is taking\n the client up on its offer.\n\n When multiple extensions of different types are present in the\n ClientHello or ServerHello messages, the extensions MAY appear in any\n order. There MUST NOT be more than one extension of the same type.\n\n Finally, note that extensions can be sent both when starting a new\n session and when requesting session resumption. Indeed, a client\n that requests session resumption does not in general know whether the\n server will accept this request, and therefore it SHOULD send the\n same extensions as it would send if it were not attempting\n resumption.\n\n\n\nDierks & Rescorla Standards Track [Page 44]\n\f\nRFC 5246 TLS August 2008\n\n\n In general, the specification of each extension type needs to\n describe the effect of the extension both during full handshake and\n session resumption. Most current TLS extensions are relevant only\n when a session is initiated: when an older session is resumed, the\n server does not process these extensions in Client Hello, and does\n not include them in Server Hello. However, some extensions may\n specify different behavior during session resumption.\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions, and some are simply refusals to support particular\n features. In general, error alerts should be used for the former,\n and a field in the server extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem.\n\n Often the fact that the extension fields are included in the\n inputs to the Finished message hashes will be sufficient, but\n extreme care is needed when the extension changes the meaning of\n messages sent in the handshake phase. Designers and implementors\n should be aware of the fact that until the handshake has been\n authenticated, active attackers can modify messages and insert,\n remove, or replace extensions.\n\n - It would be technically possible to use extensions to change major\n aspects of the design of TLS; for example the design of cipher\n suite negotiation. This is not recommended; it would be more\n appropriate to define a new version of TLS -- particularly since\n the TLS handshake algorithms have specific protection against\n version rollback attacks based on the version number, and the\n possibility of version rollback should be a significant\n consideration in any major design change.\n\n7.4.1.4.1. Signature Algorithms\n\n The client uses the \"signature_algorithms\" extension to indicate to\n the server which signature/hash algorithm pairs may be used in\n digital signatures. The \"extension_data\" field of this extension\n contains a \"supported_signature_algorithms\" value.\n\n\n\nDierks & Rescorla Standards Track [Page 45]\n\f\nRFC 5246 TLS August 2008\n\n\n enum {\n none(0), md5(1), sha1(2), sha224(3), sha256(4), sha384(5),\n sha512(6), (255)\n } HashAlgorithm;\n\n enum { anonymous(0), rsa(1), dsa(2), ecdsa(3), (255) }\n SignatureAlgorithm;\n\n struct {\n HashAlgorithm hash;\n SignatureAlgorithm signature;\n } SignatureAndHashAlgorithm;\n\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2..2^16-2>;\n\n Each SignatureAndHashAlgorithm value lists a single hash/signature\n pair that the client is willing to verify. The values are indicated\n in descending order of preference.\n\n Note: Because not all signature algorithms and hash algorithms may be\n accepted by an implementation (e.g., DSA with SHA-1, but not\n SHA-256), algorithms here are listed in pairs.\n\n hash\n This field indicates the hash algorithm which may be used. The\n values indicate support for unhashed data, MD5 [MD5], SHA-1,\n SHA-224, SHA-256, SHA-384, and SHA-512 [SHS], respectively. The\n \"none\" value is provided for future extensibility, in case of a\n signature algorithm which does not require hashing before signing.\n\n signature\n This field indicates the signature algorithm that may be used.\n The values indicate anonymous signatures, RSASSA-PKCS1-v1_5\n [PKCS1] and DSA [DSS], and ECDSA [ECDSA], respectively. The\n \"anonymous\" value is meaningless in this context but used in\n Section 7.4.3. It MUST NOT appear in this extension.\n\n The semantics of this extension are somewhat complicated because the\n cipher suite indicates permissible signature algorithms but not hash\n algorithms. Sections 7.4.2 and 7.4.3 describe the appropriate rules.\n\n If the client supports only the default hash and signature algorithms\n (listed in this section), it MAY omit the signature_algorithms\n extension. If the client does not support the default algorithms, or\n supports other hash and signature algorithms (and it is willing to\n use them for verifying messages sent by the server, i.e., server\n certificates and server key exchange), it MUST send the\n\n\n\nDierks & Rescorla Standards Track [Page 46]\n\f\nRFC 5246 TLS August 2008\n\n\n signature_algorithms extension, listing the algorithms it is willing\n to accept.\n\n If the client does not send the signature_algorithms extension, the\n server MUST do the following:\n\n - If the negotiated key exchange algorithm is one of (RSA, DHE_RSA,\n DH_RSA, RSA_PSK, ECDH_RSA, ECDHE_RSA), behave as if client had\n sent the value {sha1,rsa}.\n\n - If the negotiated key exchange algorithm is one of (DHE_DSS,\n DH_DSS), behave as if the client had sent the value {sha1,dsa}.\n\n - If the negotiated key exchange algorithm is one of (ECDH_ECDSA,\n ECDHE_ECDSA), behave as if the client had sent value {sha1,ecdsa}.\n\n Note: this is a change from TLS 1.1 where there are no explicit\n rules, but as a practical matter one can assume that the peer\n supports MD5 and SHA-1.\n\n Note: this extension is not meaningful for TLS versions prior to 1.2.\n Clients MUST NOT offer it if they are offering prior versions.\n However, even if clients do offer it, the rules specified in [TLSEXT]\n require servers to ignore extensions they do not understand.\n\n Servers MUST NOT send this extension. TLS servers MUST support\n receiving this extension.\n\n When performing session resumption, this extension is not included in\n Server Hello, and the server ignores the extension in Client Hello\n (if present).\n\n7.4.2. Server Certificate\n\n When this message will be sent:\n\n The server MUST send a Certificate message whenever the agreed-\n upon key exchange method uses certificates for authentication\n (this includes all key exchange methods defined in this document\n except DH_anon). This message will always immediately follow the\n ServerHello message.\n\n Meaning of this message:\n\n This message conveys the server's certificate chain to the client.\n\n The certificate MUST be appropriate for the negotiated cipher\n suite's key exchange algorithm and any negotiated extensions.\n\n\n\nDierks & Rescorla Standards Track [Page 47]\n\f\nRFC 5246 TLS August 2008\n\n\n Structure of this message:\n\n opaque ASN.1Cert<1..2^24-1>;\n\n struct {\n ASN.1Cert certificate_list<0..2^24-1>;\n } Certificate;\n\n certificate_list\n This is a sequence (chain) of certificates. The sender's\n certificate MUST come first in the list. Each following\n certificate MUST directly certify the one preceding it. Because\n certificate validation requires that root keys be distributed\n independently, the self-signed certificate that specifies the root\n certificate authority MAY be omitted from the chain, under the\n assumption that the remote end must already possess it in order to\n validate it in any case.\n\n The same message type and structure will be used for the client's\n response to a certificate request message. Note that a client MAY\n send no certificates if it does not have an appropriate certificate\n to send in response to the server's authentication request.\n\n Note: PKCS #7 [PKCS7] is not used as the format for the certificate\n vector because PKCS #6 [PKCS6] extended certificates are not used.\n Also, PKCS #7 defines a SET rather than a SEQUENCE, making the task\n of parsing the list more difficult.\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3, unless explicitly negotiated\n otherwise (e.g., [TLSPGP]).\n\n - The end entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected key exchange\n algorithm.\n\n Key Exchange Alg. Certificate Key Type\n\n RSA RSA public key; the certificate MUST allow the\n RSA_PSK key to be used for encryption (the\n keyEncipherment bit MUST be set if the key\n usage extension is present).\n Note: RSA_PSK is defined in [TLSPSK].\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 48]\n\f\nRFC 5246 TLS August 2008\n\n\n DHE_RSA RSA public key; the certificate MUST allow the\n ECDHE_RSA key to be used for signing (the\n digitalSignature bit MUST be set if the key\n usage extension is present) with the signature\n scheme and hash algorithm that will be employed\n in the server key exchange message.\n Note: ECDHE_RSA is defined in [TLSECC].\n\n DHE_DSS DSA public key; the certificate MUST allow the\n key to be used for signing with the hash\n algorithm that will be employed in the server\n key exchange message.\n\n DH_DSS Diffie-Hellman public key; the keyAgreement bit\n DH_RSA MUST be set if the key usage extension is\n present.\n\n ECDH_ECDSA ECDH-capable public key; the public key MUST\n ECDH_RSA use a curve and point format supported by the\n client, as described in [TLSECC].\n\n ECDHE_ECDSA ECDSA-capable public key; the certificate MUST\n allow the key to be used for signing with the\n hash algorithm that will be employed in the\n server key exchange message. The public key\n MUST use a curve and point format supported by\n the client, as described in [TLSECC].\n\n - The \"server_name\" and \"trusted_ca_keys\" extensions [TLSEXT] are\n used to guide certificate selection.\n\n If the client provided a \"signature_algorithms\" extension, then all\n certificates provided by the server MUST be signed by a\n hash/signature algorithm pair that appears in that extension. Note\n that this implies that a certificate containing a key for one\n signature algorithm MAY be signed using a different signature\n algorithm (for instance, an RSA key signed with a DSA key). This is\n a departure from TLS 1.1, which required that the algorithms be the\n same. Note that this also implies that the DH_DSS, DH_RSA,\n ECDH_ECDSA, and ECDH_RSA key exchange algorithms do not restrict the\n algorithm used to sign the certificate. Fixed DH certificates MAY be\n signed with any hash/signature algorithm pair appearing in the\n extension. The names DH_DSS, DH_RSA, ECDH_ECDSA, and ECDH_RSA are\n historical.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 49]\n\f\nRFC 5246 TLS August 2008\n\n\n If the server has multiple certificates, it chooses one of them based\n on the above-mentioned criteria (in addition to other criteria, such\n as transport layer endpoint, local configuration and preferences,\n etc.). If the server has a single certificate, it SHOULD attempt to\n validate that it meets these criteria.\n\n Note that there are certificates that use algorithms and/or algorithm\n combinations that cannot be currently used with TLS. For example, a\n certificate with RSASSA-PSS signature key (id-RSASSA-PSS OID in\n SubjectPublicKeyInfo) cannot be used because TLS defines no\n corresponding signature algorithm.\n\n As cipher suites that specify new key exchange methods are specified\n for the TLS protocol, they will imply the certificate format and the\n required encoded keying information.\n\n7.4.3. Server Key Exchange Message\n\n When this message will be sent:\n\n This message will be sent immediately after the server Certificate\n message (or the ServerHello message, if this is an anonymous\n negotiation).\n\n The ServerKeyExchange message is sent by the server only when the\n server Certificate message (if sent) does not contain enough data\n to allow the client to exchange a premaster secret. This is true\n for the following key exchange methods:\n\n DHE_DSS\n DHE_RSA\n DH_anon\n\n It is not legal to send the ServerKeyExchange message for the\n following key exchange methods:\n\n RSA\n DH_DSS\n DH_RSA\n\n Other key exchange algorithms, such as those defined in [TLSECC],\n MUST specify whether the ServerKeyExchange message is sent or not;\n and if the message is sent, its contents.\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 50]\n\f\nRFC 5246 TLS August 2008\n\n\n Meaning of this message:\n\n This message conveys cryptographic information to allow the client\n to communicate the premaster secret: a Diffie-Hellman public key\n with which the client can complete a key exchange (with the result\n being the premaster secret) or a public key for some other\n algorithm.\n\n Structure of this message:\n\n enum { dhe_dss, dhe_rsa, dh_anon, rsa, dh_dss, dh_rsa\n /* may be extended, e.g., for ECDH -- see [TLSECC] */\n } KeyExchangeAlgorithm;\n\n struct {\n opaque dh_p<1..2^16-1>;\n opaque dh_g<1..2^16-1>;\n opaque dh_Ys<1..2^16-1>;\n } ServerDHParams; /* Ephemeral DH parameters */\n\n dh_p\n The prime modulus used for the Diffie-Hellman operation.\n\n dh_g\n The generator used for the Diffie-Hellman operation.\n\n dh_Ys\n The server's Diffie-Hellman public value (g^X mod p).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 51]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n select (KeyExchangeAlgorithm) {\n case dh_anon:\n ServerDHParams params;\n case dhe_dss:\n case dhe_rsa:\n ServerDHParams params;\n digitally-signed struct {\n opaque client_random[32];\n opaque server_random[32];\n ServerDHParams params;\n } signed_params;\n case rsa:\n case dh_dss:\n case dh_rsa:\n struct {} ;\n /* message is omitted for rsa, dh_dss, and dh_rsa */\n /* may be extended, e.g., for ECDH -- see [TLSECC] */\n };\n } ServerKeyExchange;\n\n params\n The server's key exchange parameters.\n\n signed_params\n For non-anonymous key exchanges, a signature over the server's\n key exchange parameters.\n\n If the client has offered the \"signature_algorithms\" extension, the\n signature algorithm and hash algorithm MUST be a pair listed in that\n extension. Note that there is a possibility for inconsistencies\n here. For instance, the client might offer DHE_DSS key exchange but\n omit any DSA pairs from its \"signature_algorithms\" extension. In\n order to negotiate correctly, the server MUST check any candidate\n cipher suites against the \"signature_algorithms\" extension before\n selecting them. This is somewhat inelegant but is a compromise\n designed to minimize changes to the original cipher suite design.\n\n In addition, the hash and signature algorithms MUST be compatible\n with the key in the server's end-entity certificate. RSA keys MAY be\n used with any permitted hash algorithm, subject to restrictions in\n the certificate, if any.\n\n Because DSA signatures do not contain any secure indication of hash\n algorithm, there is a risk of hash substitution if multiple hashes\n may be used with any key. Currently, DSA [DSS] may only be used with\n SHA-1. Future revisions of DSS [DSS-3] are expected to allow the use\n of other digest algorithms with DSA, as well as guidance as to which\n\n\n\nDierks & Rescorla Standards Track [Page 52]\n\f\nRFC 5246 TLS August 2008\n\n\n digest algorithms should be used with each key size. In addition,\n future revisions of [PKIX] may specify mechanisms for certificates to\n indicate which digest algorithms are to be used with DSA.\n\n As additional cipher suites are defined for TLS that include new key\n exchange algorithms, the server key exchange message will be sent if\n and only if the certificate type associated with the key exchange\n algorithm does not provide enough information for the client to\n exchange a premaster secret.\n\n7.4.4. Certificate Request\n\n When this message will be sent:\n\n A non-anonymous server can optionally request a certificate from\n the client, if appropriate for the selected cipher suite. This\n message, if sent, will immediately follow the ServerKeyExchange\n message (if it is sent; otherwise, this message follows the\n server's Certificate message).\n\n Structure of this message:\n\n enum {\n rsa_sign(1), dss_sign(2), rsa_fixed_dh(3), dss_fixed_dh(4),\n rsa_ephemeral_dh_RESERVED(5), dss_ephemeral_dh_RESERVED(6),\n fortezza_dms_RESERVED(20), (255)\n } ClientCertificateType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n ClientCertificateType certificate_types<1..2^8-1>;\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2^16-1>;\n DistinguishedName certificate_authorities<0..2^16-1>;\n } CertificateRequest;\n\n certificate_types\n A list of the types of certificate types that the client may\n offer.\n\n rsa_sign a certificate containing an RSA key\n dss_sign a certificate containing a DSA key\n rsa_fixed_dh a certificate containing a static DH key.\n dss_fixed_dh a certificate containing a static DH key\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 53]\n\f\nRFC 5246 TLS August 2008\n\n\n supported_signature_algorithms\n A list of the hash/signature algorithm pairs that the server is\n able to verify, listed in descending order of preference.\n\n certificate_authorities\n A list of the distinguished names [X501] of acceptable\n certificate_authorities, represented in DER-encoded format. These\n distinguished names may specify a desired distinguished name for a\n root CA or for a subordinate CA; thus, this message can be used to\n describe known roots as well as a desired authorization space. If\n the certificate_authorities list is empty, then the client MAY\n send any certificate of the appropriate ClientCertificateType,\n unless there is some external arrangement to the contrary.\n\n The interaction of the certificate_types and\n supported_signature_algorithms fields is somewhat complicated.\n certificate_types has been present in TLS since SSLv3, but was\n somewhat underspecified. Much of its functionality is superseded by\n supported_signature_algorithms. The following rules apply:\n\n - Any certificates provided by the client MUST be signed using a\n hash/signature algorithm pair found in\n supported_signature_algorithms.\n\n - The end-entity certificate provided by the client MUST contain a\n key that is compatible with certificate_types. If the key is a\n signature key, it MUST be usable with some hash/signature\n algorithm pair in supported_signature_algorithms.\n\n - For historical reasons, the names of some client certificate types\n include the algorithm used to sign the certificate. For example,\n in earlier versions of TLS, rsa_fixed_dh meant a certificate\n signed with RSA and containing a static DH key. In TLS 1.2, this\n functionality has been obsoleted by the\n supported_signature_algorithms, and the certificate type no longer\n restricts the algorithm used to sign the certificate. For\n example, if the server sends dss_fixed_dh certificate type and\n {{sha1, dsa}, {sha1, rsa}} signature types, the client MAY reply\n with a certificate containing a static DH key, signed with RSA-\n SHA1.\n\n New ClientCertificateType values are assigned by IANA as described in\n Section 12.\n\n Note: Values listed as RESERVED may not be used. They were used in\n SSLv3.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 54]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: It is a fatal handshake_failure alert for an anonymous server\n to request client authentication.\n\n7.4.5. Server Hello Done\n\n When this message will be sent:\n\n The ServerHelloDone message is sent by the server to indicate the\n end of the ServerHello and associated messages. After sending\n this message, the server will wait for a client response.\n\n Meaning of this message:\n\n This message means that the server is done sending messages to\n support the key exchange, and the client can proceed with its\n phase of the key exchange.\n\n Upon receipt of the ServerHelloDone message, the client SHOULD\n verify that the server provided a valid certificate, if required,\n and check that the server hello parameters are acceptable.\n\n Structure of this message:\n\n struct { } ServerHelloDone;\n\n7.4.6. Client Certificate\n\n When this message will be sent:\n\n This is the first message the client can send after receiving a\n ServerHelloDone message. This message is only sent if the server\n requests a certificate. If no suitable certificate is available,\n the client MUST send a certificate message containing no\n certificates. That is, the certificate_list structure has a\n length of zero. If the client does not send any certificates, the\n server MAY at its discretion either continue the handshake without\n client authentication, or respond with a fatal handshake_failure\n alert. Also, if some aspect of the certificate chain was\n unacceptable (e.g., it was not signed by a known, trusted CA), the\n server MAY at its discretion either continue the handshake\n (considering the client unauthenticated) or send a fatal alert.\n\n Client certificates are sent using the Certificate structure\n defined in Section 7.4.2.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 55]\n\f\nRFC 5246 TLS August 2008\n\n\n Meaning of this message:\n\n This message conveys the client's certificate chain to the server;\n the server will use it when verifying the CertificateVerify\n message (when the client authentication is based on signing) or\n calculating the premaster secret (for non-ephemeral Diffie-\n Hellman). The certificate MUST be appropriate for the negotiated\n cipher suite's key exchange algorithm, and any negotiated\n extensions.\n\n In particular:\n\n - The certificate type MUST be X.509v3, unless explicitly negotiated\n otherwise (e.g., [TLSPGP]).\n\n - The end-entity certificate's public key (and associated\n restrictions) has to be compatible with the certificate types\n listed in CertificateRequest:\n\n Client Cert. Type Certificate Key Type\n\n rsa_sign RSA public key; the certificate MUST allow the\n key to be used for signing with the signature\n scheme and hash algorithm that will be\n employed in the certificate verify message.\n\n dss_sign DSA public key; the certificate MUST allow the\n key to be used for signing with the hash\n algorithm that will be employed in the\n certificate verify message.\n\n ecdsa_sign ECDSA-capable public key; the certificate MUST\n allow the key to be used for signing with the\n hash algorithm that will be employed in the\n certificate verify message; the public key\n MUST use a curve and point format supported by\n the server.\n\n rsa_fixed_dh Diffie-Hellman public key; MUST use the same\n dss_fixed_dh parameters as server's key.\n\n rsa_fixed_ecdh ECDH-capable public key; MUST use the\n ecdsa_fixed_ecdh same curve as the server's key, and MUST use a\n point format supported by the server.\n\n - If the certificate_authorities list in the certificate request\n message was non-empty, one of the certificates in the certificate\n chain SHOULD be issued by one of the listed CAs.\n\n\n\nDierks & Rescorla Standards Track [Page 56]\n\f\nRFC 5246 TLS August 2008\n\n\n - The certificates MUST be signed using an acceptable hash/\n signature algorithm pair, as described in Section 7.4.4. Note\n that this relaxes the constraints on certificate-signing\n algorithms found in prior versions of TLS.\n\n Note that, as with the server certificate, there are certificates\n that use algorithms/algorithm combinations that cannot be currently\n used with TLS.\n\n7.4.7. Client Key Exchange Message\n\n When this message will be sent:\n\n This message is always sent by the client. It MUST immediately\n follow the client certificate message, if it is sent. Otherwise,\n it MUST be the first message sent by the client after it receives\n the ServerHelloDone message.\n\n Meaning of this message:\n\n With this message, the premaster secret is set, either by direct\n transmission of the RSA-encrypted secret or by the transmission of\n Diffie-Hellman parameters that will allow each side to agree upon\n the same premaster secret.\n\n When the client is using an ephemeral Diffie-Hellman exponent,\n then this message contains the client's Diffie-Hellman public\n value. If the client is sending a certificate containing a static\n DH exponent (i.e., it is doing fixed_dh client authentication),\n then this message MUST be sent but MUST be empty.\n\n Structure of this message:\n\n The choice of messages depends on which key exchange method has\n been selected. See Section 7.4.3 for the KeyExchangeAlgorithm\n definition.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 57]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n select (KeyExchangeAlgorithm) {\n case rsa:\n EncryptedPreMasterSecret;\n case dhe_dss:\n case dhe_rsa:\n case dh_dss:\n case dh_rsa:\n case dh_anon:\n ClientDiffieHellmanPublic;\n } exchange_keys;\n } ClientKeyExchange;\n\n7.4.7.1. RSA-Encrypted Premaster Secret Message\n\n Meaning of this message:\n\n If RSA is being used for key agreement and authentication, the\n client generates a 48-byte premaster secret, encrypts it using the\n public key from the server's certificate, and sends the result in\n an encrypted premaster secret message. This structure is a\n variant of the ClientKeyExchange message and is not a message in\n itself.\n\n Structure of this message:\n\n struct {\n ProtocolVersion client_version;\n opaque random[46];\n } PreMasterSecret;\n\n client_version\n The latest (newest) version supported by the client. This is\n used to detect version rollback attacks.\n\n random\n 46 securely-generated random bytes.\n\n struct {\n public-key-encrypted PreMasterSecret pre_master_secret;\n } EncryptedPreMasterSecret;\n\n pre_master_secret\n This random value is generated by the client and is used to\n generate the master secret, as specified in Section 8.1.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 58]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: The version number in the PreMasterSecret is the version\n offered by the client in the ClientHello.client_version, not the\n version negotiated for the connection. This feature is designed to\n prevent rollback attacks. Unfortunately, some old implementations\n use the negotiated version instead, and therefore checking the\n version number may lead to failure to interoperate with such\n incorrect client implementations.\n\n Client implementations MUST always send the correct version number in\n PreMasterSecret. If ClientHello.client_version is TLS 1.1 or higher,\n server implementations MUST check the version number as described in\n the note below. If the version number is TLS 1.0 or earlier, server\n implementations SHOULD check the version number, but MAY have a\n configuration option to disable the check. Note that if the check\n fails, the PreMasterSecret SHOULD be randomized as described below.\n\n Note: Attacks discovered by Bleichenbacher [BLEI] and Klima et al.\n [KPR03] can be used to attack a TLS server that reveals whether a\n particular message, when decrypted, is properly PKCS#1 formatted,\n contains a valid PreMasterSecret structure, or has the correct\n version number.\n\n As described by Klima [KPR03], these vulnerabilities can be avoided\n by treating incorrectly formatted message blocks and/or mismatched\n version numbers in a manner indistinguishable from correctly\n formatted RSA blocks. In other words:\n\n 1. Generate a string R of 46 random bytes\n\n 2. Decrypt the message to recover the plaintext M\n\n 3. If the PKCS#1 padding is not correct, or the length of message\n M is not exactly 48 bytes:\n pre_master_secret = ClientHello.client_version || R\n else If ClientHello.client_version <= TLS 1.0, and version\n number check is explicitly disabled:\n pre_master_secret = M\n else:\n pre_master_secret = ClientHello.client_version || M[2..47]\n\n Note that explicitly constructing the pre_master_secret with the\n ClientHello.client_version produces an invalid master_secret if the\n client has sent the wrong version in the original pre_master_secret.\n\n An alternative approach is to treat a version number mismatch as a\n PKCS-1 formatting error and randomize the premaster secret\n completely:\n\n\n\n\nDierks & Rescorla Standards Track [Page 59]\n\f\nRFC 5246 TLS August 2008\n\n\n 1. Generate a string R of 48 random bytes\n\n 2. Decrypt the message to recover the plaintext M\n\n 3. If the PKCS#1 padding is not correct, or the length of message\n M is not exactly 48 bytes:\n pre_master_secret = R\n else If ClientHello.client_version <= TLS 1.0, and version\n number check is explicitly disabled:\n premaster secret = M\n else If M[0..1] != ClientHello.client_version:\n premaster secret = R\n else:\n premaster secret = M\n\n Although no practical attacks against this construction are known,\n Klima et al. [KPR03] describe some theoretical attacks, and therefore\n the first construction described is RECOMMENDED.\n\n In any case, a TLS server MUST NOT generate an alert if processing an\n RSA-encrypted premaster secret message fails, or the version number\n is not as expected. Instead, it MUST continue the handshake with a\n randomly generated premaster secret. It may be useful to log the\n real cause of failure for troubleshooting purposes; however, care\n must be taken to avoid leaking the information to an attacker\n (through, e.g., timing, log files, or other channels.)\n\n The RSAES-OAEP encryption scheme defined in [PKCS1] is more secure\n against the Bleichenbacher attack. However, for maximal\n compatibility with earlier versions of TLS, this specification uses\n the RSAES-PKCS1-v1_5 scheme. No variants of the Bleichenbacher\n attack are known to exist provided that the above recommendations are\n followed.\n\n Implementation note: Public-key-encrypted data is represented as an\n opaque vector <0..2^16-1> (see Section 4.7). Thus, the RSA-encrypted\n PreMasterSecret in a ClientKeyExchange is preceded by two length\n bytes. These bytes are redundant in the case of RSA because the\n EncryptedPreMasterSecret is the only data in the ClientKeyExchange\n and its length can therefore be unambiguously determined. The SSLv3\n specification was not clear about the encoding of public-key-\n encrypted data, and therefore many SSLv3 implementations do not\n include the length bytes -- they encode the RSA-encrypted data\n directly in the ClientKeyExchange message.\n\n This specification requires correct encoding of the\n EncryptedPreMasterSecret complete with length bytes. The resulting\n PDU is incompatible with many SSLv3 implementations. Implementors\n\n\n\nDierks & Rescorla Standards Track [Page 60]\n\f\nRFC 5246 TLS August 2008\n\n\n upgrading from SSLv3 MUST modify their implementations to generate\n and accept the correct encoding. Implementors who wish to be\n compatible with both SSLv3 and TLS should make their implementation's\n behavior dependent on the protocol version.\n\n Implementation note: It is now known that remote timing-based attacks\n on TLS are possible, at least when the client and server are on the\n same LAN. Accordingly, implementations that use static RSA keys MUST\n use RSA blinding or some other anti-timing technique, as described in\n [TIMING].\n\n7.4.7.2. Client Diffie-Hellman Public Value\n\n Meaning of this message:\n\n This structure conveys the client's Diffie-Hellman public value\n (Yc) if it was not already included in the client's certificate.\n The encoding used for Yc is determined by the enumerated\n PublicValueEncoding. This structure is a variant of the client\n key exchange message, and not a message in itself.\n\n Structure of this message:\n\n enum { implicit, explicit } PublicValueEncoding;\n\n implicit\n If the client has sent a certificate which contains a suitable\n Diffie-Hellman key (for fixed_dh client authentication), then\n Yc is implicit and does not need to be sent again. In this\n case, the client key exchange message will be sent, but it MUST\n be empty.\n\n explicit\n Yc needs to be sent.\n\n struct {\n select (PublicValueEncoding) {\n case implicit: struct { };\n case explicit: opaque dh_Yc<1..2^16-1>;\n } dh_public;\n } ClientDiffieHellmanPublic;\n\n dh_Yc\n The client's Diffie-Hellman public value (Yc).\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 61]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.8. Certificate Verify\n\n When this message will be sent:\n\n This message is used to provide explicit verification of a client\n certificate. This message is only sent following a client\n certificate that has signing capability (i.e., all certificates\n except those containing fixed Diffie-Hellman parameters). When\n sent, it MUST immediately follow the client key exchange message.\n\n Structure of this message:\n\n struct {\n digitally-signed struct {\n opaque handshake_messages[handshake_messages_length];\n }\n } CertificateVerify;\n\n Here handshake_messages refers to all handshake messages sent or\n received, starting at client hello and up to, but not including,\n this message, including the type and length fields of the\n handshake messages. This is the concatenation of all the\n Handshake structures (as defined in Section 7.4) exchanged thus\n far. Note that this requires both sides to either buffer the\n messages or compute running hashes for all potential hash\n algorithms up to the time of the CertificateVerify computation.\n Servers can minimize this computation cost by offering a\n restricted set of digest algorithms in the CertificateRequest\n message.\n\n The hash and signature algorithms used in the signature MUST be\n one of those present in the supported_signature_algorithms field\n of the CertificateRequest message. In addition, the hash and\n signature algorithms MUST be compatible with the key in the\n client's end-entity certificate. RSA keys MAY be used with any\n permitted hash algorithm, subject to restrictions in the\n certificate, if any.\n\n Because DSA signatures do not contain any secure indication of\n hash algorithm, there is a risk of hash substitution if multiple\n hashes may be used with any key. Currently, DSA [DSS] may only be\n used with SHA-1. Future revisions of DSS [DSS-3] are expected to\n allow the use of other digest algorithms with DSA, as well as\n guidance as to which digest algorithms should be used with each\n key size. In addition, future revisions of [PKIX] may specify\n mechanisms for certificates to indicate which digest algorithms\n are to be used with DSA.\n\n\n\n\nDierks & Rescorla Standards Track [Page 62]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.9. Finished\n\n When this message will be sent:\n\n A Finished message is always sent immediately after a change\n cipher spec message to verify that the key exchange and\n authentication processes were successful. It is essential that a\n change cipher spec message be received between the other handshake\n messages and the Finished message.\n\n Meaning of this message:\n\n The Finished message is the first one protected with the just\n negotiated algorithms, keys, and secrets. Recipients of Finished\n messages MUST verify that the contents are correct. Once a side\n has sent its Finished message and received and validated the\n Finished message from its peer, it may begin to send and receive\n application data over the connection.\n\n Structure of this message:\n\n struct {\n opaque verify_data[verify_data_length];\n } Finished;\n\n verify_data\n PRF(master_secret, finished_label, Hash(handshake_messages))\n [0..verify_data_length-1];\n\n finished_label\n For Finished messages sent by the client, the string\n \"client finished\". For Finished messages sent by the server,\n the string \"server finished\".\n\n Hash denotes a Hash of the handshake messages. For the PRF\n defined in Section 5, the Hash MUST be the Hash used as the basis\n for the PRF. Any cipher suite which defines a different PRF MUST\n also define the Hash to use in the Finished computation.\n\n In previous versions of TLS, the verify_data was always 12 octets\n long. In the current version of TLS, it depends on the cipher\n suite. Any cipher suite which does not explicitly specify\n verify_data_length has a verify_data_length equal to 12. This\n includes all existing cipher suites. Note that this\n representation has the same encoding as with previous versions.\n Future cipher suites MAY specify other lengths but such length\n MUST be at least 12 bytes.\n\n\n\n\nDierks & Rescorla Standards Track [Page 63]\n\f\nRFC 5246 TLS August 2008\n\n\n handshake_messages\n All of the data from all messages in this handshake (not\n including any HelloRequest messages) up to, but not including,\n this message. This is only data visible at the handshake layer\n and does not include record layer headers. This is the\n concatenation of all the Handshake structures as defined in\n Section 7.4, exchanged thus far.\n\n It is a fatal error if a Finished message is not preceded by a\n ChangeCipherSpec message at the appropriate point in the handshake.\n\n The value handshake_messages includes all handshake messages starting\n at ClientHello up to, but not including, this Finished message. This\n may be different from handshake_messages in Section 7.4.8 because it\n would include the CertificateVerify message (if sent). Also, the\n handshake_messages for the Finished message sent by the client will\n be different from that for the Finished message sent by the server,\n because the one that is sent second will include the prior one.\n\n Note: ChangeCipherSpec messages, alerts, and any other record types\n are not handshake messages and are not included in the hash\n computations. Also, HelloRequest messages are omitted from handshake\n hashes.\n\n8. Cryptographic Computations\n\n In order to begin connection protection, the TLS Record Protocol\n requires specification of a suite of algorithms, a master secret, and\n the client and server random values. The authentication, encryption,\n and MAC algorithms are determined by the cipher_suite selected by the\n server and revealed in the ServerHello message. The compression\n algorithm is negotiated in the hello messages, and the random values\n are exchanged in the hello messages. All that remains is to\n calculate the master secret.\n\n8.1. Computing the Master Secret\n\n For all key exchange methods, the same algorithm is used to convert\n the pre_master_secret into the master_secret. The pre_master_secret\n should be deleted from memory once the master_secret has been\n computed.\n\n master_secret = PRF(pre_master_secret, \"master secret\",\n ClientHello.random + ServerHello.random)\n [0..47];\n\n The master secret is always exactly 48 bytes in length. The length\n of the premaster secret will vary depending on key exchange method.\n\n\n\nDierks & Rescorla Standards Track [Page 64]\n\f\nRFC 5246 TLS August 2008\n\n\n8.1.1. RSA\n\n When RSA is used for server authentication and key exchange, a 48-\n byte pre_master_secret is generated by the client, encrypted under\n the server's public key, and sent to the server. The server uses its\n private key to decrypt the pre_master_secret. Both parties then\n convert the pre_master_secret into the master_secret, as specified\n above.\n\n8.1.2. Diffie-Hellman\n\n A conventional Diffie-Hellman computation is performed. The\n negotiated key (Z) is used as the pre_master_secret, and is converted\n into the master_secret, as specified above. Leading bytes of Z that\n contain all zero bits are stripped before it is used as the\n pre_master_secret.\n\n Note: Diffie-Hellman parameters are specified by the server and may\n be either ephemeral or contained within the server's certificate.\n\n9. Mandatory Cipher Suites\n\n In the absence of an application profile standard specifying\n otherwise, a TLS-compliant application MUST implement the cipher\n suite TLS_RSA_WITH_AES_128_CBC_SHA (see Appendix A.5 for the\n definition).\n\n10. Application Data Protocol\n\n Application data messages are carried by the record layer and are\n fragmented, compressed, and encrypted based on the current connection\n state. The messages are treated as transparent data to the record\n layer.\n\n11. Security Considerations\n\n Security issues are discussed throughout this memo, especially in\n Appendices D, E, and F.\n\n12. IANA Considerations\n\n This document uses several registries that were originally created in\n [TLS1.1]. IANA has updated these to reference this document. The\n registries and their allocation policies (unchanged from [TLS1.1])\n are listed below.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 65]\n\f\nRFC 5246 TLS August 2008\n\n\n - TLS ClientCertificateType Identifiers Registry: Future values in\n the range 0-63 (decimal) inclusive are assigned via Standards\n Action [RFC2434]. Values in the range 64-223 (decimal) inclusive\n are assigned via Specification Required [RFC2434]. Values from\n 224-255 (decimal) inclusive are reserved for Private Use\n [RFC2434].\n\n - TLS Cipher Suite Registry: Future values with the first byte in\n the range 0-191 (decimal) inclusive are assigned via Standards\n Action [RFC2434]. Values with the first byte in the range 192-254\n (decimal) are assigned via Specification Required [RFC2434].\n Values with the first byte 255 (decimal) are reserved for Private\n Use [RFC2434].\n\n - This document defines several new HMAC-SHA256-based cipher suites,\n whose values (in Appendix A.5) have been allocated from the TLS\n Cipher Suite registry.\n\n - TLS ContentType Registry: Future values are allocated via\n Standards Action [RFC2434].\n\n - TLS Alert Registry: Future values are allocated via Standards\n Action [RFC2434].\n\n - TLS HandshakeType Registry: Future values are allocated via\n Standards Action [RFC2434].\n\n This document also uses a registry originally created in [RFC4366].\n IANA has updated it to reference this document. The registry and its\n allocation policy (unchanged from [RFC4366]) is listed below:\n\n - TLS ExtensionType Registry: Future values are allocated via IETF\n Consensus [RFC2434]. IANA has updated this registry to include\n the signature_algorithms extension and its corresponding value\n (see Section 7.4.1.4).\n\n In addition, this document defines two new registries to be\n maintained by IANA:\n\n - TLS SignatureAlgorithm Registry: The registry has been initially\n populated with the values described in Section 7.4.1.4.1. Future\n values in the range 0-63 (decimal) inclusive are assigned via\n Standards Action [RFC2434]. Values in the range 64-223 (decimal)\n inclusive are assigned via Specification Required [RFC2434].\n Values from 224-255 (decimal) inclusive are reserved for Private\n Use [RFC2434].\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 66]\n\f\nRFC 5246 TLS August 2008\n\n\n - TLS HashAlgorithm Registry: The registry has been initially\n populated with the values described in Section 7.4.1.4.1. Future\n values in the range 0-63 (decimal) inclusive are assigned via\n Standards Action [RFC2434]. Values in the range 64-223 (decimal)\n inclusive are assigned via Specification Required [RFC2434].\n Values from 224-255 (decimal) inclusive are reserved for Private\n Use [RFC2434].\n\n This document also uses the TLS Compression Method Identifiers\n Registry, defined in [RFC3749]. IANA has allocated value 0 for\n the \"null\" compression method.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 67]\n\f\nRFC 5246 TLS August 2008\n\n\nAppendix A. Protocol Data Structures and Constant Values\n\n This section describes protocol types and constants.\n\nA.1. Record Layer\n\n struct {\n uint8 major;\n uint8 minor;\n } ProtocolVersion;\n\n ProtocolVersion version = { 3, 3 }; /* TLS v1.2*/\n\n enum {\n change_cipher_spec(20), alert(21), handshake(22),\n application_data(23), (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSCompressed.length];\n } TLSCompressed;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n select (SecurityParameters.cipher_type) {\n case stream: GenericStreamCipher;\n case block: GenericBlockCipher;\n case aead: GenericAEADCipher;\n } fragment;\n } TLSCiphertext;\n\n stream-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n } GenericStreamCipher;\n\n\n\n\nDierks & Rescorla Standards Track [Page 68]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n opaque IV[SecurityParameters.record_iv_length];\n block-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n uint8 padding[GenericBlockCipher.padding_length];\n uint8 padding_length;\n };\n } GenericBlockCipher;\n\n struct {\n opaque nonce_explicit[SecurityParameters.record_iv_length];\n aead-ciphered struct {\n opaque content[TLSCompressed.length];\n };\n } GenericAEADCipher;\n\nA.2. Change Cipher Specs Message\n\n struct {\n enum { change_cipher_spec(1), (255) } type;\n } ChangeCipherSpec;\n\nA.3. Alert Messages\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction_RESERVED(60),\n protocol_version(70),\n\n\n\nDierks & Rescorla Standards Track [Page 69]\n\f\nRFC 5246 TLS August 2008\n\n\n insufficient_security(71),\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110), /* new */\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\nA.4. Handshake Protocol\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20)\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type;\n uint24 length;\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n } body;\n } Handshake;\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 70]\n\f\nRFC 5246 TLS August 2008\n\n\nA.4.1. Hello Messages\n\n struct { } HelloRequest;\n\n struct {\n uint32 gmt_unix_time;\n opaque random_bytes[28];\n } Random;\n\n opaque SessionID<0..32>;\n\n uint8 CipherSuite[2];\n\n enum { null(0), (255) } CompressionMethod;\n\n struct {\n ProtocolVersion client_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suites<2..2^16-2>;\n CompressionMethod compression_methods<1..2^8-1>;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ClientHello;\n\n struct {\n ProtocolVersion server_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ServerHello;\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n\n\n\nDierks & Rescorla Standards Track [Page 71]\n\f\nRFC 5246 TLS August 2008\n\n\n enum {\n signature_algorithms(13), (65535)\n } ExtensionType;\n\n enum{\n none(0), md5(1), sha1(2), sha224(3), sha256(4), sha384(5),\n sha512(6), (255)\n } HashAlgorithm;\n enum {\n anonymous(0), rsa(1), dsa(2), ecdsa(3), (255)\n } SignatureAlgorithm;\n\n struct {\n HashAlgorithm hash;\n SignatureAlgorithm signature;\n } SignatureAndHashAlgorithm;\n\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2..2^16-1>;\n\nA.4.2. Server Authentication and Key Exchange Messages\n\n opaque ASN.1Cert<2^24-1>;\n\n struct {\n ASN.1Cert certificate_list<0..2^24-1>;\n } Certificate;\n\n enum { dhe_dss, dhe_rsa, dh_anon, rsa,dh_dss, dh_rsa\n /* may be extended, e.g., for ECDH -- see [TLSECC] */\n } KeyExchangeAlgorithm;\n\n struct {\n opaque dh_p<1..2^16-1>;\n opaque dh_g<1..2^16-1>;\n opaque dh_Ys<1..2^16-1>;\n } ServerDHParams; /* Ephemeral DH parameters */\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 72]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n select (KeyExchangeAlgorithm) {\n case dh_anon:\n ServerDHParams params;\n case dhe_dss:\n case dhe_rsa:\n ServerDHParams params;\n digitally-signed struct {\n opaque client_random[32];\n opaque server_random[32];\n ServerDHParams params;\n } signed_params;\n case rsa:\n case dh_dss:\n case dh_rsa:\n struct {} ;\n /* message is omitted for rsa, dh_dss, and dh_rsa */\n /* may be extended, e.g., for ECDH -- see [TLSECC] */\n } ServerKeyExchange;\n\n enum {\n rsa_sign(1), dss_sign(2), rsa_fixed_dh(3), dss_fixed_dh(4),\n rsa_ephemeral_dh_RESERVED(5), dss_ephemeral_dh_RESERVED(6),\n fortezza_dms_RESERVED(20),\n (255)\n } ClientCertificateType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n ClientCertificateType certificate_types<1..2^8-1>;\n DistinguishedName certificate_authorities<0..2^16-1>;\n } CertificateRequest;\n\n struct { } ServerHelloDone;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 73]\n\f\nRFC 5246 TLS August 2008\n\n\nA.4.3. Client Authentication and Key Exchange Messages\n\n struct {\n select (KeyExchangeAlgorithm) {\n case rsa:\n EncryptedPreMasterSecret;\n case dhe_dss:\n case dhe_rsa:\n case dh_dss:\n case dh_rsa:\n case dh_anon:\n ClientDiffieHellmanPublic;\n } exchange_keys;\n } ClientKeyExchange;\n\n struct {\n ProtocolVersion client_version;\n opaque random[46];\n } PreMasterSecret;\n\n struct {\n public-key-encrypted PreMasterSecret pre_master_secret;\n } EncryptedPreMasterSecret;\n\n enum { implicit, explicit } PublicValueEncoding;\n\n struct {\n select (PublicValueEncoding) {\n case implicit: struct {};\n case explicit: opaque DH_Yc<1..2^16-1>;\n } dh_public;\n } ClientDiffieHellmanPublic;\n\n struct {\n digitally-signed struct {\n opaque handshake_messages[handshake_messages_length];\n }\n } CertificateVerify;\n\nA.4.4. Handshake Finalization Message\n\n struct {\n opaque verify_data[verify_data_length];\n } Finished;\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 74]\n\f\nRFC 5246 TLS August 2008\n\n\nA.5. The Cipher Suite\n\n The following values define the cipher suite codes used in the\n ClientHello and ServerHello messages.\n\n A cipher suite defines a cipher specification supported in TLS\n Version 1.2.\n\n TLS_NULL_WITH_NULL_NULL is specified and is the initial state of a\n TLS connection during the first handshake on that channel, but MUST\n NOT be negotiated, as it provides no more protection than an\n unsecured connection.\n\n CipherSuite TLS_NULL_WITH_NULL_NULL = { 0x00,0x00 };\n\n The following CipherSuite definitions require that the server provide\n an RSA certificate that can be used for key exchange. The server may\n request any signature-capable certificate in the certificate request\n message.\n\n CipherSuite TLS_RSA_WITH_NULL_MD5 = { 0x00,0x01 };\n CipherSuite TLS_RSA_WITH_NULL_SHA = { 0x00,0x02 };\n CipherSuite TLS_RSA_WITH_NULL_SHA256 = { 0x00,0x3B };\n CipherSuite TLS_RSA_WITH_RC4_128_MD5 = { 0x00,0x04 };\n CipherSuite TLS_RSA_WITH_RC4_128_SHA = { 0x00,0x05 };\n CipherSuite TLS_RSA_WITH_3DES_EDE_CBC_SHA = { 0x00,0x0A };\n CipherSuite TLS_RSA_WITH_AES_128_CBC_SHA = { 0x00,0x2F };\n CipherSuite TLS_RSA_WITH_AES_256_CBC_SHA = { 0x00,0x35 };\n CipherSuite TLS_RSA_WITH_AES_128_CBC_SHA256 = { 0x00,0x3C };\n CipherSuite TLS_RSA_WITH_AES_256_CBC_SHA256 = { 0x00,0x3D };\n\n The following cipher suite definitions are used for server-\n authenticated (and optionally client-authenticated) Diffie-Hellman.\n DH denotes cipher suites in which the server's certificate contains\n the Diffie-Hellman parameters signed by the certificate authority\n (CA). DHE denotes ephemeral Diffie-Hellman, where the Diffie-Hellman\n parameters are signed by a signature-capable certificate, which has\n been signed by the CA. The signing algorithm used by the server is\n specified after the DHE component of the CipherSuite name. The\n server can request any signature-capable certificate from the client\n for client authentication, or it may request a Diffie-Hellman\n certificate. Any Diffie-Hellman certificate provided by the client\n must use the parameters (group and generator) described by the\n server.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 75]\n\f\nRFC 5246 TLS August 2008\n\n\n CipherSuite TLS_DH_DSS_WITH_3DES_EDE_CBC_SHA = { 0x00,0x0D };\n CipherSuite TLS_DH_RSA_WITH_3DES_EDE_CBC_SHA = { 0x00,0x10 };\n CipherSuite TLS_DHE_DSS_WITH_3DES_EDE_CBC_SHA = { 0x00,0x13 };\n CipherSuite TLS_DHE_RSA_WITH_3DES_EDE_CBC_SHA = { 0x00,0x16 };\n CipherSuite TLS_DH_DSS_WITH_AES_128_CBC_SHA = { 0x00,0x30 };\n CipherSuite TLS_DH_RSA_WITH_AES_128_CBC_SHA = { 0x00,0x31 };\n CipherSuite TLS_DHE_DSS_WITH_AES_128_CBC_SHA = { 0x00,0x32 };\n CipherSuite TLS_DHE_RSA_WITH_AES_128_CBC_SHA = { 0x00,0x33 };\n CipherSuite TLS_DH_DSS_WITH_AES_256_CBC_SHA = { 0x00,0x36 };\n CipherSuite TLS_DH_RSA_WITH_AES_256_CBC_SHA = { 0x00,0x37 };\n CipherSuite TLS_DHE_DSS_WITH_AES_256_CBC_SHA = { 0x00,0x38 };\n CipherSuite TLS_DHE_RSA_WITH_AES_256_CBC_SHA = { 0x00,0x39 };\n CipherSuite TLS_DH_DSS_WITH_AES_128_CBC_SHA256 = { 0x00,0x3E };\n CipherSuite TLS_DH_RSA_WITH_AES_128_CBC_SHA256 = { 0x00,0x3F };\n CipherSuite TLS_DHE_DSS_WITH_AES_128_CBC_SHA256 = { 0x00,0x40 };\n CipherSuite TLS_DHE_RSA_WITH_AES_128_CBC_SHA256 = { 0x00,0x67 };\n CipherSuite TLS_DH_DSS_WITH_AES_256_CBC_SHA256 = { 0x00,0x68 };\n CipherSuite TLS_DH_RSA_WITH_AES_256_CBC_SHA256 = { 0x00,0x69 };\n CipherSuite TLS_DHE_DSS_WITH_AES_256_CBC_SHA256 = { 0x00,0x6A };\n CipherSuite TLS_DHE_RSA_WITH_AES_256_CBC_SHA256 = { 0x00,0x6B };\n\n The following cipher suites are used for completely anonymous\n Diffie-Hellman communications in which neither party is\n authenticated. Note that this mode is vulnerable to man-in-the-\n middle attacks. Using this mode therefore is of limited use: These\n cipher suites MUST NOT be used by TLS 1.2 implementations unless the\n application layer has specifically requested to allow anonymous key\n exchange. (Anonymous key exchange may sometimes be acceptable, for\n example, to support opportunistic encryption when no set-up for\n authentication is in place, or when TLS is used as part of more\n complex security protocols that have other means to ensure\n authentication.)\n\n CipherSuite TLS_DH_anon_WITH_RC4_128_MD5 = { 0x00,0x18 };\n CipherSuite TLS_DH_anon_WITH_3DES_EDE_CBC_SHA = { 0x00,0x1B };\n CipherSuite TLS_DH_anon_WITH_AES_128_CBC_SHA = { 0x00,0x34 };\n CipherSuite TLS_DH_anon_WITH_AES_256_CBC_SHA = { 0x00,0x3A };\n CipherSuite TLS_DH_anon_WITH_AES_128_CBC_SHA256 = { 0x00,0x6C };\n CipherSuite TLS_DH_anon_WITH_AES_256_CBC_SHA256 = { 0x00,0x6D };\n\n Note that using non-anonymous key exchange without actually verifying\n the key exchange is essentially equivalent to anonymous key exchange,\n and the same precautions apply. While non-anonymous key exchange\n will generally involve a higher computational and communicational\n cost than anonymous key exchange, it may be in the interest of\n interoperability not to disable non-anonymous key exchange when the\n application layer is allowing anonymous key exchange.\n\n\n\n\nDierks & Rescorla Standards Track [Page 76]\n\f\nRFC 5246 TLS August 2008\n\n\n New cipher suite values have been assigned by IANA as described in\n Section 12.\n\n Note: The cipher suite values { 0x00, 0x1C } and { 0x00, 0x1D } are\n reserved to avoid collision with Fortezza-based cipher suites in\n SSL 3.\n\nA.6. The Security Parameters\n\n These security parameters are determined by the TLS Handshake\n Protocol and provided as parameters to the TLS record layer in order\n to initialize a connection state. SecurityParameters includes:\n\n enum { null(0), (255) } CompressionMethod;\n\n enum { server, client } ConnectionEnd;\n\n enum { tls_prf_sha256 } PRFAlgorithm;\n\n enum { null, rc4, 3des, aes } BulkCipherAlgorithm;\n\n enum { stream, block, aead } CipherType;\n\n enum { null, hmac_md5, hmac_sha1, hmac_sha256, hmac_sha384,\n hmac_sha512} MACAlgorithm;\n\n /* Other values may be added to the algorithms specified in\n CompressionMethod, PRFAlgorithm, BulkCipherAlgorithm, and\n MACAlgorithm. */\n\n struct {\n ConnectionEnd entity;\n PRFAlgorithm prf_algorithm;\n BulkCipherAlgorithm bulk_cipher_algorithm;\n CipherType cipher_type;\n uint8 enc_key_length;\n uint8 block_length;\n uint8 fixed_iv_length;\n uint8 record_iv_length;\n MACAlgorithm mac_algorithm;\n uint8 mac_length;\n uint8 mac_key_length;\n CompressionMethod compression_algorithm;\n opaque master_secret[48];\n opaque client_random[32];\n opaque server_random[32];\n } SecurityParameters;\n\n\n\n\nDierks & Rescorla Standards Track [Page 77]\n\f\nRFC 5246 TLS August 2008\n\n\nA.7. Changes to RFC 4492\n\n RFC 4492 [TLSECC] adds Elliptic Curve cipher suites to TLS. This\n document changes some of the structures used in that document. This\n section details the required changes for implementors of both RFC\n 4492 and TLS 1.2. Implementors of TLS 1.2 who are not implementing\n RFC 4492 do not need to read this section.\n\n This document adds a \"signature_algorithm\" field to the digitally-\n signed element in order to identify the signature and digest\n algorithms used to create a signature. This change applies to\n digital signatures formed using ECDSA as well, thus allowing ECDSA\n signatures to be used with digest algorithms other than SHA-1,\n provided such use is compatible with the certificate and any\n restrictions imposed by future revisions of [PKIX].\n\n As described in Sections 7.4.2 and 7.4.6, the restrictions on the\n signature algorithms used to sign certificates are no longer tied to\n the cipher suite (when used by the server) or the\n ClientCertificateType (when used by the client). Thus, the\n restrictions on the algorithm used to sign certificates specified in\n Sections 2 and 3 of RFC 4492 are also relaxed. As in this document,\n the restrictions on the keys in the end-entity certificate remain.\n\nAppendix B. Glossary\n\n Advanced Encryption Standard (AES)\n AES [AES] is a widely used symmetric encryption algorithm. AES is\n a block cipher with a 128-, 192-, or 256-bit keys and a 16-byte\n block size. TLS currently only supports the 128- and 256-bit key\n sizes.\n\n application protocol\n An application protocol is a protocol that normally layers\n directly on top of the transport layer (e.g., TCP/IP). Examples\n include HTTP, TELNET, FTP, and SMTP.\n\n asymmetric cipher\n See public key cryptography.\n\n authenticated encryption with additional data (AEAD)\n A symmetric encryption algorithm that simultaneously provides\n confidentiality and message integrity.\n\n authentication\n Authentication is the ability of one entity to determine the\n identity of another entity.\n\n\n\n\nDierks & Rescorla Standards Track [Page 78]\n\f\nRFC 5246 TLS August 2008\n\n\n block cipher\n A block cipher is an algorithm that operates on plaintext in\n groups of bits, called blocks. 64 bits was, and 128 bits is, a\n common block size.\n\n bulk cipher\n A symmetric encryption algorithm used to encrypt large quantities\n of data.\n\n cipher block chaining (CBC)\n CBC is a mode in which every plaintext block encrypted with a\n block cipher is first exclusive-ORed with the previous ciphertext\n block (or, in the case of the first block, with the initialization\n vector). For decryption, every block is first decrypted, then\n exclusive-ORed with the previous ciphertext block (or IV).\n\n certificate\n As part of the X.509 protocol (a.k.a. ISO Authentication\n framework), certificates are assigned by a trusted Certificate\n Authority and provide a strong binding between a party's identity\n or some other attributes and its public key.\n\n client\n The application entity that initiates a TLS connection to a\n server. This may or may not imply that the client initiated the\n underlying transport connection. The primary operational\n difference between the server and client is that the server is\n generally authenticated, while the client is only optionally\n authenticated.\n\n client write key\n The key used to encrypt data written by the client.\n\n client write MAC key\n The secret data used to authenticate data written by the client.\n\n connection\n A connection is a transport (in the OSI layering model definition)\n that provides a suitable type of service. For TLS, such\n connections are peer-to-peer relationships. The connections are\n transient. Every connection is associated with one session.\n\n Data Encryption Standard\n DES [DES] still is a very widely used symmetric encryption\n algorithm although it is considered as rather weak now. DES is a\n block cipher with a 56-bit key and an 8-byte block size. Note\n that in TLS, for key generation purposes, DES is treated as having\n an 8-byte key length (64 bits), but it still only provides 56 bits\n\n\n\nDierks & Rescorla Standards Track [Page 79]\n\f\nRFC 5246 TLS August 2008\n\n\n of protection. (The low bit of each key byte is presumed to be\n set to produce odd parity in that key byte.) DES can also be\n operated in a mode [3DES] where three independent keys and three\n encryptions are used for each block of data; this uses 168 bits of\n key (24 bytes in the TLS key generation method) and provides the\n equivalent of 112 bits of security.\n\n Digital Signature Standard (DSS)\n A standard for digital signing, including the Digital Signing\n Algorithm, approved by the National Institute of Standards and\n Technology, defined in NIST FIPS PUB 186-2, \"Digital Signature\n Standard\", published January 2000 by the U.S. Department of\n Commerce [DSS]. A significant update [DSS-3] has been drafted and\n was published in March 2006.\n\n digital signatures\n Digital signatures utilize public key cryptography and one-way\n hash functions to produce a signature of the data that can be\n authenticated, and is difficult to forge or repudiate.\n\n handshake An initial negotiation between client and server that\n establishes the parameters of their transactions.\n\n Initialization Vector (IV)\n When a block cipher is used in CBC mode, the initialization vector\n is exclusive-ORed with the first plaintext block prior to\n encryption.\n\n Message Authentication Code (MAC)\n A Message Authentication Code is a one-way hash computed from a\n message and some secret data. It is difficult to forge without\n knowing the secret data. Its purpose is to detect if the message\n has been altered.\n\n master secret\n Secure secret data used for generating encryption keys, MAC\n secrets, and IVs.\n\n MD5\n MD5 [MD5] is a hashing function that converts an arbitrarily long\n data stream into a hash of fixed size (16 bytes). Due to\n significant progress in cryptanalysis, at the time of publication\n of this document, MD5 no longer can be considered a 'secure'\n hashing function.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 80]\n\f\nRFC 5246 TLS August 2008\n\n\n public key cryptography\n A class of cryptographic techniques employing two-key ciphers.\n Messages encrypted with the public key can only be decrypted with\n the associated private key. Conversely, messages signed with the\n private key can be verified with the public key.\n\n one-way hash function\n A one-way transformation that converts an arbitrary amount of data\n into a fixed-length hash. It is computationally hard to reverse\n the transformation or to find collisions. MD5 and SHA are\n examples of one-way hash functions.\n\n RC4\n A stream cipher invented by Ron Rivest. A compatible cipher is\n described in [SCH].\n\n RSA\n A very widely used public key algorithm that can be used for\n either encryption or digital signing. [RSA]\n\n server\n The server is the application entity that responds to requests for\n connections from clients. See also \"client\".\n\n session\n A TLS session is an association between a client and a server.\n Sessions are created by the handshake protocol. Sessions define a\n set of cryptographic security parameters that can be shared among\n multiple connections. Sessions are used to avoid the expensive\n negotiation of new security parameters for each connection.\n\n session identifier\n A session identifier is a value generated by a server that\n identifies a particular session.\n\n server write key\n The key used to encrypt data written by the server.\n\n server write MAC key\n The secret data used to authenticate data written by the server.\n\n SHA\n The Secure Hash Algorithm [SHS] is defined in FIPS PUB 180-2. It\n produces a 20-byte output. Note that all references to SHA\n (without a numerical suffix) actually use the modified SHA-1\n algorithm.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 81]\n\f\nRFC 5246 TLS August 2008\n\n\n SHA-256\n The 256-bit Secure Hash Algorithm is defined in FIPS PUB 180-2.\n It produces a 32-byte output.\n\n SSL\n Netscape's Secure Socket Layer protocol [SSL3]. TLS is based on\n SSL Version 3.0.\n\n stream cipher\n An encryption algorithm that converts a key into a\n cryptographically strong keystream, which is then exclusive-ORed\n with the plaintext.\n\n symmetric cipher\n See bulk cipher.\n\n Transport Layer Security (TLS)\n This protocol; also, the Transport Layer Security working group of\n the Internet Engineering Task Force (IETF). See \"Working Group\n Information\" at the end of this document (see page 99).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 82]\n\f\nRFC 5246 TLS August 2008\n\n\nAppendix C. Cipher Suite Definitions\n\nCipher Suite Key Cipher Mac\n Exchange\n\nTLS_NULL_WITH_NULL_NULL NULL NULL NULL\nTLS_RSA_WITH_NULL_MD5 RSA NULL MD5\nTLS_RSA_WITH_NULL_SHA RSA NULL SHA\nTLS_RSA_WITH_NULL_SHA256 RSA NULL SHA256\nTLS_RSA_WITH_RC4_128_MD5 RSA RC4_128 MD5\nTLS_RSA_WITH_RC4_128_SHA RSA RC4_128 SHA\nTLS_RSA_WITH_3DES_EDE_CBC_SHA RSA 3DES_EDE_CBC SHA\nTLS_RSA_WITH_AES_128_CBC_SHA RSA AES_128_CBC SHA\nTLS_RSA_WITH_AES_256_CBC_SHA RSA AES_256_CBC SHA\nTLS_RSA_WITH_AES_128_CBC_SHA256 RSA AES_128_CBC SHA256\nTLS_RSA_WITH_AES_256_CBC_SHA256 RSA AES_256_CBC SHA256\nTLS_DH_DSS_WITH_3DES_EDE_CBC_SHA DH_DSS 3DES_EDE_CBC SHA\nTLS_DH_RSA_WITH_3DES_EDE_CBC_SHA DH_RSA 3DES_EDE_CBC SHA\nTLS_DHE_DSS_WITH_3DES_EDE_CBC_SHA DHE_DSS 3DES_EDE_CBC SHA\nTLS_DHE_RSA_WITH_3DES_EDE_CBC_SHA DHE_RSA 3DES_EDE_CBC SHA\nTLS_DH_anon_WITH_RC4_128_MD5 DH_anon RC4_128 MD5\nTLS_DH_anon_WITH_3DES_EDE_CBC_SHA DH_anon 3DES_EDE_CBC SHA\nTLS_DH_DSS_WITH_AES_128_CBC_SHA DH_DSS AES_128_CBC SHA\nTLS_DH_RSA_WITH_AES_128_CBC_SHA DH_RSA AES_128_CBC SHA\nTLS_DHE_DSS_WITH_AES_128_CBC_SHA DHE_DSS AES_128_CBC SHA\nTLS_DHE_RSA_WITH_AES_128_CBC_SHA DHE_RSA AES_128_CBC SHA\nTLS_DH_anon_WITH_AES_128_CBC_SHA DH_anon AES_128_CBC SHA\nTLS_DH_DSS_WITH_AES_256_CBC_SHA DH_DSS AES_256_CBC SHA\nTLS_DH_RSA_WITH_AES_256_CBC_SHA DH_RSA AES_256_CBC SHA\nTLS_DHE_DSS_WITH_AES_256_CBC_SHA DHE_DSS AES_256_CBC SHA\nTLS_DHE_RSA_WITH_AES_256_CBC_SHA DHE_RSA AES_256_CBC SHA\nTLS_DH_anon_WITH_AES_256_CBC_SHA DH_anon AES_256_CBC SHA\nTLS_DH_DSS_WITH_AES_128_CBC_SHA256 DH_DSS AES_128_CBC SHA256\nTLS_DH_RSA_WITH_AES_128_CBC_SHA256 DH_RSA AES_128_CBC SHA256\nTLS_DHE_DSS_WITH_AES_128_CBC_SHA256 DHE_DSS AES_128_CBC SHA256\nTLS_DHE_RSA_WITH_AES_128_CBC_SHA256 DHE_RSA AES_128_CBC SHA256\nTLS_DH_anon_WITH_AES_128_CBC_SHA256 DH_anon AES_128_CBC SHA256\nTLS_DH_DSS_WITH_AES_256_CBC_SHA256 DH_DSS AES_256_CBC SHA256\nTLS_DH_RSA_WITH_AES_256_CBC_SHA256 DH_RSA AES_256_CBC SHA256\nTLS_DHE_DSS_WITH_AES_256_CBC_SHA256 DHE_DSS AES_256_CBC SHA256\nTLS_DHE_RSA_WITH_AES_256_CBC_SHA256 DHE_RSA AES_256_CBC SHA256\nTLS_DH_anon_WITH_AES_256_CBC_SHA256 DH_anon AES_256_CBC SHA256\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 83]\n\f\nRFC 5246 TLS August 2008\n\n\n Key IV Block\nCipher Type Material Size Size\n------------ ------ -------- ---- -----\nNULL Stream 0 0 N/A\nRC4_128 Stream 16 0 N/A\n3DES_EDE_CBC Block 24 8 8\nAES_128_CBC Block 16 16 16\nAES_256_CBC Block 32 16 16\n\n\nMAC Algorithm mac_length mac_key_length\n-------- ----------- ---------- --------------\nNULL N/A 0 0\nMD5 HMAC-MD5 16 16\nSHA HMAC-SHA1 20 20\nSHA256 HMAC-SHA256 32 32\n\n Type\n Indicates whether this is a stream cipher or a block cipher\n running in CBC mode.\n\n Key Material\n The number of bytes from the key_block that are used for\n generating the write keys.\n\n IV Size\n The amount of data needed to be generated for the initialization\n vector. Zero for stream ciphers; equal to the block size for\n block ciphers (this is equal to\n SecurityParameters.record_iv_length).\n\n Block Size\n The amount of data a block cipher enciphers in one chunk; a block\n cipher running in CBC mode can only encrypt an even multiple of\n its block size.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 84]\n\f\nRFC 5246 TLS August 2008\n\n\nAppendix D. Implementation Notes\n\n The TLS protocol cannot prevent many common security mistakes. This\n section provides several recommendations to assist implementors.\n\nD.1. Random Number Generation and Seeding\n\n TLS requires a cryptographically secure pseudorandom number generator\n (PRNG). Care must be taken in designing and seeding PRNGs. PRNGs\n based on secure hash operations, most notably SHA-1, are acceptable,\n but cannot provide more security than the size of the random number\n generator state.\n\n To estimate the amount of seed material being produced, add the\n number of bits of unpredictable information in each seed byte. For\n example, keystroke timing values taken from a PC compatible's 18.2 Hz\n timer provide 1 or 2 secure bits each, even though the total size of\n the counter value is 16 bits or more. Seeding a 128-bit PRNG would\n thus require approximately 100 such timer values.\n\n [RANDOM] provides guidance on the generation of random values.\n\nD.2. Certificates and Authentication\n\n Implementations are responsible for verifying the integrity of\n certificates and should generally support certificate revocation\n messages. Certificates should always be verified to ensure proper\n signing by a trusted Certificate Authority (CA). The selection and\n addition of trusted CAs should be done very carefully. Users should\n be able to view information about the certificate and root CA.\n\nD.3. Cipher Suites\n\n TLS supports a range of key sizes and security levels, including some\n that provide no or minimal security. A proper implementation will\n probably not support many cipher suites. For instance, anonymous\n Diffie-Hellman is strongly discouraged because it cannot prevent man-\n in-the-middle attacks. Applications should also enforce minimum and\n maximum key sizes. For example, certificate chains containing 512-\n bit RSA keys or signatures are not appropriate for high-security\n applications.\n\nD.4. Implementation Pitfalls\n\n Implementation experience has shown that certain parts of earlier TLS\n specifications are not easy to understand, and have been a source of\n interoperability and security problems. Many of these areas have\n\n\n\n\nDierks & Rescorla Standards Track [Page 85]\n\f\nRFC 5246 TLS August 2008\n\n\n been clarified in this document, but this appendix contains a short\n list of the most important things that require special attention from\n implementors.\n\n TLS protocol issues:\n\n - Do you correctly handle handshake messages that are fragmented to\n multiple TLS records (see Section 6.2.1)? Including corner cases\n like a ClientHello that is split to several small fragments? Do\n you fragment handshake messages that exceed the maximum fragment\n size? In particular, the certificate and certificate request\n handshake messages can be large enough to require fragmentation.\n\n - Do you ignore the TLS record layer version number in all TLS\n records before ServerHello (see Appendix E.1)?\n\n - Do you handle TLS extensions in ClientHello correctly, including\n omitting the extensions field completely?\n\n - Do you support renegotiation, both client and server initiated?\n While renegotiation is an optional feature, supporting it is\n highly recommended.\n\n - When the server has requested a client certificate, but no\n suitable certificate is available, do you correctly send an empty\n Certificate message, instead of omitting the whole message (see\n Section 7.4.6)?\n\n Cryptographic details:\n\n - In the RSA-encrypted Premaster Secret, do you correctly send and\n verify the version number? When an error is encountered, do you\n continue the handshake to avoid the Bleichenbacher attack (see\n Section 7.4.7.1)?\n\n - What countermeasures do you use to prevent timing attacks against\n RSA decryption and signing operations (see Section 7.4.7.1)?\n\n - When verifying RSA signatures, do you accept both NULL and missing\n parameters (see Section 4.7)? Do you verify that the RSA padding\n doesn't have additional data after the hash value? [FI06]\n\n - When using Diffie-Hellman key exchange, do you correctly strip\n leading zero bytes from the negotiated key (see Section 8.1.2)?\n\n - Does your TLS client check that the Diffie-Hellman parameters sent\n by the server are acceptable (see Section F.1.1.3)?\n\n\n\n\nDierks & Rescorla Standards Track [Page 86]\n\f\nRFC 5246 TLS August 2008\n\n\n - How do you generate unpredictable IVs for CBC mode ciphers (see\n Section 6.2.3.2)?\n\n - Do you accept long CBC mode padding (up to 255 bytes; see Section\n 6.2.3.2)?\n\n - How do you address CBC mode timing attacks (Section 6.2.3.2)?\n\n - Do you use a strong and, most importantly, properly seeded random\n number generator (see Appendix D.1) for generating the premaster\n secret (for RSA key exchange), Diffie-Hellman private values, the\n DSA \"k\" parameter, and other security-critical values?\n\nAppendix E. Backward Compatibility\n\nE.1. Compatibility with TLS 1.0/1.1 and SSL 3.0\n\n Since there are various versions of TLS (1.0, 1.1, 1.2, and any\n future versions) and SSL (2.0 and 3.0), means are needed to negotiate\n the specific protocol version to use. The TLS protocol provides a\n built-in mechanism for version negotiation so as not to bother other\n protocol components with the complexities of version selection.\n\n TLS versions 1.0, 1.1, and 1.2, and SSL 3.0 are very similar, and use\n compatible ClientHello messages; thus, supporting all of them is\n relatively easy. Similarly, servers can easily handle clients trying\n to use future versions of TLS as long as the ClientHello format\n remains compatible, and the client supports the highest protocol\n version available in the server.\n\n A TLS 1.2 client who wishes to negotiate with such older servers will\n send a normal TLS 1.2 ClientHello, containing { 3, 3 } (TLS 1.2) in\n ClientHello.client_version. If the server does not support this\n version, it will respond with a ServerHello containing an older\n version number. If the client agrees to use this version, the\n negotiation will proceed as appropriate for the negotiated protocol.\n\n If the version chosen by the server is not supported by the client\n (or not acceptable), the client MUST send a \"protocol_version\" alert\n message and close the connection.\n\n If a TLS server receives a ClientHello containing a version number\n greater than the highest version supported by the server, it MUST\n reply according to the highest version supported by the server.\n\n A TLS server can also receive a ClientHello containing a version\n number smaller than the highest supported version. If the server\n wishes to negotiate with old clients, it will proceed as appropriate\n\n\n\nDierks & Rescorla Standards Track [Page 87]\n\f\nRFC 5246 TLS August 2008\n\n\n for the highest version supported by the server that is not greater\n than ClientHello.client_version. For example, if the server supports\n TLS 1.0, 1.1, and 1.2, and client_version is TLS 1.0, the server will\n proceed with a TLS 1.0 ServerHello. If server supports (or is\n willing to use) only versions greater than client_version, it MUST\n send a \"protocol_version\" alert message and close the connection.\n\n Whenever a client already knows the highest protocol version known to\n a server (for example, when resuming a session), it SHOULD initiate\n the connection in that native protocol.\n\n Note: some server implementations are known to implement version\n negotiation incorrectly. For example, there are buggy TLS 1.0\n servers that simply close the connection when the client offers a\n version newer than TLS 1.0. Also, it is known that some servers will\n refuse the connection if any TLS extensions are included in\n ClientHello. Interoperability with such buggy servers is a complex\n topic beyond the scope of this document, and may require multiple\n connection attempts by the client.\n\n Earlier versions of the TLS specification were not fully clear on\n what the record layer version number (TLSPlaintext.version) should\n contain when sending ClientHello (i.e., before it is known which\n version of the protocol will be employed). Thus, TLS servers\n compliant with this specification MUST accept any value {03,XX} as\n the record layer version number for ClientHello.\n\n TLS clients that wish to negotiate with older servers MAY send any\n value {03,XX} as the record layer version number. Typical values\n would be {03,00}, the lowest version number supported by the client,\n and the value of ClientHello.client_version. No single value will\n guarantee interoperability with all old servers, but this is a\n complex topic beyond the scope of this document.\n\nE.2. Compatibility with SSL 2.0\n\n TLS 1.2 clients that wish to support SSL 2.0 servers MUST send\n version 2.0 CLIENT-HELLO messages defined in [SSL2]. The message\n MUST contain the same version number as would be used for ordinary\n ClientHello, and MUST encode the supported TLS cipher suites in the\n CIPHER-SPECS-DATA field as described below.\n\n Warning: The ability to send version 2.0 CLIENT-HELLO messages will\n be phased out with all due haste, since the newer ClientHello format\n provides better mechanisms for moving to newer versions and\n negotiating extensions. TLS 1.2 clients SHOULD NOT support SSL 2.0.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 88]\n\f\nRFC 5246 TLS August 2008\n\n\n However, even TLS servers that do not support SSL 2.0 MAY accept\n version 2.0 CLIENT-HELLO messages. The message is presented below in\n sufficient detail for TLS server implementors; the true definition is\n still assumed to be [SSL2].\n\n For negotiation purposes, 2.0 CLIENT-HELLO is interpreted the same\n way as a ClientHello with a \"null\" compression method and no\n extensions. Note that this message MUST be sent directly on the\n wire, not wrapped as a TLS record. For the purposes of calculating\n Finished and CertificateVerify, the msg_length field is not\n considered to be a part of the handshake message.\n\n uint8 V2CipherSpec[3];\n struct {\n uint16 msg_length;\n uint8 msg_type;\n Version version;\n uint16 cipher_spec_length;\n uint16 session_id_length;\n uint16 challenge_length;\n V2CipherSpec cipher_specs[V2ClientHello.cipher_spec_length];\n opaque session_id[V2ClientHello.session_id_length];\n opaque challenge[V2ClientHello.challenge_length;\n } V2ClientHello;\n\n msg_length\n The highest bit MUST be 1; the remaining bits contain the length\n of the following data in bytes.\n\n msg_type\n This field, in conjunction with the version field, identifies a\n version 2 ClientHello message. The value MUST be 1.\n\n version\n Equal to ClientHello.client_version.\n\n cipher_spec_length\n This field is the total length of the field cipher_specs. It\n cannot be zero and MUST be a multiple of the V2CipherSpec length\n (3).\n\n session_id_length\n This field MUST have a value of zero for a client that claims to\n support TLS 1.2.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 89]\n\f\nRFC 5246 TLS August 2008\n\n\n challenge_length\n The length in bytes of the client's challenge to the server to\n authenticate itself. Historically, permissible values are between\n 16 and 32 bytes inclusive. When using the SSLv2 backward-\n compatible handshake the client SHOULD use a 32-byte challenge.\n\n cipher_specs\n This is a list of all CipherSpecs the client is willing and able\n to use. In addition to the 2.0 cipher specs defined in [SSL2],\n this includes the TLS cipher suites normally sent in\n ClientHello.cipher_suites, with each cipher suite prefixed by a\n zero byte. For example, the TLS cipher suite {0x00,0x0A} would be\n sent as {0x00,0x00,0x0A}.\n\n session_id\n This field MUST be empty.\n\n challenge\n Corresponds to ClientHello.random. If the challenge length is\n less than 32, the TLS server will pad the data with leading (note:\n not trailing) zero bytes to make it 32 bytes long.\n\n Note: Requests to resume a TLS session MUST use a TLS client hello.\n\nE.3. Avoiding Man-in-the-Middle Version Rollback\n\n When TLS clients fall back to Version 2.0 compatibility mode, they\n MUST use special PKCS#1 block formatting. This is done so that TLS\n servers will reject Version 2.0 sessions with TLS-capable clients.\n\n When a client negotiates SSL 2.0 but also supports TLS, it MUST set\n the right-hand (least-significant) 8 random bytes of the PKCS padding\n (not including the terminal null of the padding) for the RSA\n encryption of the ENCRYPTED-KEY-DATA field of the CLIENT-MASTER-KEY\n to 0x03 (the other padding bytes are random).\n\n When a TLS-capable server negotiates SSL 2.0 it SHOULD, after\n decrypting the ENCRYPTED-KEY-DATA field, check that these 8 padding\n bytes are 0x03. If they are not, the server SHOULD generate a random\n value for SECRET-KEY-DATA, and continue the handshake (which will\n eventually fail since the keys will not match). Note that reporting\n the error situation to the client could make the server vulnerable to\n attacks described in [BLEI].\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 90]\n\f\nRFC 5246 TLS August 2008\n\n\nAppendix F. Security Analysis\n\n The TLS protocol is designed to establish a secure connection between\n a client and a server communicating over an insecure channel. This\n document makes several traditional assumptions, including that\n attackers have substantial computational resources and cannot obtain\n secret information from sources outside the protocol. Attackers are\n assumed to have the ability to capture, modify, delete, replay, and\n otherwise tamper with messages sent over the communication channel.\n This appendix outlines how TLS has been designed to resist a variety\n of attacks.\n\nF.1. Handshake Protocol\n\n The handshake protocol is responsible for selecting a cipher spec and\n generating a master secret, which together comprise the primary\n cryptographic parameters associated with a secure session. The\n handshake protocol can also optionally authenticate parties who have\n certificates signed by a trusted certificate authority.\n\nF.1.1. Authentication and Key Exchange\n\n TLS supports three authentication modes: authentication of both\n parties, server authentication with an unauthenticated client, and\n total anonymity. Whenever the server is authenticated, the channel\n is secure against man-in-the-middle attacks, but completely anonymous\n sessions are inherently vulnerable to such attacks. Anonymous\n servers cannot authenticate clients. If the server is authenticated,\n its certificate message must provide a valid certificate chain\n leading to an acceptable certificate authority. Similarly,\n authenticated clients must supply an acceptable certificate to the\n server. Each party is responsible for verifying that the other's\n certificate is valid and has not expired or been revoked.\n\n The general goal of the key exchange process is to create a\n pre_master_secret known to the communicating parties and not to\n attackers. The pre_master_secret will be used to generate the\n master_secret (see Section 8.1). The master_secret is required to\n generate the Finished messages, encryption keys, and MAC keys (see\n Sections 7.4.9 and 6.3). By sending a correct Finished message,\n parties thus prove that they know the correct pre_master_secret.\n\nF.1.1.1. Anonymous Key Exchange\n\n Completely anonymous sessions can be established using Diffie-Hellman\n for key exchange. The server's public parameters are contained in\n the server key exchange message, and the client's are sent in the\n\n\n\n\nDierks & Rescorla Standards Track [Page 91]\n\f\nRFC 5246 TLS August 2008\n\n\n client key exchange message. Eavesdroppers who do not know the\n private values should not be able to find the Diffie-Hellman result\n (i.e., the pre_master_secret).\n\n Warning: Completely anonymous connections only provide protection\n against passive eavesdropping. Unless an independent tamper-proof\n channel is used to verify that the Finished messages were not\n replaced by an attacker, server authentication is required in\n environments where active man-in-the-middle attacks are a concern.\n\nF.1.1.2. RSA Key Exchange and Authentication\n\n With RSA, key exchange and server authentication are combined. The\n public key is contained in the server's certificate. Note that\n compromise of the server's static RSA key results in a loss of\n confidentiality for all sessions protected under that static key.\n TLS users desiring Perfect Forward Secrecy should use DHE cipher\n suites. The damage done by exposure of a private key can be limited\n by changing one's private key (and certificate) frequently.\n\n After verifying the server's certificate, the client encrypts a\n pre_master_secret with the server's public key. By successfully\n decoding the pre_master_secret and producing a correct Finished\n message, the server demonstrates that it knows the private key\n corresponding to the server certificate.\n\n When RSA is used for key exchange, clients are authenticated using\n the certificate verify message (see Section 7.4.8). The client signs\n a value derived from all preceding handshake messages. These\n handshake messages include the server certificate, which binds the\n signature to the server, and ServerHello.random, which binds the\n signature to the current handshake process.\n\nF.1.1.3. Diffie-Hellman Key Exchange with Authentication\n\n When Diffie-Hellman key exchange is used, the server can either\n supply a certificate containing fixed Diffie-Hellman parameters or\n use the server key exchange message to send a set of temporary\n Diffie-Hellman parameters signed with a DSA or RSA certificate.\n Temporary parameters are hashed with the hello.random values before\n signing to ensure that attackers do not replay old parameters. In\n either case, the client can verify the certificate or signature to\n ensure that the parameters belong to the server.\n\n If the client has a certificate containing fixed Diffie-Hellman\n parameters, its certificate contains the information required to\n complete the key exchange. Note that in this case the client and\n server will generate the same Diffie-Hellman result (i.e.,\n\n\n\nDierks & Rescorla Standards Track [Page 92]\n\f\nRFC 5246 TLS August 2008\n\n\n pre_master_secret) every time they communicate. To prevent the\n pre_master_secret from staying in memory any longer than necessary,\n it should be converted into the master_secret as soon as possible.\n Client Diffie-Hellman parameters must be compatible with those\n supplied by the server for the key exchange to work.\n\n If the client has a standard DSA or RSA certificate or is\n unauthenticated, it sends a set of temporary parameters to the server\n in the client key exchange message, then optionally uses a\n certificate verify message to authenticate itself.\n\n If the same DH keypair is to be used for multiple handshakes, either\n because the client or server has a certificate containing a fixed DH\n keypair or because the server is reusing DH keys, care must be taken\n to prevent small subgroup attacks. Implementations SHOULD follow the\n guidelines found in [SUBGROUP].\n\n Small subgroup attacks are most easily avoided by using one of the\n DHE cipher suites and generating a fresh DH private key (X) for each\n handshake. If a suitable base (such as 2) is chosen, g^X mod p can\n be computed very quickly; therefore, the performance cost is\n minimized. Additionally, using a fresh key for each handshake\n provides Perfect Forward Secrecy. Implementations SHOULD generate a\n new X for each handshake when using DHE cipher suites.\n\n Because TLS allows the server to provide arbitrary DH groups, the\n client should verify that the DH group is of suitable size as defined\n by local policy. The client SHOULD also verify that the DH public\n exponent appears to be of adequate size. [KEYSIZ] provides a useful\n guide to the strength of various group sizes. The server MAY choose\n to assist the client by providing a known group, such as those\n defined in [IKEALG] or [MODP]. These can be verified by simple\n comparison.\n\nF.1.2. Version Rollback Attacks\n\n Because TLS includes substantial improvements over SSL Version 2.0,\n attackers may try to make TLS-capable clients and servers fall back\n to Version 2.0. This attack can occur if (and only if) two TLS-\n capable parties use an SSL 2.0 handshake.\n\n Although the solution using non-random PKCS #1 block type 2 message\n padding is inelegant, it provides a reasonably secure way for Version\n 3.0 servers to detect the attack. This solution is not secure\n against attackers who can brute-force the key and substitute a new\n ENCRYPTED-KEY-DATA message containing the same key (but with normal\n padding) before the application-specified wait threshold has expired.\n Altering the padding of the least-significant 8 bytes of the PKCS\n\n\n\nDierks & Rescorla Standards Track [Page 93]\n\f\nRFC 5246 TLS August 2008\n\n\n padding does not impact security for the size of the signed hashes\n and RSA key lengths used in the protocol, since this is essentially\n equivalent to increasing the input block size by 8 bytes.\n\nF.1.3. Detecting Attacks Against the Handshake Protocol\n\n An attacker might try to influence the handshake exchange to make the\n parties select different encryption algorithms than they would\n normally choose.\n\n For this attack, an attacker must actively change one or more\n handshake messages. If this occurs, the client and server will\n compute different values for the handshake message hashes. As a\n result, the parties will not accept each others' Finished messages.\n Without the master_secret, the attacker cannot repair the Finished\n messages, so the attack will be discovered.\n\nF.1.4. Resuming Sessions\n\n When a connection is established by resuming a session, new\n ClientHello.random and ServerHello.random values are hashed with the\n session's master_secret. Provided that the master_secret has not\n been compromised and that the secure hash operations used to produce\n the encryption keys and MAC keys are secure, the connection should be\n secure and effectively independent from previous connections.\n Attackers cannot use known encryption keys or MAC secrets to\n compromise the master_secret without breaking the secure hash\n operations.\n\n Sessions cannot be resumed unless both the client and server agree.\n If either party suspects that the session may have been compromised,\n or that certificates may have expired or been revoked, it should\n force a full handshake. An upper limit of 24 hours is suggested for\n session ID lifetimes, since an attacker who obtains a master_secret\n may be able to impersonate the compromised party until the\n corresponding session ID is retired. Applications that may be run in\n relatively insecure environments should not write session IDs to\n stable storage.\n\nF.2. Protecting Application Data\n\n The master_secret is hashed with the ClientHello.random and\n ServerHello.random to produce unique data encryption keys and MAC\n secrets for each connection.\n\n Outgoing data is protected with a MAC before transmission. To\n prevent message replay or modification attacks, the MAC is computed\n from the MAC key, the sequence number, the message length, the\n\n\n\nDierks & Rescorla Standards Track [Page 94]\n\f\nRFC 5246 TLS August 2008\n\n\n message contents, and two fixed character strings. The message type\n field is necessary to ensure that messages intended for one TLS\n record layer client are not redirected to another. The sequence\n number ensures that attempts to delete or reorder messages will be\n detected. Since sequence numbers are 64 bits long, they should never\n overflow. Messages from one party cannot be inserted into the\n other's output, since they use independent MAC keys. Similarly, the\n server write and client write keys are independent, so stream cipher\n keys are used only once.\n\n If an attacker does break an encryption key, all messages encrypted\n with it can be read. Similarly, compromise of a MAC key can make\n message-modification attacks possible. Because MACs are also\n encrypted, message-alteration attacks generally require breaking the\n encryption algorithm as well as the MAC.\n\n Note: MAC keys may be larger than encryption keys, so messages can\n remain tamper resistant even if encryption keys are broken.\n\nF.3. Explicit IVs\n\n [CBCATT] describes a chosen plaintext attack on TLS that depends on\n knowing the IV for a record. Previous versions of TLS [TLS1.0] used\n the CBC residue of the previous record as the IV and therefore\n enabled this attack. This version uses an explicit IV in order to\n protect against this attack.\n\nF.4. Security of Composite Cipher Modes\n\n TLS secures transmitted application data via the use of symmetric\n encryption and authentication functions defined in the negotiated\n cipher suite. The objective is to protect both the integrity and\n confidentiality of the transmitted data from malicious actions by\n active attackers in the network. It turns out that the order in\n which encryption and authentication functions are applied to the data\n plays an important role for achieving this goal [ENCAUTH].\n\n The most robust method, called encrypt-then-authenticate, first\n applies encryption to the data and then applies a MAC to the\n ciphertext. This method ensures that the integrity and\n confidentiality goals are obtained with ANY pair of encryption and\n MAC functions, provided that the former is secure against chosen\n plaintext attacks and that the MAC is secure against chosen-message\n attacks. TLS uses another method, called authenticate-then-encrypt,\n in which first a MAC is computed on the plaintext and then the\n concatenation of plaintext and MAC is encrypted. This method has\n been proven secure for CERTAIN combinations of encryption functions\n and MAC functions, but it is not guaranteed to be secure in general.\n\n\n\nDierks & Rescorla Standards Track [Page 95]\n\f\nRFC 5246 TLS August 2008\n\n\n In particular, it has been shown that there exist perfectly secure\n encryption functions (secure even in the information-theoretic sense)\n that combined with any secure MAC function, fail to provide the\n confidentiality goal against an active attack. Therefore, new cipher\n suites and operation modes adopted into TLS need to be analyzed under\n the authenticate-then-encrypt method to verify that they achieve the\n stated integrity and confidentiality goals.\n\n Currently, the security of the authenticate-then-encrypt method has\n been proven for some important cases. One is the case of stream\n ciphers in which a computationally unpredictable pad of the length of\n the message, plus the length of the MAC tag, is produced using a\n pseudorandom generator and this pad is exclusive-ORed with the\n concatenation of plaintext and MAC tag. The other is the case of CBC\n mode using a secure block cipher. In this case, security can be\n shown if one applies one CBC encryption pass to the concatenation of\n plaintext and MAC and uses a new, independent, and unpredictable IV\n for each new pair of plaintext and MAC. In versions of TLS prior to\n 1.1, CBC mode was used properly EXCEPT that it used a predictable IV\n in the form of the last block of the previous ciphertext. This made\n TLS open to chosen plaintext attacks. This version of the protocol\n is immune to those attacks. For exact details in the encryption\n modes proven secure, see [ENCAUTH].\n\nF.5. Denial of Service\n\n TLS is susceptible to a number of denial-of-service (DoS) attacks.\n In particular, an attacker who initiates a large number of TCP\n connections can cause a server to consume large amounts of CPU for\n doing RSA decryption. However, because TLS is generally used over\n TCP, it is difficult for the attacker to hide his point of origin if\n proper TCP SYN randomization is used [SEQNUM] by the TCP stack.\n\n Because TLS runs over TCP, it is also susceptible to a number of DoS\n attacks on individual connections. In particular, attackers can\n forge RSTs, thereby terminating connections, or forge partial TLS\n records, thereby causing the connection to stall. These attacks\n cannot in general be defended against by a TCP-using protocol.\n Implementors or users who are concerned with this class of attack\n should use IPsec AH [AH] or ESP [ESP].\n\nF.6. Final Notes\n\n For TLS to be able to provide a secure connection, both the client\n and server systems, keys, and applications must be secure. In\n addition, the implementation must be free of security errors.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 96]\n\f\nRFC 5246 TLS August 2008\n\n\n The system is only as strong as the weakest key exchange and\n authentication algorithm supported, and only trustworthy\n cryptographic functions should be used. Short public keys and\n anonymous servers should be used with great caution. Implementations\n and users must be careful when deciding which certificates and\n certificate authorities are acceptable; a dishonest certificate\n authority can do tremendous damage.\n\nNormative References\n\n [AES] National Institute of Standards and Technology,\n \"Specification for the Advanced Encryption Standard (AES)\"\n FIPS 197. November 26, 2001.\n\n [3DES] National Institute of Standards and Technology,\n \"Recommendation for the Triple Data Encryption Algorithm\n (TDEA) Block Cipher\", NIST Special Publication 800-67, May\n 2004.\n\n [DSS] NIST FIPS PUB 186-2, \"Digital Signature Standard\",\n National Institute of Standards and Technology, U.S.\n Department of Commerce, 2000.\n\n [HMAC] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC: Keyed-\n Hashing for Message Authentication\", RFC 2104, February\n 1997.\n\n [MD5] Rivest, R., \"The MD5 Message-Digest Algorithm\", RFC 1321,\n April 1992.\n\n [PKCS1] Jonsson, J. and B. Kaliski, \"Public-Key Cryptography\n Standards (PKCS) #1: RSA Cryptography Specifications\n Version 2.1\", RFC 3447, February 2003.\n\n [PKIX] Housley, R., Polk, W., Ford, W., and D. Solo, \"Internet\n X.509 Public Key Infrastructure Certificate and\n Certificate Revocation List (CRL) Profile\", RFC 3280,\n April 2002.\n\n [SCH] B. Schneier. \"Applied Cryptography: Protocols, Algorithms,\n and Source Code in C, 2nd ed.\", Published by John Wiley &\n Sons, Inc. 1996.\n\n [SHS] NIST FIPS PUB 180-2, \"Secure Hash Standard\", National\n Institute of Standards and Technology, U.S. Department of\n Commerce, August 2002.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 97]\n\f\nRFC 5246 TLS August 2008\n\n\n [REQ] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2434] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 2434,\n October 1998.\n\n [X680] ITU-T Recommendation X.680 (2002) | ISO/IEC 8824-1:2002,\n Information technology - Abstract Syntax Notation One\n (ASN.1): Specification of basic notation.\n\n [X690] ITU-T Recommendation X.690 (2002) | ISO/IEC 8825-1:2002,\n Information technology - ASN.1 encoding Rules:\n Specification of Basic Encoding Rules (BER), Canonical\n Encoding Rules (CER) and Distinguished Encoding Rules\n (DER).\n\nInformative References\n\n [AEAD] McGrew, D., \"An Interface and Algorithms for Authenticated\n Encryption\", RFC 5116, January 2008.\n\n [AH] Kent, S., \"IP Authentication Header\", RFC 4302, December\n 2005.\n\n [BLEI] Bleichenbacher D., \"Chosen Ciphertext Attacks against\n Protocols Based on RSA Encryption Standard PKCS #1\" in\n Advances in Cryptology -- CRYPTO'98, LNCS vol. 1462,\n pages: 1-12, 1998.\n\n [CBCATT] Moeller, B., \"Security of CBC Ciphersuites in SSL/TLS:\n Problems and Countermeasures\",\n http://www.openssl.org/~bodo/tls-cbc.txt.\n\n [CBCTIME] Canvel, B., Hiltgen, A., Vaudenay, S., and M. Vuagnoux,\n \"Password Interception in a SSL/TLS Channel\", Advances in\n Cryptology -- CRYPTO 2003, LNCS vol. 2729, 2003.\n\n [CCM] \"NIST Special Publication 800-38C: The CCM Mode for\n Authentication and Confidentiality\",\n http://csrc.nist.gov/publications/nistpubs/800-38C/\n SP800-38C.pdf\n\n [DES] National Institute of Standards and Technology, \"Data\n Encryption Standard (DES)\", FIPS PUB 46-3, October 1999.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 98]\n\f\nRFC 5246 TLS August 2008\n\n\n [DSS-3] NIST FIPS PUB 186-3 Draft, \"Digital Signature Standard\",\n National Institute of Standards and Technology, U.S.\n Department of Commerce, 2006.\n\n [ECDSA] American National Standards Institute, \"Public Key\n Cryptography for the Financial Services Industry: The\n Elliptic Curve Digital Signature Algorithm (ECDSA)\", ANS\n X9.62-2005, November 2005.\n\n [ENCAUTH] Krawczyk, H., \"The Order of Encryption and Authentication\n for Protecting Communications (Or: How Secure is SSL?)\",\n Crypto 2001.\n\n [ESP] Kent, S., \"IP Encapsulating Security Payload (ESP)\", RFC\n 4303, December 2005.\n\n [FI06] Hal Finney, \"Bleichenbacher's RSA signature forgery based\n on implementation error\", [redacted-email] mailing\n list, 27 August 2006, http://www.imc.org/ietf-openpgp/\n mail-archive/msg14307.html.\n\n [GCM] Dworkin, M., NIST Special Publication 800-38D,\n \"Recommendation for Block Cipher Modes of Operation:\n Galois/Counter Mode (GCM) and GMAC\", November 2007.\n\n [IKEALG] Schiller, J., \"Cryptographic Algorithms for Use in the\n Internet Key Exchange Version 2 (IKEv2)\", RFC 4307,\n December 2005.\n\n [KEYSIZ] Orman, H. and P. Hoffman, \"Determining Strengths For\n Public Keys Used For Exchanging Symmetric Keys\", BCP 86,\n RFC 3766, April 2004.\n\n [KPR03] Klima, V., Pokorny, O., Rosa, T., \"Attacking RSA-based\n Sessions in SSL/TLS\", http://eprint.iacr.org/2003/052/,\n March 2003.\n\n [MODP] Kivinen, T. and M. Kojo, \"More Modular Exponential (MODP)\n Diffie-Hellman groups for Internet Key Exchange (IKE)\",\n RFC 3526, May 2003.\n\n [PKCS6] RSA Laboratories, \"PKCS #6: RSA Extended Certificate\n Syntax Standard\", version 1.5, November 1993.\n\n [PKCS7] RSA Laboratories, \"PKCS #7: RSA Cryptographic Message\n Syntax Standard\", version 1.5, November 1993.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 99]\n\f\nRFC 5246 TLS August 2008\n\n\n [RANDOM] Eastlake, D., 3rd, Schiller, J., and S. Crocker,\n \"Randomness Requirements for Security\", BCP 106, RFC 4086,\n June 2005.\n\n [RFC3749] Hollenbeck, S., \"Transport Layer Security Protocol\n Compression Methods\", RFC 3749, May 2004.\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, April 2006.\n\n [RSA] R. Rivest, A. Shamir, and L. M. Adleman, \"A Method for\n Obtaining Digital Signatures and Public-Key\n Cryptosystems\", Communications of the ACM, v. 21, n. 2,\n Feb 1978, pp. 120-126.\n\n [SEQNUM] Bellovin, S., \"Defending Against Sequence Number Attacks\",\n RFC 1948, May 1996.\n\n [SSL2] Hickman, Kipp, \"The SSL Protocol\", Netscape Communications\n Corp., Feb 9, 1995.\n\n [SSL3] A. Freier, P. Karlton, and P. Kocher, \"The SSL 3.0\n Protocol\", Netscape Communications Corp., Nov 18, 1996.\n\n [SUBGROUP] Zuccherato, R., \"Methods for Avoiding the \"Small-Subgroup\"\n Attacks on the Diffie-Hellman Key Agreement Method for\n S/MIME\", RFC 2785, March 2000.\n\n [TCP] Postel, J., \"Transmission Control Protocol\", STD 7, RFC\n 793, September 1981.\n\n [TIMING] Boneh, D., Brumley, D., \"Remote timing attacks are\n practical\", USENIX Security Symposium 2003.\n\n [TLSAES] Chown, P., \"Advanced Encryption Standard (AES)\n Ciphersuites for Transport Layer Security (TLS)\", RFC\n 3268, June 2002.\n\n [TLSECC] Blake-Wilson, S., Bolyard, N., Gupta, V., Hawk, C., and B.\n Moeller, \"Elliptic Curve Cryptography (ECC) Cipher Suites\n for Transport Layer Security (TLS)\", RFC 4492, May 2006.\n\n [TLSEXT] Eastlake, D., 3rd, \"Transport Layer Security (TLS)\n Extensions: Extension Definitions\", Work in Progress,\n February 2008.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 100]\n\f\nRFC 5246 TLS August 2008\n\n\n [TLSPGP] Mavrogiannopoulos, N., \"Using OpenPGP Keys for Transport\n Layer Security (TLS) Authentication\", RFC 5081, November\n 2007.\n\n [TLSPSK] Eronen, P., Ed., and H. Tschofenig, Ed., \"Pre-Shared Key\n Ciphersuites for Transport Layer Security (TLS)\", RFC\n 4279, December 2005.\n\n [TLS1.0] Dierks, T. and C. Allen, \"The TLS Protocol Version 1.0\",\n RFC 2246, January 1999.\n\n [TLS1.1] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346, April 2006.\n\n [X501] ITU-T Recommendation X.501: Information Technology - Open\n Systems Interconnection - The Directory: Models, 1993.\n\n [XDR] Eisler, M., Ed., \"XDR: External Data Representation\n Standard\", STD 67, RFC 4506, May 2006.\n\nWorking Group Information\n\n The discussion list for the IETF TLS working group is located at the\n e-mail address <[redacted-email]>. Information on the group and\n information on how to subscribe to the list is at\n \n\n Archives of the list can be found at:\n \n\nContributors\n\n Christopher Allen (co-editor of TLS 1.0)\n Alacrity Ventures\n [redacted-email]\n\n Martin Abadi\n University of California, Santa Cruz\n [redacted-email]\n\n Steven M. Bellovin\n Columbia University\n [redacted-email]\n\n Simon Blake-Wilson\n BCI\n [redacted-email]\n\n\n\n\nDierks & Rescorla Standards Track [Page 101]\n\f\nRFC 5246 TLS August 2008\n\n\n Ran Canetti\n IBM\n [redacted-email]\n\n Pete Chown\n Skygate Technology Ltd\n [redacted-email]\n\n Taher Elgamal\n [redacted-email]\n Securify\n\n Pasi Eronen\n [redacted-email]\n Nokia\n\n Anil Gangolli\n [redacted-email]\n\n Kipp Hickman\n\n Alfred Hoenes\n\n David Hopwood\n Independent Consultant\n [redacted-email]\n\n Phil Karlton (co-author of SSLv3)\n\n Paul Kocher (co-author of SSLv3)\n Cryptography Research\n [redacted-email]\n\n Hugo Krawczyk\n IBM\n [redacted-email]\n\n Jan Mikkelsen\n Transactionware\n [redacted-email]\n\n Magnus Nystrom\n RSA Security\n [redacted-email]\n\n Robert Relyea\n Netscape Communications\n [redacted-email]\n\n\n\nDierks & Rescorla Standards Track [Page 102]\n\f\nRFC 5246 TLS August 2008\n\n\n Jim Roskind\n Netscape Communications\n [redacted-email]\n\n Michael Sabin\n\n Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n Tom Weinstein\n\n Tim Wright\n Vodafone\n [redacted-email]\n\nEditors' Addresses\n\n Tim Dierks\n Independent\n EMail: [redacted-email]\n\n Eric Rescorla\n RTFM, Inc.\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 103]\n\f\nRFC 5246 TLS August 2008\n\n\nFull Copyright Statement\n\n Copyright (C) The IETF Trust (2008).\n\n This document is subject to the rights, licenses and restrictions\n contained in BCP 78, and except as set forth therein, the authors\n retain all their rights.\n\n This document and the information contained herein are provided on an\n \"AS IS\" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\n OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\n THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\n OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\n THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\n WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\n\nIntellectual Property\n\n The IETF takes no position regarding the validity or scope of any\n Intellectual Property Rights or other rights that might be claimed to\n pertain to the implementation or use of the technology described in\n this document or the extent to which any license under such rights\n might or might not be available; nor does it represent that it has\n made any independent effort to identify any such rights. Information\n on the procedures with respect to rights in RFC documents can be\n found in BCP 78 and BCP 79.\n\n Copies of IPR disclosures made to the IETF Secretariat and any\n assurances of licenses to be made available, or the result of an\n attempt made to obtain a general license or permission for the use of\n such proprietary rights by implementers or users of this\n specification can be obtained from the IETF on-line IPR repository at\n http://www.ietf.org/ipr.\n\n The IETF invites any interested party to bring to its attention any\n copyrights, patents or patent applications, or other proprietary\n rights that may cover technology that may be required to implement\n this standard. Please address the information to the IETF at\n ietf-ipr@ietf.org.\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 104]", "text_sha256": "0ba07ad7c1e04ded76e38500ef83eaa39b002af6cc226329ff75c3e8859fed85"}, {"draft_references": [], "excluded_rfc_relation_targets": [], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5705 RTFM, Inc.", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "5705"}], "kind": "rfc", "occurred_at": "2010-03-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 1, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a", "record_id": "ietf:rfc:5705", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "rfc_number": 5705, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a", "source_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "temporal_semantics": "source_event_time", "text": "Internet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 5705 RTFM, Inc.\nCategory: Standards Track March 2010\nISSN: 2070-1721\n\n\n Keying Material Exporters for Transport Layer Security (TLS)\n\nAbstract\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n to perform key establishment but then use some of the keying material\n for their own purposes. This document describes a general mechanism\n for allowing that.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/5705.\n\nCopyright Notice\n\n Copyright (c) 2010 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2\n 2. Conventions Used In This Document . . . . . . . . . . . . . . . 3\n 3. Binding to Application Contexts . . . . . . . . . . . . . . . . 3\n 4. Exporter Definition . . . . . . . . . . . . . . . . . . . . . . 4\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . . 5\n 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 6\n 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . 6\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 7\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 7\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 7\n\n1. Introduction\n\n Note: The mechanism described in this document was previously known\n as \"TLS Extractors\" but was changed to avoid a name conflict\n with the use of the term \"Extractor\" in the cryptographic\n community.\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n [RFC5246] or Datagram TLS (DTLS) [RFC4347] to perform key\n establishment but then use some of the keying material for their own\n purposes. A typical example is DTLS-SRTP [DTLS-SRTP], a key\n management scheme for the Secure Real-time Transport Protocol (SRTP)\n that uses DTLS to perform a key exchange and negotiate the SRTP\n [RFC3711] protection suite and then uses the DTLS master_secret to\n generate the SRTP keys.\n\n These applications imply a need to be able to export keying material\n (later called Exported Keying Material or EKM) from TLS/DTLS to an\n application or protocol residing at an upper layer, and to securely\n agree on the upper-layer context where the keying material will be\n used. The mechanism for exporting the keying material has the\n following requirements:\n\n o Both client and server need to be able to export the same EKM\n value.\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n o EKM values should be indistinguishable from random data to\n attackers who don't know the master_secret.\n\n o It should be possible to export multiple EKM values from the same\n TLS/DTLS association.\n\n o Knowing one EKM value should not reveal any useful information\n about the master_secret or about other EKM values.\n\n The mechanism described in this document is intended to fulfill these\n requirements. This mechanism is compatible with all versions of TLS.\n\n2. Conventions Used In This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Binding to Application Contexts\n\n In addition to using an exporter to obtain keying material, an\n application using the keying material has to securely establish the\n upper-layer context where the keying material will be used. The\n details of this context depend on the application, but it could\n include things such as algorithms and parameters that will be used\n with the keys, identifier(s) for the endpoint(s) who will use the\n keys, identifier(s) for the session(s) where the keys will be used,\n and the lifetime(s) for the context and/or keys. At a minimum, there\n should be some mechanism for signaling that an exporter will be used.\n\n This specification does not mandate a single mechanism for agreeing\n on such context; instead, there are several possibilities that can be\n used (and can complement each other). For example:\n\n o Information about the upper-layer context can be included in the\n optional data after the exporter label (see Section 4).\n\n o Information about the upper-layer context can be exchanged in TLS\n extensions included in the ClientHello and ServerHello messages.\n This approach is used in [DTLS-SRTP]. The handshake messages are\n protected by the Finished messages, so once the handshake\n completes, the peers will have the same view of the information.\n Extensions also allow a limited form of negotiation: for example,\n the TLS client could propose several alternatives for some context\n parameters, and the TLS server could select one of them.\n\n o The upper-layer protocol can include its own handshake, which can\n be protected using the keys exported by TLS.\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n No matter how the context is agreed, it is required that it has one\n part that indicates which application will use the exported keys.\n This part is the disambiguating label string (see Section 4).\n\n It is important to note that just embedding TLS messages in the\n upper-layer protocol may not automatically secure all the important\n context information, since the upper-layer messages are not covered\n by TLS Finished messages.\n\n4. Exporter Definition\n\n The output of the exporter is intended to be used in a single scope,\n which is associated with the TLS session, the label, and the context\n value.\n\n The exporter takes three input values:\n\n o a disambiguating label string,\n\n o a per-association context value provided by the application using\n the exporter, and\n\n o a length value.\n\n If no context is provided, it then computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random\n )[length]\n\n If context is provided, it computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random +\n context_value_length + context_value\n )[length]\n\n Where PRF is the TLS Pseudorandom Function in use for the session.\n The output is a pseudorandom bit string of length bytes generated\n from the master_secret. (This construction allows for\n interoperability with older exporter-type constructions which do not\n use context values, e.g., [RFC5281]).\n\n Labels here have the same definition as in TLS, i.e., an ASCII string\n with no terminating NULL. Label values beginning with \"EXPERIMENTAL\"\n MAY be used for private use without registration. All other label\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n values MUST be registered via Specification Required as described by\n RFC 5226 [RFC5226]. Note that exporter labels have the potential to\n collide with existing PRF labels. In order to prevent this, labels\n SHOULD begin with \"EXPORTER\". This is not a MUST because there are\n existing uses that have labels which do not begin with this prefix.\n\n The context value allows the application using the exporter to mix\n its own data with the TLS PRF for the exporter output. One example\n of where this might be useful is an authentication setting where the\n client credentials are valid for more than one identity; the context\n value could then be used to mix the expected identity into the keying\n material, thus preventing substitution attacks. The context value\n length is encoded as an unsigned, 16-bit quantity (uint16; see\n [RFC5246], Section 4.4) representing the length of the context value.\n The context MAY be zero length. Because the context value is mixed\n with the master_secret via the PRF, it is safe to mix confidential\n information into the exporter, provided that the master_secret will\n not be known to the attacker.\n\n5. Security Considerations\n\n The prime security requirement for exporter outputs is that they be\n independent. More formally, after a particular TLS session, if an\n adversary is allowed to choose multiple (label, context value) pairs\n and is given the output of the PRF for those values, the attacker is\n still unable to distinguish between the output of the PRF for a\n (label, context value) pair (different from the ones that it\n submitted) and a random value of the same length. In particular,\n there may be settings, such as the one described in Section 4, where\n the attacker can control the context value; such an attacker MUST NOT\n be able to predict the output of the exporter. Similarly, an\n attacker who does not know the master secret should not be able to\n distinguish valid exporter outputs from random values. The current\n set of TLS PRFs is believed to meet this objective, provided the\n master secret is randomly generated.\n\n Because an exporter produces the same value if applied twice with the\n same label to the same master_secret, it is critical that two EKM\n values generated with the same label not be used for two different\n purposes -- hence, the requirement for IANA registration. However,\n because exporters depend on the TLS PRF, it is not a threat to the\n use of an EKM value generated from one label to reveal an EKM value\n generated from another label.\n\n With certain TLS cipher suites, the TLS master secret is not\n necessarily unique to a single TLS session. In particular, with RSA\n key exchange, a malicious party acting as TLS server in one session\n and as TLS client in another session can cause those two sessions to\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n have the same TLS master secret (though the sessions must be\n established simultaneously to get adequate control of the Random\n values). Applications using the EKM need to consider this in how\n they use the EKM; in some cases, requiring the use of other cipher\n suites (such as those using a Diffie-Hellman key exchange) may be\n advisable.\n\n Designing a secure mechanism that uses exporters is not necessarily\n straightforward. This document only provides the exporter mechanism,\n but the problem of agreeing on the surrounding context and the\n meaning of the information passed to and from the exporter remains.\n Any new uses of the exporter mechanism should be subject to careful\n review.\n\n6. IANA Considerations\n\n IANA has created a TLS Exporter Label registry for this purpose. The\n initial contents of the registry are given below:\n\n Value Reference Note\n ----------------------------- --------- ----\n client finished [RFC5246] (1)\n server finished [RFC5246] (1)\n master secret [RFC5246] (1)\n key expansion [RFC5246] (1)\n client EAP encryption [RFC5216]\n ttls keying material [RFC5281]\n ttls challenge [RFC5281]\n\n Note: (1) These entries are reserved and MUST NOT be used for the\n purpose described in RFC 5705, in order to avoid confusion with\n similar, but distinct, use in RFC 5246.\n\n Future values are allocated via the RFC 5226 Specification Required\n policy. The label is a string consisting of printable ASCII\n characters. IANA MUST also verify that one label is not a prefix of\n any other label. For example, labels \"key\" or \"master secretary\" are\n forbidden.\n\n7. Acknowledgments\n\n Thanks to Pasi Eronen for valuable comments and for the contents of\n the IANA section and Section 3. Thanks to David McGrew for helpful\n discussion of the security considerations and to Vijay Gurbani and\n Alfred Hoenes for editorial comments.\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246,\n August 2008.\n\n8.2. Informative References\n\n [DTLS-SRTP] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for Secure\n Real-time Transport Protocol (SRTP)\", Work in Progress,\n February 2009.\n\n [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and\n K. Norrman, \"The Secure Real-time Transport Protocol\n (SRTP)\", RFC 3711, March 2004.\n\n [RFC4347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security\", RFC 4347, April 2006.\n\n [RFC5216] Simon, D., Aboba, B., and R. Hurst, \"The EAP-TLS\n Authentication Protocol\", RFC 5216, March 2008.\n\n [RFC5281] Funk, P. and S. Blake-Wilson, \"Extensible Authentication\n Protocol Tunneled Transport Layer Security Authenticated\n Protocol Version 0 (EAP-TTLSv0)\", RFC 5281, August 2008.\n\nAuthor's Address\n\n Eric Rescorla\n RTFM, Inc.\n 2064 Edgewood Drive\n Palo Alto, CA 94303\n USA\n\n EMail: [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 7]", "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"draft_references": [], "excluded_rfc_relation_targets": [{"kind": "obsoletes", "reason": "not_requested", "rfc_number": 2011}, {"kind": "obsoletes", "reason": "not_requested", "rfc_number": 4366}], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6066 Huawei", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "6066"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 4366 January 2011", "fact_id": "obsoletes:4366", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "4366"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 4366 January 2011", "fact_id": "obsoletes:2011", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "2011"}], "kind": "rfc", "occurred_at": "2011-01-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 3, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0", "record_id": "ietf:rfc:6066", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "rfc_number": 6066, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0", "source_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "temporal_semantics": "source_event_time", "text": "Internet Engineering Task Force (IETF) D. Eastlake 3rd\nRequest for Comments: 6066 Huawei\nObsoletes: 4366 January 2011\nCategory: Standards Track\nISSN: 2070-1721\n\n\n Transport Layer Security (TLS) Extensions: Extension Definitions\n\nAbstract\n\n This document provides specifications for existing TLS extensions.\n It is a companion document for RFC 5246, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\". The extensions specified are\n server_name, max_fragment_length, client_certificate_url,\n trusted_ca_keys, truncated_hmac, and status_request.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6066.\n\nCopyright Notice\n\n Copyright (c) 2011 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n\n\n\n\n\nEastlake Standards Track [Page 1]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction ....................................................3\n 1.1. Specific Extensions Covered ................................3\n 1.2. Conventions Used in This Document ..........................5\n 2. Extensions to the Handshake Protocol ............................5\n 3. Server Name Indication ..........................................6\n 4. Maximum Fragment Length Negotiation .............................8\n 5. Client Certificate URLs .........................................9\n 6. Trusted CA Indication ..........................................12\n 7. Truncated HMAC .................................................13\n 8. Certificate Status Request .....................................14\n 9. Error Alerts ...................................................16\n 10. IANA Considerations ...........................................17\n 10.1. pkipath MIME Type Registration ...........................17\n 10.2. Reference for TLS Alerts, TLS HandshakeTypes, and\n ExtensionTypes ...........................................19\n 11. Security Considerations .......................................19\n 11.1. Security Considerations for server_name ..................19\n 11.2. Security Considerations for max_fragment_length ..........20\n 11.3. Security Considerations for client_certificate_url .......20\n 11.4. Security Considerations for trusted_ca_keys ..............21\n 11.5. Security Considerations for truncated_hmac ...............21\n 11.6. Security Considerations for status_request ...............22\n 12. Normative References ..........................................22\n 13. Informative References ........................................23\n Appendix A. Changes from RFC 4366 .................................24\n Appendix B. Acknowledgements ......................................25\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 2]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1. Introduction\n\n The Transport Layer Security (TLS) Protocol Version 1.2 is specified\n in [RFC5246]. That specification includes the framework for\n extensions to TLS, considerations in designing such extensions (see\n Section 7.4.1.4 of [RFC5246]), and IANA Considerations for the\n allocation of new extension code points; however, it does not specify\n any particular extensions other than Signature Algorithms (see\n Section 7.4.1.4.1 of [RFC5246]).\n\n This document provides the specifications for existing TLS\n extensions. It is, for the most part, the adaptation and editing of\n material from RFC 4366, which covered TLS extensions for TLS 1.0 (RFC\n 2246) and TLS 1.1 (RFC 4346).\n\n1.1. Specific Extensions Covered\n\n The extensions described here focus on extending the functionality\n provided by the TLS protocol message formats. Other issues, such as\n the addition of new cipher suites, are deferred.\n\n The extension types defined in this document are:\n\n enum {\n server_name(0), max_fragment_length(1),\n client_certificate_url(2), trusted_ca_keys(3),\n truncated_hmac(4), status_request(5), (65535)\n } ExtensionType;\n\n Specifically, the extensions described in this document:\n\n - Allow TLS clients to provide to the TLS server the name of the\n server they are contacting. This functionality is desirable in\n order to facilitate secure connections to servers that host\n multiple 'virtual' servers at a single underlying network address.\n\n - Allow TLS clients and servers to negotiate the maximum fragment\n length to be sent. This functionality is desirable as a result of\n memory constraints among some clients, and bandwidth constraints\n among some access networks.\n\n - Allow TLS clients and servers to negotiate the use of client\n certificate URLs. This functionality is desirable in order to\n conserve memory on constrained clients.\n\n\n\n\n\n\n\nEastlake Standards Track [Page 3]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - Allow TLS clients to indicate to TLS servers which certification\n authority (CA) root keys they possess. This functionality is\n desirable in order to prevent multiple handshake failures\n involving TLS clients that are only able to store a small number\n of CA root keys due to memory limitations.\n\n - Allow TLS clients and servers to negotiate the use of truncated\n Message Authentication Codes (MACs). This functionality is\n desirable in order to conserve bandwidth in constrained access\n networks.\n\n - Allow TLS clients and servers to negotiate that the server sends\n the client certificate status information (e.g., an Online\n Certificate Status Protocol (OCSP) [RFC2560] response) during a\n TLS handshake. This functionality is desirable in order to avoid\n sending a Certificate Revocation List (CRL) over a constrained\n access network and therefore saving bandwidth.\n\n TLS clients and servers may use the extensions described in this\n document. The extensions are designed to be backwards compatible,\n meaning that TLS clients that support the extensions can talk to TLS\n servers that do not support the extensions, and vice versa.\n\n Note that any messages associated with these extensions that are sent\n during the TLS handshake MUST be included in the hash calculations\n involved in \"Finished\" messages.\n\n Note also that all the extensions defined in this document are\n relevant only when a session is initiated. A client that requests\n session resumption does not in general know whether the server will\n accept this request, and therefore it SHOULD send the same extensions\n as it would send if it were not attempting resumption. When a client\n includes one or more of the defined extension types in an extended\n client hello while requesting session resumption:\n\n - The server name indication extension MAY be used by the server\n when deciding whether or not to resume a session as described in\n Section 3.\n\n - If the resumption request is denied, the use of the extensions is\n negotiated as normal.\n\n - If, on the other hand, the older session is resumed, then the\n server MUST ignore the extensions and send a server hello\n containing none of the extension types. In this case, the\n functionality of these extensions negotiated during the original\n session initiation is applied to the resumed session.\n\n\n\n\nEastlake Standards Track [Page 4]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1.2. Conventions Used in This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n [RFC2119].\n\n2. Extensions to the Handshake Protocol\n\n This document specifies the use of two new handshake messages,\n \"CertificateURL\" and \"CertificateStatus\". These messages are\n described in Sections 5 and 8, respectively. The new handshake\n message structure therefore becomes:\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), certificate_url(21), certificate_status(22),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case certificate_url: CertificateURL;\n case certificate_status: CertificateStatus;\n } body;\n } Handshake;\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 5]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n3. Server Name Indication\n\n TLS does not provide a mechanism for a client to tell a server the\n name of the server it is contacting. It may be desirable for clients\n to provide this information to facilitate secure connections to\n servers that host multiple 'virtual' servers at a single underlying\n network address.\n\n In order to provide any of the server names, clients MAY include an\n extension of type \"server_name\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"ServerNameList\" where:\n\n struct {\n NameType name_type;\n select (name_type) {\n case host_name: HostName;\n } name;\n } ServerName;\n\n enum {\n host_name(0), (255)\n } NameType;\n\n opaque HostName<1..2^16-1>;\n\n struct {\n ServerName server_name_list<1..2^16-1>\n } ServerNameList;\n\n The ServerNameList MUST NOT contain more than one name of the same\n name_type. If the server understood the ClientHello extension but\n does not recognize the server name, the server SHOULD take one of two\n actions: either abort the handshake by sending a fatal-level\n unrecognized_name(112) alert or continue the handshake. It is NOT\n RECOMMENDED to send a warning-level unrecognized_name(112) alert,\n because the client's behavior in response to warning-level alerts is\n unpredictable. If there is a mismatch between the server name used\n by the client application and the server name of the credential\n chosen by the server, this mismatch will become apparent when the\n client application performs the server endpoint identification, at\n which point the client application will have to decide whether to\n proceed with the communication. TLS implementations are encouraged\n to make information available to application callers about warning-\n level alerts that were received or sent during a TLS handshake. Such\n information can be useful for diagnostic purposes.\n\n\n\n\n\nEastlake Standards Track [Page 6]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note: Earlier versions of this specification permitted multiple\n names of the same name_type. In practice, current client\n implementations only send one name, and the client cannot\n necessarily find out which name the server selected. Multiple\n names of the same name_type are therefore now prohibited.\n\n Currently, the only server names supported are DNS hostnames;\n however, this does not imply any dependency of TLS on DNS, and other\n name types may be added in the future (by an RFC that updates this\n document). The data structure associated with the host_name NameType\n is a variable-length vector that begins with a 16-bit length. For\n backward compatibility, all future data structures associated with\n new NameTypes MUST begin with a 16-bit length field. TLS MAY treat\n provided server names as opaque data and pass the names and types to\n the application.\n\n \"HostName\" contains the fully qualified DNS hostname of the server,\n as understood by the client. The hostname is represented as a byte\n string using ASCII encoding without a trailing dot. This allows the\n support of internationalized domain names through the use of A-labels\n defined in [RFC5890]. DNS hostnames are case-insensitive. The\n algorithm to compare hostnames is described in [RFC5890], Section\n 2.3.2.4.\n\n Literal IPv4 and IPv6 addresses are not permitted in \"HostName\".\n\n It is RECOMMENDED that clients include an extension of type\n \"server_name\" in the client hello whenever they locate a server by a\n supported name type.\n\n A server that receives a client hello containing the \"server_name\"\n extension MAY use the information contained in the extension to guide\n its selection of an appropriate certificate to return to the client,\n and/or other aspects of security policy. In this event, the server\n SHALL include an extension of type \"server_name\" in the (extended)\n server hello. The \"extension_data\" field of this extension SHALL be\n empty.\n\n When the server is deciding whether or not to accept a request to\n resume a session, the contents of a server_name extension MAY be used\n in the lookup of the session in the session cache. The client SHOULD\n include the same server_name extension in the session resumption\n request as it did in the full handshake that established the session.\n A server that implements this extension MUST NOT accept the request\n to resume the session if the server_name extension contains a\n different name. Instead, it proceeds with a full handshake to\n establish a new session. When resuming a session, the server MUST\n NOT include a server_name extension in the server hello.\n\n\n\nEastlake Standards Track [Page 7]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n If an application negotiates a server name using an application\n protocol and then upgrades to TLS, and if a server_name extension is\n sent, then the extension SHOULD contain the same name that was\n negotiated in the application protocol. If the server_name is\n established in the TLS session handshake, the client SHOULD NOT\n attempt to request a different server name at the application layer.\n\n4. Maximum Fragment Length Negotiation\n\n Without this extension, TLS specifies a fixed maximum plaintext\n fragment length of 2^14 bytes. It may be desirable for constrained\n clients to negotiate a smaller maximum fragment length due to memory\n limitations or bandwidth limitations.\n\n In order to negotiate smaller maximum fragment lengths, clients MAY\n include an extension of type \"max_fragment_length\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain:\n\n enum{\n 2^9(1), 2^10(2), 2^11(3), 2^12(4), (255)\n } MaxFragmentLength;\n\n whose value is the desired maximum fragment length. The allowed\n values for this field are: 2^9, 2^10, 2^11, and 2^12.\n\n Servers that receive an extended client hello containing a\n \"max_fragment_length\" extension MAY accept the requested maximum\n fragment length by including an extension of type\n \"max_fragment_length\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL contain a\n \"MaxFragmentLength\" whose value is the same as the requested maximum\n fragment length.\n\n If a server receives a maximum fragment length negotiation request\n for a value other than the allowed values, it MUST abort the\n handshake with an \"illegal_parameter\" alert. Similarly, if a client\n receives a maximum fragment length negotiation response that differs\n from the length it requested, it MUST also abort the handshake with\n an \"illegal_parameter\" alert.\n\n Once a maximum fragment length other than 2^14 has been successfully\n negotiated, the client and server MUST immediately begin fragmenting\n messages (including handshake messages) to ensure that no fragment\n larger than the negotiated length is sent. Note that TLS already\n requires clients and servers to support fragmentation of handshake\n messages.\n\n\n\n\nEastlake Standards Track [Page 8]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n The negotiated length applies for the duration of the session\n including session resumptions.\n\n The negotiated length limits the input that the record layer may\n process without fragmentation (that is, the maximum value of\n TLSPlaintext.length; see [RFC5246], Section 6.2.1). Note that the\n output of the record layer may be larger. For example, if the\n negotiated length is 2^9=512, then, when using currently defined\n cipher suites (those defined in [RFC5246] and [RFC2712]) and null\n compression, the record-layer output can be at most 805 bytes: 5\n bytes of headers, 512 bytes of application data, 256 bytes of\n padding, and 32 bytes of MAC. This means that in this event a TLS\n record-layer peer receiving a TLS record-layer message larger than\n 805 bytes MUST discard the message and send a \"record_overflow\"\n alert, without decrypting the message. When this extension is used\n with Datagram Transport Layer Security (DTLS), implementations SHOULD\n NOT generate record_overflow alerts unless the packet passes message\n authentication.\n\n5. Client Certificate URLs\n\n Without this extension, TLS specifies that when client authentication\n is performed, client certificates are sent by clients to servers\n during the TLS handshake. It may be desirable for constrained\n clients to send certificate URLs in place of certificates, so that\n they do not need to store their certificates and can therefore save\n memory.\n\n In order to negotiate sending certificate URLs to a server, clients\n MAY include an extension of type \"client_certificate_url\" in the\n (extended) client hello. The \"extension_data\" field of this\n extension SHALL be empty.\n\n (Note that it is necessary to negotiate the use of client certificate\n URLs in order to avoid \"breaking\" existing TLS servers.)\n\n Servers that receive an extended client hello containing a\n \"client_certificate_url\" extension MAY indicate that they are willing\n to accept certificate URLs by including an extension of type\n \"client_certificate_url\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n After negotiation of the use of client certificate URLs has been\n successfully completed (by exchanging hellos including\n \"client_certificate_url\" extensions), clients MAY send a\n \"CertificateURL\" message in place of a \"Certificate\" message as\n follows (see also Section 2):\n\n\n\n\nEastlake Standards Track [Page 9]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n enum {\n individual_certs(0), pkipath(1), (255)\n } CertChainType;\n\n struct {\n CertChainType type;\n URLAndHash url_and_hash_list<1..2^16-1>;\n } CertificateURL;\n\n struct {\n opaque url<1..2^16-1>;\n unint8 padding;\n opaque SHA1Hash[20];\n } URLAndHash;\n\n Here, \"url_and_hash_list\" contains a sequence of URLs and hashes.\n Each \"url\" MUST be an absolute URI reference according to [RFC3986]\n that can be immediately used to fetch the certificate(s).\n\n When X.509 certificates are used, there are two possibilities:\n\n - If CertificateURL.type is \"individual_certs\", each URL refers to a\n single DER-encoded X.509v3 certificate, with the URL for the\n client's certificate first.\n\n - If CertificateURL.type is \"pkipath\", the list contains a single\n URL referring to a DER-encoded certificate chain, using the type\n PkiPath described in Section 10.1.\n\n When any other certificate format is used, the specification that\n describes use of that format in TLS should define the encoding format\n of certificates or certificate chains, and any constraint on their\n ordering.\n\n The \"padding\" byte MUST be 0x01. It is present to make the structure\n backwards compatible.\n\n The hash corresponding to each URL is the SHA-1 hash of the\n certificate or certificate chain (in the case of X.509 certificates,\n the DER-encoded certificate or the DER-encoded PkiPath).\n\n Note that when a list of URLs for X.509 certificates is used, the\n ordering of URLs is the same as that used in the TLS Certificate\n message (see [RFC5246], Section 7.4.2), but opposite to the order in\n which certificates are encoded in PkiPath. In either case, the self-\n signed root certificate MAY be omitted from the chain, under the\n assumption that the server must already possess it in order to\n validate it.\n\n\n\nEastlake Standards Track [Page 10]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Servers receiving \"CertificateURL\" SHALL attempt to retrieve the\n client's certificate chain from the URLs and then process the\n certificate chain as usual. A cached copy of the content of any URL\n in the chain MAY be used, provided that the SHA-1 hash matches the\n hash of the cached copy.\n\n Servers that support this extension MUST support the 'http' URI\n scheme for certificate URLs and MAY support other schemes. Use of\n other schemes than 'http', 'https', or 'ftp' may create unexpected\n problems.\n\n If the protocol used is HTTP, then the HTTP server can be configured\n to use the Cache-Control and Expires directives described in\n [RFC2616] to specify whether and for how long certificates or\n certificate chains should be cached.\n\n The TLS server MUST NOT follow HTTP redirects when retrieving the\n certificates or certificate chain. The URLs used in this extension\n MUST NOT be chosen to depend on such redirects.\n\n If the protocol used to retrieve certificates or certificate chains\n returns a MIME-formatted response (as HTTP does), then the following\n MIME Content-Types SHALL be used: when a single X.509v3 certificate\n is returned, the Content-Type is \"application/pkix-cert\" [RFC2585],\n and when a chain of X.509v3 certificates is returned, the Content-\n Type is \"application/pkix-pkipath\" (Section 10.1).\n\n The server MUST check that the SHA-1 hash of the contents of the\n object retrieved from that URL (after decoding any MIME Content-\n Transfer-Encoding) matches the given hash. If any retrieved object\n does not have the correct SHA-1 hash, the server MUST abort the\n handshake with a bad_certificate_hash_value(114) alert. This alert\n is always fatal.\n\n Clients may choose to send either \"Certificate\" or \"CertificateURL\"\n after successfully negotiating the option to send certificate URLs.\n The option to send a certificate is included to provide flexibility\n to clients possessing multiple certificates.\n\n If a server is unable to obtain certificates in a given\n CertificateURL, it MUST send a fatal certificate_unobtainable(111)\n alert if it requires the certificates to complete the handshake. If\n the server does not require the certificates, then the server\n continues the handshake. The server MAY send a warning-level alert\n in this case. Clients receiving such an alert SHOULD log the alert\n and continue with the handshake if possible.\n\n\n\n\n\nEastlake Standards Track [Page 11]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n6. Trusted CA Indication\n\n Constrained clients that, due to memory limitations, possess only a\n small number of CA root keys may wish to indicate to servers which\n root keys they possess, in order to avoid repeated handshake\n failures.\n\n In order to indicate which CA root keys they possess, clients MAY\n include an extension of type \"trusted_ca_keys\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain \"TrustedAuthorities\" where:\n\n struct {\n TrustedAuthority trusted_authorities_list<0..2^16-1>;\n } TrustedAuthorities;\n\n struct {\n IdentifierType identifier_type;\n select (identifier_type) {\n case pre_agreed: struct {};\n case key_sha1_hash: SHA1Hash;\n case x509_name: DistinguishedName;\n case cert_sha1_hash: SHA1Hash;\n } identifier;\n } TrustedAuthority;\n\n enum {\n pre_agreed(0), key_sha1_hash(1), x509_name(2),\n cert_sha1_hash(3), (255)\n } IdentifierType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n Here, \"TrustedAuthorities\" provides a list of CA root key identifiers\n that the client possesses. Each CA root key is identified via\n either:\n\n - \"pre_agreed\": no CA root key identity supplied.\n\n - \"key_sha1_hash\": contains the SHA-1 hash of the CA root key. For\n Digital Signature Algorithm (DSA) and Elliptic Curve Digital\n Signature Algorithm (ECDSA) keys, this is the hash of the\n \"subjectPublicKey\" value. For RSA keys, the hash is of the big-\n endian byte string representation of the modulus without any\n initial zero-valued bytes. (This copies the key hash formats\n deployed in other environments.)\n\n\n\n\n\nEastlake Standards Track [Page 12]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - \"x509_name\": contains the DER-encoded X.509 DistinguishedName of\n the CA.\n\n - \"cert_sha1_hash\": contains the SHA-1 hash of a DER-encoded\n Certificate containing the CA root key.\n\n Note that clients may include none, some, or all of the CA root keys\n they possess in this extension.\n\n Note also that it is possible that a key hash or a Distinguished Name\n alone may not uniquely identify a certificate issuer (for example, if\n a particular CA has multiple key pairs). However, here we assume\n this is the case following the use of Distinguished Names to identify\n certificate issuers in TLS.\n\n The option to include no CA root keys is included to allow the client\n to indicate possession of some pre-defined set of CA root keys.\n\n Servers that receive a client hello containing the \"trusted_ca_keys\"\n extension MAY use the information contained in the extension to guide\n their selection of an appropriate certificate chain to return to the\n client. In this event, the server SHALL include an extension of type\n \"trusted_ca_keys\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n7. Truncated HMAC\n\n Currently defined TLS cipher suites use the MAC construction HMAC\n [RFC2104] to authenticate record-layer communications. In TLS, the\n entire output of the hash function is used as the MAC tag. However,\n it may be desirable in constrained environments to save bandwidth by\n truncating the output of the hash function to 80 bits when forming\n MAC tags.\n\n In order to negotiate the use of 80-bit truncated HMAC, clients MAY\n include an extension of type \"truncated_hmac\" in the extended client\n hello. The \"extension_data\" field of this extension SHALL be empty.\n\n Servers that receive an extended hello containing a \"truncated_hmac\"\n extension MAY agree to use a truncated HMAC by including an extension\n of type \"truncated_hmac\", with empty \"extension_data\", in the\n extended server hello.\n\n Note that if new cipher suites are added that do not use HMAC, and\n the session negotiates one of these cipher suites, this extension\n will have no effect. It is strongly recommended that any new cipher\n suites using other MACs consider the MAC size an integral part of the\n\n\n\n\nEastlake Standards Track [Page 13]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n cipher suite definition, taking into account both security and\n bandwidth considerations.\n\n If HMAC truncation has been successfully negotiated during a TLS\n handshake, and the negotiated cipher suite uses HMAC, both the client\n and the server pass this fact to the TLS record layer along with the\n other negotiated security parameters. Subsequently during the\n session, clients and servers MUST use truncated HMACs, calculated as\n specified in [RFC2104]. That is, SecurityParameters.mac_length is 10\n bytes, and only the first 10 bytes of the HMAC output are transmitted\n and checked. Note that this extension does not affect the\n calculation of the pseudo-random function (PRF) as part of\n handshaking or key derivation.\n\n The negotiated HMAC truncation size applies for the duration of the\n session including session resumptions.\n\n8. Certificate Status Request\n\n Constrained clients may wish to use a certificate-status protocol\n such as OCSP [RFC2560] to check the validity of server certificates,\n in order to avoid transmission of CRLs and therefore save bandwidth\n on constrained networks. This extension allows for such information\n to be sent in the TLS handshake, saving roundtrips and resources.\n\n In order to indicate their desire to receive certificate status\n information, clients MAY include an extension of type\n \"status_request\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"CertificateStatusRequest\" where:\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n } request;\n } CertificateStatusRequest;\n\n enum { ocsp(1), (255) } CertificateStatusType;\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n\n\n\nEastlake Standards Track [Page 14]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n In the OCSPStatusRequest, the \"ResponderIDs\" provides a list of OCSP\n responders that the client trusts. A zero-length \"responder_id_list\"\n sequence has the special meaning that the responders are implicitly\n known to the server, e.g., by prior arrangement. \"Extensions\" is a\n DER encoding of OCSP request extensions.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC2560]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a zero-length ASN.1 SEQUENCE, which is not\n valid for the \"Extensions\" type).\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement).\n\n Servers that receive a client hello containing the \"status_request\"\n extension MAY return a suitable certificate status response to the\n client along with their certificate. If OCSP is requested, they\n SHOULD use the information contained in the extension when selecting\n an OCSP responder and SHOULD include request_extensions in the OCSP\n request.\n\n Servers return a certificate response along with their certificate by\n sending a \"CertificateStatus\" message immediately after the\n \"Certificate\" message (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message, then the server MUST have included an\n extension of type \"status_request\" with empty \"extension_data\" in the\n extended server hello. The \"CertificateStatus\" message is conveyed\n using the handshake message type \"certificate_status\" as follows (see\n also Section 2):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<1..2^24-1>;\n\n An \"ocsp_response\" contains a complete, DER-encoded OCSP response\n (using the ASN.1 type OCSPResponse defined in [RFC2560]). Only one\n OCSP response may be sent.\n\n\n\n\nEastlake Standards Track [Page 15]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if has received a \"status_request\" extension in the\n client hello message and has sent a \"status_request\" extension in the\n server hello message.\n\n Note in addition that a server MUST NOT send the \"CertificateStatus\"\n message unless it received a \"status_request\" extension in the client\n hello message and sent a \"status_request\" extension in the server\n hello message.\n\n Clients requesting an OCSP response and receiving an OCSP response in\n a \"CertificateStatus\" message MUST check the OCSP response and abort\n the handshake if the response is not satisfactory with\n bad_certificate_status_response(113) alert. This alert is always\n fatal.\n\n9. Error Alerts\n\n Four new error alerts are defined for use with the TLS extensions\n defined in this document. To avoid \"breaking\" existing clients and\n servers, these alerts MUST NOT be sent unless the sending party has\n received an extended hello message from the party they are\n communicating with. These error alerts are conveyed using the\n following syntax. The new alerts are the last four, as indicated by\n the comments on the same line as the error alert number.\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n /* 41 is not defined, for historical reasons */\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction(60),\n protocol_version(70),\n insufficient_security(71),\n\n\n\nEastlake Standards Track [Page 16]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n certificate_unobtainable(111), /* new */\n unrecognized_name(112), /* new */\n bad_certificate_status_response(113), /* new */\n bad_certificate_hash_value(114), /* new */\n (255)\n } AlertDescription;\n\n \"certificate_unobtainable\" is described in Section 5.\n \"unrecognized_name\" is described in Section 3.\n \"bad_certificate_status_response\" is described in Section 8.\n \"bad_certificate_hash_value\" is described in Section 5.\n\n10. IANA Considerations\n\n IANA Considerations for TLS extensions and the creation of a registry\n are covered in Section 12 of [RFC5246] except for the registration of\n MIME type application/pkix-pkipath, which appears below.\n\n The IANA TLS extensions and MIME type application/pkix-pkipath\n registry entries that reference RFC 4366 have been updated to\n reference this document.\n\n10.1. pkipath MIME Type Registration\n\n MIME media type name: application\n MIME subtype name: pkix-pkipath\n Required parameters: none\n\n Optional parameters: version (default value is \"1\")\n\n Encoding considerations:\n Binary; this MIME type is a DER encoding of the ASN.1 type\n PkiPath, defined as follows:\n PkiPath ::= SEQUENCE OF Certificate\n PkiPath is used to represent a certification path. Within the\n sequence, the order of certificates is such that the subject of\n the first certificate is the issuer of the second certificate,\n etc.\n This is identical to the definition published in [X509-4th-TC1];\n note that it is different from that in [X509-4th].\n\n All Certificates MUST conform to [RFC5280]. (This should be\n interpreted as a requirement to encode only PKIX-conformant\n certificates using this type. It does not necessarily require\n\n\n\nEastlake Standards Track [Page 17]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n that all certificates that are not strictly PKIX-conformant must\n be rejected by relying parties, although the security consequences\n of accepting any such certificates should be considered\n carefully.)\n\n DER (as opposed to BER) encoding MUST be used. If this type is\n sent over a 7-bit transport, base64 encoding SHOULD be used.\n\n Security considerations:\n The security considerations of [X509-4th] and [RFC5280] (or any\n updates to them) apply, as well as those of any protocol that uses\n this type (e.g., TLS).\n\n Note that this type only specifies a certificate chain that can be\n assessed for validity according to the relying party's existing\n configuration of trusted CAs; it is not intended to be used to\n specify any change to that configuration.\n\n Interoperability considerations:\n No specific interoperability problems are known with this type,\n but for recommendations relating to X.509 certificates in general,\n see [RFC5280].\n\n Published specification: This document and [RFC5280].\n\n Applications that use this media type:\n TLS. It may also be used by other protocols or for general\n interchange of PKIX certificate chains.\n\n Additional information:\n Magic number(s): DER-encoded ASN.1 can be easily recognized.\n Further parsing is required to distinguish it from other ASN.1\n types.\n File extension(s): .pkipath\n Macintosh File Type Code(s): not specified\n\n Person & email address to contact for further information:\n Magnus Nystrom <[redacted-email]>\n\n Intended usage: COMMON\n\n Change controller: IESG <[redacted-email]>\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 18]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n10.2. Reference for TLS Alerts, TLS HandshakeTypes, and ExtensionTypes\n\n The following values in the TLS Alert Registry have been updated to\n reference this document:\n\n 111 certificate_unobtainable\n 112 unrecognized_name\n 113 bad_certificate_status_response\n 114 bad_certificate_hash_value\n\n The following values in the TLS HandshakeType Registry have been\n updated to reference this document:\n\n 21 certificate_url\n 22 certificate_status\n\n The following ExtensionType values have been updated to reference\n this document:\n\n 0 server_name\n 1 max_fragment_length\n 2 client_certificate_url\n 3 trusted_ca_keys\n 4 truncated_hmac\n 5 status_request\n\n11. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security Considerations for particular extensions\n specified in this document are given below.\n\n In general, implementers should continue to monitor the state of the\n art and address any weaknesses identified.\n\n11.1. Security Considerations for server_name\n\n If a single server hosts several domains, then clearly it is\n necessary for the owners of each domain to ensure that this satisfies\n their security needs. Apart from this, server_name does not appear\n to introduce significant security issues.\n\n Since it is possible for a client to present a different server_name\n in the application protocol, application server implementations that\n rely upon these names being the same MUST check to make sure the\n client did not present a different name in the application protocol.\n\n\n\n\n\nEastlake Standards Track [Page 19]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Implementations MUST ensure that a buffer overflow does not occur,\n whatever the values of the length fields in server_name.\n\n11.2. Security Considerations for max_fragment_length\n\n The maximum fragment length takes effect immediately, including for\n handshake messages. However, that does not introduce any security\n complications that are not already present in TLS, since TLS requires\n implementations to be able to handle fragmented handshake messages.\n\n Note that, as described in Section 4, once a non-null cipher suite\n has been activated, the effective maximum fragment length depends on\n the cipher suite and compression method, as well as on the negotiated\n max_fragment_length. This must be taken into account when sizing\n buffers and checking for buffer overflow.\n\n11.3. Security Considerations for client_certificate_url\n\n Support for client_certificate_url involves the server's acting as a\n client in another URI-scheme-dependent protocol. The server\n therefore becomes subject to many of the same security concerns that\n clients of the URI scheme are subject to, with the added concern that\n the client can attempt to prompt the server to connect to some\n (possibly weird-looking) URL.\n\n In general, this issue means that an attacker might use the server to\n indirectly attack another host that is vulnerable to some security\n flaw. It also introduces the possibility of denial-of-service\n attacks in which an attacker makes many connections to the server,\n each of which results in the server's attempting a connection to the\n target of the attack.\n\n Note that the server may be behind a firewall or otherwise able to\n access hosts that would not be directly accessible from the public\n Internet. This could exacerbate the potential security and denial-\n of-service problems described above, as well as allow the existence\n of internal hosts to be confirmed when they would otherwise be\n hidden.\n\n The detailed security concerns involved will depend on the URI\n schemes supported by the server. In the case of HTTP, the concerns\n are similar to those that apply to a publicly accessible HTTP proxy\n server. In the case of HTTPS, loops and deadlocks may be created,\n and this should be addressed. In the case of FTP, attacks arise that\n are similar to FTP bounce attacks.\n\n\n\n\n\n\nEastlake Standards Track [Page 20]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n As a result of this issue, it is RECOMMENDED that the\n client_certificate_url extension should have to be specifically\n enabled by a server administrator, rather than be enabled by default.\n It is also RECOMMENDED that URI schemes be enabled by the\n administrator individually, and only a minimal set of schemes be\n enabled. Unusual protocols that offer limited security or whose\n security is not well understood SHOULD be avoided.\n\n As discussed in [RFC3986], URLs that specify ports other than the\n default may cause problems, as may very long URLs (which are more\n likely to be useful in exploiting buffer overflow bugs).\n\n This extension continues to use SHA-1 (as in RFC 4366) and does not\n provide algorithm agility. The property required of SHA-1 in this\n case is second pre-image resistance, not collision resistance.\n Furthermore, even if second pre-image attacks against SHA-1 are found\n in the future, an attack against client_certificate_url would require\n a second pre-image that is accepted as a valid certificate by the\n server and contains the same public key.\n\n Also note that HTTP caching proxies are common on the Internet, and\n some proxies do not check for the latest version of an object\n correctly. If a request using HTTP (or another caching protocol)\n goes through a misconfigured or otherwise broken proxy, the proxy may\n return an out-of-date response.\n\n11.4. Security Considerations for trusted_ca_keys\n\n Potentially, the CA root keys a client possesses could be regarded as\n confidential information. As a result, the CA root key indication\n extension should be used with care.\n\n The use of the SHA-1 certificate hash alternative ensures that each\n certificate is specified unambiguously. This context does not\n require a cryptographic hash function, so the use of SHA-1 is\n considered acceptable, and no algorithm agility is provided.\n\n11.5. Security Considerations for truncated_hmac\n\n It is possible that truncated MACs are weaker than \"un-truncated\"\n MACs. However, no significant weaknesses are currently known or\n expected to exist for HMAC with MD5 or SHA-1, truncated to 80 bits.\n\n Note that the output length of a MAC need not be as long as the\n length of a symmetric cipher key, since forging of MAC values cannot\n be done off-line: in TLS, a single failed MAC guess will cause the\n immediate termination of the TLS session.\n\n\n\n\nEastlake Standards Track [Page 21]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Since the MAC algorithm only takes effect after all handshake\n messages that affect extension parameters have been authenticated by\n the hashes in the Finished messages, it is not possible for an active\n attacker to force negotiation of the truncated HMAC extension where\n it would not otherwise be used (to the extent that the handshake\n authentication is secure). Therefore, in the event that any security\n problems were found with truncated HMAC in the future, if either the\n client or the server for a given session were updated to take the\n problem into account, it would be able to veto use of this extension.\n\n11.6. Security Considerations for status_request\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC2560] for further details.\n\n12. Normative References\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC:\n Keyed-Hashing for Message Authentication\", RFC 2104,\n February 1997.\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and\n C. Adams, \"X.509 Internet Public Key Infrastructure\n Online Certificate Status Protocol - OCSP\", RFC 2560,\n June 1999.\n\n [RFC2585] Housley, R. and P. Hoffman, \"Internet X.509 Public Key\n Infrastructure Operational Protocols: FTP and HTTP\",\n RFC 2585, May 1999.\n\n [RFC2616] Fielding, R., Gettys, J., Mogul, J., Frystyk, H.,\n Masinter, L., Leach, P., and T. Berners-Lee,\n \"Hypertext Transfer Protocol -- HTTP/1.1\", RFC 2616,\n June 1999.\n\n [RFC3986] Berners-Lee, T., Fielding, R., and L. Masinter,\n \"Uniform Resource Identifier (URI): Generic Syntax\",\n STD 66, RFC 3986, January 2005.\n\n\n\nEastlake Standards Track [Page 22]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246, August\n 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation\n List (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC5890] Klensin, J., \"Internationalized Domain Names for\n Applications (IDNA): Definitions and Document\n Framework\", RFC 5890, August 2010.\n\n13. Informative References\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [X509-4th] ITU-T Recommendation X.509 (2000) | ISO/IEC\n 9594-8:2001, \"Information Systems - Open Systems\n Interconnection - The Directory: Public key and\n attribute certificate frameworks\".\n\n [X509-4th-TC1] ITU-T Recommendation X.509(2000) Corrigendum 1(2001) |\n ISO/IEC 9594-8:2001/Cor.1:2002, Technical Corrigendum\n 1 to ISO/IEC 9594:8:2001.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 23]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix A. Changes from RFC 4366\n\n The significant changes between RFC 4366 and this document are\n described below.\n\n RFC 4366 described both general extension mechanisms (for the TLS\n handshake and client and server hellos) as well as specific\n extensions. RFC 4366 was associated with RFC 4346, TLS 1.1. The\n client and server hello extension mechanisms have been moved into RFC\n 5246, TLS 1.2, so this document, which is associated with RFC 5246,\n includes only the handshake extension mechanisms and the specific\n extensions from RFC 4366. RFC 5246 also specifies the unknown\n extension error and new extension specification considerations, so\n that material has been removed from this document.\n\n The Server Name extension now specifies only ASCII representation,\n eliminating UTF-8. It is provided that the ServerNameList can\n contain more than only one name of any particular name_type. If a\n server name is provided but not recognized, the server should either\n continue the handshake without an error or send a fatal error.\n Sending a warning-level message is not recommended because client\n behavior will be unpredictable. Provision was added for the user\n using the server_name extension in deciding whether or not to resume\n a session. Furthermore, this extension should be the same in a\n session resumption request as it was in the full handshake that\n established the session. Such a resumption request must not be\n accepted if the server_name extension is different, but instead a\n full handshake must be done to possibly establish a new session.\n\n The Client Certificate URLs extension has been changed to make the\n presence of a hash mandatory.\n\n For the case of DTLS, the requirement to report an overflow of the\n negotiated maximum fragment length is made conditional on passing\n authentication.\n\n TLS servers are now prohibited from following HTTP redirects when\n retrieving certificates.\n\n The material was also re-organized in minor ways. For example,\n information as to which errors are fatal is moved from the \"Error\n Alerts\" section to the individual extension specifications.\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 24]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix B. Acknowledgements\n\n This document is based on material from RFC 4366 for which the\n authors were S. Blake-Wilson, M. Nystrom, D. Hopwood, J. Mikkelsen,\n and T. Wright. Other contributors include Joseph Salowey, Alexey\n Melnikov, Peter Saint-Andre, and Adrian Farrel.\n\nAuthor's Address\n\n Donald Eastlake 3rd\n Huawei\n 155 Beaver Street\n Milford, MA 01757 USA\n\n Phone: +1-508-333-2270\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 25]", "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}, {"draft_references": [], "excluded_rfc_relation_targets": [], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6961 June 2013", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "6961"}], "kind": "rfc", "occurred_at": "2013-06-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 1, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db", "record_id": "ietf:rfc:6961", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "rfc_number": 6961, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db", "source_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "temporal_semantics": "source_event_time", "text": "Internet Engineering Task Force (IETF) Y. Pettersen\nRequest for Comments: 6961 June 2013\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\n\nAbstract\n\n This document defines the Transport Layer Security (TLS) Certificate\n Status Version 2 Extension to allow clients to specify and support\n several certificate status methods. (The use of the Certificate\n Status extension is commonly referred to as \"OCSP stapling\".) Also\n defined is a new method based on the Online Certificate Status\n Protocol (OCSP) that servers can use to provide status information\n about not only the server's own certificate but also the status of\n intermediate certificates in the chain.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6961.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 1]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nCopyright Notice\n\n Copyright (c) 2013 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n1. Introduction\n\n The Transport Layer Security (TLS) Extension [RFC6066] framework\n defines, among other extensions, the Certificate Status extension\n (also referred to as \"OCSP stapling\") that clients can use to request\n the server's copy of the current status of its certificate. The\n benefits of this extension include a reduced number of roundtrips and\n network delays for the client to verify the status of the server's\n certificate and a reduced load on the certificate issuer's status\n response servers, thus solving a problem that can become significant\n when the issued certificate is presented by a frequently visited\n server.\n\n There are two problems with the existing Certificate Status\n extension. First, it does not provide functionality to request the\n status information about intermediate Certification Authority (CA)\n certificates, which means the client has to request status\n information through other methods, such as Certificate Revocation\n Lists (CRLs), introducing further delays. Second, the current format\n of the extension and requirements in the TLS protocol prevent a\n client from offering the server multiple status methods.\n\n\n\nPettersen Standards Track [Page 2]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n Many CAs are now issuing intermediate CA certificates that not only\n specify the publication point for their CRLs in a CRL Distribution\n Point [RFC5280] but also specify a URL for their OCSP [RFC6960]\n server in Authority Information Access [RFC5280]. Given that\n client-cached CRLs are frequently out of date, clients would benefit\n from using OCSP to access up-to-date status information about\n intermediate CA certificates. The benefit to the issuing CA is less\n clear, as providing the bandwidth for the OCSP responder can be\n costly, especially for CAs with many high-traffic subscriber sites,\n and this cost is a concern for many CAs. There are cases where OCSP\n requests for a single high-traffic site caused significant network\n problems for the issuing CA.\n\n Clients will benefit from the TLS server providing certificate status\n information regardless of type, not just for the server certificate\n but also for the intermediate CA certificates. Combining the status\n checks into one extension will reduce the roundtrips needed to\n complete the handshake by the client to just those needed for\n negotiating the TLS connection. Also, for the Certification\n Authorities, the load on their servers will depend on the number of\n certificates they have issued, not on the number of visitors to those\n sites. Additionally, using this extension significantly reduces\n privacy concerns around the clients informing the certificate issuer\n about which sites they are visiting.\n\n For such a new system to be introduced seamlessly, clients need to be\n able to indicate support for the existing OCSP Certificate Status\n method and a new multiple-OCSP mode.\n\n Unfortunately, the definition of the Certificate Status extension\n only allows a single Certificate Status extension to be defined in a\n single extension record in the handshake, and the TLS protocol\n [RFC5246] only allows a single record in the extension list for any\n given extension. This means that it is not possible for clients to\n indicate support for new methods while still supporting older\n methods, which would cause problems for interoperability between\n newer clients and older servers. This will not just be an issue for\n the multiple status request mode proposed above but also for any\n other future status methods that might be introduced. This will be\n true not just for the current PKIX infrastructure [RFC5280] but also\n for alternative PKI structures.\n\n The solution to this problem is to define a new extension,\n \"status_request_v2\", with an extended format that allows the client\n to indicate support for multiple status request methods. This is\n implemented using a list of CertificateStatusRequestItemV2 records in\n the extension record. As the server will select the single status\n\n\n\n\nPettersen Standards Track [Page 3]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n method based on the selected cipher suite and the certificate\n presented, no significant changes are needed in the server's\n extension format.\n\n1.1. Requirements Language\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [RFC2119].\n\n1.2. Presentation Language\n\n This document defines protocol structures using the same conventions\n and presentation language as defined in Section 4 of [RFC5246].\n\n2. Multiple Certificate Status Extension\n\n2.1. New Extension\n\n The extension defined by this document is indicated by\n \"status_request_v2\" in the ExtensionType enum (originally defined by\n [RFC6066]), which uses the following value:\n\n enum {\n status_request_v2(17), (65535)\n } ExtensionType;\n\n2.2. Multiple Certificate Status Request Record\n\n Clients that support a certificate status protocol like OCSP may send\n the \"status_request_v2\" extension to the server in order to use the\n TLS handshake to transfer such data instead of downloading it through\n separate connections. When using this extension, the\n \"extension_data\" field (defined in Section 7.4.1.4 of [RFC5246]) of\n the extension SHALL contain a CertificateStatusRequestListV2 where:\n\n struct {\n CertificateStatusType status_type;\n uint16 request_length; /* Length of request field in bytes */\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n case ocsp_multi: OCSPStatusRequest;\n } request;\n } CertificateStatusRequestItemV2;\n\n enum { ocsp(1), ocsp_multi(2), (255) } CertificateStatusType;\n\n\n\n\n\nPettersen Standards Track [Page 4]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n struct {\n CertificateStatusRequestItemV2\n certificate_status_req_list<1..2^16-1>;\n } CertificateStatusRequestListV2;\n\n In the OCSPStatusRequest (originally defined by [RFC6066]), the\n \"ResponderID\" provides a list of OCSP responders that the client\n trusts. A zero-length \"responder_id_list\" sequence has the special\n meaning that the responders are implicitly known to the server, e.g.,\n by prior arrangement, or are identified by the certificates used by\n the server. \"Extensions\" is a DER encoding [X.690] of the OCSP\n request extensions, and if the server supports the forwarding of OCSP\n request extensions, this value MUST be forwarded without\n modification.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC6960]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a DER-encoded zero-length ASN.1 SEQUENCE,\n which is not valid for the \"Extensions\" type).\n\n Servers that support a client's selection of responders using the\n ResponderID field could implement this selection by matching the\n responder ID values from the client's list with the ResponderIDs of\n known OCSP responders, either by using a binary compare of the values\n or a hash calculation and compare method.\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement). This\n has been clarified in [RFC6960].\n\n The items in the list of CertificateStatusRequestItemV2 entries are\n ordered according to the client's preference (favorite choice first).\n\n A server that receives a client hello containing the\n \"status_request_v2\" extension MAY return a suitable certificate\n status response message to the client along with the server's\n\n\n\nPettersen Standards Track [Page 5]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n certificate message. If OCSP is requested, it SHOULD use the\n information contained in the extension when selecting an OCSP\n responder and SHOULD include request_extensions in the OCSP request.\n\n The server returns a certificate status response along with its\n certificate by sending a \"CertificateStatus\" message (originally\n defined by [RFC6066]) immediately after the \"Certificate\" message\n (Section 7.4.2 of [RFC5246]) (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message in response to a \"status_request_v2\"\n request, then the server MUST have included an extension of type\n \"status_request_v2\" with empty \"extension_data\" in the extended\n server hello.\n\n The \"CertificateStatus\" message is conveyed using the handshake\n message type \"certificate_status\" (defined in [RFC6066]) as follows\n (updated from the definition in [RFC6066]):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n case ocsp_multi: OCSPResponseList;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<0..2^24-1>;\n\n struct {\n OCSPResponse ocsp_response_list<1..2^24-1>;\n } OCSPResponseList;\n\n An \"OCSPResponse\" element (originally defined by [RFC6066]) contains\n a complete, DER-encoded OCSP response (using the ASN.1 [X.680] type\n OCSPResponse defined in [RFC6960]). Only one OCSP response, with a\n length of at least one byte, may be sent for status_type \"ocsp\".\n\n An \"ocsp_response_list\" contains a list of \"OCSPResponse\" elements,\n as specified above, each containing the OCSP response for the\n matching corresponding certificate in the server's Certificate TLS\n handshake message. That is, the first entry is the OCSP response for\n the first certificate in the Certificate list, the second entry is\n the response for the second certificate, and so on. The list MAY\n contain fewer OCSP responses than there were certificates in the\n Certificate handshake message, but there MUST NOT be more responses\n than there were certificates in the list. Individual elements of the\n list MAY have a length of 0 (zero) bytes if the server does not have\n the OCSP response for that particular certificate stored, in which\n\n\n\nPettersen Standards Track [Page 6]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n case the client MUST act as if a response was not received for that\n particular certificate. If the client receives a\n \"ocsp_response_list\" that does not contain a response for one or more\n of the certificates in the completed certificate chain, the client\n SHOULD attempt to validate the certificate using an alternative\n retrieval method, such as downloading the relevant CRL; OCSP SHOULD\n in this situation only be used for the end-entity certificate, not\n intermediate CA certificates, for reasons stated above.\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if it has received a \"status_request_v2\" extension in\n the client hello message and has sent a \"status_request_v2\" extension\n in the server hello message. Additionally, note that a server MUST\n NOT send the \"CertificateStatus\" message unless it received either a\n \"status_request\" or \"status_request_v2\" extension in the client hello\n message and sent a corresponding \"status_request\" or\n \"status_request_v2\" extension in the server hello message.\n\n Clients requesting an OCSP response and receiving one or more OCSP\n responses in a \"CertificateStatus\" message MUST check the OCSP\n response(s) and abort the handshake if the response is a \"revoked\"\n status or other unacceptable responses (as determined by client\n policy) with a bad_certificate_status_response(113) alert. This\n alert is always fatal.\n\n If the OCSP response received from the server does not result in a\n definite \"good\" or \"revoked\" status, it is inconclusive. A TLS\n client in such a case MAY check the validity of the server\n certificate through other means, e.g., by directly querying the\n certificate issuer. If such processing still results in an\n inconclusive response, then the application using the TLS connection\n will have to decide whether to close the connection or not. Note\n that this problem cannot be decided by the generic TLS client code\n without information from the application. If the application doesn't\n provide any such information, then the client MUST abort the\n connection, since the server certificate has not been sufficiently\n validated.\n\n An example of where the application might wish to continue is with\n EAP-TLS (Extensible Authentication Protocol - TLS), where the\n application can use another mechanism to check the status of a\n certificate once it obtains network access. In this case, the\n application could have the client continue with the handshake, but it\n MUST NOT disclose a username and password until it has fully\n validated the server certificate.\n\n\n\n\n\n\nPettersen Standards Track [Page 7]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n3. IANA Considerations\n\n Section 2.1 defines the new TLS extension status_request_v2 (17)\n enum, which has been added to the \"ExtensionType Values\" list in the\n IANA \"Transport Layer Security (TLS) Extensions\" registry.\n\n Section 2.2 describes a TLS CertificateStatusType registry that is\n now maintained by IANA. The \"TLS Certificate Status Types\" registry\n has been created under the \"Transport Layer Security (TLS)\n Extensions\" registry. CertificateStatusType values are to be\n assigned via IETF Review as defined in [RFC5226]. The initial\n registry corresponds to the definition of \"CertificateStatusType\" in\n Section 2.2.\n\n Value Description Reference\n -----------------------------------------\n 0 Reserved [RFC6961]\n 1 ocsp [RFC6066] [RFC6961]\n 2 ocsp_multi [RFC6961]\n 3-255 Unassigned\n\n4. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security considerations for the particular extension\n specified in this document are given below. In general, implementers\n should continue to monitor the state of the art and address any\n weaknesses identified.\n\n4.1. Security Considerations for status_request_v2\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC6960] for further details.\n\n This extension allows the client to send arbitrary data to the\n server. The server implementers need to handle such data carefully\n to avoid introducing security vulnerabilities.\n\n The security considerations of [RFC6960] apply to OCSP requests and\n responses.\n\n\n\n\nPettersen Standards Track [Page 8]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n5. Acknowledgements\n\n This document is based on [RFC6066], authored by Donald Eastlake 3rd.\n\n6. References\n\n6.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246, August 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC6066] Eastlake, D., \"Transport Layer Security (TLS) Extensions:\n Extension Definitions\", RFC 6066, January 2011.\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, June 2013.\n\n [X.680] ITU-T Recommendation X.680 (2008) | ISO/IEC 8824-1:2008,\n \"Abstract Syntax Notation One (ASN.1): Specification of\n basic notation\", November 2008.\n\n [X.690] ITU-T Recommendation X.690 (2008) | ISO/IEC 8825-1:2008,\n \"ASN.1 encoding rules: Specification of Basic Encoding\n Rules (BER), Canonical Encoding Rules (CER) and\n Distinguished Encoding Rules (DER)\", November 2008.\n\n6.2. Informative References\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and C.\n Adams, \"X.509 Internet Public Key Infrastructure Online\n Certificate Status Protocol - OCSP\", RFC 2560, June 1999.\n\n\n\n\n\n\nPettersen Standards Track [Page 9]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nAuthor's Address\n\n Yngve N. Pettersen\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 10]", "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"draft_name": "draft-ietf-tls-tls13", "facts": [{"char_end": 454, "char_start": 431, "evidence_quote": "draft-ietf-tls-tls13-28", "fact_id": "draft_revision", "field": "draft_revision", "raw_char_end": 458, "raw_char_start": 435, "value": "draft-ietf-tls-tls13-28"}], "kind": "draft_revision", "occurred_at": "2018-03-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 85, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6", "record_id": "ietf:draft:draft-ietf-tls-tls13-28", "retrieval_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "revision": "28", "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6", "source_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "temporal_semantics": "source_event_time", "text": "Network Working Group E. Rescorla\nInternet-Draft RTFM, Inc.\nObsoletes: 5077, 5246, 6961 (if March 20, 2018\n approved)\nUpdates: 4492, 5705, 6066 (if approved)\nIntended status: Standards Track\nExpires: September 21, 2018\n\n\n The Transport Layer Security (TLS) Protocol Version 1.3\n draft-ietf-tls-tls13-28\n\nAbstract\n\n This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol. TLS allows client/server applications to communicate\n over the Internet in a way that is designed to prevent eavesdropping,\n tampering, and message forgery.\n\n This document updates RFCs 4492, 5705, and 6066 and it obsoletes RFCs\n 5077, 5246, and 6961. This document also specifies new requirements\n for TLS 1.2 implementations.\n\nStatus of This Memo\n\n This Internet-Draft is submitted in full conformance with the\n provisions of BCP 78 and BCP 79.\n\n Internet-Drafts are working documents of the Internet Engineering\n Task Force (IETF). Note that other groups may also distribute\n working documents as Internet-Drafts. The list of current Internet-\n Drafts is at http://datatracker.ietf.org/drafts/current/.\n\n Internet-Drafts are draft documents valid for a maximum of six months\n and may be updated, replaced, or obsoleted by other documents at any\n time. It is inappropriate to use Internet-Drafts as reference\n material or to cite them other than as \"work in progress.\"\n\n This Internet-Draft will expire on September 21, 2018.\n\nCopyright Notice\n\n Copyright (c) 2018 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n\n\n\nRescorla Expires September 21, 2018 [Page 1]\n\f\nInternet-Draft TLS March 2018\n\n\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 5\n 1.1. Conventions and Terminology . . . . . . . . . . . . . . . 6\n 1.2. Change Log . . . . . . . . . . . . . . . . . . . . . . . 7\n 1.3. Major Differences from TLS 1.2 . . . . . . . . . . . . . 16\n 1.4. Updates Affecting TLS 1.2 . . . . . . . . . . . . . . . . 17\n 2. Protocol Overview . . . . . . . . . . . . . . . . . . . . . . 18\n 2.1. Incorrect DHE Share . . . . . . . . . . . . . . . . . . . 21\n 2.2. Resumption and Pre-Shared Key (PSK) . . . . . . . . . . . 22\n 2.3. 0-RTT Data . . . . . . . . . . . . . . . . . . . . . . . 24\n 3. Presentation Language . . . . . . . . . . . . . . . . . . . . 26\n 3.1. Basic Block Size . . . . . . . . . . . . . . . . . . . . 26\n 3.2. Miscellaneous . . . . . . . . . . . . . . . . . . . . . . 26\n 3.3. Numbers . . . . . . . . . . . . . . . . . . . . . . . . . 27\n 3.4. Vectors . . . . . . . . . . . . . . . . . . . . . . . . . 27\n 3.5. Enumerateds . . . . . . . . . . . . . . . . . . . . . . . 28\n 3.6. Constructed Types . . . . . . . . . . . . . . . . . . . . 29\n 3.7. Constants . . . . . . . . . . . . . . . . . . . . . . . . 29\n 3.8. Variants . . . . . . . . . . . . . . . . . . . . . . . . 30\n 4. Handshake Protocol . . . . . . . . . . . . . . . . . . . . . 31\n 4.1. Key Exchange Messages . . . . . . . . . . . . . . . . . . 32\n 4.1.1. Cryptographic Negotiation . . . . . . . . . . . . . . 32\n 4.1.2. Client Hello . . . . . . . . . . . . . . . . . . . . 33\n 4.1.3. Server Hello . . . . . . . . . . . . . . . . . . . . 36\n 4.1.4. Hello Retry Request . . . . . . . . . . . . . . . . . 38\n 4.2. Extensions . . . . . . . . . . . . . . . . . . . . . . . 40\n 4.2.1. Supported Versions . . . . . . . . . . . . . . . . . 43\n 4.2.2. Cookie . . . . . . . . . . . . . . . . . . . . . . . 45\n\n\n\nRescorla Expires September 21, 2018 [Page 2]\n\f\nInternet-Draft TLS March 2018\n\n\n 4.2.3. Signature Algorithms . . . . . . . . . . . . . . . . 46\n 4.2.4. Certificate Authorities . . . . . . . . . . . . . . . 50\n 4.2.5. OID Filters . . . . . . . . . . . . . . . . . . . . . 50\n 4.2.6. Post-Handshake Client Authentication . . . . . . . . 51\n 4.2.7. Negotiated Groups . . . . . . . . . . . . . . . . . . 52\n 4.2.8. Key Share . . . . . . . . . . . . . . . . . . . . . . 53\n 4.2.9. Pre-Shared Key Exchange Modes . . . . . . . . . . . . 56\n 4.2.10. Early Data Indication . . . . . . . . . . . . . . . . 57\n 4.2.11. Pre-Shared Key Extension . . . . . . . . . . . . . . 60\n 4.3. Server Parameters . . . . . . . . . . . . . . . . . . . . 63\n 4.3.1. Encrypted Extensions . . . . . . . . . . . . . . . . 63\n 4.3.2. Certificate Request . . . . . . . . . . . . . . . . . 64\n 4.4. Authentication Messages . . . . . . . . . . . . . . . . . 65\n 4.4.1. The Transcript Hash . . . . . . . . . . . . . . . . . 66\n 4.4.2. Certificate . . . . . . . . . . . . . . . . . . . . . 67\n 4.4.3. Certificate Verify . . . . . . . . . . . . . . . . . 72\n 4.4.4. Finished . . . . . . . . . . . . . . . . . . . . . . 74\n 4.5. End of Early Data . . . . . . . . . . . . . . . . . . . . 75\n 4.6. Post-Handshake Messages . . . . . . . . . . . . . . . . . 76\n 4.6.1. New Session Ticket Message . . . . . . . . . . . . . 76\n 4.6.2. Post-Handshake Authentication . . . . . . . . . . . . 78\n 4.6.3. Key and IV Update . . . . . . . . . . . . . . . . . . 79\n 5. Record Protocol . . . . . . . . . . . . . . . . . . . . . . . 80\n 5.1. Record Layer . . . . . . . . . . . . . . . . . . . . . . 81\n 5.2. Record Payload Protection . . . . . . . . . . . . . . . . 83\n 5.3. Per-Record Nonce . . . . . . . . . . . . . . . . . . . . 85\n 5.4. Record Padding . . . . . . . . . . . . . . . . . . . . . 86\n 5.5. Limits on Key Usage . . . . . . . . . . . . . . . . . . . 87\n 6. Alert Protocol . . . . . . . . . . . . . . . . . . . . . . . 87\n 6.1. Closure Alerts . . . . . . . . . . . . . . . . . . . . . 89\n 6.2. Error Alerts . . . . . . . . . . . . . . . . . . . . . . 90\n 7. Cryptographic Computations . . . . . . . . . . . . . . . . . 93\n 7.1. Key Schedule . . . . . . . . . . . . . . . . . . . . . . 93\n 7.2. Updating Traffic Secrets . . . . . . . . . . . . . . . . 96\n 7.3. Traffic Key Calculation . . . . . . . . . . . . . . . . . 97\n 7.4. (EC)DHE Shared Secret Calculation . . . . . . . . . . . . 98\n 7.4.1. Finite Field Diffie-Hellman . . . . . . . . . . . . . 98\n 7.4.2. Elliptic Curve Diffie-Hellman . . . . . . . . . . . . 98\n 7.5. Exporters . . . . . . . . . . . . . . . . . . . . . . . . 99\n 8. 0-RTT and Anti-Replay . . . . . . . . . . . . . . . . . . . . 99\n 8.1. Single-Use Tickets . . . . . . . . . . . . . . . . . . . 101\n 8.2. Client Hello Recording . . . . . . . . . . . . . . . . . 101\n 8.3. Freshness Checks . . . . . . . . . . . . . . . . . . . . 102\n 9. Compliance Requirements . . . . . . . . . . . . . . . . . . . 104\n 9.1. Mandatory-to-Implement Cipher Suites . . . . . . . . . . 104\n 9.2. Mandatory-to-Implement Extensions . . . . . . . . . . . . 104\n 9.3. Protocol Invariants . . . . . . . . . . . . . . . . . . . 105\n 10. Security Considerations . . . . . . . . . . . . . . . . . . . 106\n\n\n\nRescorla Expires September 21, 2018 [Page 3]\n\f\nInternet-Draft TLS March 2018\n\n\n 11. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 107\n 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 108\n 12.1. Normative References . . . . . . . . . . . . . . . . . . 108\n 12.2. Informative References . . . . . . . . . . . . . . . . . 111\n Appendix A. State Machine . . . . . . . . . . . . . . . . . . . 119\n A.1. Client . . . . . . . . . . . . . . . . . . . . . . . . . 119\n A.2. Server . . . . . . . . . . . . . . . . . . . . . . . . . 120\n Appendix B. Protocol Data Structures and Constant Values . . . . 120\n B.1. Record Layer . . . . . . . . . . . . . . . . . . . . . . 121\n B.2. Alert Messages . . . . . . . . . . . . . . . . . . . . . 121\n B.3. Handshake Protocol . . . . . . . . . . . . . . . . . . . 123\n B.3.1. Key Exchange Messages . . . . . . . . . . . . . . . . 123\n B.3.2. Server Parameters Messages . . . . . . . . . . . . . 128\n B.3.3. Authentication Messages . . . . . . . . . . . . . . . 129\n B.3.4. Ticket Establishment . . . . . . . . . . . . . . . . 130\n B.3.5. Updating Keys . . . . . . . . . . . . . . . . . . . . 131\n B.4. Cipher Suites . . . . . . . . . . . . . . . . . . . . . . 131\n Appendix C. Implementation Notes . . . . . . . . . . . . . . . . 132\n C.1. Random Number Generation and Seeding . . . . . . . . . . 132\n C.2. Certificates and Authentication . . . . . . . . . . . . . 133\n C.3. Implementation Pitfalls . . . . . . . . . . . . . . . . . 133\n C.4. Client Tracking Prevention . . . . . . . . . . . . . . . 134\n C.5. Unauthenticated Operation . . . . . . . . . . . . . . . . 135\n Appendix D. Backward Compatibility . . . . . . . . . . . . . . . 135\n D.1. Negotiating with an older server . . . . . . . . . . . . 136\n D.2. Negotiating with an older client . . . . . . . . . . . . 137\n D.3. 0-RTT backwards compatibility . . . . . . . . . . . . . . 137\n D.4. Middlebox Compatibility Mode . . . . . . . . . . . . . . 137\n D.5. Backwards Compatibility Security Restrictions . . . . . . 138\n Appendix E. Overview of Security Properties . . . . . . . . . . 139\n E.1. Handshake . . . . . . . . . . . . . . . . . . . . . . . . 139\n E.1.1. Key Derivation and HKDF . . . . . . . . . . . . . . . 142\n E.1.2. Client Authentication . . . . . . . . . . . . . . . . 143\n E.1.3. 0-RTT . . . . . . . . . . . . . . . . . . . . . . . . 143\n E.1.4. Exporter Independence . . . . . . . . . . . . . . . . 143\n E.1.5. Post-Compromise Security . . . . . . . . . . . . . . 144\n E.1.6. External References . . . . . . . . . . . . . . . . . 144\n E.2. Record Layer . . . . . . . . . . . . . . . . . . . . . . 144\n E.2.1. External References . . . . . . . . . . . . . . . . . 145\n E.3. Traffic Analysis . . . . . . . . . . . . . . . . . . . . 145\n E.4. Side Channel Attacks . . . . . . . . . . . . . . . . . . 146\n E.5. Replay Attacks on 0-RTT . . . . . . . . . . . . . . . . . 147\n E.5.1. Replay and Exporters . . . . . . . . . . . . . . . . 148\n E.6. PSK Identity Exposure . . . . . . . . . . . . . . . . . . 149\n E.7. Attacks on Static RSA . . . . . . . . . . . . . . . . . . 149\n Appendix F. Working Group Information . . . . . . . . . . . . . 149\n Appendix G. Contributors . . . . . . . . . . . . . . . . . . . . 149\n Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 156\n\n\n\nRescorla Expires September 21, 2018 [Page 4]\n\f\nInternet-Draft TLS March 2018\n\n\n1. Introduction\n\n RFC EDITOR: PLEASE REMOVE THE FOLLOWING PARAGRAPH The source for this\n draft is maintained in GitHub. Suggested changes should be submitted\n as pull requests at https://github.com/tlswg/tls13-spec.\n Instructions are on that page as well. Editorial changes can be\n managed in GitHub, but any substantive change should be discussed on\n the TLS mailing list.\n\n The primary goal of TLS is to provide a secure channel between two\n communicating peers; the only requirement from the underlying\n transport is a reliable, in-order, data stream. Specifically, the\n secure channel should provide the following properties:\n\n - Authentication: The server side of the channel is always\n authenticated; the client side is optionally authenticated.\n Authentication can happen via asymmetric cryptography (e.g., RSA\n [RSA], ECDSA [ECDSA], EdDSA [RFC8032]) or a pre-shared key (PSK).\n\n - Confidentiality: Data sent over the channel after establishment is\n only visible to the endpoints. TLS does not hide the length of\n the data it transmits, though endpoints are able to pad TLS\n records in order to obscure lengths and improve protection against\n traffic analysis techniques.\n\n - Integrity: Data sent over the channel after establishment cannot\n be modified by attackers.\n\n These properties should be true even in the face of an attacker who\n has complete control of the network, as described in [RFC3552]. See\n Appendix E for a more complete statement of the relevant security\n properties.\n\n TLS consists of two primary components:\n\n - A handshake protocol (Section 4) that authenticates the\n communicating parties, negotiates cryptographic modes and\n parameters, and establishes shared keying material. The handshake\n protocol is designed to resist tampering; an active attacker\n should not be able to force the peers to negotiate different\n parameters than they would if the connection were not under\n attack.\n\n - A record protocol (Section 5) that uses the parameters established\n by the handshake protocol to protect traffic between the\n communicating peers. The record protocol divides traffic up into\n a series of records, each of which is independently protected\n using the traffic keys.\n\n\n\nRescorla Expires September 21, 2018 [Page 5]\n\f\nInternet-Draft TLS March 2018\n\n\n TLS is application protocol independent; higher-level protocols can\n layer on top of TLS transparently. The TLS standard, however, does\n not specify how protocols add security with TLS; how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n This document defines TLS version 1.3. While TLS 1.3 is not directly\n compatible with previous versions, all versions of TLS incorporate a\n versioning mechanism which allows clients and servers to\n interoperably negotiate a common version if one is supported by both\n peers.\n\n This document supersedes and obsoletes previous versions of TLS\n including version 1.2 [RFC5246]. It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2. Section 4.2.7 updates [RFC4492] by modifying\n the protocol attributes used to negotiate Elliptic Curves. Because\n TLS 1.3 changes the way keys are derived, it updates [RFC5705] as\n described in Section 7.5. It also changes how OCSP messages are\n carried and therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in section Section 4.4.2.1.\n\n1.1. Conventions and Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in BCP\n 14 [RFC2119] [RFC8174] when, and only when, they appear in all\n capitals, as shown here.\n\n The following terms are used:\n\n client: The endpoint initiating the TLS connection.\n\n connection: A transport-layer connection between two endpoints.\n\n endpoint: Either the client or server of the connection.\n\n handshake: An initial negotiation between client and server that\n establishes the parameters of their subsequent interactions within\n TLS.\n\n peer: An endpoint. When discussing a particular endpoint, \"peer\"\n refers to the endpoint that is not the primary subject of discussion.\n\n receiver: An endpoint that is receiving records.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 6]\n\f\nInternet-Draft TLS March 2018\n\n\n sender: An endpoint that is transmitting records.\n\n server: The endpoint which did not initiate the TLS connection.\n\n1.2. Change Log\n\n RFC EDITOR PLEASE DELETE THIS SECTION.\n\n (*) indicates changes to the wire protocol which may require\n implementations to update.\n\n draft-28\n\n Add a section on exposure of PSK identities.\n\n draft-27\n\n - SHOULD->MUST for being able to process \"supported_versions\"\n without 0x0304.\n\n - Much editorial cleanup.\n\n draft-26\n\n - Clarify that you can't negotiate pre-TLS 1.3 with\n supported_versions.\n\n draft-25\n\n - Add the header to additional data (*)\n\n - Minor clarifications.\n\n - IANA cleanup.\n\n draft-24\n\n - Require that CH2 have version 0303 (*)\n\n - Some clarifications\n\n draft-23\n\n - Renumber key_share (*)\n\n - Add a new extension and new code points to allow negotiating PSS\n separately for certificates and CertificateVerify (*)\n\n\n\n\nRescorla Expires September 21, 2018 [Page 7]\n\f\nInternet-Draft TLS March 2018\n\n\n - Slightly restrict when CCS must be accepted to make implementation\n easier.\n\n - Document protocol invariants\n\n - Add some text on the security of static RSA.\n\n draft-22\n\n - Implement changes for improved middlebox penetration (*)\n\n - Move server_certificate_type to encrypted extensions (*)\n\n - Allow resumption with a different SNI (*)\n\n - Padding extension can change on HRR (*)\n\n - Allow an empty ticket_nonce (*)\n\n - Remove requirement to immediately respond to close_notify with\n close_notify (allowing half-close)\n\n draft-21\n\n - Add a per-ticket nonce so that each ticket is associated with a\n different PSK (*).\n\n - Clarify that clients should send alerts with the handshake key if\n possible.\n\n - Update state machine to show rekeying events\n\n - Add discussion of 0-RTT and replay. Recommend that\n implementations implement some anti-replay mechanism.\n\n draft-20\n\n - Add \"post_handshake_auth\" extension to negotiate post-handshake\n authentication (*).\n\n - Shorten labels for HKDF-Expand-Label so that we can fit within one\n compression block (*).\n\n - Define how RFC 7250 works (*).\n\n - Re-enable post-handshake client authentication even when you do\n PSK. The previous prohibition was editorial error.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 8]\n\f\nInternet-Draft TLS March 2018\n\n\n - Remove cert_type and user_mapping, which don't work on TLS 1.3\n anyway.\n\n - Added the no_application_protocol alert from [RFC7301] to the list\n of extensions.\n\n - Added discussion of traffic analysis and side channel attacks.\n\n draft-19\n\n - Hash context_value input to Exporters (*)\n\n - Add an additional Derive-Secret stage to Exporters (*).\n\n - Hash ClientHello1 in the transcript when HRR is used. This\n reduces the state that needs to be carried in cookies. (*)\n\n - Restructure CertificateRequest to have the selectors in\n extensions. This also allowed defining a\n \"certificate_authorities\" extension which can be used by the\n client instead of trusted_ca_keys (*).\n\n - Tighten record framing requirements and require checking of them\n (*).\n\n - Consolidate \"ticket_early_data_info\" and \"early_data\" into a\n single extension (*).\n\n - Change end_of_early_data to be a handshake message (*).\n\n - Add pre-extract Derive-Secret stages to key schedule (*).\n\n - Remove spurious requirement to implement \"pre_shared_key\".\n\n - Clarify location of \"early_data\" from server (it goes in EE, as\n indicated by the table in S 10).\n\n - Require peer public key validation\n\n - Add state machine diagram.\n\n draft-18\n\n - Remove unnecessary resumption_psk which is the only thing expanded\n from the resumption master secret. (*).\n\n - Fix signature_algorithms entry in extensions table.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 9]\n\f\nInternet-Draft TLS March 2018\n\n\n - Restate rule from RFC 6066 that you can't resume unless SNI is the\n same.\n\n draft-17\n\n - Remove 0-RTT Finished and resumption_context, and replace with a\n psk_binder field in the PSK itself (*)\n\n - Restructure PSK key exchange negotiation modes (*)\n\n - Add max_early_data_size field to TicketEarlyDataInfo (*)\n\n - Add a 0-RTT exporter and change the transcript for the regular\n exporter (*)\n\n - Merge TicketExtensions and Extensions registry. Changes\n ticket_early_data_info code point (*)\n\n - Replace Client.key_shares in response to HRR (*)\n\n - Remove redundant labels for traffic key derivation (*)\n\n - Harmonize requirements about cipher suite matching: for resumption\n you need to match KDF but for 0-RTT you need whole cipher suite.\n This allows PSKs to actually negotiate cipher suites. (*)\n\n - Move SCT and OCSP into Certificate.extensions (*)\n\n - Explicitly allow non-offered extensions in NewSessionTicket\n\n - Explicitly allow predicting client Finished for NST\n\n - Clarify conditions for allowing 0-RTT with PSK\n\n draft-16\n\n - Revise version negotiation (*)\n\n - Change RSASSA-PSS and EdDSA SignatureScheme codepoints for better\n backwards compatibility (*)\n\n - Move HelloRetryRequest.selected_group to an extension (*)\n\n - Clarify the behavior of no exporter context and make it the same\n as an empty context.(*)\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 10]\n\f\nInternet-Draft TLS March 2018\n\n\n - New KeyUpdate format that allows for requesting/not-requesting an\n answer. This also means changes to the key schedule to support\n independent updates (*)\n\n - New certificate_required alert (*)\n\n - Forbid CertificateRequest with 0-RTT and PSK.\n\n - Relax requirement to check SNI for 0-RTT.\n\n draft-15\n\n - New negotiation syntax as discussed in Berlin (*)\n\n - Require CertificateRequest.context to be empty during handshake\n (*)\n\n - Forbid empty tickets (*)\n\n - Forbid application data messages in between post-handshake\n messages from the same flight (*)\n\n - Clean up alert guidance (*)\n\n - Clearer guidance on what is needed for TLS 1.2.\n\n - Guidance on 0-RTT time windows.\n\n - Rename a bunch of fields.\n\n - Remove old PRNG text.\n\n - Explicitly require checking that handshake records not span key\n changes.\n\n draft-14\n\n - Allow cookies to be longer (*)\n\n - Remove the \"context\" from EarlyDataIndication as it was undefined\n and nobody used it (*)\n\n - Remove 0-RTT EncryptedExtensions and replace the ticket_age\n extension with an obfuscated version. Also necessitates a change\n to NewSessionTicket (*).\n\n - Move the downgrade sentinel to the end of ServerHello.Random to\n accommodate tlsdate (*).\n\n\n\nRescorla Expires September 21, 2018 [Page 11]\n\f\nInternet-Draft TLS March 2018\n\n\n - Define ecdsa_sha1 (*).\n\n - Allow resumption even after fatal alerts. This matches current\n practice.\n\n - Remove non-closure warning alerts. Require treating unknown\n alerts as fatal.\n\n - Make the rules for accepting 0-RTT less restrictive.\n\n - Clarify 0-RTT backward-compatibility rules.\n\n - Clarify how 0-RTT and PSK identities interact.\n\n - Add a section describing the data limits for each cipher.\n\n - Major editorial restructuring.\n\n - Replace the Security Analysis section with a WIP draft.\n\n draft-13\n\n - Allow server to send SupportedGroups.\n\n - Remove 0-RTT client authentication\n\n - Remove (EC)DHE 0-RTT.\n\n - Flesh out 0-RTT PSK mode and shrink EarlyDataIndication\n\n - Turn PSK-resumption response into an index to save room\n\n - Move CertificateStatus to an extension\n\n - Extra fields in NewSessionTicket.\n\n - Restructure key schedule and add a resumption_context value.\n\n - Require DH public keys and secrets to be zero-padded to the size\n of the group.\n\n - Remove the redundant length fields in KeyShareEntry.\n\n - Define a cookie field for HRR.\n\n draft-12\n\n - Provide a list of the PSK cipher suites.\n\n\n\nRescorla Expires September 21, 2018 [Page 12]\n\f\nInternet-Draft TLS March 2018\n\n\n - Remove the ability for the ServerHello to have no extensions (this\n aligns the syntax with the text).\n\n - Clarify that the server can send application data after its first\n flight (0.5 RTT data)\n\n - Revise signature algorithm negotiation to group hash, signature\n algorithm, and curve together. This is backwards compatible.\n\n - Make ticket lifetime mandatory and limit it to a week.\n\n - Make the purpose strings lower-case. This matches how people are\n implementing for interop.\n\n - Define exporters.\n\n - Editorial cleanup\n\n draft-11\n\n - Port the CFRG curves & signatures work from RFC4492bis.\n\n - Remove sequence number and version from additional_data, which is\n now empty.\n\n - Reorder values in HkdfLabel.\n\n - Add support for version anti-downgrade mechanism.\n\n - Update IANA considerations section and relax some of the policies.\n\n - Unify authentication modes. Add post-handshake client\n authentication.\n\n - Remove early_handshake content type. Terminate 0-RTT data with an\n alert.\n\n - Reset sequence number upon key change (as proposed by Fournet et\n al.)\n\n draft-10\n\n - Remove ClientCertificateTypes field from CertificateRequest and\n add extensions.\n\n - Merge client and server key shares into a single extension.\n\n draft-09\n\n\n\nRescorla Expires September 21, 2018 [Page 13]\n\f\nInternet-Draft TLS March 2018\n\n\n - Change to RSA-PSS signatures for handshake messages.\n\n - Remove support for DSA.\n\n - Update key schedule per suggestions by Hugo, Hoeteck, and Bjoern\n Tackmann.\n\n - Add support for per-record padding.\n\n - Switch to encrypted record ContentType.\n\n - Change HKDF labeling to include protocol version and value\n lengths.\n\n - Shift the final decision to abort a handshake due to incompatible\n certificates to the client rather than having servers abort early.\n\n - Deprecate SHA-1 with signatures.\n\n - Add MTI algorithms.\n\n draft-08\n\n - Remove support for weak and lesser used named curves.\n\n - Remove support for MD5 and SHA-224 hashes with signatures.\n\n - Update lists of available AEAD cipher suites and error alerts.\n\n - Reduce maximum permitted record expansion for AEAD from 2048 to\n 256 octets.\n\n - Require digital signatures even when a previous configuration is\n used.\n\n - Merge EarlyDataIndication and KnownConfiguration.\n\n - Change code point for server_configuration to avoid collision with\n server_hello_done.\n\n - Relax certificate_list ordering requirement to match current\n practice.\n\n draft-07\n\n - Integration of semi-ephemeral DH proposal.\n\n - Add initial 0-RTT support.\n\n\n\nRescorla Expires September 21, 2018 [Page 14]\n\f\nInternet-Draft TLS March 2018\n\n\n - Remove resumption and replace with PSK + tickets.\n\n - Move ClientKeyShare into an extension.\n\n - Move to HKDF.\n\n draft-06\n\n - Prohibit RC4 negotiation for backwards compatibility.\n\n - Freeze & deprecate record layer version field.\n\n - Update format of signatures with context.\n\n - Remove explicit IV.\n\n draft-05\n\n - Prohibit SSL negotiation for backwards compatibility.\n\n - Fix which MS is used for exporters.\n\n draft-04\n\n - Modify key computations to include session hash.\n\n - Remove ChangeCipherSpec.\n\n - Renumber the new handshake messages to be somewhat more consistent\n with existing convention and to remove a duplicate registration.\n\n - Remove renegotiation.\n\n - Remove point format negotiation.\n\n draft-03\n\n - Remove GMT time.\n\n - Merge in support for ECC from RFC 4492 but without explicit\n curves.\n\n - Remove the unnecessary length field from the AD input to AEAD\n ciphers.\n\n - Rename {Client,Server}KeyExchange to {Client,Server}KeyShare.\n\n - Add an explicit HelloRetryRequest to reject the client's.\n\n\n\nRescorla Expires September 21, 2018 [Page 15]\n\f\nInternet-Draft TLS March 2018\n\n\n draft-02\n\n - Increment version number.\n\n - Rework handshake to provide 1-RTT mode.\n\n - Remove custom DHE groups.\n\n - Remove support for compression.\n\n - Remove support for static RSA and DH key exchange.\n\n - Remove support for non-AEAD ciphers.\n\n1.3. Major Differences from TLS 1.2\n\n The following is a list of the major functional differences between\n TLS 1.2 and TLS 1.3. It is not intended to be exhaustive and there\n are many minor differences.\n\n - The list of supported symmetric algorithms has been pruned of all\n algorithms that are considered legacy. Those that remain all use\n Authenticated Encryption with Associated Data (AEAD) algorithms.\n The ciphersuite concept has been changed to separate the\n authentication and key exchange mechanisms from the record\n protection algorithm (including secret key length) and a hash to\n be used with the key derivation function and HMAC.\n\n - A 0-RTT mode was added, saving a round-trip at connection setup\n for some application data, at the cost of certain security\n properties.\n\n - Static RSA and Diffie-Hellman cipher suites have been removed; all\n public-key based key exchange mechanisms now provide forward\n secrecy.\n\n - All handshake messages after the ServerHello are now encrypted.\n The newly introduced EncryptedExtension message allows various\n extensions previously sent in clear in the ServerHello to also\n enjoy confidentiality protection from active attackers.\n\n - The key derivation functions have been re-designed. The new\n design allows easier analysis by cryptographers due to their\n improved key separation properties. The HMAC-based Extract-and-\n Expand Key Derivation Function (HKDF) is used as an underlying\n primitive.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 16]\n\f\nInternet-Draft TLS March 2018\n\n\n - The handshake state machine has been significantly restructured to\n be more consistent and to remove superfluous messages such as\n ChangeCipherSpec (except when needed for middlebox compatibility).\n\n - Elliptic curve algorithms are now in the base spec and new\n signature algorithms, such as ed25519 and ed448, are included.\n TLS 1.3 removed point format negotiation in favor of a single\n point format for each curve.\n\n - Other cryptographic improvements including the removal of\n compression and custom DHE groups, changing the RSA padding to use\n RSASSA-PSS, and the removal of DSA.\n\n - The TLS 1.2 version negotiation mechanism has been deprecated in\n favor of a version list in an extension. This increases\n compatibility with existing servers that incorrectly implemented\n version negotiation.\n\n - Session resumption with and without server-side state as well as\n the PSK-based ciphersuites of earlier TLS versions have been\n replaced by a single new PSK exchange.\n\n - Updated references to point to the updated versions of RFCs, as\n appropriate (e.g., RFC 5280 rather than RFC 3280).\n\n1.4. Updates Affecting TLS 1.2\n\n This document defines several changes that optionally affect\n implementations of TLS 1.2, including those which do not also support\n TLS 1.3:\n\n - A version downgrade protection mechanism is described in\n Section 4.1.3.\n\n - RSASSA-PSS signature schemes are defined in Section 4.2.3.\n\n - The \"supported_versions\" ClientHello extension can be used to\n negotiate the version of TLS to use, in preference to the\n legacy_version field of the ClientHello.\n\n - The \"signature_algorithms_cert\" extension allows a client to\n indicate which signature algorithms it can validate in X.509\n certificates\n\n Additionally, this document clarifies some compliance requirements\n for earlier versions of TLS; see Section 9.3.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 17]\n\f\nInternet-Draft TLS March 2018\n\n\n2. Protocol Overview\n\n The cryptographic parameters used by the secure channel are produced\n by the TLS handshake protocol. This sub-protocol of TLS is used by\n the client and server when first communicating with each other. The\n handshake protocol allows peers to negotiate a protocol version,\n select cryptographic algorithms, optionally authenticate each other,\n and establish shared secret keying material. Once the handshake is\n complete, the peers use the established keys to protect the\n application layer traffic.\n\n A failure of the handshake or other protocol error triggers the\n termination of the connection, optionally preceded by an alert\n message (Section 6).\n\n TLS supports three basic key exchange modes:\n\n - (EC)DHE (Diffie-Hellman over either finite fields or elliptic\n curves)\n\n - PSK-only\n\n - PSK with (EC)DHE\n\n Figure 1 below shows the basic full TLS handshake:\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 18]\n\f\nInternet-Draft TLS March 2018\n\n\n Client Server\n\nKey ^ ClientHello\nExch | + key_share*\n | + signature_algorithms*\n | + psk_key_exchange_modes*\n v + pre_shared_key* -------->\n ServerHello ^ Key\n + key_share* | Exch\n + pre_shared_key* v\n {EncryptedExtensions} ^ Server\n {CertificateRequest*} v Params\n {Certificate*} ^\n {CertificateVerify*} | Auth\n {Finished} v\n <-------- [Application Data*]\n ^ {Certificate*}\nAuth | {CertificateVerify*}\n v {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N\n\n Figure 1: Message flow for full TLS Handshake\n\n The handshake can be thought of as having three phases (indicated in\n the diagram above):\n\n - Key Exchange: Establish shared keying material and select the\n cryptographic parameters. Everything after this phase is\n encrypted.\n\n - Server Parameters: Establish other handshake parameters (whether\n the client is authenticated, application layer protocol support,\n etc.).\n\n - Authentication: Authenticate the server (and optionally the\n client) and provide key confirmation and handshake integrity.\n\n\n\nRescorla Expires September 21, 2018 [Page 19]\n\f\nInternet-Draft TLS March 2018\n\n\n In the Key Exchange phase, the client sends the ClientHello\n (Section 4.1.2) message, which contains a random nonce\n (ClientHello.random); its offered protocol versions; a list of\n symmetric cipher/HKDF hash pairs; either a set of Diffie-Hellman key\n shares (in the \"key_share\" extension Section 4.2.8), a set of pre-\n shared key labels (in the \"pre_shared_key\" extension Section 4.2.11)\n or both; and potentially additional extensions. Additional fields\n and/or messages may also be present for middlebox compatibility.\n\n The server processes the ClientHello and determines the appropriate\n cryptographic parameters for the connection. It then responds with\n its own ServerHello (Section 4.1.3), which indicates the negotiated\n connection parameters. The combination of the ClientHello and the\n ServerHello determines the shared keys. If (EC)DHE key establishment\n is in use, then the ServerHello contains a \"key_share\" extension with\n the server's ephemeral Diffie-Hellman share; the server's share MUST\n be in the same group as one of the client's shares. If PSK key\n establishment is in use, then the ServerHello contains a\n \"pre_shared_key\" extension indicating which of the client's offered\n PSKs was selected. Note that implementations can use (EC)DHE and PSK\n together, in which case both extensions will be supplied.\n\n The server then sends two messages to establish the Server\n Parameters:\n\n EncryptedExtensions: responses to ClientHello extensions that are\n not required to determine the cryptographic parameters, other than\n those that are specific to individual certificates.\n [Section 4.3.1]\n\n CertificateRequest: if certificate-based client authentication is\n desired, the desired parameters for that certificate. This\n message is omitted if client authentication is not desired.\n [Section 4.3.2]\n\n Finally, the client and server exchange Authentication messages. TLS\n uses the same set of messages every time that certificate-based\n authentication is needed. (PSK-based authentication happens as a\n side effect of key exchange.) Specifically:\n\n Certificate: the certificate of the endpoint and any per-certificate\n extensions. This message is omitted by the server if not\n authenticating with a certificate and by the client if the server\n did not send CertificateRequest (thus indicating that the client\n should not authenticate with a certificate). Note that if raw\n public keys [RFC7250] or the cached information extension\n [RFC7924] are in use, then this message will not contain a\n\n\n\n\nRescorla Expires September 21, 2018 [Page 20]\n\f\nInternet-Draft TLS March 2018\n\n\n certificate but rather some other value corresponding to the\n server's long-term key. [Section 4.4.2]\n\n CertificateVerify: a signature over the entire handshake using the\n private key corresponding to the public key in the Certificate\n message. This message is omitted if the endpoint is not\n authenticating via a certificate. [Section 4.4.3]\n\n Finished: a MAC (Message Authentication Code) over the entire\n handshake. This message provides key confirmation, binds the\n endpoint's identity to the exchanged keys, and in PSK mode also\n authenticates the handshake. [Section 4.4.4]\n\n Upon receiving the server's messages, the client responds with its\n Authentication messages, namely Certificate and CertificateVerify (if\n requested), and Finished.\n\n At this point, the handshake is complete, and the client and server\n derive the keying material required by the record layer to exchange\n application-layer data protected through authenticated encryption.\n Application data MUST NOT be sent prior to sending the Finished\n message, except as specified in [Section 2.3]. Note that while the\n server may send application data prior to receiving the client's\n Authentication messages, any data sent at that point is, of course,\n being sent to an unauthenticated peer.\n\n2.1. Incorrect DHE Share\n\n If the client has not provided a sufficient \"key_share\" extension\n (e.g., it includes only DHE or ECDHE groups unacceptable to or\n unsupported by the server), the server corrects the mismatch with a\n HelloRetryRequest and the client needs to restart the handshake with\n an appropriate \"key_share\" extension, as shown in Figure 2. If no\n common cryptographic parameters can be negotiated, the server MUST\n abort the handshake with an appropriate alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 21]\n\f\nInternet-Draft TLS March 2018\n\n\n Client Server\n\n ClientHello\n + key_share -------->\n HelloRetryRequest\n <-------- + key_share\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 2: Message flow for a full handshake with mismatched\n parameters\n\n Note: The handshake transcript incorporates the initial ClientHello/\n HelloRetryRequest exchange; it is not reset with the new ClientHello.\n\n TLS also allows several optimized variants of the basic handshake, as\n described in the following sections.\n\n2.2. Resumption and Pre-Shared Key (PSK)\n\n Although TLS PSKs can be established out of band, PSKs can also be\n established in a previous connection and then used to establish a new\n connection (\"session resumption\" or \"resuming\" with a PSK). Once a\n handshake has completed, the server can send to the client a PSK\n identity that corresponds to a unique key derived from the initial\n handshake (see Section 4.6.1). The client can then use that PSK\n identity in future handshakes to negotiate the use of the associated\n PSK. If the server accepts the PSK, then the security context of the\n new connection is cryptographically tied to the original connection\n and the key derived from the initial handshake is used to bootstrap\n the cryptographic state instead of a full handshake. In TLS 1.2 and\n below, this functionality was provided by \"session IDs\" and \"session\n tickets\" [RFC5077]. Both mechanisms are obsoleted in TLS 1.3.\n\n PSKs can be used with (EC)DHE key exchange in order to provide\n forward secrecy in combination with shared keys, or can be used\n\n\n\nRescorla Expires September 21, 2018 [Page 22]\n\f\nInternet-Draft TLS March 2018\n\n\n alone, at the cost of losing forward secrecy for the application\n data.\n\n Figure 3 shows a pair of handshakes in which the first establishes a\n PSK and the second uses it:\n\n Client Server\n\n Initial Handshake:\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n <-------- [NewSessionTicket]\n [Application Data] <-------> [Application Data]\n\n\n Subsequent Handshake:\n ClientHello\n + key_share*\n + pre_shared_key -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n {Finished}\n <-------- [Application Data*]\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 3: Message flow for resumption and PSK\n\n As the server is authenticating via a PSK, it does not send a\n Certificate or a CertificateVerify message. When a client offers\n resumption via PSK, it SHOULD also supply a \"key_share\" extension to\n the server to allow the server to decline resumption and fall back to\n a full handshake, if needed. The server responds with a\n \"pre_shared_key\" extension to negotiate use of PSK key establishment\n\n\n\n\nRescorla Expires September 21, 2018 [Page 23]\n\f\nInternet-Draft TLS March 2018\n\n\n and can (as shown here) respond with a \"key_share\" extension to do\n (EC)DHE key establishment, thus providing forward secrecy.\n\n When PSKs are provisioned out of band, the PSK identity and the KDF\n hash algorithm to be used with the PSK MUST also be provisioned.\n\n Note: When using an out-of-band provisioned pre-shared secret, a\n critical consideration is using sufficient entropy during the key\n generation, as discussed in [RFC4086]. Deriving a shared secret\n from a password or other low-entropy sources is not secure. A\n low-entropy secret, or password, is subject to dictionary attacks\n based on the PSK binder. The specified PSK authentication is not\n a strong password-based authenticated key exchange even when used\n with Diffie-Hellman key establishment. Specifically, it does not\n prevent an attacker that can observe the handshake from performing\n a brute-force attack on the password/pre-shared key.\n\n2.3. 0-RTT Data\n\n When clients and servers share a PSK (either obtained externally or\n via a previous handshake), TLS 1.3 allows clients to send data on the\n first flight (\"early data\"). The client uses the PSK to authenticate\n the server and to encrypt the early data.\n\n As shown in Figure 4, the 0-RTT data is just added to the 1-RTT\n handshake in the first flight. The rest of the handshake uses the\n same messages as for a 1-RTT handshake with PSK resumption.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 24]\n\f\nInternet-Draft TLS March 2018\n\n\n Client Server\n\n ClientHello\n + early_data\n + key_share*\n + psk_key_exchange_modes\n + pre_shared_key\n (Application Data*) -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n + early_data*\n {Finished}\n <-------- [Application Data*]\n (EndOfEarlyData)\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n () Indicates messages protected using keys\n derived from client_early_traffic_secret.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N\n\n Figure 4: Message flow for a zero round trip handshake\n\n IMPORTANT NOTE: The security properties for 0-RTT data are weaker\n than those for other kinds of TLS data. Specifically:\n\n 1. This data is not forward secret, as it is encrypted solely under\n keys derived using the offered PSK.\n\n 2. There are no guarantees of non-replay between connections.\n Protection against replay for ordinary TLS 1.3 1-RTT data is\n provided via the server's Random value, but 0-RTT data does not\n depend on the ServerHello and therefore has weaker guarantees.\n This is especially relevant if the data is authenticated either\n with TLS client authentication or inside the application\n\n\n\nRescorla Expires September 21, 2018 [Page 25]\n\f\nInternet-Draft TLS March 2018\n\n\n protocol. The same warnings apply to any use of the\n early_exporter_master_secret.\n\n 0-RTT data cannot be duplicated within a connection (i.e., the server\n will not process the same data twice for the same connection) and an\n attacker will not be able to make 0-RTT data appear to be 1-RTT data\n (because it is protected with different keys.) Appendix E.5 contains\n a description of potential attacks and Section 8 describes mechanisms\n which the server can use to limit the impact of replay.\n\n3. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used.\n\n3.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n3.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" double\n brackets.\n\n Single-byte entities containing uninterpreted data are of type\n opaque.\n\n A type alias T' for an existing type T is defined by:\n\n T T';\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 26]\n\f\nInternet-Draft TLS March 2018\n\n\n3.3. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are formed from fixed-length series of bytes\n concatenated as described in Section 3.1 and are also unsigned. The\n following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are transmitted\n in network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n3.4. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* 3 consecutive 3-byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n\n\n\nRescorla Expires September 21, 2018 [Page 27]\n\f\nInternet-Draft TLS March 2018\n\n\n In the following example, mandatory is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 3.3). Similarly,\n longer can represent up to 800 bytes of data, or 400 uint16 elements,\n and it may be empty. Its encoding will include a two-byte actual\n length field prepended to the vector. The length of an encoded\n vector must be an exact multiple of the length of a single element\n (e.g., a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is 2 bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n3.5. Enumerateds\n\n An additional sparse data type is available called enum or\n enumerated. Each definition is a different type. Only enumerateds\n of the same type may be assigned or compared. Every element of an\n enumerated must be assigned a value, as demonstrated in the following\n example. Since the elements of the enumerated are not ordered, they\n can be assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n Future extensions or additions to the protocol may define new values.\n Implementations need to be able to parse and ignore unknown values\n unless the definition of the field states otherwise.\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4 in the current\n version of the protocol.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n\n\n\nRescorla Expires September 21, 2018 [Page 28]\n\f\nInternet-Draft TLS March 2018\n\n\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n The names assigned to enumerateds do not need to be unique. The\n numerical value can describe a range over which the same name\n applies. The value includes the minimum and maximum inclusive values\n in that range, separated by two period characters. This is\n principally useful for reserving regions of the space.\n\n enum { sad(0), meh(1..254), happy(255) } Mood;\n\n3.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax for definition is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } T;\n\n Fixed- and variable-length vector fields are allowed using the\n standard vector syntax. Structures V1 and V2 in the variants example\n below demonstrate this.\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n\n3.7. Constants\n\n Fields and variables may be assigned a fixed value using \"=\", as in:\n\n struct {\n T1 f1 = 8; /* T.f1 must always be 8 */\n T2 f2;\n } T;\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 29]\n\f\nInternet-Draft TLS March 2018\n\n\n3.8. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. Each\n arm of the select specifies the type of that variant's field and an\n optional field label. The mechanism by which the variant is selected\n at runtime is not prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1 [[fe1]];\n case e2: Te2 [[fe2]];\n ....\n case en: Ten [[fen]];\n };\n } Tv;\n\n For example:\n\n enum { apple(0), orange(1) } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n VariantTag type;\n select (VariantRecord.type) {\n case apple: V1;\n case orange: V2;\n };\n } VariantRecord;\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 30]\n\f\nInternet-Draft TLS March 2018\n\n\n4. Handshake Protocol\n\n The handshake protocol is used to negotiate the security parameters\n of a connection. Handshake messages are supplied to the TLS record\n layer, where they are encapsulated within one or more TLSPlaintext or\n TLSCiphertext structures, which are processed and transmitted as\n specified by the current active connection state.\n\n enum {\n client_hello(1),\n server_hello(2),\n new_session_ticket(4),\n end_of_early_data(5),\n encrypted_extensions(8),\n certificate(11),\n certificate_request(13),\n certificate_verify(15),\n finished(20),\n key_update(24),\n message_hash(254),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (Handshake.msg_type) {\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case end_of_early_data: EndOfEarlyData;\n case encrypted_extensions: EncryptedExtensions;\n case certificate_request: CertificateRequest;\n case certificate: Certificate;\n case certificate_verify: CertificateVerify;\n case finished: Finished;\n case new_session_ticket: NewSessionTicket;\n case key_update: KeyUpdate;\n };\n } Handshake;\n\n Protocol messages MUST be sent in the order defined in Section 4.4.1\n and shown in the diagrams in Section 2. A peer which receives a\n handshake message in an unexpected order MUST abort the handshake\n with an \"unexpected_message\" alert.\n\n New handshake message types are assigned by IANA as described in\n Section 11.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 31]\n\f\nInternet-Draft TLS March 2018\n\n\n4.1. Key Exchange Messages\n\n The key exchange messages are used to determine the security\n capabilities of the client and the server and to establish shared\n secrets including the traffic keys used to protect the rest of the\n handshake and the data.\n\n4.1.1. Cryptographic Negotiation\n\n In TLS, the cryptographic negotiation proceeds by the client offering\n the following four sets of options in its ClientHello:\n\n - A list of cipher suites which indicates the AEAD algorithm/HKDF\n hash pairs which the client supports.\n\n - A \"supported_groups\" (Section 4.2.7) extension which indicates the\n (EC)DHE groups which the client supports and a \"key_share\"\n (Section 4.2.8) extension which contains (EC)DHE shares for some\n or all of these groups.\n\n - A \"signature_algorithms\" (Section 4.2.3) extension which indicates\n the signature algorithms which the client can accept.\n\n - A \"pre_shared_key\" (Section 4.2.11) extension which contains a\n list of symmetric key identities known to the client and a\n \"psk_key_exchange_modes\" (Section 4.2.9) extension which indicates\n the key exchange modes that may be used with PSKs.\n\n If the server does not select a PSK, then the first three of these\n options are entirely orthogonal: the server independently selects a\n cipher suite, an (EC)DHE group and key share for key establishment,\n and a signature algorithm/certificate pair to authenticate itself to\n the client. If there is no overlap between the received\n \"supported_groups\" and the groups supported by the server then the\n server MUST abort the handshake with a \"handshake_failure\" or an\n \"insufficient_security\" alert.\n\n If the server selects a PSK, then it MUST also select a key\n establishment mode from the set indicated by client's\n \"psk_key_exchange_modes\" extension (at present, PSK alone or with\n (EC)DHE). Note that if the PSK can be used without (EC)DHE then non-\n overlap in the \"supported_groups\" parameters need not be fatal, as it\n is in the non-PSK case discussed in the previous paragraph.\n\n If the server selects an (EC)DHE group and the client did not offer a\n compatible \"key_share\" extension in the initial ClientHello, the\n server MUST respond with a HelloRetryRequest (Section 4.1.4) message.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 32]\n\f\nInternet-Draft TLS March 2018\n\n\n If the server successfully selects parameters and does not require a\n HelloRetryRequest, it indicates the selected parameters in the\n ServerHello as follows:\n\n - If PSK is being used, then the server will send a \"pre_shared_key\"\n extension indicating the selected key.\n\n - If PSK is not being used, then (EC)DHE and certificate-based\n authentication are always used.\n\n - When (EC)DHE is in use, the server will also provide a \"key_share\"\n extension.\n\n - When authenticating via a certificate, the server will send the\n Certificate (Section 4.4.2) and CertificateVerify (Section 4.4.3)\n messages. In TLS 1.3 as defined by this document, either a PSK or\n a certificate is always used, but not both. Future documents may\n define how to use them together.\n\n If the server is unable to negotiate a supported set of parameters\n (i.e., there is no overlap between the client and server parameters),\n it MUST abort the handshake with either a \"handshake_failure\" or\n \"insufficient_security\" fatal alert (see Section 6).\n\n4.1.2. Client Hello\n\n When a client first connects to a server, it is REQUIRED to send the\n ClientHello as its first TLS message. The client will also send a\n ClientHello when the server has responded to its ClientHello with a\n HelloRetryRequest. In that case, the client MUST send the same\n ClientHello without modification, except:\n\n - If a \"key_share\" extension was supplied in the HelloRetryRequest,\n replacing the list of shares with a list containing a single\n KeyShareEntry from the indicated group.\n\n - Removing the \"early_data\" extension (Section 4.2.10) if one was\n present. Early data is not permitted after HelloRetryRequest.\n\n - Including a \"cookie\" extension if one was provided in the\n HelloRetryRequest.\n\n - Updating the \"pre_shared_key\" extension if present by recomputing\n the \"obfuscated_ticket_age\" and binder values and (optionally)\n removing any PSKs which are incompatible with the server's\n indicated cipher suite.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 33]\n\f\nInternet-Draft TLS March 2018\n\n\n - Optionally adding, removing, or changing the length of the\n \"padding\" extension [RFC7685].\n\n - Other modifications that may be allowed by an extension defined in\n the future and present in the HelloRetryRequest.\n\n Because TLS 1.3 forbids renegotiation, if a server has negotiated TLS\n 1.3 and receives a ClientHello at any other time, it MUST terminate\n the connection with an \"unexpected_message\" alert.\n\n If a server established a TLS connection with a previous version of\n TLS and receives a TLS 1.3 ClientHello in a renegotiation, it MUST\n retain the previous protocol version. In particular, it MUST NOT\n negotiate TLS 1.3.\n\n Structure of this message:\n\n uint16 ProtocolVersion;\n opaque Random[32];\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id<0..32>;\n CipherSuite cipher_suites<2..2^16-2>;\n opaque legacy_compression_methods<1..2^8-1>;\n Extension extensions<8..2^16-1>;\n } ClientHello;\n\n legacy_version In previous versions of TLS, this field was used for\n version negotiation and represented the highest version number\n supported by the client. Experience has shown that many servers\n do not properly implement version negotiation, leading to \"version\n intolerance\" in which the server rejects an otherwise acceptable\n ClientHello with a version number higher than it supports. In TLS\n 1.3, the client indicates its version preferences in the\n \"supported_versions\" extension (Section 4.2.1) and the\n legacy_version field MUST be set to 0x0303, which is the version\n number for TLS 1.2. (See Appendix D for details about backward\n compatibility.)\n\n random 32 bytes generated by a secure random number generator. See\n Appendix C for additional information.\n\n legacy_session_id Versions of TLS before TLS 1.3 supported a\n \"session resumption\" feature which has been merged with Pre-Shared\n\n\n\nRescorla Expires September 21, 2018 [Page 34]\n\f\nInternet-Draft TLS March 2018\n\n\n Keys in this version (see Section 2.2). A client which has a\n cached session ID set by a pre-TLS 1.3 server SHOULD set this\n field to that value. In compatibility mode (see Appendix D.4)\n this field MUST be non-empty, so a client not offering a pre-TLS\n 1.3 session MUST generate a new 32-byte value. This value need\n not be random but SHOULD be unpredictable to avoid implementations\n fixating on a specific value (also known as ossification).\n Otherwise, it MUST be set as a zero length vector (i.e., a single\n zero byte length field).\n\n cipher_suites This is a list of the symmetric cipher options\n supported by the client, specifically the record protection\n algorithm (including secret key length) and a hash to be used with\n HKDF, in descending order of client preference. If the list\n contains cipher suites that the server does not recognize, support\n or wish to use, the server MUST ignore those cipher suites and\n process the remaining ones as usual. Values are defined in\n Appendix B.4. If the client is attempting a PSK key\n establishment, it SHOULD advertise at least one cipher suite\n indicating a Hash associated with the PSK.\n\n legacy_compression_methods Versions of TLS before 1.3 supported\n compression with the list of supported compression methods being\n sent in this field. For every TLS 1.3 ClientHello, this vector\n MUST contain exactly one byte, set to zero, which corresponds to\n the \"null\" compression method in prior versions of TLS. If a TLS\n 1.3 ClientHello is received with any other value in this field,\n the server MUST abort the handshake with an \"illegal_parameter\"\n alert. Note that TLS 1.3 servers might receive TLS 1.2 or prior\n ClientHellos which contain other compression methods and (if\n negotiating such a prior version) MUST follow the procedures for\n the appropriate prior version of TLS. TLS 1.3 ClientHellos are\n identified as having a legacy_version of 0x0303 and a\n supported_versions extension present with 0x0304 as the highest\n version indicated therein.\n\n extensions Clients request extended functionality from servers by\n sending data in the extensions field. The actual \"Extension\"\n format is defined in Section 4.2. In TLS 1.3, use of certain\n extensions is mandatory, as functionality is moved into extensions\n to preserve ClientHello compatibility with previous versions of\n TLS. Servers MUST ignore unrecognized extensions.\n\n All versions of TLS allow an extensions field to optionally follow\n the compression_methods field. TLS 1.3 ClientHello messages always\n contain extensions (minimally \"supported_versions\", otherwise they\n will be interpreted as TLS 1.2 ClientHello messages). However, TLS\n 1.3 servers might receive ClientHello messages without an extensions\n\n\n\nRescorla Expires September 21, 2018 [Page 35]\n\f\nInternet-Draft TLS March 2018\n\n\n field from prior versions of TLS. The presence of extensions can be\n detected by determining whether there are bytes following the\n compression_methods field at the end of the ClientHello. Note that\n this method of detecting optional data differs from the normal TLS\n method of having a variable-length field, but it is used for\n compatibility with TLS before extensions were defined. TLS 1.3\n servers will need to perform this check first and only attempt to\n negotiate TLS 1.3 if the \"supported_versions\" extension is present.\n If negotiating a version of TLS prior to 1.3, a server MUST check\n that the message either contains no data after\n legacy_compression_methods or that it contains a valid extensions\n block with no data following. If not, then it MUST abort the\n handshake with a \"decode_error\" alert.\n\n In the event that a client requests additional functionality using\n extensions, and this functionality is not supplied by the server, the\n client MAY abort the handshake.\n\n After sending the ClientHello message, the client waits for a\n ServerHello or HelloRetryRequest message. If early data is in use,\n the client may transmit early application data (Section 2.3) while\n waiting for the next handshake message.\n\n4.1.3. Server Hello\n\n The server will send this message in response to a ClientHello\n message to proceed with the handshake if it is able to negotiate an\n acceptable set of handshake parameters based on the ClientHello.\n\n Structure of this message:\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id_echo<0..32>;\n CipherSuite cipher_suite;\n uint8 legacy_compression_method = 0;\n Extension extensions<6..2^16-1>;\n } ServerHello;\n\n legacy_version In previous versions of TLS, this field was used for\n version negotiation and represented the selected version number\n for the connection. Unfortunately, some middleboxes fail when\n presented with new values. In TLS 1.3, the TLS server indicates\n its version using the \"supported_versions\" extension\n (Section 4.2.1), and the legacy_version field MUST be set to\n 0x0303, which is the version number for TLS 1.2. (See Appendix D\n for details about backward compatibility.)\n\n\n\nRescorla Expires September 21, 2018 [Page 36]\n\f\nInternet-Draft TLS March 2018\n\n\n random 32 bytes generated by a secure random number generator. See\n Appendix C for additional information. The last eight bytes MUST\n be overwritten as described below if negotiating TLS 1.2 or TLS\n 1.1, but the remaining bytes MUST be random. This structure is\n generated by the server and MUST be generated independently of the\n ClientHello.random.\n\n legacy_session_id_echo The contents of the client's\n legacy_session_id field. Note that this field is echoed even if\n the client's value corresponded to a cached pre-TLS 1.3 session\n which the server has chosen not to resume. A client which\n receives a legacy_session_id_echo field that does not match what\n it sent in the ClientHello MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n cipher_suite The single cipher suite selected by the server from the\n list in ClientHello.cipher_suites. A client which receives a\n cipher suite that was not offered MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n legacy_compression_method A single byte which MUST have the value 0.\n\n extensions A list of extensions. The ServerHello MUST only include\n extensions which are required to establish the cryptographic\n context and negotiate the protocol version. All TLS 1.3\n ServerHello messages MUST contain the \"supported_versions\"\n extension. Current ServerHello messages additionally contain\n either the \"pre_shared_key\" or \"key_share\" extensions, or both\n when using a PSK with (EC)DHE key establishment. Other extensions\n are sent separately in the EncryptedExtensions message.\n\n For reasons of backward compatibility with middleboxes (see\n Appendix D.4) the HelloRetryRequest message uses the same structure\n as the ServerHello, but with Random set to the special value of the\n SHA-256 of \"HelloRetryRequest\":\n\n CF 21 AD 74 E5 9A 61 11 BE 1D 8C 02 1E 65 B8 91\n C2 A2 11 16 7A BB 8C 5E 07 9E 09 E2 C8 A8 33 9C\n\n Upon receiving a message with type server_hello, implementations MUST\n first examine the Random value and if it matches this value, process\n it as described in Section 4.1.4).\n\n TLS 1.3 has a downgrade protection mechanism embedded in the server's\n random value. TLS 1.3 servers which negotiate TLS 1.2 or below in\n response to a ClientHello MUST set the last eight bytes of their\n Random value specially.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 37]\n\f\nInternet-Draft TLS March 2018\n\n\n If negotiating TLS 1.2, TLS 1.3 servers MUST set the last eight bytes\n of their Random value to the bytes:\n\n 44 4F 57 4E 47 52 44 01\n\n If negotiating TLS 1.1 or below, TLS 1.3 servers MUST and TLS 1.2\n servers SHOULD set the last eight bytes of their Random value to the\n bytes:\n\n 44 4F 57 4E 47 52 44 00\n\n TLS 1.3 clients receiving a ServerHello indicating TLS 1.2 or below\n MUST check that the last eight bytes are not equal to either of these\n values. TLS 1.2 clients SHOULD also check that the last eight bytes\n are not equal to the second value if the ServerHello indicates TLS\n 1.1 or below. If a match is found, the client MUST abort the\n handshake with an \"illegal_parameter\" alert. This mechanism provides\n limited protection against downgrade attacks over and above what is\n provided by the Finished exchange: because the ServerKeyExchange, a\n message present in TLS 1.2 and below, includes a signature over both\n random values, it is not possible for an active attacker to modify\n the random values without detection as long as ephemeral ciphers are\n used. It does not provide downgrade protection when static RSA is\n used.\n\n Note: This is a change from [RFC5246], so in practice many TLS 1.2\n clients and servers will not behave as specified above.\n\n A legacy TLS client performing renegotiation with TLS 1.2 or prior\n and which receives a TLS 1.3 ServerHello during renegotiation MUST\n abort the handshake with a \"protocol_version\" alert. Note that\n renegotiation is not possible when TLS 1.3 has been negotiated.\n\n RFC EDITOR: PLEASE REMOVE THE FOLLOWING PARAGRAPH Implementations of\n draft versions (see Section 4.2.1.1) of this specification SHOULD NOT\n implement this mechanism on either client and server. A pre-RFC\n client connecting to RFC servers, or vice versa, will appear to\n downgrade to TLS 1.2. With the mechanism enabled, this will cause an\n interoperability failure.\n\n4.1.4. Hello Retry Request\n\n The server will send this message in response to a ClientHello\n message if it is able to find an acceptable set of parameters but the\n ClientHello does not contain sufficient information to proceed with\n the handshake. As discussed in Section 4.1.3, the HelloRetryRequest\n has the same format as a ServerHello message, and the legacy_version,\n legacy_session_id_echo, cipher_suite, and legacy_compression methods\n\n\n\nRescorla Expires September 21, 2018 [Page 38]\n\f\nInternet-Draft TLS March 2018\n\n\n fields have the same meaning. However, for convenience we discuss\n HelloRetryRequest throughout this document as if it were a distinct\n message.\n\n The server's extensions MUST contain \"supported_versions\" and\n otherwise the server SHOULD send only the extensions necessary for\n the client to generate a correct ClientHello pair. As with\n ServerHello, a HelloRetryRequest MUST NOT contain any extensions that\n were not first offered by the client in its ClientHello, with the\n exception of optionally the \"cookie\" (see Section 4.2.2) extension.\n\n Upon receipt of a HelloRetryRequest, the client MUST check the\n legacy_version, legacy_session_id_echo, cipher_suite, and\n legacy_compression_method as specified in Section 4.1.3 and then\n process the extensions, starting with determining the version using\n \"supported_versions\". Clients MUST abort the handshake with an\n \"illegal_parameter\" alert if the HelloRetryRequest would not result\n in any change in the ClientHello. If a client receives a second\n HelloRetryRequest in the same connection (i.e., where the ClientHello\n was itself in response to a HelloRetryRequest), it MUST abort the\n handshake with an \"unexpected_message\" alert.\n\n Otherwise, the client MUST process all extensions in the\n HelloRetryRequest and send a second updated ClientHello. The\n HelloRetryRequest extensions defined in this specification are:\n\n - supported_versions (see Section 4.2.1)\n\n - cookie (see Section 4.2.2)\n\n - key_share (see Section 4.2.8)\n\n In addition, in its updated ClientHello, the client SHOULD NOT offer\n any pre-shared keys associated with a hash other than that of the\n selected cipher suite. This allows the client to avoid having to\n compute partial hash transcripts for multiple hashes in the second\n ClientHello. A client which receives a cipher suite that was not\n offered MUST abort the handshake. Servers MUST ensure that they\n negotiate the same cipher suite when receiving a conformant updated\n ClientHello (if the server selects the cipher suite as the first step\n in the negotiation, then this will happen automatically). Upon\n receiving the ServerHello, clients MUST check that the cipher suite\n supplied in the ServerHello is the same as that in the\n HelloRetryRequest and otherwise abort the handshake with an\n \"illegal_parameter\" alert.\n\n The value of selected_version in the HelloRetryRequest\n \"supported_versions\" extension MUST be retained in the ServerHello,\n\n\n\nRescorla Expires September 21, 2018 [Page 39]\n\f\nInternet-Draft TLS March 2018\n\n\n and a client MUST abort the handshake with an \"illegal_parameter\"\n alert if the value changes.\n\n4.2. Extensions\n\n A number of TLS messages contain tag-length-value encoded extensions\n structures.\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n server_name(0), /* RFC 6066 */\n max_fragment_length(1), /* RFC 6066 */\n status_request(5), /* RFC 6066 */\n supported_groups(10), /* RFC 4492, 7919 */\n signature_algorithms(13), /* [[this document]] */\n use_srtp(14), /* RFC 5764 */\n heartbeat(15), /* RFC 6520 */\n application_layer_protocol_negotiation(16), /* RFC 7301 */\n signed_certificate_timestamp(18), /* RFC 6962 */\n client_certificate_type(19), /* RFC 7250 */\n server_certificate_type(20), /* RFC 7250 */\n padding(21), /* RFC 7685 */\n pre_shared_key(41), /* [[this document]] */\n early_data(42), /* [[this document]] */\n supported_versions(43), /* [[this document]] */\n cookie(44), /* [[this document]] */\n psk_key_exchange_modes(45), /* [[this document]] */\n certificate_authorities(47), /* [[this document]] */\n oid_filters(48), /* [[this document]] */\n post_handshake_auth(49), /* [[this document]] */\n signature_algorithms_cert(50), /* [[this document]] */\n key_share(51), /* [[this document]] */\n (65535)\n } ExtensionType;\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The list of extension types is maintained by IANA as described in\n Section 11.\n\n\n\nRescorla Expires September 21, 2018 [Page 40]\n\f\nInternet-Draft TLS March 2018\n\n\n Extensions are generally structured in a request/response fashion,\n though some extensions are just indications with no corresponding\n response. The client sends its extension requests in the ClientHello\n message and the server sends its extension responses in the\n ServerHello, EncryptedExtensions, HelloRetryRequest and Certificate\n messages. The server sends extension requests in the\n CertificateRequest message which a client MAY respond to with a\n Certificate message. The server MAY also send unsolicited extensions\n in the NewSessionTicket, though the client does not respond directly\n to these.\n\n Implementations MUST NOT send extension responses if the remote\n endpoint did not send the corresponding extension requests, with the\n exception of the \"cookie\" extension in HelloRetryRequest. Upon\n receiving such an extension, an endpoint MUST abort the handshake\n with an \"unsupported_extension\" alert.\n\n The table below indicates the messages where a given extension may\n appear, using the following notation: CH (ClientHello), SH\n (ServerHello), EE (EncryptedExtensions), CT (Certificate), CR\n (CertificateRequest), NST (NewSessionTicket) and HRR\n (HelloRetryRequest). If an implementation receives an extension\n which it recognizes and which is not specified for the message in\n which it appears it MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 41]\n\f\nInternet-Draft TLS March 2018\n\n\n +--------------------------------------------------+-------------+\n | Extension | TLS 1.3 |\n +--------------------------------------------------+-------------+\n | server_name [RFC6066] | CH, EE |\n | | |\n | max_fragment_length [RFC6066] | CH, EE |\n | | |\n | status_request [RFC6066] | CH, CR, CT |\n | | |\n | supported_groups [RFC7919] | CH, EE |\n | | |\n | signature_algorithms [RFC5246] | CH, CR |\n | | |\n | use_srtp [RFC5764] | CH, EE |\n | | |\n | heartbeat [RFC6520] | CH, EE |\n | | |\n | application_layer_protocol_negotiation [RFC7301] | CH, EE |\n | | |\n | signed_certificate_timestamp [RFC6962] | CH, CR, CT |\n | | |\n | client_certificate_type [RFC7250] | CH, EE |\n | | |\n | server_certificate_type [RFC7250] | CH, EE |\n | | |\n | padding [RFC7685] | CH |\n | | |\n | key_share [[this document]] | CH, SH, HRR |\n | | |\n | pre_shared_key [[this document]] | CH, SH |\n | | |\n | psk_key_exchange_modes [[this document]] | CH |\n | | |\n | early_data [[this document]] | CH, EE, NST |\n | | |\n | cookie [[this document]] | CH, HRR |\n | | |\n | supported_versions [[this document]] | CH, SH, HRR |\n | | |\n | certificate_authorities [[this document]] | CH, CR |\n | | |\n | oid_filters [[this document]] | CR |\n | | |\n | post_handshake_auth [[this document]] | CH |\n | | |\n | signature_algorithms_cert [[this document]] | CH, CR |\n +--------------------------------------------------+-------------+\n\n\n\n\nRescorla Expires September 21, 2018 [Page 42]\n\f\nInternet-Draft TLS March 2018\n\n\n When multiple extensions of different types are present, the\n extensions MAY appear in any order, with the exception of\n \"pre_shared_key\" Section 4.2.11 which MUST be the last extension in\n the ClientHello. There MUST NOT be more than one extension of the\n same type in a given extension block.\n\n In TLS 1.3, unlike TLS 1.2, extensions are negotiated for each\n handshake even when in resumption-PSK mode. However, 0-RTT\n parameters are those negotiated in the previous handshake; mismatches\n may require rejecting 0-RTT (see Section 4.2.10).\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions (e.g., the handshake cannot continue), and some are\n simply refusals to support particular features. In general, error\n alerts should be used for the former and a field in the server\n extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem. Often the fact that the extension fields are\n included in the inputs to the Finished message hashes will be\n sufficient, but extreme care is needed when the extension changes\n the meaning of messages sent in the handshake phase. Designers\n and implementors should be aware of the fact that until the\n handshake has been authenticated, active attackers can modify\n messages and insert, remove, or replace extensions.\n\n4.2.1. Supported Versions\n\n struct {\n select (Handshake.msg_type) {\n case client_hello:\n ProtocolVersion versions<2..254>;\n\n case server_hello: /* and HelloRetryRequest */\n ProtocolVersion selected_version;\n };\n } SupportedVersions;\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 43]\n\f\nInternet-Draft TLS March 2018\n\n\n The \"supported_versions\" extension is used by the client to indicate\n which versions of TLS it supports and by the server to indicate which\n version it is using. The extension contains a list of supported\n versions in preference order, with the most preferred version first.\n Implementations of this specification MUST send this extension in the\n ClientHello containing all versions of TLS which they are prepared to\n negotiate (for this specification, that means minimally 0x0304, but\n if previous versions of TLS are allowed to be negotiated, they MUST\n be present as well).\n\n If this extension is not present, servers which are compliant with\n this specification, and which also support TLS 1.2, MUST negotiate\n TLS 1.2 or prior as specified in [RFC5246], even if\n ClientHello.legacy_version is 0x0304 or later. Servers MAY abort the\n handshake upon receiving a ClientHello with legacy_version 0x0304 or\n later.\n\n If this extension is present in the ClientHello, servers MUST NOT use\n the ClientHello.legacy_version value for version negotiation and MUST\n use only the \"supported_versions\" extension to determine client\n preferences. Servers MUST only select a version of TLS present in\n that extension and MUST ignore any unknown versions that are present\n in that extension. Note that this mechanism makes it possible to\n negotiate a version prior to TLS 1.2 if one side supports a sparse\n range. Implementations of TLS 1.3 which choose to support prior\n versions of TLS SHOULD support TLS 1.2. Servers MUST be prepared to\n receive ClientHellos that include this extension but do not include\n 0x0304 in the list of versions.\n\n A server which negotiates a version of TLS prior to TLS 1.3 MUST set\n ServerHello.version and MUST NOT send the \"supported_versions\"\n extension. A server which negotiates TLS 1.3 MUST respond by sending\n a \"supported_versions\" extension containing the selected version\n value (0x0304). It MUST set the ServerHello.legacy_version field to\n 0x0303 (TLS 1.2). Clients MUST check for this extension prior to\n processing the rest of the ServerHello (although they will have to\n parse the ServerHello in order to read the extension). If this\n extension is present, clients MUST ignore the\n ServerHello.legacy_version value and MUST use only the\n \"supported_versions\" extension to determine the selected version. If\n the \"supported_versions\" extension in the ServerHello contains a\n version not offered by the client or contains a version prior to TLS\n 1.3, the client MUST abort the handshake with an \"illegal_parameter\"\n alert.\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 44]\n\f\nInternet-Draft TLS March 2018\n\n\n4.2.1.1. Draft Version Indicator\n\n RFC EDITOR: PLEASE REMOVE THIS SECTION\n\n While the eventual version indicator for the RFC version of TLS 1.3\n will be 0x0304, implementations of draft versions of this\n specification SHOULD instead advertise 0x7f00 | draft_version in the\n ServerHello and HelloRetryRequest \"supported_versions\" extension.\n For instance, draft-17 would be encoded as the 0x7f11. This allows\n pre-RFC implementations to safely negotiate with each other, even if\n they would otherwise be incompatible.\n\n4.2.2. Cookie\n\n struct {\n opaque cookie<1..2^16-1>;\n } Cookie;\n\n Cookies serve two primary purposes:\n\n - Allowing the server to force the client to demonstrate\n reachability at their apparent network address (thus providing a\n measure of DoS protection). This is primarily useful for non-\n connection-oriented transports (see [RFC6347] for an example of\n this).\n\n - Allowing the server to offload state to the client, thus allowing\n it to send a HelloRetryRequest without storing any state. The\n server can do this by storing the hash of the ClientHello in the\n HelloRetryRequest cookie (protected with some suitable integrity\n algorithm).\n\n When sending a HelloRetryRequest, the server MAY provide a \"cookie\"\n extension to the client (this is an exception to the usual rule that\n the only extensions that may be sent are those that appear in the\n ClientHello). When sending the new ClientHello, the client MUST copy\n the contents of the extension received in the HelloRetryRequest into\n a \"cookie\" extension in the new ClientHello. Clients MUST NOT use\n cookies in their initial ClientHello in subsequent connections.\n\n When a server is operating statelessly it may receive an unprotected\n record of type change_cipher_spec between the first and second\n ClientHello (see Section 5). Since the server is not storing any\n state this will appear as if it were the first message to be\n received. Servers operating statelessly MUST ignore these records.\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 45]\n\f\nInternet-Draft TLS March 2018\n\n\n4.2.3. Signature Algorithms\n\n TLS 1.3 provides two extensions for indicating which signature\n algorithms may be used in digital signatures. The\n \"signature_algorithms_cert\" extension applies to signatures in\n certificates and the \"signature_algorithms\" extension, which\n originally appeared in TLS 1.2, applies to signatures in\n CertificateVerify messages. The keys found in certificates MUST also\n be of appropriate type for the signature algorithms they are used\n with. This is a particular issue for RSA keys and PSS signatures, as\n described below. If no \"signature_algorithms_cert\" extension is\n present, then the \"signature_algorithms\" extension also applies to\n signatures appearing in certificates. Clients which desire the\n server to authenticate itself via a certificate MUST send\n \"signature_algorithms\". If a server is authenticating via a\n certificate and the client has not sent a \"signature_algorithms\"\n extension, then the server MUST abort the handshake with a\n \"missing_extension\" alert (see Section 9.2).\n\n The \"signature_algorithms_cert\" extension was added to allow\n implementations which supported different sets of algorithms for\n certificates and in TLS itself to clearly signal their capabilities.\n TLS 1.2 implementations SHOULD also process this extension.\n Implementations which have the same policy in both cases MAY omit the\n \"signature_algorithms_cert\" extension.\n\n The \"extension_data\" field of these extensions contains a\n SignatureSchemeList value:\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 46]\n\f\nInternet-Draft TLS March 2018\n\n\n enum {\n /* RSASSA-PKCS1-v1_5 algorithms */\n rsa_pkcs1_sha256(0x0401),\n rsa_pkcs1_sha384(0x0501),\n rsa_pkcs1_sha512(0x0601),\n\n /* ECDSA algorithms */\n ecdsa_secp256r1_sha256(0x0403),\n ecdsa_secp384r1_sha384(0x0503),\n ecdsa_secp521r1_sha512(0x0603),\n\n /* RSASSA-PSS algorithms with public key OID rsaEncryption */\n rsa_pss_rsae_sha256(0x0804),\n rsa_pss_rsae_sha384(0x0805),\n rsa_pss_rsae_sha512(0x0806),\n\n /* EdDSA algorithms */\n ed25519(0x0807),\n ed448(0x0808),\n\n /* RSASSA-PSS algorithms with public key OID RSASSA-PSS */\n rsa_pss_pss_sha256(0x0809),\n rsa_pss_pss_sha384(0x080a),\n rsa_pss_pss_sha512(0x080b),\n\n /* Legacy algorithms */\n rsa_pkcs1_sha1(0x0201),\n ecdsa_sha1(0x0203),\n\n /* Reserved Code Points */\n private_use(0xFE00..0xFFFF),\n (0xFFFF)\n } SignatureScheme;\n\n struct {\n SignatureScheme supported_signature_algorithms<2..2^16-2>;\n } SignatureSchemeList;\n\n Note: This enum is named \"SignatureScheme\" because there is already a\n \"SignatureAlgorithm\" type in TLS 1.2, which this replaces. We use\n the term \"signature algorithm\" throughout the text.\n\n Each SignatureScheme value lists a single signature algorithm that\n the client is willing to verify. The values are indicated in\n descending order of preference. Note that a signature algorithm\n takes as input an arbitrary-length message, rather than a digest.\n Algorithms which traditionally act on a digest should be defined in\n TLS to first hash the input with a specified hash algorithm and then\n\n\n\nRescorla Expires September 21, 2018 [Page 47]\n\f\nInternet-Draft TLS March 2018\n\n\n proceed as usual. The code point groups listed above have the\n following meanings:\n\n RSASSA-PKCS1-v1_5 algorithms Indicates a signature algorithm using\n RSASSA-PKCS1-v1_5 [RFC8017] with the corresponding hash algorithm\n as defined in [SHS]. These values refer solely to signatures\n which appear in certificates (see Section 4.4.2.2) and are not\n defined for use in signed TLS handshake messages, although they\n MAY appear in \"signature_algorithms\" and\n \"signature_algorithms_cert\" for backward compatibility with TLS\n 1.2,\n\n ECDSA algorithms Indicates a signature algorithm using ECDSA\n [ECDSA], the corresponding curve as defined in ANSI X9.62 [X962]\n and FIPS 186-4 [DSS], and the corresponding hash algorithm as\n defined in [SHS]. The signature is represented as a DER-encoded\n [X690] ECDSA-Sig-Value structure.\n\n RSASSA-PSS RSAE algorithms Indicates a signature algorithm using\n RSASSA-PSS [RFC8017] with mask generation function 1. The digest\n used in the mask generation function and the digest being signed\n are both the corresponding hash algorithm as defined in [SHS].\n The length of the salt MUST be equal to the length of the output\n of the digest algorithm. If the public key is carried in an X.509\n certificate, it MUST use the rsaEncryption OID [RFC5280].\n\n EdDSA algorithms Indicates a signature algorithm using EdDSA as\n defined in [RFC8032] or its successors. Note that these\n correspond to the \"PureEdDSA\" algorithms and not the \"prehash\"\n variants.\n\n RSASSA-PSS PSS algorithms Indicates a signature algorithm using\n RSASSA-PSS [RFC8017] with mask generation function 1. The digest\n used in the mask generation function and the digest being signed\n are both the corresponding hash algorithm as defined in [SHS].\n The length of the salt MUST be equal to the length of the digest\n algorithm. If the public key is carried in an X.509 certificate,\n it MUST use the RSASSA-PSS OID [RFC5756]. When used in\n certificate signatures, the algorithm parameters MUST be DER\n encoded. If the corresponding public key's parameters are\n present, then the parameters in the signature MUST be identical to\n those in the public key.\n\n Legacy algorithms Indicates algorithms which are being deprecated\n because they use algorithms with known weaknesses, specifically\n SHA-1 which is used in this context with either with RSA using\n RSASSA-PKCS1-v1_5 or ECDSA. These values refer solely to\n signatures which appear in certificates (see Section 4.4.2.2) and\n\n\n\nRescorla Expires September 21, 2018 [Page 48]\n\f\nInternet-Draft TLS March 2018\n\n\n are not defined for use in signed TLS handshake messages, although\n they MAY appear in \"signature_algorithms\" and\n \"signature_algorithms_cert\" for backward compatibility with TLS\n 1.2, Endpoints SHOULD NOT negotiate these algorithms but are\n permitted to do so solely for backward compatibility. Clients\n offering these values MUST list them as the lowest priority\n (listed after all other algorithms in SignatureSchemeList). TLS\n 1.3 servers MUST NOT offer a SHA-1 signed certificate unless no\n valid certificate chain can be produced without it (see\n Section 4.4.2.2).\n\n The signatures on certificates that are self-signed or certificates\n that are trust anchors are not validated since they begin a\n certification path (see [RFC5280], Section 3.2). A certificate that\n begins a certification path MAY use a signature algorithm that is not\n advertised as being supported in the \"signature_algorithms\"\n extension.\n\n Note that TLS 1.2 defines this extension differently. TLS 1.3\n implementations willing to negotiate TLS 1.2 MUST behave in\n accordance with the requirements of [RFC5246] when negotiating that\n version. In particular:\n\n - TLS 1.2 ClientHellos MAY omit this extension.\n\n - In TLS 1.2, the extension contained hash/signature pairs. The\n pairs are encoded in two octets, so SignatureScheme values have\n been allocated to align with TLS 1.2's encoding. Some legacy\n pairs are left unallocated. These algorithms are deprecated as of\n TLS 1.3. They MUST NOT be offered or negotiated by any\n implementation. In particular, MD5 [SLOTH], SHA-224, and DSA MUST\n NOT be used.\n\n - ECDSA signature schemes align with TLS 1.2's ECDSA hash/signature\n pairs. However, the old semantics did not constrain the signing\n curve. If TLS 1.2 is negotiated, implementations MUST be prepared\n to accept a signature that uses any curve that they advertised in\n the \"supported_groups\" extension.\n\n - Implementations that advertise support for RSASSA-PSS (which is\n mandatory in TLS 1.3), MUST be prepared to accept a signature\n using that scheme even when TLS 1.2 is negotiated. In TLS 1.2,\n RSASSA-PSS is used with RSA cipher suites.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 49]\n\f\nInternet-Draft TLS March 2018\n\n\n4.2.4. Certificate Authorities\n\n The \"certificate_authorities\" extension is used to indicate the\n certificate authorities which an endpoint supports and which SHOULD\n be used by the receiving endpoint to guide certificate selection.\n\n The body of the \"certificate_authorities\" extension consists of a\n CertificateAuthoritiesExtension structure.\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n DistinguishedName authorities<3..2^16-1>;\n } CertificateAuthoritiesExtension;\n\n authorities A list of the distinguished names [X501] of acceptable\n certificate authorities, represented in DER-encoded [X690] format.\n These distinguished names specify a desired distinguished name for\n trust anchor or subordinate CA; thus, this message can be used to\n describe known trust anchors as well as a desired authorization\n space.\n\n The client MAY send the \"certificate_authorities\" extension in the\n ClientHello message. The server MAY send it in the\n CertificateRequest message.\n\n The \"trusted_ca_keys\" extension, which serves a similar purpose\n [RFC6066], but is more complicated, is not used in TLS 1.3 (although\n it may appear in ClientHello messages from clients which are offering\n prior versions of TLS).\n\n4.2.5. OID Filters\n\n The \"oid_filters\" extension allows servers to provide a set of OID/\n value pairs which it would like the client's certificate to match.\n This extension, if provided by the server, MUST only be sent in the\n CertificateRequest message.\n\n struct {\n opaque certificate_extension_oid<1..2^8-1>;\n opaque certificate_extension_values<0..2^16-1>;\n } OIDFilter;\n\n struct {\n OIDFilter filters<0..2^16-1>;\n } OIDFilterExtension;\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 50]\n\f\nInternet-Draft TLS March 2018\n\n\n filters A list of certificate extension OIDs [RFC5280] with their\n allowed value(s) and represented in DER-encoded [X690] format.\n Some certificate extension OIDs allow multiple values (e.g.,\n Extended Key Usage). If the server has included a non-empty\n filters list, the client certificate included in the response MUST\n contain all of the specified extension OIDs that the client\n recognizes. For each extension OID recognized by the client, all\n of the specified values MUST be present in the client certificate\n (but the certificate MAY have other values as well). However, the\n client MUST ignore and skip any unrecognized certificate extension\n OIDs. If the client ignored some of the required certificate\n extension OIDs and supplied a certificate that does not satisfy\n the request, the server MAY at its discretion either continue the\n connection without client authentication, or abort the handshake\n with an \"unsupported_certificate\" alert. Any given OID MUST NOT\n appear more than once in the filters list.\n\n PKIX RFCs define a variety of certificate extension OIDs and their\n corresponding value types. Depending on the type, matching\n certificate extension values are not necessarily bitwise-equal. It\n is expected that TLS implementations will rely on their PKI libraries\n to perform certificate selection using certificate extension OIDs.\n\n This document defines matching rules for two standard certificate\n extensions defined in [RFC5280]:\n\n - The Key Usage extension in a certificate matches the request when\n all key usage bits asserted in the request are also asserted in\n the Key Usage certificate extension.\n\n - The Extended Key Usage extension in a certificate matches the\n request when all key purpose OIDs present in the request are also\n found in the Extended Key Usage certificate extension. The\n special anyExtendedKeyUsage OID MUST NOT be used in the request.\n\n Separate specifications may define matching rules for other\n certificate extensions.\n\n4.2.6. Post-Handshake Client Authentication\n\n The \"post_handshake_auth\" extension is used to indicate that a client\n is willing to perform post-handshake authentication (Section 4.6.2).\n Servers MUST NOT send a post-handshake CertificateRequest to clients\n which do not offer this extension. Servers MUST NOT send this\n extension.\n\n struct {} PostHandshakeAuth;\n\n\n\n\nRescorla Expires September 21, 2018 [Page 51]\n\f\nInternet-Draft TLS March 2018\n\n\n The \"extension_data\" field of the \"post_handshake_auth\" extension is\n zero length.\n\n4.2.7. Negotiated Groups\n\n When sent by the client, the \"supported_groups\" extension indicates\n the named groups which the client supports for key exchange, ordered\n from most preferred to least preferred.\n\n Note: In versions of TLS prior to TLS 1.3, this extension was named\n \"elliptic_curves\" and only contained elliptic curve groups. See\n [RFC4492] and [RFC7919]. This extension was also used to negotiate\n ECDSA curves. Signature algorithms are now negotiated independently\n (see Section 4.2.3).\n\n The \"extension_data\" field of this extension contains a\n \"NamedGroupList\" value:\n\n enum {\n\n /* Elliptic Curve Groups (ECDHE) */\n secp256r1(0x0017), secp384r1(0x0018), secp521r1(0x0019),\n x25519(0x001D), x448(0x001E),\n\n /* Finite Field Groups (DHE) */\n ffdhe2048(0x0100), ffdhe3072(0x0101), ffdhe4096(0x0102),\n ffdhe6144(0x0103), ffdhe8192(0x0104),\n\n /* Reserved Code Points */\n ffdhe_private_use(0x01FC..0x01FF),\n ecdhe_private_use(0xFE00..0xFEFF),\n (0xFFFF)\n } NamedGroup;\n\n struct {\n NamedGroup named_group_list<2..2^16-1>;\n } NamedGroupList;\n\n Elliptic Curve Groups (ECDHE) Indicates support for the\n corresponding named curve, defined either in FIPS 186-4 [DSS] or\n in [RFC7748]. Values 0xFE00 through 0xFEFF are reserved for\n private use.\n\n Finite Field Groups (DHE) Indicates support of the corresponding\n finite field group, defined in [RFC7919]. Values 0x01FC through\n 0x01FF are reserved for private use.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 52]\n\f\nInternet-Draft TLS March 2018\n\n\n Items in named_group_list are ordered according to the client's\n preferences (most preferred choice first).\n\n As of TLS 1.3, servers are permitted to send the \"supported_groups\"\n extension to the client. Clients MUST NOT act upon any information\n found in \"supported_groups\" prior to successful completion of the\n handshake but MAY use the information learned from a successfully\n completed handshake to change what groups they use in their\n \"key_share\" extension in subsequent connections. If the server has a\n group it prefers to the ones in the \"key_share\" extension but is\n still willing to accept the ClientHello, it SHOULD send\n \"supported_groups\" to update the client's view of its preferences;\n this extension SHOULD contain all groups the server supports,\n regardless of whether they are currently supported by the client.\n\n4.2.8. Key Share\n\n The \"key_share\" extension contains the endpoint's cryptographic\n parameters.\n\n Clients MAY send an empty client_shares vector in order to request\n group selection from the server at the cost of an additional round\n trip. (see Section 4.1.4)\n\n struct {\n NamedGroup group;\n opaque key_exchange<1..2^16-1>;\n } KeyShareEntry;\n\n group The named group for the key being exchanged.\n\n key_exchange Key exchange information. The contents of this field\n are determined by the specified group and its corresponding\n definition. Finite Field Diffie-Hellman [DH] parameters are\n described in Section 4.2.8.1; Elliptic Curve Diffie-Hellman\n parameters are described in Section 4.2.8.2.\n\n In the ClientHello message, the \"extension_data\" field of this\n extension contains a \"KeyShareClientHello\" value:\n\n struct {\n KeyShareEntry client_shares<0..2^16-1>;\n } KeyShareClientHello;\n\n client_shares A list of offered KeyShareEntry values in descending\n order of client preference.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 53]\n\f\nInternet-Draft TLS March 2018\n\n\n This vector MAY be empty if the client is requesting a\n HelloRetryRequest. Each KeyShareEntry value MUST correspond to a\n group offered in the \"supported_groups\" extension and MUST appear in\n the same order. However, the values MAY be a non-contiguous subset\n of the \"supported_groups\" extension and MAY omit the most preferred\n groups. Such a situation could arise if the most preferred groups\n are new and unlikely to be supported in enough places to make\n pregenerating key shares for them efficient.\n\n Clients can offer as many KeyShareEntry values as the number of\n supported groups it is offering, each representing a single set of\n key exchange parameters. For instance, a client might offer shares\n for several elliptic curves or multiple FFDHE groups. The\n key_exchange values for each KeyShareEntry MUST be generated\n independently. Clients MUST NOT offer multiple KeyShareEntry values\n for the same group. Clients MUST NOT offer any KeyShareEntry values\n for groups not listed in the client's \"supported_groups\" extension.\n Servers MAY check for violations of these rules and abort the\n handshake with an \"illegal_parameter\" alert if one is violated.\n\n In a HelloRetryRequest message, the \"extension_data\" field of this\n extension contains a KeyShareHelloRetryRequest value:\n\n struct {\n NamedGroup selected_group;\n } KeyShareHelloRetryRequest;\n\n selected_group The mutually supported group the server intends to\n negotiate and is requesting a retried ClientHello/KeyShare for.\n\n Upon receipt of this extension in a HelloRetryRequest, the client\n MUST verify that (1) the selected_group field corresponds to a group\n which was provided in the \"supported_groups\" extension in the\n original ClientHello; and (2) the selected_group field does not\n correspond to a group which was provided in the \"key_share\" extension\n in the original ClientHello. If either of these checks fails, then\n the client MUST abort the handshake with an \"illegal_parameter\"\n alert. Otherwise, when sending the new ClientHello, the client MUST\n replace the original \"key_share\" extension with one containing only a\n new KeyShareEntry for the group indicated in the selected_group field\n of the triggering HelloRetryRequest.\n\n In a ServerHello message, the \"extension_data\" field of this\n extension contains a KeyShareServerHello value:\n\n struct {\n KeyShareEntry server_share;\n } KeyShareServerHello;\n\n\n\nRescorla Expires September 21, 2018 [Page 54]\n\f\nInternet-Draft TLS March 2018\n\n\n server_share A single KeyShareEntry value that is in the same group\n as one of the client's shares.\n\n If using (EC)DHE key establishment, servers offer exactly one\n KeyShareEntry in the ServerHello. This value MUST be in the same\n group as the KeyShareEntry value offered by the client that the\n server has selected for the negotiated key exchange. Servers MUST\n NOT send a KeyShareEntry for any group not indicated in the\n \"supported_groups\" extension and MUST NOT send a KeyShareEntry when\n using the \"psk_ke\" PskKeyExchangeMode. If using (EC)DHE key\n establishment, and a HelloRetryRequest containing a \"key_share\"\n extension was received by the client, the client MUST verify that the\n selected NamedGroup in the ServerHello is the same as that in the\n HelloRetryRequest. If this check fails, the client MUST abort the\n handshake with an \"illegal_parameter\" alert.\n\n4.2.8.1. Diffie-Hellman Parameters\n\n Diffie-Hellman [DH] parameters for both clients and servers are\n encoded in the opaque key_exchange field of a KeyShareEntry in a\n KeyShare structure. The opaque value contains the Diffie-Hellman\n public value (Y = g^X mod p) for the specified group (see [RFC7919]\n for group definitions) encoded as a big-endian integer and padded to\n the left with zeros to the size of p in bytes.\n\n Note: For a given Diffie-Hellman group, the padding results in all\n public keys having the same length.\n\n Peers MUST validate each other's public key Y by ensuring that 1 < Y\n < p-1. This check ensures that the remote peer is properly behaved\n and isn't forcing the local system into a small subgroup.\n\n4.2.8.2. ECDHE Parameters\n\n ECDHE parameters for both clients and servers are encoded in the\n opaque key_exchange field of a KeyShareEntry in a KeyShare structure.\n\n For secp256r1, secp384r1 and secp521r1, the contents are the\n serialized value of the following struct:\n\n struct {\n uint8 legacy_form = 4;\n opaque X[coordinate_length];\n opaque Y[coordinate_length];\n } UncompressedPointRepresentation;\n\n X and Y respectively are the binary representations of the x and y\n values in network byte order. There are no internal length markers,\n\n\n\nRescorla Expires September 21, 2018 [Page 55]\n\f\nInternet-Draft TLS March 2018\n\n\n so each number representation occupies as many octets as implied by\n the curve parameters. For P-256 this means that each of X and Y use\n 32 octets, padded on the left by zeros if necessary. For P-384 they\n take 48 octets each, and for P-521 they take 66 octets each.\n\n For the curves secp256r1, secp384r1 and secp521r1, peers MUST\n validate each other's public value Q by ensuring that the point is a\n valid point on the elliptic curve. The appropriate validation\n procedures are defined in Section 4.3.7 of [X962] and alternatively\n in Section 5.6.2.3 of [KEYAGREEMENT]. This process consists of three\n steps: (1) verify that Q is not the point at infinity (O), (2) verify\n that for Q = (x, y) both integers x and y are in the correct\n interval, (3) ensure that (x, y) is a correct solution to the\n elliptic curve equation. For these curves, implementers do not need\n to verify membership in the correct subgroup.\n\n For X25519 and X448, the contents of the public value are the byte\n string inputs and outputs of the corresponding functions defined in\n [RFC7748], 32 bytes for X25519 and 56 bytes for X448.\n\n Note: Versions of TLS prior to 1.3 permitted point format\n negotiation; TLS 1.3 removes this feature in favor of a single point\n format for each curve.\n\n4.2.9. Pre-Shared Key Exchange Modes\n\n In order to use PSKs, clients MUST also send a\n \"psk_key_exchange_modes\" extension. The semantics of this extension\n are that the client only supports the use of PSKs with these modes,\n which restricts both the use of PSKs offered in this ClientHello and\n those which the server might supply via NewSessionTicket.\n\n A client MUST provide a \"psk_key_exchange_modes\" extension if it\n offers a \"pre_shared_key\" extension. If clients offer\n \"pre_shared_key\" without a \"psk_key_exchange_modes\" extension,\n servers MUST abort the handshake. Servers MUST NOT select a key\n exchange mode that is not listed by the client. This extension also\n restricts the modes for use with PSK resumption; servers SHOULD NOT\n send NewSessionTicket with tickets that are not compatible with the\n advertised modes; however, if a server does so, the impact will just\n be that the client's attempts at resumption fail.\n\n The server MUST NOT send a \"psk_key_exchange_modes\" extension.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 56]\n\f\nInternet-Draft TLS March 2018\n\n\n enum { psk_ke(0), psk_dhe_ke(1), (255) } PskKeyExchangeMode;\n\n struct {\n PskKeyExchangeMode ke_modes<1..255>;\n } PskKeyExchangeModes;\n\n psk_ke PSK-only key establishment. In this mode, the server MUST\n NOT supply a \"key_share\" value.\n\n psk_dhe_ke PSK with (EC)DHE key establishment. In this mode, the\n client and server MUST supply \"key_share\" values as described in\n Section 4.2.8.\n\n Any future values that are allocated must ensure that the transmitted\n protocol messages unambiguously identify which mode was selected by\n the server; at present, this is indicated by the presence of the\n \"key_share\" in the ServerHello.\n\n4.2.10. Early Data Indication\n\n When a PSK is used and early data is allowed for that PSK, the client\n can send application data in its first flight of messages. If the\n client opts to do so, it MUST supply both the \"early_data\" extension\n as well as the \"pre_shared_key\" extension.\n\n The \"extension_data\" field of this extension contains an\n \"EarlyDataIndication\" value.\n\n struct {} Empty;\n\n struct {\n select (Handshake.msg_type) {\n case new_session_ticket: uint32 max_early_data_size;\n case client_hello: Empty;\n case encrypted_extensions: Empty;\n };\n } EarlyDataIndication;\n\n See Section 4.6.1 for the use of the max_early_data_size field.\n\n The parameters for the 0-RTT data (version, symmetric cipher suite,\n ALPN protocol, etc.) are those associated with the PSK in use. For\n externally provisioned PSKs, the associated values are those\n provisioned along with the key. For PSKs established via a\n NewSessionTicket message, the associated values are those which were\n negotiated in the connection which established the PSK. The PSK used\n to encrypt the early data MUST be the first PSK listed in the\n client's \"pre_shared_key\" extension.\n\n\n\nRescorla Expires September 21, 2018 [Page 57]\n\f\nInternet-Draft TLS March 2018\n\n\n For PSKs provisioned via NewSessionTicket, a server MUST validate\n that the ticket age for the selected PSK identity (computed by\n subtracting ticket_age_add from PskIdentity.obfuscated_ticket_age\n modulo 2^32) is within a small tolerance of the time since the ticket\n was issued (see Section 8). If it is not, the server SHOULD proceed\n with the handshake but reject 0-RTT, and SHOULD NOT take any other\n action that assumes that this ClientHello is fresh.\n\n 0-RTT messages sent in the first flight have the same (encrypted)\n content types as messages of the same type sent in other flights\n (handshake and application_data) but are protected under different\n keys. After receiving the server's Finished message, if the server\n has accepted early data, an EndOfEarlyData message will be sent to\n indicate the key change. This message will be encrypted with the\n 0-RTT traffic keys.\n\n A server which receives an \"early_data\" extension MUST behave in one\n of three ways:\n\n - Ignore the extension and return a regular 1-RTT response. The\n server then skips past early data by attempting to deprotect\n received records using the handshake traffic key, discarding\n records which fail deprotection (up to the configured\n max_early_data_size). Once a record is deprotected successfully,\n it is treated as the start of the client's second flight and the\n the server proceeds as with an ordinary 1-RTT handshake.\n\n - Request that the client send another ClientHello by responding\n with a HelloRetryRequest. A client MUST NOT include the\n \"early_data\" extension in its followup ClientHello. The server\n then ignores early data by skipping all records with external\n content type of \"application_data\" (indicating that they are\n encrypted), up to the configured max_early_data_size.\n\n - Return its own \"early_data\" extension in EncryptedExtensions,\n indicating that it intends to process the early data. It is not\n possible for the server to accept only a subset of the early data\n messages. Even though the server sends a message accepting early\n data, the actual early data itself may already be in flight by the\n time the server generates this message.\n\n In order to accept early data, the server MUST have accepted a PSK\n cipher suite and selected the first key offered in the client's\n \"pre_shared_key\" extension. In addition, it MUST verify that the\n following values are the same as those associated with the selected\n PSK:\n\n - The TLS version number\n\n\n\nRescorla Expires September 21, 2018 [Page 58]\n\f\nInternet-Draft TLS March 2018\n\n\n - The selected cipher suite\n\n - The selected ALPN [RFC7301] protocol, if any\n\n These requirements are a superset of those needed to perform a 1-RTT\n handshake using the PSK in question. For externally established\n PSKs, the associated values are those provisioned along with the key.\n For PSKs established via a NewSessionTicket message, the associated\n values are those negotiated in the connection during which the ticket\n was established.\n\n Future extensions MUST define their interaction with 0-RTT.\n\n If any of these checks fail, the server MUST NOT respond with the\n extension and must discard all the first flight data using one of the\n first two mechanisms listed above (thus falling back to 1-RTT or\n 2-RTT). If the client attempts a 0-RTT handshake but the server\n rejects it, the server will generally not have the 0-RTT record\n protection keys and must instead use trial decryption (either with\n the 1-RTT handshake keys or by looking for a cleartext ClientHello in\n the case of HelloRetryRequest) to find the first non-0-RTT message.\n\n If the server chooses to accept the \"early_data\" extension, then it\n MUST comply with the same error handling requirements specified for\n all records when processing early data records. Specifically, if the\n server fails to decrypt a 0-RTT record following an accepted\n \"early_data\" extension it MUST terminate the connection with a\n \"bad_record_mac\" alert as per Section 5.2.\n\n If the server rejects the \"early_data\" extension, the client\n application MAY opt to retransmit the application data previously\n sent in early data once the handshake has been completed. Note that\n automatic re-transmission of early data could result in assumptions\n about the status of the connection being incorrect. For instance,\n when the negotiated connection selects a different ALPN protocol from\n what was used for the early data, an application might need to\n construct different messages. Similarly, if early data assumes\n anything about the connection state, it might be sent in error after\n the handshake completes.\n\n A TLS implementation SHOULD NOT automatically re-send early data;\n applications are in a better position to decide when re-transmission\n is appropriate. A TLS implementation MUST NOT automatically re-send\n early data unless the negotiated connection selects the same ALPN\n protocol.\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 59]\n\f\nInternet-Draft TLS March 2018\n\n\n4.2.11. Pre-Shared Key Extension\n\n The \"pre_shared_key\" extension is used to negotiate the identity of\n the pre-shared key to be used with a given handshake in association\n with PSK key establishment.\n\n The \"extension_data\" field of this extension contains a\n \"PreSharedKeyExtension\" value:\n\n struct {\n opaque identity<1..2^16-1>;\n uint32 obfuscated_ticket_age;\n } PskIdentity;\n\n opaque PskBinderEntry<32..255>;\n\n struct {\n PskIdentity identities<7..2^16-1>;\n PskBinderEntry binders<33..2^16-1>;\n } OfferedPsks;\n\n struct {\n select (Handshake.msg_type) {\n case client_hello: OfferedPsks;\n case server_hello: uint16 selected_identity;\n };\n } PreSharedKeyExtension;\n\n identity A label for a key. For instance, a ticket defined in\n Appendix B.3.4 or a label for a pre-shared key established\n externally.\n\n obfuscated_ticket_age An obfuscated version of the age of the key.\n Section 4.2.11.1 describes how to form this value for identities\n established via the NewSessionTicket message. For identities\n established externally an obfuscated_ticket_age of 0 SHOULD be\n used, and servers MUST ignore the value.\n\n identities A list of the identities that the client is willing to\n negotiate with the server. If sent alongside the \"early_data\"\n extension (see Section 4.2.10), the first identity is the one used\n for 0-RTT data.\n\n binders A series of HMAC values, one for each PSK offered in the\n \"pre_shared_keys\" extension and in the same order, computed as\n described below.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 60]\n\f\nInternet-Draft TLS March 2018\n\n\n selected_identity The server's chosen identity expressed as a\n (0-based) index into the identities in the client's list.\n\n Each PSK is associated with a single Hash algorithm. For PSKs\n established via the ticket mechanism (Section 4.6.1), this is the KDF\n Hash algorithm on the connection where the ticket was established.\n For externally established PSKs, the Hash algorithm MUST be set when\n the PSK is established, or default to SHA-256 if no such algorithm is\n defined. The server MUST ensure that it selects a compatible PSK (if\n any) and cipher suite.\n\n In TLS versions prior to TLS 1.3, the Server Name Identification\n (SNI) value was intended to be associated with the session (Section 3\n of [RFC6066]), with the server being required to enforce that the SNI\n value associated with the session matches the one specified in the\n resumption handshake. However, in reality the implementations were\n not consistent on which of two supplied SNI values they would use,\n leading to the consistency requirement being de-facto enforced by the\n clients. In TLS 1.3, the SNI value is always explicitly specified in\n the resumption handshake, and there is no need for the server to\n associate an SNI value with the ticket. Clients, however, SHOULD\n store the SNI with the PSK to fulfill the requirements of\n Section 4.6.1.\n\n Implementor's note: when session resumption is the primary use case\n of PSKs the most straightforward way to implement the PSK/cipher\n suite matching requirements is to negotiate the cipher suite first\n and then exclude any incompatible PSKs. Any unknown PSKs (e.g., they\n are not in the PSK database or are encrypted with an unknown key)\n SHOULD simply be ignored. If no acceptable PSKs are found, the\n server SHOULD perform a non-PSK handshake if possible. If backwards\n compatibility is important, client provided, externally established\n PSKs SHOULD influence cipher suite selection.\n\n Prior to accepting PSK key establishment, the server MUST validate\n the corresponding binder value (see Section 4.2.11.2 below). If this\n value is not present or does not validate, the server MUST abort the\n handshake. Servers SHOULD NOT attempt to validate multiple binders;\n rather they SHOULD select a single PSK and validate solely the binder\n that corresponds to that PSK. See [Section 8.2] and [Appendix E.6]\n for the security rationale for this requirement. In order to accept\n PSK key establishment, the server sends a \"pre_shared_key\" extension\n indicating the selected identity.\n\n Clients MUST verify that the server's selected_identity is within the\n range supplied by the client, that the server selected a cipher suite\n indicating a Hash associated with the PSK and that a server\n \"key_share\" extension is present if required by the ClientHello\n\n\n\nRescorla Expires September 21, 2018 [Page 61]\n\f\nInternet-Draft TLS March 2018\n\n\n \"psk_key_exchange_modes\". If these values are not consistent the\n client MUST abort the handshake with an \"illegal_parameter\" alert.\n\n If the server supplies an \"early_data\" extension, the client MUST\n verify that the server's selected_identity is 0. If any other value\n is returned, the client MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n The \"pre_shared_key\" extension MUST be the last extension in the\n ClientHello (this facilitates implementation as described below).\n Servers MUST check that it is the last extension and otherwise fail\n the handshake with an \"illegal_parameter\" alert.\n\n4.2.11.1. Ticket Age\n\n The client's view of the age of a ticket is the time since the\n receipt of the NewSessionTicket message. Clients MUST NOT attempt to\n use tickets which have ages greater than the \"ticket_lifetime\" value\n which was provided with the ticket. The \"obfuscated_ticket_age\"\n field of each PskIdentity contains an obfuscated version of the\n ticket age formed by taking the age in milliseconds and adding the\n \"ticket_age_add\" value that was included with the ticket (see\n Section 4.6.1), modulo 2^32. This addition prevents passive\n observers from correlating connections unless tickets are reused.\n Note that the \"ticket_lifetime\" field in the NewSessionTicket message\n is in seconds but the \"obfuscated_ticket_age\" is in milliseconds.\n Because ticket lifetimes are restricted to a week, 32 bits is enough\n to represent any plausible age, even in milliseconds.\n\n4.2.11.2. PSK Binder\n\n The PSK binder value forms a binding between a PSK and the current\n handshake, as well as a binding between the handshake in which the\n PSK was generated (if via a NewSessionTicket message) and the current\n handshake. Each entry in the binders list is computed as an HMAC\n over a transcript hash (see Section 4.4.1) containing a partial\n ClientHello up to and including the PreSharedKeyExtension.identities\n field. That is, it includes all of the ClientHello but not the\n binders list itself. The length fields for the message (including\n the overall length, the length of the extensions block, and the\n length of the \"pre_shared_key\" extension) are all set as if binders\n of the correct lengths were present.\n\n The PskBinderEntry is computed in the same way as the Finished\n message (Section 4.4.4) but with the BaseKey being the binder_key\n derived via the key schedule from the corresponding PSK which is\n being offered (see Section 7.1).\n\n\n\n\nRescorla Expires September 21, 2018 [Page 62]\n\f\nInternet-Draft TLS March 2018\n\n\n If the handshake includes a HelloRetryRequest, the initial\n ClientHello and HelloRetryRequest are included in the transcript\n along with the new ClientHello. For instance, if the client sends\n ClientHello1, its binder will be computed over:\n\n Transcript-Hash(Truncate(ClientHello1))\n\n Where Truncate() removes the binders list from the ClientHello.\n\n If the server responds with HelloRetryRequest, and the client then\n sends ClientHello2, its binder will be computed over:\n\n Transcript-Hash(ClientHello1,\n HelloRetryRequest,\n Truncate(ClientHello2))\n\n The full ClientHello1/ClientHello2 is included in all other handshake\n hash computations. Note that in the first flight,\n Truncate(ClientHello1) is hashed directly, but in the second flight,\n ClientHello1 is hashed and then reinjected as a \"message_hash\"\n message, as described in Section 4.4.1.\n\n4.2.11.3. Processing Order\n\n Clients are permitted to \"stream\" 0-RTT data until they receive the\n server's Finished, only then sending the EndOfEarlyData message,\n followed by the rest of the handshake. In order to avoid deadlocks,\n when accepting \"early_data\", servers MUST process the client's\n ClientHello and then immediately send their flight of messages,\n rather than waiting for the client's EndOfEarlyData message before\n sending its ServerHello.\n\n4.3. Server Parameters\n\n The next two messages from the server, EncryptedExtensions and\n CertificateRequest, contain information from the server that\n determines the rest of the handshake. These messages are encrypted\n with keys derived from the server_handshake_traffic_secret.\n\n4.3.1. Encrypted Extensions\n\n In all handshakes, the server MUST send the EncryptedExtensions\n message immediately after the ServerHello message. This is the first\n message that is encrypted under keys derived from the\n server_handshake_traffic_secret.\n\n The EncryptedExtensions message contains extensions that can be\n protected, i.e., any which are not needed to establish the\n\n\n\nRescorla Expires September 21, 2018 [Page 63]\n\f\nInternet-Draft TLS March 2018\n\n\n cryptographic context, but which are not associated with individual\n certificates. The client MUST check EncryptedExtensions for the\n presence of any forbidden extensions and if any are found MUST abort\n the handshake with an \"illegal_parameter\" alert.\n\n Structure of this message:\n\n struct {\n Extension extensions<0..2^16-1>;\n } EncryptedExtensions;\n\n extensions A list of extensions. For more information, see the\n table in Section 4.2.\n\n4.3.2. Certificate Request\n\n A server which is authenticating with a certificate MAY optionally\n request a certificate from the client. This message, if sent, MUST\n follow EncryptedExtensions.\n\n Structure of this message:\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n Extension extensions<2..2^16-1>;\n } CertificateRequest;\n\n certificate_request_context An opaque string which identifies the\n certificate request and which will be echoed in the client's\n Certificate message. The certificate_request_context MUST be\n unique within the scope of this connection (thus preventing replay\n of client CertificateVerify messages). This field SHALL be zero\n length unless used for the post-handshake authentication exchanges\n described in Section 4.6.2. When requesting post-handshake\n authentication, the server SHOULD make the context unpredictable\n to the client (e.g., by randomly generating it) in order to\n prevent an attacker who has temporary access to the client's\n private key from pre-computing valid CertificateVerify messages.\n\n extensions A set of extensions describing the parameters of the\n certificate being requested. The \"signature_algorithms\" extension\n MUST be specified, and other extensions may optionally be included\n if defined for this message. Clients MUST ignore unrecognized\n extensions.\n\n In prior versions of TLS, the CertificateRequest message carried a\n list of signature algorithms and certificate authorities which the\n server would accept. In TLS 1.3 the former is expressed by sending\n\n\n\nRescorla Expires September 21, 2018 [Page 64]\n\f\nInternet-Draft TLS March 2018\n\n\n the \"signature_algorithms\" and optionally \"signature_algorithms_cert\"\n extensions. The latter is expressed by sending the\n \"certificate_authorities\" extension (see Section 4.2.4).\n\n Servers which are authenticating with a PSK MUST NOT send the\n CertificateRequest message in the main handshake, though they MAY\n send it in post-handshake authentication (see Section 4.6.2) provided\n that the client has sent the \"post_handshake_auth\" extension (see\n Section 4.2.6).\n\n4.4. Authentication Messages\n\n As discussed in Section 2, TLS generally uses a common set of\n messages for authentication, key confirmation, and handshake\n integrity: Certificate, CertificateVerify, and Finished. (The\n PreSharedKey binders also perform key confirmation, in a similar\n fashion.) These three messages are always sent as the last messages\n in their handshake flight. The Certificate and CertificateVerify\n messages are only sent under certain circumstances, as defined below.\n The Finished message is always sent as part of the Authentication\n block. These messages are encrypted under keys derived from\n [sender]_handshake_traffic_secret.\n\n The computations for the Authentication messages all uniformly take\n the following inputs:\n\n - The certificate and signing key to be used.\n\n - A Handshake Context consisting of the set of messages to be\n included in the transcript hash.\n\n - A base key to be used to compute a MAC key.\n\n Based on these inputs, the messages then contain:\n\n Certificate The certificate to be used for authentication, and any\n supporting certificates in the chain. Note that certificate-based\n client authentication is not available in PSK (including 0-RTT)\n flows.\n\n CertificateVerify A signature over the value Transcript-\n Hash(Handshake Context, Certificate)\n\n Finished A MAC over the value Transcript-Hash(Handshake Context,\n Certificate, CertificateVerify) using a MAC key derived from the\n base key.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 65]\n\f\nInternet-Draft TLS March 2018\n\n\n The following table defines the Handshake Context and MAC Base Key\n for each scenario:\n\n +-----------+----------------------------+--------------------------+\n | Mode | Handshake Context | Base Key |\n +-----------+----------------------------+--------------------------+\n | Server | ClientHello ... later of E | server_handshake_traffic |\n | | ncryptedExtensions/Certifi | _secret |\n | | cateRequest | |\n | | | |\n | Client | ClientHello ... later of | client_handshake_traffic |\n | | server | _secret |\n | | Finished/EndOfEarlyData | |\n | | | |\n | Post- | ClientHello ... client | client_application_traff |\n | Handshake | Finished + | ic_secret_N |\n | | CertificateRequest | |\n +-----------+----------------------------+--------------------------+\n\n4.4.1. The Transcript Hash\n\n Many of the cryptographic computations in TLS make use of a\n transcript hash. This value is computed by hashing the concatenation\n of each included handshake message, including the handshake message\n header carrying the handshake message type and length fields, but not\n including record layer headers. I.e.,\n\n Transcript-Hash(M1, M2, ... Mn) = Hash(M1 || M2 || ... || Mn)\n\n As an exception to this general rule, when the server responds to a\n ClientHello with a HelloRetryRequest, the value of ClientHello1 is\n replaced with a special synthetic handshake message of handshake type\n \"message_hash\" containing Hash(ClientHello1). I.e.,\n\n Transcript-Hash(ClientHello1, HelloRetryRequest, ... Mn) =\n Hash(message_hash || /* Handshake type */\n 00 00 Hash.length || /* Handshake message length (bytes) */\n Hash(ClientHello1) || /* Hash of ClientHello1 */\n HelloRetryRequest || ... || Mn)\n\n The reason for this construction is to allow the server to do a\n stateless HelloRetryRequest by storing just the hash of ClientHello1\n in the cookie, rather than requiring it to export the entire\n intermediate hash state (see Section 4.2.2).\n\n For concreteness, the transcript hash is always taken from the\n following sequence of handshake messages, starting at the first\n ClientHello and including only those messages that were sent:\n\n\n\nRescorla Expires September 21, 2018 [Page 66]\n\f\nInternet-Draft TLS March 2018\n\n\n ClientHello, HelloRetryRequest, ClientHello, ServerHello,\n EncryptedExtensions, server CertificateRequest, server Certificate,\n server CertificateVerify, server Finished, EndOfEarlyData, client\n Certificate, client CertificateVerify, client Finished.\n\n In general, implementations can implement the transcript by keeping a\n running transcript hash value based on the negotiated hash. Note,\n however, that subsequent post-handshake authentications do not\n include each other, just the messages through the end of the main\n handshake.\n\n4.4.2. Certificate\n\n This message conveys the endpoint's certificate chain to the peer.\n\n The server MUST send a Certificate message whenever the agreed-upon\n key exchange method uses certificates for authentication (this\n includes all key exchange methods defined in this document except\n PSK).\n\n The client MUST send a Certificate message if and only if the server\n has requested client authentication via a CertificateRequest message\n (Section 4.3.2). If the server requests client authentication but no\n suitable certificate is available, the client MUST send a Certificate\n message containing no certificates (i.e., with the \"certificate_list\"\n field having length 0). A Finished message MUST be sent regardless\n of whether the Certificate message is empty.\n\n Structure of this message:\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 67]\n\f\nInternet-Draft TLS March 2018\n\n\n /* Managed by IANA */\n enum {\n X509(0),\n RawPublicKey(2),\n (255)\n } CertificateType;\n\n struct {\n select (certificate_type) {\n case RawPublicKey:\n /* From RFC 7250 ASN.1_subjectPublicKeyInfo */\n opaque ASN1_subjectPublicKeyInfo<1..2^24-1>;\n\n case X509:\n opaque cert_data<1..2^24-1>;\n };\n Extension extensions<0..2^16-1>;\n } CertificateEntry;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n CertificateEntry certificate_list<0..2^24-1>;\n } Certificate;\n\n certificate_request_context If this message is in response to a\n CertificateRequest, the value of certificate_request_context in\n that message. Otherwise (in the case of server authentication),\n this field SHALL be zero length.\n\n certificate_list This is a sequence (chain) of CertificateEntry\n structures, each containing a single certificate and set of\n extensions.\n\n extensions: A set of extension values for the CertificateEntry. The\n \"Extension\" format is defined in Section 4.2. Valid extensions\n for server certificates at present include OCSP Status extension\n ([RFC6066]) and SignedCertificateTimestamps ([RFC6962]); future\n extensions may be defined for this message as well. Extensions in\n the Certificate message from the server MUST correspond to ones\n from the ClientHello message. Extensions in the Certificate from\n the client MUST correspond with extensions in the\n CertificateRequest message from the server. If an extension\n applies to the entire chain, it SHOULD be included in the first\n CertificateEntry.\n\n If the corresponding certificate type extension\n (\"server_certificate_type\" or \"client_certificate_type\") was not\n negotiated in Encrypted Extensions, or the X.509 certificate type was\n\n\n\nRescorla Expires September 21, 2018 [Page 68]\n\f\nInternet-Draft TLS March 2018\n\n\n negotiated, then each CertificateEntry contains a DER-encoded X.509\n certificate. The sender's certificate MUST come in the first\n CertificateEntry in the list. Each following certificate SHOULD\n directly certify the one immediately preceding it. Because\n certificate validation requires that trust anchors be distributed\n independently, a certificate that specifies a trust anchor MAY be\n omitted from the chain, provided that supported peers are known to\n possess any omitted certificates.\n\n Note: Prior to TLS 1.3, \"certificate_list\" ordering required each\n certificate to certify the one immediately preceding it; however,\n some implementations allowed some flexibility. Servers sometimes\n send both a current and deprecated intermediate for transitional\n purposes, and others are simply configured incorrectly, but these\n cases can nonetheless be validated properly. For maximum\n compatibility, all implementations SHOULD be prepared to handle\n potentially extraneous certificates and arbitrary orderings from any\n TLS version, with the exception of the end-entity certificate which\n MUST be first.\n\n If the RawPublicKey certificate type was negotiated, then the\n certificate_list MUST contain no more than one CertificateEntry,\n which contains an ASN1_subjectPublicKeyInfo value as defined in\n [RFC7250], Section 3.\n\n The OpenPGP certificate type [RFC6091] MUST NOT be used with TLS 1.3.\n\n The server's certificate_list MUST always be non-empty. A client\n will send an empty certificate_list if it does not have an\n appropriate certificate to send in response to the server's\n authentication request.\n\n4.4.2.1. OCSP Status and SCT Extensions\n\n [RFC6066] and [RFC6961] provide extensions to negotiate the server\n sending OCSP responses to the client. In TLS 1.2 and below, the\n server replies with an empty extension to indicate negotiation of\n this extension and the OCSP information is carried in a\n CertificateStatus message. In TLS 1.3, the server's OCSP information\n is carried in an extension in the CertificateEntry containing the\n associated certificate. Specifically: The body of the\n \"status_request\" extension from the server MUST be a\n CertificateStatus structure as defined in [RFC6066], which is\n interpreted as defined in [RFC6960].\n\n Note: status_request_v2 extension ([RFC6961]) is deprecated. TLS 1.3\n servers MUST NOT act upon its presence or information in it when\n processing Client Hello, in particular they MUST NOT send the\n\n\n\nRescorla Expires September 21, 2018 [Page 69]\n\f\nInternet-Draft TLS March 2018\n\n\n status_request_v2 extension in the Encrypted Extensions, Certificate\n Request or the Certificate messages. TLS 1.3 servers MUST be able to\n process Client Hello messages that include it, as it MAY be sent by\n clients that wish to use it in earlier protocol versions.\n\n A server MAY request that a client present an OCSP response with its\n certificate by sending an empty \"status_request\" extension in its\n CertificateRequest message. If the client opts to send an OCSP\n response, the body of its \"status_request\" extension MUST be a\n CertificateStatus structure as defined in [RFC6066].\n\n Similarly, [RFC6962] provides a mechanism for a server to send a\n Signed Certificate Timestamp (SCT) as an extension in the ServerHello\n in TLS 1.2 and below. In TLS 1.3, the server's SCT information is\n carried in an extension in CertificateEntry.\n\n4.4.2.2. Server Certificate Selection\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3 [RFC5280], unless explicitly\n negotiated otherwise (e.g., [RFC7250]).\n\n - The server's end-entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected authentication\n algorithm from the client's \"signature_algorithms\" extension\n (currently RSA, ECDSA, or EdDSA).\n\n - The certificate MUST allow the key to be used for signing (i.e.,\n the digitalSignature bit MUST be set if the Key Usage extension is\n present) with a signature scheme indicated in the client's\n \"signature_algorithms\"/\"signature_algorithms_cert\" extensions (see\n Section 4.2.3).\n\n - The \"server_name\" [RFC6066] and \"certificate_authorities\"\n extensions are used to guide certificate selection. As servers\n MAY require the presence of the \"server_name\" extension, clients\n SHOULD send this extension, when applicable.\n\n All certificates provided by the server MUST be signed by a signature\n algorithm advertised by the client, if it is able to provide such a\n chain (see Section 4.2.3). Certificates that are self-signed or\n certificates that are expected to be trust anchors are not validated\n as part of the chain and therefore MAY be signed with any algorithm.\n\n If the server cannot produce a certificate chain that is signed only\n via the indicated supported algorithms, then it SHOULD continue the\n handshake by sending the client a certificate chain of its choice\n\n\n\nRescorla Expires September 21, 2018 [Page 70]\n\f\nInternet-Draft TLS March 2018\n\n\n that may include algorithms that are not known to be supported by the\n client. This fallback chain SHOULD NOT use the deprecated SHA-1 hash\n algorithm in general, but MAY do so if the client's advertisement\n permits it, and MUST NOT do so otherwise.\n\n If the client cannot construct an acceptable chain using the provided\n certificates and decides to abort the handshake, then it MUST abort\n the handshake with an appropriate certificate-related alert (by\n default, \"unsupported_certificate\"; see Section 6.2 for more).\n\n If the server has multiple certificates, it chooses one of them based\n on the above-mentioned criteria (in addition to other criteria, such\n as transport layer endpoint, local configuration and preferences).\n\n4.4.2.3. Client Certificate Selection\n\n The following rules apply to certificates sent by the client:\n\n - The certificate type MUST be X.509v3 [RFC5280], unless explicitly\n negotiated otherwise (e.g., [RFC7250]).\n\n - If the \"certificate_authorities\" extension in the\n CertificateRequest message was present, at least one of the\n certificates in the certificate chain SHOULD be issued by one of\n the listed CAs.\n\n - The certificates MUST be signed using an acceptable signature\n algorithm, as described in Section 4.3.2. Note that this relaxes\n the constraints on certificate-signing algorithms found in prior\n versions of TLS.\n\n - If the CertificateRequest message contained a non-empty\n \"oid_filters\" extension, the end-entity certificate MUST match the\n extension OIDs that are recognized by the client, as described in\n Section 4.2.5.\n\n Note that, as with the server certificate, there are certificates\n that use algorithm combinations that cannot be currently used with\n TLS.\n\n4.4.2.4. Receiving a Certificate Message\n\n In general, detailed certificate validation procedures are out of\n scope for TLS (see [RFC5280]). This section provides TLS-specific\n requirements.\n\n If the server supplies an empty Certificate message, the client MUST\n abort the handshake with a \"decode_error\" alert.\n\n\n\nRescorla Expires September 21, 2018 [Page 71]\n\f\nInternet-Draft TLS March 2018\n\n\n If the client does not send any certificates (i.e., it sends an empty\n Certificate message), the server MAY at its discretion either\n continue the handshake without client authentication, or abort the\n handshake with a \"certificate_required\" alert. Also, if some aspect\n of the certificate chain was unacceptable (e.g., it was not signed by\n a known, trusted CA), the server MAY at its discretion either\n continue the handshake (considering the client unauthenticated) or\n abort the handshake.\n\n Any endpoint receiving any certificate which it would need to\n validate using any signature algorithm using an MD5 hash MUST abort\n the handshake with a \"bad_certificate\" alert. SHA-1 is deprecated\n and it is RECOMMENDED that any endpoint receiving any certificate\n which it would need to validate using any signature algorithm using a\n SHA-1 hash abort the handshake with a \"bad_certificate\" alert. For\n clarity, this means that endpoints MAY accept these algorithms for\n certificates that are self-signed or are trust anchors.\n\n All endpoints are RECOMMENDED to transition to SHA-256 or better as\n soon as possible to maintain interoperability with implementations\n currently in the process of phasing out SHA-1 support.\n\n Note that a certificate containing a key for one signature algorithm\n MAY be signed using a different signature algorithm (for instance, an\n RSA key signed with an ECDSA key).\n\n4.4.3. Certificate Verify\n\n This message is used to provide explicit proof that an endpoint\n possesses the private key corresponding to its certificate. The\n CertificateVerify message also provides integrity for the handshake\n up to this point. Servers MUST send this message when authenticating\n via a certificate. Clients MUST send this message whenever\n authenticating via a certificate (i.e., when the Certificate message\n is non-empty). When sent, this message MUST appear immediately after\n the Certificate message and immediately prior to the Finished\n message.\n\n Structure of this message:\n\n struct {\n SignatureScheme algorithm;\n opaque signature<0..2^16-1>;\n } CertificateVerify;\n\n The algorithm field specifies the signature algorithm used (see\n Section 4.2.3 for the definition of this field). The signature is a\n digital signature using that algorithm. The content that is covered\n\n\n\nRescorla Expires September 21, 2018 [Page 72]\n\f\nInternet-Draft TLS March 2018\n\n\n under the signature is the hash output as described in Section 4.4.1,\n namely:\n\n Transcript-Hash(Handshake Context, Certificate)\n\n The digital signature is then computed over the concatenation of:\n\n - A string that consists of octet 32 (0x20) repeated 64 times\n\n - The context string\n\n - A single 0 byte which serves as the separator\n\n - The content to be signed\n\n This structure is intended to prevent an attack on previous versions\n of TLS in which the ServerKeyExchange format meant that attackers\n could obtain a signature of a message with a chosen 32-byte prefix\n (ClientHello.random). The initial 64-byte pad clears that prefix\n along with the server-controlled ServerHello.random.\n\n The context string for a server signature is: \"TLS 1.3, server\n CertificateVerify\" The context string for a client signature is: \"TLS\n 1.3, client CertificateVerify\" It is used to provide separation\n between signatures made in different contexts, helping against\n potential cross-protocol attacks.\n\n For example, if the transcript hash was 32 bytes of 01 (this length\n would make sense for SHA-256), the content covered by the digital\n signature for a server CertificateVerify would be:\n\n 2020202020202020202020202020202020202020202020202020202020202020\n 2020202020202020202020202020202020202020202020202020202020202020\n 544c5320312e332c207365727665722043657274696669636174655665726966\n 79\n 00\n 0101010101010101010101010101010101010101010101010101010101010101\n\n On the sender side the process for computing the signature field of\n the CertificateVerify message takes as input:\n\n - The content covered by the digital signature\n\n - The private signing key corresponding to the certificate sent in\n the previous message\n\n If the CertificateVerify message is sent by a server, the signature\n algorithm MUST be one offered in the client's \"signature_algorithms\"\n\n\n\nRescorla Expires September 21, 2018 [Page 73]\n\f\nInternet-Draft TLS March 2018\n\n\n extension unless no valid certificate chain can be produced without\n unsupported algorithms (see Section 4.2.3).\n\n If sent by a client, the signature algorithm used in the signature\n MUST be one of those present in the supported_signature_algorithms\n field of the \"signature_algorithms\" extension in the\n CertificateRequest message.\n\n In addition, the signature algorithm MUST be compatible with the key\n in the sender's end-entity certificate. RSA signatures MUST use an\n RSASSA-PSS algorithm, regardless of whether RSASSA-PKCS1-v1_5\n algorithms appear in \"signature_algorithms\". The SHA-1 algorithm\n MUST NOT be used in any signatures of CertificateVerify messages.\n All SHA-1 signature algorithms in this specification are defined\n solely for use in legacy certificates and are not valid for\n CertificateVerify signatures.\n\n The receiver of a CertificateVerify message MUST verify the signature\n field. The verification process takes as input:\n\n - The content covered by the digital signature\n\n - The public key contained in the end-entity certificate found in\n the associated Certificate message.\n\n - The digital signature received in the signature field of the\n CertificateVerify message\n\n If the verification fails, the receiver MUST terminate the handshake\n with a \"decrypt_error\" alert.\n\n4.4.4. Finished\n\n The Finished message is the final message in the authentication\n block. It is essential for providing authentication of the handshake\n and of the computed keys.\n\n Recipients of Finished messages MUST verify that the contents are\n correct and if incorrect MUST terminate the connection with a\n \"decrypt_error\" alert.\n\n Once a side has sent its Finished message and received and validated\n the Finished message from its peer, it may begin to send and receive\n application data over the connection. There are two settings in\n which it is permitted to send data prior to receiving the peer's\n Finished:\n\n 1. Clients sending 0-RTT data as described in Section 4.2.10.\n\n\n\nRescorla Expires September 21, 2018 [Page 74]\n\f\nInternet-Draft TLS March 2018\n\n\n 2. Servers MAY send data after sending their first flight, but\n because the handshake is not yet complete, they have no assurance\n of either the peer's identity or of its liveness (i.e., the\n ClientHello might have been replayed).\n\n The key used to compute the Finished message is computed from the\n Base key defined in Section 4.4 using HKDF (see Section 7.1).\n Specifically:\n\n finished_key =\n HKDF-Expand-Label(BaseKey, \"finished\", \"\", Hash.length)\n\n Structure of this message:\n\n struct {\n opaque verify_data[Hash.length];\n } Finished;\n\n The verify_data value is computed as follows:\n\n verify_data =\n HMAC(finished_key,\n Transcript-Hash(Handshake Context,\n Certificate*, CertificateVerify*))\n\n * Only included if present.\n\n HMAC [RFC2104] uses the Hash algorithm for the handshake. As noted\n above, the HMAC input can generally be implemented by a running hash,\n i.e., just the handshake hash at this point.\n\n In previous versions of TLS, the verify_data was always 12 octets\n long. In TLS 1.3, it is the size of the HMAC output for the Hash\n used for the handshake.\n\n Note: Alerts and any other record types are not handshake messages\n and are not included in the hash computations.\n\n Any records following a Finished message MUST be encrypted under the\n appropriate application traffic key as described in Section 7.2. In\n particular, this includes any alerts sent by the server in response\n to client Certificate and CertificateVerify messages.\n\n4.5. End of Early Data\n\n struct {} EndOfEarlyData;\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 75]\n\f\nInternet-Draft TLS March 2018\n\n\n If the server sent an \"early_data\" extension, the client MUST send an\n EndOfEarlyData message after receiving the server Finished. If the\n server does not send an \"early_data\" extension, then the client MUST\n NOT send an EndOfEarlyData message. This message indicates that all\n 0-RTT application_data messages, if any, have been transmitted and\n that the following records are protected under handshake traffic\n keys. Servers MUST NOT send this message and clients receiving it\n MUST terminate the connection with an \"unexpected_message\" alert.\n This message is encrypted under keys derived from the\n client_early_traffic_secret.\n\n4.6. Post-Handshake Messages\n\n TLS also allows other messages to be sent after the main handshake.\n These messages use a handshake content type and are encrypted under\n the appropriate application traffic key.\n\n4.6.1. New Session Ticket Message\n\n At any time after the server has received the client Finished\n message, it MAY send a NewSessionTicket message. This message\n creates a unique association between the ticket value and a secret\n PSK derived from the resumption master secret (see Section 7.\n\n The client MAY use this PSK for future handshakes by including the\n ticket value in the \"pre_shared_key\" extension in its ClientHello\n (Section 4.2.11). Servers MAY send multiple tickets on a single\n connection, either immediately after each other or after specific\n events (see Appendix C.4). For instance, the server might send a new\n ticket after post-handshake authentication in order to encapsulate\n the additional client authentication state. Multiple tickets are\n useful for clients for a variety of purposes, including:\n\n - Opening multiple parallel HTTP connections.\n\n - Performing connection racing across interfaces and address\n families via, e.g., Happy Eyeballs [RFC8305] or related\n techniques.\n\n Any ticket MUST only be resumed with a cipher suite that has the same\n KDF hash algorithm as that used to establish the original connection.\n\n Clients MUST only resume if the new SNI value is valid for the server\n certificate presented in the original session, and SHOULD only resume\n if the SNI value matches the one used in the original session. The\n latter is a performance optimization: normally, there is no reason to\n expect that different servers covered by a single certificate would\n be able to accept each other's tickets, hence attempting resumption\n\n\n\nRescorla Expires September 21, 2018 [Page 76]\n\f\nInternet-Draft TLS March 2018\n\n\n in that case would waste a single-use ticket. If such an indication\n is provided (externally or by any other means), clients MAY resume\n with a different SNI value.\n\n On resumption, if reporting an SNI value to the calling application,\n implementations MUST use the value sent in the resumption ClientHello\n rather than the value sent in the previous session. Note that if a\n server implementation declines all PSK identities with different SNI\n values, these two values are always the same.\n\n Note: Although the resumption master secret depends on the client's\n second flight, servers which do not request client authentication MAY\n compute the remainder of the transcript independently and then send a\n NewSessionTicket immediately upon sending its Finished rather than\n waiting for the client Finished. This might be appropriate in cases\n where the client is expected to open multiple TLS connections in\n parallel and would benefit from the reduced overhead of a resumption\n handshake, for example.\n\n struct {\n uint32 ticket_lifetime;\n uint32 ticket_age_add;\n opaque ticket_nonce<0..255>;\n opaque ticket<1..2^16-1>;\n Extension extensions<0..2^16-2>;\n } NewSessionTicket;\n\n ticket_lifetime Indicates the lifetime in seconds as a 32-bit\n unsigned integer in network byte order from the time of ticket\n issuance. Servers MUST NOT use any value greater than 604800\n seconds (7 days). The value of zero indicates that the ticket\n should be discarded immediately. Clients MUST NOT cache tickets\n for longer than 7 days, regardless of the ticket_lifetime, and MAY\n delete tickets earlier based on local policy. A server MAY treat\n a ticket as valid for a shorter period of time than what is stated\n in the ticket_lifetime.\n\n ticket_age_add A securely generated, random 32-bit value that is\n used to obscure the age of the ticket that the client includes in\n the \"pre_shared_key\" extension. The client-side ticket age is\n added to this value modulo 2^32 to obtain the value that is\n transmitted by the client. The server MUST generate a fresh value\n for each ticket it sends.\n\n ticket_nonce A per-ticket value that is unique across all tickets\n issued on this connection.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 77]\n\f\nInternet-Draft TLS March 2018\n\n\n ticket The value of the ticket to be used as the PSK identity. The\n ticket itself is an opaque label. It MAY either be a database\n lookup key or a self-encrypted and self-authenticated value.\n Section 4 of [RFC5077] describes a recommended ticket construction\n mechanism.\n\n extensions A set of extension values for the ticket. The\n \"Extension\" format is defined in Section 4.2. Clients MUST ignore\n unrecognized extensions.\n\n The sole extension currently defined for NewSessionTicket is\n \"early_data\", indicating that the ticket may be used to send 0-RTT\n data (Section 4.2.10)). It contains the following value:\n\n max_early_data_size The maximum amount of 0-RTT data that the client\n is allowed to send when using this ticket, in bytes. Only\n Application Data payload (i.e., plaintext but not padding or the\n inner content type byte) is counted. A server receiving more than\n max_early_data_size bytes of 0-RTT data SHOULD terminate the\n connection with an \"unexpected_message\" alert. Note that servers\n that reject early data due to lack of cryptographic material will\n be unable to differentiate padding from content, so clients SHOULD\n NOT depend on being able to send large quantities of padding in\n early data records.\n\n The PSK associated with the ticket is computed as:\n\n HKDF-Expand-Label(resumption_master_secret,\n \"resumption\", ticket_nonce, Hash.length)\n\n Because the ticket_nonce value is distinct for each NewSessionTicket\n message, a different PSK will be derived for each ticket.\n\n Note that in principle it is possible to continue issuing new tickets\n which indefinitely extend the lifetime of the keying material\n originally derived from an initial non-PSK handshake (which was most\n likely tied to the peer's certificate). It is RECOMMENDED that\n implementations place limits on the total lifetime of such keying\n material; these limits should take into account the lifetime of the\n peer's certificate, the likelihood of intervening revocation, and the\n time since the peer's online CertificateVerify signature.\n\n4.6.2. Post-Handshake Authentication\n\n When the client has sent the \"post_handshake_auth\" extension (see\n Section 4.2.6), a server MAY request client authentication at any\n time after the handshake has completed by sending a\n CertificateRequest message. The client MUST respond with the\n\n\n\nRescorla Expires September 21, 2018 [Page 78]\n\f\nInternet-Draft TLS March 2018\n\n\n appropriate Authentication messages (see Section 4.4). If the client\n chooses to authenticate, it MUST send Certificate, CertificateVerify,\n and Finished. If it declines, it MUST send a Certificate message\n containing no certificates followed by Finished. All of the client's\n messages for a given response MUST appear consecutively on the wire\n with no intervening messages of other types.\n\n A client that receives a CertificateRequest message without having\n sent the \"post_handshake_auth\" extension MUST send an\n \"unexpected_message\" fatal alert.\n\n Note: Because client authentication could involve prompting the user,\n servers MUST be prepared for some delay, including receiving an\n arbitrary number of other messages between sending the\n CertificateRequest and receiving a response. In addition, clients\n which receive multiple CertificateRequests in close succession MAY\n respond to them in a different order than they were received (the\n certificate_request_context value allows the server to disambiguate\n the responses).\n\n4.6.3. Key and IV Update\n\n enum {\n update_not_requested(0), update_requested(1), (255)\n } KeyUpdateRequest;\n\n struct {\n KeyUpdateRequest request_update;\n } KeyUpdate;\n\n request_update Indicates whether the recipient of the KeyUpdate\n should respond with its own KeyUpdate. If an implementation\n receives any other value, it MUST terminate the connection with an\n \"illegal_parameter\" alert.\n\n The KeyUpdate handshake message is used to indicate that the sender\n is updating its sending cryptographic keys. This message can be sent\n by either peer after it has sent a Finished message. Implementations\n that receive a KeyUpdate message prior to receiving a Finished\n message MUST terminate the connection with an \"unexpected_message\"\n alert. After sending a KeyUpdate message, the sender SHALL send all\n its traffic using the next generation of keys, computed as described\n in Section 7.2. Upon receiving a KeyUpdate, the receiver MUST update\n its receiving keys.\n\n If the request_update field is set to \"update_requested\" then the\n receiver MUST send a KeyUpdate of its own with request_update set to\n \"update_not_requested\" prior to sending its next application data\n\n\n\nRescorla Expires September 21, 2018 [Page 79]\n\f\nInternet-Draft TLS March 2018\n\n\n record. This mechanism allows either side to force an update to the\n entire connection, but causes an implementation which receives\n multiple KeyUpdates while it is silent to respond with a single\n update. Note that implementations may receive an arbitrary number of\n messages between sending a KeyUpdate with request_update set to\n update_requested and receiving the peer's KeyUpdate, because those\n messages may already be in flight. However, because send and receive\n keys are derived from independent traffic secrets, retaining the\n receive traffic secret does not threaten the forward secrecy of data\n sent before the sender changed keys.\n\n If implementations independently send their own KeyUpdates with\n request_update set to \"update_requested\", and they cross in flight,\n then each side will also send a response, with the result that each\n side increments by two generations.\n\n Both sender and receiver MUST encrypt their KeyUpdate messages with\n the old keys. Additionally, both sides MUST enforce that a KeyUpdate\n with the old key is received before accepting any messages encrypted\n with the new key. Failure to do so may allow message truncation\n attacks.\n\n5. Record Protocol\n\n The TLS record protocol takes messages to be transmitted, fragments\n the data into manageable blocks, protects the records, and transmits\n the result. Received data is verified, decrypted, reassembled, and\n then delivered to higher-level clients.\n\n TLS records are typed, which allows multiple higher-level protocols\n to be multiplexed over the same record layer. This document\n specifies four content types: handshake, application data, alert, and\n change_cipher_spec. The change_cipher_spec record is used only for\n compatibility purposes (see Appendix D.4).\n\n An implementation may receive an unencrypted record of type\n change_cipher_spec consisting of the single byte value 0x01 at any\n time after the first ClientHello message has been sent or received\n and before the peer's Finished message has been received and MUST\n simply drop it without further processing. Note that this record may\n appear at a point at the handshake where the implementation is\n expecting protected records and so it is necessary to detect this\n condition prior to attempting to deprotect the record. An\n implementation which receives any other change_cipher_spec value or\n which receives a protected change_cipher_spec record MUST abort the\n handshake with an \"unexpected_message\" alert. A change_cipher_spec\n record received before the first ClientHello message or after the\n peer's Finished message MUST be treated as an unexpected record type\n\n\n\nRescorla Expires September 21, 2018 [Page 80]\n\f\nInternet-Draft TLS March 2018\n\n\n (though stateless servers may not be able to distinguish these cases\n from allowed cases).\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST terminate\n the connection with an \"unexpected_message\" alert. New record\n content type values are assigned by IANA in the TLS Content Type\n Registry as described in Section 11.\n\n5.1. Record Layer\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Message\n boundaries are handled differently depending on the underlying\n ContentType. Any future content types MUST specify appropriate\n rules. Note that these rules are stricter than what was enforced in\n TLS 1.2.\n\n Handshake messages MAY be coalesced into a single TLSPlaintext record\n or fragmented across several records, provided that:\n\n - Handshake messages MUST NOT be interleaved with other record\n types. That is, if a handshake message is split over two or more\n records, there MUST NOT be any other records between them.\n\n - Handshake messages MUST NOT span key changes. Implementations\n MUST verify that all messages immediately preceding a key change\n align with a record boundary; if not, then they MUST terminate the\n connection with an \"unexpected_message\" alert. Because the\n ClientHello, EndOfEarlyData, ServerHello, Finished, and KeyUpdate\n messages can immediately precede a key change, implementations\n MUST send these messages in alignment with a record boundary.\n\n Implementations MUST NOT send zero-length fragments of Handshake\n types, even if those fragments contain padding.\n\n Alert messages (Section 6) MUST NOT be fragmented across records and\n multiple Alert messages MUST NOT be coalesced into a single\n TLSPlaintext record. In other words, a record with an Alert type\n MUST contain exactly one message.\n\n Application Data messages contain data that is opaque to TLS.\n Application Data messages are always protected. Zero-length\n fragments of Application Data MAY be sent as they are potentially\n useful as a traffic analysis countermeasure. Application Data\n fragments MAY be split across multiple records or coalesced into a\n single record.\n\n\n\nRescorla Expires September 21, 2018 [Page 81]\n\f\nInternet-Draft TLS March 2018\n\n\n enum {\n invalid(0),\n change_cipher_spec(20),\n alert(21),\n handshake(22),\n application_data(23),\n (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion legacy_record_version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type The higher-level protocol used to process the enclosed\n fragment.\n\n legacy_record_version This value MUST be set to 0x0303 for all\n records generated by a TLS 1.3 implementation other than an\n initial ClientHello (i.e., one not generated after a\n HelloRetryRequest), where it MAY also be 0x0301 for compatibility\n purposes. This field is deprecated and MUST be ignored for all\n purposes. Previous versions of TLS would use other values in this\n field under some circumstances.\n\n length The length (in bytes) of the following TLSPlaintext.fragment.\n The length MUST NOT exceed 2^14 bytes. An endpoint that receives\n a record that exceeds this length MUST terminate the connection\n with a \"record_overflow\" alert.\n\n fragment The data being transmitted. This value is transparent and\n is treated as an independent block to be dealt with by the higher-\n level protocol specified by the type field.\n\n This document describes TLS 1.3, which uses the version 0x0304. This\n version value is historical, deriving from the use of 0x0301 for TLS\n 1.0 and 0x0300 for SSL 3.0. In order to maximize backwards\n compatibility, records containing an initial ClientHello SHOULD have\n version 0x0301 and a record containing a second ClientHello or a\n ServerHello MUST have version 0x0303, reflecting TLS 1.0 and TLS 1.2\n respectively. When negotiating prior versions of TLS, endpoints\n follow the procedure and requirements in Appendix D.\n\n When record protection has not yet been engaged, TLSPlaintext\n structures are written directly onto the wire. Once record\n protection has started, TLSPlaintext records are protected and sent\n\n\n\nRescorla Expires September 21, 2018 [Page 82]\n\f\nInternet-Draft TLS March 2018\n\n\n as described in the following section. Note that application data\n records MUST NOT be written to the wire unprotected (see Section 2\n for details).\n\n5.2. Record Payload Protection\n\n The record protection functions translate a TLSPlaintext structure\n into a TLSCiphertext. The deprotection functions reverse the\n process. In TLS 1.3, as opposed to previous versions of TLS, all\n ciphers are modeled as \"Authenticated Encryption with Additional\n Data\" (AEAD) [RFC5116]. AEAD functions provide an unified encryption\n and authentication operation which turns plaintext into authenticated\n ciphertext and back again. Each encrypted record consists of a\n plaintext header followed by an encrypted body, which itself contains\n a type and optional padding.\n\n struct {\n opaque content[TLSPlaintext.length];\n ContentType type;\n uint8 zeros[length_of_padding];\n } TLSInnerPlaintext;\n\n struct {\n ContentType opaque_type = application_data; /* 23 */\n ProtocolVersion legacy_record_version = 0x0303; /* TLS v1.2 */\n uint16 length;\n opaque encrypted_record[TLSCiphertext.length];\n } TLSCiphertext;\n\n content The TLSPLaintext.fragment value, containing the byte\n encoding of a handshake or an alert message, or the raw bytes of\n the application's data to send.\n\n type The TLSPlaintext.type value containing the content type of the\n record.\n\n zeros An arbitrary-length run of zero-valued bytes may appear in the\n cleartext after the type field. This provides an opportunity for\n senders to pad any TLS record by a chosen amount as long as the\n total stays within record size limits. See Section 5.4 for more\n details.\n\n opaque_type The outer opaque_type field of a TLSCiphertext record is\n always set to the value 23 (application_data) for outward\n compatibility with middleboxes accustomed to parsing previous\n versions of TLS. The actual content type of the record is found\n in TLSInnerPlaintext.type after decryption.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 83]\n\f\nInternet-Draft TLS March 2018\n\n\n legacy_record_version The legacy_record_version field is always\n 0x0303. TLS 1.3 TLSCiphertexts are not generated until after TLS\n 1.3 has been negotiated, so there are no historical compatibility\n concerns where other values might be received. Note that the\n handshake protocol including the ClientHello and ServerHello\n messages authenticates the protocol version, so this value is\n redundant.\n\n length The length (in bytes) of the following\n TLSCiphertext.encrypted_record, which is the sum of the lengths of\n the content and the padding, plus one for the inner content type,\n plus any expansion added by the AEAD algorithm. The length MUST\n NOT exceed 2^14 + 256 bytes. An endpoint that receives a record\n that exceeds this length MUST terminate the connection with a\n \"record_overflow\" alert.\n\n encrypted_record The AEAD-encrypted form of the serialized\n TLSInnerPlaintext structure.\n\n AEAD algorithms take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [RFC5116]. The key is either the\n client_write_key or the server_write_key, the nonce is derived from\n the sequence number and the client_write_iv or server_write_iv (see\n Section 5.3), and the additional data input is the record header.\n I.e.,\n\n additional_data = TLSCiphertext.opaque_type ||\n TLSCiphertext.legacy_record_version ||\n TLSCiphertext.length\n\n The plaintext input to the AEAD algorithm is the encoded\n TLSInnerPlaintext structure. Derivation of traffic keys is defined\n in Section 7.3.\n\n The AEAD output consists of the ciphertext output from the AEAD\n encryption operation. The length of the plaintext is greater than\n the corresponding TLSPlaintext.length due to the inclusion of\n TLSInnerPlaintext.type and any padding supplied by the sender. The\n length of the AEAD output will generally be larger than the\n plaintext, but by an amount that varies with the AEAD algorithm.\n Since the ciphers might incorporate padding, the amount of overhead\n could vary with different lengths of plaintext. Symbolically,\n\n AEADEncrypted =\n AEAD-Encrypt(write_key, nonce, additional_data, plaintext)\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 84]\n\f\nInternet-Draft TLS March 2018\n\n\n Then the encrypted_record field of TLSCiphertext is set to\n AEADEncrypted.\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, additional data, and the AEADEncrypted value. The output is\n either the plaintext or an error indicating that the decryption\n failed. There is no separate integrity check. That is:\n\n plaintext of encrypted_record =\n AEAD-Decrypt(peer_write_key, nonce, additional_data, AEADEncrypted)\n\n If the decryption fails, the receiver MUST terminate the connection\n with a \"bad_record_mac\" alert.\n\n An AEAD algorithm used in TLS 1.3 MUST NOT produce an expansion\n greater than 255 octets. An endpoint that receives a record from its\n peer with TLSCiphertext.length larger than 2^14 + 256 octets MUST\n terminate the connection with a \"record_overflow\" alert. This limit\n is derived from the maximum TLSInnerPlaintext length of 2^14 octets +\n 1 octet for ContentType + the maximum AEAD expansion of 255 octets.\n\n5.3. Per-Record Nonce\n\n A 64-bit sequence number is maintained separately for reading and\n writing records. The appropriate sequence number is incremented by\n one after reading or writing each record. Each sequence number is\n set to zero at the beginning of a connection and whenever the key is\n changed; the first record transmitted under a particular traffic key\n MUST use sequence number 0.\n\n Because the size of sequence numbers is 64-bit, they should not wrap.\n If a TLS implementation would need to wrap a sequence number, it MUST\n either re-key (Section 4.6.3) or terminate the connection.\n\n Each AEAD algorithm will specify a range of possible lengths for the\n per-record nonce, from N_MIN bytes to N_MAX bytes of input\n ([RFC5116]). The length of the TLS per-record nonce (iv_length) is\n set to the larger of 8 bytes and N_MIN for the AEAD algorithm (see\n [RFC5116] Section 4). An AEAD algorithm where N_MAX is less than 8\n bytes MUST NOT be used with TLS. The per-record nonce for the AEAD\n construction is formed as follows:\n\n 1. The 64-bit record sequence number is encoded in network byte\n order and padded to the left with zeros to iv_length.\n\n 2. The padded sequence number is XORed with the static\n client_write_iv or server_write_iv, depending on the role.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 85]\n\f\nInternet-Draft TLS March 2018\n\n\n The resulting quantity (of length iv_length) is used as the per-\n record nonce.\n\n Note: This is a different construction from that in TLS 1.2, which\n specified a partially explicit nonce.\n\n5.4. Record Padding\n\n All encrypted TLS records can be padded to inflate the size of the\n TLSCiphertext. This allows the sender to hide the size of the\n traffic from an observer.\n\n When generating a TLSCiphertext record, implementations MAY choose to\n pad. An unpadded record is just a record with a padding length of\n zero. Padding is a string of zero-valued bytes appended to the\n ContentType field before encryption. Implementations MUST set the\n padding octets to all zeros before encrypting.\n\n Application Data records may contain a zero-length\n TLSInnerPlaintext.content if the sender desires. This permits\n generation of plausibly-sized cover traffic in contexts where the\n presence or absence of activity may be sensitive. Implementations\n MUST NOT send Handshake or Alert records that have a zero-length\n TLSInnerPlaintext.content; if such a message is received, the\n receiving implementation MUST terminate the connection with an\n \"unexpected_message\" alert.\n\n The padding sent is automatically verified by the record protection\n mechanism; upon successful decryption of a\n TLSCiphertext.encrypted_record, the receiving implementation scans\n the field from the end toward the beginning until it finds a non-zero\n octet. This non-zero octet is the content type of the message. This\n padding scheme was selected because it allows padding of any\n encrypted TLS record by an arbitrary size (from zero up to TLS record\n size limits) without introducing new content types. The design also\n enforces all-zero padding octets, which allows for quick detection of\n padding errors.\n\n Implementations MUST limit their scanning to the cleartext returned\n from the AEAD decryption. If a receiving implementation does not\n find a non-zero octet in the cleartext, it MUST terminate the\n connection with an \"unexpected_message\" alert.\n\n The presence of padding does not change the overall record size\n limitations - the full encoded TLSInnerPlaintext MUST NOT exceed 2^14\n + 1 octets. If the maximum fragment length is reduced, as for\n example by the max_fragment_length extension from [RFC6066], then the\n\n\n\n\nRescorla Expires September 21, 2018 [Page 86]\n\f\nInternet-Draft TLS March 2018\n\n\n reduced limit applies to the full plaintext, including the content\n type and padding.\n\n Selecting a padding policy that suggests when and how much to pad is\n a complex topic and is beyond the scope of this specification. If\n the application layer protocol on top of TLS has its own padding, it\n may be preferable to pad application_data TLS records within the\n application layer. Padding for encrypted handshake and alert TLS\n records must still be handled at the TLS layer, though. Later\n documents may define padding selection algorithms or define a padding\n policy request mechanism through TLS extensions or some other means.\n\n5.5. Limits on Key Usage\n\n There are cryptographic limits on the amount of plaintext which can\n be safely encrypted under a given set of keys. [AEAD-LIMITS]\n provides an analysis of these limits under the assumption that the\n underlying primitive (AES or ChaCha20) has no weaknesses.\n Implementations SHOULD do a key update as described in Section 4.6.3\n prior to reaching these limits.\n\n For AES-GCM, up to 2^24.5 full-size records (about 24 million) may be\n encrypted on a given connection while keeping a safety margin of\n approximately 2^-57 for Authenticated Encryption (AE) security. For\n ChaCha20/Poly1305, the record sequence number would wrap before the\n safety limit is reached.\n\n6. Alert Protocol\n\n One of the content types supported by the TLS record layer is the\n alert type. Like other messages, alert messages are encrypted as\n specified by the current connection state.\n\n Alert messages convey a description of the alert and a legacy field\n that conveyed the severity of the message in previous versions of\n TLS. Alerts are divided into two classes: closure alerts and error\n alerts. In TLS 1.3, the severity is implicit in the type of alert\n being sent, and the 'level' field can safely be ignored. The\n \"close_notify\" alert is used to indicate orderly closure of one\n direction of the connection. Upon receiving such an alert, the TLS\n implementation SHOULD indicate end-of-data to the application.\n\n Error alerts indicate abortive closure of the connection (see\n Section 6.2). Upon receiving an error alert, the TLS implementation\n SHOULD indicate an error to the application and MUST NOT allow any\n further data to be sent or received on the connection. Servers and\n clients MUST forget the secret values and keys established in failed\n\n\n\n\nRescorla Expires September 21, 2018 [Page 87]\n\f\nInternet-Draft TLS March 2018\n\n\n connections, with the exception of the PSKs associated with session\n tickets, which SHOULD be discarded if possible.\n\n All the alerts listed in Section 6.2 MUST be sent with\n AlertLevel=fatal and MUST be treated as error alerts regardless of\n the AlertLevel in the message. Unknown alert types MUST be treated\n as error alerts.\n\n Note: TLS defines two generic alerts (see Section 6) to use upon\n failure to parse a message. Peers which receive a message which\n cannot be parsed according to the syntax (e.g., have a length\n extending beyond the message boundary or contain an out-of-range\n length) MUST terminate the connection with a \"decode_error\" alert.\n Peers which receive a message which is syntactically correct but\n semantically invalid (e.g., a DHE share of p - 1, or an invalid enum)\n MUST terminate the connection with an \"illegal_parameter\" alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 88]\n\f\nInternet-Draft TLS March 2018\n\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n record_overflow(22),\n handshake_failure(40),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n inappropriate_fallback(86),\n user_canceled(90),\n missing_extension(109),\n unsupported_extension(110),\n unrecognized_name(112),\n bad_certificate_status_response(113),\n unknown_psk_identity(115),\n certificate_required(116),\n no_application_protocol(120),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n6.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack.\n\n close_notify This alert notifies the recipient that the sender will\n not send any more messages on this connection. Any data received\n after a closure alert has been received MUST be ignored.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 89]\n\f\nInternet-Draft TLS March 2018\n\n\n user_canceled This alert notifies the recipient that the sender is\n canceling the handshake for some reason unrelated to a protocol\n failure. If a user cancels an operation after the handshake is\n complete, just closing the connection by sending a \"close_notify\"\n is more appropriate. This alert SHOULD be followed by a\n \"close_notify\". This alert generally has AlertLevel=warning.\n\n Either party MAY initiate a close of its write side of the connection\n by sending a \"close_notify\" alert. Any data received after a closure\n alert has been received MUST be ignored. If a transport-level close\n is received prior to a \"close_notify\", the receiver cannot know that\n all the data that was sent has been received.\n\n Each party MUST send a \"close_notify\" alert before closing its write\n side of the connection, unless it has already sent some error alert.\n This does not have any effect on its read side of the connection.\n Note that this is a change from versions of TLS prior to TLS 1.3 in\n which implementations were required to react to a \"close_notify\" by\n discarding pending writes and sending an immediate \"close_notify\"\n alert of their own. That previous requirement could cause truncation\n in the read side. Both parties need not wait to receive a\n \"close_notify\" alert before closing their read side of the\n connection, though doing so would introduce the possibility of\n truncation.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation MUST receive a \"close_notify\" alert\n before indicating end-of-data to the application-layer. No part of\n this standard should be taken to dictate the manner in which a usage\n profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing the write side of a connection\n reliably delivers pending data before destroying the transport.\n\n6.2. Error Alerts\n\n Error handling in the TLS Handshake Protocol is very simple. When an\n error is detected, the detecting party sends a message to its peer.\n Upon transmission or receipt of a fatal alert message, both parties\n MUST immediately close the connection.\n\n Whenever an implementation encounters a fatal error condition, it\n SHOULD send an appropriate fatal alert and MUST close the connection\n without sending or receiving any additional data. In the rest of\n this specification, when the phrases \"terminate the connection\" and\n \"abort the handshake\" are used without a specific alert it means that\n\n\n\nRescorla Expires September 21, 2018 [Page 90]\n\f\nInternet-Draft TLS March 2018\n\n\n the implementation SHOULD send the alert indicated by the\n descriptions below. The phrases \"terminate the connection with a X\n alert\" and \"abort the handshake with a X alert\" mean that the\n implementation MUST send alert X if it sends any alert. All alerts\n defined in this section below, as well as all unknown alerts, are\n universally considered fatal as of TLS 1.3 (see Section 6). The\n implementation SHOULD provide a way to facilitate logging the sending\n and receiving of alerts.\n\n The following error alerts are defined:\n\n unexpected_message An inappropriate message (e.g., the wrong\n handshake message, premature application data, etc.) was received.\n This alert should never be observed in communication between\n proper implementations.\n\n bad_record_mac This alert is returned if a record is received which\n cannot be deprotected. Because AEAD algorithms combine decryption\n and verification, and also to avoid side channel attacks, this\n alert is used for all deprotection failures. This alert should\n never be observed in communication between proper implementations,\n except when messages were corrupted in the network.\n\n record_overflow A TLSCiphertext record was received that had a\n length more than 2^14 + 256 bytes, or a record decrypted to a\n TLSPlaintext record with more than 2^14 bytes (or some other\n negotiated limit). This alert should never be observed in\n communication between proper implementations, except when messages\n were corrupted in the network.\n\n handshake_failure Receipt of a \"handshake_failure\" alert message\n indicates that the sender was unable to negotiate an acceptable\n set of security parameters given the options available.\n\n bad_certificate A certificate was corrupt, contained signatures that\n did not verify correctly, etc.\n\n unsupported_certificate A certificate was of an unsupported type.\n\n certificate_revoked A certificate was revoked by its signer.\n\n certificate_expired A certificate has expired or is not currently\n valid.\n\n certificate_unknown Some other (unspecified) issue arose in\n processing the certificate, rendering it unacceptable.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 91]\n\f\nInternet-Draft TLS March 2018\n\n\n illegal_parameter A field in the handshake was incorrect or\n inconsistent with other fields. This alert is used for errors\n which conform to the formal protocol syntax but are otherwise\n incorrect.\n\n unknown_ca A valid certificate chain or partial chain was received,\n but the certificate was not accepted because the CA certificate\n could not be located or could not be matched with a known trust\n anchor.\n\n access_denied A valid certificate or PSK was received, but when\n access control was applied, the sender decided not to proceed with\n negotiation.\n\n decode_error A message could not be decoded because some field was\n out of the specified range or the length of the message was\n incorrect. This alert is used for errors where the message does\n not conform to the formal protocol syntax. This alert should\n never be observed in communication between proper implementations,\n except when messages were corrupted in the network.\n\n decrypt_error A handshake (not record-layer) cryptographic operation\n failed, including being unable to correctly verify a signature or\n validate a Finished message or a PSK binder.\n\n protocol_version The protocol version the peer has attempted to\n negotiate is recognized but not supported. (see Appendix D)\n\n insufficient_security Returned instead of \"handshake_failure\" when a\n negotiation has failed specifically because the server requires\n parameters more secure than those supported by the client.\n\n internal_error An internal error unrelated to the peer or the\n correctness of the protocol (such as a memory allocation failure)\n makes it impossible to continue.\n\n inappropriate_fallback Sent by a server in response to an invalid\n connection retry attempt from a client (see [RFC7507]).\n\n missing_extension Sent by endpoints that receive a handshake message\n not containing an extension that is mandatory to send for the\n offered TLS version or other negotiated parameters.\n\n unsupported_extension Sent by endpoints receiving any handshake\n message containing an extension known to be prohibited for\n inclusion in the given handshake message, or including any\n extensions in a ServerHello or Certificate not first offered in\n the corresponding ClientHello.\n\n\n\nRescorla Expires September 21, 2018 [Page 92]\n\f\nInternet-Draft TLS March 2018\n\n\n unrecognized_name Sent by servers when no server exists identified\n by the name provided by the client via the \"server_name\" extension\n (see [RFC6066]).\n\n bad_certificate_status_response Sent by clients when an invalid or\n unacceptable OCSP response is provided by the server via the\n \"status_request\" extension (see [RFC6066]).\n\n unknown_psk_identity Sent by servers when PSK key establishment is\n desired but no acceptable PSK identity is provided by the client.\n Sending this alert is OPTIONAL; servers MAY instead choose to send\n a \"decrypt_error\" alert to merely indicate an invalid PSK\n identity.\n\n certificate_required Sent by servers when a client certificate is\n desired but none was provided by the client.\n\n no_application_protocol Sent by servers when a client\n \"application_layer_protocol_negotiation\" extension advertises only\n protocols that the server does not support (see [RFC7301]).\n\n New Alert values are assigned by IANA as described in Section 11.\n\n7. Cryptographic Computations\n\n The TLS handshake establishes one or more input secrets which are\n combined to create the actual working keying material, as detailed\n below. The key derivation process incorporates both the input\n secrets and the handshake transcript. Note that because the\n handshake transcript includes the random values from the Hello\n messages, any given handshake will have different traffic secrets,\n even if the same input secrets are used, as is the case when the same\n PSK is used for multiple connections.\n\n7.1. Key Schedule\n\n The key derivation process makes use of the HKDF-Extract and HKDF-\n Expand functions as defined for HKDF [RFC5869], as well as the\n functions defined below:\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 93]\n\f\nInternet-Draft TLS March 2018\n\n\n HKDF-Expand-Label(Secret, Label, Context, Length) =\n HKDF-Expand(Secret, HkdfLabel, Length)\n\n Where HkdfLabel is specified as:\n\n struct {\n uint16 length = Length;\n opaque label<7..255> = \"tls13 \" + Label;\n opaque context<0..255> = Context;\n } HkdfLabel;\n\n Derive-Secret(Secret, Label, Messages) =\n HKDF-Expand-Label(Secret, Label,\n Transcript-Hash(Messages), Hash.length)\n\n The Hash function used by Transcript-Hash and HKDF is the cipher\n suite hash algorithm. Hash.length is its output length in bytes.\n Messages is the concatenation of the indicated handshake messages,\n including the handshake message type and length fields, but not\n including record layer headers. Note that in some cases a zero-\n length Context (indicated by \"\") is passed to HKDF-Expand-Label. The\n Labels specified in this document are all ASCII strings, and do not\n include a trailing NUL byte.\n\n Note: with common hash functions, any label longer than 12 characters\n requires an additional iteration of the hash function to compute.\n The labels in this specification have all been chosen to fit within\n this limit.\n\n Keys are derived from two input secrets using the HKDF-Extract and\n Derive-Secret functions. The general pattern for adding a new secret\n is to use HKDF-Extract with the salt being the current secret state\n and the IKM being the new secret to be added. In this version of TLS\n 1.3, the two input secrets are:\n\n - PSK (a pre-shared key established externally or derived from the\n resumption_master_secret value from a previous connection)\n\n - (EC)DHE shared secret (Section 7.4)\n\n This produces a full key derivation schedule shown in the diagram\n below. In this diagram, the following formatting conventions apply:\n\n - HKDF-Extract is drawn as taking the Salt argument from the top and\n the IKM argument from the left, with its output to the bottom and\n the name of the output on the right.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 94]\n\f\nInternet-Draft TLS March 2018\n\n\n - Derive-Secret's Secret argument is indicated by the incoming\n arrow. For instance, the Early Secret is the Secret for\n generating the client_early_traffic_secret.\n\n - \"0\" indicates a string of Hash-lengths bytes set to 0.\n\n 0\n |\n v\n PSK -> HKDF-Extract = Early Secret\n |\n +-----> Derive-Secret(.,\n | \"ext binder\" |\n | \"res binder\",\n | \"\")\n | = binder_key\n |\n +-----> Derive-Secret(., \"c e traffic\",\n | ClientHello)\n | = client_early_traffic_secret\n |\n +-----> Derive-Secret(., \"e exp master\",\n | ClientHello)\n | = early_exporter_master_secret\n v\n Derive-Secret(., \"derived\", \"\")\n |\n v\n(EC)DHE -> HKDF-Extract = Handshake Secret\n |\n +-----> Derive-Secret(., \"c hs traffic\",\n | ClientHello...ServerHello)\n | = client_handshake_traffic_secret\n |\n +-----> Derive-Secret(., \"s hs traffic\",\n | ClientHello...ServerHello)\n | = server_handshake_traffic_secret\n v\n Derive-Secret(., \"derived\", \"\")\n |\n v\n 0 -> HKDF-Extract = Master Secret\n |\n +-----> Derive-Secret(., \"c ap traffic\",\n | ClientHello...server Finished)\n | = client_application_traffic_secret_0\n |\n +-----> Derive-Secret(., \"s ap traffic\",\n\n\n\nRescorla Expires September 21, 2018 [Page 95]\n\f\nInternet-Draft TLS March 2018\n\n\n | ClientHello...server Finished)\n | = server_application_traffic_secret_0\n |\n +-----> Derive-Secret(., \"exp master\",\n | ClientHello...server Finished)\n | = exporter_master_secret\n |\n +-----> Derive-Secret(., \"res master\",\n ClientHello...client Finished)\n = resumption_master_secret\n\n The general pattern here is that the secrets shown down the left side\n of the diagram are just raw entropy without context, whereas the\n secrets down the right side include handshake context and therefore\n can be used to derive working keys without additional context. Note\n that the different calls to Derive-Secret may take different Messages\n arguments, even with the same secret. In a 0-RTT exchange, Derive-\n Secret is called with four distinct transcripts; in a 1-RTT-only\n exchange with three distinct transcripts.\n\n If a given secret is not available, then the 0-value consisting of a\n string of Hash.length bytes set to zeros is used. Note that this\n does not mean skipping rounds, so if PSK is not in use Early Secret\n will still be HKDF-Extract(0, 0). For the computation of the\n binder_secret, the label is \"ext binder\" for external PSKs (those\n provisioned outside of TLS) and \"res binder\" for resumption PSKs\n (those provisioned as the resumption master secret of a previous\n handshake). The different labels prevent the substitution of one\n type of PSK for the other.\n\n There are multiple potential Early Secret values depending on which\n PSK the server ultimately selects. The client will need to compute\n one for each potential PSK; if no PSK is selected, it will then need\n to compute the early secret corresponding to the zero PSK.\n\n Once all the values which are to be derived from a given secret have\n been computed, that secret SHOULD be erased.\n\n7.2. Updating Traffic Secrets\n\n Once the handshake is complete, it is possible for either side to\n update its sending traffic keys using the KeyUpdate handshake message\n defined in Section 4.6.3. The next generation of traffic keys is\n computed by generating client_/server_application_traffic_secret_N+1\n from client_/server_application_traffic_secret_N as described in this\n section then re-deriving the traffic keys as described in\n Section 7.3.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 96]\n\f\nInternet-Draft TLS March 2018\n\n\n The next-generation application_traffic_secret is computed as:\n\n application_traffic_secret_N+1 =\n HKDF-Expand-Label(application_traffic_secret_N,\n \"traffic upd\", \"\", Hash.length)\n\n Once client/server_application_traffic_secret_N+1 and its associated\n traffic keys have been computed, implementations SHOULD delete\n client_/server_application_traffic_secret_N and its associated\n traffic keys.\n\n7.3. Traffic Key Calculation\n\n The traffic keying material is generated from the following input\n values:\n\n - A secret value\n\n - A purpose value indicating the specific value being generated\n\n - The length of the key being generated\n\n The traffic keying material is generated from an input traffic secret\n value using:\n\n [sender]_write_key = HKDF-Expand-Label(Secret, \"key\", \"\", key_length)\n [sender]_write_iv = HKDF-Expand-Label(Secret, \"iv\" , \"\", iv_length)\n\n [sender] denotes the sending side. The Secret value for each record\n type is shown in the table below.\n\n +-------------------+---------------------------------------+\n | Record Type | Secret |\n +-------------------+---------------------------------------+\n | 0-RTT Application | client_early_traffic_secret |\n | | |\n | Handshake | [sender]_handshake_traffic_secret |\n | | |\n | Application Data | [sender]_application_traffic_secret_N |\n +-------------------+---------------------------------------+\n\n All the traffic keying material is recomputed whenever the underlying\n Secret changes (e.g., when changing from the handshake to application\n data keys or upon a key update).\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 97]\n\f\nInternet-Draft TLS March 2018\n\n\n7.4. (EC)DHE Shared Secret Calculation\n\n7.4.1. Finite Field Diffie-Hellman\n\n For finite field groups, a conventional Diffie-Hellman [DH76]\n computation is performed. The negotiated key (Z) is converted to a\n byte string by encoding in big-endian and left padded with zeros up\n to the size of the prime. This byte string is used as the shared\n secret in the key schedule as specified above.\n\n Note that this construction differs from previous versions of TLS\n which remove leading zeros.\n\n7.4.2. Elliptic Curve Diffie-Hellman\n\n For secp256r1, secp384r1 and secp521r1, ECDH calculations (including\n parameter and key generation as well as the shared secret\n calculation) are performed according to [IEEE1363] using the ECKAS-\n DH1 scheme with the identity map as key derivation function (KDF), so\n that the shared secret is the x-coordinate of the ECDH shared secret\n elliptic curve point represented as an octet string. Note that this\n octet string (Z in IEEE 1363 terminology) as output by FE2OSP, the\n Field Element to Octet String Conversion Primitive, has constant\n length for any given field; leading zeros found in this octet string\n MUST NOT be truncated.\n\n (Note that this use of the identity KDF is a technicality. The\n complete picture is that ECDH is employed with a non-trivial KDF\n because TLS does not directly use this secret for anything other than\n for computing other secrets.)\n\n ECDH functions are used as follows:\n\n - The public key to put into the KeyShareEntry.key_exchange\n structure is the result of applying the ECDH scalar multiplication\n function to the secret key of appropriate length (into scalar\n input) and the standard public basepoint (into u-coordinate point\n input).\n\n - The ECDH shared secret is the result of applying the ECDH scalar\n multiplication function to the secret key (into scalar input) and\n the peer's public key (into u-coordinate point input). The output\n is used raw, with no processing.\n\n For X25519 and X448, implementations SHOULD use the approach\n specified in [RFC7748] to calculate the Diffie-Hellman shared secret.\n Implementations MUST check whether the computed Diffie-Hellman shared\n secret is the all-zero value and abort if so, as described in\n\n\n\nRescorla Expires September 21, 2018 [Page 98]\n\f\nInternet-Draft TLS March 2018\n\n\n Section 6 of [RFC7748]. If implementors use an alternative\n implementation of these elliptic curves, they SHOULD perform the\n additional checks specified in Section 7 of [RFC7748].\n\n7.5. Exporters\n\n [RFC5705] defines keying material exporters for TLS in terms of the\n TLS pseudorandom function (PRF). This document replaces the PRF with\n HKDF, thus requiring a new construction. The exporter interface\n remains the same.\n\n The exporter value is computed as:\n\n TLS-Exporter(label, context_value, key_length) =\n HKDF-Expand-Label(Derive-Secret(Secret, label, \"\"),\n \"exporter\", Hash(context_value), key_length)\n\n Where Secret is either the early_exporter_master_secret or the\n exporter_master_secret. Implementations MUST use the\n exporter_master_secret unless explicitly specified by the\n application. The early_exporter_master_secret is defined for use in\n settings where an exporter is needed for 0-RTT data. A separate\n interface for the early exporter is RECOMMENDED; this avoids the\n exporter user accidentally using an early exporter when a regular one\n is desired or vice versa.\n\n If no context is provided, the context_value is zero-length.\n Consequently, providing no context computes the same value as\n providing an empty context. This is a change from previous versions\n of TLS where an empty context produced a different output to an\n absent context. As of this document's publication, no allocated\n exporter label is used both with and without a context. Future\n specifications MUST NOT define a use of exporters that permit both an\n empty context and no context with the same label. New uses of\n exporters SHOULD provide a context in all exporter computations,\n though the value could be empty.\n\n Requirements for the format of exporter labels are defined in section\n 4 of [RFC5705].\n\n8. 0-RTT and Anti-Replay\n\n As noted in Section 2.3 and Appendix E.5, TLS does not provide\n inherent replay protections for 0-RTT data. There are two potential\n threats to be concerned with:\n\n - Network attackers who mount a replay attack by simply duplicating\n a flight of 0-RTT data.\n\n\n\nRescorla Expires September 21, 2018 [Page 99]\n\f\nInternet-Draft TLS March 2018\n\n\n - Network attackers who take advantage of client retry behavior to\n arrange for the server to receive multiple copies of an\n application message. This threat already exists to some extent\n because clients that value robustness respond to network errors by\n attempting to retry requests. However, 0-RTT adds an additional\n dimension for any server system which does not maintain globally\n consistent server state. Specifically, if a server system has\n multiple zones where tickets from zone A will not be accepted in\n zone B, then an attacker can duplicate a ClientHello and early\n data intended for A to both A and B. At A, the data will be\n accepted in 0-RTT, but at B the server will reject 0-RTT data and\n instead force a full handshake. If the attacker blocks the\n ServerHello from A, then the client will complete the handshake\n with B and probably retry the request, leading to duplication on\n the server system as a whole.\n\n The first class of attack can be prevented by sharing state to\n guarantee that the 0-RTT data is accepted at most once. Servers\n SHOULD provide that level of replay safety, by implementing one of\n the methods described in this section or by equivalent means. It is\n understood, however, that due to operational concerns not all\n deployments will maintain state at that level. Therefore, in normal\n operation, clients will not know which, if any, of these mechanisms\n servers actually implement and hence MUST only send early data which\n they deem safe to be replayed.\n\n In addition to the direct effects of replays, there is a class of\n attacks where even operations normally considered idempotent could be\n exploited by a large number of replays (timing attacks, resource\n limit exhaustion and others described in Appendix E.5). Those can be\n mitigated by ensuring that every 0-RTT payload can be replayed only a\n limited number of times. The server MUST ensure that any instance of\n it (be it a machine, a thread or any other entity within the relevant\n serving infrastructure) would accept 0-RTT for the same 0-RTT\n handshake at most once; this limits the number of replays to the\n number of server instances in the deployment. Such a guarantee can\n be accomplished by locally recording data from recently-received\n ClientHellos and rejecting repeats, or by any other method that\n provides the same or a stronger guarantee. The \"at most once per\n server instance\" guarantee is a minimum requirement; servers SHOULD\n limit 0-RTT replays further when feasible.\n\n The second class of attack cannot be prevented at the TLS layer and\n MUST be dealt with by any application. Note that any application\n whose clients implement any kind of retry behavior already needs to\n implement some sort of anti-replay defense.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 100]\n\f\nInternet-Draft TLS March 2018\n\n\n8.1. Single-Use Tickets\n\n The simplest form of anti-replay defense is for the server to only\n allow each session ticket to be used once. For instance, the server\n can maintain a database of all outstanding valid tickets; deleting\n each ticket from the database as it is used. If an unknown ticket is\n provided, the server would then fall back to a full handshake.\n\n If the tickets are not self-contained but rather are database keys,\n and the corresponding PSKs are deleted upon use, then connections\n established using PSKs enjoy forward secrecy. This improves security\n for all 0-RTT data and PSK usage when PSK is used without (EC)DHE.\n\n Because this mechanism requires sharing the session database between\n server nodes in environments with multiple distributed servers, it\n may be hard to achieve high rates of successful PSK 0-RTT connections\n when compared to self-encrypted tickets. Unlike session databases,\n session tickets can successfully do PSK-based session establishment\n even without consistent storage, though when 0-RTT is allowed they\n still require consistent storage for anti-replay of 0-RTT data, as\n detailed in the following section.\n\n8.2. Client Hello Recording\n\n An alternative form of anti-replay is to record a unique value\n derived from the ClientHello (generally either the random value or\n the PSK binder) and reject duplicates. Recording all ClientHellos\n causes state to grow without bound, but a server can instead record\n ClientHellos within a given time window and use the\n \"obfuscated_ticket_age\" to ensure that tickets aren't reused outside\n that window.\n\n In order to implement this, when a ClientHello is received, the\n server first verifies the PSK binder as described Section 4.2.11. It\n then computes the expected_arrival_time as described in the next\n section and rejects 0-RTT if it is outside the recording window,\n falling back to the 1-RTT handshake.\n\n If the expected arrival time is in the window, then the server checks\n to see if it has recorded a matching ClientHello. If one is found,\n it either aborts the handshake with an \"illegal_parameter\" alert or\n accepts the PSK but reject 0-RTT. If no matching ClientHello is\n found, then it accepts 0-RTT and then stores the ClientHello for as\n long as the expected_arrival_time is inside the window. Servers MAY\n also implement data stores with false positives, such as Bloom\n filters, in which case they MUST respond to apparent replay by\n rejecting 0-RTT but MUST NOT abort the handshake.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 101]\n\f\nInternet-Draft TLS March 2018\n\n\n The server MUST derive the storage key only from validated sections\n of the ClientHello. If the ClientHello contains multiple PSK\n identities, then an attacker can create multiple ClientHellos with\n different binder values for the less-preferred identity on the\n assumption that the server will not verify it, as recommended by\n Section 4.2.11. I.e., if the client sends PSKs A and B but the\n server prefers A, then the attacker can change the binder for B\n without affecting the binder for A. If the binder for B is part of\n the storage key, then this ClientHello will not appear as a\n duplicate, which will cause the ClientHello to be accepted, and may\n cause side effects such as replay cache pollution, although any 0-RTT\n data will not be decryptable because it will use different keys. If\n the validated binder or the ClientHello.random are used as the\n storage key, then this attack is not possible.\n\n Because this mechanism does not require storing all outstanding\n tickets, it may be easier to implement in distributed systems with\n high rates of resumption and 0-RTT, at the cost of potentially weaker\n anti-replay defense because of the difficulty of reliably storing and\n retrieving the received ClientHello messages. In many such systems,\n it is impractical to have globally consistent storage of all the\n received ClientHellos. In this case, the best anti-replay protection\n is provided by having a single storage zone be authoritative for a\n given ticket and refusing 0-RTT for that ticket in any other zone.\n This approach prevents simple replay by the attacker because only one\n zone will accept 0-RTT data. A weaker design is to implement\n separate storage for each zone but allow 0-RTT in any zone. This\n approach limits the number of replays to once per zone. Application\n message duplication of course remains possible with either design.\n\n When implementations are freshly started, they SHOULD reject 0-RTT as\n long as any portion of their recording window overlaps the startup\n time. Otherwise, they run the risk of accepting replays which were\n originally sent during that period.\n\n Note: If the client's clock is running much faster than the server's\n then a ClientHello may be received that is outside the window in the\n future, in which case it might be accepted for 1-RTT, causing a\n client retry, and then acceptable later for 0-RTT. This is another\n variant of the second form of attack described above.\n\n8.3. Freshness Checks\n\n Because the ClientHello indicates the time at which the client sent\n it, it is possible to efficiently determine whether a ClientHello was\n likely sent reasonably recently and only accept 0-RTT for such a\n ClientHello, otherwise falling back to a 1-RTT handshake. This is\n necessary for the ClientHello storage mechanism described in\n\n\n\nRescorla Expires September 21, 2018 [Page 102]\n\f\nInternet-Draft TLS March 2018\n\n\n Section 8.2 because otherwise the server needs to store an unlimited\n number of ClientHellos and is a useful optimization for self-\n contained single-use tickets because it allows efficient rejection of\n ClientHellos which cannot be used for 0-RTT.\n\n In order to implement this mechanism, a server needs to store the\n time that the server generated the session ticket, offset by an\n estimate of the round trip time between client and server. I.e.,\n\n adjusted_creation_time = creation_time + estimated_RTT\n\n This value can be encoded in the ticket, thus avoiding the need to\n keep state for each outstanding ticket. The server can determine the\n client's view of the age of the ticket by subtracting the ticket's\n \"ticket_age_add value\" from the \"obfuscated_ticket_age\" parameter in\n the client's \"pre_shared_key\" extension. The server can determine\n the \"expected arrival time\" of the ClientHello as:\n\n expected_arrival_time = adjusted_creation_time + clients_ticket_age\n\n When a new ClientHello is received, the expected_arrival_time is then\n compared against the current server wall clock time and if they\n differ by more than a certain amount, 0-RTT is rejected, though the\n 1-RTT handshake can be allowed to complete.\n\n There are several potential sources of error that might cause\n mismatches between the expected arrival time and the measured time.\n Variations in client and server clock rates are likely to be minimal,\n though potentially the absolute times may be off by large values.\n Network propagation delays are the most likely causes of a mismatch\n in legitimate values for elapsed time. Both the NewSessionTicket and\n ClientHello messages might be retransmitted and therefore delayed,\n which might be hidden by TCP. For clients on the Internet, this\n implies windows on the order of ten seconds to account for errors in\n clocks and variations in measurements; other deployment scenarios may\n have different needs. Clock skew distributions are not symmetric, so\n the optimal tradeoff may involve an asymmetric range of permissible\n mismatch values.\n\n Note that freshness checking alone is not sufficient to prevent\n replays because it does not detect them during the error window,\n which, depending on bandwidth and system capacity could include\n billions of replays in real-world settings. In addition, this\n freshness checking is only done at the time the ClientHello is\n received, and not when later early application data records are\n received. After early data is accepted, records may continue to be\n streamed to the server over a longer time period.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 103]\n\f\nInternet-Draft TLS March 2018\n\n\n9. Compliance Requirements\n\n9.1. Mandatory-to-Implement Cipher Suites\n\n In the absence of an application profile standard specifying\n otherwise, a TLS-compliant application MUST implement the\n TLS_AES_128_GCM_SHA256 [GCM] cipher suite and SHOULD implement the\n TLS_AES_256_GCM_SHA384 [GCM] and TLS_CHACHA20_POLY1305_SHA256\n [RFC7539] cipher suites. (see Appendix B.4)\n\n A TLS-compliant application MUST support digital signatures with\n rsa_pkcs1_sha256 (for certificates), rsa_pss_rsae_sha256 (for\n CertificateVerify and certificates), and ecdsa_secp256r1_sha256. A\n TLS-compliant application MUST support key exchange with secp256r1\n (NIST P-256) and SHOULD support key exchange with X25519 [RFC7748].\n\n9.2. Mandatory-to-Implement Extensions\n\n In the absence of an application profile standard specifying\n otherwise, a TLS-compliant application MUST implement the following\n TLS extensions:\n\n - Supported Versions (\"supported_versions\"; Section 4.2.1)\n\n - Cookie (\"cookie\"; Section 4.2.2)\n\n - Signature Algorithms (\"signature_algorithms\"; Section 4.2.3)\n\n - Signature Algorithms Certificate (\"signature_algorithms_cert\";\n Section 4.2.3)\n\n - Negotiated Groups (\"supported_groups\"; Section 4.2.7)\n\n - Key Share (\"key_share\"; Section 4.2.8)\n\n - Server Name Indication (\"server_name\"; Section 3 of [RFC6066])\n\n All implementations MUST send and use these extensions when offering\n applicable features:\n\n - \"supported_versions\" is REQUIRED for all ClientHello, ServerHello\n and HelloRetryRequest messages.\n\n - \"signature_algorithms\" is REQUIRED for certificate authentication.\n\n - \"supported_groups\" is REQUIRED for ClientHello messages using DHE\n or ECDHE key exchange.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 104]\n\f\nInternet-Draft TLS March 2018\n\n\n - \"key_share\" is REQUIRED for DHE or ECDHE key exchange.\n\n - \"pre_shared_key\" is REQUIRED for PSK key agreement.\n\n - \"psk_key_exchange_modes\" is REQUIRED for PSK key agreement.\n\n A client is considered to be attempting to negotiate using this\n specification if the ClientHello contains a \"supported_versions\"\n extension with 0x0304 contained in its body. Such a ClientHello\n message MUST meet the following requirements:\n\n - If not containing a \"pre_shared_key\" extension, it MUST contain\n both a \"signature_algorithms\" extension and a \"supported_groups\"\n extension.\n\n - If containing a \"supported_groups\" extension, it MUST also contain\n a \"key_share\" extension, and vice versa. An empty\n KeyShare.client_shares vector is permitted.\n\n Servers receiving a ClientHello which does not conform to these\n requirements MUST abort the handshake with a \"missing_extension\"\n alert.\n\n Additionally, all implementations MUST support use of the\n \"server_name\" extension with applications capable of using it.\n Servers MAY require clients to send a valid \"server_name\" extension.\n Servers requiring this extension SHOULD respond to a ClientHello\n lacking a \"server_name\" extension by terminating the connection with\n a \"missing_extension\" alert.\n\n9.3. Protocol Invariants\n\n This section describes invariants that TLS endpoints and middleboxes\n MUST follow. It also applies to earlier versions of TLS.\n\n TLS is designed to be securely and compatibly extensible. Newer\n clients or servers, when communicating with newer peers, should\n negotiate the most preferred common parameters. The TLS handshake\n provides downgrade protection: Middleboxes passing traffic between a\n newer client and newer server without terminating TLS should be\n unable to influence the handshake (see Appendix E.1). At the same\n time, deployments update at different rates, so a newer client or\n server MAY continue to support older parameters, which would allow it\n to interoperate with older endpoints.\n\n For this to work, implementations MUST correctly handle extensible\n fields:\n\n\n\n\nRescorla Expires September 21, 2018 [Page 105]\n\f\nInternet-Draft TLS March 2018\n\n\n - A client sending a ClientHello MUST support all parameters\n advertised in it. Otherwise, the server may fail to interoperate\n by selecting one of those parameters.\n\n - A server receiving a ClientHello MUST correctly ignore all\n unrecognized cipher suites, extensions, and other parameters.\n Otherwise, it may fail to interoperate with newer clients. In TLS\n 1.3, a client receiving a CertificateRequest or NewSessionTicket\n MUST also ignore all unrecognized extensions.\n\n - A middlebox which terminates a TLS connection MUST behave as a\n compliant TLS server (to the original client), including having a\n certificate which the client is willing to accept, and as a\n compliant TLS client (to the original server), including verifying\n the original server's certificate. In particular, it MUST\n generate its own ClientHello containing only parameters it\n understands, and it MUST generate a fresh ServerHello random\n value, rather than forwarding the endpoint's value.\n\n Note that TLS's protocol requirements and security analysis only\n apply to the two connections separately. Safely deploying a TLS\n terminator requires additional security considerations which are\n beyond the scope of this document.\n\n - An middlebox which forwards ClientHello parameters it does not\n understand MUST NOT process any messages beyond that ClientHello.\n It MUST forward all subsequent traffic unmodified. Otherwise, it\n may fail to interoperate with newer clients and servers.\n\n Forwarded ClientHellos may contain advertisements for features not\n supported by the middlebox, so the response may include future TLS\n additions the middlebox does not recognize. These additions MAY\n change any message beyond the ClientHello arbitrarily. In\n particular, the values sent in the ServerHello might change, the\n ServerHello format might change, and the TLSCiphertext format\n might change.\n\n The design of TLS 1.3 was constrained by widely-deployed non-\n compliant TLS middleboxes (see Appendix D.4), however it does not\n relax the invariants. Those middleboxes continue to be non-\n compliant.\n\n10. Security Considerations\n\n Security issues are discussed throughout this memo, especially in\n Appendix C, Appendix D, and Appendix E.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 106]\n\f\nInternet-Draft TLS March 2018\n\n\n11. IANA Considerations\n\n This document uses several registries that were originally created in\n [RFC4346]. IANA [SHALL update/has updated] these to reference this\n document. The registries and their allocation policies are below:\n\n - TLS Cipher Suite Registry: values with the first byte in the range\n 0-254 (decimal) are assigned via Specification Required [RFC8126].\n Values with the first byte 255 (decimal) are reserved for Private\n Use [RFC8126].\n\n IANA [SHALL add/has added] the cipher suites listed in\n Appendix B.4 to the registry. The \"Value\" and \"Description\"\n columns are taken from the table. The \"DTLS-OK\" and \"Recommended\"\n columns are both marked as \"Yes\" for each new cipher suite.\n [[This assumes [I-D.ietf-tls-iana-registry-updates] has been\n applied.]]\n\n - TLS ContentType Registry: Future values are allocated via\n Standards Action [RFC8126].\n\n - TLS Alert Registry: Future values are allocated via Standards\n Action [RFC8126]. IANA [SHALL update/has updated] this registry\n to include values for \"missing_extension\" and\n \"certificate_required\". The \"DTLS-OK\" column is marked as \"Yes\"\n for each new alert.\n\n - TLS HandshakeType Registry: Future values are allocated via\n Standards Action [RFC8126]. IANA [SHALL update/has updated] this\n registry to rename item 4 from \"NewSessionTicket\" to\n \"new_session_ticket\" and to add the\n \"hello_retry_request_RESERVED\", \"encrypted_extensions\",\n \"end_of_early_data\", \"key_update\", and \"message_hash\" values. The\n \"DTLS-OK\" are marked as \"Yes\" for each of these additions.\n\n This document also uses the TLS ExtensionType Registry originally\n created in [RFC4366]. IANA has updated it to reference this\n document. Changes to the registry follow:\n\n - IANA [SHALL update/has updated] the registration policy as\n follows:\n\n Values with the first byte in the range 0-254 (decimal) are\n assigned via Specification Required [RFC8126]. Values with the\n first byte 255 (decimal) are reserved for Private Use [RFC8126].\n\n - IANA [SHALL update/has updated] this registry to include the\n \"key_share\", \"pre_shared_key\", \"psk_key_exchange_modes\",\n\n\n\nRescorla Expires September 21, 2018 [Page 107]\n\f\nInternet-Draft TLS March 2018\n\n\n \"early_data\", \"cookie\", \"supported_versions\",\n \"certificate_authorities\", \"oid_filters\", \"post_handshake_auth\",\n and \"signature_algorithms_cert\", extensions with the values\n defined in this document and the Recommended value of \"Yes\".\n\n - IANA [SHALL update/has updated] this registry to include a \"TLS\n 1.3\" column which lists the messages in which the extension may\n appear. This column [SHALL be/has been] initially populated from\n the table in Section 4.2 with any extension not listed there\n marked as \"-\" to indicate that it is not used by TLS 1.3.\n\n In addition, this document defines two new registries to be\n maintained by IANA:\n\n - TLS SignatureScheme Registry: Values with the first byte in the\n range 0-253 (decimal) are assigned via Specification Required\n [RFC8126]. Values with the first byte 254 or 255 (decimal) are\n reserved for Private Use [RFC8126]. Values with the first byte in\n the range 0-6 or with the second byte in the range 0-3 that are\n not currently allocated are reserved for backwards compatibility.\n This registry SHALL have a \"Recommended\" column. The registry\n [shall be/ has been] initially populated with the values described\n in Section 4.2.3. The following values SHALL be marked as\n \"Recommended\": ecdsa_secp256r1_sha256, ecdsa_secp384r1_sha384,\n rsa_pss_rsae_sha256, rsa_pss_rsae_sha384, rsa_pss_rsae_sha512,\n rsa_pss_pss_sha256, rsa_pss_pss_sha384, rsa_pss_pss_sha512, and\n ed25519.\n\n - TLS PskKeyExchangeMode Registry: Values in the range 0-253\n (decimal) are assigned via Specification Required [RFC8126].\n Values with the first byte 254 or 255 (decimal) are reserved for\n Private Use [RFC8126]. This registry SHALL have a \"Recommended\"\n column. The registry [shall be/ has been] initially populated\n psk_ke (0) and psk_dhe_ke (1). Both SHALL be marked as\n \"Recommended\".\n\n12. References\n\n12.1. Normative References\n\n [DH] Diffie, W. and M. Hellman, \"New Directions in\n Cryptography\", IEEE Transactions on Information Theory,\n V.IT-22 n.6 , June 1977.\n\n [DH76] Diffie, W. and M. Hellman, \"New directions in\n cryptography\", IEEE Transactions on Information\n Theory Vol. 22, pp. 644-654, DOI 10.1109/tit.1976.1055638,\n November 1976.\n\n\n\nRescorla Expires September 21, 2018 [Page 108]\n\f\nInternet-Draft TLS March 2018\n\n\n [GCM] Dworkin, M., \"Recommendation for Block Cipher Modes of\n Operation: Galois/Counter Mode (GCM) and GMAC\",\n NIST Special Publication 800-38D, November 2007.\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC: Keyed-\n Hashing for Message Authentication\", RFC 2104,\n DOI 10.17487/RFC2104, February 1997, .\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119,\n DOI 10.17487/RFC2119, March 1997, .\n\n [RFC5116] McGrew, D., \"An Interface and Algorithms for Authenticated\n Encryption\", RFC 5116, DOI 10.17487/RFC5116, January 2008,\n .\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, DOI 10.17487/RFC5280, May 2008,\n .\n\n [RFC5705] Rescorla, E., \"Keying Material Exporters for Transport\n Layer Security (TLS)\", RFC 5705, DOI 10.17487/RFC5705,\n March 2010, .\n\n [RFC5756] Turner, S., Brown, D., Yiu, K., Housley, R., and T. Polk,\n \"Updates for RSAES-OAEP and RSASSA-PSS Algorithm\n Parameters\", RFC 5756, DOI 10.17487/RFC5756, January 2010,\n .\n\n [RFC5869] Krawczyk, H. and P. Eronen, \"HMAC-based Extract-and-Expand\n Key Derivation Function (HKDF)\", RFC 5869,\n DOI 10.17487/RFC5869, May 2010, .\n\n [RFC6066] Eastlake 3rd, D., \"Transport Layer Security (TLS)\n Extensions: Extension Definitions\", RFC 6066,\n DOI 10.17487/RFC6066, January 2011, .\n\n [RFC6655] McGrew, D. and D. Bailey, \"AES-CCM Cipher Suites for\n Transport Layer Security (TLS)\", RFC 6655,\n DOI 10.17487/RFC6655, July 2012, .\n\n\n\n\nRescorla Expires September 21, 2018 [Page 109]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, DOI 10.17487/RFC6960, June 2013,\n .\n\n [RFC6961] Pettersen, Y., \"The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\", RFC 6961,\n DOI 10.17487/RFC6961, June 2013, .\n\n [RFC6962] Laurie, B., Langley, A., and E. Kasper, \"Certificate\n Transparency\", RFC 6962, DOI 10.17487/RFC6962, June 2013,\n .\n\n [RFC6979] Pornin, T., \"Deterministic Usage of the Digital Signature\n Algorithm (DSA) and Elliptic Curve Digital Signature\n Algorithm (ECDSA)\", RFC 6979, DOI 10.17487/RFC6979, August\n 2013, .\n\n [RFC7301] Friedl, S., Popov, A., Langley, A., and E. Stephan,\n \"Transport Layer Security (TLS) Application-Layer Protocol\n Negotiation Extension\", RFC 7301, DOI 10.17487/RFC7301,\n July 2014, .\n\n [RFC7507] Moeller, B. and A. Langley, \"TLS Fallback Signaling Cipher\n Suite Value (SCSV) for Preventing Protocol Downgrade\n Attacks\", RFC 7507, DOI 10.17487/RFC7507, April 2015,\n .\n\n [RFC7539] Nir, Y. and A. Langley, \"ChaCha20 and Poly1305 for IETF\n Protocols\", RFC 7539, DOI 10.17487/RFC7539, May 2015,\n .\n\n [RFC7748] Langley, A., Hamburg, M., and S. Turner, \"Elliptic Curves\n for Security\", RFC 7748, DOI 10.17487/RFC7748, January\n 2016, .\n\n [RFC7919] Gillmor, D., \"Negotiated Finite Field Diffie-Hellman\n Ephemeral Parameters for Transport Layer Security (TLS)\",\n RFC 7919, DOI 10.17487/RFC7919, August 2016,\n .\n\n [RFC8017] Moriarty, K., Ed., Kaliski, B., Jonsson, J., and A. Rusch,\n \"PKCS #1: RSA Cryptography Specifications Version 2.2\",\n RFC 8017, DOI 10.17487/RFC8017, November 2016,\n .\n\n\n\n\nRescorla Expires September 21, 2018 [Page 110]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC8032] Josefsson, S. and I. Liusvaara, \"Edwards-Curve Digital\n Signature Algorithm (EdDSA)\", RFC 8032,\n DOI 10.17487/RFC8032, January 2017, .\n\n [RFC8126] Cotton, M., Leiba, B., and T. Narten, \"Guidelines for\n Writing an IANA Considerations Section in RFCs\", BCP 26,\n RFC 8126, DOI 10.17487/RFC8126, June 2017,\n .\n\n [RFC8174] Leiba, B., \"Ambiguity of Uppercase vs Lowercase in RFC\n 2119 Key Words\", BCP 14, RFC 8174, DOI 10.17487/RFC8174,\n May 2017, .\n\n [SHS] Dang, Q., \"Secure Hash Standard\", National Institute of\n Standards and Technology report,\n DOI 10.6028/nist.fips.180-4, July 2015.\n\n [X690] ITU-T, \"Information technology - ASN.1 encoding Rules:\n Specification of Basic Encoding Rules (BER), Canonical\n Encoding Rules (CER) and Distinguished Encoding Rules\n (DER)\", ISO/IEC 8825-1:2002, 2002.\n\n [X962] ANSI, \"Public Key Cryptography For The Financial Services\n Industry: The Elliptic Curve Digital Signature Algorithm\n (ECDSA)\", ANSI X9.62, 1998.\n\n12.2. Informative References\n\n [AEAD-LIMITS]\n Luykx, A. and K. Paterson, \"Limits on Authenticated\n Encryption Use in TLS\", 2016,\n .\n\n [Anon18] Anonymous, A., \"Secure Channels for Multiplexed Data\n Streams: Analyzing the TLS 1.3 Record Layer Without\n Elision\", In submission to CRYPTO 2018. RFC EDITOR: PLEASE\n UPDATE THIS REFERENCE AFTER FINAL NOTIFICATION\n (2018-4-29). , 2018.\n\n [BBFKZG16]\n Bhargavan, K., Brzuska, C., Fournet, C., Kohlweiss, M.,\n Zanella-Beguelin, S., and M. Green, \"Downgrade Resilience\n in Key-Exchange Protocols\", Proceedings of IEEE Symposium\n on Security and Privacy (Oakland) 2016 , 2016.\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 111]\n\f\nInternet-Draft TLS March 2018\n\n\n [BBK17] Bhargavan, K., Blanchet, B., and N. Kobeissi, \"Verified\n Models and Reference Implementations for the TLS 1.3\n Standard Candidate\", Proceedings of IEEE Symposium on\n Security and Privacy (Oakland) 2017 , 2017.\n\n [BDFKPPRSZZ16]\n Bhargavan, K., Delignat-Lavaud, A., Fournet, C.,\n Kohlweiss, M., Pan, J., Protzenko, J., Rastogi, A., Swamy,\n N., Zanella-Beguelin, S., and J. Zinzindohoue,\n \"Implementing and Proving the TLS 1.3 Record Layer\",\n Proceedings of IEEE Symposium on Security and Privacy\n (Oakland) 2017 , December 2016,\n .\n\n [Ben17a] Benjamin, D., \"Presentation before the TLS WG at IETF\n 100\", 2017,\n .\n\n [Ben17b] Benjamin, D., \"Additional TLS 1.3 results from Chrome\",\n 2017, .\n\n [BMMT15] Badertscher, C., Matt, C., Maurer, U., and B. Tackmann,\n \"Augmented Secure Channels and the Goal of the TLS 1.3\n Record Layer\", ProvSec 2015 , September 2015,\n .\n\n [BT16] Bellare, M. and B. Tackmann, \"The Multi-User Security of\n Authenticated Encryption: AES-GCM in TLS 1.3\", Proceedings\n of CRYPTO 2016 , 2016, .\n\n [CCG16] Cohn-Gordon, K., Cremers, C., and L. Garratt, \"On Post-\n Compromise Security\", IEEE Computer Security Foundations\n Symposium , 2015.\n\n [CHECKOWAY]\n Checkoway, S., Shacham, H., Maskiewicz, J., Garman, C.,\n Fried, J., Cohney, S., Green, M., Heninger, N., Weinmann,\n R., and E. Rescorla, \"A Systematic Analysis of the Juniper\n Dual EC Incident\", Proceedings of the 2016 ACM SIGSAC\n Conference on Computer and Communications Security\n - CCS'16, DOI 10.1145/2976749.2978395, 2016.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 112]\n\f\nInternet-Draft TLS March 2018\n\n\n [CHHSV17] Cremers, C., Horvat, M., Hoyland, J., van der Merwe, T.,\n and S. Scott, \"Awkward Handshake: Possible mismatch of\n client/server view on client authentication in post-\n handshake mode in Revision 18\", 2017,\n .\n\n [CHSV16] Cremers, C., Horvat, M., Scott, S., and T. van der Merwe,\n \"Automated Analysis and Verification of TLS 1.3: 0-RTT,\n Resumption and Delayed Authentication\", Proceedings of\n IEEE Symposium on Security and Privacy (Oakland) 2016 ,\n 2016, .\n\n [CK01] Canetti, R. and H. Krawczyk, \"Analysis of Key-Exchange\n Protocols and Their Use for Building Secure Channels\",\n Proceedings of Eurocrypt 2001 , 2001.\n\n [CLINIC] Miller, B., Huang, L., Joseph, A., and J. Tygar, \"I Know\n Why You Went to the Clinic: Risks and Realization of HTTPS\n Traffic Analysis\", Privacy Enhancing Technologies pp.\n 143-163, DOI 10.1007/978-3-319-08506-7_8, 2014.\n\n [DFGS15] Dowling, B., Fischlin, M., Guenther, F., and D. Stebila,\n \"A Cryptographic Analysis of the TLS 1.3 draft-10 Full and\n Pre-shared Key Handshake Protocol\", Proceedings of ACM CCS\n 2015 , 2015, .\n\n [DFGS16] Dowling, B., Fischlin, M., Guenther, F., and D. Stebila,\n \"A Cryptographic Analysis of the TLS 1.3 draft-10 Full and\n Pre-shared Key Handshake Protocol\", TRON 2016 , 2016,\n .\n\n [DOW92] Diffie, W., van Oorschot, P., and M. Wiener,\n \"Authentication and authenticated key exchanges\", Designs,\n Codes and Cryptography , 1992.\n\n [DSS] National Institute of Standards and Technology, U.S.\n Department of Commerce, \"Digital Signature Standard,\n version 4\", NIST FIPS PUB 186-4, 2013.\n\n [ECDSA] American National Standards Institute, \"Public Key\n Cryptography for the Financial Services Industry: The\n Elliptic Curve Digital Signature Algorithm (ECDSA)\",\n ANSI ANS X9.62-2005, November 2005.\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 113]\n\f\nInternet-Draft TLS March 2018\n\n\n [FG17] Fischlin, M. and F. Guenther, \"Replay Attacks on Zero\n Round-Trip Time: The Case of the TLS 1.3 Handshake\n Candidates\", Proceedings of Euro S\"P 2017 , 2017,\n .\n\n [FGSW16] Fischlin, M., Guenther, F., Schmidt, B., and B. Warinschi,\n \"Key Confirmation in Key Exchange: A Formal Treatment and\n Implications for TLS 1.3\", Proceedings of IEEE Symposium\n on Security and Privacy (Oakland) 2016 , 2016,\n .\n\n [FW15] Florian Weimer, ., \"Factoring RSA Keys With TLS Perfect\n Forward Secrecy\", September 2015.\n\n [HCJ16] Husak, M., Čermak, M., Jirsik, T., and P.\n Čeleda, \"HTTPS traffic analysis and client\n identification using passive SSL/TLS fingerprinting\",\n EURASIP Journal on Information Security Vol. 2016,\n DOI 10.1186/s13635-016-0030-7, February 2016.\n\n [HGFS15] Hlauschek, C., Gruber, M., Fankhauser, F., and C. Schanes,\n \"Prying Open Pandora's Box: KCI Attacks against TLS\",\n Proceedings of USENIX Workshop on Offensive Technologies ,\n 2015.\n\n [I-D.ietf-tls-iana-registry-updates]\n Salowey, J. and S. Turner, \"IANA Registry Updates for TLS\n and DTLS\", draft-ietf-tls-iana-registry-updates-04 (work\n in progress), February 2018.\n\n [I-D.ietf-tls-tls13-vectors]\n Thomson, M., \"Example Handshake Traces for TLS 1.3\",\n draft-ietf-tls-tls13-vectors-03 (work in progress),\n December 2017.\n\n [IEEE1363]\n IEEE, \"Standard Specifications for Public Key\n Cryptography\", IEEE 1363 , 2000.\n\n [JSS15] Jager, T., Schwenk, J., and J. Somorovsky, \"On the\n Security of TLS 1.3 and QUIC Against Weaknesses in PKCS#1\n v1.5 Encryption\", Proceedings of ACM CCS 2015 , 2015,\n .\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 114]\n\f\nInternet-Draft TLS March 2018\n\n\n [KEYAGREEMENT]\n Barker, E., Chen, L., Roginsky, A., and M. Smid,\n \"Recommendation for Pair-Wise Key Establishment Schemes\n Using Discrete Logarithm Cryptography\", National Institute\n of Standards and Technology report,\n DOI 10.6028/nist.sp.800-56ar2, May 2013.\n\n [Kraw10] Krawczyk, H., \"Cryptographic Extraction and Key\n Derivation: The HKDF Scheme\", Proceedings of CRYPTO 2010 ,\n 2010, .\n\n [Kraw16] Krawczyk, H., \"A Unilateral-to-Mutual Authentication\n Compiler for Key Exchange (with Applications to Client\n Authentication in TLS 1.3\", Proceedings of ACM CCS 2016 ,\n 2016, .\n\n [KW16] Krawczyk, H. and H. Wee, \"The OPTLS Protocol and TLS 1.3\",\n Proceedings of Euro S\"P 2016 , 2016,\n .\n\n [LXZFH16] Li, X., Xu, J., Feng, D., Zhang, Z., and H. Hu, \"Multiple\n Handshakes Security of TLS 1.3 Candidates\", Proceedings of\n IEEE Symposium on Security and Privacy (Oakland) 2016 ,\n 2016, .\n\n [Mac17] MacCarthaigh, C., \"Security Review of TLS1.3 0-RTT\", 2017,\n .\n\n [PSK-FINISHED]\n Cremers, C., Horvat, M., van der Merwe, T., and S. Scott,\n \"Revision 10: possible attack if client authentication is\n allowed during PSK\", 2015, .\n\n [REKEY] Abdalla, M. and M. Bellare, \"Increasing the Lifetime of a\n Key: A Comparative Analysis of the Security of Re-keying\n Techniques\", ASIACRYPT2000 , October 2000.\n\n [Res17a] Rescorla, E., \"Preliminary data on Firefox TLS 1.3\n Middlebox experiment\", 2017, .\n\n [Res17b] Rescorla, E., \"More compatibility measurement results\",\n 2017, .\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 115]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC3552] Rescorla, E. and B. Korver, \"Guidelines for Writing RFC\n Text on Security Considerations\", BCP 72, RFC 3552,\n DOI 10.17487/RFC3552, July 2003, .\n\n [RFC4086] Eastlake 3rd, D., Schiller, J., and S. Crocker,\n \"Randomness Requirements for Security\", BCP 106, RFC 4086,\n DOI 10.17487/RFC4086, June 2005, .\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346,\n DOI 10.17487/RFC4346, April 2006, .\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, DOI 10.17487/RFC4366, April 2006,\n .\n\n [RFC4492] Blake-Wilson, S., Bolyard, N., Gupta, V., Hawk, C., and B.\n Moeller, \"Elliptic Curve Cryptography (ECC) Cipher Suites\n for Transport Layer Security (TLS)\", RFC 4492,\n DOI 10.17487/RFC4492, May 2006, .\n\n [RFC5077] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 5077, DOI 10.17487/RFC5077,\n January 2008, .\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246,\n DOI 10.17487/RFC5246, August 2008, .\n\n [RFC5764] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for the Secure\n Real-time Transport Protocol (SRTP)\", RFC 5764,\n DOI 10.17487/RFC5764, May 2010, .\n\n [RFC5929] Altman, J., Williams, N., and L. Zhu, \"Channel Bindings\n for TLS\", RFC 5929, DOI 10.17487/RFC5929, July 2010,\n .\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 116]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC6091] Mavrogiannopoulos, N. and D. Gillmor, \"Using OpenPGP Keys\n for Transport Layer Security (TLS) Authentication\",\n RFC 6091, DOI 10.17487/RFC6091, February 2011,\n .\n\n [RFC6176] Turner, S. and T. Polk, \"Prohibiting Secure Sockets Layer\n (SSL) Version 2.0\", RFC 6176, DOI 10.17487/RFC6176, March\n 2011, .\n\n [RFC6347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security Version 1.2\", RFC 6347, DOI 10.17487/RFC6347,\n January 2012, .\n\n [RFC6520] Seggelmann, R., Tuexen, M., and M. Williams, \"Transport\n Layer Security (TLS) and Datagram Transport Layer Security\n (DTLS) Heartbeat Extension\", RFC 6520,\n DOI 10.17487/RFC6520, February 2012, .\n\n [RFC7230] Fielding, R., Ed. and J. Reschke, Ed., \"Hypertext Transfer\n Protocol (HTTP/1.1): Message Syntax and Routing\",\n RFC 7230, DOI 10.17487/RFC7230, June 2014,\n .\n\n [RFC7250] Wouters, P., Ed., Tschofenig, H., Ed., Gilmore, J.,\n Weiler, S., and T. Kivinen, \"Using Raw Public Keys in\n Transport Layer Security (TLS) and Datagram Transport\n Layer Security (DTLS)\", RFC 7250, DOI 10.17487/RFC7250,\n June 2014, .\n\n [RFC7465] Popov, A., \"Prohibiting RC4 Cipher Suites\", RFC 7465,\n DOI 10.17487/RFC7465, February 2015, .\n\n [RFC7568] Barnes, R., Thomson, M., Pironti, A., and A. Langley,\n \"Deprecating Secure Sockets Layer Version 3.0\", RFC 7568,\n DOI 10.17487/RFC7568, June 2015, .\n\n [RFC7627] Bhargavan, K., Ed., Delignat-Lavaud, A., Pironti, A.,\n Langley, A., and M. Ray, \"Transport Layer Security (TLS)\n Session Hash and Extended Master Secret Extension\",\n RFC 7627, DOI 10.17487/RFC7627, September 2015,\n .\n\n [RFC7685] Langley, A., \"A Transport Layer Security (TLS) ClientHello\n Padding Extension\", RFC 7685, DOI 10.17487/RFC7685,\n October 2015, .\n\n\n\nRescorla Expires September 21, 2018 [Page 117]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC7924] Santesson, S. and H. Tschofenig, \"Transport Layer Security\n (TLS) Cached Information Extension\", RFC 7924,\n DOI 10.17487/RFC7924, July 2016, .\n\n [RFC8305] Schinazi, D. and T. Pauly, \"Happy Eyeballs Version 2:\n Better Connectivity Using Concurrency\", RFC 8305,\n DOI 10.17487/RFC8305, December 2017, .\n\n [RSA] Rivest, R., Shamir, A., and L. Adleman, \"A Method for\n Obtaining Digital Signatures and Public-Key\n Cryptosystems\", Communications of the ACM v. 21, n. 2, pp.\n 120-126., February 1978.\n\n [SIGMA] Krawczyk, H., \"SIGMA: the 'SIGn-and-MAc' approach to\n authenticated Diffie-Hellman and its use in the IKE\n protocols\", Proceedings of CRYPTO 2003 , 2003.\n\n [SLOTH] Bhargavan, K. and G. Leurent, \"Transcript Collision\n Attacks: Breaking Authentication in TLS, IKE, and SSH\",\n Network and Distributed System Security Symposium (NDSS\n 2016) , 2016.\n\n [SSL2] Hickman, K., \"The SSL Protocol\", February 1995.\n\n [SSL3] Freier, A., Karlton, P., and P. Kocher, \"The SSL 3.0\n Protocol\", November 1996.\n\n [TIMING] Boneh, D. and D. Brumley, \"Remote timing attacks are\n practical\", USENIX Security Symposium, 2003.\n\n [X501] \"Information Technology - Open Systems Interconnection -\n The Directory: Models\", ITU-T X.501, 1993.\n\n12.3. URIs\n\n [1] mailto:[redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 118]\n\f\nInternet-Draft TLS March 2018\n\n\nAppendix A. State Machine\n\n This section provides a summary of the legal state transitions for\n the client and server handshakes. State names (in all capitals,\n e.g., START) have no formal meaning but are provided for ease of\n comprehension. Actions which are taken only in certain circumstances\n are indicated in []. The notation \"K_{send,recv} = foo\" means \"set\n the send/recv key to the given key\".\n\nA.1. Client\n\n START <----+\n Send ClientHello | | Recv HelloRetryRequest\n [K_send = early data] | |\n v |\n / WAIT_SH ----+\n | | Recv ServerHello\n | | K_recv = handshake\n Can | V\n send | WAIT_EE\n early | | Recv EncryptedExtensions\n data | +--------+--------+\n | Using | | Using certificate\n | PSK | v\n | | WAIT_CERT_CR\n | | Recv | | Recv CertificateRequest\n | | Certificate | v\n | | | WAIT_CERT\n | | | | Recv Certificate\n | | v v\n | | WAIT_CV\n | | | Recv CertificateVerify\n | +> WAIT_FINISHED <+\n | | Recv Finished\n \\ | [Send EndOfEarlyData]\n | K_send = handshake\n | [Send Certificate [+ CertificateVerify]]\n Can send | Send Finished\n app data --> | K_send = K_recv = application\n after here v\n CONNECTED\n\n Note that with the transitions as shown above, clients may send\n alerts that derive from post-ServerHello messages in the clear or\n with the early data keys. If clients need to send such alerts, they\n SHOULD first rekey to the handshake keys if possible.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 119]\n\f\nInternet-Draft TLS March 2018\n\n\nA.2. Server\n\n START <-----+\n Recv ClientHello | | Send HelloRetryRequest\n v |\n RECVD_CH ----+\n | Select parameters\n v\n NEGOTIATED\n | Send ServerHello\n | K_send = handshake\n | Send EncryptedExtensions\n | [Send CertificateRequest]\n Can send | [Send Certificate + CertificateVerify]\n app data | Send Finished\n after --> | K_send = application\n here +--------+--------+\n No 0-RTT | | 0-RTT\n | |\n K_recv = handshake | | K_recv = early data\n [Skip decrypt errors] | +------> WAIT_EOED -+\n | | Recv | | Recv EndOfEarlyData\n | | early data | | K_recv = handshake\n | +------------+ |\n | |\n +> WAIT_FLIGHT2 <--------+\n |\n +--------+--------+\n No auth | | Client auth\n | |\n | v\n | WAIT_CERT\n | Recv | | Recv Certificate\n | empty | v\n | Certificate | WAIT_CV\n | | | Recv\n | v | CertificateVerify\n +-> WAIT_FINISHED <---+\n | Recv Finished\n | K_recv = application\n v\n CONNECTED\n\nAppendix B. Protocol Data Structures and Constant Values\n\n This section provides the normative protocol types and constants\n definitions. Values listed as _RESERVED were used in previous\n versions of TLS and are listed here for completeness. TLS 1.3\n\n\n\nRescorla Expires September 21, 2018 [Page 120]\n\f\nInternet-Draft TLS March 2018\n\n\n implementations MUST NOT send them but might receive them from older\n TLS implementations.\n\nB.1. Record Layer\n\n enum {\n invalid(0),\n change_cipher_spec(20),\n alert(21),\n handshake(22),\n application_data(23),\n (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion legacy_record_version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n struct {\n opaque content[TLSPlaintext.length];\n ContentType type;\n uint8 zeros[length_of_padding];\n } TLSInnerPlaintext;\n\n struct {\n ContentType opaque_type = application_data; /* 23 */\n ProtocolVersion legacy_record_version = 0x0303; /* TLS v1.2 */\n uint16 length;\n opaque encrypted_record[TLSCiphertext.length];\n } TLSCiphertext;\n\nB.2. Alert Messages\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 121]\n\f\nInternet-Draft TLS March 2018\n\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure_RESERVED(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n inappropriate_fallback(86),\n user_canceled(90),\n no_renegotiation_RESERVED(100),\n missing_extension(109),\n unsupported_extension(110),\n certificate_unobtainable_RESERVED(111),\n unrecognized_name(112),\n bad_certificate_status_response(113),\n bad_certificate_hash_value_RESERVED(114),\n unknown_psk_identity(115),\n certificate_required(116),\n no_application_protocol(120),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 122]\n\f\nInternet-Draft TLS March 2018\n\n\nB.3. Handshake Protocol\n\n enum {\n hello_request_RESERVED(0),\n client_hello(1),\n server_hello(2),\n hello_verify_request_RESERVED(3),\n new_session_ticket(4),\n end_of_early_data(5),\n hello_retry_request_RESERVED(6),\n encrypted_extensions(8),\n certificate(11),\n server_key_exchange_RESERVED(12),\n certificate_request(13),\n server_hello_done_RESERVED(14),\n certificate_verify(15),\n client_key_exchange_RESERVED(16),\n finished(20),\n key_update(24),\n message_hash(254),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (Handshake.msg_type) {\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case end_of_early_data: EndOfEarlyData;\n case encrypted_extensions: EncryptedExtensions;\n case certificate_request: CertificateRequest;\n case certificate: Certificate;\n case certificate_verify: CertificateVerify;\n case finished: Finished;\n case new_session_ticket: NewSessionTicket;\n case key_update: KeyUpdate;\n };\n } Handshake;\n\nB.3.1. Key Exchange Messages\n\n uint16 ProtocolVersion;\n opaque Random[32];\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n struct {\n\n\n\nRescorla Expires September 21, 2018 [Page 123]\n\f\nInternet-Draft TLS March 2018\n\n\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id<0..32>;\n CipherSuite cipher_suites<2..2^16-2>;\n opaque legacy_compression_methods<1..2^8-1>;\n Extension extensions<8..2^16-1>;\n } ClientHello;\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id_echo<0..32>;\n CipherSuite cipher_suite;\n uint8 legacy_compression_method = 0;\n Extension extensions<6..2^16-1>;\n } ServerHello;\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n server_name(0), /* RFC 6066 */\n max_fragment_length(1), /* RFC 6066 */\n status_request(5), /* RFC 6066 */\n supported_groups(10), /* RFC 4492, 7919 */\n signature_algorithms(13), /* [[this document]] */\n use_srtp(14), /* RFC 5764 */\n heartbeat(15), /* RFC 6520 */\n application_layer_protocol_negotiation(16), /* RFC 7301 */\n signed_certificate_timestamp(18), /* RFC 6962 */\n client_certificate_type(19), /* RFC 7250 */\n server_certificate_type(20), /* RFC 7250 */\n padding(21), /* RFC 7685 */\n RESERVED(40), /* Used but never assigned */\n pre_shared_key(41), /* [[this document]] */\n early_data(42), /* [[this document]] */\n supported_versions(43), /* [[this document]] */\n cookie(44), /* [[this document]] */\n psk_key_exchange_modes(45), /* [[this document]] */\n RESERVED(46), /* Used but never assigned */\n certificate_authorities(47), /* [[this document]] */\n oid_filters(48), /* [[this document]] */\n post_handshake_auth(49), /* [[this document]] */\n signature_algorithms_cert(50), /* [[this document]] */\n key_share(51), /* [[this document]] */\n (65535)\n\n\n\nRescorla Expires September 21, 2018 [Page 124]\n\f\nInternet-Draft TLS March 2018\n\n\n } ExtensionType;\n\n struct {\n NamedGroup group;\n opaque key_exchange<1..2^16-1>;\n } KeyShareEntry;\n\n struct {\n KeyShareEntry client_shares<0..2^16-1>;\n } KeyShareClientHello;\n\n struct {\n NamedGroup selected_group;\n } KeyShareHelloRetryRequest;\n\n struct {\n KeyShareEntry server_share;\n } KeyShareServerHello;\n\n struct {\n uint8 legacy_form = 4;\n opaque X[coordinate_length];\n opaque Y[coordinate_length];\n } UncompressedPointRepresentation;\n\n enum { psk_ke(0), psk_dhe_ke(1), (255) } PskKeyExchangeMode;\n\n struct {\n PskKeyExchangeMode ke_modes<1..255>;\n } PskKeyExchangeModes;\n\n struct {} Empty;\n\n struct {\n select (Handshake.msg_type) {\n case new_session_ticket: uint32 max_early_data_size;\n case client_hello: Empty;\n case encrypted_extensions: Empty;\n };\n } EarlyDataIndication;\n\n struct {\n opaque identity<1..2^16-1>;\n uint32 obfuscated_ticket_age;\n } PskIdentity;\n\n opaque PskBinderEntry<32..255>;\n\n\n\n\nRescorla Expires September 21, 2018 [Page 125]\n\f\nInternet-Draft TLS March 2018\n\n\n struct {\n PskIdentity identities<7..2^16-1>;\n PskBinderEntry binders<33..2^16-1>;\n } OfferedPsks;\n\n struct {\n select (Handshake.msg_type) {\n case client_hello: OfferedPsks;\n case server_hello: uint16 selected_identity;\n };\n } PreSharedKeyExtension;\n\nB.3.1.1. Version Extension\n\n struct {\n select (Handshake.msg_type) {\n case client_hello:\n ProtocolVersion versions<2..254>;\n\n case server_hello: /* and HelloRetryRequest */\n ProtocolVersion selected_version;\n };\n } SupportedVersions;\n\nB.3.1.2. Cookie Extension\n\n struct {\n opaque cookie<1..2^16-1>;\n } Cookie;\n\nB.3.1.3. Signature Algorithm Extension\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 126]\n\f\nInternet-Draft TLS March 2018\n\n\n enum {\n /* RSASSA-PKCS1-v1_5 algorithms */\n rsa_pkcs1_sha256(0x0401),\n rsa_pkcs1_sha384(0x0501),\n rsa_pkcs1_sha512(0x0601),\n\n /* ECDSA algorithms */\n ecdsa_secp256r1_sha256(0x0403),\n ecdsa_secp384r1_sha384(0x0503),\n ecdsa_secp521r1_sha512(0x0603),\n\n /* RSASSA-PSS algorithms with public key OID rsaEncryption */\n rsa_pss_rsae_sha256(0x0804),\n rsa_pss_rsae_sha384(0x0805),\n rsa_pss_rsae_sha512(0x0806),\n\n /* EdDSA algorithms */\n ed25519(0x0807),\n ed448(0x0808),\n\n /* RSASSA-PSS algorithms with public key OID RSASSA-PSS */\n rsa_pss_pss_sha256(0x0809),\n rsa_pss_pss_sha384(0x080a),\n rsa_pss_pss_sha512(0x080b),\n\n /* Legacy algorithms */\n rsa_pkcs1_sha1(0x0201),\n ecdsa_sha1(0x0203),\n\n /* Reserved Code Points */\n obsolete_RESERVED(0x0000..0x0200),\n dsa_sha1_RESERVED(0x0202),\n obsolete_RESERVED(0x0204..0x0400),\n dsa_sha256_RESERVED(0x0402),\n obsolete_RESERVED(0x0404..0x0500),\n dsa_sha384_RESERVED(0x0502),\n obsolete_RESERVED(0x0504..0x0600),\n dsa_sha512_RESERVED(0x0602),\n obsolete_RESERVED(0x0604..0x06FF),\n private_use(0xFE00..0xFFFF),\n (0xFFFF)\n } SignatureScheme;\n\n struct {\n SignatureScheme supported_signature_algorithms<2..2^16-2>;\n } SignatureSchemeList;\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 127]\n\f\nInternet-Draft TLS March 2018\n\n\nB.3.1.4. Supported Groups Extension\n\n enum {\n unallocated_RESERVED(0x0000),\n\n /* Elliptic Curve Groups (ECDHE) */\n obsolete_RESERVED(0x0001..0x0016),\n secp256r1(0x0017), secp384r1(0x0018), secp521r1(0x0019),\n obsolete_RESERVED(0x001A..0x001C),\n x25519(0x001D), x448(0x001E),\n\n /* Finite Field Groups (DHE) */\n ffdhe2048(0x0100), ffdhe3072(0x0101), ffdhe4096(0x0102),\n ffdhe6144(0x0103), ffdhe8192(0x0104),\n\n /* Reserved Code Points */\n ffdhe_private_use(0x01FC..0x01FF),\n ecdhe_private_use(0xFE00..0xFEFF),\n obsolete_RESERVED(0xFF01..0xFF02),\n (0xFFFF)\n } NamedGroup;\n\n struct {\n NamedGroup named_group_list<2..2^16-1>;\n } NamedGroupList;\n\n Values within \"obsolete_RESERVED\" ranges are used in previous\n versions of TLS and MUST NOT be offered or negotiated by TLS 1.3\n implementations. The obsolete curves have various known/theoretical\n weaknesses or have had very little usage, in some cases only due to\n unintentional server configuration issues. They are no longer\n considered appropriate for general use and should be assumed to be\n potentially unsafe. The set of curves specified here is sufficient\n for interoperability with all currently deployed and properly\n configured TLS implementations.\n\nB.3.2. Server Parameters Messages\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 128]\n\f\nInternet-Draft TLS March 2018\n\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n DistinguishedName authorities<3..2^16-1>;\n } CertificateAuthoritiesExtension;\n\n struct {\n opaque certificate_extension_oid<1..2^8-1>;\n opaque certificate_extension_values<0..2^16-1>;\n } OIDFilter;\n\n struct {\n OIDFilter filters<0..2^16-1>;\n } OIDFilterExtension;\n\n struct {} PostHandshakeAuth;\n\n struct {\n Extension extensions<0..2^16-1>;\n } EncryptedExtensions;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n Extension extensions<2..2^16-1>;\n } CertificateRequest;\n\nB.3.3. Authentication Messages\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 129]\n\f\nInternet-Draft TLS March 2018\n\n\n /* Managed by IANA */\n enum {\n X509(0),\n OpenPGP_RESERVED(1),\n RawPublicKey(2),\n (255)\n } CertificateType;\n\n struct {\n select (certificate_type) {\n case RawPublicKey:\n /* From RFC 7250 ASN.1_subjectPublicKeyInfo */\n opaque ASN1_subjectPublicKeyInfo<1..2^24-1>;\n\n case X509:\n opaque cert_data<1..2^24-1>;\n };\n Extension extensions<0..2^16-1>;\n } CertificateEntry;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n CertificateEntry certificate_list<0..2^24-1>;\n } Certificate;\n\n struct {\n SignatureScheme algorithm;\n opaque signature<0..2^16-1>;\n } CertificateVerify;\n\n struct {\n opaque verify_data[Hash.length];\n } Finished;\n\nB.3.4. Ticket Establishment\n\n struct {\n uint32 ticket_lifetime;\n uint32 ticket_age_add;\n opaque ticket_nonce<0..255>;\n opaque ticket<1..2^16-1>;\n Extension extensions<0..2^16-2>;\n } NewSessionTicket;\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 130]\n\f\nInternet-Draft TLS March 2018\n\n\nB.3.5. Updating Keys\n\n struct {} EndOfEarlyData;\n\n enum {\n update_not_requested(0), update_requested(1), (255)\n } KeyUpdateRequest;\n\n struct {\n KeyUpdateRequest request_update;\n } KeyUpdate;\n\nB.4. Cipher Suites\n\n A symmetric cipher suite defines the pair of the AEAD algorithm and\n hash algorithm to be used with HKDF. Cipher suite names follow the\n naming convention:\n\n CipherSuite TLS_AEAD_HASH = VALUE;\n\n +-----------+------------------------------------------------+\n | Component | Contents |\n +-----------+------------------------------------------------+\n | TLS | The string \"TLS\" |\n | | |\n | AEAD | The AEAD algorithm used for record protection |\n | | |\n | HASH | The hash algorithm used with HKDF |\n | | |\n | VALUE | The two byte ID assigned for this cipher suite |\n +-----------+------------------------------------------------+\n\n This specification defines the following cipher suites for use with\n TLS 1.3.\n\n +------------------------------+-------------+\n | Description | Value |\n +------------------------------+-------------+\n | TLS_AES_128_GCM_SHA256 | {0x13,0x01} |\n | | |\n | TLS_AES_256_GCM_SHA384 | {0x13,0x02} |\n | | |\n | TLS_CHACHA20_POLY1305_SHA256 | {0x13,0x03} |\n | | |\n | TLS_AES_128_CCM_SHA256 | {0x13,0x04} |\n | | |\n | TLS_AES_128_CCM_8_SHA256 | {0x13,0x05} |\n +------------------------------+-------------+\n\n\n\nRescorla Expires September 21, 2018 [Page 131]\n\f\nInternet-Draft TLS March 2018\n\n\n The corresponding AEAD algorithms AEAD_AES_128_GCM, AEAD_AES_256_GCM,\n and AEAD_AES_128_CCM are defined in [RFC5116].\n AEAD_CHACHA20_POLY1305 is defined in [RFC7539]. AEAD_AES_128_CCM_8\n is defined in [RFC6655]. The corresponding hash algorithms are\n defined in [SHS].\n\n Although TLS 1.3 uses the same cipher suite space as previous\n versions of TLS, TLS 1.3 cipher suites are defined differently, only\n specifying the symmetric ciphers, and cannot be used for TLS 1.2.\n Similarly, TLS 1.2 and lower cipher suites cannot be used with TLS\n 1.3.\n\n New cipher suite values are assigned by IANA as described in\n Section 11.\n\nAppendix C. Implementation Notes\n\n The TLS protocol cannot prevent many common security mistakes. This\n section provides several recommendations to assist implementors.\n [I-D.ietf-tls-tls13-vectors] provides test vectors for TLS 1.3\n handshakes.\n\nC.1. Random Number Generation and Seeding\n\n TLS requires a cryptographically secure pseudorandom number generator\n (CSPRNG). In most cases, the operating system provides an\n appropriate facility such as /dev/urandom, which should be used\n absent other (performance) concerns. It is RECOMMENDED to use an\n existing CSPRNG implementation in preference to crafting a new one.\n Many adequate cryptographic libraries are already available under\n favorable license terms. Should those prove unsatisfactory,\n [RFC4086] provides guidance on the generation of random values.\n\n TLS uses random values both in public protocol fields such as the\n public Random values in the ClientHello and ServerHello and to\n generate keying material. With a properly functioning CSPRNG, this\n does not present a security problem as it is not feasible to\n determine the CSPRNG state from its output. However, with a broken\n CSPRNG, it may be possible for an attacker to use the public output\n to determine the CSPRNG internal state and thereby predict the keying\n material, as documented in [CHECKOWAY]. Implementations can provide\n extra security against this form of attack by using separate CSPRNGs\n to generate public and private values.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 132]\n\f\nInternet-Draft TLS March 2018\n\n\nC.2. Certificates and Authentication\n\n Implementations are responsible for verifying the integrity of\n certificates and should generally support certificate revocation\n messages. Absent a specific indication from an application profile,\n Certificates should always be verified to ensure proper signing by a\n trusted Certificate Authority (CA). The selection and addition of\n trust anchors should be done very carefully. Users should be able to\n view information about the certificate and trust anchor.\n Applications SHOULD also enforce minimum and maximum key sizes. For\n example, certification paths containing keys or signatures weaker\n than 2048-bit RSA or 224-bit ECDSA are not appropriate for secure\n applications.\n\nC.3. Implementation Pitfalls\n\n Implementation experience has shown that certain parts of earlier TLS\n specifications are not easy to understand and have been a source of\n interoperability and security problems. Many of these areas have\n been clarified in this document but this appendix contains a short\n list of the most important things that require special attention from\n implementors.\n\n TLS protocol issues:\n\n - Do you correctly handle handshake messages that are fragmented to\n multiple TLS records (see Section 5.1)? Including corner cases\n like a ClientHello that is split to several small fragments? Do\n you fragment handshake messages that exceed the maximum fragment\n size? In particular, the Certificate and CertificateRequest\n handshake messages can be large enough to require fragmentation.\n\n - Do you ignore the TLS record layer version number in all\n unencrypted TLS records? (see Appendix D)\n\n - Have you ensured that all support for SSL, RC4, EXPORT ciphers,\n and MD5 (via the \"signature_algorithms\" extension) is completely\n removed from all possible configurations that support TLS 1.3 or\n later, and that attempts to use these obsolete capabilities fail\n correctly? (see Appendix D)\n\n - Do you handle TLS extensions in ClientHello correctly, including\n unknown extensions?\n\n - When the server has requested a client certificate, but no\n suitable certificate is available, do you correctly send an empty\n Certificate message, instead of omitting the whole message (see\n Section 4.4.2.3)?\n\n\n\nRescorla Expires September 21, 2018 [Page 133]\n\f\nInternet-Draft TLS March 2018\n\n\n - When processing the plaintext fragment produced by AEAD-Decrypt\n and scanning from the end for the ContentType, do you avoid\n scanning past the start of the cleartext in the event that the\n peer has sent a malformed plaintext of all-zeros?\n\n - Do you properly ignore unrecognized cipher suites (Section 4.1.2),\n hello extensions (Section 4.2), named groups (Section 4.2.7), key\n shares (Section 4.2.8), supported versions (Section 4.2.1), and\n signature algorithms (Section 4.2.3) in the ClientHello?\n\n - As a server, do you send a HelloRetryRequest to clients which\n support a compatible (EC)DHE group but do not predict it in the\n \"key_share\" extension? As a client, do you correctly handle a\n HelloRetryRequest from the server?\n\n Cryptographic details:\n\n - What countermeasures do you use to prevent timing attacks\n [TIMING]?\n\n - When using Diffie-Hellman key exchange, do you correctly preserve\n leading zero bytes in the negotiated key (see Section 7.4.1)?\n\n - Does your TLS client check that the Diffie-Hellman parameters sent\n by the server are acceptable, (see Section 4.2.8.1)?\n\n - Do you use a strong and, most importantly, properly seeded random\n number generator (see Appendix C.1) when generating Diffie-Hellman\n private values, the ECDSA \"k\" parameter, and other security-\n critical values? It is RECOMMENDED that implementations implement\n \"deterministic ECDSA\" as specified in [RFC6979].\n\n - Do you zero-pad Diffie-Hellman public key values to the group size\n (see Section 4.2.8.1)?\n\n - Do you verify signatures after making them to protect against RSA-\n CRT key leaks? [FW15]\n\nC.4. Client Tracking Prevention\n\n Clients SHOULD NOT reuse a ticket for multiple connections. Reuse of\n a ticket allows passive observers to correlate different connections.\n Servers that issue tickets SHOULD offer at least as many tickets as\n the number of connections that a client might use; for example, a web\n browser using HTTP/1.1 [RFC7230] might open six connections to a\n server. Servers SHOULD issue new tickets with every connection.\n This ensures that clients are always able to use a new ticket when\n creating a new connection.\n\n\n\nRescorla Expires September 21, 2018 [Page 134]\n\f\nInternet-Draft TLS March 2018\n\n\nC.5. Unauthenticated Operation\n\n Previous versions of TLS offered explicitly unauthenticated cipher\n suites based on anonymous Diffie-Hellman. These modes have been\n deprecated in TLS 1.3. However, it is still possible to negotiate\n parameters that do not provide verifiable server authentication by\n several methods, including:\n\n - Raw public keys [RFC7250].\n\n - Using a public key contained in a certificate but without\n validation of the certificate chain or any of its contents.\n\n Either technique used alone is vulnerable to man-in-the-middle\n attacks and therefore unsafe for general use. However, it is also\n possible to bind such connections to an external authentication\n mechanism via out-of-band validation of the server's public key,\n trust on first use, or a mechanism such as channel bindings (though\n the channel bindings described in [RFC5929] are not defined for TLS\n 1.3). If no such mechanism is used, then the connection has no\n protection against active man-in-the-middle attack; applications MUST\n NOT use TLS in such a way absent explicit configuration or a specific\n application profile.\n\nAppendix D. Backward Compatibility\n\n The TLS protocol provides a built-in mechanism for version\n negotiation between endpoints potentially supporting different\n versions of TLS.\n\n TLS 1.x and SSL 3.0 use compatible ClientHello messages. Servers can\n also handle clients trying to use future versions of TLS as long as\n the ClientHello format remains compatible and there is at least one\n protocol version supported by both the client and the server.\n\n Prior versions of TLS used the record layer version number\n (TLSPlaintext.legacy_record_version and\n TLSCiphertext.legacy_record_version) for various purposes. As of TLS\n 1.3, this field is deprecated. The value of\n TLSPlaintext.legacy_record_version MUST be ignored by all\n implementations. The value of TLSCiphertext.legacy_record_version is\n included in the additional data for deprotection but MAY otherwise be\n ignored or MAY be validated to match the fixed constant value.\n Version negotiation is performed using only the handshake versions\n (ClientHello.legacy_version, ServerHello.legacy_version, as well as\n the ClientHello, HelloRetryRequest and ServerHello\n \"supported_versions\" extensions). In order to maximize\n interoperability with older endpoints, implementations that negotiate\n\n\n\nRescorla Expires September 21, 2018 [Page 135]\n\f\nInternet-Draft TLS March 2018\n\n\n the use of TLS 1.0-1.2 SHOULD set the record layer version number to\n the negotiated version for the ServerHello and all records\n thereafter.\n\n For maximum compatibility with previously non-standard behavior and\n misconfigured deployments, all implementations SHOULD support\n validation of certification paths based on the expectations in this\n document, even when handling prior TLS versions' handshakes. (see\n Section 4.4.2.2)\n\n TLS 1.2 and prior supported an \"Extended Master Secret\" [RFC7627]\n extension which digested large parts of the handshake transcript into\n the master secret. Because TLS 1.3 always hashes in the transcript\n up to the server CertificateVerify, implementations which support\n both TLS 1.3 and earlier versions SHOULD indicate the use of the\n Extended Master Secret extension in their APIs whenever TLS 1.3 is\n used.\n\nD.1. Negotiating with an older server\n\n A TLS 1.3 client who wishes to negotiate with servers that do not\n support TLS 1.3 will send a normal TLS 1.3 ClientHello containing\n 0x0303 (TLS 1.2) in ClientHello.legacy_version but with the correct\n version(s) in the \"supported_versions\" extension. If the server does\n not support TLS 1.3 it will respond with a ServerHello containing an\n older version number. If the client agrees to use this version, the\n negotiation will proceed as appropriate for the negotiated protocol.\n A client using a ticket for resumption SHOULD initiate the connection\n using the version that was previously negotiated.\n\n Note that 0-RTT data is not compatible with older servers and SHOULD\n NOT be sent absent knowledge that the server supports TLS 1.3. See\n Appendix D.3.\n\n If the version chosen by the server is not supported by the client\n (or not acceptable), the client MUST abort the handshake with a\n \"protocol_version\" alert.\n\n Some legacy server implementations are known to not implement the TLS\n specification properly and might abort connections upon encountering\n TLS extensions or versions which they are not aware of.\n Interoperability with buggy servers is a complex topic beyond the\n scope of this document. Multiple connection attempts may be required\n in order to negotiate a backwards compatible connection; however,\n this practice is vulnerable to downgrade attacks and is NOT\n RECOMMENDED.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 136]\n\f\nInternet-Draft TLS March 2018\n\n\nD.2. Negotiating with an older client\n\n A TLS server can also receive a ClientHello indicating a version\n number smaller than its highest supported version. If the\n \"supported_versions\" extension is present, the server MUST negotiate\n using that extension as described in Section 4.2.1. If the\n \"supported_versions\" extension is not present, the server MUST\n negotiate the minimum of ClientHello.legacy_version and TLS 1.2. For\n example, if the server supports TLS 1.0, 1.1, and 1.2, and\n legacy_version is TLS 1.0, the server will proceed with a TLS 1.0\n ServerHello. If the \"supported_versions\" extension is absent and the\n server only supports versions greater than\n ClientHello.legacy_version, the server MUST abort the handshake with\n a \"protocol_version\" alert.\n\n Note that earlier versions of TLS did not clearly specify the record\n layer version number value in all cases\n (TLSPlaintext.legacy_record_version). Servers will receive various\n TLS 1.x versions in this field, but its value MUST always be ignored.\n\nD.3. 0-RTT backwards compatibility\n\n 0-RTT data is not compatible with older servers. An older server\n will respond to the ClientHello with an older ServerHello, but it\n will not correctly skip the 0-RTT data and will fail to complete the\n handshake. This can cause issues when a client attempts to use\n 0-RTT, particularly against multi-server deployments. For example, a\n deployment could deploy TLS 1.3 gradually with some servers\n implementing TLS 1.3 and some implementing TLS 1.2, or a TLS 1.3\n deployment could be downgraded to TLS 1.2.\n\n A client that attempts to send 0-RTT data MUST fail a connection if\n it receives a ServerHello with TLS 1.2 or older. A client that\n attempts to repair this error SHOULD NOT send a TLS 1.2 ClientHello,\n but instead send a TLS 1.3 ClientHello without 0-RTT data.\n\n To avoid this error condition, multi-server deployments SHOULD ensure\n a uniform and stable deployment of TLS 1.3 without 0-RTT prior to\n enabling 0-RTT.\n\nD.4. Middlebox Compatibility Mode\n\n Field measurements [Ben17a], [Ben17b], [Res17a], [Res17b] have found\n that a significant number of middleboxes misbehave when a TLS client/\n server pair negotiates TLS 1.3. Implementations can increase the\n chance of making connections through those middleboxes by making the\n TLS 1.3 handshake look more like a TLS 1.2 handshake:\n\n\n\n\nRescorla Expires September 21, 2018 [Page 137]\n\f\nInternet-Draft TLS March 2018\n\n\n - The client always provides a non-empty session ID in the\n ClientHello, as described in the legacy_session_id section of\n Section 4.1.2.\n\n - If not offering early data, the client sends a dummy\n change_cipher_spec record (see the third paragraph of Section 5.1)\n immediately before its second flight. This may either be before\n its second ClientHello or before its encrypted handshake flight.\n If offering early data, the record is placed immediately after the\n first ClientHello.\n\n - The server sends a dummy change_cipher_spec record immediately\n after its first handshake message. This may either be after a\n ServerHello or a HelloRetryRequest.\n\n When put together, these changes make the TLS 1.3 handshake resemble\n TLS 1.2 session resumption, which improves the chance of successfully\n connecting through middleboxes. This \"compatibility mode\" is\n partially negotiated: The client can opt to provide a session ID or\n not and the server has to echo it. Either side can send\n change_cipher_spec at any time during the handshake, as they must be\n ignored by the peer, but if the client sends a non-empty session ID,\n the server MUST send the change_cipher_spec as described in this\n section.\n\nD.5. Backwards Compatibility Security Restrictions\n\n Implementations negotiating use of older versions of TLS SHOULD\n prefer forward secret and AEAD cipher suites, when available.\n\n The security of RC4 cipher suites is considered insufficient for the\n reasons cited in [RFC7465]. Implementations MUST NOT offer or\n negotiate RC4 cipher suites for any version of TLS for any reason.\n\n Old versions of TLS permitted the use of very low strength ciphers.\n Ciphers with a strength less than 112 bits MUST NOT be offered or\n negotiated for any version of TLS for any reason.\n\n The security of SSL 3.0 [SSL3] is considered insufficient for the\n reasons enumerated in [RFC7568], and it MUST NOT be negotiated for\n any reason.\n\n The security of SSL 2.0 [SSL2] is considered insufficient for the\n reasons enumerated in [RFC6176], and it MUST NOT be negotiated for\n any reason.\n\n Implementations MUST NOT send an SSL version 2.0 compatible CLIENT-\n HELLO. Implementations MUST NOT negotiate TLS 1.3 or later using an\n\n\n\nRescorla Expires September 21, 2018 [Page 138]\n\f\nInternet-Draft TLS March 2018\n\n\n SSL version 2.0 compatible CLIENT-HELLO. Implementations are NOT\n RECOMMENDED to accept an SSL version 2.0 compatible CLIENT-HELLO in\n order to negotiate older versions of TLS.\n\n Implementations MUST NOT send a ClientHello.legacy_version or\n ServerHello.legacy_version set to 0x0300 or less. Any endpoint\n receiving a Hello message with ClientHello.legacy_version or\n ServerHello.legacy_version set to 0x0300 MUST abort the handshake\n with a \"protocol_version\" alert.\n\n Implementations MUST NOT send any records with a version less than\n 0x0300. Implementations SHOULD NOT accept any records with a version\n less than 0x0300 (but may inadvertently do so if the record version\n number is ignored completely).\n\n Implementations MUST NOT use the Truncated HMAC extension, defined in\n Section 7 of [RFC6066], as it is not applicable to AEAD algorithms\n and has been shown to be insecure in some scenarios.\n\nAppendix E. Overview of Security Properties\n\n A complete security analysis of TLS is outside the scope of this\n document. In this section, we provide an informal description the\n desired properties as well as references to more detailed work in the\n research literature which provides more formal definitions.\n\n We cover properties of the handshake separately from those of the\n record layer.\n\nE.1. Handshake\n\n The TLS handshake is an Authenticated Key Exchange (AKE) protocol\n which is intended to provide both one-way authenticated (server-only)\n and mutually authenticated (client and server) functionality. At the\n completion of the handshake, each side outputs its view of the\n following values:\n\n - A set of \"session keys\" (the various secrets derived from the\n master secret) from which can be derived a set of working keys.\n\n - A set of cryptographic parameters (algorithms, etc.)\n\n - The identities of the communicating parties.\n\n We assume the attacker to be an active network attacker, which means\n it has complete control over the network used to communicate between\n the parties [RFC3552]. Even under these conditions, the handshake\n should provide the properties listed below. Note that these\n\n\n\nRescorla Expires September 21, 2018 [Page 139]\n\f\nInternet-Draft TLS March 2018\n\n\n properties are not necessarily independent, but reflect the protocol\n consumers' needs.\n\n Establishing the same session keys. The handshake needs to output\n the same set of session keys on both sides of the handshake,\n provided that it completes successfully on each endpoint (See\n [CK01]; defn 1, part 1).\n\n Secrecy of the session keys. The shared session keys should be known\n only to the communicating parties and not to the attacker (See\n [CK01]; defn 1, part 2). Note that in a unilaterally\n authenticated connection, the attacker can establish its own\n session keys with the server, but those session keys are distinct\n from those established by the client.\n\n Peer Authentication. The client's view of the peer identity should\n reflect the server's identity. If the client is authenticated,\n the server's view of the peer identity should match the client's\n identity.\n\n Uniqueness of the session keys: Any two distinct handshakes should\n produce distinct, unrelated session keys. Individual session keys\n produced by a handshake should also be distinct and independent.\n\n Downgrade protection. The cryptographic parameters should be the\n same on both sides and should be the same as if the peers had been\n communicating in the absence of an attack (See [BBFKZG16]; defns 8\n and 9}).\n\n Forward secret with respect to long-term keys If the long-term\n keying material (in this case the signature keys in certificate-\n based authentication modes or the external/resumption PSK in PSK\n with (EC)DHE modes) is compromised after the handshake is\n complete, this does not compromise the security of the session key\n (See [DOW92]), as long as the session key itself has been erased.\n The forward secrecy property is not satisfied when PSK is used in\n the \"psk_ke\" PskKeyExchangeMode.\n\n Key Compromise Impersonation (KCI) resistance In a mutually-\n authenticated connection with certificates, compromising the long-\n term secret of one actor should not break that actor's\n authentication of their peer in the given connection (see\n [HGFS15]). For example, if a client's signature key is\n compromised, it should not be possible to impersonate arbitrary\n servers to that client in subsequent handshakes.\n\n Protection of endpoint identities. The server's identity\n (certificate) should be protected against passive attackers. The\n\n\n\nRescorla Expires September 21, 2018 [Page 140]\n\f\nInternet-Draft TLS March 2018\n\n\n client's identity should be protected against both passive and\n active attackers.\n\n Informally, the signature-based modes of TLS 1.3 provide for the\n establishment of a unique, secret, shared key established by an\n (EC)DHE key exchange and authenticated by the server's signature over\n the handshake transcript, as well as tied to the server's identity by\n a MAC. If the client is authenticated by a certificate, it also\n signs over the handshake transcript and provides a MAC tied to both\n identities. [SIGMA] describes the design and analysis of this type\n of key exchange protocol. If fresh (EC)DHE keys are used for each\n connection, then the output keys are forward secret.\n\n The external PSK and resumption PSK bootstrap from a long-term shared\n secret into a unique per-connection set of short-term session keys.\n This secret may have been established in a previous handshake. If\n PSK with (EC)DHE key establishment is used, these session keys will\n also be forward secret. The resumption PSK has been designed so that\n the resumption master secret computed by connection N and needed to\n form connection N+1 is separate from the traffic keys used by\n connection N, thus providing forward secrecy between the connections.\n In addition, if multiple tickets are established on the same\n connection, they are associated with different keys, so compromise of\n the PSK associated with one ticket does not lead to the compromise of\n connections established with PSKs associated with other tickets.\n This property is most interesting if tickets are stored in a database\n (and so can be deleted) rather than if they are self-encrypted.\n\n The PSK binder value forms a binding between a PSK and the current\n handshake, as well as between the session where the PSK was\n established and the current session. This binding transitively\n includes the original handshake transcript, because that transcript\n is digested into the values which produce the Resumption Master\n Secret. This requires that both the KDF used to produce the\n resumption master secret and the MAC used to compute the binder be\n collision resistant. See Appendix E.1.1 for more on this. Note: The\n binder does not cover the binder values from other PSKs, though they\n are included in the Finished MAC.\n\n Note: TLS does not currently permit the server to send a\n certificate_request message in non-certificate-based handshakes\n (e.g., PSK). If this restriction were to be relaxed in future, the\n client's signature would not cover the server's certificate directly.\n However, if the PSK was established through a NewSessionTicket, the\n client's signature would transitively cover the server's certificate\n through the PSK binder. [PSK-FINISHED] describes a concrete attack\n on constructions that do not bind to the server's certificate (see\n also [Kraw16]). It is unsafe to use certificate-based client\n\n\n\nRescorla Expires September 21, 2018 [Page 141]\n\f\nInternet-Draft TLS March 2018\n\n\n authentication when the client might potentially share the same PSK/\n key-id pair with two different endpoints. Implementations MUST NOT\n combine external PSKs with certificate-based authentication of either\n the client or the server unless negotiated by some extension.\n\n If an exporter is used, then it produces values which are unique and\n secret (because they are generated from a unique session key).\n Exporters computed with different labels and contexts are\n computationally independent, so it is not feasible to compute one\n from another or the session secret from the exported value. Note:\n exporters can produce arbitrary-length values. If exporters are to\n be used as channel bindings, the exported value MUST be large enough\n to provide collision resistance. The exporters provided in TLS 1.3\n are derived from the same handshake contexts as the early traffic\n keys and the application traffic keys respectively, and thus have\n similar security properties. Note that they do not include the\n client's certificate; future applications which wish to bind to the\n client's certificate may need to define a new exporter that includes\n the full handshake transcript.\n\n For all handshake modes, the Finished MAC (and where present, the\n signature), prevents downgrade attacks. In addition, the use of\n certain bytes in the random nonces as described in Section 4.1.3\n allows the detection of downgrade to previous TLS versions. See\n [BBFKZG16] for more detail on TLS 1.3 and downgrade.\n\n As soon as the client and the server have exchanged enough\n information to establish shared keys, the remainder of the handshake\n is encrypted, thus providing protection against passive attackers,\n even if the computed shared key is not authenticated. Because the\n server authenticates before the client, the client can ensure that if\n it authenticates to the server, it only reveals its identity to an\n authenticated server. Note that implementations must use the\n provided record padding mechanism during the handshake to avoid\n leaking information about the identities due to length. The client's\n proposed PSK identities are not encrypted, nor is the one that the\n server selects.\n\nE.1.1. Key Derivation and HKDF\n\n Key derivation in TLS 1.3 uses the HKDF function defined in [RFC5869]\n and its two components, HKDF-Extract and HKDF-Expand. The full\n rationale for the HKDF construction can be found in [Kraw10] and the\n rationale for the way it is used in TLS 1.3 in [KW16]. Throughout\n this document, each application of HKDF-Extract is followed by one or\n more invocations of HKDF-Expand. This ordering should always be\n followed (including in future revisions of this document), in\n particular, one SHOULD NOT use an output of HKDF-Extract as an input\n\n\n\nRescorla Expires September 21, 2018 [Page 142]\n\f\nInternet-Draft TLS March 2018\n\n\n to another application of HKDF-Extract without an HKDF-Expand in\n between. Consecutive applications of HKDF-Expand are allowed as long\n as these are differentiated via the key and/or the labels.\n\n Note that HKDF-Expand implements a pseudorandom function (PRF) with\n both inputs and outputs of variable length. In some of the uses of\n HKDF in this document (e.g., for generating exporters and the\n resumption_master_secret), it is necessary that the application of\n HKDF-Expand be collision-resistant, namely, it should be infeasible\n to find two different inputs to HKDF-Expand that output the same\n value. This requires the underlying hash function to be collision\n resistant and the output length from HKDF-Expand to be of size at\n least 256 bits (or as much as needed for the hash function to prevent\n finding collisions).\n\nE.1.2. Client Authentication\n\n A client that has sent authentication data to a server, either during\n the handshake or in post-handshake authentication, cannot be sure if\n the server afterwards considers the client to be authenticated or\n not. If the client needs to determine if the server considers the\n connection to be unilaterally or mutually authenticated, this has to\n be provisioned by the application layer. See [CHHSV17] for details.\n In addition, the analysis of post-handshake authentication from\n [Kraw16] shows that the client identified by the certificate sent in\n the post-handshake phase possesses the traffic key. This party is\n therefore the client that participated in the original handshake or\n one to whom the original client delegated the traffic key (assuming\n that the traffic key has not been compromised).\n\nE.1.3. 0-RTT\n\n The 0-RTT mode of operation generally provides similar security\n properties as 1-RTT data, with the two exceptions that the 0-RTT\n encryption keys do not provide full forward secrecy and that the\n server is not able to guarantee uniqueness of the handshake (non-\n replayability) without keeping potentially undue amounts of state.\n See Section 8 for mechanisms to limit the exposure to replay.\n\nE.1.4. Exporter Independence\n\n The exporter_master_secret and early_exporter_master_secret are\n derived to be independent of the traffic keys and therefore do not\n represent a threat to the security of traffic encrypted with those\n keys. However, because these secrets can be used to compute any\n exporter value, they SHOULD be erased as soon as possible. If the\n total set of exporter labels is known, then implementations SHOULD\n pre-compute the inner Derive-Secret stage of the exporter computation\n\n\n\nRescorla Expires September 21, 2018 [Page 143]\n\f\nInternet-Draft TLS March 2018\n\n\n for all those labels, then erase the [early_]exporter_master_secret,\n followed by each inner values as soon as it is known that it will not\n be needed again.\n\nE.1.5. Post-Compromise Security\n\n TLS does not provide security for handshakes which take place after\n the peer's long-term secret (signature key or external PSK) is\n compromised. It therefore does not provide post-compromise security\n [CCG16], sometimes also referred to as backwards or future secrecy.\n This is in contrast to KCI resistance, which describes the security\n guarantees that a party has after its own long-term secret has been\n compromised.\n\nE.1.6. External References\n\n The reader should refer to the following references for analysis of\n the TLS handshake: [DFGS15] [CHSV16] [DFGS16] [KW16] [Kraw16]\n [FGSW16] [LXZFH16] [FG17] [BBK17].\n\nE.2. Record Layer\n\n The record layer depends on the handshake producing strong traffic\n secrets which can be used to derive bidirectional encryption keys and\n nonces. Assuming that is true, and the keys are used for no more\n data than indicated in Section 5.5 then the record layer should\n provide the following guarantees:\n\n Confidentiality. An attacker should not be able to determine the\n plaintext contents of a given record.\n\n Integrity. An attacker should not be able to craft a new record\n which is different from an existing record which will be accepted\n by the receiver.\n\n Order protection/non-replayability An attacker should not be able to\n cause the receiver to accept a record which it has already\n accepted or cause the receiver to accept record N+1 without having\n first processed record N.\n\n Length concealment. Given a record with a given external length, the\n attacker should not be able to determine the amount of the record\n that is content versus padding.\n\n Forward secrecy after key change. If the traffic key update\n mechanism described in Section 4.6.3 has been used and the\n previous generation key is deleted, an attacker who compromises\n\n\n\n\nRescorla Expires September 21, 2018 [Page 144]\n\f\nInternet-Draft TLS March 2018\n\n\n the endpoint should not be able to decrypt traffic encrypted with\n the old key.\n\n Informally, TLS 1.3 provides these properties by AEAD-protecting the\n plaintext with a strong key. AEAD encryption [RFC5116] provides\n confidentiality and integrity for the data. Non-replayability is\n provided by using a separate nonce for each record, with the nonce\n being derived from the record sequence number (Section 5.3), with the\n sequence number being maintained independently at both sides thus\n records which are delivered out of order result in AEAD deprotection\n failures. In order to prevent mass cryptanalysis when the same\n plaintext is repeatedly encrypted by different users under the same\n key (as is commonly the case for HTTP), the nonce is formed by mixing\n the sequence number with a secret per-connection initialization\n vector derived along with the traffic keys. See [BT16] for analysis\n of this construction.\n\n The re-keying technique in TLS 1.3 (see Section 7.2) follows the\n construction of the serial generator in [REKEY], which shows that re-\n keying can allow keys to be used for a larger number of encryptions\n than without re-keying. This relies on the security of the HKDF-\n Expand-Label function as a pseudorandom function (PRF). In addition,\n as long as this function is truly one way, it is not possible to\n compute traffic keys from prior to a key change (forward secrecy).\n\n TLS does not provide security for data which is communicated on a\n connection after a traffic secret of that connection is compromised.\n That is, TLS does not provide post-compromise security/future\n secrecy/backward secrecy with respect to the traffic secret. Indeed,\n an attacker who learns a traffic secret can compute all future\n traffic secrets on that connection. Systems which want such\n guarantees need to do a fresh handshake and establish a new\n connection with an (EC)DHE exchange.\n\nE.2.1. External References\n\n The reader should refer to the following references for analysis of\n the TLS record layer: [BMMT15] [BT16] [BDFKPPRSZZ16] [BBK17]\n [Anon18].\n\nE.3. Traffic Analysis\n\n TLS is susceptible to a variety of traffic analysis attacks based on\n observing the length and timing of encrypted packets [CLINIC]\n [HCJ16]. This is particularly easy when there is a small set of\n possible messages to be distinguished, such as for a video server\n hosting a fixed corpus of content, but still provides usable\n information even in more complicated scenarios.\n\n\n\nRescorla Expires September 21, 2018 [Page 145]\n\f\nInternet-Draft TLS March 2018\n\n\n TLS does not provide any specific defenses against this form of\n attack but does include a padding mechanism for use by applications:\n The plaintext protected by the AEAD function consists of content plus\n variable-length padding, which allows the application to produce\n arbitrary length encrypted records as well as padding-only cover\n traffic to conceal the difference between periods of transmission and\n periods of silence. Because the padding is encrypted alongside the\n actual content, an attacker cannot directly determine the length of\n the padding, but may be able to measure it indirectly by the use of\n timing channels exposed during record processing (i.e., seeing how\n long it takes to process a record or trickling in records to see\n which ones elicit a response from the server). In general, it is not\n known how to remove all of these channels because even a constant\n time padding removal function will likely feed the content into data-\n dependent functions. At minimum, a fully constant time server or\n client would require close cooperation with the application layer\n protocol implementation, including making that higher level protocol\n constant time.\n\n Note: Robust traffic analysis defences will likely lead to inferior\n performance due to delay in transmitting packets and increased\n traffic volume.\n\nE.4. Side Channel Attacks\n\n In general, TLS does not have specific defenses against side-channel\n attacks (i.e., those which attack the communications via secondary\n channels such as timing) leaving those to the implementation of the\n relevant cryptographic primitives. However, certain features of TLS\n are designed to make it easier to write side-channel resistant code:\n\n - Unlike previous versions of TLS which used a composite MAC-then-\n encrypt structure, TLS 1.3 only uses AEAD algorithms, allowing\n implementations to use self-contained constant-time\n implementations of those primitives.\n\n - TLS uses a uniform \"bad_record_mac\" alert for all decryption\n errors, which is intended to prevent an attacker from gaining\n piecewise insight into portions of the message. Additional\n resistance is provided by terminating the connection on such\n errors; a new connection will have different cryptographic\n material, preventing attacks against the cryptographic primitives\n that require multiple trials.\n\n Information leakage through side channels can occur at layers above\n TLS, in application protocols and the applications that use them.\n Resistance to side-channel attacks depends on applications and\n\n\n\n\nRescorla Expires September 21, 2018 [Page 146]\n\f\nInternet-Draft TLS March 2018\n\n\n application protocols separately ensuring that confidential\n information is not inadvertently leaked.\n\nE.5. Replay Attacks on 0-RTT\n\n Replayable 0-RTT data presents a number of security threats to TLS-\n using applications, unless those applications are specifically\n engineered to be safe under replay (minimally, this means idempotent,\n but in many cases may also require other stronger conditions, such as\n constant-time response). Potential attacks include:\n\n - Duplication of actions which cause side effects (e.g., purchasing\n an item or transferring money) to be duplicated, thus harming the\n site or the user.\n\n - Attackers can store and replay 0-RTT messages in order to re-order\n them with respect to other messages (e.g., moving a delete to\n after a create).\n\n - Exploiting cache timing behavior to discover the content of 0-RTT\n messages by replaying a 0-RTT message to a different cache node\n and then using a separate connection to measure request latency,\n to see if the two requests address the same resource.\n\n If data can be replayed a large number of times, additional attacks\n become possible, such as making repeated measurements of the speed of\n cryptographic operations. In addition, they may be able to overload\n rate-limiting systems. For further description of these attacks, see\n [Mac17].\n\n Ultimately, servers have the responsibility to protect themselves\n against attacks employing 0-RTT data replication. The mechanisms\n described in Section 8 are intended to prevent replay at the TLS\n layer but do not provide complete protection against receiving\n multiple copies of client data. TLS 1.3 falls back to the 1-RTT\n handshake when the server does not have any information about the\n client, e.g., because it is in a different cluster which does not\n share state or because the ticket has been deleted as described in\n Section 8.1. If the application layer protocol retransmits data in\n this setting, then it is possible for an attacker to induce message\n duplication by sending the ClientHello to both the original cluster\n (which processes the data immediately) and another cluster which will\n fall back to 1-RTT and process the data upon application layer\n replay. The scale of this attack is limited by the client's\n willingness to retry transactions and therefore only allows a limited\n amount of duplication, with each copy appearing as a new connection\n at the server.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 147]\n\f\nInternet-Draft TLS March 2018\n\n\n If implemented correctly, the mechanisms described in Section 8.1 and\n Section 8.2 prevent a replayed ClientHello and its associated 0-RTT\n data from being accepted multiple times by any cluster with\n consistent state; for servers which limit the use of 0-RTT to one\n cluster for a single ticket, then a given ClientHello and its\n associated 0-RTT data will only be accepted once. However, if state\n is not completely consistent, then an attacker might be able to have\n multiple copies of the data be accepted during the replication\n window. Because clients do not know the exact details of server\n behavior, they MUST NOT send messages in early data which are not\n safe to have replayed and which they would not be willing to retry\n across multiple 1-RTT connections.\n\n Application protocols MUST NOT use 0-RTT data without a profile that\n defines its use. That profile needs to identify which messages or\n interactions are safe to use with 0-RTT and how to handle the\n situation when the server rejects 0-RTT and falls back to 1-RTT.\n\n In addition, to avoid accidental misuse, TLS implementations MUST NOT\n enable 0-RTT (either sending or accepting) unless specifically\n requested by the application and MUST NOT automatically resend 0-RTT\n data if it is rejected by the server unless instructed by the\n application. Server-side applications may wish to implement special\n processing for 0-RTT data for some kinds of application traffic\n (e.g., abort the connection, request that data be resent at the\n application layer, or delay processing until the handshake\n completes). In order to allow applications to implement this kind of\n processing, TLS implementations MUST provide a way for the\n application to determine if the handshake has completed.\n\nE.5.1. Replay and Exporters\n\n Replays of the ClientHello produce the same early exporter, thus\n requiring additional care by applications which use these exporters.\n In particular, if these exporters are used as an authentication\n channel binding (e.g., by signing the output of the exporter) an\n attacker who compromises the PSK can transplant authenticators\n between connections without compromising the authentication key.\n\n In addition, the early exporter SHOULD NOT be used to generate\n server-to-client encryption keys because that would entail the reuse\n of those keys. This parallels the use of the early application\n traffic keys only in the client-to-server direction.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 148]\n\f\nInternet-Draft TLS March 2018\n\n\nE.6. PSK Identity Exposure\n\n Because implementations respond to an invalid PSK binder by aborting\n the handshake, it may be possible for an attacker to verify whether a\n given PSK identity is valid. Specifically, if a server accepts both\n external PSK and certificate-based handshakes, a valid PSK identity\n will result in a failed handshake, whereas an invalid identity will\n just be skipped and result in a successful certificate handshake.\n Servers which solely support PSK handshakes may be able to resist\n this form of attack by treating the cases where there is no valid PSK\n identity and where there is an identity but it has an invalid binder\n identically.\n\nE.7. Attacks on Static RSA\n\n Although TLS 1.3 does not use RSA key transport and so is not\n directly susceptible to Bleichenbacher-type attacks, if TLS 1.3\n servers also support static RSA in the context of previous versions\n of TLS, then it may be possible to impersonate the server for TLS 1.3\n connections [JSS15]. TLS 1.3 implementations can prevent this attack\n by disabling support for static RSA across all versions of TLS. In\n principle, implementations might also be able to separate\n certificates with different keyUsage bits for static RSA decryption\n and RSA signature, but this technique relies on clients refusing to\n accept signatures using keys in certificates that do not have the\n digitalSignature bit set, and many clients do not enforce this\n restriction.\n\nAppendix F. Working Group Information\n\n The discussion list for the IETF TLS working group is located at the\n e-mail address [redacted-email] [1]. Information on the group and\n information on how to subscribe to the list is at\n https://www.ietf.org/mailman/listinfo/tls\n\n Archives of the list can be found at: https://www.ietf.org/mail-\n archive/web/tls/current/index.html\n\nAppendix G. Contributors\n\n - Martin Abadi\n University of California, Santa Cruz\n [redacted-email]\n\n - Christopher Allen (co-editor of TLS 1.0)\n Alacrity Ventures\n [redacted-email]\n\n\n\n\nRescorla Expires September 21, 2018 [Page 149]\n\f\nInternet-Draft TLS March 2018\n\n\n - Richard Barnes\n Cisco\n [redacted-email]\n\n - Steven M. Bellovin\n Columbia University\n [redacted-email]\n\n - David Benjamin\n Google\n [redacted-email]\n\n - Benjamin Beurdouche\n INRIA & Microsoft Research\n [redacted-email]\n\n - Karthikeyan Bhargavan (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n - Simon Blake-Wilson (co-author of [RFC4492])\n BCI\n [redacted-email]\n\n - Nelson Bolyard (co-author of [RFC4492])\n Sun Microsystems, Inc.\n [redacted-email]\n\n - Ran Canetti\n IBM\n [redacted-email]\n\n - Matt Caswell\n OpenSSL\n [redacted-email]\n\n - Stephen Checkoway\n University of Illinois at Chicago\n [redacted-email]\n\n - Pete Chown\n Skygate Technology Ltd\n [redacted-email]\n\n - Katriel Cohn-Gordon\n University of Oxford\n [redacted-email]\n\n\n\n\nRescorla Expires September 21, 2018 [Page 150]\n\f\nInternet-Draft TLS March 2018\n\n\n - Cas Cremers\n University of Oxford\n [redacted-email]\n\n - Antoine Delignat-Lavaud (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n - Tim Dierks (co-editor of TLS 1.0, 1.1, and 1.2)\n Independent\n [redacted-email]\n\n - Roelof DuToit\n Symantec Corporation\n [redacted-email]\n\n - Taher Elgamal\n Securify\n [redacted-email]\n\n - Pasi Eronen\n Nokia\n [redacted-email]\n\n - Cedric Fournet\n Microsoft\n [redacted-email]\n\n - Anil Gangolli\n [redacted-email]\n\n - David M. Garrett\n [redacted-email]\n\n - Illya Gerasymchuk\n Independent\n [redacted-email]\n\n - Alessandro Ghedini\n Cloudflare Inc.\n [redacted-email]\n\n - Daniel Kahn Gillmor\n ACLU\n [redacted-email]\n\n - Matthew Green\n Johns Hopkins University\n\n\n\nRescorla Expires September 21, 2018 [Page 151]\n\f\nInternet-Draft TLS March 2018\n\n\n [redacted-email]\n\n - Jens Guballa\n ETAS\n [redacted-email]\n\n - Felix Guenther\n TU Darmstadt\n [redacted-email]\n\n - Vipul Gupta (co-author of [RFC4492])\n Sun Microsystems Laboratories\n [redacted-email]\n\n - Chris Hawk (co-author of [RFC4492])\n Corriente Networks LLC\n [redacted-email]\n\n - Kipp Hickman\n\n - Alfred Hoenes\n\n - David Hopwood\n Independent Consultant\n [redacted-email]\n\n - Marko Horvat\n MPI-SWS\n [redacted-email]\n\n - Jonathan Hoyland\n Royal Holloway, University of London [redacted-email]\n\n - Subodh Iyengar\n Facebook\n [redacted-email]\n\n - Benjamin Kaduk\n Akamai\n [redacted-email]\n\n - Hubert Kario\n Red Hat Inc.\n [redacted-email]\n\n - Phil Karlton (co-author of SSL 3.0)\n\n - Leon Klingele\n\n\n\nRescorla Expires September 21, 2018 [Page 152]\n\f\nInternet-Draft TLS March 2018\n\n\n Independent\n [redacted-email]\n\n - Paul Kocher (co-author of SSL 3.0)\n Cryptography Research\n [redacted-email]\n\n - Hugo Krawczyk\n IBM\n [redacted-email]\n\n - Adam Langley (co-author of [RFC7627])\n Google\n [redacted-email]\n\n - Olivier Levillain\n ANSSI\n [redacted-email]\n\n - Xiaoyin Liu\n University of North Carolina at Chapel Hill\n [redacted-email]\n\n - Ilari Liusvaara\n Independent\n [redacted-email]\n\n - Atul Luykx\n K.U. Leuven\n [redacted-email]\n\n - Colm MacCarthaigh\n Amazon Web Services\n [redacted-email]\n\n - Carl Mehner\n USAA\n [redacted-email]\n\n - Jan Mikkelsen\n Transactionware\n [redacted-email]\n\n - Bodo Moeller (co-author of [RFC4492])\n Google\n [redacted-email]\n\n - Kyle Nekritz\n\n\n\nRescorla Expires September 21, 2018 [Page 153]\n\f\nInternet-Draft TLS March 2018\n\n\n Facebook\n [redacted-email]\n\n - Erik Nygren\n Akamai Technologies\n [redacted-email]\n\n - Magnus Nystrom\n Microsoft\n [redacted-email]\n\n - Kazuho Oku\n DeNA Co., Ltd.\n [redacted-email]\n\n - Kenny Paterson\n Royal Holloway, University of London\n [redacted-email]\n\n - Alfredo Pironti (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n - Andrei Popov\n Microsoft\n [redacted-email]\n\n - Marsh Ray (co-author of [RFC7627])\n Microsoft\n [redacted-email]\n\n - Robert Relyea\n Netscape Communications\n [redacted-email]\n\n - Kyle Rose\n Akamai Technologies\n [redacted-email]\n\n - Jim Roskind\n Amazon\n [redacted-email]\n\n - Michael Sabin\n\n - Joe Salowey\n Tableau Software\n [redacted-email]\n\n\n\nRescorla Expires September 21, 2018 [Page 154]\n\f\nInternet-Draft TLS March 2018\n\n\n - Rich Salz\n Akamai\n [redacted-email]\n\n - David Schinazi\n Apple Inc.\n [redacted-email]\n\n - Sam Scott\n Royal Holloway, University of London\n [redacted-email]\n\n - Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n - Brian Smith\n Independent\n [redacted-email]\n\n - Brian Sniffen\n Akamai Technologies\n [redacted-email]\n\n - Nick Sullivan\n Cloudflare Inc.\n [redacted-email]\n\n - Bjoern Tackmann\n University of California, San Diego\n [redacted-email]\n\n - Tim Taubert\n Mozilla\n [redacted-email]\n\n - Martin Thomson\n Mozilla\n [redacted-email]\n\n - Sean Turner\n sn3rd\n [redacted-email]\n\n - Steven Valdez\n Google\n [redacted-email]\n\n\n\n\nRescorla Expires September 21, 2018 [Page 155]\n\f\nInternet-Draft TLS March 2018\n\n\n - Filippo Valsorda\n Cloudflare Inc.\n [redacted-email]\n\n - Thyla van der Merwe\n Royal Holloway, University of London\n [redacted-email]\n\n - Victor Vasiliev\n Google\n [redacted-email]\n\n - Tom Weinstein\n\n - Hoeteck Wee\n Ecole Normale Superieure, Paris\n [redacted-email]\n\n - David Wong\n NCC Group\n [redacted-email]\n\n - Christopher A. Wood\n Apple Inc.\n [redacted-email]\n\n - Tim Wright\n Vodafone\n [redacted-email]\n\n - Peter Wu\n Independent\n [redacted-email]\n\n - Kazu Yamamoto\n Internet Initiative Japan Inc.\n [redacted-email]\n\nAuthor's Address\n\n Eric Rescorla\n RTFM, Inc.\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 156]", "text_sha256": "a7cda7524edf08714e498aaf4b2a684ee1b5fdf0c9648505eb54649b7cfb0b92"}, {"draft_references": ["draft-ietf-tls-tls13-vectors-06"], "excluded_rfc_relation_targets": [{"kind": "obsoletes", "reason": "not_requested", "rfc_number": 2018}], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "8446"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_id": "obsoletes:5077", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "5077"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_id": "obsoletes:5246", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "5246"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_id": "obsoletes:6961", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "6961"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_id": "obsoletes:2018", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "2018"}, {"char_end": 238, "char_start": 219, "evidence_quote": "Updates: 5705, 6066", "fact_id": "updates:5705", "field": "updates", "raw_char_end": 244, "raw_char_start": 225, "value": "5705"}, {"char_end": 238, "char_start": 219, "evidence_quote": "Updates: 5705, 6066", "fact_id": "updates:6066", "field": "updates", "raw_char_end": 244, "raw_char_start": 225, "value": "6066"}, {"char_end": 261161, "char_start": 261130, "evidence_quote": "draft-ietf-tls-tls13-vectors-06", "fact_id": "draft_reference:draft-ietf-tls-tls13-vectors-06", "field": "draft_reference", "raw_char_end": 261167, "raw_char_start": 261136, "value": "draft-ietf-tls-tls13-vectors-06"}], "kind": "rfc", "occurred_at": "2018-08-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 86, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd", "record_id": "ietf:rfc:8446", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "rfc_number": 8446, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd", "source_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "temporal_semantics": "source_event_time", "text": "Internet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 8446 Mozilla\nObsoletes: 5077, 5246, 6961 August 2018\nUpdates: 5705, 6066\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS) Protocol Version 1.3\n\nAbstract\n\n This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol. TLS allows client/server applications to communicate\n over the Internet in a way that is designed to prevent eavesdropping,\n tampering, and message forgery.\n\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077,\n 5246, and 6961. This document also specifies new requirements for\n TLS 1.2 implementations.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 7841.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n https://www.rfc-editor.org/info/rfc8446.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 8446 TLS August 2018\n\n\nCopyright Notice\n\n Copyright (c) 2018 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (https://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 8446 TLS August 2018\n\n\nTable of Contents\n\n 1. Introduction ....................................................6\n 1.1. Conventions and Terminology ................................7\n 1.2. Major Differences from TLS 1.2 .............................8\n 1.3. Updates Affecting TLS 1.2 ..................................9\n 2. Protocol Overview ..............................................10\n 2.1. Incorrect DHE Share .......................................14\n 2.2. Resumption and Pre-Shared Key (PSK) .......................15\n 2.3. 0-RTT Data ................................................17\n 3. Presentation Language ..........................................19\n 3.1. Basic Block Size ..........................................19\n 3.2. Miscellaneous .............................................20\n 3.3. Numbers ...................................................20\n 3.4. Vectors ...................................................20\n 3.5. Enumerateds ...............................................21\n 3.6. Constructed Types .........................................22\n 3.7. Constants .................................................23\n 3.8. Variants ..................................................23\n 4. Handshake Protocol .............................................24\n 4.1. Key Exchange Messages .....................................25\n 4.1.1. Cryptographic Negotiation ..........................26\n 4.1.2. Client Hello .......................................27\n 4.1.3. Server Hello .......................................31\n 4.1.4. Hello Retry Request ................................33\n 4.2. Extensions ................................................35\n 4.2.1. Supported Versions .................................39\n 4.2.2. Cookie .............................................40\n 4.2.3. Signature Algorithms ...............................41\n 4.2.4. Certificate Authorities ............................45\n 4.2.5. OID Filters ........................................45\n 4.2.6. Post-Handshake Client Authentication ...............47\n 4.2.7. Supported Groups ...................................47\n 4.2.8. Key Share ..........................................48\n 4.2.9. Pre-Shared Key Exchange Modes ......................51\n 4.2.10. Early Data Indication .............................52\n 4.2.11. Pre-Shared Key Extension ..........................55\n 4.3. Server Parameters .........................................59\n 4.3.1. Encrypted Extensions ...............................60\n 4.3.2. Certificate Request ................................60\n 4.4. Authentication Messages ...................................61\n 4.4.1. The Transcript Hash ................................63\n 4.4.2. Certificate ........................................64\n 4.4.3. Certificate Verify .................................69\n 4.4.4. Finished ...........................................71\n 4.5. End of Early Data .........................................72\n\n\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 8446 TLS August 2018\n\n\n 4.6. Post-Handshake Messages ...................................73\n 4.6.1. New Session Ticket Message .........................73\n 4.6.2. Post-Handshake Authentication ......................75\n 4.6.3. Key and Initialization Vector Update ...............76\n 5. Record Protocol ................................................77\n 5.1. Record Layer ..............................................78\n 5.2. Record Payload Protection .................................80\n 5.3. Per-Record Nonce ..........................................82\n 5.4. Record Padding ............................................83\n 5.5. Limits on Key Usage .......................................84\n 6. Alert Protocol .................................................85\n 6.1. Closure Alerts ............................................87\n 6.2. Error Alerts ..............................................88\n 7. Cryptographic Computations .....................................90\n 7.1. Key Schedule ..............................................91\n 7.2. Updating Traffic Secrets ..................................94\n 7.3. Traffic Key Calculation ...................................95\n 7.4. (EC)DHE Shared Secret Calculation .........................95\n 7.4.1. Finite Field Diffie-Hellman ........................95\n 7.4.2. Elliptic Curve Diffie-Hellman ......................96\n 7.5. Exporters .................................................97\n 8. 0-RTT and Anti-Replay ..........................................98\n 8.1. Single-Use Tickets ........................................99\n 8.2. Client Hello Recording ....................................99\n 8.3. Freshness Checks .........................................101\n 9. Compliance Requirements .......................................102\n 9.1. Mandatory-to-Implement Cipher Suites .....................102\n 9.2. Mandatory-to-Implement Extensions ........................103\n 9.3. Protocol Invariants ......................................104\n 10. Security Considerations ......................................106\n 11. IANA Considerations ..........................................106\n 12. References ...................................................109\n 12.1. Normative References ....................................109\n 12.2. Informative References ..................................112\n Appendix A. State Machine ........................................120\n A.1. Client ....................................................120\n A.2. Server ....................................................121\n Appendix B. Protocol Data Structures and Constant Values .........122\n B.1. Record Layer ..............................................122\n B.2. Alert Messages ............................................123\n B.3. Handshake Protocol ........................................124\n B.3.1. Key Exchange Messages .................................125\n B.3.2. Server Parameters Messages ............................131\n B.3.3. Authentication Messages ...............................132\n B.3.4. Ticket Establishment ..................................132\n B.3.5. Updating Keys .........................................133\n B.4. Cipher Suites .............................................133\n\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 8446 TLS August 2018\n\n\n Appendix C. Implementation Notes .................................134\n C.1. Random Number Generation and Seeding ......................134\n C.2. Certificates and Authentication ...........................135\n C.3. Implementation Pitfalls ...................................135\n C.4. Client Tracking Prevention ................................137\n C.5. Unauthenticated Operation .................................137\n Appendix D. Backward Compatibility ...............................138\n D.1. Negotiating with an Older Server ..........................139\n D.2. Negotiating with an Older Client ..........................139\n D.3. 0-RTT Backward Compatibility ..............................140\n D.4. Middlebox Compatibility Mode ..............................140\n D.5. Security Restrictions Related to Backward Compatibility ...141\n Appendix E. Overview of Security Properties ......................142\n E.1. Handshake .................................................142\n E.1.1. Key Derivation and HKDF ...............................145\n E.1.2. Client Authentication .................................146\n E.1.3. 0-RTT .................................................146\n E.1.4. Exporter Independence .................................146\n E.1.5. Post-Compromise Security ..............................146\n E.1.6. External References ...................................147\n E.2. Record Layer ..............................................147\n E.2.1. External References ...................................148\n E.3. Traffic Analysis ..........................................148\n E.4. Side-Channel Attacks ......................................149\n E.5. Replay Attacks on 0-RTT ...................................150\n E.5.1. Replay and Exporters ..................................151\n E.6. PSK Identity Exposure .....................................152\n E.7. Sharing PSKs ..............................................152\n E.8. Attacks on Static RSA .....................................152\n Contributors .....................................................153\n Author's Address .................................................160\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 8446 TLS August 2018\n\n\n1. Introduction\n\n The primary goal of TLS is to provide a secure channel between two\n communicating peers; the only requirement from the underlying\n transport is a reliable, in-order data stream. Specifically, the\n secure channel should provide the following properties:\n\n - Authentication: The server side of the channel is always\n authenticated; the client side is optionally authenticated.\n Authentication can happen via asymmetric cryptography (e.g., RSA\n [RSA], the Elliptic Curve Digital Signature Algorithm (ECDSA)\n [ECDSA], or the Edwards-Curve Digital Signature Algorithm (EdDSA)\n [RFC8032]) or a symmetric pre-shared key (PSK).\n\n - Confidentiality: Data sent over the channel after establishment is\n only visible to the endpoints. TLS does not hide the length of\n the data it transmits, though endpoints are able to pad TLS\n records in order to obscure lengths and improve protection against\n traffic analysis techniques.\n\n - Integrity: Data sent over the channel after establishment cannot\n be modified by attackers without detection.\n\n These properties should be true even in the face of an attacker who\n has complete control of the network, as described in [RFC3552]. See\n Appendix E for a more complete statement of the relevant security\n properties.\n\n TLS consists of two primary components:\n\n - A handshake protocol (Section 4) that authenticates the\n communicating parties, negotiates cryptographic modes and\n parameters, and establishes shared keying material. The handshake\n protocol is designed to resist tampering; an active attacker\n should not be able to force the peers to negotiate different\n parameters than they would if the connection were not under\n attack.\n\n - A record protocol (Section 5) that uses the parameters established\n by the handshake protocol to protect traffic between the\n communicating peers. The record protocol divides traffic up into\n a series of records, each of which is independently protected\n using the traffic keys.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 8446 TLS August 2018\n\n\n TLS is application protocol independent; higher-level protocols can\n layer on top of TLS transparently. The TLS standard, however, does\n not specify how protocols add security with TLS; how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n This document defines TLS version 1.3. While TLS 1.3 is not directly\n compatible with previous versions, all versions of TLS incorporate a\n versioning mechanism which allows clients and servers to\n interoperably negotiate a common version if one is supported by both\n peers.\n\n This document supersedes and obsoletes previous versions of TLS,\n including version 1.2 [RFC5246]. It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2. Because TLS 1.3 changes the way keys are\n derived, it updates [RFC5705] as described in Section 7.5. It also\n changes how Online Certificate Status Protocol (OCSP) messages are\n carried and therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in Section 4.4.2.1.\n\n1.1. Conventions and Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all\n capitals, as shown here.\n\n The following terms are used:\n\n client: The endpoint initiating the TLS connection.\n\n connection: A transport-layer connection between two endpoints.\n\n endpoint: Either the client or server of the connection.\n\n handshake: An initial negotiation between client and server that\n establishes the parameters of their subsequent interactions\n within TLS.\n\n peer: An endpoint. When discussing a particular endpoint, \"peer\"\n refers to the endpoint that is not the primary subject of\n discussion.\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\f\nRFC 8446 TLS August 2018\n\n\n receiver: An endpoint that is receiving records.\n\n sender: An endpoint that is transmitting records.\n\n server: The endpoint that did not initiate the TLS connection.\n\n1.2. Major Differences from TLS 1.2\n\n The following is a list of the major functional differences between\n TLS 1.2 and TLS 1.3. It is not intended to be exhaustive, and there\n are many minor differences.\n\n - The list of supported symmetric encryption algorithms has been\n pruned of all algorithms that are considered legacy. Those that\n remain are all Authenticated Encryption with Associated Data\n (AEAD) algorithms. The cipher suite concept has been changed to\n separate the authentication and key exchange mechanisms from the\n record protection algorithm (including secret key length) and a\n hash to be used with both the key derivation function and\n handshake message authentication code (MAC).\n\n - A zero round-trip time (0-RTT) mode was added, saving a round trip\n at connection setup for some application data, at the cost of\n certain security properties.\n\n - Static RSA and Diffie-Hellman cipher suites have been removed; all\n public-key based key exchange mechanisms now provide forward\n secrecy.\n\n - All handshake messages after the ServerHello are now encrypted.\n The newly introduced EncryptedExtensions message allows various\n extensions previously sent in the clear in the ServerHello to also\n enjoy confidentiality protection.\n\n - The key derivation functions have been redesigned. The new design\n allows easier analysis by cryptographers due to their improved key\n separation properties. The HMAC-based Extract-and-Expand Key\n Derivation Function (HKDF) is used as an underlying primitive.\n\n - The handshake state machine has been significantly restructured to\n be more consistent and to remove superfluous messages such as\n ChangeCipherSpec (except when needed for middlebox compatibility).\n\n - Elliptic curve algorithms are now in the base spec, and new\n signature algorithms, such as EdDSA, are included. TLS 1.3\n removed point format negotiation in favor of a single point format\n for each curve.\n\n\n\n\nRescorla Standards Track [Page 8]\n\f\nRFC 8446 TLS August 2018\n\n\n - Other cryptographic improvements were made, including changing the\n RSA padding to use the RSA Probabilistic Signature Scheme\n (RSASSA-PSS), and the removal of compression, the Digital\n Signature Algorithm (DSA), and custom Ephemeral Diffie-Hellman\n (DHE) groups.\n\n - The TLS 1.2 version negotiation mechanism has been deprecated in\n favor of a version list in an extension. This increases\n compatibility with existing servers that incorrectly implemented\n version negotiation.\n\n - Session resumption with and without server-side state as well as\n the PSK-based cipher suites of earlier TLS versions have been\n replaced by a single new PSK exchange.\n\n - References have been updated to point to the updated versions of\n RFCs, as appropriate (e.g., RFC 5280 rather than RFC 3280).\n\n1.3. Updates Affecting TLS 1.2\n\n This document defines several changes that optionally affect\n implementations of TLS 1.2, including those which do not also support\n TLS 1.3:\n\n - A version downgrade protection mechanism is described in\n Section 4.1.3.\n\n - RSASSA-PSS signature schemes are defined in Section 4.2.3.\n\n - The \"supported_versions\" ClientHello extension can be used to\n negotiate the version of TLS to use, in preference to the\n legacy_version field of the ClientHello.\n\n - The \"signature_algorithms_cert\" extension allows a client to\n indicate which signature algorithms it can validate in X.509\n certificates.\n\n Additionally, this document clarifies some compliance requirements\n for earlier versions of TLS; see Section 9.3.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 9]\n\f\nRFC 8446 TLS August 2018\n\n\n2. Protocol Overview\n\n The cryptographic parameters used by the secure channel are produced\n by the TLS handshake protocol. This sub-protocol of TLS is used by\n the client and server when first communicating with each other. The\n handshake protocol allows peers to negotiate a protocol version,\n select cryptographic algorithms, optionally authenticate each other,\n and establish shared secret keying material. Once the handshake is\n complete, the peers use the established keys to protect the\n application-layer traffic.\n\n A failure of the handshake or other protocol error triggers the\n termination of the connection, optionally preceded by an alert\n message (Section 6).\n\n TLS supports three basic key exchange modes:\n\n - (EC)DHE (Diffie-Hellman over either finite fields or elliptic\n curves)\n\n - PSK-only\n\n - PSK with (EC)DHE\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 10]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 1 below shows the basic full TLS handshake:\n\n Client Server\n\nKey ^ ClientHello\nExch | + key_share*\n | + signature_algorithms*\n | + psk_key_exchange_modes*\n v + pre_shared_key* -------->\n ServerHello ^ Key\n + key_share* | Exch\n + pre_shared_key* v\n {EncryptedExtensions} ^ Server\n {CertificateRequest*} v Params\n {Certificate*} ^\n {CertificateVerify*} | Auth\n {Finished} v\n <-------- [Application Data*]\n ^ {Certificate*}\nAuth | {CertificateVerify*}\n v {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 1: Message Flow for Full TLS Handshake\n\n The handshake can be thought of as having three phases (indicated in\n the diagram above):\n\n - Key Exchange: Establish shared keying material and select the\n cryptographic parameters. Everything after this phase is\n encrypted.\n\n - Server Parameters: Establish other handshake parameters\n (whether the client is authenticated, application-layer protocol\n support, etc.).\n\n\n\n\nRescorla Standards Track [Page 11]\n\f\nRFC 8446 TLS August 2018\n\n\n - Authentication: Authenticate the server (and, optionally, the\n client) and provide key confirmation and handshake integrity.\n\n In the Key Exchange phase, the client sends the ClientHello\n (Section 4.1.2) message, which contains a random nonce\n (ClientHello.random); its offered protocol versions; a list of\n symmetric cipher/HKDF hash pairs; either a set of Diffie-Hellman key\n shares (in the \"key_share\" (Section 4.2.8) extension), a set of\n pre-shared key labels (in the \"pre_shared_key\" (Section 4.2.11)\n extension), or both; and potentially additional extensions.\n Additional fields and/or messages may also be present for middlebox\n compatibility.\n\n The server processes the ClientHello and determines the appropriate\n cryptographic parameters for the connection. It then responds with\n its own ServerHello (Section 4.1.3), which indicates the negotiated\n connection parameters. The combination of the ClientHello and the\n ServerHello determines the shared keys. If (EC)DHE key establishment\n is in use, then the ServerHello contains a \"key_share\" extension with\n the server's ephemeral Diffie-Hellman share; the server's share MUST\n be in the same group as one of the client's shares. If PSK key\n establishment is in use, then the ServerHello contains a\n \"pre_shared_key\" extension indicating which of the client's offered\n PSKs was selected. Note that implementations can use (EC)DHE and PSK\n together, in which case both extensions will be supplied.\n\n The server then sends two messages to establish the Server\n Parameters:\n\n EncryptedExtensions: responses to ClientHello extensions that are\n not required to determine the cryptographic parameters, other than\n those that are specific to individual certificates.\n [Section 4.3.1]\n\n CertificateRequest: if certificate-based client authentication is\n desired, the desired parameters for that certificate. This\n message is omitted if client authentication is not desired.\n [Section 4.3.2]\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 12]\n\f\nRFC 8446 TLS August 2018\n\n\n Finally, the client and server exchange Authentication messages. TLS\n uses the same set of messages every time that certificate-based\n authentication is needed. (PSK-based authentication happens as a\n side effect of key exchange.) Specifically:\n\n Certificate: The certificate of the endpoint and any per-certificate\n extensions. This message is omitted by the server if not\n authenticating with a certificate and by the client if the server\n did not send CertificateRequest (thus indicating that the client\n should not authenticate with a certificate). Note that if raw\n public keys [RFC7250] or the cached information extension\n [RFC7924] are in use, then this message will not contain a\n certificate but rather some other value corresponding to the\n server's long-term key. [Section 4.4.2]\n\n CertificateVerify: A signature over the entire handshake using the\n private key corresponding to the public key in the Certificate\n message. This message is omitted if the endpoint is not\n authenticating via a certificate. [Section 4.4.3]\n\n Finished: A MAC (Message Authentication Code) over the entire\n handshake. This message provides key confirmation, binds the\n endpoint's identity to the exchanged keys, and in PSK mode also\n authenticates the handshake. [Section 4.4.4]\n\n Upon receiving the server's messages, the client responds with its\n Authentication messages, namely Certificate and CertificateVerify (if\n requested), and Finished.\n\n At this point, the handshake is complete, and the client and server\n derive the keying material required by the record layer to exchange\n application-layer data protected through authenticated encryption.\n Application Data MUST NOT be sent prior to sending the Finished\n message, except as specified in Section 2.3. Note that while the\n server may send Application Data prior to receiving the client's\n Authentication messages, any data sent at that point is, of course,\n being sent to an unauthenticated peer.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 13]\n\f\nRFC 8446 TLS August 2018\n\n\n2.1. Incorrect DHE Share\n\n If the client has not provided a sufficient \"key_share\" extension\n (e.g., it includes only DHE or ECDHE groups unacceptable to or\n unsupported by the server), the server corrects the mismatch with a\n HelloRetryRequest and the client needs to restart the handshake with\n an appropriate \"key_share\" extension, as shown in Figure 2. If no\n common cryptographic parameters can be negotiated, the server MUST\n abort the handshake with an appropriate alert.\n\n Client Server\n\n ClientHello\n + key_share -------->\n HelloRetryRequest\n <-------- + key_share\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 2: Message Flow for a Full Handshake with\n Mismatched Parameters\n\n Note: The handshake transcript incorporates the initial\n ClientHello/HelloRetryRequest exchange; it is not reset with the\n new ClientHello.\n\n TLS also allows several optimized variants of the basic handshake, as\n described in the following sections.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 14]\n\f\nRFC 8446 TLS August 2018\n\n\n2.2. Resumption and Pre-Shared Key (PSK)\n\n Although TLS PSKs can be established out of band, PSKs can also be\n established in a previous connection and then used to establish a new\n connection (\"session resumption\" or \"resuming\" with a PSK). Once a\n handshake has completed, the server can send the client a PSK\n identity that corresponds to a unique key derived from the initial\n handshake (see Section 4.6.1). The client can then use that PSK\n identity in future handshakes to negotiate the use of the associated\n PSK. If the server accepts the PSK, then the security context of the\n new connection is cryptographically tied to the original connection\n and the key derived from the initial handshake is used to bootstrap\n the cryptographic state instead of a full handshake. In TLS 1.2 and\n below, this functionality was provided by \"session IDs\" and \"session\n tickets\" [RFC5077]. Both mechanisms are obsoleted in TLS 1.3.\n\n PSKs can be used with (EC)DHE key exchange in order to provide\n forward secrecy in combination with shared keys, or can be used\n alone, at the cost of losing forward secrecy for the application\n data.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 15]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 3 shows a pair of handshakes in which the first handshake\n establishes a PSK and the second handshake uses it:\n\n Client Server\n\n Initial Handshake:\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n <-------- [NewSessionTicket]\n [Application Data] <-------> [Application Data]\n\n\n Subsequent Handshake:\n ClientHello\n + key_share*\n + pre_shared_key -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n {Finished}\n <-------- [Application Data*]\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 3: Message Flow for Resumption and PSK\n\n As the server is authenticating via a PSK, it does not send a\n Certificate or a CertificateVerify message. When a client offers\n resumption via a PSK, it SHOULD also supply a \"key_share\" extension\n to the server to allow the server to decline resumption and fall back\n to a full handshake, if needed. The server responds with a\n \"pre_shared_key\" extension to negotiate the use of PSK key\n establishment and can (as shown here) respond with a \"key_share\"\n extension to do (EC)DHE key establishment, thus providing forward\n secrecy.\n\n\n\n\nRescorla Standards Track [Page 16]\n\f\nRFC 8446 TLS August 2018\n\n\n When PSKs are provisioned out of band, the PSK identity and the KDF\n hash algorithm to be used with the PSK MUST also be provisioned.\n\n Note: When using an out-of-band provisioned pre-shared secret, a\n critical consideration is using sufficient entropy during the key\n generation, as discussed in [RFC4086]. Deriving a shared secret\n from a password or other low-entropy sources is not secure. A\n low-entropy secret, or password, is subject to dictionary attacks\n based on the PSK binder. The specified PSK authentication is not\n a strong password-based authenticated key exchange even when used\n with Diffie-Hellman key establishment. Specifically, it does not\n prevent an attacker that can observe the handshake from performing\n a brute-force attack on the password/pre-shared key.\n\n2.3. 0-RTT Data\n\n When clients and servers share a PSK (either obtained externally or\n via a previous handshake), TLS 1.3 allows clients to send data on the\n first flight (\"early data\"). The client uses the PSK to authenticate\n the server and to encrypt the early data.\n\n As shown in Figure 4, the 0-RTT data is just added to the 1-RTT\n handshake in the first flight. The rest of the handshake uses the\n same messages as for a 1-RTT handshake with PSK resumption.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 17]\n\f\nRFC 8446 TLS August 2018\n\n\n Client Server\n\n ClientHello\n + early_data\n + key_share*\n + psk_key_exchange_modes\n + pre_shared_key\n (Application Data*) -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n + early_data*\n {Finished}\n <-------- [Application Data*]\n (EndOfEarlyData)\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n () Indicates messages protected using keys\n derived from a client_early_traffic_secret.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 4: Message Flow for a 0-RTT Handshake\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 18]\n\f\nRFC 8446 TLS August 2018\n\n\n IMPORTANT NOTE: The security properties for 0-RTT data are weaker\n than those for other kinds of TLS data. Specifically:\n\n 1. This data is not forward secret, as it is encrypted solely under\n keys derived using the offered PSK.\n\n 2. There are no guarantees of non-replay between connections.\n Protection against replay for ordinary TLS 1.3 1-RTT data is\n provided via the server's Random value, but 0-RTT data does not\n depend on the ServerHello and therefore has weaker guarantees.\n This is especially relevant if the data is authenticated either\n with TLS client authentication or inside the application\n protocol. The same warnings apply to any use of the\n early_exporter_master_secret.\n\n 0-RTT data cannot be duplicated within a connection (i.e., the server\n will not process the same data twice for the same connection), and an\n attacker will not be able to make 0-RTT data appear to be 1-RTT data\n (because it is protected with different keys). Appendix E.5 contains\n a description of potential attacks, and Section 8 describes\n mechanisms which the server can use to limit the impact of replay.\n\n3. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used.\n\n3.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple-byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n following example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 19]\n\f\nRFC 8446 TLS August 2018\n\n\n3.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" (double\n brackets).\n\n Single-byte entities containing uninterpreted data are of\n type opaque.\n\n A type alias T' for an existing type T is defined by:\n\n T T';\n\n3.3. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are constructed from a fixed-length series of\n bytes concatenated as described in Section 3.1 and are also unsigned.\n The following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are transmitted\n in network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n3.4. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 20]\n\f\nRFC 8446 TLS August 2018\n\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* three consecutive 3-byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n In the following example, \"mandatory\" is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 3.3). Similarly,\n \"longer\" can represent up to 800 bytes of data, or 400 uint16\n elements, and it may be empty. Its encoding will include a two-byte\n actual length field prepended to the vector. The length of an\n encoded vector must be an exact multiple of the length of a single\n element (e.g., a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is two bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n3.5. Enumerateds\n\n An additional sparse data type, called \"enum\" or \"enumerated\", is\n available. Each definition is a different type. Only enumerateds of\n the same type may be assigned or compared. Every element of an\n enumerated must be assigned a value, as demonstrated in the following\n example. Since the elements of the enumerated are not ordered, they\n can be assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n Future extensions or additions to the protocol may define new values.\n Implementations need to be able to parse and ignore unknown values\n unless the definition of the field states otherwise.\n\n\n\n\n\nRescorla Standards Track [Page 21]\n\f\nRFC 8446 TLS August 2018\n\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4 in the current\n version of the protocol.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n The names assigned to enumerateds do not need to be unique. The\n numerical value can describe a range over which the same name\n applies. The value includes the minimum and maximum inclusive values\n in that range, separated by two period characters. This is\n principally useful for reserving regions of the space.\n\n enum { sad(0), meh(1..254), happy(255) } Mood;\n\n3.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax used for definitions is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } T;\n\n Fixed- and variable-length vector fields are allowed using the\n standard vector syntax. Structures V1 and V2 in the variants example\n (Section 3.8) demonstrate this.\n\n\n\nRescorla Standards Track [Page 22]\n\f\nRFC 8446 TLS August 2018\n\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n\n3.7. Constants\n\n Fields and variables may be assigned a fixed value using \"=\", as in:\n\n struct {\n T1 f1 = 8; /* T.f1 must always be 8 */\n T2 f2;\n } T;\n\n3.8. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. Each\n arm of the select (below) specifies the type of that variant's field\n and an optional field label. The mechanism by which the variant is\n selected at runtime is not prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1 [[fe1]];\n case e2: Te2 [[fe2]];\n ....\n case en: Ten [[fen]];\n };\n } Tv;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 23]\n\f\nRFC 8446 TLS August 2018\n\n\n For example:\n\n enum { apple(0), orange(1) } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n VariantTag type;\n select (VariantRecord.type) {\n case apple: V1;\n case orange: V2;\n };\n } VariantRecord;\n\n4. Handshake Protocol\n\n The handshake protocol is used to negotiate the security parameters\n of a connection. Handshake messages are supplied to the TLS record\n layer, where they are encapsulated within one or more TLSPlaintext or\n TLSCiphertext structures which are processed and transmitted as\n specified by the current active connection state.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 24]\n\f\nRFC 8446 TLS August 2018\n\n\n enum {\n client_hello(1),\n server_hello(2),\n new_session_ticket(4),\n end_of_early_data(5),\n encrypted_extensions(8),\n certificate(11),\n certificate_request(13),\n certificate_verify(15),\n finished(20),\n key_update(24),\n message_hash(254),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* remaining bytes in message */\n select (Handshake.msg_type) {\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case end_of_early_data: EndOfEarlyData;\n case encrypted_extensions: EncryptedExtensions;\n case certificate_request: CertificateRequest;\n case certificate: Certificate;\n case certificate_verify: CertificateVerify;\n case finished: Finished;\n case new_session_ticket: NewSessionTicket;\n case key_update: KeyUpdate;\n };\n } Handshake;\n\n Protocol messages MUST be sent in the order defined in Section 4.4.1\n and shown in the diagrams in Section 2. A peer which receives a\n handshake message in an unexpected order MUST abort the handshake\n with an \"unexpected_message\" alert.\n\n New handshake message types are assigned by IANA as described in\n Section 11.\n\n4.1. Key Exchange Messages\n\n The key exchange messages are used to determine the security\n capabilities of the client and the server and to establish shared\n secrets, including the traffic keys used to protect the rest of the\n handshake and the data.\n\n\n\n\n\nRescorla Standards Track [Page 25]\n\f\nRFC 8446 TLS August 2018\n\n\n4.1.1. Cryptographic Negotiation\n\n In TLS, the cryptographic negotiation proceeds by the client offering\n the following four sets of options in its ClientHello:\n\n - A list of cipher suites which indicates the AEAD algorithm/HKDF\n hash pairs which the client supports.\n\n - A \"supported_groups\" (Section 4.2.7) extension which indicates the\n (EC)DHE groups which the client supports and a \"key_share\"\n (Section 4.2.8) extension which contains (EC)DHE shares for some\n or all of these groups.\n\n - A \"signature_algorithms\" (Section 4.2.3) extension which indicates\n the signature algorithms which the client can accept. A\n \"signature_algorithms_cert\" extension (Section 4.2.3) may also be\n added to indicate certificate-specific signature algorithms.\n\n - A \"pre_shared_key\" (Section 4.2.11) extension which contains a\n list of symmetric key identities known to the client and a\n \"psk_key_exchange_modes\" (Section 4.2.9) extension which indicates\n the key exchange modes that may be used with PSKs.\n\n If the server does not select a PSK, then the first three of these\n options are entirely orthogonal: the server independently selects a\n cipher suite, an (EC)DHE group and key share for key establishment,\n and a signature algorithm/certificate pair to authenticate itself to\n the client. If there is no overlap between the received\n \"supported_groups\" and the groups supported by the server, then the\n server MUST abort the handshake with a \"handshake_failure\" or an\n \"insufficient_security\" alert.\n\n If the server selects a PSK, then it MUST also select a key\n establishment mode from the set indicated by the client's\n \"psk_key_exchange_modes\" extension (at present, PSK alone or with\n (EC)DHE). Note that if the PSK can be used without (EC)DHE, then\n non-overlap in the \"supported_groups\" parameters need not be fatal,\n as it is in the non-PSK case discussed in the previous paragraph.\n\n If the server selects an (EC)DHE group and the client did not offer a\n compatible \"key_share\" extension in the initial ClientHello, the\n server MUST respond with a HelloRetryRequest (Section 4.1.4) message.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 26]\n\f\nRFC 8446 TLS August 2018\n\n\n If the server successfully selects parameters and does not require a\n HelloRetryRequest, it indicates the selected parameters in the\n ServerHello as follows:\n\n - If PSK is being used, then the server will send a \"pre_shared_key\"\n extension indicating the selected key.\n\n - When (EC)DHE is in use, the server will also provide a \"key_share\"\n extension. If PSK is not being used, then (EC)DHE and\n certificate-based authentication are always used.\n\n - When authenticating via a certificate, the server will send the\n Certificate (Section 4.4.2) and CertificateVerify (Section 4.4.3)\n messages. In TLS 1.3 as defined by this document, either a PSK or\n a certificate is always used, but not both. Future documents may\n define how to use them together.\n\n If the server is unable to negotiate a supported set of parameters\n (i.e., there is no overlap between the client and server parameters),\n it MUST abort the handshake with either a \"handshake_failure\" or\n \"insufficient_security\" fatal alert (see Section 6).\n\n4.1.2. Client Hello\n\n When a client first connects to a server, it is REQUIRED to send the\n ClientHello as its first TLS message. The client will also send a\n ClientHello when the server has responded to its ClientHello with a\n HelloRetryRequest. In that case, the client MUST send the same\n ClientHello without modification, except as follows:\n\n - If a \"key_share\" extension was supplied in the HelloRetryRequest,\n replacing the list of shares with a list containing a single\n KeyShareEntry from the indicated group.\n\n - Removing the \"early_data\" extension (Section 4.2.10) if one was\n present. Early data is not permitted after a HelloRetryRequest.\n\n - Including a \"cookie\" extension if one was provided in the\n HelloRetryRequest.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 27]\n\f\nRFC 8446 TLS August 2018\n\n\n - Updating the \"pre_shared_key\" extension if present by recomputing\n the \"obfuscated_ticket_age\" and binder values and (optionally)\n removing any PSKs which are incompatible with the server's\n indicated cipher suite.\n\n - Optionally adding, removing, or changing the length of the\n \"padding\" extension [RFC7685].\n\n - Other modifications that may be allowed by an extension defined in\n the future and present in the HelloRetryRequest.\n\n Because TLS 1.3 forbids renegotiation, if a server has negotiated\n TLS 1.3 and receives a ClientHello at any other time, it MUST\n terminate the connection with an \"unexpected_message\" alert.\n\n If a server established a TLS connection with a previous version of\n TLS and receives a TLS 1.3 ClientHello in a renegotiation, it MUST\n retain the previous protocol version. In particular, it MUST NOT\n negotiate TLS 1.3.\n\n Structure of this message:\n\n uint16 ProtocolVersion;\n opaque Random[32];\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id<0..32>;\n CipherSuite cipher_suites<2..2^16-2>;\n opaque legacy_compression_methods<1..2^8-1>;\n Extension extensions<8..2^16-1>;\n } ClientHello;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 28]\n\f\nRFC 8446 TLS August 2018\n\n\n legacy_version: In previous versions of TLS, this field was used for\n version negotiation and represented the highest version number\n supported by the client. Experience has shown that many servers\n do not properly implement version negotiation, leading to \"version\n intolerance\" in which the server rejects an otherwise acceptable\n ClientHello with a version number higher than it supports. In\n TLS 1.3, the client indicates its version preferences in the\n \"supported_versions\" extension (Section 4.2.1) and the\n legacy_version field MUST be set to 0x0303, which is the version\n number for TLS 1.2. TLS 1.3 ClientHellos are identified as having\n a legacy_version of 0x0303 and a supported_versions extension\n present with 0x0304 as the highest version indicated therein.\n (See Appendix D for details about backward compatibility.)\n\n random: 32 bytes generated by a secure random number generator. See\n Appendix C for additional information.\n\n legacy_session_id: Versions of TLS before TLS 1.3 supported a\n \"session resumption\" feature which has been merged with pre-shared\n keys in this version (see Section 2.2). A client which has a\n cached session ID set by a pre-TLS 1.3 server SHOULD set this\n field to that value. In compatibility mode (see Appendix D.4),\n this field MUST be non-empty, so a client not offering a\n pre-TLS 1.3 session MUST generate a new 32-byte value. This value\n need not be random but SHOULD be unpredictable to avoid\n implementations fixating on a specific value (also known as\n ossification). Otherwise, it MUST be set as a zero-length vector\n (i.e., a zero-valued single byte length field).\n\n cipher_suites: A list of the symmetric cipher options supported by\n the client, specifically the record protection algorithm\n (including secret key length) and a hash to be used with HKDF, in\n descending order of client preference. Values are defined in\n Appendix B.4. If the list contains cipher suites that the server\n does not recognize, support, or wish to use, the server MUST\n ignore those cipher suites and process the remaining ones as\n usual. If the client is attempting a PSK key establishment, it\n SHOULD advertise at least one cipher suite indicating a Hash\n associated with the PSK.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 29]\n\f\nRFC 8446 TLS August 2018\n\n\n legacy_compression_methods: Versions of TLS before 1.3 supported\n compression with the list of supported compression methods being\n sent in this field. For every TLS 1.3 ClientHello, this vector\n MUST contain exactly one byte, set to zero, which corresponds to\n the \"null\" compression method in prior versions of TLS. If a\n TLS 1.3 ClientHello is received with any other value in this\n field, the server MUST abort the handshake with an\n \"illegal_parameter\" alert. Note that TLS 1.3 servers might\n receive TLS 1.2 or prior ClientHellos which contain other\n compression methods and (if negotiating such a prior version) MUST\n follow the procedures for the appropriate prior version of TLS.\n\n extensions: Clients request extended functionality from servers by\n sending data in the extensions field. The actual \"Extension\"\n format is defined in Section 4.2. In TLS 1.3, the use of certain\n extensions is mandatory, as functionality has moved into\n extensions to preserve ClientHello compatibility with previous\n versions of TLS. Servers MUST ignore unrecognized extensions.\n\n All versions of TLS allow an extensions field to optionally follow\n the compression_methods field. TLS 1.3 ClientHello messages always\n contain extensions (minimally \"supported_versions\", otherwise, they\n will be interpreted as TLS 1.2 ClientHello messages). However,\n TLS 1.3 servers might receive ClientHello messages without an\n extensions field from prior versions of TLS. The presence of\n extensions can be detected by determining whether there are bytes\n following the compression_methods field at the end of the\n ClientHello. Note that this method of detecting optional data\n differs from the normal TLS method of having a variable-length field,\n but it is used for compatibility with TLS before extensions were\n defined. TLS 1.3 servers will need to perform this check first and\n only attempt to negotiate TLS 1.3 if the \"supported_versions\"\n extension is present. If negotiating a version of TLS prior to 1.3,\n a server MUST check that the message either contains no data after\n legacy_compression_methods or that it contains a valid extensions\n block with no data following. If not, then it MUST abort the\n handshake with a \"decode_error\" alert.\n\n In the event that a client requests additional functionality using\n extensions and this functionality is not supplied by the server, the\n client MAY abort the handshake.\n\n After sending the ClientHello message, the client waits for a\n ServerHello or HelloRetryRequest message. If early data is in use,\n the client may transmit early Application Data (Section 2.3) while\n waiting for the next handshake message.\n\n\n\n\n\nRescorla Standards Track [Page 30]\n\f\nRFC 8446 TLS August 2018\n\n\n4.1.3. Server Hello\n\n The server will send this message in response to a ClientHello\n message to proceed with the handshake if it is able to negotiate an\n acceptable set of handshake parameters based on the ClientHello.\n\n Structure of this message:\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id_echo<0..32>;\n CipherSuite cipher_suite;\n uint8 legacy_compression_method = 0;\n Extension extensions<6..2^16-1>;\n } ServerHello;\n\n legacy_version: In previous versions of TLS, this field was used for\n version negotiation and represented the selected version number\n for the connection. Unfortunately, some middleboxes fail when\n presented with new values. In TLS 1.3, the TLS server indicates\n its version using the \"supported_versions\" extension\n (Section 4.2.1), and the legacy_version field MUST be set to\n 0x0303, which is the version number for TLS 1.2. (See Appendix D\n for details about backward compatibility.)\n\n random: 32 bytes generated by a secure random number generator. See\n Appendix C for additional information. The last 8 bytes MUST be\n overwritten as described below if negotiating TLS 1.2 or TLS 1.1,\n but the remaining bytes MUST be random. This structure is\n generated by the server and MUST be generated independently of the\n ClientHello.random.\n\n legacy_session_id_echo: The contents of the client's\n legacy_session_id field. Note that this field is echoed even if\n the client's value corresponded to a cached pre-TLS 1.3 session\n which the server has chosen not to resume. A client which\n receives a legacy_session_id_echo field that does not match what\n it sent in the ClientHello MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n cipher_suite: The single cipher suite selected by the server from\n the list in ClientHello.cipher_suites. A client which receives a\n cipher suite that was not offered MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n legacy_compression_method: A single byte which MUST have the\n value 0.\n\n\n\nRescorla Standards Track [Page 31]\n\f\nRFC 8446 TLS August 2018\n\n\n extensions: A list of extensions. The ServerHello MUST only include\n extensions which are required to establish the cryptographic\n context and negotiate the protocol version. All TLS 1.3\n ServerHello messages MUST contain the \"supported_versions\"\n extension. Current ServerHello messages additionally contain\n either the \"pre_shared_key\" extension or the \"key_share\"\n extension, or both (when using a PSK with (EC)DHE key\n establishment). Other extensions (see Section 4.2) are sent\n separately in the EncryptedExtensions message.\n\n For reasons of backward compatibility with middleboxes (see\n Appendix D.4), the HelloRetryRequest message uses the same structure\n as the ServerHello, but with Random set to the special value of the\n SHA-256 of \"HelloRetryRequest\":\n\n CF 21 AD 74 E5 9A 61 11 BE 1D 8C 02 1E 65 B8 91\n C2 A2 11 16 7A BB 8C 5E 07 9E 09 E2 C8 A8 33 9C\n\n Upon receiving a message with type server_hello, implementations MUST\n first examine the Random value and, if it matches this value, process\n it as described in Section 4.1.4).\n\n TLS 1.3 has a downgrade protection mechanism embedded in the server's\n random value. TLS 1.3 servers which negotiate TLS 1.2 or below in\n response to a ClientHello MUST set the last 8 bytes of their Random\n value specially in their ServerHello.\n\n If negotiating TLS 1.2, TLS 1.3 servers MUST set the last 8 bytes of\n their Random value to the bytes:\n\n 44 4F 57 4E 47 52 44 01\n\n If negotiating TLS 1.1 or below, TLS 1.3 servers MUST, and TLS 1.2\n servers SHOULD, set the last 8 bytes of their ServerHello.Random\n value to the bytes:\n\n 44 4F 57 4E 47 52 44 00\n\n TLS 1.3 clients receiving a ServerHello indicating TLS 1.2 or below\n MUST check that the last 8 bytes are not equal to either of these\n values. TLS 1.2 clients SHOULD also check that the last 8 bytes are\n not equal to the second value if the ServerHello indicates TLS 1.1 or\n below. If a match is found, the client MUST abort the handshake with\n an \"illegal_parameter\" alert. This mechanism provides limited\n protection against downgrade attacks over and above what is provided\n by the Finished exchange: because the ServerKeyExchange, a message\n present in TLS 1.2 and below, includes a signature over both random\n values, it is not possible for an active attacker to modify the\n\n\n\nRescorla Standards Track [Page 32]\n\f\nRFC 8446 TLS August 2018\n\n\n random values without detection as long as ephemeral ciphers are\n used. It does not provide downgrade protection when static RSA\n is used.\n\n Note: This is a change from [RFC5246], so in practice many TLS 1.2\n clients and servers will not behave as specified above.\n\n A legacy TLS client performing renegotiation with TLS 1.2 or prior\n and which receives a TLS 1.3 ServerHello during renegotiation MUST\n abort the handshake with a \"protocol_version\" alert. Note that\n renegotiation is not possible when TLS 1.3 has been negotiated.\n\n4.1.4. Hello Retry Request\n\n The server will send this message in response to a ClientHello\n message if it is able to find an acceptable set of parameters but the\n ClientHello does not contain sufficient information to proceed with\n the handshake. As discussed in Section 4.1.3, the HelloRetryRequest\n has the same format as a ServerHello message, and the legacy_version,\n legacy_session_id_echo, cipher_suite, and legacy_compression_method\n fields have the same meaning. However, for convenience we discuss\n \"HelloRetryRequest\" throughout this document as if it were a distinct\n message.\n\n The server's extensions MUST contain \"supported_versions\".\n Additionally, it SHOULD contain the minimal set of extensions\n necessary for the client to generate a correct ClientHello pair. As\n with the ServerHello, a HelloRetryRequest MUST NOT contain any\n extensions that were not first offered by the client in its\n ClientHello, with the exception of optionally the \"cookie\" (see\n Section 4.2.2) extension.\n\n Upon receipt of a HelloRetryRequest, the client MUST check the\n legacy_version, legacy_session_id_echo, cipher_suite, and\n legacy_compression_method as specified in Section 4.1.3 and then\n process the extensions, starting with determining the version using\n \"supported_versions\". Clients MUST abort the handshake with an\n \"illegal_parameter\" alert if the HelloRetryRequest would not result\n in any change in the ClientHello. If a client receives a second\n HelloRetryRequest in the same connection (i.e., where the ClientHello\n was itself in response to a HelloRetryRequest), it MUST abort the\n handshake with an \"unexpected_message\" alert.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 33]\n\f\nRFC 8446 TLS August 2018\n\n\n Otherwise, the client MUST process all extensions in the\n HelloRetryRequest and send a second updated ClientHello. The\n HelloRetryRequest extensions defined in this specification are:\n\n - supported_versions (see Section 4.2.1)\n\n - cookie (see Section 4.2.2)\n\n - key_share (see Section 4.2.8)\n\n A client which receives a cipher suite that was not offered MUST\n abort the handshake. Servers MUST ensure that they negotiate the\n same cipher suite when receiving a conformant updated ClientHello (if\n the server selects the cipher suite as the first step in the\n negotiation, then this will happen automatically). Upon receiving\n the ServerHello, clients MUST check that the cipher suite supplied in\n the ServerHello is the same as that in the HelloRetryRequest and\n otherwise abort the handshake with an \"illegal_parameter\" alert.\n\n In addition, in its updated ClientHello, the client SHOULD NOT offer\n any pre-shared keys associated with a hash other than that of the\n selected cipher suite. This allows the client to avoid having to\n compute partial hash transcripts for multiple hashes in the second\n ClientHello.\n\n The value of selected_version in the HelloRetryRequest\n \"supported_versions\" extension MUST be retained in the ServerHello,\n and a client MUST abort the handshake with an \"illegal_parameter\"\n alert if the value changes.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 34]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2. Extensions\n\n A number of TLS messages contain tag-length-value encoded extensions\n structures.\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n server_name(0), /* RFC 6066 */\n max_fragment_length(1), /* RFC 6066 */\n status_request(5), /* RFC 6066 */\n supported_groups(10), /* RFC 8422, 7919 */\n signature_algorithms(13), /* RFC 8446 */\n use_srtp(14), /* RFC 5764 */\n heartbeat(15), /* RFC 6520 */\n application_layer_protocol_negotiation(16), /* RFC 7301 */\n signed_certificate_timestamp(18), /* RFC 6962 */\n client_certificate_type(19), /* RFC 7250 */\n server_certificate_type(20), /* RFC 7250 */\n padding(21), /* RFC 7685 */\n pre_shared_key(41), /* RFC 8446 */\n early_data(42), /* RFC 8446 */\n supported_versions(43), /* RFC 8446 */\n cookie(44), /* RFC 8446 */\n psk_key_exchange_modes(45), /* RFC 8446 */\n certificate_authorities(47), /* RFC 8446 */\n oid_filters(48), /* RFC 8446 */\n post_handshake_auth(49), /* RFC 8446 */\n signature_algorithms_cert(50), /* RFC 8446 */\n key_share(51), /* RFC 8446 */\n (65535)\n } ExtensionType;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 35]\n\f\nRFC 8446 TLS August 2018\n\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The list of extension types is maintained by IANA as described in\n Section 11.\n\n Extensions are generally structured in a request/response fashion,\n though some extensions are just indications with no corresponding\n response. The client sends its extension requests in the ClientHello\n message, and the server sends its extension responses in the\n ServerHello, EncryptedExtensions, HelloRetryRequest, and Certificate\n messages. The server sends extension requests in the\n CertificateRequest message which a client MAY respond to with a\n Certificate message. The server MAY also send unsolicited extensions\n in the NewSessionTicket, though the client does not respond directly\n to these.\n\n Implementations MUST NOT send extension responses if the remote\n endpoint did not send the corresponding extension requests, with the\n exception of the \"cookie\" extension in the HelloRetryRequest. Upon\n receiving such an extension, an endpoint MUST abort the handshake\n with an \"unsupported_extension\" alert.\n\n The table below indicates the messages where a given extension may\n appear, using the following notation: CH (ClientHello),\n SH (ServerHello), EE (EncryptedExtensions), CT (Certificate),\n CR (CertificateRequest), NST (NewSessionTicket), and\n HRR (HelloRetryRequest). If an implementation receives an extension\n which it recognizes and which is not specified for the message in\n which it appears, it MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 36]\n\f\nRFC 8446 TLS August 2018\n\n\n +--------------------------------------------------+-------------+\n | Extension | TLS 1.3 |\n +--------------------------------------------------+-------------+\n | server_name [RFC6066] | CH, EE |\n | | |\n | max_fragment_length [RFC6066] | CH, EE |\n | | |\n | status_request [RFC6066] | CH, CR, CT |\n | | |\n | supported_groups [RFC7919] | CH, EE |\n | | |\n | signature_algorithms (RFC 8446) | CH, CR |\n | | |\n | use_srtp [RFC5764] | CH, EE |\n | | |\n | heartbeat [RFC6520] | CH, EE |\n | | |\n | application_layer_protocol_negotiation [RFC7301] | CH, EE |\n | | |\n | signed_certificate_timestamp [RFC6962] | CH, CR, CT |\n | | |\n | client_certificate_type [RFC7250] | CH, EE |\n | | |\n | server_certificate_type [RFC7250] | CH, EE |\n | | |\n | padding [RFC7685] | CH |\n | | |\n | key_share (RFC 8446) | CH, SH, HRR |\n | | |\n | pre_shared_key (RFC 8446) | CH, SH |\n | | |\n | psk_key_exchange_modes (RFC 8446) | CH |\n | | |\n | early_data (RFC 8446) | CH, EE, NST |\n | | |\n | cookie (RFC 8446) | CH, HRR |\n | | |\n | supported_versions (RFC 8446) | CH, SH, HRR |\n | | |\n | certificate_authorities (RFC 8446) | CH, CR |\n | | |\n | oid_filters (RFC 8446) | CR |\n | | |\n | post_handshake_auth (RFC 8446) | CH |\n | | |\n | signature_algorithms_cert (RFC 8446) | CH, CR |\n +--------------------------------------------------+-------------+\n\n\n\n\nRescorla Standards Track [Page 37]\n\f\nRFC 8446 TLS August 2018\n\n\n When multiple extensions of different types are present, the\n extensions MAY appear in any order, with the exception of\n \"pre_shared_key\" (Section 4.2.11) which MUST be the last extension in\n the ClientHello (but can appear anywhere in the ServerHello\n extensions block). There MUST NOT be more than one extension of the\n same type in a given extension block.\n\n In TLS 1.3, unlike TLS 1.2, extensions are negotiated for each\n handshake even when in resumption-PSK mode. However, 0-RTT\n parameters are those negotiated in the previous handshake; mismatches\n may require rejecting 0-RTT (see Section 4.2.10).\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions (e.g., the handshake cannot continue), and some are\n simply refusals to support particular features. In general, error\n alerts should be used for the former and a field in the server\n extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem. Often the fact that the extension fields are\n included in the inputs to the Finished message hashes will be\n sufficient, but extreme care is needed when the extension changes\n the meaning of messages sent in the handshake phase. Designers\n and implementors should be aware of the fact that until the\n handshake has been authenticated, active attackers can modify\n messages and insert, remove, or replace extensions.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 38]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.1. Supported Versions\n\n struct {\n select (Handshake.msg_type) {\n case client_hello:\n ProtocolVersion versions<2..254>;\n\n case server_hello: /* and HelloRetryRequest */\n ProtocolVersion selected_version;\n };\n } SupportedVersions;\n\n The \"supported_versions\" extension is used by the client to indicate\n which versions of TLS it supports and by the server to indicate which\n version it is using. The extension contains a list of supported\n versions in preference order, with the most preferred version first.\n Implementations of this specification MUST send this extension in the\n ClientHello containing all versions of TLS which they are prepared to\n negotiate (for this specification, that means minimally 0x0304, but\n if previous versions of TLS are allowed to be negotiated, they MUST\n be present as well).\n\n If this extension is not present, servers which are compliant with\n this specification and which also support TLS 1.2 MUST negotiate\n TLS 1.2 or prior as specified in [RFC5246], even if\n ClientHello.legacy_version is 0x0304 or later. Servers MAY abort the\n handshake upon receiving a ClientHello with legacy_version 0x0304 or\n later.\n\n If this extension is present in the ClientHello, servers MUST NOT use\n the ClientHello.legacy_version value for version negotiation and MUST\n use only the \"supported_versions\" extension to determine client\n preferences. Servers MUST only select a version of TLS present in\n that extension and MUST ignore any unknown versions that are present\n in that extension. Note that this mechanism makes it possible to\n negotiate a version prior to TLS 1.2 if one side supports a sparse\n range. Implementations of TLS 1.3 which choose to support prior\n versions of TLS SHOULD support TLS 1.2. Servers MUST be prepared to\n receive ClientHellos that include this extension but do not include\n 0x0304 in the list of versions.\n\n A server which negotiates a version of TLS prior to TLS 1.3 MUST set\n ServerHello.version and MUST NOT send the \"supported_versions\"\n extension. A server which negotiates TLS 1.3 MUST respond by sending\n a \"supported_versions\" extension containing the selected version\n value (0x0304). It MUST set the ServerHello.legacy_version field to\n 0x0303 (TLS 1.2). Clients MUST check for this extension prior to\n processing the rest of the ServerHello (although they will have to\n\n\n\nRescorla Standards Track [Page 39]\n\f\nRFC 8446 TLS August 2018\n\n\n parse the ServerHello in order to read the extension). If this\n extension is present, clients MUST ignore the\n ServerHello.legacy_version value and MUST use only the\n \"supported_versions\" extension to determine the selected version. If\n the \"supported_versions\" extension in the ServerHello contains a\n version not offered by the client or contains a version prior to\n TLS 1.3, the client MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n4.2.2. Cookie\n\n struct {\n opaque cookie<1..2^16-1>;\n } Cookie;\n\n Cookies serve two primary purposes:\n\n - Allowing the server to force the client to demonstrate\n reachability at their apparent network address (thus providing a\n measure of DoS protection). This is primarily useful for\n non-connection-oriented transports (see [RFC6347] for an example\n of this).\n\n - Allowing the server to offload state to the client, thus allowing\n it to send a HelloRetryRequest without storing any state. The\n server can do this by storing the hash of the ClientHello in the\n HelloRetryRequest cookie (protected with some suitable integrity\n protection algorithm).\n\n When sending a HelloRetryRequest, the server MAY provide a \"cookie\"\n extension to the client (this is an exception to the usual rule that\n the only extensions that may be sent are those that appear in the\n ClientHello). When sending the new ClientHello, the client MUST copy\n the contents of the extension received in the HelloRetryRequest into\n a \"cookie\" extension in the new ClientHello. Clients MUST NOT use\n cookies in their initial ClientHello in subsequent connections.\n\n When a server is operating statelessly, it may receive an unprotected\n record of type change_cipher_spec between the first and second\n ClientHello (see Section 5). Since the server is not storing any\n state, this will appear as if it were the first message to be\n received. Servers operating statelessly MUST ignore these records.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 40]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.3. Signature Algorithms\n\n TLS 1.3 provides two extensions for indicating which signature\n algorithms may be used in digital signatures. The\n \"signature_algorithms_cert\" extension applies to signatures in\n certificates, and the \"signature_algorithms\" extension, which\n originally appeared in TLS 1.2, applies to signatures in\n CertificateVerify messages. The keys found in certificates MUST also\n be of appropriate type for the signature algorithms they are used\n with. This is a particular issue for RSA keys and PSS signatures, as\n described below. If no \"signature_algorithms_cert\" extension is\n present, then the \"signature_algorithms\" extension also applies to\n signatures appearing in certificates. Clients which desire the\n server to authenticate itself via a certificate MUST send the\n \"signature_algorithms\" extension. If a server is authenticating via\n a certificate and the client has not sent a \"signature_algorithms\"\n extension, then the server MUST abort the handshake with a\n \"missing_extension\" alert (see Section 9.2).\n\n The \"signature_algorithms_cert\" extension was added to allow\n implementations which supported different sets of algorithms for\n certificates and in TLS itself to clearly signal their capabilities.\n TLS 1.2 implementations SHOULD also process this extension.\n Implementations which have the same policy in both cases MAY omit the\n \"signature_algorithms_cert\" extension.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 41]\n\f\nRFC 8446 TLS August 2018\n\n\n The \"extension_data\" field of these extensions contains a\n SignatureSchemeList value:\n\n enum {\n /* RSASSA-PKCS1-v1_5 algorithms */\n rsa_pkcs1_sha256(0x0401),\n rsa_pkcs1_sha384(0x0501),\n rsa_pkcs1_sha512(0x0601),\n\n /* ECDSA algorithms */\n ecdsa_secp256r1_sha256(0x0403),\n ecdsa_secp384r1_sha384(0x0503),\n ecdsa_secp521r1_sha512(0x0603),\n\n /* RSASSA-PSS algorithms with public key OID rsaEncryption */\n rsa_pss_rsae_sha256(0x0804),\n rsa_pss_rsae_sha384(0x0805),\n rsa_pss_rsae_sha512(0x0806),\n\n /* EdDSA algorithms */\n ed25519(0x0807),\n ed448(0x0808),\n\n /* RSASSA-PSS algorithms with public key OID RSASSA-PSS */\n rsa_pss_pss_sha256(0x0809),\n rsa_pss_pss_sha384(0x080a),\n rsa_pss_pss_sha512(0x080b),\n\n /* Legacy algorithms */\n rsa_pkcs1_sha1(0x0201),\n ecdsa_sha1(0x0203),\n\n /* Reserved Code Points */\n private_use(0xFE00..0xFFFF),\n (0xFFFF)\n } SignatureScheme;\n\n struct {\n SignatureScheme supported_signature_algorithms<2..2^16-2>;\n } SignatureSchemeList;\n\n Note: This enum is named \"SignatureScheme\" because there is already a\n \"SignatureAlgorithm\" type in TLS 1.2, which this replaces. We use\n the term \"signature algorithm\" throughout the text.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 42]\n\f\nRFC 8446 TLS August 2018\n\n\n Each SignatureScheme value lists a single signature algorithm that\n the client is willing to verify. The values are indicated in\n descending order of preference. Note that a signature algorithm\n takes as input an arbitrary-length message, rather than a digest.\n Algorithms which traditionally act on a digest should be defined in\n TLS to first hash the input with a specified hash algorithm and then\n proceed as usual. The code point groups listed above have the\n following meanings:\n\n RSASSA-PKCS1-v1_5 algorithms: Indicates a signature algorithm using\n RSASSA-PKCS1-v1_5 [RFC8017] with the corresponding hash algorithm\n as defined in [SHS]. These values refer solely to signatures\n which appear in certificates (see Section 4.4.2.2) and are not\n defined for use in signed TLS handshake messages, although they\n MAY appear in \"signature_algorithms\" and\n \"signature_algorithms_cert\" for backward compatibility with\n TLS 1.2.\n\n ECDSA algorithms: Indicates a signature algorithm using ECDSA\n [ECDSA], the corresponding curve as defined in ANSI X9.62 [ECDSA]\n and FIPS 186-4 [DSS], and the corresponding hash algorithm as\n defined in [SHS]. The signature is represented as a DER-encoded\n [X690] ECDSA-Sig-Value structure.\n\n RSASSA-PSS RSAE algorithms: Indicates a signature algorithm using\n RSASSA-PSS [RFC8017] with mask generation function 1. The digest\n used in the mask generation function and the digest being signed\n are both the corresponding hash algorithm as defined in [SHS].\n The length of the Salt MUST be equal to the length of the output\n of the digest algorithm. If the public key is carried in an X.509\n certificate, it MUST use the rsaEncryption OID [RFC5280].\n\n EdDSA algorithms: Indicates a signature algorithm using EdDSA as\n defined in [RFC8032] or its successors. Note that these\n correspond to the \"PureEdDSA\" algorithms and not the \"prehash\"\n variants.\n\n RSASSA-PSS PSS algorithms: Indicates a signature algorithm using\n RSASSA-PSS [RFC8017] with mask generation function 1. The digest\n used in the mask generation function and the digest being signed\n are both the corresponding hash algorithm as defined in [SHS].\n The length of the Salt MUST be equal to the length of the digest\n algorithm. If the public key is carried in an X.509 certificate,\n it MUST use the RSASSA-PSS OID [RFC5756]. When used in\n certificate signatures, the algorithm parameters MUST be DER\n encoded. If the corresponding public key's parameters are\n present, then the parameters in the signature MUST be identical to\n those in the public key.\n\n\n\nRescorla Standards Track [Page 43]\n\f\nRFC 8446 TLS August 2018\n\n\n Legacy algorithms: Indicates algorithms which are being deprecated\n because they use algorithms with known weaknesses, specifically\n SHA-1 which is used in this context with either (1) RSA using\n RSASSA-PKCS1-v1_5 or (2) ECDSA. These values refer solely to\n signatures which appear in certificates (see Section 4.4.2.2) and\n are not defined for use in signed TLS handshake messages, although\n they MAY appear in \"signature_algorithms\" and\n \"signature_algorithms_cert\" for backward compatibility with\n TLS 1.2. Endpoints SHOULD NOT negotiate these algorithms but are\n permitted to do so solely for backward compatibility. Clients\n offering these values MUST list them as the lowest priority\n (listed after all other algorithms in SignatureSchemeList).\n TLS 1.3 servers MUST NOT offer a SHA-1 signed certificate unless\n no valid certificate chain can be produced without it (see\n Section 4.4.2.2).\n\n The signatures on certificates that are self-signed or certificates\n that are trust anchors are not validated, since they begin a\n certification path (see [RFC5280], Section 3.2). A certificate that\n begins a certification path MAY use a signature algorithm that is not\n advertised as being supported in the \"signature_algorithms\"\n extension.\n\n Note that TLS 1.2 defines this extension differently. TLS 1.3\n implementations willing to negotiate TLS 1.2 MUST behave in\n accordance with the requirements of [RFC5246] when negotiating that\n version. In particular:\n\n - TLS 1.2 ClientHellos MAY omit this extension.\n\n - In TLS 1.2, the extension contained hash/signature pairs. The\n pairs are encoded in two octets, so SignatureScheme values have\n been allocated to align with TLS 1.2's encoding. Some legacy\n pairs are left unallocated. These algorithms are deprecated as of\n TLS 1.3. They MUST NOT be offered or negotiated by any\n implementation. In particular, MD5 [SLOTH], SHA-224, and DSA\n MUST NOT be used.\n\n - ECDSA signature schemes align with TLS 1.2's ECDSA hash/signature\n pairs. However, the old semantics did not constrain the signing\n curve. If TLS 1.2 is negotiated, implementations MUST be prepared\n to accept a signature that uses any curve that they advertised in\n the \"supported_groups\" extension.\n\n - Implementations that advertise support for RSASSA-PSS (which is\n mandatory in TLS 1.3) MUST be prepared to accept a signature using\n that scheme even when TLS 1.2 is negotiated. In TLS 1.2,\n RSASSA-PSS is used with RSA cipher suites.\n\n\n\nRescorla Standards Track [Page 44]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.4. Certificate Authorities\n\n The \"certificate_authorities\" extension is used to indicate the\n certificate authorities (CAs) which an endpoint supports and which\n SHOULD be used by the receiving endpoint to guide certificate\n selection.\n\n The body of the \"certificate_authorities\" extension consists of a\n CertificateAuthoritiesExtension structure.\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n DistinguishedName authorities<3..2^16-1>;\n } CertificateAuthoritiesExtension;\n\n authorities: A list of the distinguished names [X501] of acceptable\n certificate authorities, represented in DER-encoded [X690] format.\n These distinguished names specify a desired distinguished name for\n a trust anchor or subordinate CA; thus, this message can be used\n to describe known trust anchors as well as a desired authorization\n space.\n\n The client MAY send the \"certificate_authorities\" extension in the\n ClientHello message. The server MAY send it in the\n CertificateRequest message.\n\n The \"trusted_ca_keys\" extension [RFC6066], which serves a similar\n purpose but is more complicated, is not used in TLS 1.3 (although it\n may appear in ClientHello messages from clients which are offering\n prior versions of TLS).\n\n4.2.5. OID Filters\n\n The \"oid_filters\" extension allows servers to provide a set of\n OID/value pairs which it would like the client's certificate to\n match. This extension, if provided by the server, MUST only be sent\n in the CertificateRequest message.\n\n struct {\n opaque certificate_extension_oid<1..2^8-1>;\n opaque certificate_extension_values<0..2^16-1>;\n } OIDFilter;\n\n struct {\n OIDFilter filters<0..2^16-1>;\n } OIDFilterExtension;\n\n\n\n\nRescorla Standards Track [Page 45]\n\f\nRFC 8446 TLS August 2018\n\n\n filters: A list of certificate extension OIDs [RFC5280] with their\n allowed value(s) and represented in DER-encoded [X690] format.\n Some certificate extension OIDs allow multiple values (e.g.,\n Extended Key Usage). If the server has included a non-empty\n filters list, the client certificate included in the response MUST\n contain all of the specified extension OIDs that the client\n recognizes. For each extension OID recognized by the client, all\n of the specified values MUST be present in the client certificate\n (but the certificate MAY have other values as well). However, the\n client MUST ignore and skip any unrecognized certificate extension\n OIDs. If the client ignored some of the required certificate\n extension OIDs and supplied a certificate that does not satisfy\n the request, the server MAY at its discretion either continue the\n connection without client authentication or abort the handshake\n with an \"unsupported_certificate\" alert. Any given OID MUST NOT\n appear more than once in the filters list.\n\n PKIX RFCs define a variety of certificate extension OIDs and their\n corresponding value types. Depending on the type, matching\n certificate extension values are not necessarily bitwise-equal. It\n is expected that TLS implementations will rely on their PKI libraries\n to perform certificate selection using certificate extension OIDs.\n\n This document defines matching rules for two standard certificate\n extensions defined in [RFC5280]:\n\n - The Key Usage extension in a certificate matches the request when\n all key usage bits asserted in the request are also asserted in\n the Key Usage certificate extension.\n\n - The Extended Key Usage extension in a certificate matches the\n request when all key purpose OIDs present in the request are also\n found in the Extended Key Usage certificate extension. The\n special anyExtendedKeyUsage OID MUST NOT be used in the request.\n\n Separate specifications may define matching rules for other\n certificate extensions.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 46]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.6. Post-Handshake Client Authentication\n\n The \"post_handshake_auth\" extension is used to indicate that a client\n is willing to perform post-handshake authentication (Section 4.6.2).\n Servers MUST NOT send a post-handshake CertificateRequest to clients\n which do not offer this extension. Servers MUST NOT send this\n extension.\n\n struct {} PostHandshakeAuth;\n\n The \"extension_data\" field of the \"post_handshake_auth\" extension is\n zero length.\n\n4.2.7. Supported Groups\n\n When sent by the client, the \"supported_groups\" extension indicates\n the named groups which the client supports for key exchange, ordered\n from most preferred to least preferred.\n\n Note: In versions of TLS prior to TLS 1.3, this extension was named\n \"elliptic_curves\" and only contained elliptic curve groups. See\n [RFC8422] and [RFC7919]. This extension was also used to negotiate\n ECDSA curves. Signature algorithms are now negotiated independently\n (see Section 4.2.3).\n\n The \"extension_data\" field of this extension contains a\n \"NamedGroupList\" value:\n\n enum {\n\n /* Elliptic Curve Groups (ECDHE) */\n secp256r1(0x0017), secp384r1(0x0018), secp521r1(0x0019),\n x25519(0x001D), x448(0x001E),\n\n /* Finite Field Groups (DHE) */\n ffdhe2048(0x0100), ffdhe3072(0x0101), ffdhe4096(0x0102),\n ffdhe6144(0x0103), ffdhe8192(0x0104),\n\n /* Reserved Code Points */\n ffdhe_private_use(0x01FC..0x01FF),\n ecdhe_private_use(0xFE00..0xFEFF),\n (0xFFFF)\n } NamedGroup;\n\n struct {\n NamedGroup named_group_list<2..2^16-1>;\n } NamedGroupList;\n\n\n\n\nRescorla Standards Track [Page 47]\n\f\nRFC 8446 TLS August 2018\n\n\n Elliptic Curve Groups (ECDHE): Indicates support for the\n corresponding named curve, defined in either FIPS 186-4 [DSS] or\n [RFC7748]. Values 0xFE00 through 0xFEFF are reserved for\n Private Use [RFC8126].\n\n Finite Field Groups (DHE): Indicates support for the corresponding\n finite field group, defined in [RFC7919]. Values 0x01FC through\n 0x01FF are reserved for Private Use.\n\n Items in named_group_list are ordered according to the sender's\n preferences (most preferred choice first).\n\n As of TLS 1.3, servers are permitted to send the \"supported_groups\"\n extension to the client. Clients MUST NOT act upon any information\n found in \"supported_groups\" prior to successful completion of the\n handshake but MAY use the information learned from a successfully\n completed handshake to change what groups they use in their\n \"key_share\" extension in subsequent connections. If the server has a\n group it prefers to the ones in the \"key_share\" extension but is\n still willing to accept the ClientHello, it SHOULD send\n \"supported_groups\" to update the client's view of its preferences;\n this extension SHOULD contain all groups the server supports,\n regardless of whether they are currently supported by the client.\n\n4.2.8. Key Share\n\n The \"key_share\" extension contains the endpoint's cryptographic\n parameters.\n\n Clients MAY send an empty client_shares vector in order to request\n group selection from the server, at the cost of an additional round\n trip (see Section 4.1.4).\n\n struct {\n NamedGroup group;\n opaque key_exchange<1..2^16-1>;\n } KeyShareEntry;\n\n group: The named group for the key being exchanged.\n\n key_exchange: Key exchange information. The contents of this field\n are determined by the specified group and its corresponding\n definition. Finite Field Diffie-Hellman [DH76] parameters are\n described in Section 4.2.8.1; Elliptic Curve Diffie-Hellman\n parameters are described in Section 4.2.8.2.\n\n\n\n\n\n\nRescorla Standards Track [Page 48]\n\f\nRFC 8446 TLS August 2018\n\n\n In the ClientHello message, the \"extension_data\" field of this\n extension contains a \"KeyShareClientHello\" value:\n\n struct {\n KeyShareEntry client_shares<0..2^16-1>;\n } KeyShareClientHello;\n\n client_shares: A list of offered KeyShareEntry values in descending\n order of client preference.\n\n This vector MAY be empty if the client is requesting a\n HelloRetryRequest. Each KeyShareEntry value MUST correspond to a\n group offered in the \"supported_groups\" extension and MUST appear in\n the same order. However, the values MAY be a non-contiguous subset\n of the \"supported_groups\" extension and MAY omit the most preferred\n groups. Such a situation could arise if the most preferred groups\n are new and unlikely to be supported in enough places to make\n pregenerating key shares for them efficient.\n\n Clients can offer as many KeyShareEntry values as the number of\n supported groups it is offering, each representing a single set of\n key exchange parameters. For instance, a client might offer shares\n for several elliptic curves or multiple FFDHE groups. The\n key_exchange values for each KeyShareEntry MUST be generated\n independently. Clients MUST NOT offer multiple KeyShareEntry values\n for the same group. Clients MUST NOT offer any KeyShareEntry values\n for groups not listed in the client's \"supported_groups\" extension.\n Servers MAY check for violations of these rules and abort the\n handshake with an \"illegal_parameter\" alert if one is violated.\n\n In a HelloRetryRequest message, the \"extension_data\" field of this\n extension contains a KeyShareHelloRetryRequest value:\n\n struct {\n NamedGroup selected_group;\n } KeyShareHelloRetryRequest;\n\n selected_group: The mutually supported group the server intends to\n negotiate and is requesting a retried ClientHello/KeyShare for.\n\n Upon receipt of this extension in a HelloRetryRequest, the client\n MUST verify that (1) the selected_group field corresponds to a group\n which was provided in the \"supported_groups\" extension in the\n original ClientHello and (2) the selected_group field does not\n correspond to a group which was provided in the \"key_share\" extension\n in the original ClientHello. If either of these checks fails, then\n the client MUST abort the handshake with an \"illegal_parameter\"\n alert. Otherwise, when sending the new ClientHello, the client MUST\n\n\n\nRescorla Standards Track [Page 49]\n\f\nRFC 8446 TLS August 2018\n\n\n replace the original \"key_share\" extension with one containing only a\n new KeyShareEntry for the group indicated in the selected_group field\n of the triggering HelloRetryRequest.\n\n In a ServerHello message, the \"extension_data\" field of this\n extension contains a KeyShareServerHello value:\n\n struct {\n KeyShareEntry server_share;\n } KeyShareServerHello;\n\n server_share: A single KeyShareEntry value that is in the same group\n as one of the client's shares.\n\n If using (EC)DHE key establishment, servers offer exactly one\n KeyShareEntry in the ServerHello. This value MUST be in the same\n group as the KeyShareEntry value offered by the client that the\n server has selected for the negotiated key exchange. Servers\n MUST NOT send a KeyShareEntry for any group not indicated in the\n client's \"supported_groups\" extension and MUST NOT send a\n KeyShareEntry when using the \"psk_ke\" PskKeyExchangeMode. If using\n (EC)DHE key establishment and a HelloRetryRequest containing a\n \"key_share\" extension was received by the client, the client MUST\n verify that the selected NamedGroup in the ServerHello is the same as\n that in the HelloRetryRequest. If this check fails, the client MUST\n abort the handshake with an \"illegal_parameter\" alert.\n\n4.2.8.1. Diffie-Hellman Parameters\n\n Diffie-Hellman [DH76] parameters for both clients and servers are\n encoded in the opaque key_exchange field of a KeyShareEntry in a\n KeyShare structure. The opaque value contains the Diffie-Hellman\n public value (Y = g^X mod p) for the specified group (see [RFC7919]\n for group definitions) encoded as a big-endian integer and padded to\n the left with zeros to the size of p in bytes.\n\n Note: For a given Diffie-Hellman group, the padding results in all\n public keys having the same length.\n\n Peers MUST validate each other's public key Y by ensuring that 1 < Y\n < p-1. This check ensures that the remote peer is properly behaved\n and isn't forcing the local system into a small subgroup.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 50]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.8.2. ECDHE Parameters\n\n ECDHE parameters for both clients and servers are encoded in the\n opaque key_exchange field of a KeyShareEntry in a KeyShare structure.\n\n For secp256r1, secp384r1, and secp521r1, the contents are the\n serialized value of the following struct:\n\n struct {\n uint8 legacy_form = 4;\n opaque X[coordinate_length];\n opaque Y[coordinate_length];\n } UncompressedPointRepresentation;\n\n X and Y, respectively, are the binary representations of the x and y\n values in network byte order. There are no internal length markers,\n so each number representation occupies as many octets as implied by\n the curve parameters. For P-256, this means that each of X and Y use\n 32 octets, padded on the left by zeros if necessary. For P-384, they\n take 48 octets each. For P-521, they take 66 octets each.\n\n For the curves secp256r1, secp384r1, and secp521r1, peers MUST\n validate each other's public value Q by ensuring that the point is a\n valid point on the elliptic curve. The appropriate validation\n procedures are defined in Section 4.3.7 of [ECDSA] and alternatively\n in Section 5.6.2.3 of [KEYAGREEMENT]. This process consists of three\n steps: (1) verify that Q is not the point at infinity (O), (2) verify\n that for Q = (x, y) both integers x and y are in the correct\n interval, and (3) ensure that (x, y) is a correct solution to the\n elliptic curve equation. For these curves, implementors do not need\n to verify membership in the correct subgroup.\n\n For X25519 and X448, the contents of the public value are the byte\n string inputs and outputs of the corresponding functions defined in\n [RFC7748]: 32 bytes for X25519 and 56 bytes for X448.\n\n Note: Versions of TLS prior to 1.3 permitted point format\n negotiation; TLS 1.3 removes this feature in favor of a single point\n format for each curve.\n\n4.2.9. Pre-Shared Key Exchange Modes\n\n In order to use PSKs, clients MUST also send a\n \"psk_key_exchange_modes\" extension. The semantics of this extension\n are that the client only supports the use of PSKs with these modes,\n which restricts both the use of PSKs offered in this ClientHello and\n those which the server might supply via NewSessionTicket.\n\n\n\n\nRescorla Standards Track [Page 51]\n\f\nRFC 8446 TLS August 2018\n\n\n A client MUST provide a \"psk_key_exchange_modes\" extension if it\n offers a \"pre_shared_key\" extension. If clients offer\n \"pre_shared_key\" without a \"psk_key_exchange_modes\" extension,\n servers MUST abort the handshake. Servers MUST NOT select a key\n exchange mode that is not listed by the client. This extension also\n restricts the modes for use with PSK resumption. Servers SHOULD NOT\n send NewSessionTicket with tickets that are not compatible with the\n advertised modes; however, if a server does so, the impact will just\n be that the client's attempts at resumption fail.\n\n The server MUST NOT send a \"psk_key_exchange_modes\" extension.\n\n enum { psk_ke(0), psk_dhe_ke(1), (255) } PskKeyExchangeMode;\n\n struct {\n PskKeyExchangeMode ke_modes<1..255>;\n } PskKeyExchangeModes;\n\n psk_ke: PSK-only key establishment. In this mode, the server\n MUST NOT supply a \"key_share\" value.\n\n psk_dhe_ke: PSK with (EC)DHE key establishment. In this mode, the\n client and server MUST supply \"key_share\" values as described in\n Section 4.2.8.\n\n Any future values that are allocated must ensure that the transmitted\n protocol messages unambiguously identify which mode was selected by\n the server; at present, this is indicated by the presence of the\n \"key_share\" in the ServerHello.\n\n4.2.10. Early Data Indication\n\n When a PSK is used and early data is allowed for that PSK, the client\n can send Application Data in its first flight of messages. If the\n client opts to do so, it MUST supply both the \"pre_shared_key\" and\n \"early_data\" extensions.\n\n The \"extension_data\" field of this extension contains an\n \"EarlyDataIndication\" value.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 52]\n\f\nRFC 8446 TLS August 2018\n\n\n struct {} Empty;\n\n struct {\n select (Handshake.msg_type) {\n case new_session_ticket: uint32 max_early_data_size;\n case client_hello: Empty;\n case encrypted_extensions: Empty;\n };\n } EarlyDataIndication;\n\n See Section 4.6.1 for details regarding the use of the\n max_early_data_size field.\n\n The parameters for the 0-RTT data (version, symmetric cipher suite,\n Application-Layer Protocol Negotiation (ALPN) [RFC7301] protocol,\n etc.) are those associated with the PSK in use. For externally\n provisioned PSKs, the associated values are those provisioned along\n with the key. For PSKs established via a NewSessionTicket message,\n the associated values are those which were negotiated in the\n connection which established the PSK. The PSK used to encrypt the\n early data MUST be the first PSK listed in the client's\n \"pre_shared_key\" extension.\n\n For PSKs provisioned via NewSessionTicket, a server MUST validate\n that the ticket age for the selected PSK identity (computed by\n subtracting ticket_age_add from PskIdentity.obfuscated_ticket_age\n modulo 2^32) is within a small tolerance of the time since the ticket\n was issued (see Section 8). If it is not, the server SHOULD proceed\n with the handshake but reject 0-RTT, and SHOULD NOT take any other\n action that assumes that this ClientHello is fresh.\n\n 0-RTT messages sent in the first flight have the same (encrypted)\n content types as messages of the same type sent in other flights\n (handshake and application_data) but are protected under different\n keys. After receiving the server's Finished message, if the server\n has accepted early data, an EndOfEarlyData message will be sent to\n indicate the key change. This message will be encrypted with the\n 0-RTT traffic keys.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 53]\n\f\nRFC 8446 TLS August 2018\n\n\n A server which receives an \"early_data\" extension MUST behave in one\n of three ways:\n\n - Ignore the extension and return a regular 1-RTT response. The\n server then skips past early data by attempting to deprotect\n received records using the handshake traffic key, discarding\n records which fail deprotection (up to the configured\n max_early_data_size). Once a record is deprotected successfully,\n it is treated as the start of the client's second flight and the\n server proceeds as with an ordinary 1-RTT handshake.\n\n - Request that the client send another ClientHello by responding\n with a HelloRetryRequest. A client MUST NOT include the\n \"early_data\" extension in its followup ClientHello. The server\n then ignores early data by skipping all records with an external\n content type of \"application_data\" (indicating that they are\n encrypted), up to the configured max_early_data_size.\n\n - Return its own \"early_data\" extension in EncryptedExtensions,\n indicating that it intends to process the early data. It is not\n possible for the server to accept only a subset of the early data\n messages. Even though the server sends a message accepting early\n data, the actual early data itself may already be in flight by the\n time the server generates this message.\n\n In order to accept early data, the server MUST have accepted a PSK\n cipher suite and selected the first key offered in the client's\n \"pre_shared_key\" extension. In addition, it MUST verify that the\n following values are the same as those associated with the\n selected PSK:\n\n - The TLS version number\n\n - The selected cipher suite\n\n - The selected ALPN [RFC7301] protocol, if any\n\n These requirements are a superset of those needed to perform a 1-RTT\n handshake using the PSK in question. For externally established\n PSKs, the associated values are those provisioned along with the key.\n For PSKs established via a NewSessionTicket message, the associated\n values are those negotiated in the connection during which the ticket\n was established.\n\n Future extensions MUST define their interaction with 0-RTT.\n\n\n\n\n\n\nRescorla Standards Track [Page 54]\n\f\nRFC 8446 TLS August 2018\n\n\n If any of these checks fail, the server MUST NOT respond with the\n extension and must discard all the first-flight data using one of the\n first two mechanisms listed above (thus falling back to 1-RTT or\n 2-RTT). If the client attempts a 0-RTT handshake but the server\n rejects it, the server will generally not have the 0-RTT record\n protection keys and must instead use trial decryption (either with\n the 1-RTT handshake keys or by looking for a cleartext ClientHello in\n the case of a HelloRetryRequest) to find the first non-0-RTT message.\n\n If the server chooses to accept the \"early_data\" extension, then it\n MUST comply with the same error-handling requirements specified for\n all records when processing early data records. Specifically, if the\n server fails to decrypt a 0-RTT record following an accepted\n \"early_data\" extension, it MUST terminate the connection with a\n \"bad_record_mac\" alert as per Section 5.2.\n\n If the server rejects the \"early_data\" extension, the client\n application MAY opt to retransmit the Application Data previously\n sent in early data once the handshake has been completed. Note that\n automatic retransmission of early data could result in incorrect\n assumptions regarding the status of the connection. For instance,\n when the negotiated connection selects a different ALPN protocol from\n what was used for the early data, an application might need to\n construct different messages. Similarly, if early data assumes\n anything about the connection state, it might be sent in error after\n the handshake completes.\n\n A TLS implementation SHOULD NOT automatically resend early data;\n applications are in a better position to decide when retransmission\n is appropriate. A TLS implementation MUST NOT automatically resend\n early data unless the negotiated connection selects the same ALPN\n protocol.\n\n4.2.11. Pre-Shared Key Extension\n\n The \"pre_shared_key\" extension is used to negotiate the identity of\n the pre-shared key to be used with a given handshake in association\n with PSK key establishment.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 55]\n\f\nRFC 8446 TLS August 2018\n\n\n The \"extension_data\" field of this extension contains a\n \"PreSharedKeyExtension\" value:\n\n struct {\n opaque identity<1..2^16-1>;\n uint32 obfuscated_ticket_age;\n } PskIdentity;\n\n opaque PskBinderEntry<32..255>;\n\n struct {\n PskIdentity identities<7..2^16-1>;\n PskBinderEntry binders<33..2^16-1>;\n } OfferedPsks;\n\n struct {\n select (Handshake.msg_type) {\n case client_hello: OfferedPsks;\n case server_hello: uint16 selected_identity;\n };\n } PreSharedKeyExtension;\n\n identity: A label for a key. For instance, a ticket (as defined in\n Appendix B.3.4) or a label for a pre-shared key established\n externally.\n\n obfuscated_ticket_age: An obfuscated version of the age of the key.\n Section 4.2.11.1 describes how to form this value for identities\n established via the NewSessionTicket message. For identities\n established externally, an obfuscated_ticket_age of 0 SHOULD be\n used, and servers MUST ignore the value.\n\n identities: A list of the identities that the client is willing to\n negotiate with the server. If sent alongside the \"early_data\"\n extension (see Section 4.2.10), the first identity is the one used\n for 0-RTT data.\n\n binders: A series of HMAC values, one for each value in the\n identities list and in the same order, computed as described\n below.\n\n selected_identity: The server's chosen identity expressed as a\n (0-based) index into the identities in the client's list.\n\n Each PSK is associated with a single Hash algorithm. For PSKs\n established via the ticket mechanism (Section 4.6.1), this is the KDF\n Hash algorithm on the connection where the ticket was established.\n For externally established PSKs, the Hash algorithm MUST be set when\n\n\n\nRescorla Standards Track [Page 56]\n\f\nRFC 8446 TLS August 2018\n\n\n the PSK is established or default to SHA-256 if no such algorithm is\n defined. The server MUST ensure that it selects a compatible PSK\n (if any) and cipher suite.\n\n In TLS versions prior to TLS 1.3, the Server Name Identification\n (SNI) value was intended to be associated with the session (Section 3\n of [RFC6066]), with the server being required to enforce that the SNI\n value associated with the session matches the one specified in the\n resumption handshake. However, in reality the implementations were\n not consistent on which of two supplied SNI values they would use,\n leading to the consistency requirement being de facto enforced by the\n clients. In TLS 1.3, the SNI value is always explicitly specified in\n the resumption handshake, and there is no need for the server to\n associate an SNI value with the ticket. Clients, however, SHOULD\n store the SNI with the PSK to fulfill the requirements of\n Section 4.6.1.\n\n Implementor's note: When session resumption is the primary use case\n of PSKs, the most straightforward way to implement the PSK/cipher\n suite matching requirements is to negotiate the cipher suite first\n and then exclude any incompatible PSKs. Any unknown PSKs (e.g., ones\n not in the PSK database or encrypted with an unknown key) SHOULD\n simply be ignored. If no acceptable PSKs are found, the server\n SHOULD perform a non-PSK handshake if possible. If backward\n compatibility is important, client-provided, externally established\n PSKs SHOULD influence cipher suite selection.\n\n Prior to accepting PSK key establishment, the server MUST validate\n the corresponding binder value (see Section 4.2.11.2 below). If this\n value is not present or does not validate, the server MUST abort the\n handshake. Servers SHOULD NOT attempt to validate multiple binders;\n rather, they SHOULD select a single PSK and validate solely the\n binder that corresponds to that PSK. See Section 8.2 and\n Appendix E.6 for the security rationale for this requirement. In\n order to accept PSK key establishment, the server sends a\n \"pre_shared_key\" extension indicating the selected identity.\n\n Clients MUST verify that the server's selected_identity is within the\n range supplied by the client, that the server selected a cipher suite\n indicating a Hash associated with the PSK, and that a server\n \"key_share\" extension is present if required by the ClientHello\n \"psk_key_exchange_modes\" extension. If these values are not\n consistent, the client MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 57]\n\f\nRFC 8446 TLS August 2018\n\n\n If the server supplies an \"early_data\" extension, the client MUST\n verify that the server's selected_identity is 0. If any other value\n is returned, the client MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n The \"pre_shared_key\" extension MUST be the last extension in the\n ClientHello (this facilitates implementation as described below).\n Servers MUST check that it is the last extension and otherwise fail\n the handshake with an \"illegal_parameter\" alert.\n\n4.2.11.1. Ticket Age\n\n The client's view of the age of a ticket is the time since the\n receipt of the NewSessionTicket message. Clients MUST NOT attempt to\n use tickets which have ages greater than the \"ticket_lifetime\" value\n which was provided with the ticket. The \"obfuscated_ticket_age\"\n field of each PskIdentity contains an obfuscated version of the\n ticket age formed by taking the age in milliseconds and adding the\n \"ticket_age_add\" value that was included with the ticket (see\n Section 4.6.1), modulo 2^32. This addition prevents passive\n observers from correlating connections unless tickets are reused.\n Note that the \"ticket_lifetime\" field in the NewSessionTicket message\n is in seconds but the \"obfuscated_ticket_age\" is in milliseconds.\n Because ticket lifetimes are restricted to a week, 32 bits is enough\n to represent any plausible age, even in milliseconds.\n\n4.2.11.2. PSK Binder\n\n The PSK binder value forms a binding between a PSK and the current\n handshake, as well as a binding between the handshake in which the\n PSK was generated (if via a NewSessionTicket message) and the current\n handshake. Each entry in the binders list is computed as an HMAC\n over a transcript hash (see Section 4.4.1) containing a partial\n ClientHello up to and including the PreSharedKeyExtension.identities\n field. That is, it includes all of the ClientHello but not the\n binders list itself. The length fields for the message (including\n the overall length, the length of the extensions block, and the\n length of the \"pre_shared_key\" extension) are all set as if binders\n of the correct lengths were present.\n\n The PskBinderEntry is computed in the same way as the Finished\n message (Section 4.4.4) but with the BaseKey being the binder_key\n derived via the key schedule from the corresponding PSK which is\n being offered (see Section 7.1).\n\n\n\n\n\n\n\nRescorla Standards Track [Page 58]\n\f\nRFC 8446 TLS August 2018\n\n\n If the handshake includes a HelloRetryRequest, the initial\n ClientHello and HelloRetryRequest are included in the transcript\n along with the new ClientHello. For instance, if the client sends\n ClientHello1, its binder will be computed over:\n\n Transcript-Hash(Truncate(ClientHello1))\n\n Where Truncate() removes the binders list from the ClientHello.\n\n If the server responds with a HelloRetryRequest and the client then\n sends ClientHello2, its binder will be computed over:\n\n Transcript-Hash(ClientHello1,\n HelloRetryRequest,\n Truncate(ClientHello2))\n\n The full ClientHello1/ClientHello2 is included in all other handshake\n hash computations. Note that in the first flight,\n Truncate(ClientHello1) is hashed directly, but in the second flight,\n ClientHello1 is hashed and then reinjected as a \"message_hash\"\n message, as described in Section 4.4.1.\n\n4.2.11.3. Processing Order\n\n Clients are permitted to \"stream\" 0-RTT data until they receive the\n server's Finished, only then sending the EndOfEarlyData message,\n followed by the rest of the handshake. In order to avoid deadlocks,\n when accepting \"early_data\", servers MUST process the client's\n ClientHello and then immediately send their flight of messages,\n rather than waiting for the client's EndOfEarlyData message before\n sending its ServerHello.\n\n4.3. Server Parameters\n\n The next two messages from the server, EncryptedExtensions and\n CertificateRequest, contain information from the server that\n determines the rest of the handshake. These messages are encrypted\n with keys derived from the server_handshake_traffic_secret.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 59]\n\f\nRFC 8446 TLS August 2018\n\n\n4.3.1. Encrypted Extensions\n\n In all handshakes, the server MUST send the EncryptedExtensions\n message immediately after the ServerHello message. This is the first\n message that is encrypted under keys derived from the\n server_handshake_traffic_secret.\n\n The EncryptedExtensions message contains extensions that can be\n protected, i.e., any which are not needed to establish the\n cryptographic context but which are not associated with individual\n certificates. The client MUST check EncryptedExtensions for the\n presence of any forbidden extensions and if any are found MUST abort\n the handshake with an \"illegal_parameter\" alert.\n\n Structure of this message:\n\n struct {\n Extension extensions<0..2^16-1>;\n } EncryptedExtensions;\n\n extensions: A list of extensions. For more information, see the\n table in Section 4.2.\n\n4.3.2. Certificate Request\n\n A server which is authenticating with a certificate MAY optionally\n request a certificate from the client. This message, if sent, MUST\n follow EncryptedExtensions.\n\n Structure of this message:\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n Extension extensions<2..2^16-1>;\n } CertificateRequest;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 60]\n\f\nRFC 8446 TLS August 2018\n\n\n certificate_request_context: An opaque string which identifies the\n certificate request and which will be echoed in the client's\n Certificate message. The certificate_request_context MUST be\n unique within the scope of this connection (thus preventing replay\n of client CertificateVerify messages). This field SHALL be zero\n length unless used for the post-handshake authentication exchanges\n described in Section 4.6.2. When requesting post-handshake\n authentication, the server SHOULD make the context unpredictable\n to the client (e.g., by randomly generating it) in order to\n prevent an attacker who has temporary access to the client's\n private key from pre-computing valid CertificateVerify messages.\n\n extensions: A set of extensions describing the parameters of the\n certificate being requested. The \"signature_algorithms\" extension\n MUST be specified, and other extensions may optionally be included\n if defined for this message. Clients MUST ignore unrecognized\n extensions.\n\n In prior versions of TLS, the CertificateRequest message carried a\n list of signature algorithms and certificate authorities which the\n server would accept. In TLS 1.3, the former is expressed by sending\n the \"signature_algorithms\" and optionally \"signature_algorithms_cert\"\n extensions. The latter is expressed by sending the\n \"certificate_authorities\" extension (see Section 4.2.4).\n\n Servers which are authenticating with a PSK MUST NOT send the\n CertificateRequest message in the main handshake, though they MAY\n send it in post-handshake authentication (see Section 4.6.2) provided\n that the client has sent the \"post_handshake_auth\" extension (see\n Section 4.2.6).\n\n4.4. Authentication Messages\n\n As discussed in Section 2, TLS generally uses a common set of\n messages for authentication, key confirmation, and handshake\n integrity: Certificate, CertificateVerify, and Finished. (The PSK\n binders also perform key confirmation, in a similar fashion.) These\n three messages are always sent as the last messages in their\n handshake flight. The Certificate and CertificateVerify messages are\n only sent under certain circumstances, as defined below. The\n Finished message is always sent as part of the Authentication Block.\n These messages are encrypted under keys derived from the\n [sender]_handshake_traffic_secret.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 61]\n\f\nRFC 8446 TLS August 2018\n\n\n The computations for the Authentication messages all uniformly take\n the following inputs:\n\n - The certificate and signing key to be used.\n\n - A Handshake Context consisting of the set of messages to be\n included in the transcript hash.\n\n - A Base Key to be used to compute a MAC key.\n\n Based on these inputs, the messages then contain:\n\n Certificate: The certificate to be used for authentication, and any\n supporting certificates in the chain. Note that certificate-based\n client authentication is not available in PSK handshake flows\n (including 0-RTT).\n\n CertificateVerify: A signature over the value\n Transcript-Hash(Handshake Context, Certificate).\n\n Finished: A MAC over the value Transcript-Hash(Handshake Context,\n Certificate, CertificateVerify) using a MAC key derived from the\n Base Key.\n\n The following table defines the Handshake Context and MAC Base Key\n for each scenario:\n\n +-----------+-------------------------+-----------------------------+\n | Mode | Handshake Context | Base Key |\n +-----------+-------------------------+-----------------------------+\n | Server | ClientHello ... later | server_handshake_traffic_ |\n | | of EncryptedExtensions/ | secret |\n | | CertificateRequest | |\n | | | |\n | Client | ClientHello ... later | client_handshake_traffic_ |\n | | of server | secret |\n | | Finished/EndOfEarlyData | |\n | | | |\n | Post- | ClientHello ... client | client_application_traffic_ |\n | Handshake | Finished + | secret_N |\n | | CertificateRequest | |\n +-----------+-------------------------+-----------------------------+\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 62]\n\f\nRFC 8446 TLS August 2018\n\n\n4.4.1. The Transcript Hash\n\n Many of the cryptographic computations in TLS make use of a\n transcript hash. This value is computed by hashing the concatenation\n of each included handshake message, including the handshake message\n header carrying the handshake message type and length fields, but not\n including record layer headers. I.e.,\n\n Transcript-Hash(M1, M2, ... Mn) = Hash(M1 || M2 || ... || Mn)\n\n As an exception to this general rule, when the server responds to a\n ClientHello with a HelloRetryRequest, the value of ClientHello1 is\n replaced with a special synthetic handshake message of handshake type\n \"message_hash\" containing Hash(ClientHello1). I.e.,\n\n Transcript-Hash(ClientHello1, HelloRetryRequest, ... Mn) =\n Hash(message_hash || /* Handshake type */\n 00 00 Hash.length || /* Handshake message length (bytes) */\n Hash(ClientHello1) || /* Hash of ClientHello1 */\n HelloRetryRequest || ... || Mn)\n\n The reason for this construction is to allow the server to do a\n stateless HelloRetryRequest by storing just the hash of ClientHello1\n in the cookie, rather than requiring it to export the entire\n intermediate hash state (see Section 4.2.2).\n\n For concreteness, the transcript hash is always taken from the\n following sequence of handshake messages, starting at the first\n ClientHello and including only those messages that were sent:\n ClientHello, HelloRetryRequest, ClientHello, ServerHello,\n EncryptedExtensions, server CertificateRequest, server Certificate,\n server CertificateVerify, server Finished, EndOfEarlyData, client\n Certificate, client CertificateVerify, client Finished.\n\n In general, implementations can implement the transcript by keeping a\n running transcript hash value based on the negotiated hash. Note,\n however, that subsequent post-handshake authentications do not\n include each other, just the messages through the end of the main\n handshake.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 63]\n\f\nRFC 8446 TLS August 2018\n\n\n4.4.2. Certificate\n\n This message conveys the endpoint's certificate chain to the peer.\n\n The server MUST send a Certificate message whenever the agreed-upon\n key exchange method uses certificates for authentication (this\n includes all key exchange methods defined in this document\n except PSK).\n\n The client MUST send a Certificate message if and only if the server\n has requested client authentication via a CertificateRequest message\n (Section 4.3.2). If the server requests client authentication but no\n suitable certificate is available, the client MUST send a Certificate\n message containing no certificates (i.e., with the \"certificate_list\"\n field having length 0). A Finished message MUST be sent regardless\n of whether the Certificate message is empty.\n\n Structure of this message:\n\n enum {\n X509(0),\n RawPublicKey(2),\n (255)\n } CertificateType;\n\n struct {\n select (certificate_type) {\n case RawPublicKey:\n /* From RFC 7250 ASN.1_subjectPublicKeyInfo */\n opaque ASN1_subjectPublicKeyInfo<1..2^24-1>;\n\n case X509:\n opaque cert_data<1..2^24-1>;\n };\n Extension extensions<0..2^16-1>;\n } CertificateEntry;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n CertificateEntry certificate_list<0..2^24-1>;\n } Certificate;\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 64]\n\f\nRFC 8446 TLS August 2018\n\n\n certificate_request_context: If this message is in response to a\n CertificateRequest, the value of certificate_request_context in\n that message. Otherwise (in the case of server authentication),\n this field SHALL be zero length.\n\n certificate_list: A sequence (chain) of CertificateEntry structures,\n each containing a single certificate and set of extensions.\n\n extensions: A set of extension values for the CertificateEntry. The\n \"Extension\" format is defined in Section 4.2. Valid extensions\n for server certificates at present include the OCSP Status\n extension [RFC6066] and the SignedCertificateTimestamp extension\n [RFC6962]; future extensions may be defined for this message as\n well. Extensions in the Certificate message from the server MUST\n correspond to ones from the ClientHello message. Extensions in\n the Certificate message from the client MUST correspond to\n extensions in the CertificateRequest message from the server. If\n an extension applies to the entire chain, it SHOULD be included in\n the first CertificateEntry.\n\n If the corresponding certificate type extension\n (\"server_certificate_type\" or \"client_certificate_type\") was not\n negotiated in EncryptedExtensions, or the X.509 certificate type was\n negotiated, then each CertificateEntry contains a DER-encoded X.509\n certificate. The sender's certificate MUST come in the first\n CertificateEntry in the list. Each following certificate SHOULD\n directly certify the one immediately preceding it. Because\n certificate validation requires that trust anchors be distributed\n independently, a certificate that specifies a trust anchor MAY be\n omitted from the chain, provided that supported peers are known to\n possess any omitted certificates.\n\n Note: Prior to TLS 1.3, \"certificate_list\" ordering required each\n certificate to certify the one immediately preceding it; however,\n some implementations allowed some flexibility. Servers sometimes\n send both a current and deprecated intermediate for transitional\n purposes, and others are simply configured incorrectly, but these\n cases can nonetheless be validated properly. For maximum\n compatibility, all implementations SHOULD be prepared to handle\n potentially extraneous certificates and arbitrary orderings from any\n TLS version, with the exception of the end-entity certificate which\n MUST be first.\n\n If the RawPublicKey certificate type was negotiated, then the\n certificate_list MUST contain no more than one CertificateEntry,\n which contains an ASN1_subjectPublicKeyInfo value as defined in\n [RFC7250], Section 3.\n\n\n\n\nRescorla Standards Track [Page 65]\n\f\nRFC 8446 TLS August 2018\n\n\n The OpenPGP certificate type [RFC6091] MUST NOT be used with TLS 1.3.\n\n The server's certificate_list MUST always be non-empty. A client\n will send an empty certificate_list if it does not have an\n appropriate certificate to send in response to the server's\n authentication request.\n\n4.4.2.1. OCSP Status and SCT Extensions\n\n [RFC6066] and [RFC6961] provide extensions to negotiate the server\n sending OCSP responses to the client. In TLS 1.2 and below, the\n server replies with an empty extension to indicate negotiation of\n this extension and the OCSP information is carried in a\n CertificateStatus message. In TLS 1.3, the server's OCSP information\n is carried in an extension in the CertificateEntry containing the\n associated certificate. Specifically, the body of the\n \"status_request\" extension from the server MUST be a\n CertificateStatus structure as defined in [RFC6066], which is\n interpreted as defined in [RFC6960].\n\n Note: The status_request_v2 extension [RFC6961] is deprecated.\n TLS 1.3 servers MUST NOT act upon its presence or information in it\n when processing ClientHello messages; in particular, they MUST NOT\n send the status_request_v2 extension in the EncryptedExtensions,\n CertificateRequest, or Certificate messages. TLS 1.3 servers MUST be\n able to process ClientHello messages that include it, as it MAY be\n sent by clients that wish to use it in earlier protocol versions.\n\n A server MAY request that a client present an OCSP response with its\n certificate by sending an empty \"status_request\" extension in its\n CertificateRequest message. If the client opts to send an OCSP\n response, the body of its \"status_request\" extension MUST be a\n CertificateStatus structure as defined in [RFC6066].\n\n Similarly, [RFC6962] provides a mechanism for a server to send a\n Signed Certificate Timestamp (SCT) as an extension in the ServerHello\n in TLS 1.2 and below. In TLS 1.3, the server's SCT information is\n carried in an extension in the CertificateEntry.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 66]\n\f\nRFC 8446 TLS August 2018\n\n\n4.4.2.2. Server Certificate Selection\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3 [RFC5280], unless explicitly\n negotiated otherwise (e.g., [RFC7250]).\n\n - The server's end-entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected authentication\n algorithm from the client's \"signature_algorithms\" extension\n (currently RSA, ECDSA, or EdDSA).\n\n - The certificate MUST allow the key to be used for signing (i.e.,\n the digitalSignature bit MUST be set if the Key Usage extension is\n present) with a signature scheme indicated in the client's\n \"signature_algorithms\"/\"signature_algorithms_cert\" extensions (see\n Section 4.2.3).\n\n - The \"server_name\" [RFC6066] and \"certificate_authorities\"\n extensions are used to guide certificate selection. As servers\n MAY require the presence of the \"server_name\" extension, clients\n SHOULD send this extension, when applicable.\n\n All certificates provided by the server MUST be signed by a signature\n algorithm advertised by the client if it is able to provide such a\n chain (see Section 4.2.3). Certificates that are self-signed or\n certificates that are expected to be trust anchors are not validated\n as part of the chain and therefore MAY be signed with any algorithm.\n\n If the server cannot produce a certificate chain that is signed only\n via the indicated supported algorithms, then it SHOULD continue the\n handshake by sending the client a certificate chain of its choice\n that may include algorithms that are not known to be supported by the\n client. This fallback chain SHOULD NOT use the deprecated SHA-1 hash\n algorithm in general, but MAY do so if the client's advertisement\n permits it, and MUST NOT do so otherwise.\n\n If the client cannot construct an acceptable chain using the provided\n certificates and decides to abort the handshake, then it MUST abort\n the handshake with an appropriate certificate-related alert (by\n default, \"unsupported_certificate\"; see Section 6.2 for more\n information).\n\n If the server has multiple certificates, it chooses one of them based\n on the above-mentioned criteria (in addition to other criteria, such\n as transport-layer endpoint, local configuration, and preferences).\n\n\n\n\n\nRescorla Standards Track [Page 67]\n\f\nRFC 8446 TLS August 2018\n\n\n4.4.2.3. Client Certificate Selection\n\n The following rules apply to certificates sent by the client:\n\n - The certificate type MUST be X.509v3 [RFC5280], unless explicitly\n negotiated otherwise (e.g., [RFC7250]).\n\n - If the \"certificate_authorities\" extension in the\n CertificateRequest message was present, at least one of the\n certificates in the certificate chain SHOULD be issued by one of\n the listed CAs.\n\n - The certificates MUST be signed using an acceptable signature\n algorithm, as described in Section 4.3.2. Note that this relaxes\n the constraints on certificate-signing algorithms found in prior\n versions of TLS.\n\n - If the CertificateRequest message contained a non-empty\n \"oid_filters\" extension, the end-entity certificate MUST match the\n extension OIDs that are recognized by the client, as described in\n Section 4.2.5.\n\n4.4.2.4. Receiving a Certificate Message\n\n In general, detailed certificate validation procedures are out of\n scope for TLS (see [RFC5280]). This section provides TLS-specific\n requirements.\n\n If the server supplies an empty Certificate message, the client MUST\n abort the handshake with a \"decode_error\" alert.\n\n If the client does not send any certificates (i.e., it sends an empty\n Certificate message), the server MAY at its discretion either\n continue the handshake without client authentication or abort the\n handshake with a \"certificate_required\" alert. Also, if some aspect\n of the certificate chain was unacceptable (e.g., it was not signed by\n a known, trusted CA), the server MAY at its discretion either\n continue the handshake (considering the client unauthenticated) or\n abort the handshake.\n\n Any endpoint receiving any certificate which it would need to\n validate using any signature algorithm using an MD5 hash MUST abort\n the handshake with a \"bad_certificate\" alert. SHA-1 is deprecated,\n and it is RECOMMENDED that any endpoint receiving any certificate\n which it would need to validate using any signature algorithm using a\n SHA-1 hash abort the handshake with a \"bad_certificate\" alert. For\n clarity, this means that endpoints can accept these algorithms for\n certificates that are self-signed or are trust anchors.\n\n\n\nRescorla Standards Track [Page 68]\n\f\nRFC 8446 TLS August 2018\n\n\n All endpoints are RECOMMENDED to transition to SHA-256 or better as\n soon as possible to maintain interoperability with implementations\n currently in the process of phasing out SHA-1 support.\n\n Note that a certificate containing a key for one signature algorithm\n MAY be signed using a different signature algorithm (for instance, an\n RSA key signed with an ECDSA key).\n\n4.4.3. Certificate Verify\n\n This message is used to provide explicit proof that an endpoint\n possesses the private key corresponding to its certificate. The\n CertificateVerify message also provides integrity for the handshake\n up to this point. Servers MUST send this message when authenticating\n via a certificate. Clients MUST send this message whenever\n authenticating via a certificate (i.e., when the Certificate message\n is non-empty). When sent, this message MUST appear immediately after\n the Certificate message and immediately prior to the Finished\n message.\n\n Structure of this message:\n\n struct {\n SignatureScheme algorithm;\n opaque signature<0..2^16-1>;\n } CertificateVerify;\n\n The algorithm field specifies the signature algorithm used (see\n Section 4.2.3 for the definition of this type). The signature is a\n digital signature using that algorithm. The content that is covered\n under the signature is the hash output as described in Section 4.4.1,\n namely:\n\n Transcript-Hash(Handshake Context, Certificate)\n\n The digital signature is then computed over the concatenation of:\n\n - A string that consists of octet 32 (0x20) repeated 64 times\n\n - The context string\n\n - A single 0 byte which serves as the separator\n\n - The content to be signed\n\n\n\n\n\n\n\nRescorla Standards Track [Page 69]\n\f\nRFC 8446 TLS August 2018\n\n\n This structure is intended to prevent an attack on previous versions\n of TLS in which the ServerKeyExchange format meant that attackers\n could obtain a signature of a message with a chosen 32-byte prefix\n (ClientHello.random). The initial 64-byte pad clears that prefix\n along with the server-controlled ServerHello.random.\n\n The context string for a server signature is\n \"TLS 1.3, server CertificateVerify\". The context string for a\n client signature is \"TLS 1.3, client CertificateVerify\". It is\n used to provide separation between signatures made in different\n contexts, helping against potential cross-protocol attacks.\n\n For example, if the transcript hash was 32 bytes of 01 (this length\n would make sense for SHA-256), the content covered by the digital\n signature for a server CertificateVerify would be:\n\n 2020202020202020202020202020202020202020202020202020202020202020\n 2020202020202020202020202020202020202020202020202020202020202020\n 544c5320312e332c207365727665722043657274696669636174655665726966\n 79\n 00\n 0101010101010101010101010101010101010101010101010101010101010101\n\n On the sender side, the process for computing the signature field of\n the CertificateVerify message takes as input:\n\n - The content covered by the digital signature\n\n - The private signing key corresponding to the certificate sent in\n the previous message\n\n If the CertificateVerify message is sent by a server, the signature\n algorithm MUST be one offered in the client's \"signature_algorithms\"\n extension unless no valid certificate chain can be produced without\n unsupported algorithms (see Section 4.2.3).\n\n If sent by a client, the signature algorithm used in the signature\n MUST be one of those present in the supported_signature_algorithms\n field of the \"signature_algorithms\" extension in the\n CertificateRequest message.\n\n In addition, the signature algorithm MUST be compatible with the key\n in the sender's end-entity certificate. RSA signatures MUST use an\n RSASSA-PSS algorithm, regardless of whether RSASSA-PKCS1-v1_5\n algorithms appear in \"signature_algorithms\". The SHA-1 algorithm\n MUST NOT be used in any signatures of CertificateVerify messages.\n\n\n\n\n\nRescorla Standards Track [Page 70]\n\f\nRFC 8446 TLS August 2018\n\n\n All SHA-1 signature algorithms in this specification are defined\n solely for use in legacy certificates and are not valid for\n CertificateVerify signatures.\n\n The receiver of a CertificateVerify message MUST verify the signature\n field. The verification process takes as input:\n\n - The content covered by the digital signature\n\n - The public key contained in the end-entity certificate found in\n the associated Certificate message\n\n - The digital signature received in the signature field of the\n CertificateVerify message\n\n If the verification fails, the receiver MUST terminate the handshake\n with a \"decrypt_error\" alert.\n\n4.4.4. Finished\n\n The Finished message is the final message in the Authentication\n Block. It is essential for providing authentication of the handshake\n and of the computed keys.\n\n Recipients of Finished messages MUST verify that the contents are\n correct and if incorrect MUST terminate the connection with a\n \"decrypt_error\" alert.\n\n Once a side has sent its Finished message and has received and\n validated the Finished message from its peer, it may begin to send\n and receive Application Data over the connection. There are two\n settings in which it is permitted to send data prior to receiving the\n peer's Finished:\n\n 1. Clients sending 0-RTT data as described in Section 4.2.10.\n\n 2. Servers MAY send data after sending their first flight, but\n because the handshake is not yet complete, they have no assurance\n of either the peer's identity or its liveness (i.e., the\n ClientHello might have been replayed).\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 71]\n\f\nRFC 8446 TLS August 2018\n\n\n The key used to compute the Finished message is computed from the\n Base Key defined in Section 4.4 using HKDF (see Section 7.1).\n Specifically:\n\n finished_key =\n HKDF-Expand-Label(BaseKey, \"finished\", \"\", Hash.length)\n\n Structure of this message:\n\n struct {\n opaque verify_data[Hash.length];\n } Finished;\n\n The verify_data value is computed as follows:\n\n verify_data =\n HMAC(finished_key,\n Transcript-Hash(Handshake Context,\n Certificate*, CertificateVerify*))\n\n * Only included if present.\n\n HMAC [RFC2104] uses the Hash algorithm for the handshake. As noted\n above, the HMAC input can generally be implemented by a running hash,\n i.e., just the handshake hash at this point.\n\n In previous versions of TLS, the verify_data was always 12 octets\n long. In TLS 1.3, it is the size of the HMAC output for the Hash\n used for the handshake.\n\n Note: Alerts and any other non-handshake record types are not\n handshake messages and are not included in the hash computations.\n\n Any records following a Finished message MUST be encrypted under the\n appropriate application traffic key as described in Section 7.2. In\n particular, this includes any alerts sent by the server in response\n to client Certificate and CertificateVerify messages.\n\n4.5. End of Early Data\n\n struct {} EndOfEarlyData;\n\n If the server sent an \"early_data\" extension in EncryptedExtensions,\n the client MUST send an EndOfEarlyData message after receiving the\n server Finished. If the server does not send an \"early_data\"\n extension in EncryptedExtensions, then the client MUST NOT send an\n EndOfEarlyData message. This message indicates that all 0-RTT\n application_data messages, if any, have been transmitted and that the\n\n\n\nRescorla Standards Track [Page 72]\n\f\nRFC 8446 TLS August 2018\n\n\n following records are protected under handshake traffic keys.\n Servers MUST NOT send this message, and clients receiving it MUST\n terminate the connection with an \"unexpected_message\" alert. This\n message is encrypted under keys derived from the\n client_early_traffic_secret.\n\n4.6. Post-Handshake Messages\n\n TLS also allows other messages to be sent after the main handshake.\n These messages use a handshake content type and are encrypted under\n the appropriate application traffic key.\n\n4.6.1. New Session Ticket Message\n\n At any time after the server has received the client Finished\n message, it MAY send a NewSessionTicket message. This message\n creates a unique association between the ticket value and a secret\n PSK derived from the resumption master secret (see Section 7).\n\n The client MAY use this PSK for future handshakes by including the\n ticket value in the \"pre_shared_key\" extension in its ClientHello\n (Section 4.2.11). Servers MAY send multiple tickets on a single\n connection, either immediately after each other or after specific\n events (see Appendix C.4). For instance, the server might send a new\n ticket after post-handshake authentication in order to encapsulate\n the additional client authentication state. Multiple tickets are\n useful for clients for a variety of purposes, including:\n\n - Opening multiple parallel HTTP connections.\n\n - Performing connection racing across interfaces and address\n families via (for example) Happy Eyeballs [RFC8305] or related\n techniques.\n\n Any ticket MUST only be resumed with a cipher suite that has the same\n KDF hash algorithm as that used to establish the original connection.\n\n Clients MUST only resume if the new SNI value is valid for the server\n certificate presented in the original session and SHOULD only resume\n if the SNI value matches the one used in the original session. The\n latter is a performance optimization: normally, there is no reason to\n expect that different servers covered by a single certificate would\n be able to accept each other's tickets; hence, attempting resumption\n in that case would waste a single-use ticket. If such an indication\n is provided (externally or by any other means), clients MAY resume\n with a different SNI value.\n\n\n\n\n\nRescorla Standards Track [Page 73]\n\f\nRFC 8446 TLS August 2018\n\n\n On resumption, if reporting an SNI value to the calling application,\n implementations MUST use the value sent in the resumption ClientHello\n rather than the value sent in the previous session. Note that if a\n server implementation declines all PSK identities with different SNI\n values, these two values are always the same.\n\n Note: Although the resumption master secret depends on the client's\n second flight, a server which does not request client authentication\n MAY compute the remainder of the transcript independently and then\n send a NewSessionTicket immediately upon sending its Finished rather\n than waiting for the client Finished. This might be appropriate in\n cases where the client is expected to open multiple TLS connections\n in parallel and would benefit from the reduced overhead of a\n resumption handshake, for example.\n\n struct {\n uint32 ticket_lifetime;\n uint32 ticket_age_add;\n opaque ticket_nonce<0..255>;\n opaque ticket<1..2^16-1>;\n Extension extensions<0..2^16-2>;\n } NewSessionTicket;\n\n ticket_lifetime: Indicates the lifetime in seconds as a 32-bit\n unsigned integer in network byte order from the time of ticket\n issuance. Servers MUST NOT use any value greater than\n 604800 seconds (7 days). The value of zero indicates that the\n ticket should be discarded immediately. Clients MUST NOT cache\n tickets for longer than 7 days, regardless of the ticket_lifetime,\n and MAY delete tickets earlier based on local policy. A server\n MAY treat a ticket as valid for a shorter period of time than what\n is stated in the ticket_lifetime.\n\n ticket_age_add: A securely generated, random 32-bit value that is\n used to obscure the age of the ticket that the client includes in\n the \"pre_shared_key\" extension. The client-side ticket age is\n added to this value modulo 2^32 to obtain the value that is\n transmitted by the client. The server MUST generate a fresh value\n for each ticket it sends.\n\n ticket_nonce: A per-ticket value that is unique across all tickets\n issued on this connection.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 74]\n\f\nRFC 8446 TLS August 2018\n\n\n ticket: The value of the ticket to be used as the PSK identity. The\n ticket itself is an opaque label. It MAY be either a database\n lookup key or a self-encrypted and self-authenticated value.\n\n extensions: A set of extension values for the ticket. The\n \"Extension\" format is defined in Section 4.2. Clients MUST ignore\n unrecognized extensions.\n\n The sole extension currently defined for NewSessionTicket is\n \"early_data\", indicating that the ticket may be used to send 0-RTT\n data (Section 4.2.10). It contains the following value:\n\n max_early_data_size: The maximum amount of 0-RTT data that the\n client is allowed to send when using this ticket, in bytes. Only\n Application Data payload (i.e., plaintext but not padding or the\n inner content type byte) is counted. A server receiving more than\n max_early_data_size bytes of 0-RTT data SHOULD terminate the\n connection with an \"unexpected_message\" alert. Note that servers\n that reject early data due to lack of cryptographic material will\n be unable to differentiate padding from content, so clients\n SHOULD NOT depend on being able to send large quantities of\n padding in early data records.\n\n The PSK associated with the ticket is computed as:\n\n HKDF-Expand-Label(resumption_master_secret,\n \"resumption\", ticket_nonce, Hash.length)\n\n Because the ticket_nonce value is distinct for each NewSessionTicket\n message, a different PSK will be derived for each ticket.\n\n Note that in principle it is possible to continue issuing new tickets\n which indefinitely extend the lifetime of the keying material\n originally derived from an initial non-PSK handshake (which was most\n likely tied to the peer's certificate). It is RECOMMENDED that\n implementations place limits on the total lifetime of such keying\n material; these limits should take into account the lifetime of the\n peer's certificate, the likelihood of intervening revocation, and the\n time since the peer's online CertificateVerify signature.\n\n4.6.2. Post-Handshake Authentication\n\n When the client has sent the \"post_handshake_auth\" extension (see\n Section 4.2.6), a server MAY request client authentication at any\n time after the handshake has completed by sending a\n CertificateRequest message. The client MUST respond with the\n appropriate Authentication messages (see Section 4.4). If the client\n chooses to authenticate, it MUST send Certificate, CertificateVerify,\n\n\n\nRescorla Standards Track [Page 75]\n\f\nRFC 8446 TLS August 2018\n\n\n and Finished. If it declines, it MUST send a Certificate message\n containing no certificates followed by Finished. All of the client's\n messages for a given response MUST appear consecutively on the wire\n with no intervening messages of other types.\n\n A client that receives a CertificateRequest message without having\n sent the \"post_handshake_auth\" extension MUST send an\n \"unexpected_message\" fatal alert.\n\n Note: Because client authentication could involve prompting the user,\n servers MUST be prepared for some delay, including receiving an\n arbitrary number of other messages between sending the\n CertificateRequest and receiving a response. In addition, clients\n which receive multiple CertificateRequests in close succession MAY\n respond to them in a different order than they were received (the\n certificate_request_context value allows the server to disambiguate\n the responses).\n\n4.6.3. Key and Initialization Vector Update\n\n The KeyUpdate handshake message is used to indicate that the sender\n is updating its sending cryptographic keys. This message can be sent\n by either peer after it has sent a Finished message. Implementations\n that receive a KeyUpdate message prior to receiving a Finished\n message MUST terminate the connection with an \"unexpected_message\"\n alert. After sending a KeyUpdate message, the sender SHALL send all\n its traffic using the next generation of keys, computed as described\n in Section 7.2. Upon receiving a KeyUpdate, the receiver MUST update\n its receiving keys.\n\n enum {\n update_not_requested(0), update_requested(1), (255)\n } KeyUpdateRequest;\n\n struct {\n KeyUpdateRequest request_update;\n } KeyUpdate;\n\n request_update: Indicates whether the recipient of the KeyUpdate\n should respond with its own KeyUpdate. If an implementation\n receives any other value, it MUST terminate the connection with an\n \"illegal_parameter\" alert.\n\n If the request_update field is set to \"update_requested\", then the\n receiver MUST send a KeyUpdate of its own with request_update set to\n \"update_not_requested\" prior to sending its next Application Data\n record. This mechanism allows either side to force an update to the\n entire connection, but causes an implementation which receives\n\n\n\nRescorla Standards Track [Page 76]\n\f\nRFC 8446 TLS August 2018\n\n\n multiple KeyUpdates while it is silent to respond with a single\n update. Note that implementations may receive an arbitrary number of\n messages between sending a KeyUpdate with request_update set to\n \"update_requested\" and receiving the peer's KeyUpdate, because those\n messages may already be in flight. However, because send and receive\n keys are derived from independent traffic secrets, retaining the\n receive traffic secret does not threaten the forward secrecy of data\n sent before the sender changed keys.\n\n If implementations independently send their own KeyUpdates with\n request_update set to \"update_requested\" and they cross in flight,\n then each side will also send a response, with the result that each\n side increments by two generations.\n\n Both sender and receiver MUST encrypt their KeyUpdate messages with\n the old keys. Additionally, both sides MUST enforce that a KeyUpdate\n with the old key is received before accepting any messages encrypted\n with the new key. Failure to do so may allow message truncation\n attacks.\n\n5. Record Protocol\n\n The TLS record protocol takes messages to be transmitted, fragments\n the data into manageable blocks, protects the records, and transmits\n the result. Received data is verified, decrypted, reassembled, and\n then delivered to higher-level clients.\n\n TLS records are typed, which allows multiple higher-level protocols\n to be multiplexed over the same record layer. This document\n specifies four content types: handshake, application_data, alert, and\n change_cipher_spec. The change_cipher_spec record is used only for\n compatibility purposes (see Appendix D.4).\n\n An implementation may receive an unencrypted record of type\n change_cipher_spec consisting of the single byte value 0x01 at any\n time after the first ClientHello message has been sent or received\n and before the peer's Finished message has been received and MUST\n simply drop it without further processing. Note that this record may\n appear at a point at the handshake where the implementation is\n expecting protected records, and so it is necessary to detect this\n condition prior to attempting to deprotect the record. An\n implementation which receives any other change_cipher_spec value or\n which receives a protected change_cipher_spec record MUST abort the\n handshake with an \"unexpected_message\" alert. If an implementation\n detects a change_cipher_spec record received before the first\n ClientHello message or after the peer's Finished message, it MUST be\n treated as an unexpected record type (though stateless servers may\n not be able to distinguish these cases from allowed cases).\n\n\n\nRescorla Standards Track [Page 77]\n\f\nRFC 8446 TLS August 2018\n\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST terminate\n the connection with an \"unexpected_message\" alert. New record\n content type values are assigned by IANA in the TLS ContentType\n registry as described in Section 11.\n\n5.1. Record Layer\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Message\n boundaries are handled differently depending on the underlying\n ContentType. Any future content types MUST specify appropriate\n rules. Note that these rules are stricter than what was enforced in\n TLS 1.2.\n\n Handshake messages MAY be coalesced into a single TLSPlaintext record\n or fragmented across several records, provided that:\n\n - Handshake messages MUST NOT be interleaved with other record\n types. That is, if a handshake message is split over two or more\n records, there MUST NOT be any other records between them.\n\n - Handshake messages MUST NOT span key changes. Implementations\n MUST verify that all messages immediately preceding a key change\n align with a record boundary; if not, then they MUST terminate the\n connection with an \"unexpected_message\" alert. Because the\n ClientHello, EndOfEarlyData, ServerHello, Finished, and KeyUpdate\n messages can immediately precede a key change, implementations\n MUST send these messages in alignment with a record boundary.\n\n Implementations MUST NOT send zero-length fragments of Handshake\n types, even if those fragments contain padding.\n\n Alert messages (Section 6) MUST NOT be fragmented across records, and\n multiple alert messages MUST NOT be coalesced into a single\n TLSPlaintext record. In other words, a record with an Alert type\n MUST contain exactly one message.\n\n Application Data messages contain data that is opaque to TLS.\n Application Data messages are always protected. Zero-length\n fragments of Application Data MAY be sent, as they are potentially\n useful as a traffic analysis countermeasure. Application Data\n fragments MAY be split across multiple records or coalesced into a\n single record.\n\n\n\n\n\n\nRescorla Standards Track [Page 78]\n\f\nRFC 8446 TLS August 2018\n\n\n enum {\n invalid(0),\n change_cipher_spec(20),\n alert(21),\n handshake(22),\n application_data(23),\n (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion legacy_record_version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type: The higher-level protocol used to process the enclosed\n fragment.\n\n legacy_record_version: MUST be set to 0x0303 for all records\n generated by a TLS 1.3 implementation other than an initial\n ClientHello (i.e., one not generated after a HelloRetryRequest),\n where it MAY also be 0x0301 for compatibility purposes. This\n field is deprecated and MUST be ignored for all purposes.\n Previous versions of TLS would use other values in this field\n under some circumstances.\n\n length: The length (in bytes) of the following\n TLSPlaintext.fragment. The length MUST NOT exceed 2^14 bytes. An\n endpoint that receives a record that exceeds this length MUST\n terminate the connection with a \"record_overflow\" alert.\n\n fragment: The data being transmitted. This value is transparent and\n is treated as an independent block to be dealt with by the higher-\n level protocol specified by the type field.\n\n This document describes TLS 1.3, which uses the version 0x0304. This\n version value is historical, deriving from the use of 0x0301 for\n TLS 1.0 and 0x0300 for SSL 3.0. In order to maximize backward\n compatibility, a record containing an initial ClientHello SHOULD have\n version 0x0301 (reflecting TLS 1.0) and a record containing a second\n ClientHello or a ServerHello MUST have version 0x0303 (reflecting\n TLS 1.2). When negotiating prior versions of TLS, endpoints follow\n the procedure and requirements provided in Appendix D.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 79]\n\f\nRFC 8446 TLS August 2018\n\n\n When record protection has not yet been engaged, TLSPlaintext\n structures are written directly onto the wire. Once record\n protection has started, TLSPlaintext records are protected and sent\n as described in the following section. Note that Application Data\n records MUST NOT be written to the wire unprotected (see Section 2\n for details).\n\n5.2. Record Payload Protection\n\n The record protection functions translate a TLSPlaintext structure\n into a TLSCiphertext structure. The deprotection functions reverse\n the process. In TLS 1.3, as opposed to previous versions of TLS, all\n ciphers are modeled as \"Authenticated Encryption with Associated\n Data\" (AEAD) [RFC5116]. AEAD functions provide a unified encryption\n and authentication operation which turns plaintext into authenticated\n ciphertext and back again. Each encrypted record consists of a\n plaintext header followed by an encrypted body, which itself contains\n a type and optional padding.\n\n struct {\n opaque content[TLSPlaintext.length];\n ContentType type;\n uint8 zeros[length_of_padding];\n } TLSInnerPlaintext;\n\n struct {\n ContentType opaque_type = application_data; /* 23 */\n ProtocolVersion legacy_record_version = 0x0303; /* TLS v1.2 */\n uint16 length;\n opaque encrypted_record[TLSCiphertext.length];\n } TLSCiphertext;\n\n content: The TLSPlaintext.fragment value, containing the byte\n encoding of a handshake or an alert message, or the raw bytes of\n the application's data to send.\n\n type: The TLSPlaintext.type value containing the content type of the\n record.\n\n zeros: An arbitrary-length run of zero-valued bytes may appear in\n the cleartext after the type field. This provides an opportunity\n for senders to pad any TLS record by a chosen amount as long as\n the total stays within record size limits. See Section 5.4 for\n more details.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 80]\n\f\nRFC 8446 TLS August 2018\n\n\n opaque_type: The outer opaque_type field of a TLSCiphertext record\n is always set to the value 23 (application_data) for outward\n compatibility with middleboxes accustomed to parsing previous\n versions of TLS. The actual content type of the record is found\n in TLSInnerPlaintext.type after decryption.\n\n legacy_record_version: The legacy_record_version field is always\n 0x0303. TLS 1.3 TLSCiphertexts are not generated until after\n TLS 1.3 has been negotiated, so there are no historical\n compatibility concerns where other values might be received. Note\n that the handshake protocol, including the ClientHello and\n ServerHello messages, authenticates the protocol version, so this\n value is redundant.\n\n length: The length (in bytes) of the following\n TLSCiphertext.encrypted_record, which is the sum of the lengths of\n the content and the padding, plus one for the inner content type,\n plus any expansion added by the AEAD algorithm. The length\n MUST NOT exceed 2^14 + 256 bytes. An endpoint that receives a\n record that exceeds this length MUST terminate the connection with\n a \"record_overflow\" alert.\n\n encrypted_record: The AEAD-encrypted form of the serialized\n TLSInnerPlaintext structure.\n\n AEAD algorithms take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [RFC5116]. The key is either the\n client_write_key or the server_write_key, the nonce is derived from\n the sequence number and the client_write_iv or server_write_iv (see\n Section 5.3), and the additional data input is the record header.\n\n I.e.,\n\n additional_data = TLSCiphertext.opaque_type ||\n TLSCiphertext.legacy_record_version ||\n TLSCiphertext.length\n\n The plaintext input to the AEAD algorithm is the encoded\n TLSInnerPlaintext structure. Derivation of traffic keys is defined\n in Section 7.3.\n\n The AEAD output consists of the ciphertext output from the AEAD\n encryption operation. The length of the plaintext is greater than\n the corresponding TLSPlaintext.length due to the inclusion of\n TLSInnerPlaintext.type and any padding supplied by the sender. The\n length of the AEAD output will generally be larger than the\n plaintext, but by an amount that varies with the AEAD algorithm.\n\n\n\nRescorla Standards Track [Page 81]\n\f\nRFC 8446 TLS August 2018\n\n\n Since the ciphers might incorporate padding, the amount of overhead\n could vary with different lengths of plaintext. Symbolically,\n\n AEADEncrypted =\n AEAD-Encrypt(write_key, nonce, additional_data, plaintext)\n\n The encrypted_record field of TLSCiphertext is set to AEADEncrypted.\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, additional data, and the AEADEncrypted value. The output is\n either the plaintext or an error indicating that the decryption\n failed. There is no separate integrity check. Symbolically,\n\n plaintext of encrypted_record =\n AEAD-Decrypt(peer_write_key, nonce,\n additional_data, AEADEncrypted)\n\n If the decryption fails, the receiver MUST terminate the connection\n with a \"bad_record_mac\" alert.\n\n An AEAD algorithm used in TLS 1.3 MUST NOT produce an expansion\n greater than 255 octets. An endpoint that receives a record from its\n peer with TLSCiphertext.length larger than 2^14 + 256 octets MUST\n terminate the connection with a \"record_overflow\" alert. This limit\n is derived from the maximum TLSInnerPlaintext length of 2^14 octets +\n 1 octet for ContentType + the maximum AEAD expansion of 255 octets.\n\n5.3. Per-Record Nonce\n\n A 64-bit sequence number is maintained separately for reading and\n writing records. The appropriate sequence number is incremented by\n one after reading or writing each record. Each sequence number is\n set to zero at the beginning of a connection and whenever the key is\n changed; the first record transmitted under a particular traffic key\n MUST use sequence number 0.\n\n Because the size of sequence numbers is 64-bit, they should not wrap.\n If a TLS implementation would need to wrap a sequence number, it MUST\n either rekey (Section 4.6.3) or terminate the connection.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 82]\n\f\nRFC 8446 TLS August 2018\n\n\n Each AEAD algorithm will specify a range of possible lengths for the\n per-record nonce, from N_MIN bytes to N_MAX bytes of input [RFC5116].\n The length of the TLS per-record nonce (iv_length) is set to the\n larger of 8 bytes and N_MIN for the AEAD algorithm (see [RFC5116],\n Section 4). An AEAD algorithm where N_MAX is less than 8 bytes\n MUST NOT be used with TLS. The per-record nonce for the AEAD\n construction is formed as follows:\n\n 1. The 64-bit record sequence number is encoded in network byte\n order and padded to the left with zeros to iv_length.\n\n 2. The padded sequence number is XORed with either the static\n client_write_iv or server_write_iv (depending on the role).\n\n The resulting quantity (of length iv_length) is used as the\n per-record nonce.\n\n Note: This is a different construction from that in TLS 1.2, which\n specified a partially explicit nonce.\n\n5.4. Record Padding\n\n All encrypted TLS records can be padded to inflate the size of the\n TLSCiphertext. This allows the sender to hide the size of the\n traffic from an observer.\n\n When generating a TLSCiphertext record, implementations MAY choose to\n pad. An unpadded record is just a record with a padding length of\n zero. Padding is a string of zero-valued bytes appended to the\n ContentType field before encryption. Implementations MUST set the\n padding octets to all zeros before encrypting.\n\n Application Data records may contain a zero-length\n TLSInnerPlaintext.content if the sender desires. This permits\n generation of plausibly sized cover traffic in contexts where the\n presence or absence of activity may be sensitive. Implementations\n MUST NOT send Handshake and Alert records that have a zero-length\n TLSInnerPlaintext.content; if such a message is received, the\n receiving implementation MUST terminate the connection with an\n \"unexpected_message\" alert.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 83]\n\f\nRFC 8446 TLS August 2018\n\n\n The padding sent is automatically verified by the record protection\n mechanism; upon successful decryption of a\n TLSCiphertext.encrypted_record, the receiving implementation scans\n the field from the end toward the beginning until it finds a non-zero\n octet. This non-zero octet is the content type of the message. This\n padding scheme was selected because it allows padding of any\n encrypted TLS record by an arbitrary size (from zero up to TLS record\n size limits) without introducing new content types. The design also\n enforces all-zero padding octets, which allows for quick detection of\n padding errors.\n\n Implementations MUST limit their scanning to the cleartext returned\n from the AEAD decryption. If a receiving implementation does not\n find a non-zero octet in the cleartext, it MUST terminate the\n connection with an \"unexpected_message\" alert.\n\n The presence of padding does not change the overall record size\n limitations: the full encoded TLSInnerPlaintext MUST NOT exceed 2^14\n + 1 octets. If the maximum fragment length is reduced -- as, for\n example, by the record_size_limit extension from [RFC8449] -- then\n the reduced limit applies to the full plaintext, including the\n content type and padding.\n\n Selecting a padding policy that suggests when and how much to pad is\n a complex topic and is beyond the scope of this specification. If\n the application-layer protocol on top of TLS has its own padding, it\n may be preferable to pad Application Data TLS records within the\n application layer. Padding for encrypted Handshake or Alert records\n must still be handled at the TLS layer, though. Later documents may\n define padding selection algorithms or define a padding policy\n request mechanism through TLS extensions or some other means.\n\n5.5. Limits on Key Usage\n\n There are cryptographic limits on the amount of plaintext which can\n be safely encrypted under a given set of keys. [AEAD-LIMITS]\n provides an analysis of these limits under the assumption that the\n underlying primitive (AES or ChaCha20) has no weaknesses.\n Implementations SHOULD do a key update as described in Section 4.6.3\n prior to reaching these limits.\n\n For AES-GCM, up to 2^24.5 full-size records (about 24 million) may be\n encrypted on a given connection while keeping a safety margin of\n approximately 2^-57 for Authenticated Encryption (AE) security. For\n ChaCha20/Poly1305, the record sequence number would wrap before the\n safety limit is reached.\n\n\n\n\n\nRescorla Standards Track [Page 84]\n\f\nRFC 8446 TLS August 2018\n\n\n6. Alert Protocol\n\n TLS provides an Alert content type to indicate closure information\n and errors. Like other messages, alert messages are encrypted as\n specified by the current connection state.\n\n Alert messages convey a description of the alert and a legacy field\n that conveyed the severity level of the message in previous versions\n of TLS. Alerts are divided into two classes: closure alerts and\n error alerts. In TLS 1.3, the severity is implicit in the type of\n alert being sent, and the \"level\" field can safely be ignored. The\n \"close_notify\" alert is used to indicate orderly closure of one\n direction of the connection. Upon receiving such an alert, the TLS\n implementation SHOULD indicate end-of-data to the application.\n\n Error alerts indicate abortive closure of the connection (see\n Section 6.2). Upon receiving an error alert, the TLS implementation\n SHOULD indicate an error to the application and MUST NOT allow any\n further data to be sent or received on the connection. Servers and\n clients MUST forget the secret values and keys established in failed\n connections, with the exception of the PSKs associated with session\n tickets, which SHOULD be discarded if possible.\n\n All the alerts listed in Section 6.2 MUST be sent with\n AlertLevel=fatal and MUST be treated as error alerts when received\n regardless of the AlertLevel in the message. Unknown Alert types\n MUST be treated as error alerts.\n\n Note: TLS defines two generic alerts (see Section 6) to use upon\n failure to parse a message. Peers which receive a message which\n cannot be parsed according to the syntax (e.g., have a length\n extending beyond the message boundary or contain an out-of-range\n length) MUST terminate the connection with a \"decode_error\" alert.\n Peers which receive a message which is syntactically correct but\n semantically invalid (e.g., a DHE share of p - 1, or an invalid enum)\n MUST terminate the connection with an \"illegal_parameter\" alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 85]\n\f\nRFC 8446 TLS August 2018\n\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n record_overflow(22),\n handshake_failure(40),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n inappropriate_fallback(86),\n user_canceled(90),\n missing_extension(109),\n unsupported_extension(110),\n unrecognized_name(112),\n bad_certificate_status_response(113),\n unknown_psk_identity(115),\n certificate_required(116),\n no_application_protocol(120),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 86]\n\f\nRFC 8446 TLS August 2018\n\n\n6.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack.\n\n close_notify: This alert notifies the recipient that the sender will\n not send any more messages on this connection. Any data received\n after a closure alert has been received MUST be ignored.\n\n user_canceled: This alert notifies the recipient that the sender is\n canceling the handshake for some reason unrelated to a protocol\n failure. If a user cancels an operation after the handshake is\n complete, just closing the connection by sending a \"close_notify\"\n is more appropriate. This alert SHOULD be followed by a\n \"close_notify\". This alert generally has AlertLevel=warning.\n\n Either party MAY initiate a close of its write side of the connection\n by sending a \"close_notify\" alert. Any data received after a closure\n alert has been received MUST be ignored. If a transport-level close\n is received prior to a \"close_notify\", the receiver cannot know that\n all the data that was sent has been received.\n\n Each party MUST send a \"close_notify\" alert before closing its write\n side of the connection, unless it has already sent some error alert.\n This does not have any effect on its read side of the connection.\n Note that this is a change from versions of TLS prior to TLS 1.3 in\n which implementations were required to react to a \"close_notify\" by\n discarding pending writes and sending an immediate \"close_notify\"\n alert of their own. That previous requirement could cause truncation\n in the read side. Both parties need not wait to receive a\n \"close_notify\" alert before closing their read side of the\n connection, though doing so would introduce the possibility of\n truncation.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation MUST receive a \"close_notify\" alert\n before indicating end-of-data to the application layer. No part of\n this standard should be taken to dictate the manner in which a usage\n profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing the write side of a connection\n reliably delivers pending data before destroying the transport.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 87]\n\f\nRFC 8446 TLS August 2018\n\n\n6.2. Error Alerts\n\n Error handling in TLS is very simple. When an error is detected, the\n detecting party sends a message to its peer. Upon transmission or\n receipt of a fatal alert message, both parties MUST immediately close\n the connection.\n\n Whenever an implementation encounters a fatal error condition, it\n SHOULD send an appropriate fatal alert and MUST close the connection\n without sending or receiving any additional data. In the rest of\n this specification, when the phrases \"terminate the connection\" and\n \"abort the handshake\" are used without a specific alert it means that\n the implementation SHOULD send the alert indicated by the\n descriptions below. The phrases \"terminate the connection with an X\n alert\" and \"abort the handshake with an X alert\" mean that the\n implementation MUST send alert X if it sends any alert. All alerts\n defined below in this section, as well as all unknown alerts, are\n universally considered fatal as of TLS 1.3 (see Section 6). The\n implementation SHOULD provide a way to facilitate logging the sending\n and receiving of alerts.\n\n The following error alerts are defined:\n\n unexpected_message: An inappropriate message (e.g., the wrong\n handshake message, premature Application Data, etc.) was received.\n This alert should never be observed in communication between\n proper implementations.\n\n bad_record_mac: This alert is returned if a record is received which\n cannot be deprotected. Because AEAD algorithms combine decryption\n and verification, and also to avoid side-channel attacks, this\n alert is used for all deprotection failures. This alert should\n never be observed in communication between proper implementations,\n except when messages were corrupted in the network.\n\n record_overflow: A TLSCiphertext record was received that had a\n length more than 2^14 + 256 bytes, or a record decrypted to a\n TLSPlaintext record with more than 2^14 bytes (or some other\n negotiated limit). This alert should never be observed in\n communication between proper implementations, except when messages\n were corrupted in the network.\n\n handshake_failure: Receipt of a \"handshake_failure\" alert message\n indicates that the sender was unable to negotiate an acceptable\n set of security parameters given the options available.\n\n bad_certificate: A certificate was corrupt, contained signatures\n that did not verify correctly, etc.\n\n\n\nRescorla Standards Track [Page 88]\n\f\nRFC 8446 TLS August 2018\n\n\n unsupported_certificate: A certificate was of an unsupported type.\n\n certificate_revoked: A certificate was revoked by its signer.\n\n certificate_expired: A certificate has expired or is not currently\n valid.\n\n certificate_unknown: Some other (unspecified) issue arose in\n processing the certificate, rendering it unacceptable.\n\n illegal_parameter: A field in the handshake was incorrect or\n inconsistent with other fields. This alert is used for errors\n which conform to the formal protocol syntax but are otherwise\n incorrect.\n\n unknown_ca: A valid certificate chain or partial chain was received,\n but the certificate was not accepted because the CA certificate\n could not be located or could not be matched with a known trust\n anchor.\n\n access_denied: A valid certificate or PSK was received, but when\n access control was applied, the sender decided not to proceed with\n negotiation.\n\n decode_error: A message could not be decoded because some field was\n out of the specified range or the length of the message was\n incorrect. This alert is used for errors where the message does\n not conform to the formal protocol syntax. This alert should\n never be observed in communication between proper implementations,\n except when messages were corrupted in the network.\n\n decrypt_error: A handshake (not record layer) cryptographic\n operation failed, including being unable to correctly verify a\n signature or validate a Finished message or a PSK binder.\n\n protocol_version: The protocol version the peer has attempted to\n negotiate is recognized but not supported (see Appendix D).\n\n insufficient_security: Returned instead of \"handshake_failure\" when\n a negotiation has failed specifically because the server requires\n parameters more secure than those supported by the client.\n\n internal_error: An internal error unrelated to the peer or the\n correctness of the protocol (such as a memory allocation failure)\n makes it impossible to continue.\n\n inappropriate_fallback: Sent by a server in response to an invalid\n connection retry attempt from a client (see [RFC7507]).\n\n\n\nRescorla Standards Track [Page 89]\n\f\nRFC 8446 TLS August 2018\n\n\n missing_extension: Sent by endpoints that receive a handshake\n message not containing an extension that is mandatory to send for\n the offered TLS version or other negotiated parameters.\n\n unsupported_extension: Sent by endpoints receiving any handshake\n message containing an extension known to be prohibited for\n inclusion in the given handshake message, or including any\n extensions in a ServerHello or Certificate not first offered in\n the corresponding ClientHello or CertificateRequest.\n\n unrecognized_name: Sent by servers when no server exists identified\n by the name provided by the client via the \"server_name\" extension\n (see [RFC6066]).\n\n bad_certificate_status_response: Sent by clients when an invalid or\n unacceptable OCSP response is provided by the server via the\n \"status_request\" extension (see [RFC6066]).\n\n unknown_psk_identity: Sent by servers when PSK key establishment is\n desired but no acceptable PSK identity is provided by the client.\n Sending this alert is OPTIONAL; servers MAY instead choose to send\n a \"decrypt_error\" alert to merely indicate an invalid PSK\n identity.\n\n certificate_required: Sent by servers when a client certificate is\n desired but none was provided by the client.\n\n no_application_protocol: Sent by servers when a client\n \"application_layer_protocol_negotiation\" extension advertises only\n protocols that the server does not support (see [RFC7301]).\n\n New Alert values are assigned by IANA as described in Section 11.\n\n7. Cryptographic Computations\n\n The TLS handshake establishes one or more input secrets which are\n combined to create the actual working keying material, as detailed\n below. The key derivation process incorporates both the input\n secrets and the handshake transcript. Note that because the\n handshake transcript includes the random values from the Hello\n messages, any given handshake will have different traffic secrets,\n even if the same input secrets are used, as is the case when the same\n PSK is used for multiple connections.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 90]\n\f\nRFC 8446 TLS August 2018\n\n\n7.1. Key Schedule\n\n The key derivation process makes use of the HKDF-Extract and\n HKDF-Expand functions as defined for HKDF [RFC5869], as well as the\n functions defined below:\n\n HKDF-Expand-Label(Secret, Label, Context, Length) =\n HKDF-Expand(Secret, HkdfLabel, Length)\n\n Where HkdfLabel is specified as:\n\n struct {\n uint16 length = Length;\n opaque label<7..255> = \"tls13 \" + Label;\n opaque context<0..255> = Context;\n } HkdfLabel;\n\n Derive-Secret(Secret, Label, Messages) =\n HKDF-Expand-Label(Secret, Label,\n Transcript-Hash(Messages), Hash.length)\n\n The Hash function used by Transcript-Hash and HKDF is the cipher\n suite hash algorithm. Hash.length is its output length in bytes.\n Messages is the concatenation of the indicated handshake messages,\n including the handshake message type and length fields, but not\n including record layer headers. Note that in some cases a zero-\n length Context (indicated by \"\") is passed to HKDF-Expand-Label. The\n labels specified in this document are all ASCII strings and do not\n include a trailing NUL byte.\n\n Note: With common hash functions, any label longer than 12 characters\n requires an additional iteration of the hash function to compute.\n The labels in this specification have all been chosen to fit within\n this limit.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 91]\n\f\nRFC 8446 TLS August 2018\n\n\n Keys are derived from two input secrets using the HKDF-Extract and\n Derive-Secret functions. The general pattern for adding a new secret\n is to use HKDF-Extract with the Salt being the current secret state\n and the Input Keying Material (IKM) being the new secret to be added.\n In this version of TLS 1.3, the two input secrets are:\n\n - PSK (a pre-shared key established externally or derived from the\n resumption_master_secret value from a previous connection)\n\n - (EC)DHE shared secret (Section 7.4)\n\n This produces a full key derivation schedule shown in the diagram\n below. In this diagram, the following formatting conventions apply:\n\n - HKDF-Extract is drawn as taking the Salt argument from the top and\n the IKM argument from the left, with its output to the bottom and\n the name of the output on the right.\n\n - Derive-Secret's Secret argument is indicated by the incoming\n arrow. For instance, the Early Secret is the Secret for\n generating the client_early_traffic_secret.\n\n - \"0\" indicates a string of Hash.length bytes set to zero.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 92]\n\f\nRFC 8446 TLS August 2018\n\n\n 0\n |\n v\n PSK -> HKDF-Extract = Early Secret\n |\n +-----> Derive-Secret(., \"ext binder\" | \"res binder\", \"\")\n | = binder_key\n |\n +-----> Derive-Secret(., \"c e traffic\", ClientHello)\n | = client_early_traffic_secret\n |\n +-----> Derive-Secret(., \"e exp master\", ClientHello)\n | = early_exporter_master_secret\n v\n Derive-Secret(., \"derived\", \"\")\n |\n v\n (EC)DHE -> HKDF-Extract = Handshake Secret\n |\n +-----> Derive-Secret(., \"c hs traffic\",\n | ClientHello...ServerHello)\n | = client_handshake_traffic_secret\n |\n +-----> Derive-Secret(., \"s hs traffic\",\n | ClientHello...ServerHello)\n | = server_handshake_traffic_secret\n v\n Derive-Secret(., \"derived\", \"\")\n |\n v\n 0 -> HKDF-Extract = Master Secret\n |\n +-----> Derive-Secret(., \"c ap traffic\",\n | ClientHello...server Finished)\n | = client_application_traffic_secret_0\n |\n +-----> Derive-Secret(., \"s ap traffic\",\n | ClientHello...server Finished)\n | = server_application_traffic_secret_0\n |\n +-----> Derive-Secret(., \"exp master\",\n | ClientHello...server Finished)\n | = exporter_master_secret\n |\n +-----> Derive-Secret(., \"res master\",\n ClientHello...client Finished)\n = resumption_master_secret\n\n\n\n\nRescorla Standards Track [Page 93]\n\f\nRFC 8446 TLS August 2018\n\n\n The general pattern here is that the secrets shown down the left side\n of the diagram are just raw entropy without context, whereas the\n secrets down the right side include Handshake Context and therefore\n can be used to derive working keys without additional context. Note\n that the different calls to Derive-Secret may take different Messages\n arguments, even with the same secret. In a 0-RTT exchange,\n Derive-Secret is called with four distinct transcripts; in a\n 1-RTT-only exchange, it is called with three distinct transcripts.\n\n If a given secret is not available, then the 0-value consisting of a\n string of Hash.length bytes set to zeros is used. Note that this\n does not mean skipping rounds, so if PSK is not in use, Early Secret\n will still be HKDF-Extract(0, 0). For the computation of the\n binder_key, the label is \"ext binder\" for external PSKs (those\n provisioned outside of TLS) and \"res binder\" for resumption PSKs\n (those provisioned as the resumption master secret of a previous\n handshake). The different labels prevent the substitution of one\n type of PSK for the other.\n\n There are multiple potential Early Secret values, depending on which\n PSK the server ultimately selects. The client will need to compute\n one for each potential PSK; if no PSK is selected, it will then need\n to compute the Early Secret corresponding to the zero PSK.\n\n Once all the values which are to be derived from a given secret have\n been computed, that secret SHOULD be erased.\n\n7.2. Updating Traffic Secrets\n\n Once the handshake is complete, it is possible for either side to\n update its sending traffic keys using the KeyUpdate handshake message\n defined in Section 4.6.3. The next generation of traffic keys is\n computed by generating client_/server_application_traffic_secret_N+1\n from client_/server_application_traffic_secret_N as described in this\n section and then re-deriving the traffic keys as described in\n Section 7.3.\n\n The next-generation application_traffic_secret is computed as:\n\n application_traffic_secret_N+1 =\n HKDF-Expand-Label(application_traffic_secret_N,\n \"traffic upd\", \"\", Hash.length)\n\n Once client_/server_application_traffic_secret_N+1 and its associated\n traffic keys have been computed, implementations SHOULD delete\n client_/server_application_traffic_secret_N and its associated\n traffic keys.\n\n\n\n\nRescorla Standards Track [Page 94]\n\f\nRFC 8446 TLS August 2018\n\n\n7.3. Traffic Key Calculation\n\n The traffic keying material is generated from the following input\n values:\n\n - A secret value\n\n - A purpose value indicating the specific value being generated\n\n - The length of the key being generated\n\n The traffic keying material is generated from an input traffic secret\n value using:\n\n [sender]_write_key = HKDF-Expand-Label(Secret, \"key\", \"\", key_length)\n [sender]_write_iv = HKDF-Expand-Label(Secret, \"iv\", \"\", iv_length)\n\n [sender] denotes the sending side. The value of Secret for each\n record type is shown in the table below.\n\n +-------------------+---------------------------------------+\n | Record Type | Secret |\n +-------------------+---------------------------------------+\n | 0-RTT Application | client_early_traffic_secret |\n | | |\n | Handshake | [sender]_handshake_traffic_secret |\n | | |\n | Application Data | [sender]_application_traffic_secret_N |\n +-------------------+---------------------------------------+\n\n All the traffic keying material is recomputed whenever the underlying\n Secret changes (e.g., when changing from the handshake to Application\n Data keys or upon a key update).\n\n7.4. (EC)DHE Shared Secret Calculation\n\n7.4.1. Finite Field Diffie-Hellman\n\n For finite field groups, a conventional Diffie-Hellman [DH76]\n computation is performed. The negotiated key (Z) is converted to a\n byte string by encoding in big-endian form and left-padded with zeros\n up to the size of the prime. This byte string is used as the shared\n secret in the key schedule as specified above.\n\n Note that this construction differs from previous versions of TLS\n which removed leading zeros.\n\n\n\n\n\nRescorla Standards Track [Page 95]\n\f\nRFC 8446 TLS August 2018\n\n\n7.4.2. Elliptic Curve Diffie-Hellman\n\n For secp256r1, secp384r1, and secp521r1, ECDH calculations (including\n parameter and key generation as well as the shared secret\n calculation) are performed according to [IEEE1363] using the\n ECKAS-DH1 scheme with the identity map as the key derivation function\n (KDF), so that the shared secret is the x-coordinate of the ECDH\n shared secret elliptic curve point represented as an octet string.\n Note that this octet string (\"Z\" in IEEE 1363 terminology) as output\n by FE2OSP (the Field Element to Octet String Conversion Primitive)\n has constant length for any given field; leading zeros found in this\n octet string MUST NOT be truncated.\n\n (Note that this use of the identity KDF is a technicality. The\n complete picture is that ECDH is employed with a non-trivial KDF\n because TLS does not directly use this secret for anything other than\n for computing other secrets.)\n\n For X25519 and X448, the ECDH calculations are as follows:\n\n - The public key to put into the KeyShareEntry.key_exchange\n structure is the result of applying the ECDH scalar multiplication\n function to the secret key of appropriate length (into scalar\n input) and the standard public basepoint (into u-coordinate point\n input).\n\n - The ECDH shared secret is the result of applying the ECDH scalar\n multiplication function to the secret key (into scalar input) and\n the peer's public key (into u-coordinate point input). The output\n is used raw, with no processing.\n\n For these curves, implementations SHOULD use the approach specified\n in [RFC7748] to calculate the Diffie-Hellman shared secret.\n Implementations MUST check whether the computed Diffie-Hellman shared\n secret is the all-zero value and abort if so, as described in\n Section 6 of [RFC7748]. If implementors use an alternative\n implementation of these elliptic curves, they SHOULD perform the\n additional checks specified in Section 7 of [RFC7748].\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 96]\n\f\nRFC 8446 TLS August 2018\n\n\n7.5. Exporters\n\n [RFC5705] defines keying material exporters for TLS in terms of the\n TLS pseudorandom function (PRF). This document replaces the PRF with\n HKDF, thus requiring a new construction. The exporter interface\n remains the same.\n\n The exporter value is computed as:\n\n TLS-Exporter(label, context_value, key_length) =\n HKDF-Expand-Label(Derive-Secret(Secret, label, \"\"),\n \"exporter\", Hash(context_value), key_length)\n\n Where Secret is either the early_exporter_master_secret or the\n exporter_master_secret. Implementations MUST use the\n exporter_master_secret unless explicitly specified by the\n application. The early_exporter_master_secret is defined for use in\n settings where an exporter is needed for 0-RTT data. A separate\n interface for the early exporter is RECOMMENDED; this avoids the\n exporter user accidentally using an early exporter when a regular one\n is desired or vice versa.\n\n If no context is provided, the context_value is zero length.\n Consequently, providing no context computes the same value as\n providing an empty context. This is a change from previous versions\n of TLS where an empty context produced a different output than an\n absent context. As of this document's publication, no allocated\n exporter label is used both with and without a context. Future\n specifications MUST NOT define a use of exporters that permit both an\n empty context and no context with the same label. New uses of\n exporters SHOULD provide a context in all exporter computations,\n though the value could be empty.\n\n Requirements for the format of exporter labels are defined in\n Section 4 of [RFC5705].\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 97]\n\f\nRFC 8446 TLS August 2018\n\n\n8. 0-RTT and Anti-Replay\n\n As noted in Section 2.3 and Appendix E.5, TLS does not provide\n inherent replay protections for 0-RTT data. There are two potential\n threats to be concerned with:\n\n - Network attackers who mount a replay attack by simply duplicating\n a flight of 0-RTT data.\n\n - Network attackers who take advantage of client retry behavior to\n arrange for the server to receive multiple copies of an\n application message. This threat already exists to some extent\n because clients that value robustness respond to network errors by\n attempting to retry requests. However, 0-RTT adds an additional\n dimension for any server system which does not maintain globally\n consistent server state. Specifically, if a server system has\n multiple zones where tickets from zone A will not be accepted in\n zone B, then an attacker can duplicate a ClientHello and early\n data intended for A to both A and B. At A, the data will be\n accepted in 0-RTT, but at B the server will reject 0-RTT data and\n instead force a full handshake. If the attacker blocks the\n ServerHello from A, then the client will complete the handshake\n with B and probably retry the request, leading to duplication on\n the server system as a whole.\n\n The first class of attack can be prevented by sharing state to\n guarantee that the 0-RTT data is accepted at most once. Servers\n SHOULD provide that level of replay safety by implementing one of the\n methods described in this section or by equivalent means. It is\n understood, however, that due to operational concerns not all\n deployments will maintain state at that level. Therefore, in normal\n operation, clients will not know which, if any, of these mechanisms\n servers actually implement and hence MUST only send early data which\n they deem safe to be replayed.\n\n In addition to the direct effects of replays, there is a class of\n attacks where even operations normally considered idempotent could be\n exploited by a large number of replays (timing attacks, resource\n limit exhaustion and others, as described in Appendix E.5). Those\n can be mitigated by ensuring that every 0-RTT payload can be replayed\n only a limited number of times. The server MUST ensure that any\n instance of it (be it a machine, a thread, or any other entity within\n the relevant serving infrastructure) would accept 0-RTT for the same\n 0-RTT handshake at most once; this limits the number of replays to\n the number of server instances in the deployment. Such a guarantee\n can be accomplished by locally recording data from recently received\n ClientHellos and rejecting repeats, or by any other method that\n\n\n\n\nRescorla Standards Track [Page 98]\n\f\nRFC 8446 TLS August 2018\n\n\n provides the same or a stronger guarantee. The \"at most once per\n server instance\" guarantee is a minimum requirement; servers SHOULD\n limit 0-RTT replays further when feasible.\n\n The second class of attack cannot be prevented at the TLS layer and\n MUST be dealt with by any application. Note that any application\n whose clients implement any kind of retry behavior already needs to\n implement some sort of anti-replay defense.\n\n8.1. Single-Use Tickets\n\n The simplest form of anti-replay defense is for the server to only\n allow each session ticket to be used once. For instance, the server\n can maintain a database of all outstanding valid tickets, deleting\n each ticket from the database as it is used. If an unknown ticket is\n provided, the server would then fall back to a full handshake.\n\n If the tickets are not self-contained but rather are database keys,\n and the corresponding PSKs are deleted upon use, then connections\n established using PSKs enjoy forward secrecy. This improves security\n for all 0-RTT data and PSK usage when PSK is used without (EC)DHE.\n\n Because this mechanism requires sharing the session database between\n server nodes in environments with multiple distributed servers, it\n may be hard to achieve high rates of successful PSK 0-RTT connections\n when compared to self-encrypted tickets. Unlike session databases,\n session tickets can successfully do PSK-based session establishment\n even without consistent storage, though when 0-RTT is allowed they\n still require consistent storage for anti-replay of 0-RTT data, as\n detailed in the following section.\n\n8.2. Client Hello Recording\n\n An alternative form of anti-replay is to record a unique value\n derived from the ClientHello (generally either the random value or\n the PSK binder) and reject duplicates. Recording all ClientHellos\n causes state to grow without bound, but a server can instead record\n ClientHellos within a given time window and use the\n \"obfuscated_ticket_age\" to ensure that tickets aren't reused outside\n that window.\n\n In order to implement this, when a ClientHello is received, the\n server first verifies the PSK binder as described in Section 4.2.11.\n It then computes the expected_arrival_time as described in the next\n section and rejects 0-RTT if it is outside the recording window,\n falling back to the 1-RTT handshake.\n\n\n\n\n\nRescorla Standards Track [Page 99]\n\f\nRFC 8446 TLS August 2018\n\n\n If the expected_arrival_time is in the window, then the server checks\n to see if it has recorded a matching ClientHello. If one is found,\n it either aborts the handshake with an \"illegal_parameter\" alert or\n accepts the PSK but rejects 0-RTT. If no matching ClientHello is\n found, then it accepts 0-RTT and then stores the ClientHello for as\n long as the expected_arrival_time is inside the window. Servers MAY\n also implement data stores with false positives, such as Bloom\n filters, in which case they MUST respond to apparent replay by\n rejecting 0-RTT but MUST NOT abort the handshake.\n\n The server MUST derive the storage key only from validated sections\n of the ClientHello. If the ClientHello contains multiple PSK\n identities, then an attacker can create multiple ClientHellos with\n different binder values for the less-preferred identity on the\n assumption that the server will not verify it (as recommended by\n Section 4.2.11). I.e., if the client sends PSKs A and B but the\n server prefers A, then the attacker can change the binder for B\n without affecting the binder for A. If the binder for B is part of\n the storage key, then this ClientHello will not appear as a\n duplicate, which will cause the ClientHello to be accepted, and may\n cause side effects such as replay cache pollution, although any 0-RTT\n data will not be decryptable because it will use different keys. If\n the validated binder or the ClientHello.random is used as the storage\n key, then this attack is not possible.\n\n Because this mechanism does not require storing all outstanding\n tickets, it may be easier to implement in distributed systems with\n high rates of resumption and 0-RTT, at the cost of potentially weaker\n anti-replay defense because of the difficulty of reliably storing and\n retrieving the received ClientHello messages. In many such systems,\n it is impractical to have globally consistent storage of all the\n received ClientHellos. In this case, the best anti-replay protection\n is provided by having a single storage zone be authoritative for a\n given ticket and refusing 0-RTT for that ticket in any other zone.\n This approach prevents simple replay by the attacker because only one\n zone will accept 0-RTT data. A weaker design is to implement\n separate storage for each zone but allow 0-RTT in any zone. This\n approach limits the number of replays to once per zone. Application\n message duplication of course remains possible with either design.\n\n When implementations are freshly started, they SHOULD reject 0-RTT as\n long as any portion of their recording window overlaps the startup\n time. Otherwise, they run the risk of accepting replays which were\n originally sent during that period.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 100]\n\f\nRFC 8446 TLS August 2018\n\n\n Note: If the client's clock is running much faster than the server's,\n then a ClientHello may be received that is outside the window in the\n future, in which case it might be accepted for 1-RTT, causing a\n client retry, and then acceptable later for 0-RTT. This is another\n variant of the second form of attack described in Section 8.\n\n8.3. Freshness Checks\n\n Because the ClientHello indicates the time at which the client sent\n it, it is possible to efficiently determine whether a ClientHello was\n likely sent reasonably recently and only accept 0-RTT for such a\n ClientHello, otherwise falling back to a 1-RTT handshake. This is\n necessary for the ClientHello storage mechanism described in\n Section 8.2 because otherwise the server needs to store an unlimited\n number of ClientHellos, and is a useful optimization for self-\n contained single-use tickets because it allows efficient rejection of\n ClientHellos which cannot be used for 0-RTT.\n\n In order to implement this mechanism, a server needs to store the\n time that the server generated the session ticket, offset by an\n estimate of the round-trip time between client and server. I.e.,\n\n adjusted_creation_time = creation_time + estimated_RTT\n\n This value can be encoded in the ticket, thus avoiding the need to\n keep state for each outstanding ticket. The server can determine the\n client's view of the age of the ticket by subtracting the ticket's\n \"ticket_age_add\" value from the \"obfuscated_ticket_age\" parameter in\n the client's \"pre_shared_key\" extension. The server can determine\n the expected_arrival_time of the ClientHello as:\n\n expected_arrival_time = adjusted_creation_time + clients_ticket_age\n\n When a new ClientHello is received, the expected_arrival_time is then\n compared against the current server wall clock time and if they\n differ by more than a certain amount, 0-RTT is rejected, though the\n 1-RTT handshake can be allowed to complete.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 101]\n\f\nRFC 8446 TLS August 2018\n\n\n There are several potential sources of error that might cause\n mismatches between the expected_arrival_time and the measured time.\n Variations in client and server clock rates are likely to be minimal,\n though potentially the absolute times may be off by large values.\n Network propagation delays are the most likely causes of a mismatch\n in legitimate values for elapsed time. Both the NewSessionTicket and\n ClientHello messages might be retransmitted and therefore delayed,\n which might be hidden by TCP. For clients on the Internet, this\n implies windows on the order of ten seconds to account for errors in\n clocks and variations in measurements; other deployment scenarios may\n have different needs. Clock skew distributions are not symmetric, so\n the optimal tradeoff may involve an asymmetric range of permissible\n mismatch values.\n\n Note that freshness checking alone is not sufficient to prevent\n replays because it does not detect them during the error window,\n which -- depending on bandwidth and system capacity -- could include\n billions of replays in real-world settings. In addition, this\n freshness checking is only done at the time the ClientHello is\n received and not when subsequent early Application Data records are\n received. After early data is accepted, records may continue to be\n streamed to the server over a longer time period.\n\n9. Compliance Requirements\n\n9.1. Mandatory-to-Implement Cipher Suites\n\n In the absence of an application profile standard specifying\n otherwise:\n\n A TLS-compliant application MUST implement the TLS_AES_128_GCM_SHA256\n [GCM] cipher suite and SHOULD implement the TLS_AES_256_GCM_SHA384\n [GCM] and TLS_CHACHA20_POLY1305_SHA256 [RFC8439] cipher suites (see\n Appendix B.4).\n\n A TLS-compliant application MUST support digital signatures with\n rsa_pkcs1_sha256 (for certificates), rsa_pss_rsae_sha256 (for\n CertificateVerify and certificates), and ecdsa_secp256r1_sha256. A\n TLS-compliant application MUST support key exchange with secp256r1\n (NIST P-256) and SHOULD support key exchange with X25519 [RFC7748].\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 102]\n\f\nRFC 8446 TLS August 2018\n\n\n9.2. Mandatory-to-Implement Extensions\n\n In the absence of an application profile standard specifying\n otherwise, a TLS-compliant application MUST implement the following\n TLS extensions:\n\n - Supported Versions (\"supported_versions\"; Section 4.2.1)\n\n - Cookie (\"cookie\"; Section 4.2.2)\n\n - Signature Algorithms (\"signature_algorithms\"; Section 4.2.3)\n\n - Signature Algorithms Certificate (\"signature_algorithms_cert\";\n Section 4.2.3)\n\n - Negotiated Groups (\"supported_groups\"; Section 4.2.7)\n\n - Key Share (\"key_share\"; Section 4.2.8)\n\n - Server Name Indication (\"server_name\"; Section 3 of [RFC6066])\n\n All implementations MUST send and use these extensions when offering\n applicable features:\n\n - \"supported_versions\" is REQUIRED for all ClientHello, ServerHello,\n and HelloRetryRequest messages.\n\n - \"signature_algorithms\" is REQUIRED for certificate authentication.\n\n - \"supported_groups\" is REQUIRED for ClientHello messages using DHE\n or ECDHE key exchange.\n\n - \"key_share\" is REQUIRED for DHE or ECDHE key exchange.\n\n - \"pre_shared_key\" is REQUIRED for PSK key agreement.\n\n - \"psk_key_exchange_modes\" is REQUIRED for PSK key agreement.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 103]\n\f\nRFC 8446 TLS August 2018\n\n\n A client is considered to be attempting to negotiate using this\n specification if the ClientHello contains a \"supported_versions\"\n extension with 0x0304 contained in its body. Such a ClientHello\n message MUST meet the following requirements:\n\n - If not containing a \"pre_shared_key\" extension, it MUST contain\n both a \"signature_algorithms\" extension and a \"supported_groups\"\n extension.\n\n - If containing a \"supported_groups\" extension, it MUST also contain\n a \"key_share\" extension, and vice versa. An empty\n KeyShare.client_shares vector is permitted.\n\n Servers receiving a ClientHello which does not conform to these\n requirements MUST abort the handshake with a \"missing_extension\"\n alert.\n\n Additionally, all implementations MUST support the use of the\n \"server_name\" extension with applications capable of using it.\n Servers MAY require clients to send a valid \"server_name\" extension.\n Servers requiring this extension SHOULD respond to a ClientHello\n lacking a \"server_name\" extension by terminating the connection with\n a \"missing_extension\" alert.\n\n9.3. Protocol Invariants\n\n This section describes invariants that TLS endpoints and middleboxes\n MUST follow. It also applies to earlier versions of TLS.\n\n TLS is designed to be securely and compatibly extensible. Newer\n clients or servers, when communicating with newer peers, should\n negotiate the most preferred common parameters. The TLS handshake\n provides downgrade protection: Middleboxes passing traffic between a\n newer client and newer server without terminating TLS should be\n unable to influence the handshake (see Appendix E.1). At the same\n time, deployments update at different rates, so a newer client or\n server MAY continue to support older parameters, which would allow it\n to interoperate with older endpoints.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 104]\n\f\nRFC 8446 TLS August 2018\n\n\n For this to work, implementations MUST correctly handle extensible\n fields:\n\n - A client sending a ClientHello MUST support all parameters\n advertised in it. Otherwise, the server may fail to interoperate\n by selecting one of those parameters.\n\n - A server receiving a ClientHello MUST correctly ignore all\n unrecognized cipher suites, extensions, and other parameters.\n Otherwise, it may fail to interoperate with newer clients. In\n TLS 1.3, a client receiving a CertificateRequest or\n NewSessionTicket MUST also ignore all unrecognized extensions.\n\n - A middlebox which terminates a TLS connection MUST behave as a\n compliant TLS server (to the original client), including having a\n certificate which the client is willing to accept, and also as a\n compliant TLS client (to the original server), including verifying\n the original server's certificate. In particular, it MUST\n generate its own ClientHello containing only parameters it\n understands, and it MUST generate a fresh ServerHello random\n value, rather than forwarding the endpoint's value.\n\n Note that TLS's protocol requirements and security analysis only\n apply to the two connections separately. Safely deploying a TLS\n terminator requires additional security considerations which are\n beyond the scope of this document.\n\n - A middlebox which forwards ClientHello parameters it does not\n understand MUST NOT process any messages beyond that ClientHello.\n It MUST forward all subsequent traffic unmodified. Otherwise, it\n may fail to interoperate with newer clients and servers.\n\n Forwarded ClientHellos may contain advertisements for features not\n supported by the middlebox, so the response may include future TLS\n additions the middlebox does not recognize. These additions MAY\n change any message beyond the ClientHello arbitrarily. In\n particular, the values sent in the ServerHello might change, the\n ServerHello format might change, and the TLSCiphertext format\n might change.\n\n The design of TLS 1.3 was constrained by widely deployed\n non-compliant TLS middleboxes (see Appendix D.4); however, it does\n not relax the invariants. Those middleboxes continue to be\n non-compliant.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 105]\n\f\nRFC 8446 TLS August 2018\n\n\n10. Security Considerations\n\n Security issues are discussed throughout this memo, especially in\n Appendices C, D, and E.\n\n11. IANA Considerations\n\n This document uses several registries that were originally created in\n [RFC4346] and updated in [RFC8447]. IANA has updated these to\n reference this document. The registries and their allocation\n policies are below:\n\n - TLS Cipher Suites registry: values with the first byte in the\n range 0-254 (decimal) are assigned via Specification Required\n [RFC8126]. Values with the first byte 255 (decimal) are reserved\n for Private Use [RFC8126].\n\n IANA has added the cipher suites listed in Appendix B.4 to the\n registry. The \"Value\" and \"Description\" columns are taken from\n the table. The \"DTLS-OK\" and \"Recommended\" columns are both\n marked as \"Y\" for each new cipher suite.\n\n - TLS ContentType registry: Future values are allocated via\n Standards Action [RFC8126].\n\n - TLS Alerts registry: Future values are allocated via Standards\n Action [RFC8126]. IANA has populated this registry with the\n values from Appendix B.2. The \"DTLS-OK\" column is marked as \"Y\"\n for all such values. Values marked as \"_RESERVED\" have comments\n describing their previous usage.\n\n - TLS HandshakeType registry: Future values are allocated via\n Standards Action [RFC8126]. IANA has updated this registry to\n rename item 4 from \"NewSessionTicket\" to \"new_session_ticket\" and\n populated this registry with the values from Appendix B.3. The\n \"DTLS-OK\" column is marked as \"Y\" for all such values. Values\n marked \"_RESERVED\" have comments describing their previous or\n temporary usage.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 106]\n\f\nRFC 8446 TLS August 2018\n\n\n This document also uses the TLS ExtensionType Values registry\n originally created in [RFC4366]. IANA has updated it to reference\n this document. Changes to the registry follow:\n\n - IANA has updated the registration policy as follows:\n\n Values with the first byte in the range 0-254 (decimal) are\n assigned via Specification Required [RFC8126]. Values with the\n first byte 255 (decimal) are reserved for Private Use [RFC8126].\n\n - IANA has updated this registry to include the \"key_share\",\n \"pre_shared_key\", \"psk_key_exchange_modes\", \"early_data\",\n \"cookie\", \"supported_versions\", \"certificate_authorities\",\n \"oid_filters\", \"post_handshake_auth\", and\n \"signature_algorithms_cert\" extensions with the values defined in\n this document and the \"Recommended\" value of \"Y\".\n\n - IANA has updated this registry to include a \"TLS 1.3\" column which\n lists the messages in which the extension may appear. This column\n has been initially populated from the table in Section 4.2, with\n any extension not listed there marked as \"-\" to indicate that it\n is not used by TLS 1.3.\n\n This document updates an entry in the TLS Certificate Types registry\n originally created in [RFC6091] and updated in [RFC8447]. IANA has\n updated the entry for value 1 to have the name \"OpenPGP_RESERVED\",\n \"Recommended\" value \"N\", and comment \"Used in TLS versions prior\n to 1.3.\"\n\n This document updates an entry in the TLS Certificate Status Types\n registry originally created in [RFC6961]. IANA has updated the entry\n for value 2 to have the name \"ocsp_multi_RESERVED\" and comment \"Used\n in TLS versions prior to 1.3\".\n\n This document updates two entries in the TLS Supported Groups\n registry (created under a different name by [RFC4492]; now maintained\n by [RFC8422]) and updated by [RFC7919] and [RFC8447]. The entries\n for values 29 and 30 (x25519 and x448) have been updated to also\n refer to this document.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 107]\n\f\nRFC 8446 TLS August 2018\n\n\n In addition, this document defines two new registries that are\n maintained by IANA:\n\n - TLS SignatureScheme registry: Values with the first byte in the\n range 0-253 (decimal) are assigned via Specification Required\n [RFC8126]. Values with the first byte 254 or 255 (decimal) are\n reserved for Private Use [RFC8126]. Values with the first byte in\n the range 0-6 or with the second byte in the range 0-3 that are\n not currently allocated are reserved for backward compatibility.\n This registry has a \"Recommended\" column. The registry has been\n initially populated with the values described in Section 4.2.3.\n The following values are marked as \"Recommended\":\n ecdsa_secp256r1_sha256, ecdsa_secp384r1_sha384,\n rsa_pss_rsae_sha256, rsa_pss_rsae_sha384, rsa_pss_rsae_sha512,\n rsa_pss_pss_sha256, rsa_pss_pss_sha384, rsa_pss_pss_sha512, and\n ed25519. The \"Recommended\" column is assigned a value of \"N\"\n unless explicitly requested, and adding a value with a\n \"Recommended\" value of \"Y\" requires Standards Action [RFC8126].\n IESG Approval is REQUIRED for a Y->N transition.\n\n - TLS PskKeyExchangeMode registry: Values in the range 0-253\n (decimal) are assigned via Specification Required [RFC8126].\n The values 254 and 255 (decimal) are reserved for Private Use\n [RFC8126]. This registry has a \"Recommended\" column. The\n registry has been initially populated with psk_ke (0) and\n psk_dhe_ke (1). Both are marked as \"Recommended\". The\n \"Recommended\" column is assigned a value of \"N\" unless explicitly\n requested, and adding a value with a \"Recommended\" value of \"Y\"\n requires Standards Action [RFC8126]. IESG Approval is REQUIRED\n for a Y->N transition.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 108]\n\f\nRFC 8446 TLS August 2018\n\n\n12. References\n\n12.1. Normative References\n\n [DH76] Diffie, W. and M. Hellman, \"New directions in\n cryptography\", IEEE Transactions on Information\n Theory, Vol. 22 No. 6, pp. 644-654,\n DOI 10.1109/TIT.1976.1055638, November 1976.\n\n [ECDSA] American National Standards Institute, \"Public Key\n Cryptography for the Financial Services Industry: The\n Elliptic Curve Digital Signature Algorithm (ECDSA)\",\n ANSI ANS X9.62-2005, November 2005.\n\n [GCM] Dworkin, M., \"Recommendation for Block Cipher Modes of\n Operation: Galois/Counter Mode (GCM) and GMAC\",\n NIST Special Publication 800-38D,\n DOI 10.6028/NIST.SP.800-38D, November 2007.\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC: Keyed-\n Hashing for Message Authentication\", RFC 2104,\n DOI 10.17487/RFC2104, February 1997,\n .\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119,\n DOI 10.17487/RFC2119, March 1997,\n .\n\n [RFC5116] McGrew, D., \"An Interface and Algorithms for Authenticated\n Encryption\", RFC 5116, DOI 10.17487/RFC5116, January 2008,\n .\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, DOI 10.17487/RFC5280, May 2008,\n .\n\n [RFC5705] Rescorla, E., \"Keying Material Exporters for Transport\n Layer Security (TLS)\", RFC 5705, DOI 10.17487/RFC5705,\n March 2010, .\n\n [RFC5756] Turner, S., Brown, D., Yiu, K., Housley, R., and T. Polk,\n \"Updates for RSAES-OAEP and RSASSA-PSS Algorithm\n Parameters\", RFC 5756, DOI 10.17487/RFC5756, January 2010,\n .\n\n\n\n\nRescorla Standards Track [Page 109]\n\f\nRFC 8446 TLS August 2018\n\n\n [RFC5869] Krawczyk, H. and P. Eronen, \"HMAC-based Extract-and-Expand\n Key Derivation Function (HKDF)\", RFC 5869,\n DOI 10.17487/RFC5869, May 2010,\n .\n\n [RFC6066] Eastlake 3rd, D., \"Transport Layer Security (TLS)\n Extensions: Extension Definitions\", RFC 6066,\n DOI 10.17487/RFC6066, January 2011,\n .\n\n [RFC6655] McGrew, D. and D. Bailey, \"AES-CCM Cipher Suites for\n Transport Layer Security (TLS)\", RFC 6655,\n DOI 10.17487/RFC6655, July 2012,\n .\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, DOI 10.17487/RFC6960, June 2013,\n .\n\n [RFC6961] Pettersen, Y., \"The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\", RFC 6961,\n DOI 10.17487/RFC6961, June 2013,\n .\n\n [RFC6962] Laurie, B., Langley, A., and E. Kasper, \"Certificate\n Transparency\", RFC 6962, DOI 10.17487/RFC6962, June 2013,\n .\n\n [RFC6979] Pornin, T., \"Deterministic Usage of the Digital Signature\n Algorithm (DSA) and Elliptic Curve Digital Signature\n Algorithm (ECDSA)\", RFC 6979, DOI 10.17487/RFC6979,\n August 2013, .\n\n [RFC7301] Friedl, S., Popov, A., Langley, A., and E. Stephan,\n \"Transport Layer Security (TLS) Application-Layer Protocol\n Negotiation Extension\", RFC 7301, DOI 10.17487/RFC7301,\n July 2014, .\n\n [RFC7507] Moeller, B. and A. Langley, \"TLS Fallback Signaling Cipher\n Suite Value (SCSV) for Preventing Protocol Downgrade\n Attacks\", RFC 7507, DOI 10.17487/RFC7507, April 2015,\n .\n\n [RFC7748] Langley, A., Hamburg, M., and S. Turner, \"Elliptic Curves\n for Security\", RFC 7748, DOI 10.17487/RFC7748,\n January 2016, .\n\n\n\nRescorla Standards Track [Page 110]\n\f\nRFC 8446 TLS August 2018\n\n\n [RFC7919] Gillmor, D., \"Negotiated Finite Field Diffie-Hellman\n Ephemeral Parameters for Transport Layer Security (TLS)\",\n RFC 7919, DOI 10.17487/RFC7919, August 2016,\n .\n\n [RFC8017] Moriarty, K., Ed., Kaliski, B., Jonsson, J., and A. Rusch,\n \"PKCS #1: RSA Cryptography Specifications Version 2.2\",\n RFC 8017, DOI 10.17487/RFC8017, November 2016,\n .\n\n [RFC8032] Josefsson, S. and I. Liusvaara, \"Edwards-Curve Digital\n Signature Algorithm (EdDSA)\", RFC 8032,\n DOI 10.17487/RFC8032, January 2017,\n .\n\n [RFC8126] Cotton, M., Leiba, B., and T. Narten, \"Guidelines for\n Writing an IANA Considerations Section in RFCs\", BCP 26,\n RFC 8126, DOI 10.17487/RFC8126, June 2017,\n .\n\n [RFC8174] Leiba, B., \"Ambiguity of Uppercase vs Lowercase in\n RFC 2119 Key Words\", BCP 14, RFC 8174,\n DOI 10.17487/RFC8174, May 2017,\n .\n\n [RFC8439] Nir, Y. and A. Langley, \"ChaCha20 and Poly1305 for IETF\n Protocols\", RFC 8439, DOI 10.17487/RFC8439, June 2018,\n .\n\n [SHS] Dang, Q., \"Secure Hash Standard (SHS)\", National Institute\n of Standards and Technology report,\n DOI 10.6028/NIST.FIPS.180-4, August 2015.\n\n [X690] ITU-T, \"Information technology -- ASN.1 encoding rules:\n Specification of Basic Encoding Rules (BER), Canonical\n Encoding Rules (CER) and Distinguished Encoding Rules\n (DER)\", ISO/IEC 8825-1:2015, November 2015.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 111]\n\f\nRFC 8446 TLS August 2018\n\n\n12.2. Informative References\n\n [AEAD-LIMITS]\n Luykx, A. and K. Paterson, \"Limits on Authenticated\n Encryption Use in TLS\", August 2017,\n .\n\n [BBFGKZ16]\n Bhargavan, K., Brzuska, C., Fournet, C., Green, M.,\n Kohlweiss, M., and S. Zanella-Beguelin, \"Downgrade\n Resilience in Key-Exchange Protocols\", Proceedings of IEEE\n Symposium on Security and Privacy (San Jose),\n DOI 10.1109/SP.2016.37, May 2016.\n\n [BBK17] Bhargavan, K., Blanchet, B., and N. Kobeissi, \"Verified\n Models and Reference Implementations for the TLS 1.3\n Standard Candidate\", Proceedings of IEEE Symposium on\n Security and Privacy (San Jose), DOI 10.1109/SP.2017.26,\n May 2017.\n\n [BDFKPPRSZZ16]\n Bhargavan, K., Delignat-Lavaud, A., Fournet, C.,\n Kohlweiss, M., Pan, J., Protzenko, J., Rastogi, A., Swamy,\n N., Zanella-Beguelin, S., and J. Zinzindohoue,\n \"Implementing and Proving the TLS 1.3 Record Layer\",\n Proceedings of IEEE Symposium on Security and Privacy (San\n Jose), May 2017, .\n\n [Ben17a] Benjamin, D., \"Presentation before the TLS WG at\n IETF 100\", November 2017,\n .\n\n [Ben17b] Benjamin, D., \"Additional TLS 1.3 results from Chrome\",\n message to the TLS mailing list, 18 December 2017,\n .\n\n [Blei98] Bleichenbacher, D., \"Chosen Ciphertext Attacks against\n Protocols Based on RSA Encryption Standard PKCS #1\",\n Proceedings of CRYPTO '98, 1998.\n\n [BMMRT15] Badertscher, C., Matt, C., Maurer, U., Rogaway, P., and B.\n Tackmann, \"Augmented Secure Channels and the Goal of the\n TLS 1.3 Record Layer\", ProvSec 2015, September 2015,\n .\n\n\n\n\n\nRescorla Standards Track [Page 112]\n\f\nRFC 8446 TLS August 2018\n\n\n [BT16] Bellare, M. and B. Tackmann, \"The Multi-User Security of\n Authenticated Encryption: AES-GCM in TLS 1.3\", Proceedings\n of CRYPTO 2016, July 2016,\n .\n\n [CCG16] Cohn-Gordon, K., Cremers, C., and L. Garratt, \"On\n Post-compromise Security\", IEEE Computer Security\n Foundations Symposium, DOI 10.1109/CSF.2016.19, July 2015.\n\n [CHECKOWAY]\n Checkoway, S., Maskiewicz, J., Garman, C., Fried, J.,\n Cohney, S., Green, M., Heninger, N., Weinmann, R.,\n Rescorla, E., and H. Shacham, \"A Systematic Analysis of\n the Juniper Dual EC Incident\", Proceedings of the 2016 ACM\n SIGSAC Conference on Computer and Communications Security\n - CCS '16, DOI 10.1145/2976749.2978395, October 2016.\n\n [CHHSV17] Cremers, C., Horvat, M., Hoyland, J., Scott, S., and T.\n van der Merwe, \"Awkward Handshake: Possible mismatch of\n client/server view on client authentication in\n post-handshake mode in Revision 18\", message to the TLS\n mailing list, 10 February 2017, .\n\n [CHSV16] Cremers, C., Horvat, M., Scott, S., and T. van der Merwe,\n \"Automated Analysis and Verification of TLS 1.3: 0-RTT,\n Resumption and Delayed Authentication\", Proceedings of\n IEEE Symposium on Security and Privacy (San Jose),\n DOI 10.1109/SP.2016.35, May 2016,\n .\n\n [CK01] Canetti, R. and H. Krawczyk, \"Analysis of Key-Exchange\n Protocols and Their Use for Building Secure Channels\",\n Proceedings of Eurocrypt 2001,\n DOI 10.1007/3-540-44987-6_28, April 2001.\n\n [CLINIC] Miller, B., Huang, L., Joseph, A., and J. Tygar, \"I Know\n Why You Went to the Clinic: Risks and Realization of HTTPS\n Traffic Analysis\", Privacy Enhancing Technologies, pp.\n 143-163, DOI 10.1007/978-3-319-08506-7_8, 2014.\n\n [DFGS15] Dowling, B., Fischlin, M., Guenther, F., and D. Stebila,\n \"A Cryptographic Analysis of the TLS 1.3 Handshake\n Protocol Candidates\", Proceedings of ACM CCS 2015,\n October 2015, .\n\n\n\n\n\n\nRescorla Standards Track [Page 113]\n\f\nRFC 8446 TLS August 2018\n\n\n [DFGS16] Dowling, B., Fischlin, M., Guenther, F., and D. Stebila,\n \"A Cryptographic Analysis of the TLS 1.3 Full and\n Pre-shared Key Handshake Protocol\", TRON 2016,\n February 2016, .\n\n [DOW92] Diffie, W., van Oorschot, P., and M. Wiener,\n \"Authentication and authenticated key exchanges\", Designs,\n Codes and Cryptography, DOI 10.1007/BF00124891, June 1992.\n\n [DSS] National Institute of Standards and Technology, U.S.\n Department of Commerce, \"Digital Signature Standard\n (DSS)\", NIST FIPS PUB 186-4, DOI 10.6028/NIST.FIPS.186-4,\n July 2013.\n\n [FG17] Fischlin, M. and F. Guenther, \"Replay Attacks on Zero\n Round-Trip Time: The Case of the TLS 1.3 Handshake\n Candidates\", Proceedings of EuroS&P 2017, April 2017,\n .\n\n [FGSW16] Fischlin, M., Guenther, F., Schmidt, B., and B. Warinschi,\n \"Key Confirmation in Key Exchange: A Formal Treatment and\n Implications for TLS 1.3\", Proceedings of IEEE Symposium\n on Security and Privacy (San Jose),\n DOI 10.1109/SP.2016.34, May 2016,\n .\n\n [FW15] Weimer, F., \"Factoring RSA Keys With TLS Perfect Forward\n Secrecy\", September 2015.\n\n [HCJC16] Husak, M., Cermak, M., Jirsik, T., and P. Celeda, \"HTTPS\n traffic analysis and client identification using passive\n SSL/TLS fingerprinting\", EURASIP Journal on Information\n Security, Vol. 2016, DOI 10.1186/s13635-016-0030-7,\n February 2016.\n\n [HGFS15] Hlauschek, C., Gruber, M., Fankhauser, F., and C. Schanes,\n \"Prying Open Pandora's Box: KCI Attacks against TLS\",\n Proceedings of USENIX Workshop on Offensive Technologies,\n August 2015.\n\n [IEEE1363]\n IEEE, \"IEEE Standard Specifications for Public Key\n Cryptography\", IEEE Std. 1363-2000,\n DOI 10.1109/IEEESTD.2000.92292.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 114]\n\f\nRFC 8446 TLS August 2018\n\n\n [JSS15] Jager, T., Schwenk, J., and J. Somorovsky, \"On the\n Security of TLS 1.3 and QUIC Against Weaknesses in PKCS#1\n v1.5 Encryption\", Proceedings of ACM CCS 2015,\n DOI 10.1145/2810103.2813657, October 2015,\n .\n\n [KEYAGREEMENT]\n Barker, E., Chen, L., Roginsky, A., Vassilev, A., and R.\n Davis, \"Recommendation for Pair-Wise Key Establishment\n Schemes Using Discrete Logarithm Cryptography\", National\n Institute of Standards and Technology,\n DOI 10.6028/NIST.SP.800-56Ar3, April 2018.\n\n [Kraw10] Krawczyk, H., \"Cryptographic Extraction and Key\n Derivation: The HKDF Scheme\", Proceedings of CRYPTO 2010,\n August 2010, .\n\n [Kraw16] Krawczyk, H., \"A Unilateral-to-Mutual Authentication\n Compiler for Key Exchange (with Applications to Client\n Authentication in TLS 1.3\", Proceedings of ACM CCS 2016,\n October 2016, .\n\n [KW16] Krawczyk, H. and H. Wee, \"The OPTLS Protocol and TLS 1.3\",\n Proceedings of EuroS&P 2016, March 2016,\n .\n\n [LXZFH16] Li, X., Xu, J., Zhang, Z., Feng, D., and H. Hu, \"Multiple\n Handshakes Security of TLS 1.3 Candidates\", Proceedings of\n IEEE Symposium on Security and Privacy (San Jose),\n DOI 10.1109/SP.2016.36, May 2016,\n .\n\n [Mac17] MacCarthaigh, C., \"Security Review of TLS1.3 0-RTT\",\n March 2017, .\n\n [PS18] Patton, C. and T. Shrimpton, \"Partially specified\n channels: The TLS 1.3 record layer without elision\", 2018,\n .\n\n [PSK-FINISHED]\n Scott, S., Cremers, C., Horvat, M., and T. van der Merwe,\n \"Revision 10: possible attack if client authentication is\n allowed during PSK\", message to the TLS mailing list,\n 31 October 2015, .\n\n\n\n\nRescorla Standards Track [Page 115]\n\f\nRFC 8446 TLS August 2018\n\n\n [REKEY] Abdalla, M. and M. Bellare, \"Increasing the Lifetime of a\n Key: A Comparative Analysis of the Security of Re-keying\n Techniques\", ASIACRYPT 2000, DOI 10.1007/3-540-44448-3_42,\n October 2000.\n\n [Res17a] Rescorla, E., \"Preliminary data on Firefox TLS 1.3\n Middlebox experiment\", message to the TLS mailing list,\n 5 December 2017, .\n\n [Res17b] Rescorla, E., \"More compatibility measurement results\",\n message to the TLS mailing list, 22 December 2017,\n .\n\n [RFC3552] Rescorla, E. and B. Korver, \"Guidelines for Writing RFC\n Text on Security Considerations\", BCP 72, RFC 3552,\n DOI 10.17487/RFC3552, July 2003,\n .\n\n [RFC4086] Eastlake 3rd, D., Schiller, J., and S. Crocker,\n \"Randomness Requirements for Security\", BCP 106, RFC 4086,\n DOI 10.17487/RFC4086, June 2005,\n .\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346,\n DOI 10.17487/RFC4346, April 2006,\n .\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, DOI 10.17487/RFC4366, April 2006,\n .\n\n [RFC4492] Blake-Wilson, S., Bolyard, N., Gupta, V., Hawk, C., and B.\n Moeller, \"Elliptic Curve Cryptography (ECC) Cipher Suites\n for Transport Layer Security (TLS)\", RFC 4492,\n DOI 10.17487/RFC4492, May 2006,\n .\n\n [RFC5077] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 5077, DOI 10.17487/RFC5077,\n January 2008, .\n\n\n\n\n\n\nRescorla Standards Track [Page 116]\n\f\nRFC 8446 TLS August 2018\n\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246,\n DOI 10.17487/RFC5246, August 2008,\n .\n\n [RFC5764] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for the Secure\n Real-time Transport Protocol (SRTP)\", RFC 5764,\n DOI 10.17487/RFC5764, May 2010,\n .\n\n [RFC5929] Altman, J., Williams, N., and L. Zhu, \"Channel Bindings\n for TLS\", RFC 5929, DOI 10.17487/RFC5929, July 2010,\n .\n\n [RFC6091] Mavrogiannopoulos, N. and D. Gillmor, \"Using OpenPGP Keys\n for Transport Layer Security (TLS) Authentication\",\n RFC 6091, DOI 10.17487/RFC6091, February 2011,\n .\n\n [RFC6101] Freier, A., Karlton, P., and P. Kocher, \"The Secure\n Sockets Layer (SSL) Protocol Version 3.0\", RFC 6101,\n DOI 10.17487/RFC6101, August 2011,\n .\n\n [RFC6176] Turner, S. and T. Polk, \"Prohibiting Secure Sockets Layer\n (SSL) Version 2.0\", RFC 6176, DOI 10.17487/RFC6176,\n March 2011, .\n\n [RFC6347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security Version 1.2\", RFC 6347, DOI 10.17487/RFC6347,\n January 2012, .\n\n [RFC6520] Seggelmann, R., Tuexen, M., and M. Williams, \"Transport\n Layer Security (TLS) and Datagram Transport Layer Security\n (DTLS) Heartbeat Extension\", RFC 6520,\n DOI 10.17487/RFC6520, February 2012,\n .\n\n [RFC7230] Fielding, R., Ed. and J. Reschke, Ed., \"Hypertext Transfer\n Protocol (HTTP/1.1): Message Syntax and Routing\",\n RFC 7230, DOI 10.17487/RFC7230, June 2014,\n .\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 117]\n\f\nRFC 8446 TLS August 2018\n\n\n [RFC7250] Wouters, P., Ed., Tschofenig, H., Ed., Gilmore, J.,\n Weiler, S., and T. Kivinen, \"Using Raw Public Keys in\n Transport Layer Security (TLS) and Datagram Transport\n Layer Security (DTLS)\", RFC 7250, DOI 10.17487/RFC7250,\n June 2014, .\n\n [RFC7465] Popov, A., \"Prohibiting RC4 Cipher Suites\", RFC 7465,\n DOI 10.17487/RFC7465, February 2015,\n .\n\n [RFC7568] Barnes, R., Thomson, M., Pironti, A., and A. Langley,\n \"Deprecating Secure Sockets Layer Version 3.0\", RFC 7568,\n DOI 10.17487/RFC7568, June 2015,\n .\n\n [RFC7627] Bhargavan, K., Ed., Delignat-Lavaud, A., Pironti, A.,\n Langley, A., and M. Ray, \"Transport Layer Security (TLS)\n Session Hash and Extended Master Secret Extension\",\n RFC 7627, DOI 10.17487/RFC7627, September 2015,\n .\n\n [RFC7685] Langley, A., \"A Transport Layer Security (TLS) ClientHello\n Padding Extension\", RFC 7685, DOI 10.17487/RFC7685,\n October 2015, .\n\n [RFC7924] Santesson, S. and H. Tschofenig, \"Transport Layer Security\n (TLS) Cached Information Extension\", RFC 7924,\n DOI 10.17487/RFC7924, July 2016,\n .\n\n [RFC8305] Schinazi, D. and T. Pauly, \"Happy Eyeballs Version 2:\n Better Connectivity Using Concurrency\", RFC 8305,\n DOI 10.17487/RFC8305, December 2017,\n .\n\n [RFC8422] Nir, Y., Josefsson, S., and M. Pegourie-Gonnard, \"Elliptic\n Curve Cryptography (ECC) Cipher Suites for Transport Layer\n Security (TLS) Versions 1.2 and Earlier\", RFC 8422,\n DOI 10.17487/RFC8422, August 2018,\n .\n\n [RFC8447] Salowey, J. and S. Turner, \"IANA Registry Updates for TLS\n and DTLS\", RFC 8447, DOI 10.17487/RFC8447, August 2018,\n .\n\n [RFC8449] Thomson, M., \"Record Size Limit Extension for TLS\",\n RFC 8449, DOI 10.17487/RFC8449, August 2018,\n .\n\n\n\nRescorla Standards Track [Page 118]\n\f\nRFC 8446 TLS August 2018\n\n\n [RSA] Rivest, R., Shamir, A., and L. Adleman, \"A Method for\n Obtaining Digital Signatures and Public-Key\n Cryptosystems\", Communications of the ACM, Vol. 21 No. 2,\n pp. 120-126, DOI 10.1145/359340.359342, February 1978.\n\n [SIGMA] Krawczyk, H., \"SIGMA: The 'SIGn-and-MAc' Approach to\n Authenticated Diffie-Hellman and its Use in the IKE\n Protocols\", Proceedings of CRYPTO 2003,\n DOI 10.1007/978-3-540-45146-4_24, August 2003.\n\n [SLOTH] Bhargavan, K. and G. Leurent, \"Transcript Collision\n Attacks: Breaking Authentication in TLS, IKE, and SSH\",\n Network and Distributed System Security\n Symposium (NDSS 2016), DOI 10.14722/ndss.2016.23418,\n February 2016.\n\n [SSL2] Hickman, K., \"The SSL Protocol\", February 1995.\n\n [TIMING] Boneh, D. and D. Brumley, \"Remote Timing Attacks Are\n Practical\", USENIX Security Symposium, August 2003.\n\n [TLS13-TRACES]\n Thomson, M., \"Example Handshake Traces for TLS 1.3\", Work\n in Progress, draft-ietf-tls-tls13-vectors-06, July 2018.\n\n [X501] ITU-T, \"Information Technology - Open Systems\n Interconnection - The Directory: Models\", ITU-T X.501,\n October 2016, .\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 119]\n\f\nRFC 8446 TLS August 2018\n\n\nAppendix A. State Machine\n\n This appendix provides a summary of the legal state transitions for\n the client and server handshakes. State names (in all capitals,\n e.g., START) have no formal meaning but are provided for ease of\n comprehension. Actions which are taken only in certain circumstances\n are indicated in []. The notation \"K_{send,recv} = foo\" means \"set\n the send/recv key to the given key\".\n\nA.1. Client\n\n START <----+\n Send ClientHello | | Recv HelloRetryRequest\n [K_send = early data] | |\n v |\n / WAIT_SH ----+\n | | Recv ServerHello\n | | K_recv = handshake\n Can | V\n send | WAIT_EE\n early | | Recv EncryptedExtensions\n data | +--------+--------+\n | Using | | Using certificate\n | PSK | v\n | | WAIT_CERT_CR\n | | Recv | | Recv CertificateRequest\n | | Certificate | v\n | | | WAIT_CERT\n | | | | Recv Certificate\n | | v v\n | | WAIT_CV\n | | | Recv CertificateVerify\n | +> WAIT_FINISHED <+\n | | Recv Finished\n \\ | [Send EndOfEarlyData]\n | K_send = handshake\n | [Send Certificate [+ CertificateVerify]]\n Can send | Send Finished\n app data --> | K_send = K_recv = application\n after here v\n CONNECTED\n\n Note that with the transitions as shown above, clients may send\n alerts that derive from post-ServerHello messages in the clear or\n with the early data keys. If clients need to send such alerts, they\n SHOULD first rekey to the handshake keys if possible.\n\n\n\n\n\nRescorla Standards Track [Page 120]\n\f\nRFC 8446 TLS August 2018\n\n\nA.2. Server\n\n START <-----+\n Recv ClientHello | | Send HelloRetryRequest\n v |\n RECVD_CH ----+\n | Select parameters\n v\n NEGOTIATED\n | Send ServerHello\n | K_send = handshake\n | Send EncryptedExtensions\n | [Send CertificateRequest]\n Can send | [Send Certificate + CertificateVerify]\n app data | Send Finished\n after --> | K_send = application\n here +--------+--------+\n No 0-RTT | | 0-RTT\n | |\n K_recv = handshake | | K_recv = early data\n [Skip decrypt errors] | +------> WAIT_EOED -+\n | | Recv | | Recv EndOfEarlyData\n | | early data | | K_recv = handshake\n | +------------+ |\n | |\n +> WAIT_FLIGHT2 <--------+\n |\n +--------+--------+\n No auth | | Client auth\n | |\n | v\n | WAIT_CERT\n | Recv | | Recv Certificate\n | empty | v\n | Certificate | WAIT_CV\n | | | Recv\n | v | CertificateVerify\n +-> WAIT_FINISHED <---+\n | Recv Finished\n | K_recv = application\n v\n CONNECTED\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 121]\n\f\nRFC 8446 TLS August 2018\n\n\nAppendix B. Protocol Data Structures and Constant Values\n\n This appendix provides the normative protocol types and the\n definitions for constants. Values listed as \"_RESERVED\" were used in\n previous versions of TLS and are listed here for completeness.\n TLS 1.3 implementations MUST NOT send them but might receive them\n from older TLS implementations.\n\nB.1. Record Layer\n\n enum {\n invalid(0),\n change_cipher_spec(20),\n alert(21),\n handshake(22),\n application_data(23),\n heartbeat(24), /* RFC 6520 */\n (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion legacy_record_version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n struct {\n opaque content[TLSPlaintext.length];\n ContentType type;\n uint8 zeros[length_of_padding];\n } TLSInnerPlaintext;\n\n struct {\n ContentType opaque_type = application_data; /* 23 */\n ProtocolVersion legacy_record_version = 0x0303; /* TLS v1.2 */\n uint16 length;\n opaque encrypted_record[TLSCiphertext.length];\n } TLSCiphertext;\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 122]\n\f\nRFC 8446 TLS August 2018\n\n\nB.2. Alert Messages\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure_RESERVED(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n inappropriate_fallback(86),\n user_canceled(90),\n no_renegotiation_RESERVED(100),\n missing_extension(109),\n unsupported_extension(110),\n certificate_unobtainable_RESERVED(111),\n unrecognized_name(112),\n bad_certificate_status_response(113),\n bad_certificate_hash_value_RESERVED(114),\n unknown_psk_identity(115),\n certificate_required(116),\n no_application_protocol(120),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n\n\n\n\nRescorla Standards Track [Page 123]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3. Handshake Protocol\n\n enum {\n hello_request_RESERVED(0),\n client_hello(1),\n server_hello(2),\n hello_verify_request_RESERVED(3),\n new_session_ticket(4),\n end_of_early_data(5),\n hello_retry_request_RESERVED(6),\n encrypted_extensions(8),\n certificate(11),\n server_key_exchange_RESERVED(12),\n certificate_request(13),\n server_hello_done_RESERVED(14),\n certificate_verify(15),\n client_key_exchange_RESERVED(16),\n finished(20),\n certificate_url_RESERVED(21),\n certificate_status_RESERVED(22),\n supplemental_data_RESERVED(23),\n key_update(24),\n message_hash(254),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (Handshake.msg_type) {\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case end_of_early_data: EndOfEarlyData;\n case encrypted_extensions: EncryptedExtensions;\n case certificate_request: CertificateRequest;\n case certificate: Certificate;\n case certificate_verify: CertificateVerify;\n case finished: Finished;\n case new_session_ticket: NewSessionTicket;\n case key_update: KeyUpdate;\n };\n } Handshake;\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 124]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.1. Key Exchange Messages\n\n uint16 ProtocolVersion;\n opaque Random[32];\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id<0..32>;\n CipherSuite cipher_suites<2..2^16-2>;\n opaque legacy_compression_methods<1..2^8-1>;\n Extension extensions<8..2^16-1>;\n } ClientHello;\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id_echo<0..32>;\n CipherSuite cipher_suite;\n uint8 legacy_compression_method = 0;\n Extension extensions<6..2^16-1>;\n } ServerHello;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 125]\n\f\nRFC 8446 TLS August 2018\n\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n server_name(0), /* RFC 6066 */\n max_fragment_length(1), /* RFC 6066 */\n status_request(5), /* RFC 6066 */\n supported_groups(10), /* RFC 8422, 7919 */\n signature_algorithms(13), /* RFC 8446 */\n use_srtp(14), /* RFC 5764 */\n heartbeat(15), /* RFC 6520 */\n application_layer_protocol_negotiation(16), /* RFC 7301 */\n signed_certificate_timestamp(18), /* RFC 6962 */\n client_certificate_type(19), /* RFC 7250 */\n server_certificate_type(20), /* RFC 7250 */\n padding(21), /* RFC 7685 */\n RESERVED(40), /* Used but never\n assigned */\n pre_shared_key(41), /* RFC 8446 */\n early_data(42), /* RFC 8446 */\n supported_versions(43), /* RFC 8446 */\n cookie(44), /* RFC 8446 */\n psk_key_exchange_modes(45), /* RFC 8446 */\n RESERVED(46), /* Used but never\n assigned */\n certificate_authorities(47), /* RFC 8446 */\n oid_filters(48), /* RFC 8446 */\n post_handshake_auth(49), /* RFC 8446 */\n signature_algorithms_cert(50), /* RFC 8446 */\n key_share(51), /* RFC 8446 */\n (65535)\n } ExtensionType;\n\n struct {\n NamedGroup group;\n opaque key_exchange<1..2^16-1>;\n } KeyShareEntry;\n\n struct {\n KeyShareEntry client_shares<0..2^16-1>;\n } KeyShareClientHello;\n\n struct {\n NamedGroup selected_group;\n } KeyShareHelloRetryRequest;\n\n\n\n\nRescorla Standards Track [Page 126]\n\f\nRFC 8446 TLS August 2018\n\n\n struct {\n KeyShareEntry server_share;\n } KeyShareServerHello;\n\n struct {\n uint8 legacy_form = 4;\n opaque X[coordinate_length];\n opaque Y[coordinate_length];\n } UncompressedPointRepresentation;\n\n enum { psk_ke(0), psk_dhe_ke(1), (255) } PskKeyExchangeMode;\n\n struct {\n PskKeyExchangeMode ke_modes<1..255>;\n } PskKeyExchangeModes;\n\n struct {} Empty;\n\n struct {\n select (Handshake.msg_type) {\n case new_session_ticket: uint32 max_early_data_size;\n case client_hello: Empty;\n case encrypted_extensions: Empty;\n };\n } EarlyDataIndication;\n\n struct {\n opaque identity<1..2^16-1>;\n uint32 obfuscated_ticket_age;\n } PskIdentity;\n\n opaque PskBinderEntry<32..255>;\n\n struct {\n PskIdentity identities<7..2^16-1>;\n PskBinderEntry binders<33..2^16-1>;\n } OfferedPsks;\n\n struct {\n select (Handshake.msg_type) {\n case client_hello: OfferedPsks;\n case server_hello: uint16 selected_identity;\n };\n } PreSharedKeyExtension;\n\n\n\n\n\n\n\nRescorla Standards Track [Page 127]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.1.1. Version Extension\n\n struct {\n select (Handshake.msg_type) {\n case client_hello:\n ProtocolVersion versions<2..254>;\n\n case server_hello: /* and HelloRetryRequest */\n ProtocolVersion selected_version;\n };\n } SupportedVersions;\n\nB.3.1.2. Cookie Extension\n\n struct {\n opaque cookie<1..2^16-1>;\n } Cookie;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 128]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.1.3. Signature Algorithm Extension\n\n enum {\n /* RSASSA-PKCS1-v1_5 algorithms */\n rsa_pkcs1_sha256(0x0401),\n rsa_pkcs1_sha384(0x0501),\n rsa_pkcs1_sha512(0x0601),\n\n /* ECDSA algorithms */\n ecdsa_secp256r1_sha256(0x0403),\n ecdsa_secp384r1_sha384(0x0503),\n ecdsa_secp521r1_sha512(0x0603),\n\n /* RSASSA-PSS algorithms with public key OID rsaEncryption */\n rsa_pss_rsae_sha256(0x0804),\n rsa_pss_rsae_sha384(0x0805),\n rsa_pss_rsae_sha512(0x0806),\n\n /* EdDSA algorithms */\n ed25519(0x0807),\n ed448(0x0808),\n\n /* RSASSA-PSS algorithms with public key OID RSASSA-PSS */\n rsa_pss_pss_sha256(0x0809),\n rsa_pss_pss_sha384(0x080a),\n rsa_pss_pss_sha512(0x080b),\n\n /* Legacy algorithms */\n rsa_pkcs1_sha1(0x0201),\n ecdsa_sha1(0x0203),\n\n /* Reserved Code Points */\n obsolete_RESERVED(0x0000..0x0200),\n dsa_sha1_RESERVED(0x0202),\n obsolete_RESERVED(0x0204..0x0400),\n dsa_sha256_RESERVED(0x0402),\n obsolete_RESERVED(0x0404..0x0500),\n dsa_sha384_RESERVED(0x0502),\n obsolete_RESERVED(0x0504..0x0600),\n dsa_sha512_RESERVED(0x0602),\n obsolete_RESERVED(0x0604..0x06FF),\n private_use(0xFE00..0xFFFF),\n (0xFFFF)\n } SignatureScheme;\n\n struct {\n SignatureScheme supported_signature_algorithms<2..2^16-2>;\n } SignatureSchemeList;\n\n\n\nRescorla Standards Track [Page 129]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.1.4. Supported Groups Extension\n\n enum {\n unallocated_RESERVED(0x0000),\n\n /* Elliptic Curve Groups (ECDHE) */\n obsolete_RESERVED(0x0001..0x0016),\n secp256r1(0x0017), secp384r1(0x0018), secp521r1(0x0019),\n obsolete_RESERVED(0x001A..0x001C),\n x25519(0x001D), x448(0x001E),\n\n /* Finite Field Groups (DHE) */\n ffdhe2048(0x0100), ffdhe3072(0x0101), ffdhe4096(0x0102),\n ffdhe6144(0x0103), ffdhe8192(0x0104),\n\n /* Reserved Code Points */\n ffdhe_private_use(0x01FC..0x01FF),\n ecdhe_private_use(0xFE00..0xFEFF),\n obsolete_RESERVED(0xFF01..0xFF02),\n (0xFFFF)\n } NamedGroup;\n\n struct {\n NamedGroup named_group_list<2..2^16-1>;\n } NamedGroupList;\n\n Values within \"obsolete_RESERVED\" ranges are used in previous\n versions of TLS and MUST NOT be offered or negotiated by TLS 1.3\n implementations. The obsolete curves have various known/theoretical\n weaknesses or have had very little usage, in some cases only due to\n unintentional server configuration issues. They are no longer\n considered appropriate for general use and should be assumed to be\n potentially unsafe. The set of curves specified here is sufficient\n for interoperability with all currently deployed and properly\n configured TLS implementations.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 130]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.2. Server Parameters Messages\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n DistinguishedName authorities<3..2^16-1>;\n } CertificateAuthoritiesExtension;\n\n struct {\n opaque certificate_extension_oid<1..2^8-1>;\n opaque certificate_extension_values<0..2^16-1>;\n } OIDFilter;\n\n struct {\n OIDFilter filters<0..2^16-1>;\n } OIDFilterExtension;\n\n struct {} PostHandshakeAuth;\n\n struct {\n Extension extensions<0..2^16-1>;\n } EncryptedExtensions;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n Extension extensions<2..2^16-1>;\n } CertificateRequest;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 131]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.3. Authentication Messages\n\n enum {\n X509(0),\n OpenPGP_RESERVED(1),\n RawPublicKey(2),\n (255)\n } CertificateType;\n\n struct {\n select (certificate_type) {\n case RawPublicKey:\n /* From RFC 7250 ASN.1_subjectPublicKeyInfo */\n opaque ASN1_subjectPublicKeyInfo<1..2^24-1>;\n\n case X509:\n opaque cert_data<1..2^24-1>;\n };\n Extension extensions<0..2^16-1>;\n } CertificateEntry;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n CertificateEntry certificate_list<0..2^24-1>;\n } Certificate;\n\n struct {\n SignatureScheme algorithm;\n opaque signature<0..2^16-1>;\n } CertificateVerify;\n\n struct {\n opaque verify_data[Hash.length];\n } Finished;\n\nB.3.4. Ticket Establishment\n\n struct {\n uint32 ticket_lifetime;\n uint32 ticket_age_add;\n opaque ticket_nonce<0..255>;\n opaque ticket<1..2^16-1>;\n Extension extensions<0..2^16-2>;\n } NewSessionTicket;\n\n\n\n\n\n\n\nRescorla Standards Track [Page 132]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.5. Updating Keys\n\n struct {} EndOfEarlyData;\n\n enum {\n update_not_requested(0), update_requested(1), (255)\n } KeyUpdateRequest;\n\n struct {\n KeyUpdateRequest request_update;\n } KeyUpdate;\n\nB.4. Cipher Suites\n\n A symmetric cipher suite defines the pair of the AEAD algorithm and\n hash algorithm to be used with HKDF. Cipher suite names follow the\n naming convention:\n\n CipherSuite TLS_AEAD_HASH = VALUE;\n\n +-----------+------------------------------------------------+\n | Component | Contents |\n +-----------+------------------------------------------------+\n | TLS | The string \"TLS\" |\n | | |\n | AEAD | The AEAD algorithm used for record protection |\n | | |\n | HASH | The hash algorithm used with HKDF |\n | | |\n | VALUE | The two-byte ID assigned for this cipher suite |\n +-----------+------------------------------------------------+\n\n This specification defines the following cipher suites for use with\n TLS 1.3.\n\n +------------------------------+-------------+\n | Description | Value |\n +------------------------------+-------------+\n | TLS_AES_128_GCM_SHA256 | {0x13,0x01} |\n | | |\n | TLS_AES_256_GCM_SHA384 | {0x13,0x02} |\n | | |\n | TLS_CHACHA20_POLY1305_SHA256 | {0x13,0x03} |\n | | |\n | TLS_AES_128_CCM_SHA256 | {0x13,0x04} |\n | | |\n | TLS_AES_128_CCM_8_SHA256 | {0x13,0x05} |\n +------------------------------+-------------+\n\n\n\nRescorla Standards Track [Page 133]\n\f\nRFC 8446 TLS August 2018\n\n\n The corresponding AEAD algorithms AEAD_AES_128_GCM, AEAD_AES_256_GCM,\n and AEAD_AES_128_CCM are defined in [RFC5116].\n AEAD_CHACHA20_POLY1305 is defined in [RFC8439]. AEAD_AES_128_CCM_8\n is defined in [RFC6655]. The corresponding hash algorithms are\n defined in [SHS].\n\n Although TLS 1.3 uses the same cipher suite space as previous\n versions of TLS, TLS 1.3 cipher suites are defined differently, only\n specifying the symmetric ciphers, and cannot be used for TLS 1.2.\n Similarly, cipher suites for TLS 1.2 and lower cannot be used with\n TLS 1.3.\n\n New cipher suite values are assigned by IANA as described in\n Section 11.\n\nAppendix C. Implementation Notes\n\n The TLS protocol cannot prevent many common security mistakes. This\n appendix provides several recommendations to assist implementors.\n [TLS13-TRACES] provides test vectors for TLS 1.3 handshakes.\n\nC.1. Random Number Generation and Seeding\n\n TLS requires a cryptographically secure pseudorandom number generator\n (CSPRNG). In most cases, the operating system provides an\n appropriate facility such as /dev/urandom, which should be used\n absent other (e.g., performance) concerns. It is RECOMMENDED to use\n an existing CSPRNG implementation in preference to crafting a new\n one. Many adequate cryptographic libraries are already available\n under favorable license terms. Should those prove unsatisfactory,\n [RFC4086] provides guidance on the generation of random values.\n\n TLS uses random values (1) in public protocol fields such as the\n public Random values in the ClientHello and ServerHello and (2) to\n generate keying material. With a properly functioning CSPRNG, this\n does not present a security problem, as it is not feasible to\n determine the CSPRNG state from its output. However, with a broken\n CSPRNG, it may be possible for an attacker to use the public output\n to determine the CSPRNG internal state and thereby predict the keying\n material, as documented in [CHECKOWAY]. Implementations can provide\n extra security against this form of attack by using separate CSPRNGs\n to generate public and private values.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 134]\n\f\nRFC 8446 TLS August 2018\n\n\nC.2. Certificates and Authentication\n\n Implementations are responsible for verifying the integrity of\n certificates and should generally support certificate revocation\n messages. Absent a specific indication from an application profile,\n certificates should always be verified to ensure proper signing by a\n trusted certificate authority (CA). The selection and addition of\n trust anchors should be done very carefully. Users should be able to\n view information about the certificate and trust anchor.\n Applications SHOULD also enforce minimum and maximum key sizes. For\n example, certification paths containing keys or signatures weaker\n than 2048-bit RSA or 224-bit ECDSA are not appropriate for secure\n applications.\n\nC.3. Implementation Pitfalls\n\n Implementation experience has shown that certain parts of earlier TLS\n specifications are not easy to understand and have been a source of\n interoperability and security problems. Many of these areas have\n been clarified in this document, but this appendix contains a short\n list of the most important things that require special attention from\n implementors.\n\n TLS protocol issues:\n\n - Do you correctly handle handshake messages that are fragmented to\n multiple TLS records (see Section 5.1)? Do you correctly handle\n corner cases like a ClientHello that is split into several small\n fragments? Do you fragment handshake messages that exceed the\n maximum fragment size? In particular, the Certificate and\n CertificateRequest handshake messages can be large enough to\n require fragmentation.\n\n - Do you ignore the TLS record layer version number in all\n unencrypted TLS records (see Appendix D)?\n\n - Have you ensured that all support for SSL, RC4, EXPORT ciphers,\n and MD5 (via the \"signature_algorithms\" extension) is completely\n removed from all possible configurations that support TLS 1.3 or\n later, and that attempts to use these obsolete capabilities fail\n correctly (see Appendix D)?\n\n - Do you handle TLS extensions in ClientHellos correctly, including\n unknown extensions?\n\n\n\n\n\n\n\nRescorla Standards Track [Page 135]\n\f\nRFC 8446 TLS August 2018\n\n\n - When the server has requested a client certificate but no suitable\n certificate is available, do you correctly send an empty\n Certificate message, instead of omitting the whole message (see\n Section 4.4.2)?\n\n - When processing the plaintext fragment produced by AEAD-Decrypt\n and scanning from the end for the ContentType, do you avoid\n scanning past the start of the cleartext in the event that the\n peer has sent a malformed plaintext of all zeros?\n\n - Do you properly ignore unrecognized cipher suites (Section 4.1.2),\n hello extensions (Section 4.2), named groups (Section 4.2.7), key\n shares (Section 4.2.8), supported versions (Section 4.2.1), and\n signature algorithms (Section 4.2.3) in the ClientHello?\n\n - As a server, do you send a HelloRetryRequest to clients which\n support a compatible (EC)DHE group but do not predict it in the\n \"key_share\" extension? As a client, do you correctly handle a\n HelloRetryRequest from the server?\n\n Cryptographic details:\n\n - What countermeasures do you use to prevent timing attacks\n [TIMING]?\n\n - When using Diffie-Hellman key exchange, do you correctly preserve\n leading zero bytes in the negotiated key (see Section 7.4.1)?\n\n - Does your TLS client check that the Diffie-Hellman parameters sent\n by the server are acceptable (see Section 4.2.8.1)?\n\n - Do you use a strong and, most importantly, properly seeded random\n number generator (see Appendix C.1) when generating Diffie-Hellman\n private values, the ECDSA \"k\" parameter, and other security-\n critical values? It is RECOMMENDED that implementations implement\n \"deterministic ECDSA\" as specified in [RFC6979].\n\n - Do you zero-pad Diffie-Hellman public key values and shared\n secrets to the group size (see Section 4.2.8.1 and Section 7.4.1)?\n\n - Do you verify signatures after making them, to protect against\n RSA-CRT key leaks [FW15]?\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 136]\n\f\nRFC 8446 TLS August 2018\n\n\nC.4. Client Tracking Prevention\n\n Clients SHOULD NOT reuse a ticket for multiple connections. Reuse of\n a ticket allows passive observers to correlate different connections.\n Servers that issue tickets SHOULD offer at least as many tickets as\n the number of connections that a client might use; for example, a web\n browser using HTTP/1.1 [RFC7230] might open six connections to a\n server. Servers SHOULD issue new tickets with every connection.\n This ensures that clients are always able to use a new ticket when\n creating a new connection.\n\nC.5. Unauthenticated Operation\n\n Previous versions of TLS offered explicitly unauthenticated cipher\n suites based on anonymous Diffie-Hellman. These modes have been\n deprecated in TLS 1.3. However, it is still possible to negotiate\n parameters that do not provide verifiable server authentication by\n several methods, including:\n\n - Raw public keys [RFC7250].\n\n - Using a public key contained in a certificate but without\n validation of the certificate chain or any of its contents.\n\n Either technique used alone is vulnerable to man-in-the-middle\n attacks and therefore unsafe for general use. However, it is also\n possible to bind such connections to an external authentication\n mechanism via out-of-band validation of the server's public key,\n trust on first use, or a mechanism such as channel bindings (though\n the channel bindings described in [RFC5929] are not defined for\n TLS 1.3). If no such mechanism is used, then the connection has no\n protection against active man-in-the-middle attack; applications\n MUST NOT use TLS in such a way absent explicit configuration or a\n specific application profile.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 137]\n\f\nRFC 8446 TLS August 2018\n\n\nAppendix D. Backward Compatibility\n\n The TLS protocol provides a built-in mechanism for version\n negotiation between endpoints potentially supporting different\n versions of TLS.\n\n TLS 1.x and SSL 3.0 use compatible ClientHello messages. Servers can\n also handle clients trying to use future versions of TLS as long as\n the ClientHello format remains compatible and there is at least one\n protocol version supported by both the client and the server.\n\n Prior versions of TLS used the record layer version number\n (TLSPlaintext.legacy_record_version and\n TLSCiphertext.legacy_record_version) for various purposes. As of\n TLS 1.3, this field is deprecated. The value of\n TLSPlaintext.legacy_record_version MUST be ignored by all\n implementations. The value of TLSCiphertext.legacy_record_version is\n included in the additional data for deprotection but MAY otherwise be\n ignored or MAY be validated to match the fixed constant value.\n Version negotiation is performed using only the handshake versions\n (ClientHello.legacy_version and ServerHello.legacy_version, as well\n as the ClientHello, HelloRetryRequest, and ServerHello\n \"supported_versions\" extensions). In order to maximize\n interoperability with older endpoints, implementations that negotiate\n the use of TLS 1.0-1.2 SHOULD set the record layer version number to\n the negotiated version for the ServerHello and all records\n thereafter.\n\n For maximum compatibility with previously non-standard behavior and\n misconfigured deployments, all implementations SHOULD support\n validation of certification paths based on the expectations in this\n document, even when handling prior TLS versions' handshakes (see\n Section 4.4.2.2).\n\n TLS 1.2 and prior supported an \"Extended Master Secret\" [RFC7627]\n extension which digested large parts of the handshake transcript into\n the master secret. Because TLS 1.3 always hashes in the transcript\n up to the server Finished, implementations which support both TLS 1.3\n and earlier versions SHOULD indicate the use of the Extended Master\n Secret extension in their APIs whenever TLS 1.3 is used.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 138]\n\f\nRFC 8446 TLS August 2018\n\n\nD.1. Negotiating with an Older Server\n\n A TLS 1.3 client who wishes to negotiate with servers that do not\n support TLS 1.3 will send a normal TLS 1.3 ClientHello containing\n 0x0303 (TLS 1.2) in ClientHello.legacy_version but with the correct\n version(s) in the \"supported_versions\" extension. If the server does\n not support TLS 1.3, it will respond with a ServerHello containing an\n older version number. If the client agrees to use this version, the\n negotiation will proceed as appropriate for the negotiated protocol.\n A client using a ticket for resumption SHOULD initiate the connection\n using the version that was previously negotiated.\n\n Note that 0-RTT data is not compatible with older servers and\n SHOULD NOT be sent absent knowledge that the server supports TLS 1.3.\n See Appendix D.3.\n\n If the version chosen by the server is not supported by the client\n (or is not acceptable), the client MUST abort the handshake with a\n \"protocol_version\" alert.\n\n Some legacy server implementations are known to not implement the TLS\n specification properly and might abort connections upon encountering\n TLS extensions or versions which they are not aware of.\n Interoperability with buggy servers is a complex topic beyond the\n scope of this document. Multiple connection attempts may be required\n in order to negotiate a backward-compatible connection; however, this\n practice is vulnerable to downgrade attacks and is NOT RECOMMENDED.\n\nD.2. Negotiating with an Older Client\n\n A TLS server can also receive a ClientHello indicating a version\n number smaller than its highest supported version. If the\n \"supported_versions\" extension is present, the server MUST negotiate\n using that extension as described in Section 4.2.1. If the\n \"supported_versions\" extension is not present, the server MUST\n negotiate the minimum of ClientHello.legacy_version and TLS 1.2. For\n example, if the server supports TLS 1.0, 1.1, and 1.2, and\n legacy_version is TLS 1.0, the server will proceed with a TLS 1.0\n ServerHello. If the \"supported_versions\" extension is absent and the\n server only supports versions greater than\n ClientHello.legacy_version, the server MUST abort the handshake with\n a \"protocol_version\" alert.\n\n Note that earlier versions of TLS did not clearly specify the record\n layer version number value in all cases\n (TLSPlaintext.legacy_record_version). Servers will receive various\n TLS 1.x versions in this field, but its value MUST always be ignored.\n\n\n\n\nRescorla Standards Track [Page 139]\n\f\nRFC 8446 TLS August 2018\n\n\nD.3. 0-RTT Backward Compatibility\n\n 0-RTT data is not compatible with older servers. An older server\n will respond to the ClientHello with an older ServerHello, but it\n will not correctly skip the 0-RTT data and will fail to complete the\n handshake. This can cause issues when a client attempts to use\n 0-RTT, particularly against multi-server deployments. For example, a\n deployment could deploy TLS 1.3 gradually with some servers\n implementing TLS 1.3 and some implementing TLS 1.2, or a TLS 1.3\n deployment could be downgraded to TLS 1.2.\n\n A client that attempts to send 0-RTT data MUST fail a connection if\n it receives a ServerHello with TLS 1.2 or older. It can then retry\n the connection with 0-RTT disabled. To avoid a downgrade attack, the\n client SHOULD NOT disable TLS 1.3, only 0-RTT.\n\n To avoid this error condition, multi-server deployments SHOULD ensure\n a uniform and stable deployment of TLS 1.3 without 0-RTT prior to\n enabling 0-RTT.\n\nD.4. Middlebox Compatibility Mode\n\n Field measurements [Ben17a] [Ben17b] [Res17a] [Res17b] have found\n that a significant number of middleboxes misbehave when a TLS\n client/server pair negotiates TLS 1.3. Implementations can increase\n the chance of making connections through those middleboxes by making\n the TLS 1.3 handshake look more like a TLS 1.2 handshake:\n\n - The client always provides a non-empty session ID in the\n ClientHello, as described in the legacy_session_id section of\n Section 4.1.2.\n\n - If not offering early data, the client sends a dummy\n change_cipher_spec record (see the third paragraph of Section 5)\n immediately before its second flight. This may either be before\n its second ClientHello or before its encrypted handshake flight.\n If offering early data, the record is placed immediately after the\n first ClientHello.\n\n - The server sends a dummy change_cipher_spec record immediately\n after its first handshake message. This may either be after a\n ServerHello or a HelloRetryRequest.\n\n When put together, these changes make the TLS 1.3 handshake resemble\n TLS 1.2 session resumption, which improves the chance of successfully\n connecting through middleboxes. This \"compatibility mode\" is\n partially negotiated: the client can opt to provide a session ID or\n not, and the server has to echo it. Either side can send\n\n\n\nRescorla Standards Track [Page 140]\n\f\nRFC 8446 TLS August 2018\n\n\n change_cipher_spec at any time during the handshake, as they must be\n ignored by the peer, but if the client sends a non-empty session ID,\n the server MUST send the change_cipher_spec as described in this\n appendix.\n\nD.5. Security Restrictions Related to Backward Compatibility\n\n Implementations negotiating the use of older versions of TLS SHOULD\n prefer forward secret and AEAD cipher suites, when available.\n\n The security of RC4 cipher suites is considered insufficient for the\n reasons cited in [RFC7465]. Implementations MUST NOT offer or\n negotiate RC4 cipher suites for any version of TLS for any reason.\n\n Old versions of TLS permitted the use of very low strength ciphers.\n Ciphers with a strength less than 112 bits MUST NOT be offered or\n negotiated for any version of TLS for any reason.\n\n The security of SSL 3.0 [RFC6101] is considered insufficient for the\n reasons enumerated in [RFC7568], and it MUST NOT be negotiated for\n any reason.\n\n The security of SSL 2.0 [SSL2] is considered insufficient for the\n reasons enumerated in [RFC6176], and it MUST NOT be negotiated for\n any reason.\n\n Implementations MUST NOT send an SSL version 2.0 compatible\n CLIENT-HELLO. Implementations MUST NOT negotiate TLS 1.3 or later\n using an SSL version 2.0 compatible CLIENT-HELLO. Implementations\n are NOT RECOMMENDED to accept an SSL version 2.0 compatible\n CLIENT-HELLO in order to negotiate older versions of TLS.\n\n Implementations MUST NOT send a ClientHello.legacy_version or\n ServerHello.legacy_version set to 0x0300 or less. Any endpoint\n receiving a Hello message with ClientHello.legacy_version or\n ServerHello.legacy_version set to 0x0300 MUST abort the handshake\n with a \"protocol_version\" alert.\n\n Implementations MUST NOT send any records with a version less than\n 0x0300. Implementations SHOULD NOT accept any records with a version\n less than 0x0300 (but may inadvertently do so if the record version\n number is ignored completely).\n\n Implementations MUST NOT use the Truncated HMAC extension, defined in\n Section 7 of [RFC6066], as it is not applicable to AEAD algorithms\n and has been shown to be insecure in some scenarios.\n\n\n\n\n\nRescorla Standards Track [Page 141]\n\f\nRFC 8446 TLS August 2018\n\n\nAppendix E. Overview of Security Properties\n\n A complete security analysis of TLS is outside the scope of this\n document. In this appendix, we provide an informal description of\n the desired properties as well as references to more detailed work in\n the research literature which provides more formal definitions.\n\n We cover properties of the handshake separately from those of the\n record layer.\n\nE.1. Handshake\n\n The TLS handshake is an Authenticated Key Exchange (AKE) protocol\n which is intended to provide both one-way authenticated (server-only)\n and mutually authenticated (client and server) functionality. At the\n completion of the handshake, each side outputs its view of the\n following values:\n\n - A set of \"session keys\" (the various secrets derived from the\n master secret) from which can be derived a set of working keys.\n\n - A set of cryptographic parameters (algorithms, etc.).\n\n - The identities of the communicating parties.\n\n We assume the attacker to be an active network attacker, which means\n it has complete control over the network used to communicate between\n the parties [RFC3552]. Even under these conditions, the handshake\n should provide the properties listed below. Note that these\n properties are not necessarily independent, but reflect the protocol\n consumers' needs.\n\n Establishing the same session keys: The handshake needs to output\n the same set of session keys on both sides of the handshake,\n provided that it completes successfully on each endpoint (see\n [CK01], Definition 1, part 1).\n\n Secrecy of the session keys: The shared session keys should be known\n only to the communicating parties and not to the attacker (see\n [CK01], Definition 1, part 2). Note that in a unilaterally\n authenticated connection, the attacker can establish its own\n session keys with the server, but those session keys are distinct\n from those established by the client.\n\n Peer authentication: The client's view of the peer identity should\n reflect the server's identity. If the client is authenticated,\n the server's view of the peer identity should match the client's\n identity.\n\n\n\nRescorla Standards Track [Page 142]\n\f\nRFC 8446 TLS August 2018\n\n\n Uniqueness of the session keys: Any two distinct handshakes should\n produce distinct, unrelated session keys. Individual session keys\n produced by a handshake should also be distinct and independent.\n\n Downgrade protection: The cryptographic parameters should be the\n same on both sides and should be the same as if the peers had been\n communicating in the absence of an attack (see [BBFGKZ16],\n Definitions 8 and 9).\n\n Forward secret with respect to long-term keys: If the long-term\n keying material (in this case the signature keys in certificate-\n based authentication modes or the external/resumption PSK in PSK\n with (EC)DHE modes) is compromised after the handshake is\n complete, this does not compromise the security of the session key\n (see [DOW92]), as long as the session key itself has been erased.\n The forward secrecy property is not satisfied when PSK is used in\n the \"psk_ke\" PskKeyExchangeMode.\n\n Key Compromise Impersonation (KCI) resistance: In a mutually\n authenticated connection with certificates, compromising the\n long-term secret of one actor should not break that actor's\n authentication of their peer in the given connection (see\n [HGFS15]). For example, if a client's signature key is\n compromised, it should not be possible to impersonate arbitrary\n servers to that client in subsequent handshakes.\n\n Protection of endpoint identities: The server's identity\n (certificate) should be protected against passive attackers. The\n client's identity should be protected against both passive and\n active attackers.\n\n Informally, the signature-based modes of TLS 1.3 provide for the\n establishment of a unique, secret, shared key established by an\n (EC)DHE key exchange and authenticated by the server's signature over\n the handshake transcript, as well as tied to the server's identity by\n a MAC. If the client is authenticated by a certificate, it also\n signs over the handshake transcript and provides a MAC tied to both\n identities. [SIGMA] describes the design and analysis of this type\n of key exchange protocol. If fresh (EC)DHE keys are used for each\n connection, then the output keys are forward secret.\n\n The external PSK and resumption PSK bootstrap from a long-term shared\n secret into a unique per-connection set of short-term session keys.\n This secret may have been established in a previous handshake. If\n PSK with (EC)DHE key establishment is used, these session keys will\n also be forward secret. The resumption PSK has been designed so that\n the resumption master secret computed by connection N and needed to\n form connection N+1 is separate from the traffic keys used by\n\n\n\nRescorla Standards Track [Page 143]\n\f\nRFC 8446 TLS August 2018\n\n\n connection N, thus providing forward secrecy between the connections.\n In addition, if multiple tickets are established on the same\n connection, they are associated with different keys, so compromise of\n the PSK associated with one ticket does not lead to the compromise of\n connections established with PSKs associated with other tickets.\n This property is most interesting if tickets are stored in a database\n (and so can be deleted) rather than if they are self-encrypted.\n\n The PSK binder value forms a binding between a PSK and the current\n handshake, as well as between the session where the PSK was\n established and the current session. This binding transitively\n includes the original handshake transcript, because that transcript\n is digested into the values which produce the resumption master\n secret. This requires that both the KDF used to produce the\n resumption master secret and the MAC used to compute the binder be\n collision resistant. See Appendix E.1.1 for more on this. Note: The\n binder does not cover the binder values from other PSKs, though they\n are included in the Finished MAC.\n\n TLS does not currently permit the server to send a\n certificate_request message in non-certificate-based handshakes\n (e.g., PSK). If this restriction were to be relaxed in future, the\n client's signature would not cover the server's certificate directly.\n However, if the PSK was established through a NewSessionTicket, the\n client's signature would transitively cover the server's certificate\n through the PSK binder. [PSK-FINISHED] describes a concrete attack\n on constructions that do not bind to the server's certificate (see\n also [Kraw16]). It is unsafe to use certificate-based client\n authentication when the client might potentially share the same\n PSK/key-id pair with two different endpoints. Implementations\n MUST NOT combine external PSKs with certificate-based authentication\n of either the client or the server unless negotiated by some\n extension.\n\n If an exporter is used, then it produces values which are unique and\n secret (because they are generated from a unique session key).\n Exporters computed with different labels and contexts are\n computationally independent, so it is not feasible to compute one\n from another or the session secret from the exported value.\n Note: Exporters can produce arbitrary-length values; if exporters are\n to be used as channel bindings, the exported value MUST be large\n enough to provide collision resistance. The exporters provided in\n TLS 1.3 are derived from the same Handshake Contexts as the early\n traffic keys and the application traffic keys, respectively, and thus\n have similar security properties. Note that they do not include the\n client's certificate; future applications which wish to bind to the\n client's certificate may need to define a new exporter that includes\n the full handshake transcript.\n\n\n\nRescorla Standards Track [Page 144]\n\f\nRFC 8446 TLS August 2018\n\n\n For all handshake modes, the Finished MAC (and, where present, the\n signature) prevents downgrade attacks. In addition, the use of\n certain bytes in the random nonces as described in Section 4.1.3\n allows the detection of downgrade to previous TLS versions. See\n [BBFGKZ16] for more details on TLS 1.3 and downgrade.\n\n As soon as the client and the server have exchanged enough\n information to establish shared keys, the remainder of the handshake\n is encrypted, thus providing protection against passive attackers,\n even if the computed shared key is not authenticated. Because the\n server authenticates before the client, the client can ensure that if\n it authenticates to the server, it only reveals its identity to an\n authenticated server. Note that implementations must use the\n provided record-padding mechanism during the handshake to avoid\n leaking information about the identities due to length. The client's\n proposed PSK identities are not encrypted, nor is the one that the\n server selects.\n\nE.1.1. Key Derivation and HKDF\n\n Key derivation in TLS 1.3 uses HKDF as defined in [RFC5869] and its\n two components, HKDF-Extract and HKDF-Expand. The full rationale for\n the HKDF construction can be found in [Kraw10] and the rationale for\n the way it is used in TLS 1.3 in [KW16]. Throughout this document,\n each application of HKDF-Extract is followed by one or more\n invocations of HKDF-Expand. This ordering should always be followed\n (including in future revisions of this document); in particular, one\n SHOULD NOT use an output of HKDF-Extract as an input to another\n application of HKDF-Extract without an HKDF-Expand in between.\n Multiple applications of HKDF-Expand to some of the same inputs are\n allowed as long as these are differentiated via the key and/or the\n labels.\n\n Note that HKDF-Expand implements a pseudorandom function (PRF) with\n both inputs and outputs of variable length. In some of the uses of\n HKDF in this document (e.g., for generating exporters and the\n resumption_master_secret), it is necessary that the application of\n HKDF-Expand be collision resistant; namely, it should be infeasible\n to find two different inputs to HKDF-Expand that output the same\n value. This requires the underlying hash function to be collision\n resistant and the output length from HKDF-Expand to be of size at\n least 256 bits (or as much as needed for the hash function to prevent\n finding collisions).\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 145]\n\f\nRFC 8446 TLS August 2018\n\n\nE.1.2. Client Authentication\n\n A client that has sent authentication data to a server, either during\n the handshake or in post-handshake authentication, cannot be sure\n whether the server afterwards considers the client to be\n authenticated or not. If the client needs to determine if the server\n considers the connection to be unilaterally or mutually\n authenticated, this has to be provisioned by the application layer.\n See [CHHSV17] for details. In addition, the analysis of\n post-handshake authentication from [Kraw16] shows that the client\n identified by the certificate sent in the post-handshake phase\n possesses the traffic key. This party is therefore the client that\n participated in the original handshake or one to whom the original\n client delegated the traffic key (assuming that the traffic key has\n not been compromised).\n\nE.1.3. 0-RTT\n\n The 0-RTT mode of operation generally provides security properties\n similar to those of 1-RTT data, with the two exceptions that the\n 0-RTT encryption keys do not provide full forward secrecy and that\n the server is not able to guarantee uniqueness of the handshake\n (non-replayability) without keeping potentially undue amounts of\n state. See Section 8 for mechanisms to limit the exposure to replay.\n\nE.1.4. Exporter Independence\n\n The exporter_master_secret and early_exporter_master_secret are\n derived to be independent of the traffic keys and therefore do not\n represent a threat to the security of traffic encrypted with those\n keys. However, because these secrets can be used to compute any\n exporter value, they SHOULD be erased as soon as possible. If the\n total set of exporter labels is known, then implementations SHOULD\n pre-compute the inner Derive-Secret stage of the exporter computation\n for all those labels, then erase the [early_]exporter_master_secret,\n followed by each inner value as soon as it is known that it will not\n be needed again.\n\nE.1.5. Post-Compromise Security\n\n TLS does not provide security for handshakes which take place after\n the peer's long-term secret (signature key or external PSK) is\n compromised. It therefore does not provide post-compromise security\n [CCG16], sometimes also referred to as backward or future secrecy.\n This is in contrast to KCI resistance, which describes the security\n guarantees that a party has after its own long-term secret has been\n compromised.\n\n\n\n\nRescorla Standards Track [Page 146]\n\f\nRFC 8446 TLS August 2018\n\n\nE.1.6. External References\n\n The reader should refer to the following references for analysis of\n the TLS handshake: [DFGS15], [CHSV16], [DFGS16], [KW16], [Kraw16],\n [FGSW16], [LXZFH16], [FG17], and [BBK17].\n\nE.2. Record Layer\n\n The record layer depends on the handshake producing strong traffic\n secrets which can be used to derive bidirectional encryption keys and\n nonces. Assuming that is true, and the keys are used for no more\n data than indicated in Section 5.5, then the record layer should\n provide the following guarantees:\n\n Confidentiality: An attacker should not be able to determine the\n plaintext contents of a given record.\n\n Integrity: An attacker should not be able to craft a new record\n which is different from an existing record which will be accepted\n by the receiver.\n\n Order protection/non-replayability: An attacker should not be able\n to cause the receiver to accept a record which it has already\n accepted or cause the receiver to accept record N+1 without having\n first processed record N.\n\n Length concealment: Given a record with a given external length, the\n attacker should not be able to determine the amount of the record\n that is content versus padding.\n\n Forward secrecy after key change: If the traffic key update\n mechanism described in Section 4.6.3 has been used and the\n previous generation key is deleted, an attacker who compromises\n the endpoint should not be able to decrypt traffic encrypted with\n the old key.\n\n Informally, TLS 1.3 provides these properties by AEAD-protecting the\n plaintext with a strong key. AEAD encryption [RFC5116] provides\n confidentiality and integrity for the data. Non-replayability is\n provided by using a separate nonce for each record, with the nonce\n being derived from the record sequence number (Section 5.3), with the\n sequence number being maintained independently at both sides; thus,\n records which are delivered out of order result in AEAD deprotection\n failures. In order to prevent mass cryptanalysis when the same\n plaintext is repeatedly encrypted by different users under the same\n key (as is commonly the case for HTTP), the nonce is formed by mixing\n\n\n\n\n\nRescorla Standards Track [Page 147]\n\f\nRFC 8446 TLS August 2018\n\n\n the sequence number with a secret per-connection initialization\n vector derived along with the traffic keys. See [BT16] for analysis\n of this construction.\n\n The rekeying technique in TLS 1.3 (see Section 7.2) follows the\n construction of the serial generator as discussed in [REKEY], which\n shows that rekeying can allow keys to be used for a larger number of\n encryptions than without rekeying. This relies on the security of\n the HKDF-Expand-Label function as a pseudorandom function (PRF). In\n addition, as long as this function is truly one way, it is not\n possible to compute traffic keys from prior to a key change (forward\n secrecy).\n\n TLS does not provide security for data which is communicated on a\n connection after a traffic secret of that connection is compromised.\n That is, TLS does not provide post-compromise security/future\n secrecy/backward secrecy with respect to the traffic secret. Indeed,\n an attacker who learns a traffic secret can compute all future\n traffic secrets on that connection. Systems which want such\n guarantees need to do a fresh handshake and establish a new\n connection with an (EC)DHE exchange.\n\nE.2.1. External References\n\n The reader should refer to the following references for analysis of\n the TLS record layer: [BMMRT15], [BT16], [BDFKPPRSZZ16], [BBK17], and\n [PS18].\n\nE.3. Traffic Analysis\n\n TLS is susceptible to a variety of traffic analysis attacks based on\n observing the length and timing of encrypted packets [CLINIC]\n [HCJC16]. This is particularly easy when there is a small set of\n possible messages to be distinguished, such as for a video server\n hosting a fixed corpus of content, but still provides usable\n information even in more complicated scenarios.\n\n TLS does not provide any specific defenses against this form of\n attack but does include a padding mechanism for use by applications:\n The plaintext protected by the AEAD function consists of content plus\n variable-length padding, which allows the application to produce\n arbitrary-length encrypted records as well as padding-only cover\n traffic to conceal the difference between periods of transmission and\n periods of silence. Because the padding is encrypted alongside the\n actual content, an attacker cannot directly determine the length of\n the padding but may be able to measure it indirectly by the use of\n timing channels exposed during record processing (i.e., seeing how\n long it takes to process a record or trickling in records to see\n\n\n\nRescorla Standards Track [Page 148]\n\f\nRFC 8446 TLS August 2018\n\n\n which ones elicit a response from the server). In general, it is not\n known how to remove all of these channels because even a\n constant-time padding removal function will likely feed the content\n into data-dependent functions. At minimum, a fully constant-time\n server or client would require close cooperation with the\n application-layer protocol implementation, including making that\n higher-level protocol constant time.\n\n Note: Robust traffic analysis defenses will likely lead to inferior\n performance due to delays in transmitting packets and increased\n traffic volume.\n\nE.4. Side-Channel Attacks\n\n In general, TLS does not have specific defenses against side-channel\n attacks (i.e., those which attack the communications via secondary\n channels such as timing), leaving those to the implementation of the\n relevant cryptographic primitives. However, certain features of TLS\n are designed to make it easier to write side-channel resistant code:\n\n - Unlike previous versions of TLS which used a composite MAC-then-\n encrypt structure, TLS 1.3 only uses AEAD algorithms, allowing\n implementations to use self-contained constant-time\n implementations of those primitives.\n\n - TLS uses a uniform \"bad_record_mac\" alert for all decryption\n errors, which is intended to prevent an attacker from gaining\n piecewise insight into portions of the message. Additional\n resistance is provided by terminating the connection on such\n errors; a new connection will have different cryptographic\n material, preventing attacks against the cryptographic primitives\n that require multiple trials.\n\n Information leakage through side channels can occur at layers above\n TLS, in application protocols and the applications that use them.\n Resistance to side-channel attacks depends on applications and\n application protocols separately ensuring that confidential\n information is not inadvertently leaked.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 149]\n\f\nRFC 8446 TLS August 2018\n\n\nE.5. Replay Attacks on 0-RTT\n\n Replayable 0-RTT data presents a number of security threats to TLS-\n using applications, unless those applications are specifically\n engineered to be safe under replay (minimally, this means idempotent,\n but in many cases may also require other stronger conditions, such as\n constant-time response). Potential attacks include:\n\n - Duplication of actions which cause side effects (e.g., purchasing\n an item or transferring money) to be duplicated, thus harming the\n site or the user.\n\n - Attackers can store and replay 0-RTT messages in order to reorder\n them with respect to other messages (e.g., moving a delete to\n after a create).\n\n - Exploiting cache timing behavior to discover the content of 0-RTT\n messages by replaying a 0-RTT message to a different cache node\n and then using a separate connection to measure request latency,\n to see if the two requests address the same resource.\n\n If data can be replayed a large number of times, additional attacks\n become possible, such as making repeated measurements of the speed of\n cryptographic operations. In addition, they may be able to overload\n rate-limiting systems. For a further description of these attacks,\n see [Mac17].\n\n Ultimately, servers have the responsibility to protect themselves\n against attacks employing 0-RTT data replication. The mechanisms\n described in Section 8 are intended to prevent replay at the TLS\n layer but do not provide complete protection against receiving\n multiple copies of client data. TLS 1.3 falls back to the 1-RTT\n handshake when the server does not have any information about the\n client, e.g., because it is in a different cluster which does not\n share state or because the ticket has been deleted as described in\n Section 8.1. If the application-layer protocol retransmits data in\n this setting, then it is possible for an attacker to induce message\n duplication by sending the ClientHello to both the original cluster\n (which processes the data immediately) and another cluster which will\n fall back to 1-RTT and process the data upon application-layer\n replay. The scale of this attack is limited by the client's\n willingness to retry transactions and therefore only allows a limited\n amount of duplication, with each copy appearing as a new connection\n at the server.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 150]\n\f\nRFC 8446 TLS August 2018\n\n\n If implemented correctly, the mechanisms described in Sections 8.1\n and 8.2 prevent a replayed ClientHello and its associated 0-RTT data\n from being accepted multiple times by any cluster with consistent\n state; for servers which limit the use of 0-RTT to one cluster for a\n single ticket, then a given ClientHello and its associated 0-RTT data\n will only be accepted once. However, if state is not completely\n consistent, then an attacker might be able to have multiple copies of\n the data be accepted during the replication window. Because clients\n do not know the exact details of server behavior, they MUST NOT send\n messages in early data which are not safe to have replayed and which\n they would not be willing to retry across multiple 1-RTT connections.\n\n Application protocols MUST NOT use 0-RTT data without a profile that\n defines its use. That profile needs to identify which messages or\n interactions are safe to use with 0-RTT and how to handle the\n situation when the server rejects 0-RTT and falls back to 1-RTT.\n\n In addition, to avoid accidental misuse, TLS implementations MUST NOT\n enable 0-RTT (either sending or accepting) unless specifically\n requested by the application and MUST NOT automatically resend 0-RTT\n data if it is rejected by the server unless instructed by the\n application. Server-side applications may wish to implement special\n processing for 0-RTT data for some kinds of application traffic\n (e.g., abort the connection, request that data be resent at the\n application layer, or delay processing until the handshake\n completes). In order to allow applications to implement this kind of\n processing, TLS implementations MUST provide a way for the\n application to determine if the handshake has completed.\n\nE.5.1. Replay and Exporters\n\n Replays of the ClientHello produce the same early exporter, thus\n requiring additional care by applications which use these exporters.\n In particular, if these exporters are used as an authentication\n channel binding (e.g., by signing the output of the exporter), an\n attacker who compromises the PSK can transplant authenticators\n between connections without compromising the authentication key.\n\n In addition, the early exporter SHOULD NOT be used to generate\n server-to-client encryption keys because that would entail the reuse\n of those keys. This parallels the use of the early application\n traffic keys only in the client-to-server direction.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 151]\n\f\nRFC 8446 TLS August 2018\n\n\nE.6. PSK Identity Exposure\n\n Because implementations respond to an invalid PSK binder by aborting\n the handshake, it may be possible for an attacker to verify whether a\n given PSK identity is valid. Specifically, if a server accepts both\n external-PSK handshakes and certificate-based handshakes, a valid PSK\n identity will result in a failed handshake, whereas an invalid\n identity will just be skipped and result in a successful certificate\n handshake. Servers which solely support PSK handshakes may be able\n to resist this form of attack by treating the cases where there is no\n valid PSK identity and where there is an identity but it has an\n invalid binder identically.\n\nE.7. Sharing PSKs\n\n TLS 1.3 takes a conservative approach to PSKs by binding them to a\n specific KDF. By contrast, TLS 1.2 allows PSKs to be used with any\n hash function and the TLS 1.2 PRF. Thus, any PSK which is used with\n both TLS 1.2 and TLS 1.3 must be used with only one hash in TLS 1.3,\n which is less than optimal if users want to provision a single PSK.\n The constructions in TLS 1.2 and TLS 1.3 are different, although they\n are both based on HMAC. While there is no known way in which the\n same PSK might produce related output in both versions, only limited\n analysis has been done. Implementations can ensure safety from\n cross-protocol related output by not reusing PSKs between TLS 1.3 and\n TLS 1.2.\n\nE.8. Attacks on Static RSA\n\n Although TLS 1.3 does not use RSA key transport and so is not\n directly susceptible to Bleichenbacher-type attacks [Blei98], if TLS\n 1.3 servers also support static RSA in the context of previous\n versions of TLS, then it may be possible to impersonate the server\n for TLS 1.3 connections [JSS15]. TLS 1.3 implementations can prevent\n this attack by disabling support for static RSA across all versions\n of TLS. In principle, implementations might also be able to separate\n certificates with different keyUsage bits for static RSA decryption\n and RSA signature, but this technique relies on clients refusing to\n accept signatures using keys in certificates that do not have the\n digitalSignature bit set, and many clients do not enforce this\n restriction.\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 152]\n\f\nRFC 8446 TLS August 2018\n\n\nContributors\n\n Martin Abadi\n University of California, Santa Cruz\n [redacted-email]\n\n Christopher Allen\n (co-editor of TLS 1.0)\n Alacrity Ventures\n [redacted-email]\n\n Richard Barnes\n Cisco\n [redacted-email]\n\n Steven M. Bellovin\n Columbia University\n [redacted-email]\n\n David Benjamin\n Google\n [redacted-email]\n\n Benjamin Beurdouche\n INRIA & Microsoft Research\n [redacted-email]\n\n Karthikeyan Bhargavan\n (editor of [RFC7627])\n INRIA\n [redacted-email]\n\n Simon Blake-Wilson\n (co-author of [RFC4492])\n BCI\n [redacted-email]\n\n Nelson Bolyard\n (co-author of [RFC4492])\n Sun Microsystems, Inc.\n [redacted-email]\n\n Ran Canetti\n IBM\n [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 153]\n\f\nRFC 8446 TLS August 2018\n\n\n Matt Caswell\n OpenSSL\n [redacted-email]\n\n Stephen Checkoway\n University of Illinois at Chicago\n [redacted-email]\n\n Pete Chown\n Skygate Technology Ltd\n [redacted-email]\n\n Katriel Cohn-Gordon\n University of Oxford\n [redacted-email]\n\n Cas Cremers\n University of Oxford\n [redacted-email]\n\n Antoine Delignat-Lavaud\n (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n Tim Dierks\n (co-author of TLS 1.0, co-editor of TLS 1.1 and 1.2)\n Independent\n [redacted-email]\n\n Roelof DuToit\n Symantec Corporation\n [redacted-email]\n\n Taher Elgamal\n Securify\n [redacted-email]\n\n Pasi Eronen\n Nokia\n [redacted-email]\n\n Cedric Fournet\n Microsoft\n [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 154]\n\f\nRFC 8446 TLS August 2018\n\n\n Anil Gangolli\n [redacted-email]\n\n David M. Garrett\n [redacted-email]\n\n Illya Gerasymchuk\n Independent\n [redacted-email]\n\n Alessandro Ghedini\n Cloudflare Inc.\n [redacted-email]\n\n Daniel Kahn Gillmor\n ACLU\n [redacted-email]\n\n Matthew Green\n Johns Hopkins University\n [redacted-email]\n\n Jens Guballa\n ETAS\n [redacted-email]\n\n Felix Guenther\n TU Darmstadt\n [redacted-email]\n\n Vipul Gupta\n (co-author of [RFC4492])\n Sun Microsystems Laboratories\n [redacted-email]\n\n Chris Hawk\n (co-author of [RFC4492])\n Corriente Networks LLC\n [redacted-email]\n\n Kipp Hickman\n\n Alfred Hoenes\n\n David Hopwood\n Independent Consultant\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 155]\n\f\nRFC 8446 TLS August 2018\n\n\n Marko Horvat\n MPI-SWS\n [redacted-email]\n\n Jonathan Hoyland\n Royal Holloway, University of London\n [redacted-email]\n\n Subodh Iyengar\n Facebook\n [redacted-email]\n\n Benjamin Kaduk\n Akamai Technologies\n [redacted-email]\n\n Hubert Kario\n Red Hat Inc.\n [redacted-email]\n\n Phil Karlton\n (co-author of SSL 3.0)\n\n Leon Klingele\n Independent\n [redacted-email]\n\n Paul Kocher\n (co-author of SSL 3.0)\n Cryptography Research\n [redacted-email]\n\n Hugo Krawczyk\n IBM\n [redacted-email]\n\n Adam Langley\n (co-author of [RFC7627])\n Google\n [redacted-email]\n\n Olivier Levillain\n ANSSI\n [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 156]\n\f\nRFC 8446 TLS August 2018\n\n\n Xiaoyin Liu\n University of North Carolina at Chapel Hill\n [redacted-email]\n\n Ilari Liusvaara\n Independent\n [redacted-email]\n\n Atul Luykx\n K.U. Leuven\n [redacted-email]\n\n Colm MacCarthaigh\n Amazon Web Services\n [redacted-email]\n\n Carl Mehner\n USAA\n [redacted-email]\n\n Jan Mikkelsen\n Transactionware\n [redacted-email]\n\n Bodo Moeller\n (co-author of [RFC4492])\n Google\n [redacted-email]\n\n Kyle Nekritz\n Facebook\n [redacted-email]\n\n Erik Nygren\n Akamai Technologies\n [redacted-email]\n\n Magnus Nystrom\n Microsoft\n [redacted-email]\n\n Kazuho Oku\n DeNA Co., Ltd.\n [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 157]\n\f\nRFC 8446 TLS August 2018\n\n\n Kenny Paterson\n Royal Holloway, University of London\n [redacted-email]\n\n Christopher Patton\n University of Florida\n [redacted-email]\n\n Alfredo Pironti\n (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n Andrei Popov\n Microsoft\n [redacted-email]\n\n Marsh Ray\n (co-author of [RFC7627])\n Microsoft\n [redacted-email]\n\n Robert Relyea\n Netscape Communications\n [redacted-email]\n\n Kyle Rose\n Akamai Technologies\n [redacted-email]\n\n Jim Roskind\n Amazon\n [redacted-email]\n\n Michael Sabin\n\n Joe Salowey\n Tableau Software\n [redacted-email]\n\n Rich Salz\n Akamai\n [redacted-email]\n\n David Schinazi\n Apple Inc.\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 158]\n\f\nRFC 8446 TLS August 2018\n\n\n Sam Scott\n Royal Holloway, University of London\n [redacted-email]\n\n Thomas Shrimpton\n University of Florida\n [redacted-email]\n\n Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n Brian Smith\n Independent\n [redacted-email]\n\n Brian Sniffen\n Akamai Technologies\n [redacted-email]\n\n Nick Sullivan\n Cloudflare Inc.\n [redacted-email]\n\n Bjoern Tackmann\n University of California, San Diego\n [redacted-email]\n\n Tim Taubert\n Mozilla\n [redacted-email]\n\n Martin Thomson\n Mozilla\n [redacted-email]\n\n Hannes Tschofenig\n Arm Limited\n [redacted-email]\n\n Sean Turner\n sn3rd\n [redacted-email]\n\n Steven Valdez\n Google\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 159]\n\f\nRFC 8446 TLS August 2018\n\n\n Filippo Valsorda\n Cloudflare Inc.\n [redacted-email]\n\n Thyla van der Merwe\n Royal Holloway, University of London\n [redacted-email]\n\n Victor Vasiliev\n Google\n [redacted-email]\n\n Hoeteck Wee\n Ecole Normale Superieure, Paris\n [redacted-email]\n\n Tom Weinstein\n\n David Wong\n NCC Group\n [redacted-email]\n\n Christopher A. Wood\n Apple Inc.\n [redacted-email]\n\n Tim Wright\n Vodafone\n [redacted-email]\n\n Peter Wu\n Independent\n [redacted-email]\n\n Kazu Yamamoto\n Internet Initiative Japan Inc.\n [redacted-email]\n\nAuthor's Address\n\n Eric Rescorla\n Mozilla\n\n Email: [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 160]", "text_sha256": "d596886622bceee8500b3f3c9d869d947b4e70a2f05ea65fdd75666b46debd81"}], "relations": [{"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5077 H. Zhou", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5077", "source_sha256": "e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef"}, {"char_end": 3675, "char_start": 3419, "evidence_quote": "{\"id\": 1269006, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269006/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}", "fact_ids": ["informative_reference:5077"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "informative_reference", "relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5077"}, {"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5246 Independent", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5246", "source_sha256": "2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb"}, {"char_end": 3417, "char_start": 3161, "evidence_quote": "{\"id\": 1269005, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269005/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}", "fact_ids": ["informative_reference:5246"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "informative_reference", "relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5246"}, {"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5705 RTFM, Inc.", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5705", "source_sha256": "29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a"}, {"char_end": 5223, "char_start": 4967, "evidence_quote": "{\"id\": 1269016, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269016/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}", "fact_ids": ["normative_reference:5705"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "normative_reference", "relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5705"}, {"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6066 Huawei", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:6066", "source_sha256": "6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0"}, {"char_end": 12717, "char_start": 12466, "evidence_quote": "{\"id\": 1591472, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591472/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}", "fact_ids": ["normative_reference:6066"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "normative_reference", "relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:6066"}, {"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6961 June 2013", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:6961", "source_sha256": "c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db"}, {"char_end": 10425, "char_start": 10169, "evidence_quote": "{\"id\": 1269043, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269043/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}", "fact_ids": ["normative_reference:6961"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "normative_reference", "relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:6961"}, {"evidence": [{"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_ids": ["obsoletes:5077"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5077 H. Zhou", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5077", "source_sha256": "e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef"}], "kind": "obsoletes", "relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5077"}, {"evidence": [{"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_ids": ["obsoletes:5246"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5246 Independent", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5246", "source_sha256": "2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb"}], "kind": "obsoletes", "relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5246"}, {"evidence": [{"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_ids": ["obsoletes:6961"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6961 June 2013", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:6961", "source_sha256": "c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db"}], "kind": "obsoletes", "relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:6961"}, {"evidence": [{"char_end": 454, "char_start": 431, "evidence_quote": "draft-ietf-tls-tls13-28", "fact_ids": ["draft_revision"], "record_id": "ietf:draft:draft-ietf-tls-tls13-28", "source_sha256": "9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 13945, "char_start": 13673, "evidence_quote": "{\"id\": 1295893, \"originaltargetaliasname\": \"rfc8446\", \"relationship\": \"/api/v1/name/docrelationshipname/became_rfc/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1295893/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8446/\"}", "fact_ids": ["became_rfc"], "record_id": "ietf:datatracker-relation:draft-ietf-tls-tls13:rfc8446", "source_sha256": "3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131"}], "kind": "published_as", "relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "target_record_id": "ietf:rfc:8446"}, {"evidence": [{"char_end": 238, "char_start": 219, "evidence_quote": "Updates: 5705, 6066", "fact_ids": ["updates:5705"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5705 RTFM, Inc.", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5705", "source_sha256": "29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a"}], "kind": "updates", "relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5705"}, {"evidence": [{"char_end": 238, "char_start": 219, "evidence_quote": "Updates: 5705, 6066", "fact_ids": ["updates:6066"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6066 Huawei", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:6066", "source_sha256": "6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0"}], "kind": "updates", "relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:6066"}], "schema_version": "longworld.ietf-workflow-manifest.v1", "semantic_facts_train_ready": false, "source_status": "public_api_export", "supporting_records": [{"draft_name": "draft-ietf-tls-tls13", "facts": [{"char_end": 673, "char_start": 643, "evidence_quote": "\"name\": \"draft-ietf-tls-tls13\"", "fact_id": "draft_name", "field": "draft_name", "raw_char_end": 673, "raw_char_start": 643, "value": "draft-ietf-tls-tls13"}], "kind": "datatracker_document", "occurred_at": "2020-03-07T08:26:41Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 2, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:a632cb66584c8f5d364d909b76eed121c1a9d00e341908463e019efad0faee0f", "record_id": "ietf:datatracker:draft-ietf-tls-tls13", "retrieval_url": "https://datatracker.ietf.org/api/v1/doc/document/draft-ietf-tls-tls13/", "revision": "28", "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "a632cb66584c8f5d364d909b76eed121c1a9d00e341908463e019efad0faee0f", "source_url": "https://datatracker.ietf.org/api/v1/doc/document/draft-ietf-tls-tls13/", "temporal_semantics": "source_event_time", "text": "{\"abstract\": \"This document specifies version 1.3 of the Transport Layer Security (TLS) protocol. TLS allows client/server applications to communicate over the Internet in a way that is designed to prevent eavesdropping, tampering, and message forgery.\\n\\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077, 5246, and 6961. This document also specifies new requirements for TLS 1.2 implementations.\", \"ad\": \"/api/v1/person/person/104851/\", \"expires\": \"2018-09-21T21:26:15Z\", \"external_url\": \"\", \"group\": \"/api/v1/group/group/1326/\", \"id\": 55251, \"intended_std_level\": \"/api/v1/name/intendedstdlevelname/ps/\", \"keywords\": \"[]\", \"name\": \"draft-ietf-tls-tls13\", \"note\": \"\", \"notify\": \"Sean Turner <[redacted-email]>\", \"pages\": 160, \"resource_uri\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"rev\": \"28\", \"rfc\": null, \"rfc_number\": null, \"shepherd\": \"/api/v1/person/email/[redacted-email]/\", \"states\": [\"/api/v1/doc/state/3/\", \"/api/v1/doc/state/108/\", \"/api/v1/doc/state/112/\", \"/api/v1/doc/state/7/\", \"/api/v1/doc/state/44/\"], \"std_level\": \"/api/v1/name/stdlevelname/ps/\", \"stream\": \"/api/v1/name/streamname/ietf/\", \"submissions\": [\"/api/v1/submit/submission/59397/\", \"/api/v1/submit/submission/59398/\", \"/api/v1/submit/submission/61729/\", \"/api/v1/submit/submission/64520/\", \"/api/v1/submit/submission/65644/\", \"/api/v1/submit/submission/67695/\", \"/api/v1/submit/submission/70123/\", \"/api/v1/submit/submission/70989/\", \"/api/v1/submit/submission/71776/\", \"/api/v1/submit/submission/72481/\", \"/api/v1/submit/submission/73392/\", \"/api/v1/submit/submission/74860/\", \"/api/v1/submit/submission/77260/\", \"/api/v1/submit/submission/78266/\", \"/api/v1/submit/submission/79977/\", \"/api/v1/submit/submission/80556/\", \"/api/v1/submit/submission/81099/\", \"/api/v1/submit/submission/81687/\", \"/api/v1/submit/submission/81857/\", \"/api/v1/submit/submission/84956/\", \"/api/v1/submit/submission/86252/\", \"/api/v1/submit/submission/92891/\", \"/api/v1/submit/submission/88268/\", \"/api/v1/submit/submission/91422/\", \"/api/v1/submit/submission/92066/\", \"/api/v1/submit/submission/93465/\", \"/api/v1/submit/submission/93544/\", \"/api/v1/submit/submission/94182/\", \"/api/v1/submit/submission/94319/\"], \"tags\": [], \"time\": \"2020-03-07T08:26:41Z\", \"title\": \"The Transport Layer Security (TLS) Protocol Version 1.3\", \"type\": \"/api/v1/name/doctypename/draft/\", \"uploaded_filename\": \"\", \"words\": 35738}", "text_sha256": "e51b82d004e563f85becf3bc398f5dea916940ec55b92cf7b728f78a625a293b"}, {"facts": [{"char_end": 708, "char_start": 691, "evidence_quote": "\"name\": \"rfc8446\"", "fact_id": "rfc_name", "field": "rfc_name", "raw_char_end": 708, "raw_char_start": 691, "value": "rfc8446"}], "kind": "datatracker_document", "occurred_at": "2026-05-20T15:34:35Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 0, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:89b01010d1ff76dc260539d5a453dde0b116d857db43c9e09d4a3dc062c9ce31", "record_id": "ietf:datatracker:rfc8446", "retrieval_url": "https://datatracker.ietf.org/api/v1/doc/document/rfc8446/", "rfc_number": 8446, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "89b01010d1ff76dc260539d5a453dde0b116d857db43c9e09d4a3dc062c9ce31", "source_url": "https://datatracker.ietf.org/api/v1/doc/document/rfc8446/", "temporal_semantics": "source_event_time", "text": "{\"abstract\": \"This document specifies version 1.3 of the Transport Layer Security (TLS) protocol. TLS allows client/server applications to communicate over the Internet in a way that is designed to prevent eavesdropping, tampering, and message forgery.\\n\\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077, 5246, and 6961. This document also specifies new requirements for TLS 1.2 implementations.\", \"ad\": \"/api/v1/person/person/104851/\", \"expires\": null, \"external_url\": \"\", \"group\": \"/api/v1/group/group/1326/\", \"id\": 130263, \"intended_std_level\": null, \"keywords\": \"['international data algorithm', 'symmetric', 'transport protocol layer', 'authentication', 'privacy']\", \"name\": \"rfc8446\", \"note\": \"\", \"notify\": \"\", \"pages\": 160, \"resource_uri\": \"/api/v1/doc/document/rfc8446/\", \"rev\": \"\", \"rfc\": \"8446\", \"rfc_number\": 8446, \"shepherd\": null, \"states\": [\"/api/v1/doc/state/177/\"], \"std_level\": \"/api/v1/name/stdlevelname/ps/\", \"stream\": \"/api/v1/name/streamname/ietf/\", \"submissions\": [], \"tags\": [\"/api/v1/name/doctagname/errata/\", \"/api/v1/name/doctagname/verified-errata/\"], \"time\": \"2026-05-20T15:34:35Z\", \"title\": \"The Transport Layer Security (TLS) Protocol Version 1.3\", \"type\": \"/api/v1/name/doctypename/rfc/\", \"uploaded_filename\": \"\", \"words\": 35738}", "text_sha256": "89b01010d1ff76dc260539d5a453dde0b116d857db43c9e09d4a3dc062c9ce31"}, {"draft_name": "draft-ietf-tls-tls13", "facts": [{"char_end": 13945, "char_start": 13673, "evidence_quote": "{\"id\": 1295893, \"originaltargetaliasname\": \"rfc8446\", \"relationship\": \"/api/v1/name/docrelationshipname/became_rfc/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1295893/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8446/\"}", "fact_id": "became_rfc", "field": "became_rfc", "raw_char_end": 13945, "raw_char_start": 13673, "value": "draft-ietf-tls-tls13->RFC 8446"}, {"char_end": 3140, "char_start": 2873, "evidence_quote": "{\"id\": 630806, \"originaltargetaliasname\": \"rfc6066\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630806/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}", "fact_id": "informative_reference:6066", "field": "informative_reference", "raw_char_end": 3140, "raw_char_start": 2873, "value": "/api/v1/doc/document/rfc6066/"}, {"char_end": 3409, "char_start": 3142, "evidence_quote": "{\"id\": 630807, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630807/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}", "fact_id": "informative_reference:5246", "field": "informative_reference", "raw_char_end": 3409, "raw_char_start": 3142, "value": "/api/v1/doc/document/rfc5246/"}, {"char_end": 3678, "char_start": 3411, "evidence_quote": "{\"id\": 630805, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630805/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}", "fact_id": "informative_reference:5077", "field": "informative_reference", "raw_char_end": 3678, "raw_char_start": 3411, "value": "/api/v1/doc/document/rfc5077/"}, {"char_end": 6098, "char_start": 5831, "evidence_quote": "{\"id\": 630795, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630795/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}", "fact_id": "normative_reference:5705", "field": "normative_reference", "raw_char_end": 6098, "raw_char_start": 5831, "value": "/api/v1/doc/document/rfc5705/"}, {"char_end": 12864, "char_start": 12597, "evidence_quote": "{\"id\": 630768, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630768/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}", "fact_id": "normative_reference:6961", "field": "normative_reference", "raw_char_end": 12864, "raw_char_start": 12597, "value": "/api/v1/doc/document/rfc6961/"}], "kind": "datatracker_relation", "occurred_at": "2026-09-07T04:14:32.295913Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 0, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131", "record_id": "ietf:datatracker-relation:draft-ietf-tls-tls13:rfc8446", "retrieval_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=draft-ietf-tls-tls13&limit=100", "rfc_number": 8446, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131", "source_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=draft-ietf-tls-tls13&limit=100", "temporal_semantics": "retrieval_observation_only", "text": "{\"meta\": {\"limit\": 100, \"next\": null, \"offset\": 0, \"previous\": null, \"total_count\": 51}, \"objects\": [{\"id\": 630809, \"originaltargetaliasname\": \"bcp72\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630809/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/bcp72/\"}, {\"id\": 630802, \"originaltargetaliasname\": \"bcp26\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630802/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/bcp26/\"}, {\"id\": 630782, \"originaltargetaliasname\": \"bcp106\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630782/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/bcp106/\"}, {\"id\": 630774, \"originaltargetaliasname\": \"bcp14\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630774/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/bcp14/\"}, {\"id\": 282421, \"originaltargetaliasname\": \"draft-ietf-tls-rfc5246-bis\", \"relationship\": \"/api/v1/name/docrelationshipname/replaces/\", \"resource_uri\": \"/api/v1/doc/relateddocument/282421/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/draft-ietf-tls-rfc5246-bis/\"}, {\"id\": 630813, \"originaltargetaliasname\": \"rfc7568\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630813/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7568/\"}, {\"id\": 630810, \"originaltargetaliasname\": \"draft-ietf-tls-iana-registry-updates\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630810/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/draft-ietf-tls-iana-registry-updates/\"}, {\"id\": 630812, \"originaltargetaliasname\": \"rfc5116\", \"relationship\": \"/api/v1/name/docrelationshipname/refunk/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630812/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5116/\"}, {\"id\": 630811, \"originaltargetaliasname\": \"rfc5280\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630811/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5280/\"}, {\"id\": 630808, \"originaltargetaliasname\": \"rfc5756\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630808/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5756/\"}, {\"id\": 630806, \"originaltargetaliasname\": \"rfc6066\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630806/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}, {\"id\": 630807, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630807/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}, {\"id\": 630805, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630805/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}, {\"id\": 630804, \"originaltargetaliasname\": \"rfc8174\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630804/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8174/\"}, {\"id\": 630803, \"originaltargetaliasname\": \"rfc8032\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630803/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8032/\"}, {\"id\": 630801, \"originaltargetaliasname\": \"rfc2119\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630801/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc2119/\"}, {\"id\": 630800, \"originaltargetaliasname\": \"rfc6347\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630800/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6347/\"}, {\"id\": 630799, \"originaltargetaliasname\": \"rfc4492\", \"relationship\": \"/api/v1/name/docrelationshipname/refunk/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630799/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc4492/\"}, {\"id\": 630798, \"originaltargetaliasname\": \"rfc7539\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630798/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7539/\"}, {\"id\": 630797, \"originaltargetaliasname\": \"rfc7301\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630797/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7301/\"}, {\"id\": 630796, \"originaltargetaliasname\": \"rfc6176\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630796/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6176/\"}, {\"id\": 630795, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630795/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}, {\"id\": 630794, \"originaltargetaliasname\": \"rfc6979\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630794/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6979/\"}, {\"id\": 630793, \"originaltargetaliasname\": \"rfc5764\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630793/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5764/\"}, {\"id\": 630792, \"originaltargetaliasname\": \"rfc8017\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630792/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8017/\"}, {\"id\": 630791, \"originaltargetaliasname\": \"draft-ietf-tls-tls13-vectors\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630791/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/draft-ietf-tls-tls13-vectors/\"}, {\"id\": 630790, \"originaltargetaliasname\": \"rfc6520\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630790/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6520/\"}, {\"id\": 630789, \"originaltargetaliasname\": \"rfc5869\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630789/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5869/\"}, {\"id\": 630788, \"originaltargetaliasname\": \"rfc6960\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630788/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6960/\"}, {\"id\": 630787, \"originaltargetaliasname\": \"rfc7748\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630787/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7748/\"}, {\"id\": 630786, \"originaltargetaliasname\": \"rfc8126\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630786/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8126/\"}, {\"id\": 630785, \"originaltargetaliasname\": \"rfc7924\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630785/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7924/\"}, {\"id\": 630784, \"originaltargetaliasname\": \"rfc5929\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630784/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5929/\"}, {\"id\": 630783, \"originaltargetaliasname\": \"rfc8305\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630783/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8305/\"}, {\"id\": 630781, \"originaltargetaliasname\": \"rfc6091\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630781/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6091/\"}, {\"id\": 630780, \"originaltargetaliasname\": \"rfc4086\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630780/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc4086/\"}, {\"id\": 630779, \"originaltargetaliasname\": \"rfc7627\", \"relationship\": \"/api/v1/name/docrelationshipname/refunk/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630779/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7627/\"}, {\"id\": 630778, \"originaltargetaliasname\": \"rfc6655\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630778/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6655/\"}, {\"id\": 630777, \"originaltargetaliasname\": \"rfc7250\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630777/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7250/\"}, {\"id\": 630776, \"originaltargetaliasname\": \"rfc7507\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630776/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7507/\"}, {\"id\": 630775, \"originaltargetaliasname\": \"rfc7685\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630775/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7685/\"}, {\"id\": 630773, \"originaltargetaliasname\": \"rfc2104\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630773/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc2104/\"}, {\"id\": 630771, \"originaltargetaliasname\": \"rfc4346\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630771/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc4346/\"}, {\"id\": 630772, \"originaltargetaliasname\": \"rfc4366\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630772/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc4366/\"}, {\"id\": 630770, \"originaltargetaliasname\": \"rfc6962\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630770/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6962/\"}, {\"id\": 630769, \"originaltargetaliasname\": \"rfc3552\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630769/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc3552/\"}, {\"id\": 630768, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630768/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}, {\"id\": 630767, \"originaltargetaliasname\": \"rfc7919\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630767/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7919/\"}, {\"id\": 630766, \"originaltargetaliasname\": \"rfc7465\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630766/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7465/\"}, {\"id\": 630814, \"originaltargetaliasname\": \"rfc7230\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630814/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7230/\"}, {\"id\": 1295893, \"originaltargetaliasname\": \"rfc8446\", \"relationship\": \"/api/v1/name/docrelationshipname/became_rfc/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1295893/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8446/\"}]}", "text_sha256": "3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131"}, {"facts": [{"char_end": 3417, "char_start": 3161, "evidence_quote": "{\"id\": 1269005, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269005/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}", "fact_id": "informative_reference:5246", "field": "informative_reference", "raw_char_end": 3417, "raw_char_start": 3161, "value": "/api/v1/doc/document/rfc5246/"}, {"char_end": 3675, "char_start": 3419, "evidence_quote": "{\"id\": 1269006, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269006/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}", "fact_id": "informative_reference:5077", "field": "informative_reference", "raw_char_end": 3675, "raw_char_start": 3419, "value": "/api/v1/doc/document/rfc5077/"}, {"char_end": 5223, "char_start": 4967, "evidence_quote": "{\"id\": 1269016, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269016/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}", "fact_id": "normative_reference:5705", "field": "normative_reference", "raw_char_end": 5223, "raw_char_start": 4967, "value": "/api/v1/doc/document/rfc5705/"}, {"char_end": 10425, "char_start": 10169, "evidence_quote": "{\"id\": 1269043, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269043/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}", "fact_id": "normative_reference:6961", "field": "normative_reference", "raw_char_end": 10425, "raw_char_start": 10169, "value": "/api/v1/doc/document/rfc6961/"}, {"char_end": 12717, "char_start": 12466, "evidence_quote": "{\"id\": 1591472, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591472/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}", "fact_id": "normative_reference:6066", "field": "normative_reference", "raw_char_end": 12717, "raw_char_start": 12466, "value": "/api/v1/doc/document/rfc6066/"}], "kind": "datatracker_relation", "occurred_at": "2026-09-07T04:14:41.148823Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 0, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7", "record_id": "ietf:datatracker-relation:rfc8446", "retrieval_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=rfc8446&limit=100", "rfc_number": 8446, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7", "source_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=rfc8446&limit=100", "temporal_semantics": "retrieval_observation_only", "text": "{\"meta\": {\"limit\": 100, \"next\": null, \"offset\": 0, \"previous\": null, \"total_count\": 57}, \"objects\": [{\"id\": 1269002, \"originaltargetaliasname\": \"bcp72\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269002/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/bcp72/\"}, {\"id\": 1269009, \"originaltargetaliasname\": \"bcp26\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269009/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/bcp26/\"}, {\"id\": 1269029, \"originaltargetaliasname\": \"bcp106\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269029/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/bcp106/\"}, {\"id\": 1269037, \"originaltargetaliasname\": \"bcp14\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269037/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/bcp14/\"}, {\"id\": 666521, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/obs/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666521/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}, {\"id\": 666523, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/updates/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666523/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}, {\"id\": 666520, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/obs/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666520/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}, {\"id\": 666524, \"originaltargetaliasname\": \"rfc6066\", \"relationship\": \"/api/v1/name/docrelationshipname/updates/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666524/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}, {\"id\": 666522, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/obs/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666522/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}, {\"id\": 1268998, \"originaltargetaliasname\": \"rfc7568\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1268998/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7568/\"}, {\"id\": 1269001, \"originaltargetaliasname\": \"rfc5280\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269001/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5280/\"}, {\"id\": 1269003, \"originaltargetaliasname\": \"rfc5756\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269003/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5756/\"}, {\"id\": 1269005, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269005/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}, {\"id\": 1269006, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269006/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}, {\"id\": 1269007, \"originaltargetaliasname\": \"rfc8174\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269007/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8174/\"}, {\"id\": 1269008, \"originaltargetaliasname\": \"rfc8032\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269008/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8032/\"}, {\"id\": 1269010, \"originaltargetaliasname\": \"rfc2119\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269010/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc2119/\"}, {\"id\": 1269011, \"originaltargetaliasname\": \"rfc6347\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269011/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6347/\"}, {\"id\": 1269015, \"originaltargetaliasname\": \"rfc6176\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269015/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6176/\"}, {\"id\": 1269016, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269016/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}, {\"id\": 1269018, \"originaltargetaliasname\": \"rfc5764\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269018/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5764/\"}, {\"id\": 1269019, \"originaltargetaliasname\": \"rfc8017\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269019/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8017/\"}, {\"id\": 1269020, \"originaltargetaliasname\": \"draft-ietf-tls-tls13-vectors\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269020/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/draft-ietf-tls-tls13-vectors/\"}, {\"id\": 1269021, \"originaltargetaliasname\": \"rfc6520\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269021/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6520/\"}, {\"id\": 1269023, \"originaltargetaliasname\": \"rfc6960\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269023/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6960/\"}, {\"id\": 1269024, \"originaltargetaliasname\": \"rfc7748\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269024/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7748/\"}, {\"id\": 1269025, \"originaltargetaliasname\": \"rfc8126\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269025/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8126/\"}, {\"id\": 1269026, \"originaltargetaliasname\": \"rfc7924\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269026/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7924/\"}, {\"id\": 1269027, \"originaltargetaliasname\": \"rfc5929\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269027/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5929/\"}, {\"id\": 1269028, \"originaltargetaliasname\": \"rfc8305\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269028/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8305/\"}, {\"id\": 1269030, \"originaltargetaliasname\": \"rfc6091\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269030/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6091/\"}, {\"id\": 1269031, \"originaltargetaliasname\": \"rfc4086\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269031/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc4086/\"}, {\"id\": 1269034, \"originaltargetaliasname\": \"rfc7250\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269034/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7250/\"}, {\"id\": 1269035, \"originaltargetaliasname\": \"rfc7507\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269035/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7507/\"}, {\"id\": 1269036, \"originaltargetaliasname\": \"rfc7685\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269036/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7685/\"}, {\"id\": 1269038, \"originaltargetaliasname\": \"rfc2104\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269038/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc2104/\"}, {\"id\": 1269039, \"originaltargetaliasname\": \"rfc4346\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269039/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc4346/\"}, {\"id\": 1269040, \"originaltargetaliasname\": \"rfc4366\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269040/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc4366/\"}, {\"id\": 1269042, \"originaltargetaliasname\": \"rfc3552\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269042/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc3552/\"}, {\"id\": 1269043, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269043/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}, {\"id\": 1269044, \"originaltargetaliasname\": \"rfc7919\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269044/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7919/\"}, {\"id\": 1269045, \"originaltargetaliasname\": \"rfc7465\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269045/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7465/\"}, {\"id\": 1269046, \"originaltargetaliasname\": \"rfc7230\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269046/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7230/\"}, {\"id\": 1591467, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591467/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5869/\"}, {\"id\": 1591468, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591468/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6655/\"}, {\"id\": 1591469, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591469/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7301/\"}, {\"id\": 1591470, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591470/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6962/\"}, {\"id\": 1591471, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591471/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6979/\"}, {\"id\": 1591472, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591472/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}, {\"id\": 1591473, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591473/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8449/\"}, {\"id\": 1591474, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591474/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5116/\"}, {\"id\": 1591475, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591475/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8422/\"}, {\"id\": 1591476, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591476/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc4492/\"}, {\"id\": 1591477, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591477/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8439/\"}, {\"id\": 1591478, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591478/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8447/\"}, {\"id\": 1591479, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591479/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7627/\"}, {\"id\": 1591480, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591480/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6101/\"}]}", "text_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}]}, "source_manifest_sha256": "287ca4031b223ad818d21781c575905d5ed30eb7e4e0e38fef196cf1ca1865b4"}, "length_bucket": "64k", "local_probe_trust_isolation": "non_independent_local_diagnostic", "motif": "tls13_handshake_succession+obsoletes_updates_closure", "pipeline_capabilities": {"generic_promotion": true, "generic_strict_replay": true}, "production_eligible": false, "promoted": true, "promotion": {"candidate_sha256": "b9ffa84d0b0c4f27ec539da80a163bd8f89346a6400a51d7652440c323dd61d6", "dense_audit_sha256": "9b72d2270d99637f687703da54d985c0b10cf9755010352b634a7eb0bfb342c9", "dense_chunking": {"aggregation": "max_similarity", "max_tokens": 192, "overlap_tokens": 32, "strategy": "tokenizer_token_windows"}, "dense_model_backend": "sentence-transformers-6.0.0", "dense_model_id": "sentence-transformers/all-MiniLM-L6-v2", "dense_model_provider": "huggingface", "dense_model_revision": "1110a243fdf4706b3f48f1d95db1a4f5529b4d41", "dense_ranking_sha256": "3d2a8fb375fb7f15e5d4f03a10a16f4113bc075c539066e56b1cb1d8863a8548", "dense_score_metric": "dot_product", "dense_top_k": 3, "real_source_verified": true, "release_selection_sha256": "46c412e55a5e391ac46af172f4a11623a01c552075d5ba2d80f6604f48b5f89d", "schema_version": "train-ready-promotion-v2", "strict_replay_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "strict_replay_revision": "longworld-strict-replay-v5", "task_candidate_content_commitment": {"content_sha256": "8d7ad3403e486450be310812510f33cc1ce4cd63105676b5d60fd9e2e94335ec", "length_bucket": "64k", "view": "full", "world_id": "ietf-tls13-handshake-succession-v4"}, "task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "sidecar_schema_version": "longworld.task-replay-sidecar.v3"}, "task_semantic_commitment_sha256": "2a663ac9d67af388be26c067991445dbb2eacca593c56829c9ca45ed70649209", "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb"}, "promotion_eligible": true, "query_id": "ietf-tls13-handshake-succession-v4:64k:standard:full", "query_timing": "first", "question": "Resolve the effective TLS 1.3 handshake protocol succession at 2026-01-31T00:00:00Z for this scenario (canonical JSON): {\"protocol\":\"tls\",\"publication\":\"rfc8446\",\"succession\":\"obsoletes_and_updates\",\"version\":\"1.3\"}. Return exactly one JSON object with these keys in this order: [\"current_protocol\",\"obsoletes_session_tickets\",\"obsoletes_tls12\",\"obsoletes_ocsp_stapling\",\"updates_exporters\",\"updates_extensions\"]. Use this exact per-field output codebook (canonical JSON): {\"current_protocol\":{\"code\":\"TLS_1_3_RFC8446\",\"meaning\":\"RFC 8446 is TLS protocol version 1.3\"},\"obsoletes_ocsp_stapling\":{\"code\":\"RFC6961\",\"meaning\":\"RFC 8446 obsoletes multiple certificate status requests\"},\"obsoletes_session_tickets\":{\"code\":\"RFC5077\",\"meaning\":\"RFC 8446 obsoletes TLS session resumption tickets\"},\"obsoletes_tls12\":{\"code\":\"RFC5246\",\"meaning\":\"RFC 8446 obsoletes TLS protocol version 1.2\"},\"updates_exporters\":{\"code\":\"RFC5705\",\"meaning\":\"RFC 8446 updates TLS keying material exporters\"},\"updates_extensions\":{\"code\":\"RFC6066\",\"meaning\":\"RFC 8446 updates TLS extension definitions\"}}. Resolve each field independently from the supplied RFC graph; use the string \"UNKNOWN\" only for a field whose required evidence or relation is absent.", "real_source_family_ids": ["ietf_standards"], "real_source_token_ratio": 0.994230152794102, "real_source_verified": true, "real_source_workflow_ids": ["task-source:73e262c5a2939b145819"], "schema_version": "longworld.ietf-tls13-generation-candidate.v1", "semantic_base_task_id": "250a7cfbf1fb2c0a24b9", "semantic_growth_group_id": "ietf-tls13-handshake-succession-v4|ietf-tls13-handshake-succession", "semantic_tokens": {"causal_supporting": 16788, "event_bearing": 48341, "generic_background": 0, "internal": 65134, "measurement_basis": "exact_pinned_tokenizer_replay_relation_path_whole_artifact_upper_bound", "proof_bearing": 48341}, "source_binding": {"signed_manifest_sha256": "287ca4031b223ad818d21781c575905d5ed30eb7e4e0e38fef196cf1ca1865b4"}, "source_family_ids": ["ietf_standards"], "source_origins": ["real_public"], "source_record_ids": ["ietf:draft:draft-ietf-tls-tls13-28", "ietf:rfc:5077", "ietf:rfc:5246", "ietf:rfc:5705", "ietf:rfc:6066", "ietf:rfc:6961", "ietf:rfc:8446"], "source_record_ids_by_artifact": {"ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2": ["ietf:draft:draft-ietf-tls-tls13-28"], "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8": ["ietf:rfc:5077"], "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681": ["ietf:rfc:5246"], "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1": ["ietf:rfc:5246"], "ietf:rfc:5246:chars:0072013-0109924:596a422d1201": ["ietf:rfc:5246"], "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44": ["ietf:rfc:5705"], "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44": ["ietf:rfc:6066"], "ietf:rfc:6961:chars:0000000-0021461:bde82964b134": ["ietf:rfc:6961"], "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4": ["ietf:rfc:8446"], "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093": ["ietf:rfc:8446"]}, "source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2018-03-01T00:00:00Z", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_relation_kind": "published_as"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "updates"}], "source_relation_id": "6ee046ba0ec12a545607", "source_relation_ids": ["ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "ietf:updates:ietf:rfc:8446:ietf:rfc:6066"], "split": "train", "split_strategy": "world", "strict_replay_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "strict_support_event_count": 6, "task_proof_receipt": {"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "artifact_aligned_windows": {"16k": {"enumeration_sha256": "0dfdf47f3975b6557021b0ac9fa141d5b0a482dbe8b837666f40decdfcb359d1", "max_artifact_count": 3, "max_span_tokens": 16332, "status": "insufficient", "strict_window_count": 19, "window_tokens": 16384}, "4k": {"enumeration_sha256": "c17059cf0a6db37f54932b489e8b206cba9297c08835f29fb38c17d5547143a3", "max_artifact_count": 1, "max_span_tokens": 4053, "status": "insufficient", "strict_window_count": 3, "window_tokens": 4096}, "8k": {"enumeration_sha256": "9827342ea228a6bdd0a545fb68b476c6abfdb1407ae92f5a47381679372bb2e6", "max_artifact_count": 1, "max_span_tokens": 8187, "status": "insufficient", "strict_window_count": 8, "window_tokens": 8192}}, "artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "end_token": 9837, "start_token": 0, "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "end_token": 10302, "start_token": 9841, "text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "end_token": 18494, "start_token": 10307, "text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "end_token": 26639, "start_token": 18498, "text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "end_token": 34776, "start_token": 26643, "text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "end_token": 34883, "start_token": 34780, "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "end_token": 43029, "start_token": 34887, "text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "end_token": 48214, "start_token": 43033, "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "end_token": 52272, "start_token": 48219, "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "end_token": 65134, "start_token": 52277, "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}], "checks": {"answer_surface_free": true, "artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_top3_prefixes_insufficient": true, "body_and_classification_valid": true, "counterfactual_changes_answer": true, "empty_selection_insufficient": true, "exact_token_count_recomputed": true, "full_counterfactual_sufficient": true, "full_replay_sufficient": true, "lexical_tfidf_top3_prefixes_insufficient": true, "minimal_counterfactual_sufficient": true, "minimal_replay_sufficient": true, "minimum_complexity_met": true, "no_shortcut": true, "raw_token_executable_windows_insufficient": true, "remove_one_fails": true, "single_essential_insufficient": true, "source_relations_replayed": true}, "essential_artifact_ids": ["ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"], "proof_input_sha256": "b9ffa84d0b0c4f27ec539da80a163bd8f89346a6400a51d7652440c323dd61d6", "raw_token_offset_windows": {"16k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "4f0b1843910fd703db5d39e76976b5f85e5a61ed1666bf8ab3295611fd0b2b45", "max_artifact_count": 5, "monotonicity_proven": false, "semantic_state_count": 16, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "215935ba84770df4367dad9a46f1ec1181cdc7e4b093098c18f7b2990a0cd5a9", "exact_raw_executable": {"enumeration_sha256": "215935ba84770df4367dad9a46f1ec1181cdc7e4b093098c18f7b2990a0cd5a9", "exhaustive_start_count": 48751, "semantic_state_count": 17, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 17, "window_tokens": 16384}, "4k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "02fa86d64b71a9758c4ff96b5d522ad44bc885838009f1e86e9be94861702d27", "max_artifact_count": 3, "monotonicity_proven": false, "semantic_state_count": 19, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "4c39eda337b321f26a594cbd77edb4ae2fe1bcdcf8a543e1d44e9130cc544740", "exact_raw_executable": {"enumeration_sha256": "4c39eda337b321f26a594cbd77edb4ae2fe1bcdcf8a543e1d44e9130cc544740", "exhaustive_start_count": 61039, "semantic_state_count": 4, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 4, "window_tokens": 4096}, "8k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "74e2a505eb70d9fcdb90b3f195cd69a2713f6b9c3a76c121b450dbc516debccf", "max_artifact_count": 3, "monotonicity_proven": false, "semantic_state_count": 19, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "649f2e876f8201780e2eb55d677038a44195e5a513717b76679ee51fe67baecb", "exact_raw_executable": {"enumeration_sha256": "649f2e876f8201780e2eb55d677038a44195e5a513717b76679ee51fe67baecb", "exhaustive_start_count": 56943, "semantic_state_count": 9, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 9, "window_tokens": 8192}}, "receipt_sha256": "40397fb32e82c7dcf296f3d3f53620ad1d3be0908173b1b96fd226c9c0e4922d", "replay": {"counterfactual_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "empty_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "full_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "minimal_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "remove_one": [{"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}], "single_essential": [{"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}]}, "retrieval": {"bm25": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "score": "117.4899053298914"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "score": "85.661410221096119"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "score": "73.286337280605935"}]}, "lexical_tfidf": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "c1af44132ab8172e277a7aeb52229454aa629ca8cbfe8e7ea88b299909f9abb7"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "score": "0.42605918578853391"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "score": "0.41657976134259173"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "score": "0.32513082216618144"}]}}, "schema_version": "longworld.task-proof-receipt.v5", "sidecar_schema_version": "longworld.task-replay-sidecar.v3", "task_replay_sidecar_sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "tokenizer": {"asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "document_context_tokens": 65134, "model_id": "Qwen/Qwen3.5-4B", "revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "serialized_context_tokens": 65513}, "window_scope": "exact_raw_slice_replay_with_separate_intersection_upper_bound"}, "task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "sidecar_schema_version": "longworld.task-replay-sidecar.v3"}, "task_view_projection": {"adapter_context_header": null, "adapter_replay_answer_sha256": "a627dbccb9ab1c93890f576e49c0775aaf1e09dabcad115191e8972295391099", "artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}], "chronology": [], "derivation_revision": "longworld.task-view-derivation.v4", "document_context_sha256": "705fef2065813e6c2df73740c0c29b453a2adfe59f3999ca76bfa6ccd4847823", "dossier_id": "a00388a585cd2f4d89b3", "materialized_replay_answer_sha256": "a627dbccb9ab1c93890f576e49c0775aaf1e09dabcad115191e8972295391099", "parent_artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "provenance_id": "source-span-sha256:e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef:0:41965:80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49", "source_origin": "real_public", "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "provenance_id": "source-span-sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd:335784:337410:0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5", "source_origin": "real_public", "text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:0:35597:ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6", "source_origin": "real_public", "text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "provenance_id": "source-span-sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd:0:38698:61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e", "source_origin": "real_public", "text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:35597:72013:3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69", "source_origin": "real_public", "text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "provenance_id": "source-span-sha256:9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6:0:342:d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b", "source_origin": "real_public", "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:72013:109924:596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef", "source_origin": "real_public", "text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "provenance_id": "source-span-sha256:c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db:0:21461:bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a", "source_origin": "real_public", "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "provenance_id": "source-span-sha256:29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a:0:16341:4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473", "source_origin": "real_public", "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "provenance_id": "source-span-sha256:6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0:0:55067:a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813", "source_origin": "real_public", "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}], "parent_candidate_sha256": "3a56e24f453b328fb4635d85eaa9c95fe023fe8be199a806064111960bea07df", "parent_source_binding_sha256": "fc34a690fcb8ffffb64df92f1df724ea75d43ce88c10daf432fb085fee09ddbe", "parent_task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "e77fe08f470f227b433656f0fac14e677653bdb1f53faa939175c5c883608e76", "sidecar_schema_version": "longworld.task-replay-sidecar.v1"}, "schema_version": "longworld.task-view-projection.v1", "source_token_measurement_receipt": {"final_prompt_tokens": 65513, "measurement_basis": "final_prompt_real_source_marginal", "parent_artifact_token_contributions": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "parent_text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49", "retained_as_real": true, "token_contribution": 9837}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "parent_text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6", "retained_as_real": true, "token_contribution": 8192}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "parent_text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69", "retained_as_real": true, "token_contribution": 8137}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "parent_text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef", "retained_as_real": true, "token_contribution": 8146}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "parent_text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473", "retained_as_real": true, "token_contribution": 4057}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "parent_text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813", "retained_as_real": true, "token_contribution": 12862}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "parent_text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a", "retained_as_real": true, "token_contribution": 5186}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "parent_text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b", "retained_as_real": true, "token_contribution": 108}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "parent_text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e", "retained_as_real": true, "token_contribution": 8145}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "parent_text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5", "retained_as_real": true, "token_contribution": 464}], "parent_document_context_tokens": 65134, "parent_real_source_token_ratio": 1.0, "projected_document_context_tokens": 65134, "real_source_marginal_tokens": 65135, "real_source_token_ratio": 0.994230152794102, "retained_parent_document_tokens": 65134, "schema_version": "longworld.source-token-measurement-receipt.v2", "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "tokenizer_model_id": "Qwen/Qwen3.5-4B", "tokenizer_revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "without_real_prompt_tokens": 378}, "view": "full"}, "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "tokenizer_context_tokens": 65513, "tokenizer_model_id": "Qwen/Qwen3.5-4B", "tokenizer_revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "train_ready": true, "training_objective": "sft", "trust_scope": "local_probe", "trust_valid_for_production": false, "verification": {"artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_topk_insufficient": true, "candidate_mode": false, "closed_book_unsolved": true, "contiguous_windows_insufficient": true, "counterfactual_changes_answer": true, "counterfactual_replay_sufficient": true, "distractor_invariance_gold": true, "embedding_topk_insufficient": true, "essential_single_doc_insufficient": true, "essential_surface_gold_free": true, "essential_text_grounded": true, "full_sufficient": true, "lexical_tfidf_topk_insufficient": true, "local_window_insufficient": true, "min_complexity": true, "minimal_sufficient": true, "no_shortcut": true, "production_mode": true, "question_only_unsolved": true, "remove_one_fails": true, "schema_ok": true, "semantic_sufficient": true, "strict_executable_sufficient": true, "surface_match": true}, "verified_derived_order_relation_edges": [{"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2018-03-01T00:00:00Z", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_relation_kind": "published_as"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "updates"}], "view": "full", "view_verification": {"classification_ok": true, "content_gate_eligible": true, "essential_present": true, "expected_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "global_proof_green": true, "production_eligible": false, "semantic_text_grounded": true, "strict_replay_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}"}, "workflow_ids": ["task-source:73e262c5a2939b145819"], "workflow_kinds": ["real_source_derived"], "world_id": "ietf-tls13-handshake-succession-v4"} +{"actual_context_tokens": 65513, "answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "answer_program_id": "ietf.tls13_handshake_succession.v1", "artifact_classification": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef:0:41965:80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49", "source_char_end": 41965, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:5077", "source_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "evidence_role": "causal_supporting", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:0:35597:ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6", "source_char_end": 35597, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:5246", "source_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "evidence_role": "causal_supporting", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:35597:72013:3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69", "source_char_end": 72013, "source_char_start": 35597, "source_origin": "real_public", "source_record_id": "ietf:rfc:5246", "source_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:72013:109924:596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef", "source_char_end": 109924, "source_char_start": 72013, "source_origin": "real_public", "source_record_id": "ietf:rfc:5246", "source_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a:0:16341:4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473", "source_char_end": 16341, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:5705", "source_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0:0:55067:a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813", "source_char_end": 55067, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:6066", "source_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db:0:21461:bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a", "source_char_end": 21461, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:6961", "source_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6:0:342:d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b", "source_char_end": 342, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "source_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "evidence_role": "causal_gold", "provenance_id": "source-span-sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd:0:38698:61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e", "source_char_end": 38698, "source_char_start": 0, "source_origin": "real_public", "source_record_id": "ietf:rfc:8446", "source_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "evidence_role": "causal_supporting", "provenance_id": "source-span-sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd:335784:337410:0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5", "source_char_end": 337410, "source_char_start": 335784, "source_origin": "real_public", "source_record_id": "ietf:rfc:8446", "source_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "workflow_id": "task-source:73e262c5a2939b145819", "workflow_kind": "real_source_derived"}], "attestation": {"digest": "a9cf2451613581d301c5d8023c8538095d8a1971c76f6ff985ad050a4c0c356f", "environment": "probe", "key_id": "probe-promotion-c9a5dab33e62d6c983d4696be812c27f", "purpose": "sft_row", "role": "promotion", "scheme": "hmac-sha256-v2"}, "authentic_source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}], "authentic_source_relation_id": "e0dcb1ad1463850449f0", "base_task_id": "776021e7d8e7f94c336a", "cf_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "composition_method": "causal_timeline", "content_gate_eligible": true, "context": "Question:\nResolve the effective TLS 1.3 handshake protocol succession at 2026-01-31T00:00:00Z for this scenario (canonical JSON): {\"protocol\":\"tls\",\"publication\":\"rfc8446\",\"succession\":\"obsoletes_and_updates\",\"version\":\"1.3\"}. Return exactly one JSON object with these keys in this order: [\"current_protocol\",\"obsoletes_session_tickets\",\"obsoletes_tls12\",\"obsoletes_ocsp_stapling\",\"updates_exporters\",\"updates_extensions\"]. Use this exact per-field output codebook (canonical JSON): {\"current_protocol\":{\"code\":\"TLS_1_3_RFC8446\",\"meaning\":\"RFC 8446 is TLS protocol version 1.3\"},\"obsoletes_ocsp_stapling\":{\"code\":\"RFC6961\",\"meaning\":\"RFC 8446 obsoletes multiple certificate status requests\"},\"obsoletes_session_tickets\":{\"code\":\"RFC5077\",\"meaning\":\"RFC 8446 obsoletes TLS session resumption tickets\"},\"obsoletes_tls12\":{\"code\":\"RFC5246\",\"meaning\":\"RFC 8446 obsoletes TLS protocol version 1.2\"},\"updates_exporters\":{\"code\":\"RFC5705\",\"meaning\":\"RFC 8446 updates TLS keying material exporters\"},\"updates_extensions\":{\"code\":\"RFC6066\",\"meaning\":\"RFC 8446 updates TLS extension definitions\"}}. Resolve each field independently from the supplied RFC graph; use the string \"UNKNOWN\" only for a field whose required evidence or relation is absent.\n\nContext (internal records):\nNetwork Working Group J. Salowey\nRequest for Comments: 5077 H. Zhou\nObsoletes: 4507 Cisco Systems\nCategory: Standards Track P. Eronen\n Nokia\n H. Tschofenig\n Nokia Siemens Networks\n January 2008\n\n\n Transport Layer Security (TLS) Session Resumption without\n Server-Side State\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document describes a mechanism that enables the Transport Layer\n Security (TLS) server to resume sessions and avoid keeping per-client\n session state. The TLS server encapsulates the session state into a\n ticket and forwards it to the client. The client can subsequently\n resume a session using the obtained ticket. This document obsoletes\n RFC 4507.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 1]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3. Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . 4\n 3.2. SessionTicket TLS Extension . . . . . . . . . . . . . . . 7\n 3.3. NewSessionTicket Handshake Message . . . . . . . . . . . . 8\n 3.4. Interaction with TLS Session ID . . . . . . . . . . . . . 9\n 4. Recommended Ticket Construction . . . . . . . . . . . . . . . 10\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . 12\n 5.1. Invalidating Sessions . . . . . . . . . . . . . . . . . . 12\n 5.2. Stolen Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.3. Forged Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.4. Denial of Service Attacks . . . . . . . . . . . . . . . . 12\n 5.5. Ticket Protection Key Management . . . . . . . . . . . . . 13\n 5.6. Ticket Lifetime . . . . . . . . . . . . . . . . . . . . . 13\n 5.7. Alternate Ticket Formats and Distribution Schemes . . . . 13\n 5.8. Identity Privacy, Anonymity, and Unlinkability . . . . . . 14\n 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 14\n 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 15\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 15\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 15\n Appendix A. Discussion of Changes to RFC 4507 . . . . . . . . . . 17\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 2]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n1. Introduction\n\n This document defines a way to resume a Transport Layer Security\n (TLS) session without requiring session-specific state at the TLS\n server. This mechanism may be used with any TLS ciphersuite. This\n document applies to both TLS 1.0 defined in [RFC2246], and TLS 1.1\n defined in [RFC4346]. The mechanism makes use of TLS extensions\n defined in [RFC4366] and defines a new TLS message type.\n\n This mechanism is useful in the following situations:\n\n 1. servers that handle a large number of transactions from different\n users\n\n 2. servers that desire to cache sessions for a long time\n\n 3. ability to load balance requests across servers\n\n 4. embedded servers with little memory\n\n This document obsoletes RFC 4507 [RFC4507] to correct an error in the\n encoding that caused the specification to differ from deployed\n implementations. At the time of this writing, there are no known\n implementations that follow the encoding specified in RFC 4507. This\n update to RFC 4507 aligns the document with currently deployed\n implementations. More details of the change are given in Appendix A.\n\n2. Terminology\n\n Within this document, the term 'ticket' refers to a cryptographically\n protected data structure that is created and consumed by the server\n to rebuild session-specific state.\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Protocol\n\n This specification describes a mechanism to distribute encrypted\n session-state information to the client in the form of a ticket and a\n mechanism to present the ticket back to the server. The ticket is\n created by a TLS server and sent to a TLS client. The TLS client\n presents the ticket to the TLS server to resume a session.\n Implementations of this specification are expected to support both\n mechanisms. Other specifications can take advantage of the session\n tickets, perhaps specifying alternative means for distribution or\n selection. For example, a separate specification may describe an\n\n\n\nSalowey, et al. Standards Track [Page 3]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n alternate way to distribute a ticket and use the TLS extension in\n this document to resume the session. This behavior is beyond the\n scope of the document and would need to be described in a separate\n specification.\n\n3.1. Overview\n\n The client indicates that it supports this mechanism by including a\n SessionTicket TLS extension in the ClientHello message. The\n extension will be empty if the client does not already possess a\n ticket for the server. The server sends an empty SessionTicket\n extension to indicate that it will send a new session ticket using\n the NewSessionTicket handshake message. The extension is described\n in Section 3.2.\n\n If the server wants to use this mechanism, it stores its session\n state (such as ciphersuite and master secret) to a ticket that is\n encrypted and integrity-protected by a key known only to the server.\n The ticket is distributed to the client using the NewSessionTicket\n TLS handshake message described in Section 3.3. This message is sent\n during the TLS handshake before the ChangeCipherSpec message, after\n the server has successfully verified the client's Finished message.\n\n Client Server\n\n ClientHello\n (empty SessionTicket extension)-------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1: Message Flow for Full Handshake Issuing New Session Ticket\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 4]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n The client caches this ticket along with the master secret and other\n parameters associated with the current session. When the client\n wishes to resume the session, it includes the ticket in the\n SessionTicket extension within the ClientHello message. Appendix A\n provides a detailed description of the encoding of the extension and\n changes from RFC 4507. The server then decrypts the received ticket,\n verifies the ticket's validity, retrieves the session state from the\n contents of the ticket, and uses this state to resume the session.\n The interaction with the TLS Session ID is described in Section 3.4.\n If the server successfully verifies the client's ticket, then it may\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello.\n\n Client Server\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2: Message Flow for Abbreviated Handshake Using New Session\n Ticket\n\n A recommended ticket format is given in Section 4.\n\n If the server cannot or does not want to honor the ticket, then it\n can initiate a full handshake with the client.\n\n In the case that the server does not wish to issue a new ticket at\n this time, it just completes the handshake without including a\n SessionTicket extension or NewSessionTicket handshake message. This\n is shown below (this flow is identical to Figure 1 in RFC 4346,\n except for the SessionTicket extension in the first message):\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 5]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 3: Message Flow for Server Completing Full Handshake Without\n Issuing New Session Ticket\n\n It is also permissible to have an exchange similar to Figure 3 using\n the abbreviated handshake defined in Figure 2 of RFC 4346, where the\n client uses the SessionTicket extension to resume the session, but\n the server does not wish to issue a new ticket, and therefore does\n not send a SessionTicket extension.\n\n If the server rejects the ticket, it may still wish to issue a new\n ticket after performing the full handshake as shown below (this flow\n is identical to Figure 1, except the SessionTicket extension in the\n ClientHello is not empty):\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 6]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 4: Message Flow for Server Rejecting Ticket, Performing Full\n Handshake, and Issuing New Session Ticket\n\n3.2. SessionTicket TLS Extension\n\n The SessionTicket TLS extension is based on [RFC4366]. The format of\n the ticket is an opaque structure used to carry session-specific\n state information. This extension may be sent in the ClientHello and\n ServerHello.\n\n If the client possesses a ticket that it wants to use to resume a\n session, then it includes the ticket in the SessionTicket extension\n in the ClientHello. If the client does not have a ticket and is\n prepared to receive one in the NewSessionTicket handshake message,\n then it MUST include a zero-length ticket in the SessionTicket\n extension. If the client is not prepared to receive a ticket in the\n NewSessionTicket handshake message, then it MUST NOT include a\n SessionTicket extension unless it is sending a non-empty ticket it\n received through some other means from the server.\n\n The server uses a zero-length SessionTicket extension to indicate to\n the client that it will send a new session ticket using the\n NewSessionTicket handshake message described in Section 3.3. The\n server MUST send this extension in the ServerHello if it wishes to\n issue a new ticket to the client using the NewSessionTicket handshake\n message. The server MUST NOT send this extension if it does not\n receive one in the ClientHello.\n\n\n\n\nSalowey, et al. Standards Track [Page 7]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n If the server fails to verify the ticket, then it falls back to\n performing a full handshake. If the ticket is accepted by the server\n but the handshake fails, the client SHOULD delete the ticket.\n\n The SessionTicket extension has been assigned the number 35. The\n extension_data field of SessionTicket extension contains the ticket.\n\n3.3. NewSessionTicket Handshake Message\n\n This message is sent by the server during the TLS handshake before\n the ChangeCipherSpec message. This message MUST be sent if the\n server included a SessionTicket extension in the ServerHello. This\n message MUST NOT be sent if the server did not include a\n SessionTicket extension in the ServerHello. This message is included\n in the hash used to create and verify the Finished message. In the\n case of a full handshake, the server MUST verify the client's\n Finished message before sending the ticket. The client MUST NOT\n treat the ticket as valid until it has verified the server's Finished\n message. If the server determines that it does not want to include a\n ticket after it has included the SessionTicket extension in the\n ServerHello, then it sends a zero-length ticket in the\n NewSessionTicket handshake message.\n\n If the server successfully verifies the client's ticket, then it MAY\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello in the abbreviated handshake. The client\n should start using the new ticket as soon as possible after it\n verifies the server's Finished message for new connections. Note\n that since the updated ticket is issued before the handshake\n completes, it is possible that the client may not put the new ticket\n into use before it initiates new connections. The server MUST NOT\n assume that the client actually received the updated ticket until it\n successfully verifies the client's Finished message.\n\n The NewSessionTicket handshake message has been assigned the number 4\n and its definition is given at the end of this section. The\n ticket_lifetime_hint field contains a hint from the server about how\n long the ticket should be stored. The value indicates the lifetime\n in seconds as a 32-bit unsigned integer in network byte order\n relative to when the ticket is received. A value of zero is reserved\n to indicate that the lifetime of the ticket is unspecified. A client\n SHOULD delete the ticket and associated state when the time expires.\n It MAY delete the ticket earlier based on local policy. A server MAY\n treat a ticket as valid for a shorter or longer period of time than\n what is stated in the ticket_lifetime_hint.\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 8]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n HandshakeType msg_type;\n uint24 length;\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case session_ticket: NewSessionTicket; /* NEW */\n } body;\n } Handshake;\n\n\n struct {\n uint32 ticket_lifetime_hint;\n opaque ticket<0..2^16-1>;\n } NewSessionTicket;\n\n3.4. Interaction with TLS Session ID\n\n If a server is planning on issuing a session ticket to a client that\n does not present one, it SHOULD include an empty Session ID in the\n ServerHello. If the server rejects the ticket and falls back to the\n full handshake then it may include a non-empty Session ID to indicate\n its support for stateful session resumption. If the client receives\n a session ticket from the server, then it discards any Session ID\n that was sent in the ServerHello.\n\n When presenting a ticket, the client MAY generate and include a\n Session ID in the TLS ClientHello. If the server accepts the ticket\n and the Session ID is not empty, then it MUST respond with the same\n Session ID present in the ClientHello. This allows the client to\n easily differentiate when the server is resuming a session from when\n it is falling back to a full handshake. Since the client generates a\n Session ID, the server MUST NOT rely upon the Session ID having a\n particular value when validating the ticket. If a ticket is\n presented by the client, the server MUST NOT attempt to use the\n Session ID in the ClientHello for stateful session resumption.\n Alternatively, the client MAY include an empty Session ID in the\n ClientHello. In this case, the client ignores the Session ID sent in\n the ServerHello and determines if the server is resuming a session by\n the subsequent handshake messages.\n\n\n\nSalowey, et al. Standards Track [Page 9]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n4. Recommended Ticket Construction\n\n This section describes a recommended format and protection for the\n ticket. Note that the ticket is opaque to the client, so the\n structure is not subject to interoperability concerns, and\n implementations may diverge from this format. If implementations do\n diverge from this format, they must take security concerns seriously.\n Clients MUST NOT examine the ticket under the assumption that it\n complies with this document.\n\n The server uses two different keys: one 128-bit key for Advanced\n Encryption Standard (AES) [AES] in Cipher Block Chaining (CBC) mode\n [CBC] encryption and one 256-bit key for HMAC-SHA-256 [RFC4634].\n\n The ticket is structured as follows:\n\n struct {\n opaque key_name[16];\n opaque iv[16];\n opaque encrypted_state<0..2^16-1>;\n opaque mac[32];\n } ticket;\n\n Here, key_name serves to identify a particular set of keys used to\n protect the ticket. It enables the server to easily recognize\n tickets it has issued. The key_name should be randomly generated to\n avoid collisions between servers. One possibility is to generate new\n random keys and key_name every time the server is started.\n\n The actual state information in encrypted_state is encrypted using\n 128-bit AES in CBC mode with the given IV. The Message\n Authentication Code (MAC) is calculated using HMAC-SHA-256 over\n key_name (16 octets) and IV (16 octets), followed by the length of\n the encrypted_state field (2 octets) and its contents (variable\n length).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 10]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n ProtocolVersion protocol_version;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n opaque master_secret[48];\n ClientIdentity client_identity;\n uint32 timestamp;\n } StatePlaintext;\n\n enum {\n anonymous(0),\n certificate_based(1),\n psk(2)\n } ClientAuthenticationType;\n\n struct {\n ClientAuthenticationType client_authentication_type;\n select (ClientAuthenticationType) {\n case anonymous: struct {};\n case certificate_based:\n ASN.1Cert certificate_list<0..2^24-1>;\n case psk:\n opaque psk_identity<0..2^16-1>; /* from [RFC4279] */\n };\n } ClientIdentity;\n\n The structure StatePlaintext stores the TLS session state including\n the master_secret. The timestamp within this structure allows the\n TLS server to expire tickets. To cover the authentication and key\n exchange protocols provided by TLS, the ClientIdentity structure\n contains the authentication type of the client used in the initial\n exchange (see ClientAuthenticationType). To offer the TLS server\n with the same capabilities for authentication and authorization, a\n certificate list is included in case of public-key-based\n authentication. The TLS server is therefore able to inspect a number\n of different attributes within these certificates. A specific\n implementation might choose to store a subset of this information or\n additional information. Other authentication mechanisms, such as\n Kerberos [RFC2712], would require different client identity data.\n Other TLS extensions may require the inclusion of additional data in\n the StatePlaintext structure.\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 11]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5. Security Considerations\n\n This section addresses security issues related to the usage of a\n ticket. Tickets must be authenticated and encrypted to prevent\n modification or eavesdropping by an attacker. Several attacks\n described below will be possible if this is not carefully done.\n\n Implementations should take care to ensure that the processing of\n tickets does not increase the chance of denial of service as\n described below.\n\n5.1. Invalidating Sessions\n\n The TLS specification requires that TLS sessions be invalidated when\n errors occur. [CSSC] discusses the security implications of this in\n detail. In the analysis within this paper, failure to invalidate\n sessions does not pose a security risk. This is because the TLS\n handshake uses a non-reversible function to derive keys for a session\n so information about one session does not provide an advantage to\n attack the master secret or a different session. If a session\n invalidation scheme is used, the implementation should verify the\n integrity of the ticket before using the contents to invalidate a\n session to ensure that an attacker cannot invalidate a chosen\n session.\n\n5.2. Stolen Tickets\n\n An eavesdropper or man-in-the-middle may obtain the ticket and\n attempt to use it to establish a session with the server; however,\n since the ticket is encrypted and the attacker does not know the\n secret key, a stolen ticket does not help an attacker resume a\n session. A TLS server MUST use strong encryption and integrity\n protection for the ticket to prevent an attacker from using a brute\n force mechanism to obtain the ticket's contents.\n\n5.3. Forged Tickets\n\n A malicious user could forge or alter a ticket in order to resume a\n session, to extend its lifetime, to impersonate another user, or to\n gain additional privileges. This attack is not possible if the\n ticket is protected using a strong integrity protection algorithm\n such as a keyed HMAC-SHA-256.\n\n5.4. Denial of Service Attacks\n\n The key_name field defined in the recommended ticket format helps the\n server efficiently reject tickets that it did not issue. However, an\n adversary could store or generate a large number of tickets to send\n\n\n\nSalowey, et al. Standards Track [Page 12]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n to the TLS server for verification. To minimize the possibility of a\n denial of service, the verification of the ticket should be\n lightweight (e.g., using efficient symmetric key cryptographic\n algorithms).\n\n5.5. Ticket Protection Key Management\n\n A full description of the management of the keys used to protect the\n ticket is beyond the scope of this document. A list of RECOMMENDED\n practices is given below.\n\n o The keys should be generated securely following the randomness\n recommendations in [RFC4086].\n\n o The keys and cryptographic protection algorithms should be at\n least 128 bits in strength. Some ciphersuites and applications\n may require cryptographic protection greater than 128 bits in\n strength.\n\n o The keys should not be used for any purpose other than generating\n and verifying tickets.\n\n o The keys should be changed regularly.\n\n o The keys should be changed if the ticket format or cryptographic\n protection algorithms change.\n\n5.6. Ticket Lifetime\n\n The TLS server controls the lifetime of the ticket. Servers\n determine the acceptable lifetime based on the operational and\n security requirements of the environments in which they are deployed.\n The ticket lifetime may be longer than the 24-hour lifetime\n recommended in [RFC4346]. TLS clients may be given a hint of the\n lifetime of the ticket. Since the lifetime of a ticket may be\n unspecified, a client has its own local policy that determines when\n it discards tickets.\n\n5.7. Alternate Ticket Formats and Distribution Schemes\n\n If the ticket format or distribution scheme defined in this document\n is not used, then great care must be taken in analyzing the security\n of the solution. In particular, if confidential information, such as\n a secret key, is transferred to the client, it MUST be done using\n secure communication so as to prevent attackers from obtaining or\n modifying the key. Also, the ticket MUST have its integrity and\n confidentiality protected with strong cryptographic techniques to\n prevent a breach in the security of the system.\n\n\n\nSalowey, et al. Standards Track [Page 13]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5.8. Identity Privacy, Anonymity, and Unlinkability\n\n This document mandates that the content of the ticket is\n confidentiality protected in order to avoid leakage of its content,\n such as user-relevant information. As such, it prevents disclosure\n of potentially sensitive information carried within the ticket.\n\n The initial handshake exchange, which was used to obtain the ticket,\n might not provide identity confidentiality of the client based on the\n properties of TLS. Another relevant security threat is the ability\n for an on-path adversary to observe multiple TLS handshakes where the\n same ticket is used, therefore concluding they belong to the same\n communication endpoints. Application designers that use the ticket\n mechanism described in this document should consider that\n unlinkability [ANON] is not necessarily provided.\n\n While a full discussion of these topics is beyond the scope of this\n document, it should be noted that it is possible to issue a ticket\n using a TLS renegotiation handshake that occurs after a secure tunnel\n has been established by a previous handshake. This may help address\n some privacy and unlinkability issues in some environments.\n\n6. Acknowledgements\n\n The authors would like to thank the following people for their help\n with preparing and reviewing this document: Eric Rescorla, Mohamad\n Badra, Tim Dierks, Nelson Bolyard, Nancy Cam-Winget, David McGrew,\n Rob Dugal, Russ Housley, Amir Herzberg, Bernard Aboba, and members of\n the TLS working group.\n\n [CSSC] describes a solution that is very similar to the one described\n in this document and gives a detailed analysis of the security\n considerations involved. [RFC2712] describes a mechanism for using\n Kerberos [RFC4120] in TLS ciphersuites, which helped inspire the use\n of tickets to avoid server state. [RFC4851] makes use of a similar\n mechanism to avoid maintaining server state for the cryptographic\n tunnel. [SC97] also investigates the concept of stateless sessions.\n\n The authors would also like to thank Jan Nordqvist, who found the\n encoding error in RFC 4507, corrected by this document. In addition\n Nagendra Modadugu, Wan-Teh Chang, and Michael D'Errico provided\n useful feedback during the review of this document.\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 14]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n7. IANA Considerations\n\n IANA has assigned a TLS extension number of 35 to the SessionTicket\n TLS extension from the TLS registry of ExtensionType values defined\n in [RFC4366].\n\n IANA has assigned a TLS HandshakeType number 4 to the\n NewSessionTicket handshake type from the TLS registry of\n HandshakeType values defined in [RFC4346].\n\n This document does not require any actions or assignments from IANA.\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2246] Dierks, T. and C. Allen, \"The TLS Protocol Version 1.0\",\n RFC 2246, January 1999.\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346, April 2006.\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, April 2006.\n\n [RFC4507] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 4507, May 2006.\n\n8.2. Informative References\n\n [AES] National Institute of Standards and Technology, \"Advanced\n Encryption Standard (AES)\", Federal Information Processing\n Standards (FIPS) Publication 197, November 2001.\n\n [ANON] Pfitzmann, A. and M. Hansen, \"Anonymity, Unlinkability,\n Unobservability, Pseudonymity, and Identity Management - A\n Consolidated Proposal for Terminology\", http://\n dud.inf.tu-dresden.de/literatur/\n Anon_Terminology_v0.26-1.pdf Version 0.26, December 2005.\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 15]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n [CBC] National Institute of Standards and Technology,\n \"Recommendation for Block Cipher Modes of Operation -\n Methods and Techniques\", NIST Special Publication 800-38A,\n December 2001.\n\n [CSSC] Shacham, H., Boneh, D., and E. Rescorla, \"Client-side\n caching for TLS\", Transactions on Information and System\n Security (TISSEC) , Volume 7, Issue 4, November 2004.\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [RFC4086] Eastlake, D., Schiller, J., and S. Crocker, \"Randomness\n Requirements for Security\", BCP 106, RFC 4086, June 2005.\n\n [RFC4120] Neuman, C., Yu, T., Hartman, S., and K. Raeburn, \"The\n Kerberos Network Authentication Service (V5)\", RFC 4120,\n July 2005.\n\n [RFC4279] Eronen, P. and H. Tschofenig, \"Pre-Shared Key Ciphersuites\n for Transport Layer Security (TLS)\", RFC 4279,\n December 2005.\n\n [RFC4634] Eastlake, D. and T. Hansen, \"US Secure Hash Algorithms\n (SHA and HMAC-SHA)\", RFC 4634, July 2006.\n\n [RFC4851] Cam-Winget, N., McGrew, D., Salowey, J., and H. Zhou, \"The\n Flexible Authentication via Secure Tunneling Extensible\n Authentication Protocol Method (EAP-FAST)\", RFC 4851,\n May 2007.\n\n [SC97] Aura, T. and P. Nikander, \"Stateless Connections\",\n Proceedings of the First International Conference on\n Information and Communication Security (ICICS '97) , 1997.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 16]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAppendix A. Discussion of Changes to RFC 4507\n\n RFC 4507 [RFC4507] defines a mechanism to resume a TLS session\n without maintaining server side state by specifying an encrypted\n ticket that is maintained on the client. The client presents this\n ticket to the server in a SessionTicket hello extension. The\n encoding in RFC 4507 used the XDR style encoding specified in TLS\n [RFC4346].\n\n An error in the encoding caused the specification to differ from\n deployed implementations. At the time of this writing there are no\n known implementations that follow the encoding specified in RFC 4507.\n This update to RFC 4507 aligns the document with these currently\n deployed implementations.\n\n Erroneous encoding in RFC 4507 resulted in two length fields; one for\n the extension contents and one for the ticket itself. Hence, for a\n ticket that is 256 bytes long and begins with the hex value FF FF,\n the encoding of the extension would be as follows according to RFC\n 4507:\n\n 00 23 Ticket Extension type 35\n 01 02 Length of extension contents\n 01 00 Length of ticket\n FF FF .. .. Actual ticket\n\n The update proposed in this document reflects what implementations\n actually encode, namely it removes the redundant length field. So,\n for a ticket that is 256 bytes long and begins with the hex value FF\n FF, the encoding of the extension would be as follows according to\n this update:\n\n 00 23 Extension type 35\n 01 00 Length of extension contents (ticket)\n FF FF .. .. Actual ticket\n\n A server implemented according to RFC 4507 receiving a ticket\n extension from a client conforming to this document would interpret\n the first two bytes of the ticket as the length of this ticket. This\n will result in either an inconsistent length field or in the\n processing of a ticket missing the first two bytes. In the first\n case, the server should reject the request based on a malformed\n length. In the second case, the server should reject the ticket\n based on a malformed ticket, incorrect key version, or failed\n decryption. A server implementation based on this update receiving\n an RFC 4507 extension would interpret the first length field as the\n\n\n\n\n\nSalowey, et al. Standards Track [Page 17]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n length of the ticket and include the second two length bytes as the\n first bytes in the ticket, resulting in the ticket being rejected\n based on a malformed ticket, incorrect key version, or failed\n decryption.\n\n Note that the encoding of an empty SessionTicket extension was\n ambiguous in RFC 4507. An RFC 4507 implementation may have encoded\n it as:\n\n 00 23 Extension type 35\n 00 02 Length of extension contents\n 00 00 Length of ticket\n\n or it may have encoded it the same way as this update:\n\n 00 23 Extension type 35\n 00 00 Length of extension contents\n\n A server wishing to support RFC 4507 clients should respond to an\n empty SessionTicket extension encoded the same way as it received it.\n\n A server implementation can construct tickets such that it can detect\n an RFC 4507 implementation, if one existed, by including a cookie at\n the beginning of the tickets that can be differentiated from a valid\n length. For example, if an implementation constructed tickets to\n start with the hex values FF FF, then it could determine where the\n ticket begins and determine the length correctly from the type of\n length fields present.\n\n This document makes a few additional changes to RFC 4507 listed\n below.\n\n o Clarifying that the server can allow session resumption using a\n ticket without issuing a new ticket in Section 3.1.\n\n o Clarifying that the lifetime is relative to when the ticket is\n received in section 3.3.\n\n o Clarifying that the NewSessionTicket handshake message is included\n in the hash generated for the Finished messages in Section 3.3.\n\n o Clarifying the interaction with TLS Session ID in Section 3.4.\n\n o Recommending the use of SHA-256 for the integrity protection of\n the ticket in Section 4.\n\n o Clarifying that additional data can be included in the\n StatePlaintext structure in Section 4.\n\n\n\nSalowey, et al. Standards Track [Page 18]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAuthors' Addresses\n\n Joseph Salowey\n Cisco Systems\n 2901 3rd Ave\n Seattle, WA 98121\n US\n\n EMail: [redacted-email]\n\n\n Hao Zhou\n Cisco Systems\n 4125 Highlander Parkway\n Richfield, OH 44286\n US\n\n EMail: [redacted-email]\n\n\n Pasi Eronen\n Nokia Research Center\n P.O. Box 407\n FIN-00045 Nokia Group\n Finland\n\n EMail: [redacted-email]\n\n\n Hannes Tschofenig\n Nokia Siemens Networks\n Otto-Hahn-Ring 6\n Munich, Bayern 81739\n Germany\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 19]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nFull Copyright Statement\n\n Copyright (C) The IETF Trust (2008).\n\n This document is subject to the rights, licenses and restrictions\n contained in BCP 78, and except as set forth therein, the authors\n retain all their rights.\n\n This document and the information contained herein are provided on an\n \"AS IS\" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\n OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\n THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\n OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\n THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\n WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\n\nIntellectual Property\n\n The IETF takes no position regarding the validity or scope of any\n Intellectual Property Rights or other rights that might be claimed to\n pertain to the implementation or use of the technology described in\n this document or the extent to which any license under such rights\n might or might not be available; nor does it represent that it has\n made any independent effort to identify any such rights. Information\n on the procedures with respect to rights in RFC documents can be\n found in BCP 78 and BCP 79.\n\n Copies of IPR disclosures made to the IETF Secretariat and any\n assurances of licenses to be made available, or the result of an\n attempt made to obtain a general license or permission for the use of\n such proprietary rights by implementers or users of this\n specification can be obtained from the IETF on-line IPR repository at\n http://www.ietf.org/ipr.\n\n The IETF invites any interested party to bring to its attention any\n copyrights, patents or patent applications, or other proprietary\n rights that may cover technology that may be required to implement\n this standard. Please address the information to the IETF at\n ietf-ipr@ietf.org.\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 20]\n\n===== DOCUMENT =====\n\nNetwork Working Group T. Dierks\nRequest for Comments: 5246 Independent\nObsoletes: 3268, 4346, 4366 E. Rescorla\nUpdates: 4492 RTFM, Inc.\nCategory: Standards Track August 2008\n\n\n The Transport Layer Security (TLS) Protocol\n Version 1.2\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document specifies Version 1.2 of the Transport Layer Security\n (TLS) protocol. The TLS protocol provides communications security\n over the Internet. The protocol allows client/server applications to\n communicate in a way that is designed to prevent eavesdropping,\n tampering, or message forgery.\n\nTable of Contents\n\n 1. Introduction ....................................................4\n 1.1. Requirements Terminology ...................................5\n 1.2. Major Differences from TLS 1.1 .............................5\n 2. Goals ...........................................................6\n 3. Goals of This Document ..........................................7\n 4. Presentation Language ...........................................7\n 4.1. Basic Block Size ...........................................7\n 4.2. Miscellaneous ..............................................8\n 4.3. Vectors ....................................................8\n 4.4. Numbers ....................................................9\n 4.5. Enumerateds ................................................9\n 4.6. Constructed Types .........................................10\n 4.6.1. Variants ...........................................10\n 4.7. Cryptographic Attributes ..................................12\n 4.8. Constants .................................................14\n 5. HMAC and the Pseudorandom Function .............................14\n 6. The TLS Record Protocol ........................................15\n 6.1. Connection States .........................................16\n 6.2. Record Layer ..............................................19\n 6.2.1. Fragmentation ......................................19\n\n\n\nDierks & Rescorla Standards Track [Page 1]\n\f\nRFC 5246 TLS August 2008\n\n\n 6.2.2. Record Compression and Decompression ...............20\n 6.2.3. Record Payload Protection ..........................21\n 6.2.3.1. Null or Standard Stream Cipher ............22\n 6.2.3.2. CBC Block Cipher ..........................22\n 6.2.3.3. AEAD Ciphers ..............................24\n 6.3. Key Calculation ...........................................25\n 7. The TLS Handshaking Protocols ..................................26\n 7.1. Change Cipher Spec Protocol ...............................27\n 7.2. Alert Protocol ............................................28\n 7.2.1. Closure Alerts .....................................29\n 7.2.2. Error Alerts .......................................30\n 7.3. Handshake Protocol Overview ...............................33\n 7.4. Handshake Protocol ........................................37\n 7.4.1. Hello Messages .....................................38\n 7.4.1.1. Hello Request .............................38\n 7.4.1.2. Client Hello ..............................39\n 7.4.1.3. Server Hello ..............................42\n 7.4.1.4. Hello Extensions ..........................44\n 7.4.1.4.1. Signature Algorithms ...........45\n 7.4.2. Server Certificate .................................47\n 7.4.3. Server Key Exchange Message ........................50\n 7.4.4. Certificate Request ................................53\n 7.4.5. Server Hello Done ..................................55\n 7.4.6. Client Certificate .................................55\n 7.4.7. Client Key Exchange Message ........................57\n 7.4.7.1. RSA-Encrypted Premaster Secret Message ....58\n 7.4.7.2. Client Diffie-Hellman Public Value ........61\n 7.4.8. Certificate Verify .................................62\n 7.4.9. Finished ...........................................63\n 8. Cryptographic Computations .....................................64\n 8.1. Computing the Master Secret ...............................64\n 8.1.1. RSA ................................................65\n 8.1.2. Diffie-Hellman .....................................65\n 9. Mandatory Cipher Suites ........................................65\n 10. Application Data Protocol .....................................65\n 11. Security Considerations .......................................65\n 12. IANA Considerations ...........................................65\n Appendix A. Protocol Data Structures and Constant Values ..........68\n A.1. Record Layer ..............................................68\n A.2. Change Cipher Specs Message ...............................69\n A.3. Alert Messages ............................................69\n A.4. Handshake Protocol ........................................70\n A.4.1. Hello Messages .....................................71\n A.4.2. Server Authentication and Key Exchange Messages ....72\n A.4.3. Client Authentication and Key Exchange Messages ....74\n A.4.4. Handshake Finalization Message .....................74\n A.5. The Cipher Suite ..........................................75\n A.6. The Security Parameters ...................................77\n\n\n\nDierks & Rescorla Standards Track [Page 2]\n\f\nRFC 5246 TLS August 2008\n\n\n A.7. Changes to RFC 4492 .......................................78\n Appendix B. Glossary ..............................................78\n Appendix C. Cipher Suite Definitions ..............................83\n Appendix D. Implementation Notes ..................................85\n D.1. Random Number Generation and Seeding ......................85\n D.2. Certificates and Authentication ...........................85\n D.3. Cipher Suites .............................................85\n D.4. Implementation Pitfalls ...................................85\n Appendix E. Backward Compatibility ................................87\n E.1. Compatibility with TLS 1.0/1.1 and SSL 3.0 ................87\n E.2. Compatibility with SSL 2.0 ................................88\n E.3. Avoiding Man-in-the-Middle Version Rollback ...............90\n Appendix F. Security Analysis .....................................91\n F.1. Handshake Protocol ........................................91\n F.1.1. Authentication and Key Exchange ....................91\n F.1.1.1. Anonymous Key Exchange ....................91\n F.1.1.2. RSA Key Exchange and Authentication .......92\n F.1.1.3. Diffie-Hellman Key Exchange with\n Authentication ............................92\n F.1.2. Version Rollback Attacks ...........................93\n F.1.3. Detecting Attacks Against the Handshake Protocol ...94\n F.1.4. Resuming Sessions ..................................94\n F.2. Protecting Application Data ...............................94\n F.3. Explicit IVs ..............................................95\n F.4. Security of Composite Cipher Modes ........................95\n F.5. Denial of Service .........................................96\n F.6. Final Notes ...............................................96\n Normative References ..............................................97\n Informative References ............................................98\n Working Group Information ........................................101\n Contributors .....................................................101\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 3]\n\f\nRFC 5246 TLS August 2008\n\n\n1. Introduction\n\n The primary goal of the TLS protocol is to provide privacy and data\n integrity between two communicating applications. The protocol is\n composed of two layers: the TLS Record Protocol and the TLS Handshake\n Protocol. At the lowest level, layered on top of some reliable\n transport protocol (e.g., TCP [TCP]), is the TLS Record Protocol.\n The TLS Record Protocol provides connection security that has two\n basic properties:\n\n - The connection is private. Symmetric cryptography is used for\n data encryption (e.g., AES [AES], RC4 [SCH], etc.). The keys for\n this symmetric encryption are generated uniquely for each\n connection and are based on a secret negotiated by another\n protocol (such as the TLS Handshake Protocol). The Record\n Protocol can also be used without encryption.\n\n - The connection is reliable. Message transport includes a message\n integrity check using a keyed MAC. Secure hash functions (e.g.,\n SHA-1, etc.) are used for MAC computations. The Record Protocol\n can operate without a MAC, but is generally only used in this mode\n while another protocol is using the Record Protocol as a transport\n for negotiating security parameters.\n\n The TLS Record Protocol is used for encapsulation of various higher-\n level protocols. One such encapsulated protocol, the TLS Handshake\n Protocol, allows the server and client to authenticate each other and\n to negotiate an encryption algorithm and cryptographic keys before\n the application protocol transmits or receives its first byte of\n data. The TLS Handshake Protocol provides connection security that\n has three basic properties:\n\n - The peer's identity can be authenticated using asymmetric, or\n public key, cryptography (e.g., RSA [RSA], DSA [DSS], etc.). This\n authentication can be made optional, but is generally required for\n at least one of the peers.\n\n - The negotiation of a shared secret is secure: the negotiated\n secret is unavailable to eavesdroppers, and for any authenticated\n connection the secret cannot be obtained, even by an attacker who\n can place himself in the middle of the connection.\n\n - The negotiation is reliable: no attacker can modify the\n negotiation communication without being detected by the parties to\n the communication.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 4]\n\f\nRFC 5246 TLS August 2008\n\n\n One advantage of TLS is that it is application protocol independent.\n Higher-level protocols can layer on top of the TLS protocol\n transparently. The TLS standard, however, does not specify how\n protocols add security with TLS; the decisions on how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n1.1. Requirements Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [REQ].\n\n1.2. Major Differences from TLS 1.1\n\n This document is a revision of the TLS 1.1 [TLS1.1] protocol which\n contains improved flexibility, particularly for negotiation of\n cryptographic algorithms. The major changes are:\n\n - The MD5/SHA-1 combination in the pseudorandom function (PRF) has\n been replaced with cipher-suite-specified PRFs. All cipher suites\n in this document use P_SHA256.\n\n - The MD5/SHA-1 combination in the digitally-signed element has been\n replaced with a single hash. Signed elements now include a field\n that explicitly specifies the hash algorithm used.\n\n - Substantial cleanup to the client's and server's ability to\n specify which hash and signature algorithms they will accept.\n Note that this also relaxes some of the constraints on signature\n and hash algorithms from previous versions of TLS.\n\n - Addition of support for authenticated encryption with additional\n data modes.\n\n - TLS Extensions definition and AES Cipher Suites were merged in\n from external [TLSEXT] and [TLSAES].\n\n - Tighter checking of EncryptedPreMasterSecret version numbers.\n\n - Tightened up a number of requirements.\n\n - Verify_data length now depends on the cipher suite (default is\n still 12).\n\n - Cleaned up description of Bleichenbacher/Klima attack defenses.\n\n\n\n\nDierks & Rescorla Standards Track [Page 5]\n\f\nRFC 5246 TLS August 2008\n\n\n - Alerts MUST now be sent in many cases.\n\n - After a certificate_request, if no certificates are available,\n clients now MUST send an empty certificate list.\n\n - TLS_RSA_WITH_AES_128_CBC_SHA is now the mandatory to implement\n cipher suite.\n\n - Added HMAC-SHA256 cipher suites.\n\n - Removed IDEA and DES cipher suites. They are now deprecated and\n will be documented in a separate document.\n\n - Support for the SSLv2 backward-compatible hello is now a MAY, not\n a SHOULD, with sending it a SHOULD NOT. Support will probably\n become a SHOULD NOT in the future.\n\n - Added limited \"fall-through\" to the presentation language to allow\n multiple case arms to have the same encoding.\n\n - Added an Implementation Pitfalls sections\n\n - The usual clarifications and editorial work.\n\n2. Goals\n\n The goals of the TLS protocol, in order of priority, are as follows:\n\n 1. Cryptographic security: TLS should be used to establish a secure\n connection between two parties.\n\n 2. Interoperability: Independent programmers should be able to\n develop applications utilizing TLS that can successfully exchange\n cryptographic parameters without knowledge of one another's code.\n\n 3. Extensibility: TLS seeks to provide a framework into which new\n public key and bulk encryption methods can be incorporated as\n necessary. This will also accomplish two sub-goals: preventing\n the need to create a new protocol (and risking the introduction of\n possible new weaknesses) and avoiding the need to implement an\n entire new security library.\n\n 4. Relative efficiency: Cryptographic operations tend to be highly\n CPU intensive, particularly public key operations. For this\n reason, the TLS protocol has incorporated an optional session\n caching scheme to reduce the number of connections that need to be\n established from scratch. Additionally, care has been taken to\n reduce network activity.\n\n\n\nDierks & Rescorla Standards Track [Page 6]\n\f\nRFC 5246 TLS August 2008\n\n\n3. Goals of This Document\n\n This document and the TLS protocol itself are based on the SSL 3.0\n Protocol Specification as published by Netscape. The differences\n between this protocol and SSL 3.0 are not dramatic, but they are\n significant enough that the various versions of TLS and SSL 3.0 do\n not interoperate (although each protocol incorporates a mechanism by\n which an implementation can back down to prior versions). This\n document is intended primarily for readers who will be implementing\n the protocol and for those doing cryptographic analysis of it. The\n specification has been written with this in mind, and it is intended\n to reflect the needs of those two groups. For that reason, many of\n the algorithm-dependent data structures and rules are included in the\n body of the text (as opposed to in an appendix), providing easier\n access to them.\n\n This document is not intended to supply any details of service\n definition or of interface definition, although it does cover select\n areas of policy as they are required for the maintenance of solid\n security.\n\n4. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used. The syntax draws from\n several sources in its structure. Although it resembles the\n programming language \"C\" in its syntax and XDR [XDR] in both its\n syntax and intent, it would be risky to draw too many parallels. The\n purpose of this presentation language is to document TLS only; it has\n no general application beyond that particular goal.\n\n4.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 7]\n\f\nRFC 5246 TLS August 2008\n\n\n4.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" double\n brackets.\n\n Single-byte entities containing uninterpreted data are of type\n opaque.\n\n4.3. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* 3 consecutive 3 byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n In the following example, mandatory is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 4.4). On the\n other hand, longer can represent up to 800 bytes of data, or 400\n uint16 elements, and it may be empty. Its encoding will include a\n\n\n\nDierks & Rescorla Standards Track [Page 8]\n\f\nRFC 5246 TLS August 2008\n\n\n two-byte actual length field prepended to the vector. The length of\n an encoded vector must be an even multiple of the length of a single\n element (for example, a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is 2 bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n4.4. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are formed from fixed-length series of bytes\n concatenated as described in Section 4.1 and are also unsigned. The\n following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are stored in\n network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n Note that in some cases (e.g., DH parameters) it is necessary to\n represent integers as opaque vectors. In such cases, they are\n represented as unsigned integers (i.e., leading zero octets are not\n required even if the most significant bit is set).\n\n4.5. Enumerateds\n\n An additional sparse data type is available called enum. A field of\n type enum can only assume the values declared in the definition.\n Each definition is a different type. Only enumerateds of the same\n type may be assigned or compared. Every element of an enumerated\n must be assigned a value, as demonstrated in the following example.\n Since the elements of the enumerated are not ordered, they can be\n assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n\n\n\nDierks & Rescorla Standards Track [Page 9]\n\f\nRFC 5246 TLS August 2008\n\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n For enumerateds that are never converted to external representation,\n the numerical information may be omitted.\n\n enum { low, medium, high } Amount;\n\n4.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax for definition is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } [[T]];\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n Structure definitions may be embedded.\n\n4.6.1. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. There\n must be a case arm for every element of the enumeration declared in\n the select. Case arms have limited fall-through: if two case arms\n follow in immediate succession with no fields in between, then they\n\n\n\nDierks & Rescorla Standards Track [Page 10]\n\f\nRFC 5246 TLS August 2008\n\n\n both contain the same fields. Thus, in the example below, \"orange\"\n and \"banana\" both contain V2. Note that this is a new piece of\n syntax in TLS 1.2.\n\n The body of the variant structure may be given a label for reference.\n The mechanism by which the variant is selected at runtime is not\n prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1;\n case e2: Te2;\n case e3: case e4: Te3;\n ....\n case en: Ten;\n } [[fv]];\n } [[Tv]];\n\n For example:\n\n enum { apple, orange, banana } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n select (VariantTag) { /* value of selector is implicit */\n case apple:\n V1; /* VariantBody, tag = apple */\n case orange:\n case banana:\n V2; /* VariantBody, tag = orange or banana */\n } variant_body; /* optional label on variant */\n } VariantRecord;\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 11]\n\f\nRFC 5246 TLS August 2008\n\n\n4.7. Cryptographic Attributes\n\n The five cryptographic operations -- digital signing, stream cipher\n encryption, block cipher encryption, authenticated encryption with\n additional data (AEAD) encryption, and public key encryption -- are\n designated digitally-signed, stream-ciphered, block-ciphered, aead-\n ciphered, and public-key-encrypted, respectively. A field's\n cryptographic processing is specified by prepending an appropriate\n key word designation before the field's type specification.\n Cryptographic keys are implied by the current session state (see\n Section 6.1).\n\n A digitally-signed element is encoded as a struct DigitallySigned:\n\n struct {\n SignatureAndHashAlgorithm algorithm;\n opaque signature<0..2^16-1>;\n } DigitallySigned;\n\n The algorithm field specifies the algorithm used (see Section\n 7.4.1.4.1 for the definition of this field). Note that the\n introduction of the algorithm field is a change from previous\n versions. The signature is a digital signature using those\n algorithms over the contents of the element. The contents themselves\n do not appear on the wire but are simply calculated. The length of\n the signature is specified by the signing algorithm and key.\n\n In RSA signing, the opaque vector contains the signature generated\n using the RSASSA-PKCS1-v1_5 signature scheme defined in [PKCS1]. As\n discussed in [PKCS1], the DigestInfo MUST be DER-encoded [X680]\n [X690]. For hash algorithms without parameters (which includes\n SHA-1), the DigestInfo.AlgorithmIdentifier.parameters field MUST be\n NULL, but implementations MUST accept both without parameters and\n with NULL parameters. Note that earlier versions of TLS used a\n different RSA signature scheme that did not include a DigestInfo\n encoding.\n\n In DSA, the 20 bytes of the SHA-1 hash are run directly through the\n Digital Signing Algorithm with no additional hashing. This produces\n two values, r and s. The DSA signature is an opaque vector, as\n above, the contents of which are the DER encoding of:\n\n Dss-Sig-Value ::= SEQUENCE {\n r INTEGER,\n s INTEGER\n }\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 12]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: In current terminology, DSA refers to the Digital Signature\n Algorithm and DSS refers to the NIST standard. In the original SSL\n and TLS specs, \"DSS\" was used universally. This document uses \"DSA\"\n to refer to the algorithm, \"DSS\" to refer to the standard, and it\n uses \"DSS\" in the code point definitions for historical continuity.\n\n In stream cipher encryption, the plaintext is exclusive-ORed with an\n identical amount of output generated from a cryptographically secure\n keyed pseudorandom number generator.\n\n In block cipher encryption, every block of plaintext encrypts to a\n block of ciphertext. All block cipher encryption is done in CBC\n (Cipher Block Chaining) mode, and all items that are block-ciphered\n will be an exact multiple of the cipher block length.\n\n In AEAD encryption, the plaintext is simultaneously encrypted and\n integrity protected. The input may be of any length, and aead-\n ciphered output is generally larger than the input in order to\n accommodate the integrity check value.\n\n In public key encryption, a public key algorithm is used to encrypt\n data in such a way that it can be decrypted only with the matching\n private key. A public-key-encrypted element is encoded as an opaque\n vector <0..2^16-1>, where the length is specified by the encryption\n algorithm and key.\n\n RSA encryption is done using the RSAES-PKCS1-v1_5 encryption scheme\n defined in [PKCS1].\n\n In the following example\n\n stream-ciphered struct {\n uint8 field1;\n uint8 field2;\n digitally-signed opaque {\n uint8 field3<0..255>;\n uint8 field4;\n };\n } UserType;\n\n The contents of the inner struct (field3 and field4) are used as\n input for the signature/hash algorithm, and then the entire structure\n is encrypted with a stream cipher. The length of this structure, in\n bytes, would be equal to two bytes for field1 and field2, plus two\n bytes for the signature and hash algorithm, plus two bytes for the\n length of the signature, plus the length of the output of the signing\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 13]\n\f\nRFC 5246 TLS August 2008\n\n\n algorithm. The length of the signature is known because the\n algorithm and key used for the signing are known prior to encoding or\n decoding this structure.\n\n4.8. Constants\n\n Typed constants can be defined for purposes of specification by\n declaring a symbol of the desired type and assigning values to it.\n\n Under-specified types (opaque, variable-length vectors, and\n structures that contain opaque) cannot be assigned values. No fields\n of a multi-element structure or vector may be elided.\n\n For example:\n\n struct {\n uint8 f1;\n uint8 f2;\n } Example1;\n\n Example1 ex1 = {1, 4}; /* assigns f1 = 1, f2 = 4 */\n\n5. HMAC and the Pseudorandom Function\n\n The TLS record layer uses a keyed Message Authentication Code (MAC)\n to protect message integrity. The cipher suites defined in this\n document use a construction known as HMAC, described in [HMAC], which\n is based on a hash function. Other cipher suites MAY define their\n own MAC constructions, if needed.\n\n In addition, a construction is required to do expansion of secrets\n into blocks of data for the purposes of key generation or validation.\n This pseudorandom function (PRF) takes as input a secret, a seed, and\n an identifying label and produces an output of arbitrary length.\n\n In this section, we define one PRF, based on HMAC. This PRF with the\n SHA-256 hash function is used for all cipher suites defined in this\n document and in TLS documents published prior to this document when\n TLS 1.2 is negotiated. New cipher suites MUST explicitly specify a\n PRF and, in general, SHOULD use the TLS PRF with SHA-256 or a\n stronger standard hash function.\n\n First, we define a data expansion function, P_hash(secret, data),\n that uses a single hash function to expand a secret and seed into an\n arbitrary quantity of output:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 14]\n\f\nRFC 5246 TLS August 2008\n\n\n P_hash(secret, seed) = HMAC_hash(secret, A(1) + seed) +\n HMAC_hash(secret, A(2) + seed) +\n HMAC_hash(secret, A(3) + seed) + ...\n\n where + indicates concatenation.\n\n A() is defined as:\n\n A(0) = seed\n A(i) = HMAC_hash(secret, A(i-1))\n\n P_hash can be iterated as many times as necessary to produce the\n required quantity of data. For example, if P_SHA256 is being used to\n create 80 bytes of data, it will have to be iterated three times\n (through A(3)), creating 96 bytes of output data; the last 16 bytes\n of the final iteration will then be discarded, leaving 80 bytes of\n output data.\n\n TLS's PRF is created by applying P_hash to the secret as:\n\n PRF(secret, label, seed) = P_(secret, label + seed)\n\n The label is an ASCII string. It should be included in the exact\n form it is given without a length byte or trailing null character.\n For example, the label \"slithy toves\" would be processed by hashing\n the following bytes:\n\n 73 6C 69 74 68 79 20 74 6F 76 65 73\n\n6. The TLS Record Protocol\n\n The TLS Record Protocol is a layered protocol. At each layer,\n messages may include fields for length, description, and content.\n The Record Protocol takes messages to be transmitted, fragments the\n data into manageable blocks, optionally compresses the data, applies\n a MAC, encrypts, and transmits the result. Received data is\n decrypted, verified, decompressed, reassembled, and then delivered to\n higher-level clients.\n\n Four protocols that use the record protocol are described in this\n document: the handshake protocol, the alert protocol, the change\n cipher spec protocol, and the application data protocol. In order to\n allow extension of the TLS protocol, additional record content types\n can be supported by the record protocol. New record content type\n values are assigned by IANA in the TLS Content Type Registry as\n described in Section 12.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 15]\n\f\nRFC 5246 TLS August 2008\n\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST send an\n unexpected_message alert.\n\n Any protocol designed for use over TLS must be carefully designed to\n deal with all possible attacks against it. As a practical matter,\n this means that the protocol designer must be aware of what security\n properties TLS does and does not provide and cannot safely rely on\n the latter.\n\n Note in particular that type and length of a record are not protected\n by encryption. If this information is itself sensitive, application\n designers may wish to take steps (padding, cover traffic) to minimize\n information leakage.\n\n6.1. Connection States\n\n\n\n===== DOCUMENT =====\n\n A TLS connection state is the operating environment of the TLS Record\n Protocol. It specifies a compression algorithm, an encryption\n algorithm, and a MAC algorithm. In addition, the parameters for\n these algorithms are known: the MAC key and the bulk encryption keys\n for the connection in both the read and the write directions.\n Logically, there are always four connection states outstanding: the\n current read and write states, and the pending read and write states.\n All records are processed under the current read and write states.\n The security parameters for the pending states can be set by the TLS\n Handshake Protocol, and the ChangeCipherSpec can selectively make\n either of the pending states current, in which case the appropriate\n current state is disposed of and replaced with the pending state; the\n pending state is then reinitialized to an empty state. It is illegal\n to make a state that has not been initialized with security\n parameters a current state. The initial current state always\n specifies that no encryption, compression, or MAC will be used.\n\n The security parameters for a TLS Connection read and write state are\n set by providing the following values:\n\n connection end\n Whether this entity is considered the \"client\" or the \"server\" in\n this connection.\n\n PRF algorithm\n An algorithm used to generate keys from the master secret (see\n Sections 5 and 6.3).\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 16]\n\f\nRFC 5246 TLS August 2008\n\n\n bulk encryption algorithm\n An algorithm to be used for bulk encryption. This specification\n includes the key size of this algorithm, whether it is a block,\n stream, or AEAD cipher, the block size of the cipher (if\n appropriate), and the lengths of explicit and implicit\n initialization vectors (or nonces).\n\n MAC algorithm\n An algorithm to be used for message authentication. This\n specification includes the size of the value returned by the MAC\n algorithm.\n\n compression algorithm\n An algorithm to be used for data compression. This specification\n must include all information the algorithm requires to do\n compression.\n\n master secret\n A 48-byte secret shared between the two peers in the connection.\n\n client random\n A 32-byte value provided by the client.\n\n server random\n A 32-byte value provided by the server.\n\n These parameters are defined in the presentation language as:\n\n enum { server, client } ConnectionEnd;\n\n enum { tls_prf_sha256 } PRFAlgorithm;\n\n enum { null, rc4, 3des, aes }\n BulkCipherAlgorithm;\n\n enum { stream, block, aead } CipherType;\n\n enum { null, hmac_md5, hmac_sha1, hmac_sha256,\n hmac_sha384, hmac_sha512} MACAlgorithm;\n\n enum { null(0), (255) } CompressionMethod;\n\n /* The algorithms specified in CompressionMethod, PRFAlgorithm,\n BulkCipherAlgorithm, and MACAlgorithm may be added to. */\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 17]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ConnectionEnd entity;\n PRFAlgorithm prf_algorithm;\n BulkCipherAlgorithm bulk_cipher_algorithm;\n CipherType cipher_type;\n uint8 enc_key_length;\n uint8 block_length;\n uint8 fixed_iv_length;\n uint8 record_iv_length;\n MACAlgorithm mac_algorithm;\n uint8 mac_length;\n uint8 mac_key_length;\n CompressionMethod compression_algorithm;\n opaque master_secret[48];\n opaque client_random[32];\n opaque server_random[32];\n } SecurityParameters;\n\n The record layer will use the security parameters to generate the\n following six items (some of which are not required by all ciphers,\n and are thus empty):\n\n client write MAC key\n server write MAC key\n client write encryption key\n server write encryption key\n client write IV\n server write IV\n\n The client write parameters are used by the server when receiving and\n processing records and vice versa. The algorithm used for generating\n these items from the security parameters is described in Section 6.3.\n\n Once the security parameters have been set and the keys have been\n generated, the connection states can be instantiated by making them\n the current states. These current states MUST be updated for each\n record processed. Each connection state includes the following\n elements:\n\n compression state\n The current state of the compression algorithm.\n\n cipher state\n The current state of the encryption algorithm. This will consist\n of the scheduled key for that connection. For stream ciphers,\n this will also contain whatever state information is necessary to\n allow the stream to continue to encrypt or decrypt data.\n\n\n\n\nDierks & Rescorla Standards Track [Page 18]\n\f\nRFC 5246 TLS August 2008\n\n\n MAC key\n The MAC key for this connection, as generated above.\n\n sequence number\n Each connection state contains a sequence number, which is\n maintained separately for read and write states. The sequence\n number MUST be set to zero whenever a connection state is made the\n active state. Sequence numbers are of type uint64 and may not\n exceed 2^64-1. Sequence numbers do not wrap. If a TLS\n implementation would need to wrap a sequence number, it must\n renegotiate instead. A sequence number is incremented after each\n record: specifically, the first record transmitted under a\n particular connection state MUST use sequence number 0.\n\n6.2. Record Layer\n\n The TLS record layer receives uninterpreted data from higher layers\n in non-empty blocks of arbitrary size.\n\n6.2.1. Fragmentation\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Client\n message boundaries are not preserved in the record layer (i.e.,\n multiple client messages of the same ContentType MAY be coalesced\n into a single TLSPlaintext record, or a single message MAY be\n fragmented across several records).\n\n struct {\n uint8 major;\n uint8 minor;\n } ProtocolVersion;\n\n enum {\n change_cipher_spec(20), alert(21), handshake(22),\n application_data(23), (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type\n The higher-level protocol used to process the enclosed fragment.\n\n\n\n\nDierks & Rescorla Standards Track [Page 19]\n\f\nRFC 5246 TLS August 2008\n\n\n version\n The version of the protocol being employed. This document\n describes TLS Version 1.2, which uses the version { 3, 3 }. The\n version value 3.3 is historical, deriving from the use of {3, 1}\n for TLS 1.0. (See Appendix A.1.) Note that a client that\n supports multiple versions of TLS may not know what version will\n be employed before it receives the ServerHello. See Appendix E\n for discussion about what record layer version number should be\n employed for ClientHello.\n\n length\n The length (in bytes) of the following TLSPlaintext.fragment. The\n length MUST NOT exceed 2^14.\n\n fragment\n The application data. This data is transparent and treated as an\n independent block to be dealt with by the higher-level protocol\n specified by the type field.\n\n Implementations MUST NOT send zero-length fragments of Handshake,\n Alert, or ChangeCipherSpec content types. Zero-length fragments of\n Application data MAY be sent as they are potentially useful as a\n traffic analysis countermeasure.\n\n Note: Data of different TLS record layer content types MAY be\n interleaved. Application data is generally of lower precedence for\n transmission than other content types. However, records MUST be\n delivered to the network in the same order as they are protected by\n the record layer. Recipients MUST receive and process interleaved\n application layer traffic during handshakes subsequent to the first\n one on a connection.\n\n6.2.2. Record Compression and Decompression\n\n All records are compressed using the compression algorithm defined in\n the current session state. There is always an active compression\n algorithm; however, initially it is defined as\n CompressionMethod.null. The compression algorithm translates a\n TLSPlaintext structure into a TLSCompressed structure. Compression\n functions are initialized with default state information whenever a\n connection state is made active. [RFC3749] describes compression\n algorithms for TLS.\n\n Compression must be lossless and may not increase the content length\n by more than 1024 bytes. If the decompression function encounters a\n TLSCompressed.fragment that would decompress to a length in excess of\n 2^14 bytes, it MUST report a fatal decompression failure error.\n\n\n\n\nDierks & Rescorla Standards Track [Page 20]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ContentType type; /* same as TLSPlaintext.type */\n ProtocolVersion version;/* same as TLSPlaintext.version */\n uint16 length;\n opaque fragment[TLSCompressed.length];\n } TLSCompressed;\n\n length\n The length (in bytes) of the following TLSCompressed.fragment.\n The length MUST NOT exceed 2^14 + 1024.\n\n fragment\n The compressed form of TLSPlaintext.fragment.\n\n Note: A CompressionMethod.null operation is an identity operation;\n no fields are altered.\n\n Implementation note: Decompression functions are responsible for\n ensuring that messages cannot cause internal buffer overflows.\n\n6.2.3. Record Payload Protection\n\n The encryption and MAC functions translate a TLSCompressed\n structure into a TLSCiphertext. The decryption functions reverse\n the process. The MAC of the record also includes a sequence\n number so that missing, extra, or repeated messages are\n detectable.\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n select (SecurityParameters.cipher_type) {\n case stream: GenericStreamCipher;\n case block: GenericBlockCipher;\n case aead: GenericAEADCipher;\n } fragment;\n } TLSCiphertext;\n\n type\n The type field is identical to TLSCompressed.type.\n\n version\n The version field is identical to TLSCompressed.version.\n\n length\n The length (in bytes) of the following TLSCiphertext.fragment.\n The length MUST NOT exceed 2^14 + 2048.\n\n\n\nDierks & Rescorla Standards Track [Page 21]\n\f\nRFC 5246 TLS August 2008\n\n\n fragment\n The encrypted form of TLSCompressed.fragment, with the MAC.\n\n6.2.3.1. Null or Standard Stream Cipher\n\n Stream ciphers (including BulkCipherAlgorithm.null; see Appendix A.6)\n convert TLSCompressed.fragment structures to and from stream\n TLSCiphertext.fragment structures.\n\n stream-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n } GenericStreamCipher;\n\n The MAC is generated as:\n\n MAC(MAC_write_key, seq_num +\n TLSCompressed.type +\n TLSCompressed.version +\n TLSCompressed.length +\n TLSCompressed.fragment);\n\n where \"+\" denotes concatenation.\n\n seq_num\n The sequence number for this record.\n\n MAC\n The MAC algorithm specified by SecurityParameters.mac_algorithm.\n\n Note that the MAC is computed before encryption. The stream cipher\n encrypts the entire block, including the MAC. For stream ciphers\n that do not use a synchronization vector (such as RC4), the stream\n cipher state from the end of one record is simply used on the\n subsequent packet. If the cipher suite is TLS_NULL_WITH_NULL_NULL,\n encryption consists of the identity operation (i.e., the data is not\n encrypted, and the MAC size is zero, implying that no MAC is used).\n For both null and stream ciphers, TLSCiphertext.length is\n TLSCompressed.length plus SecurityParameters.mac_length.\n\n6.2.3.2. CBC Block Cipher\n\n For block ciphers (such as 3DES or AES), the encryption and MAC\n functions convert TLSCompressed.fragment structures to and from block\n TLSCiphertext.fragment structures.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 22]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n opaque IV[SecurityParameters.record_iv_length];\n block-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n uint8 padding[GenericBlockCipher.padding_length];\n uint8 padding_length;\n };\n } GenericBlockCipher;\n\n The MAC is generated as described in Section 6.2.3.1.\n\n IV\n The Initialization Vector (IV) SHOULD be chosen at random, and\n MUST be unpredictable. Note that in versions of TLS prior to 1.1,\n there was no IV field, and the last ciphertext block of the\n previous record (the \"CBC residue\") was used as the IV. This was\n changed to prevent the attacks described in [CBCATT]. For block\n ciphers, the IV length is of length\n SecurityParameters.record_iv_length, which is equal to the\n SecurityParameters.block_size.\n\n padding\n Padding that is added to force the length of the plaintext to be\n an integral multiple of the block cipher's block length. The\n padding MAY be any length up to 255 bytes, as long as it results\n in the TLSCiphertext.length being an integral multiple of the\n block length. Lengths longer than necessary might be desirable to\n frustrate attacks on a protocol that are based on analysis of the\n lengths of exchanged messages. Each uint8 in the padding data\n vector MUST be filled with the padding length value. The receiver\n MUST check this padding and MUST use the bad_record_mac alert to\n indicate padding errors.\n\n padding_length\n The padding length MUST be such that the total size of the\n GenericBlockCipher structure is a multiple of the cipher's block\n length. Legal values range from zero to 255, inclusive. This\n length specifies the length of the padding field exclusive of the\n padding_length field itself.\n\n The encrypted data length (TLSCiphertext.length) is one more than the\n sum of SecurityParameters.block_length, TLSCompressed.length,\n SecurityParameters.mac_length, and padding_length.\n\n Example: If the block length is 8 bytes, the content length\n (TLSCompressed.length) is 61 bytes, and the MAC length is 20 bytes,\n then the length before padding is 82 bytes (this does not include the\n\n\n\nDierks & Rescorla Standards Track [Page 23]\n\f\nRFC 5246 TLS August 2008\n\n\n IV. Thus, the padding length modulo 8 must be equal to 6 in order to\n make the total length an even multiple of 8 bytes (the block length).\n The padding length can be 6, 14, 22, and so on, through 254. If the\n padding length were the minimum necessary, 6, the padding would be 6\n bytes, each containing the value 6. Thus, the last 8 octets of the\n GenericBlockCipher before block encryption would be xx 06 06 06 06 06\n 06 06, where xx is the last octet of the MAC.\n\n Note: With block ciphers in CBC mode (Cipher Block Chaining), it is\n critical that the entire plaintext of the record be known before any\n ciphertext is transmitted. Otherwise, it is possible for the\n attacker to mount the attack described in [CBCATT].\n\n Implementation note: Canvel et al. [CBCTIME] have demonstrated a\n timing attack on CBC padding based on the time required to compute\n the MAC. In order to defend against this attack, implementations\n MUST ensure that record processing time is essentially the same\n whether or not the padding is correct. In general, the best way to\n do this is to compute the MAC even if the padding is incorrect, and\n only then reject the packet. For instance, if the pad appears to be\n incorrect, the implementation might assume a zero-length pad and then\n compute the MAC. This leaves a small timing channel, since MAC\n performance depends to some extent on the size of the data fragment,\n but it is not believed to be large enough to be exploitable, due to\n the large block size of existing MACs and the small size of the\n timing signal.\n\n6.2.3.3. AEAD Ciphers\n\n For AEAD [AEAD] ciphers (such as [CCM] or [GCM]), the AEAD function\n converts TLSCompressed.fragment structures to and from AEAD\n TLSCiphertext.fragment structures.\n\n struct {\n opaque nonce_explicit[SecurityParameters.record_iv_length];\n aead-ciphered struct {\n opaque content[TLSCompressed.length];\n };\n } GenericAEADCipher;\n\n AEAD ciphers take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [AEAD]. The key is either the\n client_write_key or the server_write_key. No MAC key is used.\n\n Each AEAD cipher suite MUST specify how the nonce supplied to the\n AEAD operation is constructed, and what is the length of the\n GenericAEADCipher.nonce_explicit part. In many cases, it is\n\n\n\nDierks & Rescorla Standards Track [Page 24]\n\f\nRFC 5246 TLS August 2008\n\n\n appropriate to use the partially implicit nonce technique described\n in Section 3.2.1 of [AEAD]; with record_iv_length being the length of\n the explicit part. In this case, the implicit part SHOULD be derived\n from key_block as client_write_iv and server_write_iv (as described\n in Section 6.3), and the explicit part is included in\n GenericAEAEDCipher.nonce_explicit.\n\n The plaintext is the TLSCompressed.fragment.\n\n The additional authenticated data, which we denote as\n additional_data, is defined as follows:\n\n additional_data = seq_num + TLSCompressed.type +\n TLSCompressed.version + TLSCompressed.length;\n\n where \"+\" denotes concatenation.\n\n The aead_output consists of the ciphertext output by the AEAD\n encryption operation. The length will generally be larger than\n TLSCompressed.length, but by an amount that varies with the AEAD\n cipher. Since the ciphers might incorporate padding, the amount of\n overhead could vary with different TLSCompressed.length values. Each\n AEAD cipher MUST NOT produce an expansion of greater than 1024 bytes.\n Symbolically,\n\n AEADEncrypted = AEAD-Encrypt(write_key, nonce, plaintext,\n additional_data)\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, the \"additional_data\", and the AEADEncrypted value. The\n output is either the plaintext or an error indicating that the\n decryption failed. There is no separate integrity check. That is:\n\n TLSCompressed.fragment = AEAD-Decrypt(write_key, nonce,\n AEADEncrypted,\n additional_data)\n\n If the decryption fails, a fatal bad_record_mac alert MUST be\n generated.\n\n6.3. Key Calculation\n\n The Record Protocol requires an algorithm to generate keys required\n by the current connection state (see Appendix A.6) from the security\n parameters provided by the handshake protocol.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 25]\n\f\nRFC 5246 TLS August 2008\n\n\n The master secret is expanded into a sequence of secure bytes, which\n is then split to a client write MAC key, a server write MAC key, a\n client write encryption key, and a server write encryption key. Each\n of these is generated from the byte sequence in that order. Unused\n values are empty. Some AEAD ciphers may additionally require a\n client write IV and a server write IV (see Section 6.2.3.3).\n\n When keys and MAC keys are generated, the master secret is used as an\n entropy source.\n\n To generate the key material, compute\n\n key_block = PRF(SecurityParameters.master_secret,\n \"key expansion\",\n SecurityParameters.server_random +\n SecurityParameters.client_random);\n\n until enough output has been generated. Then, the key_block is\n partitioned as follows:\n\n client_write_MAC_key[SecurityParameters.mac_key_length]\n server_write_MAC_key[SecurityParameters.mac_key_length]\n client_write_key[SecurityParameters.enc_key_length]\n server_write_key[SecurityParameters.enc_key_length]\n client_write_IV[SecurityParameters.fixed_iv_length]\n server_write_IV[SecurityParameters.fixed_iv_length]\n\n Currently, the client_write_IV and server_write_IV are only generated\n for implicit nonce techniques as described in Section 3.2.1 of\n [AEAD].\n\n Implementation note: The currently defined cipher suite which\n requires the most material is AES_256_CBC_SHA256. It requires 2 x 32\n byte keys and 2 x 32 byte MAC keys, for a total 128 bytes of key\n material.\n\n7. The TLS Handshaking Protocols\n\n TLS has three subprotocols that are used to allow peers to agree upon\n security parameters for the record layer, to authenticate themselves,\n to instantiate negotiated security parameters, and to report error\n conditions to each other.\n\n The Handshake Protocol is responsible for negotiating a session,\n which consists of the following items:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 26]\n\f\nRFC 5246 TLS August 2008\n\n\n session identifier\n An arbitrary byte sequence chosen by the server to identify an\n active or resumable session state.\n\n peer certificate\n X509v3 [PKIX] certificate of the peer. This element of the state\n may be null.\n\n compression method\n The algorithm used to compress data prior to encryption.\n\n cipher spec\n Specifies the pseudorandom function (PRF) used to generate keying\n material, the bulk data encryption algorithm (such as null, AES,\n etc.) and the MAC algorithm (such as HMAC-SHA1). It also defines\n cryptographic attributes such as the mac_length. (See Appendix\n A.6 for formal definition.)\n\n master secret\n 48-byte secret shared between the client and server.\n\n is resumable\n A flag indicating whether the session can be used to initiate new\n connections.\n\n These items are then used to create security parameters for use by\n the record layer when protecting application data. Many connections\n can be instantiated using the same session through the resumption\n feature of the TLS Handshake Protocol.\n\n7.1. Change Cipher Spec Protocol\n\n The change cipher spec protocol exists to signal transitions in\n ciphering strategies. The protocol consists of a single message,\n which is encrypted and compressed under the current (not the pending)\n connection state. The message consists of a single byte of value 1.\n\n struct {\n enum { change_cipher_spec(1), (255) } type;\n } ChangeCipherSpec;\n\n The ChangeCipherSpec message is sent by both the client and the\n server to notify the receiving party that subsequent records will be\n protected under the newly negotiated CipherSpec and keys. Reception\n of this message causes the receiver to instruct the record layer to\n immediately copy the read pending state into the read current state.\n Immediately after sending this message, the sender MUST instruct the\n record layer to make the write pending state the write active state.\n\n\n\nDierks & Rescorla Standards Track [Page 27]\n\f\nRFC 5246 TLS August 2008\n\n\n (See Section 6.1.) The ChangeCipherSpec message is sent during the\n handshake after the security parameters have been agreed upon, but\n before the verifying Finished message is sent.\n\n Note: If a rehandshake occurs while data is flowing on a connection,\n the communicating parties may continue to send data using the old\n CipherSpec. However, once the ChangeCipherSpec has been sent, the\n new CipherSpec MUST be used. The first side to send the\n ChangeCipherSpec does not know that the other side has finished\n computing the new keying material (e.g., if it has to perform a\n time-consuming public key operation). Thus, a small window of time,\n during which the recipient must buffer the data, MAY exist. In\n practice, with modern machines this interval is likely to be fairly\n short.\n\n7.2. Alert Protocol\n\n One of the content types supported by the TLS record layer is the\n alert type. Alert messages convey the severity of the message\n (warning or fatal) and a description of the alert. Alert messages\n with a level of fatal result in the immediate termination of the\n connection. In this case, other connections corresponding to the\n session may continue, but the session identifier MUST be invalidated,\n preventing the failed session from being used to establish new\n connections. Like other messages, alert messages are encrypted and\n compressed, as specified by the current connection state.\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n\n\n\nDierks & Rescorla Standards Track [Page 28]\n\f\nRFC 5246 TLS August 2008\n\n\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n7.2.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack. Either party may\n initiate the exchange of closing messages.\n\n close_notify\n This message notifies the recipient that the sender will not send\n any more messages on this connection. Note that as of TLS 1.1,\n failure to properly close a connection no longer requires that a\n session not be resumed. This is a change from TLS 1.0 to conform\n with widespread implementation practice.\n\n Either party may initiate a close by sending a close_notify alert.\n Any data received after a closure alert is ignored.\n\n Unless some other fatal alert has been transmitted, each party is\n required to send a close_notify alert before closing the write side\n of the connection. The other party MUST respond with a close_notify\n alert of its own and close down the connection immediately,\n discarding any pending writes. It is not required for the initiator\n of the close to wait for the responding close_notify alert before\n closing the read side of the connection.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation must receive the responding\n close_notify alert before indicating to the application layer that\n the TLS connection has ended. If the application protocol will not\n transfer any additional data, but will only close the underlying\n transport connection, then the implementation MAY choose to close the\n transport without waiting for the responding close_notify. No part\n\n\n\n\nDierks & Rescorla Standards Track [Page 29]\n\f\nRFC 5246 TLS August 2008\n\n\n of this standard should be taken to dictate the manner in which a\n usage profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing a connection reliably delivers\n pending data before destroying the transport.\n\n7.2.2. Error Alerts\n\n Error handling in the TLS Handshake protocol is very simple. When an\n error is detected, the detecting party sends a message to the other\n party. Upon transmission or receipt of a fatal alert message, both\n parties immediately close the connection. Servers and clients MUST\n forget any session-identifiers, keys, and secrets associated with a\n failed connection. Thus, any connection terminated with a fatal\n alert MUST NOT be resumed.\n\n Whenever an implementation encounters a condition which is defined as\n a fatal alert, it MUST send the appropriate alert prior to closing\n the connection. For all errors where an alert level is not\n explicitly specified, the sending party MAY determine at its\n discretion whether to treat this as a fatal error or not. If the\n implementation chooses to send an alert but intends to close the\n connection immediately afterwards, it MUST send that alert at the\n fatal alert level.\n\n If an alert with a level of warning is sent and received, generally\n the connection can continue normally. If the receiving party decides\n not to proceed with the connection (e.g., after having received a\n no_renegotiation alert that it is not willing to accept), it SHOULD\n send a fatal alert to terminate the connection. Given this, the\n sending party cannot, in general, know how the receiving party will\n behave. Therefore, warning alerts are not very useful when the\n sending party wants to continue the connection, and thus are\n sometimes omitted. For example, if a peer decides to accept an\n expired certificate (perhaps after confirming this with the user) and\n wants to continue the connection, it would not generally send a\n certificate_expired alert.\n\n The following error alerts are defined:\n\n unexpected_message\n An inappropriate message was received. This alert is always fatal\n and should never be observed in communication between proper\n implementations.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 30]\n\f\nRFC 5246 TLS August 2008\n\n\n bad_record_mac\n This alert is returned if a record is received with an incorrect\n MAC. This alert also MUST be returned if an alert is sent because\n a TLSCiphertext decrypted in an invalid way: either it wasn't an\n even multiple of the block length, or its padding values, when\n checked, weren't correct. This message is always fatal and should\n never be observed in communication between proper implementations\n (except when messages were corrupted in the network).\n\n decryption_failed_RESERVED\n This alert was used in some earlier versions of TLS, and may have\n permitted certain attacks against the CBC mode [CBCATT]. It MUST\n NOT be sent by compliant implementations.\n\n record_overflow\n A TLSCiphertext record was received that had a length more than\n 2^14+2048 bytes, or a record decrypted to a TLSCompressed record\n with more than 2^14+1024 bytes. This message is always fatal and\n should never be observed in communication between proper\n implementations (except when messages were corrupted in the\n network).\n\n decompression_failure\n The decompression function received improper input (e.g., data\n that would expand to excessive length). This message is always\n fatal and should never be observed in communication between proper\n implementations.\n\n handshake_failure\n Reception of a handshake_failure alert message indicates that the\n sender was unable to negotiate an acceptable set of security\n parameters given the options available. This is a fatal error.\n\n no_certificate_RESERVED\n This alert was used in SSLv3 but not any version of TLS. It MUST\n NOT be sent by compliant implementations.\n\n bad_certificate\n A certificate was corrupt, contained signatures that did not\n verify correctly, etc.\n\n unsupported_certificate\n A certificate was of an unsupported type.\n\n certificate_revoked\n A certificate was revoked by its signer.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 31]\n\f\nRFC 5246 TLS August 2008\n\n\n certificate_expired\n A certificate has expired or is not currently valid.\n\n certificate_unknown\n Some other (unspecified) issue arose in processing the\n certificate, rendering it unacceptable.\n\n illegal_parameter\n A field in the handshake was out of range or inconsistent with\n other fields. This message is always fatal.\n\n unknown_ca\n A valid certificate chain or partial chain was received, but the\n certificate was not accepted because the CA certificate could not\n be located or couldn't be matched with a known, trusted CA. This\n message is always fatal.\n\n access_denied\n A valid certificate was received, but when access control was\n applied, the sender decided not to proceed with negotiation. This\n message is always fatal.\n\n decode_error\n A message could not be decoded because some field was out of the\n specified range or the length of the message was incorrect. This\n message is always fatal and should never be observed in\n communication between proper implementations (except when messages\n were corrupted in the network).\n\n decrypt_error\n A handshake cryptographic operation failed, including being unable\n to correctly verify a signature or validate a Finished message.\n This message is always fatal.\n\n export_restriction_RESERVED\n This alert was used in some earlier versions of TLS. It MUST NOT\n be sent by compliant implementations.\n\n protocol_version\n The protocol version the client has attempted to negotiate is\n recognized but not supported. (For example, old protocol versions\n might be avoided for security reasons.) This message is always\n fatal.\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 32]\n\f\nRFC 5246 TLS August 2008\n\n\n insufficient_security\n Returned instead of handshake_failure when a negotiation has\n failed specifically because the server requires ciphers more\n secure than those supported by the client. This message is always\n fatal.\n\n internal_error\n An internal error unrelated to the peer or the correctness of the\n protocol (such as a memory allocation failure) makes it impossible\n to continue. This message is always fatal.\n\n\n\n===== DOCUMENT =====\n\n user_canceled\n This handshake is being canceled for some reason unrelated to a\n protocol failure. If the user cancels an operation after the\n handshake is complete, just closing the connection by sending a\n close_notify is more appropriate. This alert should be followed\n by a close_notify. This message is generally a warning.\n\n no_renegotiation\n Sent by the client in response to a hello request or by the server\n in response to a client hello after initial handshaking. Either\n of these would normally lead to renegotiation; when that is not\n appropriate, the recipient should respond with this alert. At\n that point, the original requester can decide whether to proceed\n with the connection. One case where this would be appropriate is\n where a server has spawned a process to satisfy a request; the\n process might receive security parameters (key length,\n authentication, etc.) at startup, and it might be difficult to\n communicate changes to these parameters after that point. This\n message is always a warning.\n\n unsupported_extension\n sent by clients that receive an extended server hello containing\n an extension that they did not put in the corresponding client\n hello. This message is always fatal.\n\n New Alert values are assigned by IANA as described in Section 12.\n\n7.3. Handshake Protocol Overview\n\n The cryptographic parameters of the session state are produced by the\n TLS Handshake Protocol, which operates on top of the TLS record\n layer. When a TLS client and server first start communicating, they\n agree on a protocol version, select cryptographic algorithms,\n optionally authenticate each other, and use public-key encryption\n techniques to generate shared secrets.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 33]\n\f\nRFC 5246 TLS August 2008\n\n\n The TLS Handshake Protocol involves the following steps:\n\n - Exchange hello messages to agree on algorithms, exchange random\n values, and check for session resumption.\n\n - Exchange the necessary cryptographic parameters to allow the\n client and server to agree on a premaster secret.\n\n - Exchange certificates and cryptographic information to allow the\n client and server to authenticate themselves.\n\n - Generate a master secret from the premaster secret and exchanged\n random values.\n\n - Provide security parameters to the record layer.\n\n - Allow the client and server to verify that their peer has\n calculated the same security parameters and that the handshake\n occurred without tampering by an attacker.\n\n Note that higher layers should not be overly reliant on whether TLS\n always negotiates the strongest possible connection between two\n peers. There are a number of ways in which a man-in-the-middle\n attacker can attempt to make two entities drop down to the least\n secure method they support. The protocol has been designed to\n minimize this risk, but there are still attacks available: for\n example, an attacker could block access to the port a secure service\n runs on, or attempt to get the peers to negotiate an unauthenticated\n connection. The fundamental rule is that higher levels must be\n cognizant of what their security requirements are and never transmit\n information over a channel less secure than what they require. The\n TLS protocol is secure in that any cipher suite offers its promised\n level of security: if you negotiate 3DES with a 1024-bit RSA key\n exchange with a host whose certificate you have verified, you can\n expect to be that secure.\n\n These goals are achieved by the handshake protocol, which can be\n summarized as follows: The client sends a ClientHello message to\n which the server must respond with a ServerHello message, or else a\n fatal error will occur and the connection will fail. The ClientHello\n and ServerHello are used to establish security enhancement\n capabilities between client and server. The ClientHello and\n ServerHello establish the following attributes: Protocol Version,\n Session ID, Cipher Suite, and Compression Method. Additionally, two\n random values are generated and exchanged: ClientHello.random and\n ServerHello.random.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 34]\n\f\nRFC 5246 TLS August 2008\n\n\n The actual key exchange uses up to four messages: the server\n Certificate, the ServerKeyExchange, the client Certificate, and the\n ClientKeyExchange. New key exchange methods can be created by\n specifying a format for these messages and by defining the use of the\n messages to allow the client and server to agree upon a shared\n secret. This secret MUST be quite long; currently defined key\n exchange methods exchange secrets that range from 46 bytes upwards.\n\n Following the hello messages, the server will send its certificate in\n a Certificate message if it is to be authenticated. Additionally, a\n ServerKeyExchange message may be sent, if it is required (e.g., if\n the server has no certificate, or if its certificate is for signing\n only). If the server is authenticated, it may request a certificate\n from the client, if that is appropriate to the cipher suite selected.\n Next, the server will send the ServerHelloDone message, indicating\n that the hello-message phase of the handshake is complete. The\n server will then wait for a client response. If the server has sent\n a CertificateRequest message, the client MUST send the Certificate\n message. The ClientKeyExchange message is now sent, and the content\n of that message will depend on the public key algorithm selected\n between the ClientHello and the ServerHello. If the client has sent\n a certificate with signing ability, a digitally-signed\n CertificateVerify message is sent to explicitly verify possession of\n the private key in the certificate.\n\n At this point, a ChangeCipherSpec message is sent by the client, and\n the client copies the pending Cipher Spec into the current Cipher\n Spec. The client then immediately sends the Finished message under\n the new algorithms, keys, and secrets. In response, the server will\n send its own ChangeCipherSpec message, transfer the pending to the\n current Cipher Spec, and send its Finished message under the new\n Cipher Spec. At this point, the handshake is complete, and the\n client and server may begin to exchange application layer data. (See\n flow chart below.) Application data MUST NOT be sent prior to the\n completion of the first handshake (before a cipher suite other than\n TLS_NULL_WITH_NULL_NULL is established).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 35]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1. Message flow for a full handshake\n\n * Indicates optional or situation-dependent messages that are not\n always sent.\n\n Note: To help avoid pipeline stalls, ChangeCipherSpec is an\n independent TLS protocol content type, and is not actually a TLS\n handshake message.\n\n When the client and server decide to resume a previous session or\n duplicate an existing session (instead of negotiating new security\n parameters), the message flow is as follows:\n\n The client sends a ClientHello using the Session ID of the session to\n be resumed. The server then checks its session cache for a match.\n If a match is found, and the server is willing to re-establish the\n connection under the specified session state, it will send a\n ServerHello with the same Session ID value. At this point, both\n client and server MUST send ChangeCipherSpec messages and proceed\n directly to Finished messages. Once the re-establishment is\n complete, the client and server MAY begin to exchange application\n layer data. (See flow chart below.) If a Session ID match is not\n found, the server generates a new session ID, and the TLS client and\n server perform a full handshake.\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 36]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2. Message flow for an abbreviated handshake\n\n The contents and significance of each message will be presented in\n detail in the following sections.\n\n7.4. Handshake Protocol\n\n The TLS Handshake Protocol is one of the defined higher-level clients\n of the TLS Record Protocol. This protocol is used to negotiate the\n secure attributes of a session. Handshake messages are supplied to\n the TLS record layer, where they are encapsulated within one or more\n TLSPlaintext structures, which are processed and transmitted as\n specified by the current active session state.\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n } body;\n } Handshake;\n\n\n\nDierks & Rescorla Standards Track [Page 37]\n\f\nRFC 5246 TLS August 2008\n\n\n The handshake protocol messages are presented below in the order they\n MUST be sent; sending handshake messages in an unexpected order\n results in a fatal error. Unneeded handshake messages can be\n omitted, however. Note one exception to the ordering: the\n Certificate message is used twice in the handshake (from server to\n client, then from client to server), but described only in its first\n position. The one message that is not bound by these ordering rules\n is the HelloRequest message, which can be sent at any time, but which\n SHOULD be ignored by the client if it arrives in the middle of a\n handshake.\n\n New handshake message types are assigned by IANA as described in\n Section 12.\n\n7.4.1. Hello Messages\n\n The hello phase messages are used to exchange security enhancement\n capabilities between the client and server. When a new session\n begins, the record layer's connection state encryption, hash, and\n compression algorithms are initialized to null. The current\n connection state is used for renegotiation messages.\n\n7.4.1.1. Hello Request\n\n When this message will be sent:\n\n The HelloRequest message MAY be sent by the server at any time.\n\n Meaning of this message:\n\n HelloRequest is a simple notification that the client should begin\n the negotiation process anew. In response, the client should send\n a ClientHello message when convenient. This message is not\n intended to establish which side is the client or server but\n merely to initiate a new negotiation. Servers SHOULD NOT send a\n HelloRequest immediately upon the client's initial connection. It\n is the client's job to send a ClientHello at that time.\n\n This message will be ignored by the client if the client is\n currently negotiating a session. This message MAY be ignored by\n the client if it does not wish to renegotiate a session, or the\n client may, if it wishes, respond with a no_renegotiation alert.\n Since handshake messages are intended to have transmission\n precedence over application data, it is expected that the\n negotiation will begin before no more than a few records are\n received from the client. If the server sends a HelloRequest but\n does not receive a ClientHello in response, it may close the\n connection with a fatal alert.\n\n\n\nDierks & Rescorla Standards Track [Page 38]\n\f\nRFC 5246 TLS August 2008\n\n\n After sending a HelloRequest, servers SHOULD NOT repeat the\n request until the subsequent handshake negotiation is complete.\n\n Structure of this message:\n\n struct { } HelloRequest;\n\n This message MUST NOT be included in the message hashes that are\n maintained throughout the handshake and used in the Finished messages\n and the certificate verify message.\n\n7.4.1.2. Client Hello\n\n When this message will be sent:\n\n When a client first connects to a server, it is required to send\n the ClientHello as its first message. The client can also send a\n ClientHello in response to a HelloRequest or on its own initiative\n in order to renegotiate the security parameters in an existing\n connection.\n\n Structure of this message:\n\n The ClientHello message includes a random structure, which is used\n later in the protocol.\n\n struct {\n uint32 gmt_unix_time;\n opaque random_bytes[28];\n } Random;\n\n gmt_unix_time\n The current time and date in standard UNIX 32-bit format\n (seconds since the midnight starting Jan 1, 1970, UTC, ignoring\n leap seconds) according to the sender's internal clock. Clocks\n are not required to be set correctly by the basic TLS protocol;\n higher-level or application protocols may define additional\n requirements. Note that, for historical reasons, the data\n element is named using GMT, the predecessor of the current\n worldwide time base, UTC.\n\n random_bytes\n 28 bytes generated by a secure random number generator.\n\n The ClientHello message includes a variable-length session\n identifier. If not empty, the value identifies a session between the\n same client and server whose security parameters the client wishes to\n reuse. The session identifier MAY be from an earlier connection,\n\n\n\nDierks & Rescorla Standards Track [Page 39]\n\f\nRFC 5246 TLS August 2008\n\n\n this connection, or from another currently active connection. The\n second option is useful if the client only wishes to update the\n random structures and derived values of a connection, and the third\n option makes it possible to establish several independent secure\n connections without repeating the full handshake protocol. These\n independent connections may occur sequentially or simultaneously; a\n SessionID becomes valid when the handshake negotiating it completes\n with the exchange of Finished messages and persists until it is\n removed due to aging or because a fatal error was encountered on a\n connection associated with the session. The actual contents of the\n SessionID are defined by the server.\n\n opaque SessionID<0..32>;\n\n Warning: Because the SessionID is transmitted without encryption or\n immediate MAC protection, servers MUST NOT place confidential\n information in session identifiers or let the contents of fake\n session identifiers cause any breach of security. (Note that the\n content of the handshake as a whole, including the SessionID, is\n protected by the Finished messages exchanged at the end of the\n handshake.)\n\n The cipher suite list, passed from the client to the server in the\n ClientHello message, contains the combinations of cryptographic\n algorithms supported by the client in order of the client's\n preference (favorite choice first). Each cipher suite defines a key\n exchange algorithm, a bulk encryption algorithm (including secret key\n length), a MAC algorithm, and a PRF. The server will select a cipher\n suite or, if no acceptable choices are presented, return a handshake\n failure alert and close the connection. If the list contains cipher\n suites the server does not recognize, support, or wish to use, the\n server MUST ignore those cipher suites, and process the remaining\n ones as usual.\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n The ClientHello includes a list of compression algorithms supported\n by the client, ordered according to the client's preference.\n\n enum { null(0), (255) } CompressionMethod;\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 40]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ProtocolVersion client_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suites<2..2^16-2>;\n CompressionMethod compression_methods<1..2^8-1>;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ClientHello;\n\n TLS allows extensions to follow the compression_methods field in an\n extensions block. The presence of extensions can be detected by\n determining whether there are bytes following the compression_methods\n at the end of the ClientHello. Note that this method of detecting\n optional data differs from the normal TLS method of having a\n variable-length field, but it is used for compatibility with TLS\n before extensions were defined.\n\n client_version\n The version of the TLS protocol by which the client wishes to\n communicate during this session. This SHOULD be the latest\n (highest valued) version supported by the client. For this\n version of the specification, the version will be 3.3 (see\n Appendix E for details about backward compatibility).\n\n random\n A client-generated random structure.\n\n session_id\n The ID of a session the client wishes to use for this connection.\n This field is empty if no session_id is available, or if the\n client wishes to generate new security parameters.\n\n cipher_suites\n This is a list of the cryptographic options supported by the\n client, with the client's first preference first. If the\n session_id field is not empty (implying a session resumption\n request), this vector MUST include at least the cipher_suite from\n that session. Values are defined in Appendix A.5.\n\n compression_methods\n This is a list of the compression methods supported by the client,\n sorted by client preference. If the session_id field is not empty\n (implying a session resumption request), it MUST include the\n\n\n\nDierks & Rescorla Standards Track [Page 41]\n\f\nRFC 5246 TLS August 2008\n\n\n compression_method from that session. This vector MUST contain,\n and all implementations MUST support, CompressionMethod.null.\n Thus, a client and server will always be able to agree on a\n compression method.\n\n extensions\n Clients MAY request extended functionality from servers by sending\n data in the extensions field. The actual \"Extension\" format is\n defined in Section 7.4.1.4.\n\n In the event that a client requests additional functionality using\n extensions, and this functionality is not supplied by the server, the\n client MAY abort the handshake. A server MUST accept ClientHello\n messages both with and without the extensions field, and (as for all\n other messages) it MUST check that the amount of data in the message\n precisely matches one of these formats; if not, then it MUST send a\n fatal \"decode_error\" alert.\n\n After sending the ClientHello message, the client waits for a\n ServerHello message. Any handshake message returned by the server,\n except for a HelloRequest, is treated as a fatal error.\n\n7.4.1.3. Server Hello\n\n When this message will be sent:\n\n The server will send this message in response to a ClientHello\n message when it was able to find an acceptable set of algorithms.\n If it cannot find such a match, it will respond with a handshake\n failure alert.\n\n Structure of this message:\n\n struct {\n ProtocolVersion server_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ServerHello;\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 42]\n\f\nRFC 5246 TLS August 2008\n\n\n The presence of extensions can be detected by determining whether\n there are bytes following the compression_method field at the end of\n the ServerHello.\n\n server_version\n This field will contain the lower of that suggested by the client\n in the client hello and the highest supported by the server. For\n this version of the specification, the version is 3.3. (See\n Appendix E for details about backward compatibility.)\n\n random\n This structure is generated by the server and MUST be\n independently generated from the ClientHello.random.\n\n session_id\n This is the identity of the session corresponding to this\n connection. If the ClientHello.session_id was non-empty, the\n server will look in its session cache for a match. If a match is\n found and the server is willing to establish the new connection\n using the specified session state, the server will respond with\n the same value as was supplied by the client. This indicates a\n resumed session and dictates that the parties must proceed\n directly to the Finished messages. Otherwise, this field will\n contain a different value identifying the new session. The server\n may return an empty session_id to indicate that the session will\n not be cached and therefore cannot be resumed. If a session is\n resumed, it must be resumed using the same cipher suite it was\n originally negotiated with. Note that there is no requirement\n that the server resume any session even if it had formerly\n provided a session_id. Clients MUST be prepared to do a full\n negotiation -- including negotiating new cipher suites -- during\n any handshake.\n\n cipher_suite\n The single cipher suite selected by the server from the list in\n ClientHello.cipher_suites. For resumed sessions, this field is\n the value from the state of the session being resumed.\n\n compression_method\n The single compression algorithm selected by the server from the\n list in ClientHello.compression_methods. For resumed sessions,\n this field is the value from the resumed session state.\n\n extensions\n A list of extensions. Note that only extensions offered by the\n client can appear in the server's list.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 43]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.1.4. Hello Extensions\n\n The extension format is:\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n signature_algorithms(13), (65535)\n } ExtensionType;\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The initial set of extensions is defined in a companion document\n [TLSEXT]. The list of extension types is maintained by IANA as\n described in Section 12.\n\n An extension type MUST NOT appear in the ServerHello unless the same\n extension type appeared in the corresponding ClientHello. If a\n client receives an extension type in ServerHello that it did not\n request in the associated ClientHello, it MUST abort the handshake\n with an unsupported_extension fatal alert.\n\n Nonetheless, \"server-oriented\" extensions may be provided in the\n future within this framework. Such an extension (say, of type x)\n would require the client to first send an extension of type x in a\n ClientHello with empty extension_data to indicate that it supports\n the extension type. In this case, the client is offering the\n capability to understand the extension type, and the server is taking\n the client up on its offer.\n\n When multiple extensions of different types are present in the\n ClientHello or ServerHello messages, the extensions MAY appear in any\n order. There MUST NOT be more than one extension of the same type.\n\n Finally, note that extensions can be sent both when starting a new\n session and when requesting session resumption. Indeed, a client\n that requests session resumption does not in general know whether the\n server will accept this request, and therefore it SHOULD send the\n same extensions as it would send if it were not attempting\n resumption.\n\n\n\nDierks & Rescorla Standards Track [Page 44]\n\f\nRFC 5246 TLS August 2008\n\n\n In general, the specification of each extension type needs to\n describe the effect of the extension both during full handshake and\n session resumption. Most current TLS extensions are relevant only\n when a session is initiated: when an older session is resumed, the\n server does not process these extensions in Client Hello, and does\n not include them in Server Hello. However, some extensions may\n specify different behavior during session resumption.\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions, and some are simply refusals to support particular\n features. In general, error alerts should be used for the former,\n and a field in the server extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem.\n\n Often the fact that the extension fields are included in the\n inputs to the Finished message hashes will be sufficient, but\n extreme care is needed when the extension changes the meaning of\n messages sent in the handshake phase. Designers and implementors\n should be aware of the fact that until the handshake has been\n authenticated, active attackers can modify messages and insert,\n remove, or replace extensions.\n\n - It would be technically possible to use extensions to change major\n aspects of the design of TLS; for example the design of cipher\n suite negotiation. This is not recommended; it would be more\n appropriate to define a new version of TLS -- particularly since\n the TLS handshake algorithms have specific protection against\n version rollback attacks based on the version number, and the\n possibility of version rollback should be a significant\n consideration in any major design change.\n\n7.4.1.4.1. Signature Algorithms\n\n The client uses the \"signature_algorithms\" extension to indicate to\n the server which signature/hash algorithm pairs may be used in\n digital signatures. The \"extension_data\" field of this extension\n contains a \"supported_signature_algorithms\" value.\n\n\n\nDierks & Rescorla Standards Track [Page 45]\n\f\nRFC 5246 TLS August 2008\n\n\n enum {\n none(0), md5(1), sha1(2), sha224(3), sha256(4), sha384(5),\n sha512(6), (255)\n } HashAlgorithm;\n\n enum { anonymous(0), rsa(1), dsa(2), ecdsa(3), (255) }\n SignatureAlgorithm;\n\n struct {\n HashAlgorithm hash;\n SignatureAlgorithm signature;\n } SignatureAndHashAlgorithm;\n\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2..2^16-2>;\n\n Each SignatureAndHashAlgorithm value lists a single hash/signature\n pair that the client is willing to verify. The values are indicated\n in descending order of preference.\n\n Note: Because not all signature algorithms and hash algorithms may be\n accepted by an implementation (e.g., DSA with SHA-1, but not\n SHA-256), algorithms here are listed in pairs.\n\n hash\n This field indicates the hash algorithm which may be used. The\n values indicate support for unhashed data, MD5 [MD5], SHA-1,\n SHA-224, SHA-256, SHA-384, and SHA-512 [SHS], respectively. The\n \"none\" value is provided for future extensibility, in case of a\n signature algorithm which does not require hashing before signing.\n\n signature\n This field indicates the signature algorithm that may be used.\n The values indicate anonymous signatures, RSASSA-PKCS1-v1_5\n [PKCS1] and DSA [DSS], and ECDSA [ECDSA], respectively. The\n \"anonymous\" value is meaningless in this context but used in\n Section 7.4.3. It MUST NOT appear in this extension.\n\n The semantics of this extension are somewhat complicated because the\n cipher suite indicates permissible signature algorithms but not hash\n algorithms. Sections 7.4.2 and 7.4.3 describe the appropriate rules.\n\n If the client supports only the default hash and signature algorithms\n (listed in this section), it MAY omit the signature_algorithms\n extension. If the client does not support the default algorithms, or\n supports other hash and signature algorithms (and it is willing to\n use them for verifying messages sent by the server, i.e., server\n certificates and server key exchange), it MUST send the\n\n\n\nDierks & Rescorla Standards Track [Page 46]\n\f\nRFC 5246 TLS August 2008\n\n\n signature_algorithms extension, listing the algorithms it is willing\n to accept.\n\n If the client does not send the signature_algorithms extension, the\n server MUST do the following:\n\n - If the negotiated key exchange algorithm is one of (RSA, DHE_RSA,\n DH_RSA, RSA_PSK, ECDH_RSA, ECDHE_RSA), behave as if client had\n sent the value {sha1,rsa}.\n\n - If the negotiated key exchange algorithm is one of (DHE_DSS,\n DH_DSS), behave as if the client had sent the value {sha1,dsa}.\n\n - If the negotiated key exchange algorithm is one of (ECDH_ECDSA,\n ECDHE_ECDSA), behave as if the client had sent value {sha1,ecdsa}.\n\n Note: this is a change from TLS 1.1 where there are no explicit\n rules, but as a practical matter one can assume that the peer\n supports MD5 and SHA-1.\n\n Note: this extension is not meaningful for TLS versions prior to 1.2.\n Clients MUST NOT offer it if they are offering prior versions.\n However, even if clients do offer it, the rules specified in [TLSEXT]\n require servers to ignore extensions they do not understand.\n\n Servers MUST NOT send this extension. TLS servers MUST support\n receiving this extension.\n\n When performing session resumption, this extension is not included in\n Server Hello, and the server ignores the extension in Client Hello\n (if present).\n\n7.4.2. Server Certificate\n\n When this message will be sent:\n\n The server MUST send a Certificate message whenever the agreed-\n upon key exchange method uses certificates for authentication\n (this includes all key exchange methods defined in this document\n except DH_anon). This message will always immediately follow the\n ServerHello message.\n\n Meaning of this message:\n\n This message conveys the server's certificate chain to the client.\n\n The certificate MUST be appropriate for the negotiated cipher\n suite's key exchange algorithm and any negotiated extensions.\n\n\n\nDierks & Rescorla Standards Track [Page 47]\n\f\nRFC 5246 TLS August 2008\n\n\n Structure of this message:\n\n opaque ASN.1Cert<1..2^24-1>;\n\n struct {\n ASN.1Cert certificate_list<0..2^24-1>;\n } Certificate;\n\n certificate_list\n This is a sequence (chain) of certificates. The sender's\n certificate MUST come first in the list. Each following\n certificate MUST directly certify the one preceding it. Because\n certificate validation requires that root keys be distributed\n independently, the self-signed certificate that specifies the root\n certificate authority MAY be omitted from the chain, under the\n assumption that the remote end must already possess it in order to\n validate it in any case.\n\n The same message type and structure will be used for the client's\n response to a certificate request message. Note that a client MAY\n send no certificates if it does not have an appropriate certificate\n to send in response to the server's authentication request.\n\n Note: PKCS #7 [PKCS7] is not used as the format for the certificate\n vector because PKCS #6 [PKCS6] extended certificates are not used.\n Also, PKCS #7 defines a SET rather than a SEQUENCE, making the task\n of parsing the list more difficult.\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3, unless explicitly negotiated\n otherwise (e.g., [TLSPGP]).\n\n - The end entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected key exchange\n algorithm.\n\n Key Exchange Alg. Certificate Key Type\n\n RSA RSA public key; the certificate MUST allow the\n RSA_PSK key to be used for encryption (the\n keyEncipherment bit MUST be set if the key\n usage extension is present).\n Note: RSA_PSK is defined in [TLSPSK].\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 48]\n\f\nRFC 5246 TLS August 2008\n\n\n DHE_RSA RSA public key; the certificate MUST allow the\n ECDHE_RSA key to be used for signing (the\n digitalSignature bit MUST be set if the key\n usage extension is present) with the signature\n scheme and hash algorithm that will be employed\n in the server key exchange message.\n Note: ECDHE_RSA is defined in [TLSECC].\n\n DHE_DSS DSA public key; the certificate MUST allow the\n key to be used for signing with the hash\n algorithm that will be employed in the server\n key exchange message.\n\n DH_DSS Diffie-Hellman public key; the keyAgreement bit\n DH_RSA MUST be set if the key usage extension is\n present.\n\n ECDH_ECDSA ECDH-capable public key; the public key MUST\n ECDH_RSA use a curve and point format supported by the\n client, as described in [TLSECC].\n\n ECDHE_ECDSA ECDSA-capable public key; the certificate MUST\n allow the key to be used for signing with the\n hash algorithm that will be employed in the\n server key exchange message. The public key\n MUST use a curve and point format supported by\n the client, as described in [TLSECC].\n\n - The \"server_name\" and \"trusted_ca_keys\" extensions [TLSEXT] are\n used to guide certificate selection.\n\n\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 5705 RTFM, Inc.\nCategory: Standards Track March 2010\nISSN: 2070-1721\n\n\n Keying Material Exporters for Transport Layer Security (TLS)\n\nAbstract\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n to perform key establishment but then use some of the keying material\n for their own purposes. This document describes a general mechanism\n for allowing that.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/5705.\n\nCopyright Notice\n\n Copyright (c) 2010 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2\n 2. Conventions Used In This Document . . . . . . . . . . . . . . . 3\n 3. Binding to Application Contexts . . . . . . . . . . . . . . . . 3\n 4. Exporter Definition . . . . . . . . . . . . . . . . . . . . . . 4\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . . 5\n 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 6\n 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . 6\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 7\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 7\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 7\n\n1. Introduction\n\n Note: The mechanism described in this document was previously known\n as \"TLS Extractors\" but was changed to avoid a name conflict\n with the use of the term \"Extractor\" in the cryptographic\n community.\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n [RFC5246] or Datagram TLS (DTLS) [RFC4347] to perform key\n establishment but then use some of the keying material for their own\n purposes. A typical example is DTLS-SRTP [DTLS-SRTP], a key\n management scheme for the Secure Real-time Transport Protocol (SRTP)\n that uses DTLS to perform a key exchange and negotiate the SRTP\n [RFC3711] protection suite and then uses the DTLS master_secret to\n generate the SRTP keys.\n\n These applications imply a need to be able to export keying material\n (later called Exported Keying Material or EKM) from TLS/DTLS to an\n application or protocol residing at an upper layer, and to securely\n agree on the upper-layer context where the keying material will be\n used. The mechanism for exporting the keying material has the\n following requirements:\n\n o Both client and server need to be able to export the same EKM\n value.\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n o EKM values should be indistinguishable from random data to\n attackers who don't know the master_secret.\n\n o It should be possible to export multiple EKM values from the same\n TLS/DTLS association.\n\n o Knowing one EKM value should not reveal any useful information\n about the master_secret or about other EKM values.\n\n The mechanism described in this document is intended to fulfill these\n requirements. This mechanism is compatible with all versions of TLS.\n\n2. Conventions Used In This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Binding to Application Contexts\n\n In addition to using an exporter to obtain keying material, an\n application using the keying material has to securely establish the\n upper-layer context where the keying material will be used. The\n details of this context depend on the application, but it could\n include things such as algorithms and parameters that will be used\n with the keys, identifier(s) for the endpoint(s) who will use the\n keys, identifier(s) for the session(s) where the keys will be used,\n and the lifetime(s) for the context and/or keys. At a minimum, there\n should be some mechanism for signaling that an exporter will be used.\n\n This specification does not mandate a single mechanism for agreeing\n on such context; instead, there are several possibilities that can be\n used (and can complement each other). For example:\n\n o Information about the upper-layer context can be included in the\n optional data after the exporter label (see Section 4).\n\n o Information about the upper-layer context can be exchanged in TLS\n extensions included in the ClientHello and ServerHello messages.\n This approach is used in [DTLS-SRTP]. The handshake messages are\n protected by the Finished messages, so once the handshake\n completes, the peers will have the same view of the information.\n Extensions also allow a limited form of negotiation: for example,\n the TLS client could propose several alternatives for some context\n parameters, and the TLS server could select one of them.\n\n o The upper-layer protocol can include its own handshake, which can\n be protected using the keys exported by TLS.\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n No matter how the context is agreed, it is required that it has one\n part that indicates which application will use the exported keys.\n This part is the disambiguating label string (see Section 4).\n\n It is important to note that just embedding TLS messages in the\n upper-layer protocol may not automatically secure all the important\n context information, since the upper-layer messages are not covered\n by TLS Finished messages.\n\n4. Exporter Definition\n\n The output of the exporter is intended to be used in a single scope,\n which is associated with the TLS session, the label, and the context\n value.\n\n The exporter takes three input values:\n\n o a disambiguating label string,\n\n o a per-association context value provided by the application using\n the exporter, and\n\n o a length value.\n\n If no context is provided, it then computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random\n )[length]\n\n If context is provided, it computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random +\n context_value_length + context_value\n )[length]\n\n Where PRF is the TLS Pseudorandom Function in use for the session.\n The output is a pseudorandom bit string of length bytes generated\n from the master_secret. (This construction allows for\n interoperability with older exporter-type constructions which do not\n use context values, e.g., [RFC5281]).\n\n Labels here have the same definition as in TLS, i.e., an ASCII string\n with no terminating NULL. Label values beginning with \"EXPERIMENTAL\"\n MAY be used for private use without registration. All other label\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n values MUST be registered via Specification Required as described by\n RFC 5226 [RFC5226]. Note that exporter labels have the potential to\n collide with existing PRF labels. In order to prevent this, labels\n SHOULD begin with \"EXPORTER\". This is not a MUST because there are\n existing uses that have labels which do not begin with this prefix.\n\n The context value allows the application using the exporter to mix\n its own data with the TLS PRF for the exporter output. One example\n of where this might be useful is an authentication setting where the\n client credentials are valid for more than one identity; the context\n value could then be used to mix the expected identity into the keying\n material, thus preventing substitution attacks. The context value\n length is encoded as an unsigned, 16-bit quantity (uint16; see\n [RFC5246], Section 4.4) representing the length of the context value.\n The context MAY be zero length. Because the context value is mixed\n with the master_secret via the PRF, it is safe to mix confidential\n information into the exporter, provided that the master_secret will\n not be known to the attacker.\n\n5. Security Considerations\n\n The prime security requirement for exporter outputs is that they be\n independent. More formally, after a particular TLS session, if an\n adversary is allowed to choose multiple (label, context value) pairs\n and is given the output of the PRF for those values, the attacker is\n still unable to distinguish between the output of the PRF for a\n (label, context value) pair (different from the ones that it\n submitted) and a random value of the same length. In particular,\n there may be settings, such as the one described in Section 4, where\n the attacker can control the context value; such an attacker MUST NOT\n be able to predict the output of the exporter. Similarly, an\n attacker who does not know the master secret should not be able to\n distinguish valid exporter outputs from random values. The current\n set of TLS PRFs is believed to meet this objective, provided the\n master secret is randomly generated.\n\n Because an exporter produces the same value if applied twice with the\n same label to the same master_secret, it is critical that two EKM\n values generated with the same label not be used for two different\n purposes -- hence, the requirement for IANA registration. However,\n because exporters depend on the TLS PRF, it is not a threat to the\n use of an EKM value generated from one label to reveal an EKM value\n generated from another label.\n\n With certain TLS cipher suites, the TLS master secret is not\n necessarily unique to a single TLS session. In particular, with RSA\n key exchange, a malicious party acting as TLS server in one session\n and as TLS client in another session can cause those two sessions to\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n have the same TLS master secret (though the sessions must be\n established simultaneously to get adequate control of the Random\n values). Applications using the EKM need to consider this in how\n they use the EKM; in some cases, requiring the use of other cipher\n suites (such as those using a Diffie-Hellman key exchange) may be\n advisable.\n\n Designing a secure mechanism that uses exporters is not necessarily\n straightforward. This document only provides the exporter mechanism,\n but the problem of agreeing on the surrounding context and the\n meaning of the information passed to and from the exporter remains.\n Any new uses of the exporter mechanism should be subject to careful\n review.\n\n6. IANA Considerations\n\n IANA has created a TLS Exporter Label registry for this purpose. The\n initial contents of the registry are given below:\n\n Value Reference Note\n ----------------------------- --------- ----\n client finished [RFC5246] (1)\n server finished [RFC5246] (1)\n master secret [RFC5246] (1)\n key expansion [RFC5246] (1)\n client EAP encryption [RFC5216]\n ttls keying material [RFC5281]\n ttls challenge [RFC5281]\n\n Note: (1) These entries are reserved and MUST NOT be used for the\n purpose described in RFC 5705, in order to avoid confusion with\n similar, but distinct, use in RFC 5246.\n\n Future values are allocated via the RFC 5226 Specification Required\n policy. The label is a string consisting of printable ASCII\n characters. IANA MUST also verify that one label is not a prefix of\n any other label. For example, labels \"key\" or \"master secretary\" are\n forbidden.\n\n7. Acknowledgments\n\n Thanks to Pasi Eronen for valuable comments and for the contents of\n the IANA section and Section 3. Thanks to David McGrew for helpful\n discussion of the security considerations and to Vijay Gurbani and\n Alfred Hoenes for editorial comments.\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246,\n August 2008.\n\n8.2. Informative References\n\n [DTLS-SRTP] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for Secure\n Real-time Transport Protocol (SRTP)\", Work in Progress,\n February 2009.\n\n [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and\n K. Norrman, \"The Secure Real-time Transport Protocol\n (SRTP)\", RFC 3711, March 2004.\n\n [RFC4347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security\", RFC 4347, April 2006.\n\n [RFC5216] Simon, D., Aboba, B., and R. Hurst, \"The EAP-TLS\n Authentication Protocol\", RFC 5216, March 2008.\n\n [RFC5281] Funk, P. and S. Blake-Wilson, \"Extensible Authentication\n Protocol Tunneled Transport Layer Security Authenticated\n Protocol Version 0 (EAP-TTLSv0)\", RFC 5281, August 2008.\n\nAuthor's Address\n\n Eric Rescorla\n RTFM, Inc.\n 2064 Edgewood Drive\n Palo Alto, CA 94303\n USA\n\n EMail: [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) D. Eastlake 3rd\nRequest for Comments: 6066 Huawei\nObsoletes: 4366 January 2011\nCategory: Standards Track\nISSN: 2070-1721\n\n\n Transport Layer Security (TLS) Extensions: Extension Definitions\n\nAbstract\n\n This document provides specifications for existing TLS extensions.\n It is a companion document for RFC 5246, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\". The extensions specified are\n server_name, max_fragment_length, client_certificate_url,\n trusted_ca_keys, truncated_hmac, and status_request.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6066.\n\nCopyright Notice\n\n Copyright (c) 2011 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n\n\n\n\n\nEastlake Standards Track [Page 1]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction ....................................................3\n 1.1. Specific Extensions Covered ................................3\n 1.2. Conventions Used in This Document ..........................5\n 2. Extensions to the Handshake Protocol ............................5\n 3. Server Name Indication ..........................................6\n 4. Maximum Fragment Length Negotiation .............................8\n 5. Client Certificate URLs .........................................9\n 6. Trusted CA Indication ..........................................12\n 7. Truncated HMAC .................................................13\n 8. Certificate Status Request .....................................14\n 9. Error Alerts ...................................................16\n 10. IANA Considerations ...........................................17\n 10.1. pkipath MIME Type Registration ...........................17\n 10.2. Reference for TLS Alerts, TLS HandshakeTypes, and\n ExtensionTypes ...........................................19\n 11. Security Considerations .......................................19\n 11.1. Security Considerations for server_name ..................19\n 11.2. Security Considerations for max_fragment_length ..........20\n 11.3. Security Considerations for client_certificate_url .......20\n 11.4. Security Considerations for trusted_ca_keys ..............21\n 11.5. Security Considerations for truncated_hmac ...............21\n 11.6. Security Considerations for status_request ...............22\n 12. Normative References ..........................................22\n 13. Informative References ........................................23\n Appendix A. Changes from RFC 4366 .................................24\n Appendix B. Acknowledgements ......................................25\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 2]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1. Introduction\n\n The Transport Layer Security (TLS) Protocol Version 1.2 is specified\n in [RFC5246]. That specification includes the framework for\n extensions to TLS, considerations in designing such extensions (see\n Section 7.4.1.4 of [RFC5246]), and IANA Considerations for the\n allocation of new extension code points; however, it does not specify\n any particular extensions other than Signature Algorithms (see\n Section 7.4.1.4.1 of [RFC5246]).\n\n This document provides the specifications for existing TLS\n extensions. It is, for the most part, the adaptation and editing of\n material from RFC 4366, which covered TLS extensions for TLS 1.0 (RFC\n 2246) and TLS 1.1 (RFC 4346).\n\n1.1. Specific Extensions Covered\n\n The extensions described here focus on extending the functionality\n provided by the TLS protocol message formats. Other issues, such as\n the addition of new cipher suites, are deferred.\n\n The extension types defined in this document are:\n\n enum {\n server_name(0), max_fragment_length(1),\n client_certificate_url(2), trusted_ca_keys(3),\n truncated_hmac(4), status_request(5), (65535)\n } ExtensionType;\n\n Specifically, the extensions described in this document:\n\n - Allow TLS clients to provide to the TLS server the name of the\n server they are contacting. This functionality is desirable in\n order to facilitate secure connections to servers that host\n multiple 'virtual' servers at a single underlying network address.\n\n - Allow TLS clients and servers to negotiate the maximum fragment\n length to be sent. This functionality is desirable as a result of\n memory constraints among some clients, and bandwidth constraints\n among some access networks.\n\n - Allow TLS clients and servers to negotiate the use of client\n certificate URLs. This functionality is desirable in order to\n conserve memory on constrained clients.\n\n\n\n\n\n\n\nEastlake Standards Track [Page 3]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - Allow TLS clients to indicate to TLS servers which certification\n authority (CA) root keys they possess. This functionality is\n desirable in order to prevent multiple handshake failures\n involving TLS clients that are only able to store a small number\n of CA root keys due to memory limitations.\n\n - Allow TLS clients and servers to negotiate the use of truncated\n Message Authentication Codes (MACs). This functionality is\n desirable in order to conserve bandwidth in constrained access\n networks.\n\n - Allow TLS clients and servers to negotiate that the server sends\n the client certificate status information (e.g., an Online\n Certificate Status Protocol (OCSP) [RFC2560] response) during a\n TLS handshake. This functionality is desirable in order to avoid\n sending a Certificate Revocation List (CRL) over a constrained\n access network and therefore saving bandwidth.\n\n TLS clients and servers may use the extensions described in this\n document. The extensions are designed to be backwards compatible,\n meaning that TLS clients that support the extensions can talk to TLS\n servers that do not support the extensions, and vice versa.\n\n Note that any messages associated with these extensions that are sent\n during the TLS handshake MUST be included in the hash calculations\n involved in \"Finished\" messages.\n\n Note also that all the extensions defined in this document are\n relevant only when a session is initiated. A client that requests\n session resumption does not in general know whether the server will\n accept this request, and therefore it SHOULD send the same extensions\n as it would send if it were not attempting resumption. When a client\n includes one or more of the defined extension types in an extended\n client hello while requesting session resumption:\n\n - The server name indication extension MAY be used by the server\n when deciding whether or not to resume a session as described in\n Section 3.\n\n - If the resumption request is denied, the use of the extensions is\n negotiated as normal.\n\n - If, on the other hand, the older session is resumed, then the\n server MUST ignore the extensions and send a server hello\n containing none of the extension types. In this case, the\n functionality of these extensions negotiated during the original\n session initiation is applied to the resumed session.\n\n\n\n\nEastlake Standards Track [Page 4]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1.2. Conventions Used in This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n [RFC2119].\n\n2. Extensions to the Handshake Protocol\n\n This document specifies the use of two new handshake messages,\n \"CertificateURL\" and \"CertificateStatus\". These messages are\n described in Sections 5 and 8, respectively. The new handshake\n message structure therefore becomes:\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), certificate_url(21), certificate_status(22),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case certificate_url: CertificateURL;\n case certificate_status: CertificateStatus;\n } body;\n } Handshake;\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 5]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n3. Server Name Indication\n\n TLS does not provide a mechanism for a client to tell a server the\n name of the server it is contacting. It may be desirable for clients\n to provide this information to facilitate secure connections to\n servers that host multiple 'virtual' servers at a single underlying\n network address.\n\n In order to provide any of the server names, clients MAY include an\n extension of type \"server_name\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"ServerNameList\" where:\n\n struct {\n NameType name_type;\n select (name_type) {\n case host_name: HostName;\n } name;\n } ServerName;\n\n enum {\n host_name(0), (255)\n } NameType;\n\n opaque HostName<1..2^16-1>;\n\n struct {\n ServerName server_name_list<1..2^16-1>\n } ServerNameList;\n\n The ServerNameList MUST NOT contain more than one name of the same\n name_type. If the server understood the ClientHello extension but\n does not recognize the server name, the server SHOULD take one of two\n actions: either abort the handshake by sending a fatal-level\n unrecognized_name(112) alert or continue the handshake. It is NOT\n RECOMMENDED to send a warning-level unrecognized_name(112) alert,\n because the client's behavior in response to warning-level alerts is\n unpredictable. If there is a mismatch between the server name used\n by the client application and the server name of the credential\n chosen by the server, this mismatch will become apparent when the\n client application performs the server endpoint identification, at\n which point the client application will have to decide whether to\n proceed with the communication. TLS implementations are encouraged\n to make information available to application callers about warning-\n level alerts that were received or sent during a TLS handshake. Such\n information can be useful for diagnostic purposes.\n\n\n\n\n\nEastlake Standards Track [Page 6]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note: Earlier versions of this specification permitted multiple\n names of the same name_type. In practice, current client\n implementations only send one name, and the client cannot\n necessarily find out which name the server selected. Multiple\n names of the same name_type are therefore now prohibited.\n\n Currently, the only server names supported are DNS hostnames;\n however, this does not imply any dependency of TLS on DNS, and other\n name types may be added in the future (by an RFC that updates this\n document). The data structure associated with the host_name NameType\n is a variable-length vector that begins with a 16-bit length. For\n backward compatibility, all future data structures associated with\n new NameTypes MUST begin with a 16-bit length field. TLS MAY treat\n provided server names as opaque data and pass the names and types to\n the application.\n\n \"HostName\" contains the fully qualified DNS hostname of the server,\n as understood by the client. The hostname is represented as a byte\n string using ASCII encoding without a trailing dot. This allows the\n support of internationalized domain names through the use of A-labels\n defined in [RFC5890]. DNS hostnames are case-insensitive. The\n algorithm to compare hostnames is described in [RFC5890], Section\n 2.3.2.4.\n\n Literal IPv4 and IPv6 addresses are not permitted in \"HostName\".\n\n It is RECOMMENDED that clients include an extension of type\n \"server_name\" in the client hello whenever they locate a server by a\n supported name type.\n\n A server that receives a client hello containing the \"server_name\"\n extension MAY use the information contained in the extension to guide\n its selection of an appropriate certificate to return to the client,\n and/or other aspects of security policy. In this event, the server\n SHALL include an extension of type \"server_name\" in the (extended)\n server hello. The \"extension_data\" field of this extension SHALL be\n empty.\n\n When the server is deciding whether or not to accept a request to\n resume a session, the contents of a server_name extension MAY be used\n in the lookup of the session in the session cache. The client SHOULD\n include the same server_name extension in the session resumption\n request as it did in the full handshake that established the session.\n A server that implements this extension MUST NOT accept the request\n to resume the session if the server_name extension contains a\n different name. Instead, it proceeds with a full handshake to\n establish a new session. When resuming a session, the server MUST\n NOT include a server_name extension in the server hello.\n\n\n\nEastlake Standards Track [Page 7]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n If an application negotiates a server name using an application\n protocol and then upgrades to TLS, and if a server_name extension is\n sent, then the extension SHOULD contain the same name that was\n negotiated in the application protocol. If the server_name is\n established in the TLS session handshake, the client SHOULD NOT\n attempt to request a different server name at the application layer.\n\n4. Maximum Fragment Length Negotiation\n\n Without this extension, TLS specifies a fixed maximum plaintext\n fragment length of 2^14 bytes. It may be desirable for constrained\n clients to negotiate a smaller maximum fragment length due to memory\n limitations or bandwidth limitations.\n\n In order to negotiate smaller maximum fragment lengths, clients MAY\n include an extension of type \"max_fragment_length\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain:\n\n enum{\n 2^9(1), 2^10(2), 2^11(3), 2^12(4), (255)\n } MaxFragmentLength;\n\n whose value is the desired maximum fragment length. The allowed\n values for this field are: 2^9, 2^10, 2^11, and 2^12.\n\n Servers that receive an extended client hello containing a\n \"max_fragment_length\" extension MAY accept the requested maximum\n fragment length by including an extension of type\n \"max_fragment_length\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL contain a\n \"MaxFragmentLength\" whose value is the same as the requested maximum\n fragment length.\n\n If a server receives a maximum fragment length negotiation request\n for a value other than the allowed values, it MUST abort the\n handshake with an \"illegal_parameter\" alert. Similarly, if a client\n receives a maximum fragment length negotiation response that differs\n from the length it requested, it MUST also abort the handshake with\n an \"illegal_parameter\" alert.\n\n Once a maximum fragment length other than 2^14 has been successfully\n negotiated, the client and server MUST immediately begin fragmenting\n messages (including handshake messages) to ensure that no fragment\n larger than the negotiated length is sent. Note that TLS already\n requires clients and servers to support fragmentation of handshake\n messages.\n\n\n\n\nEastlake Standards Track [Page 8]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n The negotiated length applies for the duration of the session\n including session resumptions.\n\n The negotiated length limits the input that the record layer may\n process without fragmentation (that is, the maximum value of\n TLSPlaintext.length; see [RFC5246], Section 6.2.1). Note that the\n output of the record layer may be larger. For example, if the\n negotiated length is 2^9=512, then, when using currently defined\n cipher suites (those defined in [RFC5246] and [RFC2712]) and null\n compression, the record-layer output can be at most 805 bytes: 5\n bytes of headers, 512 bytes of application data, 256 bytes of\n padding, and 32 bytes of MAC. This means that in this event a TLS\n record-layer peer receiving a TLS record-layer message larger than\n 805 bytes MUST discard the message and send a \"record_overflow\"\n alert, without decrypting the message. When this extension is used\n with Datagram Transport Layer Security (DTLS), implementations SHOULD\n NOT generate record_overflow alerts unless the packet passes message\n authentication.\n\n5. Client Certificate URLs\n\n Without this extension, TLS specifies that when client authentication\n is performed, client certificates are sent by clients to servers\n during the TLS handshake. It may be desirable for constrained\n clients to send certificate URLs in place of certificates, so that\n they do not need to store their certificates and can therefore save\n memory.\n\n In order to negotiate sending certificate URLs to a server, clients\n MAY include an extension of type \"client_certificate_url\" in the\n (extended) client hello. The \"extension_data\" field of this\n extension SHALL be empty.\n\n (Note that it is necessary to negotiate the use of client certificate\n URLs in order to avoid \"breaking\" existing TLS servers.)\n\n Servers that receive an extended client hello containing a\n \"client_certificate_url\" extension MAY indicate that they are willing\n to accept certificate URLs by including an extension of type\n \"client_certificate_url\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n After negotiation of the use of client certificate URLs has been\n successfully completed (by exchanging hellos including\n \"client_certificate_url\" extensions), clients MAY send a\n \"CertificateURL\" message in place of a \"Certificate\" message as\n follows (see also Section 2):\n\n\n\n\nEastlake Standards Track [Page 9]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n enum {\n individual_certs(0), pkipath(1), (255)\n } CertChainType;\n\n struct {\n CertChainType type;\n URLAndHash url_and_hash_list<1..2^16-1>;\n } CertificateURL;\n\n struct {\n opaque url<1..2^16-1>;\n unint8 padding;\n opaque SHA1Hash[20];\n } URLAndHash;\n\n Here, \"url_and_hash_list\" contains a sequence of URLs and hashes.\n Each \"url\" MUST be an absolute URI reference according to [RFC3986]\n that can be immediately used to fetch the certificate(s).\n\n When X.509 certificates are used, there are two possibilities:\n\n - If CertificateURL.type is \"individual_certs\", each URL refers to a\n single DER-encoded X.509v3 certificate, with the URL for the\n client's certificate first.\n\n - If CertificateURL.type is \"pkipath\", the list contains a single\n URL referring to a DER-encoded certificate chain, using the type\n PkiPath described in Section 10.1.\n\n When any other certificate format is used, the specification that\n describes use of that format in TLS should define the encoding format\n of certificates or certificate chains, and any constraint on their\n ordering.\n\n The \"padding\" byte MUST be 0x01. It is present to make the structure\n backwards compatible.\n\n The hash corresponding to each URL is the SHA-1 hash of the\n certificate or certificate chain (in the case of X.509 certificates,\n the DER-encoded certificate or the DER-encoded PkiPath).\n\n Note that when a list of URLs for X.509 certificates is used, the\n ordering of URLs is the same as that used in the TLS Certificate\n message (see [RFC5246], Section 7.4.2), but opposite to the order in\n which certificates are encoded in PkiPath. In either case, the self-\n signed root certificate MAY be omitted from the chain, under the\n assumption that the server must already possess it in order to\n validate it.\n\n\n\nEastlake Standards Track [Page 10]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Servers receiving \"CertificateURL\" SHALL attempt to retrieve the\n client's certificate chain from the URLs and then process the\n certificate chain as usual. A cached copy of the content of any URL\n in the chain MAY be used, provided that the SHA-1 hash matches the\n hash of the cached copy.\n\n Servers that support this extension MUST support the 'http' URI\n scheme for certificate URLs and MAY support other schemes. Use of\n other schemes than 'http', 'https', or 'ftp' may create unexpected\n problems.\n\n If the protocol used is HTTP, then the HTTP server can be configured\n to use the Cache-Control and Expires directives described in\n [RFC2616] to specify whether and for how long certificates or\n certificate chains should be cached.\n\n The TLS server MUST NOT follow HTTP redirects when retrieving the\n certificates or certificate chain. The URLs used in this extension\n MUST NOT be chosen to depend on such redirects.\n\n If the protocol used to retrieve certificates or certificate chains\n returns a MIME-formatted response (as HTTP does), then the following\n MIME Content-Types SHALL be used: when a single X.509v3 certificate\n is returned, the Content-Type is \"application/pkix-cert\" [RFC2585],\n and when a chain of X.509v3 certificates is returned, the Content-\n Type is \"application/pkix-pkipath\" (Section 10.1).\n\n The server MUST check that the SHA-1 hash of the contents of the\n object retrieved from that URL (after decoding any MIME Content-\n Transfer-Encoding) matches the given hash. If any retrieved object\n does not have the correct SHA-1 hash, the server MUST abort the\n handshake with a bad_certificate_hash_value(114) alert. This alert\n is always fatal.\n\n Clients may choose to send either \"Certificate\" or \"CertificateURL\"\n after successfully negotiating the option to send certificate URLs.\n The option to send a certificate is included to provide flexibility\n to clients possessing multiple certificates.\n\n If a server is unable to obtain certificates in a given\n CertificateURL, it MUST send a fatal certificate_unobtainable(111)\n alert if it requires the certificates to complete the handshake. If\n the server does not require the certificates, then the server\n continues the handshake. The server MAY send a warning-level alert\n in this case. Clients receiving such an alert SHOULD log the alert\n and continue with the handshake if possible.\n\n\n\n\n\nEastlake Standards Track [Page 11]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n6. Trusted CA Indication\n\n Constrained clients that, due to memory limitations, possess only a\n small number of CA root keys may wish to indicate to servers which\n root keys they possess, in order to avoid repeated handshake\n failures.\n\n In order to indicate which CA root keys they possess, clients MAY\n include an extension of type \"trusted_ca_keys\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain \"TrustedAuthorities\" where:\n\n struct {\n TrustedAuthority trusted_authorities_list<0..2^16-1>;\n } TrustedAuthorities;\n\n struct {\n IdentifierType identifier_type;\n select (identifier_type) {\n case pre_agreed: struct {};\n case key_sha1_hash: SHA1Hash;\n case x509_name: DistinguishedName;\n case cert_sha1_hash: SHA1Hash;\n } identifier;\n } TrustedAuthority;\n\n enum {\n pre_agreed(0), key_sha1_hash(1), x509_name(2),\n cert_sha1_hash(3), (255)\n } IdentifierType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n Here, \"TrustedAuthorities\" provides a list of CA root key identifiers\n that the client possesses. Each CA root key is identified via\n either:\n\n - \"pre_agreed\": no CA root key identity supplied.\n\n - \"key_sha1_hash\": contains the SHA-1 hash of the CA root key. For\n Digital Signature Algorithm (DSA) and Elliptic Curve Digital\n Signature Algorithm (ECDSA) keys, this is the hash of the\n \"subjectPublicKey\" value. For RSA keys, the hash is of the big-\n endian byte string representation of the modulus without any\n initial zero-valued bytes. (This copies the key hash formats\n deployed in other environments.)\n\n\n\n\n\nEastlake Standards Track [Page 12]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - \"x509_name\": contains the DER-encoded X.509 DistinguishedName of\n the CA.\n\n - \"cert_sha1_hash\": contains the SHA-1 hash of a DER-encoded\n Certificate containing the CA root key.\n\n Note that clients may include none, some, or all of the CA root keys\n they possess in this extension.\n\n Note also that it is possible that a key hash or a Distinguished Name\n alone may not uniquely identify a certificate issuer (for example, if\n a particular CA has multiple key pairs). However, here we assume\n this is the case following the use of Distinguished Names to identify\n certificate issuers in TLS.\n\n The option to include no CA root keys is included to allow the client\n to indicate possession of some pre-defined set of CA root keys.\n\n Servers that receive a client hello containing the \"trusted_ca_keys\"\n extension MAY use the information contained in the extension to guide\n their selection of an appropriate certificate chain to return to the\n client. In this event, the server SHALL include an extension of type\n \"trusted_ca_keys\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n7. Truncated HMAC\n\n Currently defined TLS cipher suites use the MAC construction HMAC\n [RFC2104] to authenticate record-layer communications. In TLS, the\n entire output of the hash function is used as the MAC tag. However,\n it may be desirable in constrained environments to save bandwidth by\n truncating the output of the hash function to 80 bits when forming\n MAC tags.\n\n In order to negotiate the use of 80-bit truncated HMAC, clients MAY\n include an extension of type \"truncated_hmac\" in the extended client\n hello. The \"extension_data\" field of this extension SHALL be empty.\n\n Servers that receive an extended hello containing a \"truncated_hmac\"\n extension MAY agree to use a truncated HMAC by including an extension\n of type \"truncated_hmac\", with empty \"extension_data\", in the\n extended server hello.\n\n Note that if new cipher suites are added that do not use HMAC, and\n the session negotiates one of these cipher suites, this extension\n will have no effect. It is strongly recommended that any new cipher\n suites using other MACs consider the MAC size an integral part of the\n\n\n\n\nEastlake Standards Track [Page 13]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n cipher suite definition, taking into account both security and\n bandwidth considerations.\n\n If HMAC truncation has been successfully negotiated during a TLS\n handshake, and the negotiated cipher suite uses HMAC, both the client\n and the server pass this fact to the TLS record layer along with the\n other negotiated security parameters. Subsequently during the\n session, clients and servers MUST use truncated HMACs, calculated as\n specified in [RFC2104]. That is, SecurityParameters.mac_length is 10\n bytes, and only the first 10 bytes of the HMAC output are transmitted\n and checked. Note that this extension does not affect the\n calculation of the pseudo-random function (PRF) as part of\n handshaking or key derivation.\n\n The negotiated HMAC truncation size applies for the duration of the\n session including session resumptions.\n\n8. Certificate Status Request\n\n Constrained clients may wish to use a certificate-status protocol\n such as OCSP [RFC2560] to check the validity of server certificates,\n in order to avoid transmission of CRLs and therefore save bandwidth\n on constrained networks. This extension allows for such information\n to be sent in the TLS handshake, saving roundtrips and resources.\n\n In order to indicate their desire to receive certificate status\n information, clients MAY include an extension of type\n \"status_request\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"CertificateStatusRequest\" where:\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n } request;\n } CertificateStatusRequest;\n\n enum { ocsp(1), (255) } CertificateStatusType;\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n\n\n\nEastlake Standards Track [Page 14]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n In the OCSPStatusRequest, the \"ResponderIDs\" provides a list of OCSP\n responders that the client trusts. A zero-length \"responder_id_list\"\n sequence has the special meaning that the responders are implicitly\n known to the server, e.g., by prior arrangement. \"Extensions\" is a\n DER encoding of OCSP request extensions.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC2560]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a zero-length ASN.1 SEQUENCE, which is not\n valid for the \"Extensions\" type).\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement).\n\n Servers that receive a client hello containing the \"status_request\"\n extension MAY return a suitable certificate status response to the\n client along with their certificate. If OCSP is requested, they\n SHOULD use the information contained in the extension when selecting\n an OCSP responder and SHOULD include request_extensions in the OCSP\n request.\n\n Servers return a certificate response along with their certificate by\n sending a \"CertificateStatus\" message immediately after the\n \"Certificate\" message (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message, then the server MUST have included an\n extension of type \"status_request\" with empty \"extension_data\" in the\n extended server hello. The \"CertificateStatus\" message is conveyed\n using the handshake message type \"certificate_status\" as follows (see\n also Section 2):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<1..2^24-1>;\n\n An \"ocsp_response\" contains a complete, DER-encoded OCSP response\n (using the ASN.1 type OCSPResponse defined in [RFC2560]). Only one\n OCSP response may be sent.\n\n\n\n\nEastlake Standards Track [Page 15]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if has received a \"status_request\" extension in the\n client hello message and has sent a \"status_request\" extension in the\n server hello message.\n\n Note in addition that a server MUST NOT send the \"CertificateStatus\"\n message unless it received a \"status_request\" extension in the client\n hello message and sent a \"status_request\" extension in the server\n hello message.\n\n Clients requesting an OCSP response and receiving an OCSP response in\n a \"CertificateStatus\" message MUST check the OCSP response and abort\n the handshake if the response is not satisfactory with\n bad_certificate_status_response(113) alert. This alert is always\n fatal.\n\n9. Error Alerts\n\n Four new error alerts are defined for use with the TLS extensions\n defined in this document. To avoid \"breaking\" existing clients and\n servers, these alerts MUST NOT be sent unless the sending party has\n received an extended hello message from the party they are\n communicating with. These error alerts are conveyed using the\n following syntax. The new alerts are the last four, as indicated by\n the comments on the same line as the error alert number.\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n /* 41 is not defined, for historical reasons */\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction(60),\n protocol_version(70),\n insufficient_security(71),\n\n\n\nEastlake Standards Track [Page 16]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n certificate_unobtainable(111), /* new */\n unrecognized_name(112), /* new */\n bad_certificate_status_response(113), /* new */\n bad_certificate_hash_value(114), /* new */\n (255)\n } AlertDescription;\n\n \"certificate_unobtainable\" is described in Section 5.\n \"unrecognized_name\" is described in Section 3.\n \"bad_certificate_status_response\" is described in Section 8.\n \"bad_certificate_hash_value\" is described in Section 5.\n\n10. IANA Considerations\n\n IANA Considerations for TLS extensions and the creation of a registry\n are covered in Section 12 of [RFC5246] except for the registration of\n MIME type application/pkix-pkipath, which appears below.\n\n The IANA TLS extensions and MIME type application/pkix-pkipath\n registry entries that reference RFC 4366 have been updated to\n reference this document.\n\n10.1. pkipath MIME Type Registration\n\n MIME media type name: application\n MIME subtype name: pkix-pkipath\n Required parameters: none\n\n Optional parameters: version (default value is \"1\")\n\n Encoding considerations:\n Binary; this MIME type is a DER encoding of the ASN.1 type\n PkiPath, defined as follows:\n PkiPath ::= SEQUENCE OF Certificate\n PkiPath is used to represent a certification path. Within the\n sequence, the order of certificates is such that the subject of\n the first certificate is the issuer of the second certificate,\n etc.\n This is identical to the definition published in [X509-4th-TC1];\n note that it is different from that in [X509-4th].\n\n All Certificates MUST conform to [RFC5280]. (This should be\n interpreted as a requirement to encode only PKIX-conformant\n certificates using this type. It does not necessarily require\n\n\n\nEastlake Standards Track [Page 17]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n that all certificates that are not strictly PKIX-conformant must\n be rejected by relying parties, although the security consequences\n of accepting any such certificates should be considered\n carefully.)\n\n DER (as opposed to BER) encoding MUST be used. If this type is\n sent over a 7-bit transport, base64 encoding SHOULD be used.\n\n Security considerations:\n The security considerations of [X509-4th] and [RFC5280] (or any\n updates to them) apply, as well as those of any protocol that uses\n this type (e.g., TLS).\n\n Note that this type only specifies a certificate chain that can be\n assessed for validity according to the relying party's existing\n configuration of trusted CAs; it is not intended to be used to\n specify any change to that configuration.\n\n Interoperability considerations:\n No specific interoperability problems are known with this type,\n but for recommendations relating to X.509 certificates in general,\n see [RFC5280].\n\n Published specification: This document and [RFC5280].\n\n Applications that use this media type:\n TLS. It may also be used by other protocols or for general\n interchange of PKIX certificate chains.\n\n Additional information:\n Magic number(s): DER-encoded ASN.1 can be easily recognized.\n Further parsing is required to distinguish it from other ASN.1\n types.\n File extension(s): .pkipath\n Macintosh File Type Code(s): not specified\n\n Person & email address to contact for further information:\n Magnus Nystrom <[redacted-email]>\n\n Intended usage: COMMON\n\n Change controller: IESG <[redacted-email]>\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 18]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n10.2. Reference for TLS Alerts, TLS HandshakeTypes, and ExtensionTypes\n\n The following values in the TLS Alert Registry have been updated to\n reference this document:\n\n 111 certificate_unobtainable\n 112 unrecognized_name\n 113 bad_certificate_status_response\n 114 bad_certificate_hash_value\n\n The following values in the TLS HandshakeType Registry have been\n updated to reference this document:\n\n 21 certificate_url\n 22 certificate_status\n\n The following ExtensionType values have been updated to reference\n this document:\n\n 0 server_name\n 1 max_fragment_length\n 2 client_certificate_url\n 3 trusted_ca_keys\n 4 truncated_hmac\n 5 status_request\n\n11. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security Considerations for particular extensions\n specified in this document are given below.\n\n In general, implementers should continue to monitor the state of the\n art and address any weaknesses identified.\n\n11.1. Security Considerations for server_name\n\n If a single server hosts several domains, then clearly it is\n necessary for the owners of each domain to ensure that this satisfies\n their security needs. Apart from this, server_name does not appear\n to introduce significant security issues.\n\n Since it is possible for a client to present a different server_name\n in the application protocol, application server implementations that\n rely upon these names being the same MUST check to make sure the\n client did not present a different name in the application protocol.\n\n\n\n\n\nEastlake Standards Track [Page 19]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Implementations MUST ensure that a buffer overflow does not occur,\n whatever the values of the length fields in server_name.\n\n11.2. Security Considerations for max_fragment_length\n\n The maximum fragment length takes effect immediately, including for\n handshake messages. However, that does not introduce any security\n complications that are not already present in TLS, since TLS requires\n implementations to be able to handle fragmented handshake messages.\n\n Note that, as described in Section 4, once a non-null cipher suite\n has been activated, the effective maximum fragment length depends on\n the cipher suite and compression method, as well as on the negotiated\n max_fragment_length. This must be taken into account when sizing\n buffers and checking for buffer overflow.\n\n11.3. Security Considerations for client_certificate_url\n\n Support for client_certificate_url involves the server's acting as a\n client in another URI-scheme-dependent protocol. The server\n therefore becomes subject to many of the same security concerns that\n clients of the URI scheme are subject to, with the added concern that\n the client can attempt to prompt the server to connect to some\n (possibly weird-looking) URL.\n\n In general, this issue means that an attacker might use the server to\n indirectly attack another host that is vulnerable to some security\n flaw. It also introduces the possibility of denial-of-service\n attacks in which an attacker makes many connections to the server,\n each of which results in the server's attempting a connection to the\n target of the attack.\n\n Note that the server may be behind a firewall or otherwise able to\n access hosts that would not be directly accessible from the public\n Internet. This could exacerbate the potential security and denial-\n of-service problems described above, as well as allow the existence\n of internal hosts to be confirmed when they would otherwise be\n hidden.\n\n The detailed security concerns involved will depend on the URI\n schemes supported by the server. In the case of HTTP, the concerns\n are similar to those that apply to a publicly accessible HTTP proxy\n server. In the case of HTTPS, loops and deadlocks may be created,\n and this should be addressed. In the case of FTP, attacks arise that\n are similar to FTP bounce attacks.\n\n\n\n\n\n\nEastlake Standards Track [Page 20]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n As a result of this issue, it is RECOMMENDED that the\n client_certificate_url extension should have to be specifically\n enabled by a server administrator, rather than be enabled by default.\n It is also RECOMMENDED that URI schemes be enabled by the\n administrator individually, and only a minimal set of schemes be\n enabled. Unusual protocols that offer limited security or whose\n security is not well understood SHOULD be avoided.\n\n As discussed in [RFC3986], URLs that specify ports other than the\n default may cause problems, as may very long URLs (which are more\n likely to be useful in exploiting buffer overflow bugs).\n\n This extension continues to use SHA-1 (as in RFC 4366) and does not\n provide algorithm agility. The property required of SHA-1 in this\n case is second pre-image resistance, not collision resistance.\n Furthermore, even if second pre-image attacks against SHA-1 are found\n in the future, an attack against client_certificate_url would require\n a second pre-image that is accepted as a valid certificate by the\n server and contains the same public key.\n\n Also note that HTTP caching proxies are common on the Internet, and\n some proxies do not check for the latest version of an object\n correctly. If a request using HTTP (or another caching protocol)\n goes through a misconfigured or otherwise broken proxy, the proxy may\n return an out-of-date response.\n\n11.4. Security Considerations for trusted_ca_keys\n\n Potentially, the CA root keys a client possesses could be regarded as\n confidential information. As a result, the CA root key indication\n extension should be used with care.\n\n The use of the SHA-1 certificate hash alternative ensures that each\n certificate is specified unambiguously. This context does not\n require a cryptographic hash function, so the use of SHA-1 is\n considered acceptable, and no algorithm agility is provided.\n\n11.5. Security Considerations for truncated_hmac\n\n It is possible that truncated MACs are weaker than \"un-truncated\"\n MACs. However, no significant weaknesses are currently known or\n expected to exist for HMAC with MD5 or SHA-1, truncated to 80 bits.\n\n Note that the output length of a MAC need not be as long as the\n length of a symmetric cipher key, since forging of MAC values cannot\n be done off-line: in TLS, a single failed MAC guess will cause the\n immediate termination of the TLS session.\n\n\n\n\nEastlake Standards Track [Page 21]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Since the MAC algorithm only takes effect after all handshake\n messages that affect extension parameters have been authenticated by\n the hashes in the Finished messages, it is not possible for an active\n attacker to force negotiation of the truncated HMAC extension where\n it would not otherwise be used (to the extent that the handshake\n authentication is secure). Therefore, in the event that any security\n problems were found with truncated HMAC in the future, if either the\n client or the server for a given session were updated to take the\n problem into account, it would be able to veto use of this extension.\n\n11.6. Security Considerations for status_request\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC2560] for further details.\n\n12. Normative References\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC:\n Keyed-Hashing for Message Authentication\", RFC 2104,\n February 1997.\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and\n C. Adams, \"X.509 Internet Public Key Infrastructure\n Online Certificate Status Protocol - OCSP\", RFC 2560,\n June 1999.\n\n [RFC2585] Housley, R. and P. Hoffman, \"Internet X.509 Public Key\n Infrastructure Operational Protocols: FTP and HTTP\",\n RFC 2585, May 1999.\n\n [RFC2616] Fielding, R., Gettys, J., Mogul, J., Frystyk, H.,\n Masinter, L., Leach, P., and T. Berners-Lee,\n \"Hypertext Transfer Protocol -- HTTP/1.1\", RFC 2616,\n June 1999.\n\n [RFC3986] Berners-Lee, T., Fielding, R., and L. Masinter,\n \"Uniform Resource Identifier (URI): Generic Syntax\",\n STD 66, RFC 3986, January 2005.\n\n\n\nEastlake Standards Track [Page 22]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246, August\n 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation\n List (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC5890] Klensin, J., \"Internationalized Domain Names for\n Applications (IDNA): Definitions and Document\n Framework\", RFC 5890, August 2010.\n\n13. Informative References\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [X509-4th] ITU-T Recommendation X.509 (2000) | ISO/IEC\n 9594-8:2001, \"Information Systems - Open Systems\n Interconnection - The Directory: Public key and\n attribute certificate frameworks\".\n\n [X509-4th-TC1] ITU-T Recommendation X.509(2000) Corrigendum 1(2001) |\n ISO/IEC 9594-8:2001/Cor.1:2002, Technical Corrigendum\n 1 to ISO/IEC 9594:8:2001.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 23]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix A. Changes from RFC 4366\n\n The significant changes between RFC 4366 and this document are\n described below.\n\n RFC 4366 described both general extension mechanisms (for the TLS\n handshake and client and server hellos) as well as specific\n extensions. RFC 4366 was associated with RFC 4346, TLS 1.1. The\n client and server hello extension mechanisms have been moved into RFC\n 5246, TLS 1.2, so this document, which is associated with RFC 5246,\n includes only the handshake extension mechanisms and the specific\n extensions from RFC 4366. RFC 5246 also specifies the unknown\n extension error and new extension specification considerations, so\n that material has been removed from this document.\n\n The Server Name extension now specifies only ASCII representation,\n eliminating UTF-8. It is provided that the ServerNameList can\n contain more than only one name of any particular name_type. If a\n server name is provided but not recognized, the server should either\n continue the handshake without an error or send a fatal error.\n Sending a warning-level message is not recommended because client\n behavior will be unpredictable. Provision was added for the user\n using the server_name extension in deciding whether or not to resume\n a session. Furthermore, this extension should be the same in a\n session resumption request as it was in the full handshake that\n established the session. Such a resumption request must not be\n accepted if the server_name extension is different, but instead a\n full handshake must be done to possibly establish a new session.\n\n The Client Certificate URLs extension has been changed to make the\n presence of a hash mandatory.\n\n For the case of DTLS, the requirement to report an overflow of the\n negotiated maximum fragment length is made conditional on passing\n authentication.\n\n TLS servers are now prohibited from following HTTP redirects when\n retrieving certificates.\n\n The material was also re-organized in minor ways. For example,\n information as to which errors are fatal is moved from the \"Error\n Alerts\" section to the individual extension specifications.\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 24]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix B. Acknowledgements\n\n This document is based on material from RFC 4366 for which the\n authors were S. Blake-Wilson, M. Nystrom, D. Hopwood, J. Mikkelsen,\n and T. Wright. Other contributors include Joseph Salowey, Alexey\n Melnikov, Peter Saint-Andre, and Adrian Farrel.\n\nAuthor's Address\n\n Donald Eastlake 3rd\n Huawei\n 155 Beaver Street\n Milford, MA 01757 USA\n\n Phone: +1-508-333-2270\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 25]\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) Y. Pettersen\nRequest for Comments: 6961 June 2013\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\n\nAbstract\n\n This document defines the Transport Layer Security (TLS) Certificate\n Status Version 2 Extension to allow clients to specify and support\n several certificate status methods. (The use of the Certificate\n Status extension is commonly referred to as \"OCSP stapling\".) Also\n defined is a new method based on the Online Certificate Status\n Protocol (OCSP) that servers can use to provide status information\n about not only the server's own certificate but also the status of\n intermediate certificates in the chain.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6961.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 1]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nCopyright Notice\n\n Copyright (c) 2013 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n1. Introduction\n\n The Transport Layer Security (TLS) Extension [RFC6066] framework\n defines, among other extensions, the Certificate Status extension\n (also referred to as \"OCSP stapling\") that clients can use to request\n the server's copy of the current status of its certificate. The\n benefits of this extension include a reduced number of roundtrips and\n network delays for the client to verify the status of the server's\n certificate and a reduced load on the certificate issuer's status\n response servers, thus solving a problem that can become significant\n when the issued certificate is presented by a frequently visited\n server.\n\n There are two problems with the existing Certificate Status\n extension. First, it does not provide functionality to request the\n status information about intermediate Certification Authority (CA)\n certificates, which means the client has to request status\n information through other methods, such as Certificate Revocation\n Lists (CRLs), introducing further delays. Second, the current format\n of the extension and requirements in the TLS protocol prevent a\n client from offering the server multiple status methods.\n\n\n\nPettersen Standards Track [Page 2]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n Many CAs are now issuing intermediate CA certificates that not only\n specify the publication point for their CRLs in a CRL Distribution\n Point [RFC5280] but also specify a URL for their OCSP [RFC6960]\n server in Authority Information Access [RFC5280]. Given that\n client-cached CRLs are frequently out of date, clients would benefit\n from using OCSP to access up-to-date status information about\n intermediate CA certificates. The benefit to the issuing CA is less\n clear, as providing the bandwidth for the OCSP responder can be\n costly, especially for CAs with many high-traffic subscriber sites,\n and this cost is a concern for many CAs. There are cases where OCSP\n requests for a single high-traffic site caused significant network\n problems for the issuing CA.\n\n Clients will benefit from the TLS server providing certificate status\n information regardless of type, not just for the server certificate\n but also for the intermediate CA certificates. Combining the status\n checks into one extension will reduce the roundtrips needed to\n complete the handshake by the client to just those needed for\n negotiating the TLS connection. Also, for the Certification\n Authorities, the load on their servers will depend on the number of\n certificates they have issued, not on the number of visitors to those\n sites. Additionally, using this extension significantly reduces\n privacy concerns around the clients informing the certificate issuer\n about which sites they are visiting.\n\n For such a new system to be introduced seamlessly, clients need to be\n able to indicate support for the existing OCSP Certificate Status\n method and a new multiple-OCSP mode.\n\n Unfortunately, the definition of the Certificate Status extension\n only allows a single Certificate Status extension to be defined in a\n single extension record in the handshake, and the TLS protocol\n [RFC5246] only allows a single record in the extension list for any\n given extension. This means that it is not possible for clients to\n indicate support for new methods while still supporting older\n methods, which would cause problems for interoperability between\n newer clients and older servers. This will not just be an issue for\n the multiple status request mode proposed above but also for any\n other future status methods that might be introduced. This will be\n true not just for the current PKIX infrastructure [RFC5280] but also\n for alternative PKI structures.\n\n The solution to this problem is to define a new extension,\n \"status_request_v2\", with an extended format that allows the client\n to indicate support for multiple status request methods. This is\n implemented using a list of CertificateStatusRequestItemV2 records in\n the extension record. As the server will select the single status\n\n\n\n\nPettersen Standards Track [Page 3]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n method based on the selected cipher suite and the certificate\n presented, no significant changes are needed in the server's\n extension format.\n\n1.1. Requirements Language\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [RFC2119].\n\n1.2. Presentation Language\n\n This document defines protocol structures using the same conventions\n and presentation language as defined in Section 4 of [RFC5246].\n\n2. Multiple Certificate Status Extension\n\n2.1. New Extension\n\n The extension defined by this document is indicated by\n \"status_request_v2\" in the ExtensionType enum (originally defined by\n [RFC6066]), which uses the following value:\n\n enum {\n status_request_v2(17), (65535)\n } ExtensionType;\n\n2.2. Multiple Certificate Status Request Record\n\n Clients that support a certificate status protocol like OCSP may send\n the \"status_request_v2\" extension to the server in order to use the\n TLS handshake to transfer such data instead of downloading it through\n separate connections. When using this extension, the\n \"extension_data\" field (defined in Section 7.4.1.4 of [RFC5246]) of\n the extension SHALL contain a CertificateStatusRequestListV2 where:\n\n struct {\n CertificateStatusType status_type;\n uint16 request_length; /* Length of request field in bytes */\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n case ocsp_multi: OCSPStatusRequest;\n } request;\n } CertificateStatusRequestItemV2;\n\n enum { ocsp(1), ocsp_multi(2), (255) } CertificateStatusType;\n\n\n\n\n\nPettersen Standards Track [Page 4]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n struct {\n CertificateStatusRequestItemV2\n certificate_status_req_list<1..2^16-1>;\n } CertificateStatusRequestListV2;\n\n In the OCSPStatusRequest (originally defined by [RFC6066]), the\n \"ResponderID\" provides a list of OCSP responders that the client\n trusts. A zero-length \"responder_id_list\" sequence has the special\n meaning that the responders are implicitly known to the server, e.g.,\n by prior arrangement, or are identified by the certificates used by\n the server. \"Extensions\" is a DER encoding [X.690] of the OCSP\n request extensions, and if the server supports the forwarding of OCSP\n request extensions, this value MUST be forwarded without\n modification.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC6960]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a DER-encoded zero-length ASN.1 SEQUENCE,\n which is not valid for the \"Extensions\" type).\n\n Servers that support a client's selection of responders using the\n ResponderID field could implement this selection by matching the\n responder ID values from the client's list with the ResponderIDs of\n known OCSP responders, either by using a binary compare of the values\n or a hash calculation and compare method.\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement). This\n has been clarified in [RFC6960].\n\n The items in the list of CertificateStatusRequestItemV2 entries are\n ordered according to the client's preference (favorite choice first).\n\n A server that receives a client hello containing the\n \"status_request_v2\" extension MAY return a suitable certificate\n status response message to the client along with the server's\n\n\n\nPettersen Standards Track [Page 5]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n certificate message. If OCSP is requested, it SHOULD use the\n information contained in the extension when selecting an OCSP\n responder and SHOULD include request_extensions in the OCSP request.\n\n The server returns a certificate status response along with its\n certificate by sending a \"CertificateStatus\" message (originally\n defined by [RFC6066]) immediately after the \"Certificate\" message\n (Section 7.4.2 of [RFC5246]) (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message in response to a \"status_request_v2\"\n request, then the server MUST have included an extension of type\n \"status_request_v2\" with empty \"extension_data\" in the extended\n server hello.\n\n The \"CertificateStatus\" message is conveyed using the handshake\n message type \"certificate_status\" (defined in [RFC6066]) as follows\n (updated from the definition in [RFC6066]):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n case ocsp_multi: OCSPResponseList;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<0..2^24-1>;\n\n struct {\n OCSPResponse ocsp_response_list<1..2^24-1>;\n } OCSPResponseList;\n\n An \"OCSPResponse\" element (originally defined by [RFC6066]) contains\n a complete, DER-encoded OCSP response (using the ASN.1 [X.680] type\n OCSPResponse defined in [RFC6960]). Only one OCSP response, with a\n length of at least one byte, may be sent for status_type \"ocsp\".\n\n An \"ocsp_response_list\" contains a list of \"OCSPResponse\" elements,\n as specified above, each containing the OCSP response for the\n matching corresponding certificate in the server's Certificate TLS\n handshake message. That is, the first entry is the OCSP response for\n the first certificate in the Certificate list, the second entry is\n the response for the second certificate, and so on. The list MAY\n contain fewer OCSP responses than there were certificates in the\n Certificate handshake message, but there MUST NOT be more responses\n than there were certificates in the list. Individual elements of the\n list MAY have a length of 0 (zero) bytes if the server does not have\n the OCSP response for that particular certificate stored, in which\n\n\n\nPettersen Standards Track [Page 6]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n case the client MUST act as if a response was not received for that\n particular certificate. If the client receives a\n \"ocsp_response_list\" that does not contain a response for one or more\n of the certificates in the completed certificate chain, the client\n SHOULD attempt to validate the certificate using an alternative\n retrieval method, such as downloading the relevant CRL; OCSP SHOULD\n in this situation only be used for the end-entity certificate, not\n intermediate CA certificates, for reasons stated above.\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if it has received a \"status_request_v2\" extension in\n the client hello message and has sent a \"status_request_v2\" extension\n in the server hello message. Additionally, note that a server MUST\n NOT send the \"CertificateStatus\" message unless it received either a\n \"status_request\" or \"status_request_v2\" extension in the client hello\n message and sent a corresponding \"status_request\" or\n \"status_request_v2\" extension in the server hello message.\n\n Clients requesting an OCSP response and receiving one or more OCSP\n responses in a \"CertificateStatus\" message MUST check the OCSP\n response(s) and abort the handshake if the response is a \"revoked\"\n status or other unacceptable responses (as determined by client\n policy) with a bad_certificate_status_response(113) alert. This\n alert is always fatal.\n\n If the OCSP response received from the server does not result in a\n definite \"good\" or \"revoked\" status, it is inconclusive. A TLS\n client in such a case MAY check the validity of the server\n certificate through other means, e.g., by directly querying the\n certificate issuer. If such processing still results in an\n inconclusive response, then the application using the TLS connection\n will have to decide whether to close the connection or not. Note\n that this problem cannot be decided by the generic TLS client code\n without information from the application. If the application doesn't\n provide any such information, then the client MUST abort the\n connection, since the server certificate has not been sufficiently\n validated.\n\n An example of where the application might wish to continue is with\n EAP-TLS (Extensible Authentication Protocol - TLS), where the\n application can use another mechanism to check the status of a\n certificate once it obtains network access. In this case, the\n application could have the client continue with the handshake, but it\n MUST NOT disclose a username and password until it has fully\n validated the server certificate.\n\n\n\n\n\n\nPettersen Standards Track [Page 7]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n3. IANA Considerations\n\n Section 2.1 defines the new TLS extension status_request_v2 (17)\n enum, which has been added to the \"ExtensionType Values\" list in the\n IANA \"Transport Layer Security (TLS) Extensions\" registry.\n\n Section 2.2 describes a TLS CertificateStatusType registry that is\n now maintained by IANA. The \"TLS Certificate Status Types\" registry\n has been created under the \"Transport Layer Security (TLS)\n Extensions\" registry. CertificateStatusType values are to be\n assigned via IETF Review as defined in [RFC5226]. The initial\n registry corresponds to the definition of \"CertificateStatusType\" in\n Section 2.2.\n\n Value Description Reference\n -----------------------------------------\n 0 Reserved [RFC6961]\n 1 ocsp [RFC6066] [RFC6961]\n 2 ocsp_multi [RFC6961]\n 3-255 Unassigned\n\n4. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security considerations for the particular extension\n specified in this document are given below. In general, implementers\n should continue to monitor the state of the art and address any\n weaknesses identified.\n\n4.1. Security Considerations for status_request_v2\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC6960] for further details.\n\n This extension allows the client to send arbitrary data to the\n server. The server implementers need to handle such data carefully\n to avoid introducing security vulnerabilities.\n\n The security considerations of [RFC6960] apply to OCSP requests and\n responses.\n\n\n\n\nPettersen Standards Track [Page 8]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n5. Acknowledgements\n\n This document is based on [RFC6066], authored by Donald Eastlake 3rd.\n\n6. References\n\n6.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246, August 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC6066] Eastlake, D., \"Transport Layer Security (TLS) Extensions:\n Extension Definitions\", RFC 6066, January 2011.\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, June 2013.\n\n [X.680] ITU-T Recommendation X.680 (2008) | ISO/IEC 8824-1:2008,\n \"Abstract Syntax Notation One (ASN.1): Specification of\n basic notation\", November 2008.\n\n [X.690] ITU-T Recommendation X.690 (2008) | ISO/IEC 8825-1:2008,\n \"ASN.1 encoding rules: Specification of Basic Encoding\n Rules (BER), Canonical Encoding Rules (CER) and\n Distinguished Encoding Rules (DER)\", November 2008.\n\n6.2. Informative References\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and C.\n Adams, \"X.509 Internet Public Key Infrastructure Online\n Certificate Status Protocol - OCSP\", RFC 2560, June 1999.\n\n\n\n\n\n\nPettersen Standards Track [Page 9]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nAuthor's Address\n\n Yngve N. Pettersen\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 10]\n\n===== DOCUMENT =====\n\nNetwork Working Group E. Rescorla\nInternet-Draft RTFM, Inc.\nObsoletes: 5077, 5246, 6961 (if March 20, 2018\n approved)\nUpdates: 4492, 5705, 6066 (if approved)\nIntended status: Standards Track\nExpires: September 21, 2018\n\n\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 8446 Mozilla\nObsoletes: 5077, 5246, 6961 August 2018\nUpdates: 5705, 6066\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS) Protocol Version 1.3\n\nAbstract\n\n This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol. TLS allows client/server applications to communicate\n over the Internet in a way that is designed to prevent eavesdropping,\n tampering, and message forgery.\n\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077,\n 5246, and 6961. This document also specifies new requirements for\n TLS 1.2 implementations.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 7841.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n https://www.rfc-editor.org/info/rfc8446.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 8446 TLS August 2018\n\n\nCopyright Notice\n\n Copyright (c) 2018 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (https://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 8446 TLS August 2018\n\n\nTable of Contents\n\n 1. Introduction ....................................................6\n 1.1. Conventions and Terminology ................................7\n 1.2. Major Differences from TLS 1.2 .............................8\n 1.3. Updates Affecting TLS 1.2 ..................................9\n 2. Protocol Overview ..............................................10\n 2.1. Incorrect DHE Share .......................................14\n 2.2. Resumption and Pre-Shared Key (PSK) .......................15\n 2.3. 0-RTT Data ................................................17\n 3. Presentation Language ..........................................19\n 3.1. Basic Block Size ..........................................19\n 3.2. Miscellaneous .............................................20\n 3.3. Numbers ...................................................20\n 3.4. Vectors ...................................................20\n 3.5. Enumerateds ...............................................21\n 3.6. Constructed Types .........................................22\n 3.7. Constants .................................................23\n 3.8. Variants ..................................................23\n 4. Handshake Protocol .............................................24\n 4.1. Key Exchange Messages .....................................25\n 4.1.1. Cryptographic Negotiation ..........................26\n 4.1.2. Client Hello .......................................27\n 4.1.3. Server Hello .......................................31\n 4.1.4. Hello Retry Request ................................33\n 4.2. Extensions ................................................35\n 4.2.1. Supported Versions .................................39\n 4.2.2. Cookie .............................................40\n 4.2.3. Signature Algorithms ...............................41\n 4.2.4. Certificate Authorities ............................45\n 4.2.5. OID Filters ........................................45\n 4.2.6. Post-Handshake Client Authentication ...............47\n 4.2.7. Supported Groups ...................................47\n 4.2.8. Key Share ..........................................48\n 4.2.9. Pre-Shared Key Exchange Modes ......................51\n 4.2.10. Early Data Indication .............................52\n 4.2.11. Pre-Shared Key Extension ..........................55\n 4.3. Server Parameters .........................................59\n 4.3.1. Encrypted Extensions ...............................60\n 4.3.2. Certificate Request ................................60\n 4.4. Authentication Messages ...................................61\n 4.4.1. The Transcript Hash ................................63\n 4.4.2. Certificate ........................................64\n 4.4.3. Certificate Verify .................................69\n 4.4.4. Finished ...........................................71\n 4.5. End of Early Data .........................................72\n\n\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 8446 TLS August 2018\n\n\n 4.6. Post-Handshake Messages ...................................73\n 4.6.1. New Session Ticket Message .........................73\n 4.6.2. Post-Handshake Authentication ......................75\n 4.6.3. Key and Initialization Vector Update ...............76\n 5. Record Protocol ................................................77\n 5.1. Record Layer ..............................................78\n 5.2. Record Payload Protection .................................80\n 5.3. Per-Record Nonce ..........................................82\n 5.4. Record Padding ............................................83\n 5.5. Limits on Key Usage .......................................84\n 6. Alert Protocol .................................................85\n 6.1. Closure Alerts ............................................87\n 6.2. Error Alerts ..............................................88\n 7. Cryptographic Computations .....................................90\n 7.1. Key Schedule ..............................................91\n 7.2. Updating Traffic Secrets ..................................94\n 7.3. Traffic Key Calculation ...................................95\n 7.4. (EC)DHE Shared Secret Calculation .........................95\n 7.4.1. Finite Field Diffie-Hellman ........................95\n 7.4.2. Elliptic Curve Diffie-Hellman ......................96\n 7.5. Exporters .................................................97\n 8. 0-RTT and Anti-Replay ..........................................98\n 8.1. Single-Use Tickets ........................................99\n 8.2. Client Hello Recording ....................................99\n 8.3. Freshness Checks .........................................101\n 9. Compliance Requirements .......................................102\n 9.1. Mandatory-to-Implement Cipher Suites .....................102\n 9.2. Mandatory-to-Implement Extensions ........................103\n 9.3. Protocol Invariants ......................................104\n 10. Security Considerations ......................................106\n 11. IANA Considerations ..........................................106\n 12. References ...................................................109\n 12.1. Normative References ....................................109\n 12.2. Informative References ..................................112\n Appendix A. State Machine ........................................120\n A.1. Client ....................................................120\n A.2. Server ....................................................121\n Appendix B. Protocol Data Structures and Constant Values .........122\n B.1. Record Layer ..............................................122\n B.2. Alert Messages ............................................123\n B.3. Handshake Protocol ........................................124\n B.3.1. Key Exchange Messages .................................125\n B.3.2. Server Parameters Messages ............................131\n B.3.3. Authentication Messages ...............................132\n B.3.4. Ticket Establishment ..................................132\n B.3.5. Updating Keys .........................................133\n B.4. Cipher Suites .............................................133\n\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 8446 TLS August 2018\n\n\n Appendix C. Implementation Notes .................................134\n C.1. Random Number Generation and Seeding ......................134\n C.2. Certificates and Authentication ...........................135\n C.3. Implementation Pitfalls ...................................135\n C.4. Client Tracking Prevention ................................137\n C.5. Unauthenticated Operation .................................137\n Appendix D. Backward Compatibility ...............................138\n D.1. Negotiating with an Older Server ..........................139\n D.2. Negotiating with an Older Client ..........................139\n D.3. 0-RTT Backward Compatibility ..............................140\n D.4. Middlebox Compatibility Mode ..............................140\n D.5. Security Restrictions Related to Backward Compatibility ...141\n Appendix E. Overview of Security Properties ......................142\n E.1. Handshake .................................................142\n E.1.1. Key Derivation and HKDF ...............................145\n E.1.2. Client Authentication .................................146\n E.1.3. 0-RTT .................................................146\n E.1.4. Exporter Independence .................................146\n E.1.5. Post-Compromise Security ..............................146\n E.1.6. External References ...................................147\n E.2. Record Layer ..............................................147\n E.2.1. External References ...................................148\n E.3. Traffic Analysis ..........................................148\n E.4. Side-Channel Attacks ......................................149\n E.5. Replay Attacks on 0-RTT ...................................150\n E.5.1. Replay and Exporters ..................................151\n E.6. PSK Identity Exposure .....................................152\n E.7. Sharing PSKs ..............................................152\n E.8. Attacks on Static RSA .....................................152\n Contributors .....................................................153\n Author's Address .................................................160\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 8446 TLS August 2018\n\n\n1. Introduction\n\n The primary goal of TLS is to provide a secure channel between two\n communicating peers; the only requirement from the underlying\n transport is a reliable, in-order data stream. Specifically, the\n secure channel should provide the following properties:\n\n - Authentication: The server side of the channel is always\n authenticated; the client side is optionally authenticated.\n Authentication can happen via asymmetric cryptography (e.g., RSA\n [RSA], the Elliptic Curve Digital Signature Algorithm (ECDSA)\n [ECDSA], or the Edwards-Curve Digital Signature Algorithm (EdDSA)\n [RFC8032]) or a symmetric pre-shared key (PSK).\n\n - Confidentiality: Data sent over the channel after establishment is\n only visible to the endpoints. TLS does not hide the length of\n the data it transmits, though endpoints are able to pad TLS\n records in order to obscure lengths and improve protection against\n traffic analysis techniques.\n\n - Integrity: Data sent over the channel after establishment cannot\n be modified by attackers without detection.\n\n These properties should be true even in the face of an attacker who\n has complete control of the network, as described in [RFC3552]. See\n Appendix E for a more complete statement of the relevant security\n properties.\n\n TLS consists of two primary components:\n\n - A handshake protocol (Section 4) that authenticates the\n communicating parties, negotiates cryptographic modes and\n parameters, and establishes shared keying material. The handshake\n protocol is designed to resist tampering; an active attacker\n should not be able to force the peers to negotiate different\n parameters than they would if the connection were not under\n attack.\n\n - A record protocol (Section 5) that uses the parameters established\n by the handshake protocol to protect traffic between the\n communicating peers. The record protocol divides traffic up into\n a series of records, each of which is independently protected\n using the traffic keys.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 8446 TLS August 2018\n\n\n TLS is application protocol independent; higher-level protocols can\n layer on top of TLS transparently. The TLS standard, however, does\n not specify how protocols add security with TLS; how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n This document defines TLS version 1.3. While TLS 1.3 is not directly\n compatible with previous versions, all versions of TLS incorporate a\n versioning mechanism which allows clients and servers to\n interoperably negotiate a common version if one is supported by both\n peers.\n\n This document supersedes and obsoletes previous versions of TLS,\n including version 1.2 [RFC5246]. It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2. Because TLS 1.3 changes the way keys are\n derived, it updates [RFC5705] as described in Section 7.5. It also\n changes how Online Certificate Status Protocol (OCSP) messages are\n carried and therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in Section 4.4.2.1.\n\n1.1. Conventions and Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all\n capitals, as shown here.\n\n The following terms are used:\n\n client: The endpoint initiating the TLS connection.\n\n connection: A transport-layer connection between two endpoints.\n\n endpoint: Either the client or server of the connection.\n\n handshake: An initial negotiation between client and server that\n establishes the parameters of their subsequent interactions\n within TLS.\n\n peer: An endpoint. When discussing a particular endpoint, \"peer\"\n refers to the endpoint that is not the primary subject of\n discussion.\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\f\nRFC 8446 TLS August 2018\n\n\n receiver: An endpoint that is receiving records.\n\n sender: An endpoint that is transmitting records.\n\n server: The endpoint that did not initiate the TLS connection.\n\n1.2. Major Differences from TLS 1.2\n\n The following is a list of the major functional differences between\n TLS 1.2 and TLS 1.3. It is not intended to be exhaustive, and there\n are many minor differences.\n\n - The list of supported symmetric encryption algorithms has been\n pruned of all algorithms that are considered legacy. Those that\n remain are all Authenticated Encryption with Associated Data\n (AEAD) algorithms. The cipher suite concept has been changed to\n separate the authentication and key exchange mechanisms from the\n record protection algorithm (including secret key length) and a\n hash to be used with both the key derivation function and\n handshake message authentication code (MAC).\n\n - A zero round-trip time (0-RTT) mode was added, saving a round trip\n at connection setup for some application data, at the cost of\n certain security properties.\n\n - Static RSA and Diffie-Hellman cipher suites have been removed; all\n public-key based key exchange mechanisms now provide forward\n secrecy.\n\n - All handshake messages after the ServerHello are now encrypted.\n The newly introduced EncryptedExtensions message allows various\n extensions previously sent in the clear in the ServerHello to also\n enjoy confidentiality protection.\n\n - The key derivation functions have been redesigned. The new design\n allows easier analysis by cryptographers due to their improved key\n separation properties. The HMAC-based Extract-and-Expand Key\n Derivation Function (HKDF) is used as an underlying primitive.\n\n - The handshake state machine has been significantly restructured to\n be more consistent and to remove superfluous messages such as\n ChangeCipherSpec (except when needed for middlebox compatibility).\n\n - Elliptic curve algorithms are now in the base spec, and new\n signature algorithms, such as EdDSA, are included. TLS 1.3\n removed point format negotiation in favor of a single point format\n for each curve.\n\n\n\n\nRescorla Standards Track [Page 8]\n\f\nRFC 8446 TLS August 2018\n\n\n - Other cryptographic improvements were made, including changing the\n RSA padding to use the RSA Probabilistic Signature Scheme\n (RSASSA-PSS), and the removal of compression, the Digital\n Signature Algorithm (DSA), and custom Ephemeral Diffie-Hellman\n (DHE) groups.\n\n - The TLS 1.2 version negotiation mechanism has been deprecated in\n favor of a version list in an extension. This increases\n compatibility with existing servers that incorrectly implemented\n version negotiation.\n\n - Session resumption with and without server-side state as well as\n the PSK-based cipher suites of earlier TLS versions have been\n replaced by a single new PSK exchange.\n\n - References have been updated to point to the updated versions of\n RFCs, as appropriate (e.g., RFC 5280 rather than RFC 3280).\n\n1.3. Updates Affecting TLS 1.2\n\n This document defines several changes that optionally affect\n implementations of TLS 1.2, including those which do not also support\n TLS 1.3:\n\n - A version downgrade protection mechanism is described in\n Section 4.1.3.\n\n - RSASSA-PSS signature schemes are defined in Section 4.2.3.\n\n - The \"supported_versions\" ClientHello extension can be used to\n negotiate the version of TLS to use, in preference to the\n legacy_version field of the ClientHello.\n\n - The \"signature_algorithms_cert\" extension allows a client to\n indicate which signature algorithms it can validate in X.509\n certificates.\n\n Additionally, this document clarifies some compliance requirements\n for earlier versions of TLS; see Section 9.3.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 9]\n\f\nRFC 8446 TLS August 2018\n\n\n2. Protocol Overview\n\n The cryptographic parameters used by the secure channel are produced\n by the TLS handshake protocol. This sub-protocol of TLS is used by\n the client and server when first communicating with each other. The\n handshake protocol allows peers to negotiate a protocol version,\n select cryptographic algorithms, optionally authenticate each other,\n and establish shared secret keying material. Once the handshake is\n complete, the peers use the established keys to protect the\n application-layer traffic.\n\n A failure of the handshake or other protocol error triggers the\n termination of the connection, optionally preceded by an alert\n message (Section 6).\n\n TLS supports three basic key exchange modes:\n\n - (EC)DHE (Diffie-Hellman over either finite fields or elliptic\n curves)\n\n - PSK-only\n\n - PSK with (EC)DHE\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 10]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 1 below shows the basic full TLS handshake:\n\n Client Server\n\nKey ^ ClientHello\nExch | + key_share*\n | + signature_algorithms*\n | + psk_key_exchange_modes*\n v + pre_shared_key* -------->\n ServerHello ^ Key\n + key_share* | Exch\n + pre_shared_key* v\n {EncryptedExtensions} ^ Server\n {CertificateRequest*} v Params\n {Certificate*} ^\n {CertificateVerify*} | Auth\n {Finished} v\n <-------- [Application Data*]\n ^ {Certificate*}\nAuth | {CertificateVerify*}\n v {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 1: Message Flow for Full TLS Handshake\n\n The handshake can be thought of as having three phases (indicated in\n the diagram above):\n\n - Key Exchange: Establish shared keying material and select the\n cryptographic parameters. Everything after this phase is\n encrypted.\n\n - Server Parameters: Establish other handshake parameters\n (whether the client is authenticated, application-layer protocol\n support, etc.).\n\n\n\n\nRescorla Standards Track [Page 11]\n\f\nRFC 8446 TLS August 2018\n\n\n - Authentication: Authenticate the server (and, optionally, the\n client) and provide key confirmation and handshake integrity.\n\n In the Key Exchange phase, the client sends the ClientHello\n (Section 4.1.2) message, which contains a random nonce\n (ClientHello.random); its offered protocol versions; a list of\n symmetric cipher/HKDF hash pairs; either a set of Diffie-Hellman key\n shares (in the \"key_share\" (Section 4.2.8) extension), a set of\n pre-shared key labels (in the \"pre_shared_key\" (Section 4.2.11)\n extension), or both; and potentially additional extensions.\n Additional fields and/or messages may also be present for middlebox\n compatibility.\n\n The server processes the ClientHello and determines the appropriate\n cryptographic parameters for the connection. It then responds with\n its own ServerHello (Section 4.1.3), which indicates the negotiated\n connection parameters. The combination of the ClientHello and the\n ServerHello determines the shared keys. If (EC)DHE key establishment\n is in use, then the ServerHello contains a \"key_share\" extension with\n the server's ephemeral Diffie-Hellman share; the server's share MUST\n be in the same group as one of the client's shares. If PSK key\n establishment is in use, then the ServerHello contains a\n \"pre_shared_key\" extension indicating which of the client's offered\n PSKs was selected. Note that implementations can use (EC)DHE and PSK\n together, in which case both extensions will be supplied.\n\n The server then sends two messages to establish the Server\n Parameters:\n\n EncryptedExtensions: responses to ClientHello extensions that are\n not required to determine the cryptographic parameters, other than\n those that are specific to individual certificates.\n [Section 4.3.1]\n\n CertificateRequest: if certificate-based client authentication is\n desired, the desired parameters for that certificate. This\n message is omitted if client authentication is not desired.\n [Section 4.3.2]\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 12]\n\f\nRFC 8446 TLS August 2018\n\n\n Finally, the client and server exchange Authentication messages. TLS\n uses the same set of messages every time that certificate-based\n authentication is needed. (PSK-based authentication happens as a\n side effect of key exchange.) Specifically:\n\n Certificate: The certificate of the endpoint and any per-certificate\n extensions. This message is omitted by the server if not\n authenticating with a certificate and by the client if the server\n did not send CertificateRequest (thus indicating that the client\n should not authenticate with a certificate). Note that if raw\n public keys [RFC7250] or the cached information extension\n [RFC7924] are in use, then this message will not contain a\n certificate but rather some other value corresponding to the\n server's long-term key. [Section 4.4.2]\n\n CertificateVerify: A signature over the entire handshake using the\n private key corresponding to the public key in the Certificate\n message. This message is omitted if the endpoint is not\n authenticating via a certificate. [Section 4.4.3]\n\n Finished: A MAC (Message Authentication Code) over the entire\n handshake. This message provides key confirmation, binds the\n endpoint's identity to the exchanged keys, and in PSK mode also\n authenticates the handshake. [Section 4.4.4]\n\n Upon receiving the server's messages, the client responds with its\n Authentication messages, namely Certificate and CertificateVerify (if\n requested), and Finished.\n\n At this point, the handshake is complete, and the client and server\n derive the keying material required by the record layer to exchange\n application-layer data protected through authenticated encryption.\n Application Data MUST NOT be sent prior to sending the Finished\n message, except as specified in Section 2.3. Note that while the\n server may send Application Data prior to receiving the client's\n Authentication messages, any data sent at that point is, of course,\n being sent to an unauthenticated peer.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 13]\n\f\nRFC 8446 TLS August 2018\n\n\n2.1. Incorrect DHE Share\n\n If the client has not provided a sufficient \"key_share\" extension\n (e.g., it includes only DHE or ECDHE groups unacceptable to or\n unsupported by the server), the server corrects the mismatch with a\n HelloRetryRequest and the client needs to restart the handshake with\n an appropriate \"key_share\" extension, as shown in Figure 2. If no\n common cryptographic parameters can be negotiated, the server MUST\n abort the handshake with an appropriate alert.\n\n Client Server\n\n ClientHello\n + key_share -------->\n HelloRetryRequest\n <-------- + key_share\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 2: Message Flow for a Full Handshake with\n Mismatched Parameters\n\n Note: The handshake transcript incorporates the initial\n ClientHello/HelloRetryRequest exchange; it is not reset with the\n new ClientHello.\n\n TLS also allows several optimized variants of the basic handshake, as\n described in the following sections.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 14]\n\f\nRFC 8446 TLS August 2018\n\n\n2.2. Resumption and Pre-Shared Key (PSK)\n\n Although TLS PSKs can be established out of band, PSKs can also be\n established in a previous connection and then used to establish a new\n connection (\"session resumption\" or \"resuming\" with a PSK). Once a\n handshake has completed, the server can send the client a PSK\n identity that corresponds to a unique key derived from the initial\n handshake (see Section 4.6.1). The client can then use that PSK\n identity in future handshakes to negotiate the use of the associated\n PSK. If the server accepts the PSK, then the security context of the\n new connection is cryptographically tied to the original connection\n and the key derived from the initial handshake is used to bootstrap\n the cryptographic state instead of a full handshake. In TLS 1.2 and\n below, this functionality was provided by \"session IDs\" and \"session\n tickets\" [RFC5077]. Both mechanisms are obsoleted in TLS 1.3.\n\n PSKs can be used with (EC)DHE key exchange in order to provide\n forward secrecy in combination with shared keys, or can be used\n alone, at the cost of losing forward secrecy for the application\n data.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 15]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 3 shows a pair of handshakes in which the first handshake\n establishes a PSK and the second handshake uses it:\n\n Client Server\n\n Initial Handshake:\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n <-------- [NewSessionTicket]\n [Application Data] <-------> [Application Data]\n\n\n Subsequent Handshake:\n ClientHello\n + key_share*\n + pre_shared_key -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n {Finished}\n <-------- [Application Data*]\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 3: Message Flow for Resumption and PSK\n\n As the server is authenticating via a PSK, it does not send a\n Certificate or a CertificateVerify message. When a client offers\n resumption via a PSK, it SHOULD also supply a \"key_share\" extension\n to the server to allow the server to decline resumption and fall back\n to a full handshake, if needed. The server responds with a\n \"pre_shared_key\" extension to negotiate the use of PSK key\n establishment and can (as shown here) respond with a \"key_share\"\n extension to do (EC)DHE key establishment, thus providing forward\n secrecy.\n\n\n\n\nRescorla Standards Track [Page 16]\n\f\nRFC 8446 TLS August 2018\n\n\n When PSKs are provisioned out of band, the PSK identity and the KDF\n hash algorithm to be used with the PSK MUST also be provisioned.\n\n Note: When using an out-of-band provisioned pre-shared secret, a\n critical consideration is using sufficient entropy during the key\n generation, as discussed in [RFC4086]. Deriving a shared secret\n from a password or other low-entropy sources is not secure. A\n low-entropy secret, or password, is subject to dictionary attacks\n based on the PSK binder. The specified PSK authentication is not\n a strong password-based authenticated key exchange even when used\n with Diffie-Hellman key establishment. Specifically, it does not\n prevent an attacker that can observe the handshake from performing\n a brute-force attack on the password/pre-shared key.\n\n2.3. 0-RTT Data\n\n When clients and servers share a PSK (either obtained externally or\n via a previous handshake), TLS 1.3 allows clients to send data on the\n first flight (\"early data\"). The client uses the PSK to authenticate\n the server and to encrypt the early data.\n\n As shown in Figure 4, the 0-RTT data is just added to the 1-RTT\n handshake in the first flight. The rest of the handshake uses the\n same messages as for a 1-RTT handshake with PSK resumption.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 17]\n\f\nRFC 8446 TLS August 2018\n\n\n Client Server\n\n ClientHello\n + early_data\n + key_share*\n + psk_key_exchange_modes\n + pre_shared_key\n (Application Data*) -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n + early_data*\n {Finished}\n <-------- [Application Data*]\n (EndOfEarlyData)\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n () Indicates messages protected using keys\n derived from a client_early_traffic_secret.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 4: Message Flow for a 0-RTT Handshake\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 18]\n\f\nRFC 8446 TLS August 2018\n\n\n IMPORTANT NOTE: The security properties for 0-RTT data are weaker\n than those for other kinds of TLS data. Specifically:\n\n 1. This data is not forward secret, as it is encrypted solely under\n keys derived using the offered PSK.\n\n 2. There are no guarantees of non-replay between connections.\n Protection against replay for ordinary TLS 1.3 1-RTT data is\n provided via the server's Random value, but 0-RTT data does not\n depend on the ServerHello and therefore has weaker guarantees.\n This is especially relevant if the data is authenticated either\n with TLS client authentication or inside the application\n protocol. The same warnings apply to any use of the\n early_exporter_master_secret.\n\n\n\n===== DOCUMENT =====\n\n\nRFC 8446 TLS August 2018\n\n\n Sam Scott\n Royal Holloway, University of London\n [redacted-email]\n\n Thomas Shrimpton\n University of Florida\n [redacted-email]\n\n Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n Brian Smith\n Independent\n [redacted-email]\n\n Brian Sniffen\n Akamai Technologies\n [redacted-email]\n\n Nick Sullivan\n Cloudflare Inc.\n [redacted-email]\n\n Bjoern Tackmann\n University of California, San Diego\n [redacted-email]\n\n Tim Taubert\n Mozilla\n [redacted-email]\n\n Martin Thomson\n Mozilla\n [redacted-email]\n\n Hannes Tschofenig\n Arm Limited\n [redacted-email]\n\n Sean Turner\n sn3rd\n [redacted-email]\n\n Steven Valdez\n Google\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 159]\n\f\nRFC 8446 TLS August 2018\n\n\n Filippo Valsorda\n Cloudflare Inc.\n [redacted-email]\n\n Thyla van der Merwe\n Royal Holloway, University of London\n [redacted-email]\n\n Victor Vasiliev\n Google\n [redacted-email]\n\n Hoeteck Wee\n Ecole Normale Superieure, Paris\n [redacted-email]\n\n Tom Weinstein\n\n David Wong\n NCC Group\n [redacted-email]\n\n Christopher A. Wood\n Apple Inc.\n [redacted-email]\n\n Tim Wright\n Vodafone\n [redacted-email]\n\n Peter Wu\n Independent\n [redacted-email]\n\n Kazu Yamamoto\n Internet Initiative Japan Inc.\n [redacted-email]\n\nAuthor's Address\n\n Eric Rescorla\n Mozilla\n\n Email: [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 160]\n\nAnswer using only the documents and the scenario in the question. Return exactly the requested JSON object; use \"UNKNOWN\" only per unresolved field.", "context_source_relation_count": 22, "counterfactual_twin": {"byte_end": 447, "byte_start": 361, "char_end": 447, "char_start": 361, "evidence_id": "current_protocol", "parent_source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd", "parent_text_sha256": "d596886622bceee8500b3f3c9d869d947b4e70a2f05ea65fdd75666b46debd81", "parent_value": "This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol.", "provenance_operation": "exclude_exact_source_span", "record_id": "ietf:rfc:8446", "source_manifest_sha256": "287ca4031b223ad818d21781c575905d5ed30eb7e4e0e38fef196cf1ca1865b4", "source_origin": "synthetic_counterfactual", "text_sha256": "98ef2decaa389708a963efbd12018cad07c13f824ea89269ccae34e25ced38b7", "value": " "}, "data_stage": "train_ready", "dependency_class": "long_range_retrieval", "diagnostic_only": true, "difficulty": {"context_tokens": 65513, "evidence_span_tokens": 64670, "max_evidence_distance": 64677, "proof_depth": 2, "state_updates": 10}, "document_context": "Network Working Group J. Salowey\nRequest for Comments: 5077 H. Zhou\nObsoletes: 4507 Cisco Systems\nCategory: Standards Track P. Eronen\n Nokia\n H. Tschofenig\n Nokia Siemens Networks\n January 2008\n\n\n Transport Layer Security (TLS) Session Resumption without\n Server-Side State\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document describes a mechanism that enables the Transport Layer\n Security (TLS) server to resume sessions and avoid keeping per-client\n session state. The TLS server encapsulates the session state into a\n ticket and forwards it to the client. The client can subsequently\n resume a session using the obtained ticket. This document obsoletes\n RFC 4507.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 1]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3. Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . 4\n 3.2. SessionTicket TLS Extension . . . . . . . . . . . . . . . 7\n 3.3. NewSessionTicket Handshake Message . . . . . . . . . . . . 8\n 3.4. Interaction with TLS Session ID . . . . . . . . . . . . . 9\n 4. Recommended Ticket Construction . . . . . . . . . . . . . . . 10\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . 12\n 5.1. Invalidating Sessions . . . . . . . . . . . . . . . . . . 12\n 5.2. Stolen Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.3. Forged Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.4. Denial of Service Attacks . . . . . . . . . . . . . . . . 12\n 5.5. Ticket Protection Key Management . . . . . . . . . . . . . 13\n 5.6. Ticket Lifetime . . . . . . . . . . . . . . . . . . . . . 13\n 5.7. Alternate Ticket Formats and Distribution Schemes . . . . 13\n 5.8. Identity Privacy, Anonymity, and Unlinkability . . . . . . 14\n 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 14\n 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 15\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 15\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 15\n Appendix A. Discussion of Changes to RFC 4507 . . . . . . . . . . 17\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 2]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n1. Introduction\n\n This document defines a way to resume a Transport Layer Security\n (TLS) session without requiring session-specific state at the TLS\n server. This mechanism may be used with any TLS ciphersuite. This\n document applies to both TLS 1.0 defined in [RFC2246], and TLS 1.1\n defined in [RFC4346]. The mechanism makes use of TLS extensions\n defined in [RFC4366] and defines a new TLS message type.\n\n This mechanism is useful in the following situations:\n\n 1. servers that handle a large number of transactions from different\n users\n\n 2. servers that desire to cache sessions for a long time\n\n 3. ability to load balance requests across servers\n\n 4. embedded servers with little memory\n\n This document obsoletes RFC 4507 [RFC4507] to correct an error in the\n encoding that caused the specification to differ from deployed\n implementations. At the time of this writing, there are no known\n implementations that follow the encoding specified in RFC 4507. This\n update to RFC 4507 aligns the document with currently deployed\n implementations. More details of the change are given in Appendix A.\n\n2. Terminology\n\n Within this document, the term 'ticket' refers to a cryptographically\n protected data structure that is created and consumed by the server\n to rebuild session-specific state.\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Protocol\n\n This specification describes a mechanism to distribute encrypted\n session-state information to the client in the form of a ticket and a\n mechanism to present the ticket back to the server. The ticket is\n created by a TLS server and sent to a TLS client. The TLS client\n presents the ticket to the TLS server to resume a session.\n Implementations of this specification are expected to support both\n mechanisms. Other specifications can take advantage of the session\n tickets, perhaps specifying alternative means for distribution or\n selection. For example, a separate specification may describe an\n\n\n\nSalowey, et al. Standards Track [Page 3]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n alternate way to distribute a ticket and use the TLS extension in\n this document to resume the session. This behavior is beyond the\n scope of the document and would need to be described in a separate\n specification.\n\n3.1. Overview\n\n The client indicates that it supports this mechanism by including a\n SessionTicket TLS extension in the ClientHello message. The\n extension will be empty if the client does not already possess a\n ticket for the server. The server sends an empty SessionTicket\n extension to indicate that it will send a new session ticket using\n the NewSessionTicket handshake message. The extension is described\n in Section 3.2.\n\n If the server wants to use this mechanism, it stores its session\n state (such as ciphersuite and master secret) to a ticket that is\n encrypted and integrity-protected by a key known only to the server.\n The ticket is distributed to the client using the NewSessionTicket\n TLS handshake message described in Section 3.3. This message is sent\n during the TLS handshake before the ChangeCipherSpec message, after\n the server has successfully verified the client's Finished message.\n\n Client Server\n\n ClientHello\n (empty SessionTicket extension)-------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1: Message Flow for Full Handshake Issuing New Session Ticket\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 4]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n The client caches this ticket along with the master secret and other\n parameters associated with the current session. When the client\n wishes to resume the session, it includes the ticket in the\n SessionTicket extension within the ClientHello message. Appendix A\n provides a detailed description of the encoding of the extension and\n changes from RFC 4507. The server then decrypts the received ticket,\n verifies the ticket's validity, retrieves the session state from the\n contents of the ticket, and uses this state to resume the session.\n The interaction with the TLS Session ID is described in Section 3.4.\n If the server successfully verifies the client's ticket, then it may\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello.\n\n Client Server\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2: Message Flow for Abbreviated Handshake Using New Session\n Ticket\n\n A recommended ticket format is given in Section 4.\n\n If the server cannot or does not want to honor the ticket, then it\n can initiate a full handshake with the client.\n\n In the case that the server does not wish to issue a new ticket at\n this time, it just completes the handshake without including a\n SessionTicket extension or NewSessionTicket handshake message. This\n is shown below (this flow is identical to Figure 1 in RFC 4346,\n except for the SessionTicket extension in the first message):\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 5]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 3: Message Flow for Server Completing Full Handshake Without\n Issuing New Session Ticket\n\n It is also permissible to have an exchange similar to Figure 3 using\n the abbreviated handshake defined in Figure 2 of RFC 4346, where the\n client uses the SessionTicket extension to resume the session, but\n the server does not wish to issue a new ticket, and therefore does\n not send a SessionTicket extension.\n\n If the server rejects the ticket, it may still wish to issue a new\n ticket after performing the full handshake as shown below (this flow\n is identical to Figure 1, except the SessionTicket extension in the\n ClientHello is not empty):\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 6]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 4: Message Flow for Server Rejecting Ticket, Performing Full\n Handshake, and Issuing New Session Ticket\n\n3.2. SessionTicket TLS Extension\n\n The SessionTicket TLS extension is based on [RFC4366]. The format of\n the ticket is an opaque structure used to carry session-specific\n state information. This extension may be sent in the ClientHello and\n ServerHello.\n\n If the client possesses a ticket that it wants to use to resume a\n session, then it includes the ticket in the SessionTicket extension\n in the ClientHello. If the client does not have a ticket and is\n prepared to receive one in the NewSessionTicket handshake message,\n then it MUST include a zero-length ticket in the SessionTicket\n extension. If the client is not prepared to receive a ticket in the\n NewSessionTicket handshake message, then it MUST NOT include a\n SessionTicket extension unless it is sending a non-empty ticket it\n received through some other means from the server.\n\n The server uses a zero-length SessionTicket extension to indicate to\n the client that it will send a new session ticket using the\n NewSessionTicket handshake message described in Section 3.3. The\n server MUST send this extension in the ServerHello if it wishes to\n issue a new ticket to the client using the NewSessionTicket handshake\n message. The server MUST NOT send this extension if it does not\n receive one in the ClientHello.\n\n\n\n\nSalowey, et al. Standards Track [Page 7]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n If the server fails to verify the ticket, then it falls back to\n performing a full handshake. If the ticket is accepted by the server\n but the handshake fails, the client SHOULD delete the ticket.\n\n The SessionTicket extension has been assigned the number 35. The\n extension_data field of SessionTicket extension contains the ticket.\n\n3.3. NewSessionTicket Handshake Message\n\n This message is sent by the server during the TLS handshake before\n the ChangeCipherSpec message. This message MUST be sent if the\n server included a SessionTicket extension in the ServerHello. This\n message MUST NOT be sent if the server did not include a\n SessionTicket extension in the ServerHello. This message is included\n in the hash used to create and verify the Finished message. In the\n case of a full handshake, the server MUST verify the client's\n Finished message before sending the ticket. The client MUST NOT\n treat the ticket as valid until it has verified the server's Finished\n message. If the server determines that it does not want to include a\n ticket after it has included the SessionTicket extension in the\n ServerHello, then it sends a zero-length ticket in the\n NewSessionTicket handshake message.\n\n If the server successfully verifies the client's ticket, then it MAY\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello in the abbreviated handshake. The client\n should start using the new ticket as soon as possible after it\n verifies the server's Finished message for new connections. Note\n that since the updated ticket is issued before the handshake\n completes, it is possible that the client may not put the new ticket\n into use before it initiates new connections. The server MUST NOT\n assume that the client actually received the updated ticket until it\n successfully verifies the client's Finished message.\n\n The NewSessionTicket handshake message has been assigned the number 4\n and its definition is given at the end of this section. The\n ticket_lifetime_hint field contains a hint from the server about how\n long the ticket should be stored. The value indicates the lifetime\n in seconds as a 32-bit unsigned integer in network byte order\n relative to when the ticket is received. A value of zero is reserved\n to indicate that the lifetime of the ticket is unspecified. A client\n SHOULD delete the ticket and associated state when the time expires.\n It MAY delete the ticket earlier based on local policy. A server MAY\n treat a ticket as valid for a shorter or longer period of time than\n what is stated in the ticket_lifetime_hint.\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 8]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n HandshakeType msg_type;\n uint24 length;\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case session_ticket: NewSessionTicket; /* NEW */\n } body;\n } Handshake;\n\n\n struct {\n uint32 ticket_lifetime_hint;\n opaque ticket<0..2^16-1>;\n } NewSessionTicket;\n\n3.4. Interaction with TLS Session ID\n\n If a server is planning on issuing a session ticket to a client that\n does not present one, it SHOULD include an empty Session ID in the\n ServerHello. If the server rejects the ticket and falls back to the\n full handshake then it may include a non-empty Session ID to indicate\n its support for stateful session resumption. If the client receives\n a session ticket from the server, then it discards any Session ID\n that was sent in the ServerHello.\n\n When presenting a ticket, the client MAY generate and include a\n Session ID in the TLS ClientHello. If the server accepts the ticket\n and the Session ID is not empty, then it MUST respond with the same\n Session ID present in the ClientHello. This allows the client to\n easily differentiate when the server is resuming a session from when\n it is falling back to a full handshake. Since the client generates a\n Session ID, the server MUST NOT rely upon the Session ID having a\n particular value when validating the ticket. If a ticket is\n presented by the client, the server MUST NOT attempt to use the\n Session ID in the ClientHello for stateful session resumption.\n Alternatively, the client MAY include an empty Session ID in the\n ClientHello. In this case, the client ignores the Session ID sent in\n the ServerHello and determines if the server is resuming a session by\n the subsequent handshake messages.\n\n\n\nSalowey, et al. Standards Track [Page 9]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n4. Recommended Ticket Construction\n\n This section describes a recommended format and protection for the\n ticket. Note that the ticket is opaque to the client, so the\n structure is not subject to interoperability concerns, and\n implementations may diverge from this format. If implementations do\n diverge from this format, they must take security concerns seriously.\n Clients MUST NOT examine the ticket under the assumption that it\n complies with this document.\n\n The server uses two different keys: one 128-bit key for Advanced\n Encryption Standard (AES) [AES] in Cipher Block Chaining (CBC) mode\n [CBC] encryption and one 256-bit key for HMAC-SHA-256 [RFC4634].\n\n The ticket is structured as follows:\n\n struct {\n opaque key_name[16];\n opaque iv[16];\n opaque encrypted_state<0..2^16-1>;\n opaque mac[32];\n } ticket;\n\n Here, key_name serves to identify a particular set of keys used to\n protect the ticket. It enables the server to easily recognize\n tickets it has issued. The key_name should be randomly generated to\n avoid collisions between servers. One possibility is to generate new\n random keys and key_name every time the server is started.\n\n The actual state information in encrypted_state is encrypted using\n 128-bit AES in CBC mode with the given IV. The Message\n Authentication Code (MAC) is calculated using HMAC-SHA-256 over\n key_name (16 octets) and IV (16 octets), followed by the length of\n the encrypted_state field (2 octets) and its contents (variable\n length).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 10]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n ProtocolVersion protocol_version;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n opaque master_secret[48];\n ClientIdentity client_identity;\n uint32 timestamp;\n } StatePlaintext;\n\n enum {\n anonymous(0),\n certificate_based(1),\n psk(2)\n } ClientAuthenticationType;\n\n struct {\n ClientAuthenticationType client_authentication_type;\n select (ClientAuthenticationType) {\n case anonymous: struct {};\n case certificate_based:\n ASN.1Cert certificate_list<0..2^24-1>;\n case psk:\n opaque psk_identity<0..2^16-1>; /* from [RFC4279] */\n };\n } ClientIdentity;\n\n The structure StatePlaintext stores the TLS session state including\n the master_secret. The timestamp within this structure allows the\n TLS server to expire tickets. To cover the authentication and key\n exchange protocols provided by TLS, the ClientIdentity structure\n contains the authentication type of the client used in the initial\n exchange (see ClientAuthenticationType). To offer the TLS server\n with the same capabilities for authentication and authorization, a\n certificate list is included in case of public-key-based\n authentication. The TLS server is therefore able to inspect a number\n of different attributes within these certificates. A specific\n implementation might choose to store a subset of this information or\n additional information. Other authentication mechanisms, such as\n Kerberos [RFC2712], would require different client identity data.\n Other TLS extensions may require the inclusion of additional data in\n the StatePlaintext structure.\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 11]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5. Security Considerations\n\n This section addresses security issues related to the usage of a\n ticket. Tickets must be authenticated and encrypted to prevent\n modification or eavesdropping by an attacker. Several attacks\n described below will be possible if this is not carefully done.\n\n Implementations should take care to ensure that the processing of\n tickets does not increase the chance of denial of service as\n described below.\n\n5.1. Invalidating Sessions\n\n The TLS specification requires that TLS sessions be invalidated when\n errors occur. [CSSC] discusses the security implications of this in\n detail. In the analysis within this paper, failure to invalidate\n sessions does not pose a security risk. This is because the TLS\n handshake uses a non-reversible function to derive keys for a session\n so information about one session does not provide an advantage to\n attack the master secret or a different session. If a session\n invalidation scheme is used, the implementation should verify the\n integrity of the ticket before using the contents to invalidate a\n session to ensure that an attacker cannot invalidate a chosen\n session.\n\n5.2. Stolen Tickets\n\n An eavesdropper or man-in-the-middle may obtain the ticket and\n attempt to use it to establish a session with the server; however,\n since the ticket is encrypted and the attacker does not know the\n secret key, a stolen ticket does not help an attacker resume a\n session. A TLS server MUST use strong encryption and integrity\n protection for the ticket to prevent an attacker from using a brute\n force mechanism to obtain the ticket's contents.\n\n5.3. Forged Tickets\n\n A malicious user could forge or alter a ticket in order to resume a\n session, to extend its lifetime, to impersonate another user, or to\n gain additional privileges. This attack is not possible if the\n ticket is protected using a strong integrity protection algorithm\n such as a keyed HMAC-SHA-256.\n\n5.4. Denial of Service Attacks\n\n The key_name field defined in the recommended ticket format helps the\n server efficiently reject tickets that it did not issue. However, an\n adversary could store or generate a large number of tickets to send\n\n\n\nSalowey, et al. Standards Track [Page 12]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n to the TLS server for verification. To minimize the possibility of a\n denial of service, the verification of the ticket should be\n lightweight (e.g., using efficient symmetric key cryptographic\n algorithms).\n\n5.5. Ticket Protection Key Management\n\n A full description of the management of the keys used to protect the\n ticket is beyond the scope of this document. A list of RECOMMENDED\n practices is given below.\n\n o The keys should be generated securely following the randomness\n recommendations in [RFC4086].\n\n o The keys and cryptographic protection algorithms should be at\n least 128 bits in strength. Some ciphersuites and applications\n may require cryptographic protection greater than 128 bits in\n strength.\n\n o The keys should not be used for any purpose other than generating\n and verifying tickets.\n\n o The keys should be changed regularly.\n\n o The keys should be changed if the ticket format or cryptographic\n protection algorithms change.\n\n5.6. Ticket Lifetime\n\n The TLS server controls the lifetime of the ticket. Servers\n determine the acceptable lifetime based on the operational and\n security requirements of the environments in which they are deployed.\n The ticket lifetime may be longer than the 24-hour lifetime\n recommended in [RFC4346]. TLS clients may be given a hint of the\n lifetime of the ticket. Since the lifetime of a ticket may be\n unspecified, a client has its own local policy that determines when\n it discards tickets.\n\n5.7. Alternate Ticket Formats and Distribution Schemes\n\n If the ticket format or distribution scheme defined in this document\n is not used, then great care must be taken in analyzing the security\n of the solution. In particular, if confidential information, such as\n a secret key, is transferred to the client, it MUST be done using\n secure communication so as to prevent attackers from obtaining or\n modifying the key. Also, the ticket MUST have its integrity and\n confidentiality protected with strong cryptographic techniques to\n prevent a breach in the security of the system.\n\n\n\nSalowey, et al. Standards Track [Page 13]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5.8. Identity Privacy, Anonymity, and Unlinkability\n\n This document mandates that the content of the ticket is\n confidentiality protected in order to avoid leakage of its content,\n such as user-relevant information. As such, it prevents disclosure\n of potentially sensitive information carried within the ticket.\n\n The initial handshake exchange, which was used to obtain the ticket,\n might not provide identity confidentiality of the client based on the\n properties of TLS. Another relevant security threat is the ability\n for an on-path adversary to observe multiple TLS handshakes where the\n same ticket is used, therefore concluding they belong to the same\n communication endpoints. Application designers that use the ticket\n mechanism described in this document should consider that\n unlinkability [ANON] is not necessarily provided.\n\n While a full discussion of these topics is beyond the scope of this\n document, it should be noted that it is possible to issue a ticket\n using a TLS renegotiation handshake that occurs after a secure tunnel\n has been established by a previous handshake. This may help address\n some privacy and unlinkability issues in some environments.\n\n6. Acknowledgements\n\n The authors would like to thank the following people for their help\n with preparing and reviewing this document: Eric Rescorla, Mohamad\n Badra, Tim Dierks, Nelson Bolyard, Nancy Cam-Winget, David McGrew,\n Rob Dugal, Russ Housley, Amir Herzberg, Bernard Aboba, and members of\n the TLS working group.\n\n [CSSC] describes a solution that is very similar to the one described\n in this document and gives a detailed analysis of the security\n considerations involved. [RFC2712] describes a mechanism for using\n Kerberos [RFC4120] in TLS ciphersuites, which helped inspire the use\n of tickets to avoid server state. [RFC4851] makes use of a similar\n mechanism to avoid maintaining server state for the cryptographic\n tunnel. [SC97] also investigates the concept of stateless sessions.\n\n The authors would also like to thank Jan Nordqvist, who found the\n encoding error in RFC 4507, corrected by this document. In addition\n Nagendra Modadugu, Wan-Teh Chang, and Michael D'Errico provided\n useful feedback during the review of this document.\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 14]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n7. IANA Considerations\n\n IANA has assigned a TLS extension number of 35 to the SessionTicket\n TLS extension from the TLS registry of ExtensionType values defined\n in [RFC4366].\n\n IANA has assigned a TLS HandshakeType number 4 to the\n NewSessionTicket handshake type from the TLS registry of\n HandshakeType values defined in [RFC4346].\n\n This document does not require any actions or assignments from IANA.\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2246] Dierks, T. and C. Allen, \"The TLS Protocol Version 1.0\",\n RFC 2246, January 1999.\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346, April 2006.\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, April 2006.\n\n [RFC4507] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 4507, May 2006.\n\n8.2. Informative References\n\n [AES] National Institute of Standards and Technology, \"Advanced\n Encryption Standard (AES)\", Federal Information Processing\n Standards (FIPS) Publication 197, November 2001.\n\n [ANON] Pfitzmann, A. and M. Hansen, \"Anonymity, Unlinkability,\n Unobservability, Pseudonymity, and Identity Management - A\n Consolidated Proposal for Terminology\", http://\n dud.inf.tu-dresden.de/literatur/\n Anon_Terminology_v0.26-1.pdf Version 0.26, December 2005.\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 15]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n [CBC] National Institute of Standards and Technology,\n \"Recommendation for Block Cipher Modes of Operation -\n Methods and Techniques\", NIST Special Publication 800-38A,\n December 2001.\n\n [CSSC] Shacham, H., Boneh, D., and E. Rescorla, \"Client-side\n caching for TLS\", Transactions on Information and System\n Security (TISSEC) , Volume 7, Issue 4, November 2004.\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [RFC4086] Eastlake, D., Schiller, J., and S. Crocker, \"Randomness\n Requirements for Security\", BCP 106, RFC 4086, June 2005.\n\n [RFC4120] Neuman, C., Yu, T., Hartman, S., and K. Raeburn, \"The\n Kerberos Network Authentication Service (V5)\", RFC 4120,\n July 2005.\n\n [RFC4279] Eronen, P. and H. Tschofenig, \"Pre-Shared Key Ciphersuites\n for Transport Layer Security (TLS)\", RFC 4279,\n December 2005.\n\n [RFC4634] Eastlake, D. and T. Hansen, \"US Secure Hash Algorithms\n (SHA and HMAC-SHA)\", RFC 4634, July 2006.\n\n [RFC4851] Cam-Winget, N., McGrew, D., Salowey, J., and H. Zhou, \"The\n Flexible Authentication via Secure Tunneling Extensible\n Authentication Protocol Method (EAP-FAST)\", RFC 4851,\n May 2007.\n\n [SC97] Aura, T. and P. Nikander, \"Stateless Connections\",\n Proceedings of the First International Conference on\n Information and Communication Security (ICICS '97) , 1997.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 16]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAppendix A. Discussion of Changes to RFC 4507\n\n RFC 4507 [RFC4507] defines a mechanism to resume a TLS session\n without maintaining server side state by specifying an encrypted\n ticket that is maintained on the client. The client presents this\n ticket to the server in a SessionTicket hello extension. The\n encoding in RFC 4507 used the XDR style encoding specified in TLS\n [RFC4346].\n\n An error in the encoding caused the specification to differ from\n deployed implementations. At the time of this writing there are no\n known implementations that follow the encoding specified in RFC 4507.\n This update to RFC 4507 aligns the document with these currently\n deployed implementations.\n\n Erroneous encoding in RFC 4507 resulted in two length fields; one for\n the extension contents and one for the ticket itself. Hence, for a\n ticket that is 256 bytes long and begins with the hex value FF FF,\n the encoding of the extension would be as follows according to RFC\n 4507:\n\n 00 23 Ticket Extension type 35\n 01 02 Length of extension contents\n 01 00 Length of ticket\n FF FF .. .. Actual ticket\n\n The update proposed in this document reflects what implementations\n actually encode, namely it removes the redundant length field. So,\n for a ticket that is 256 bytes long and begins with the hex value FF\n FF, the encoding of the extension would be as follows according to\n this update:\n\n 00 23 Extension type 35\n 01 00 Length of extension contents (ticket)\n FF FF .. .. Actual ticket\n\n A server implemented according to RFC 4507 receiving a ticket\n extension from a client conforming to this document would interpret\n the first two bytes of the ticket as the length of this ticket. This\n will result in either an inconsistent length field or in the\n processing of a ticket missing the first two bytes. In the first\n case, the server should reject the request based on a malformed\n length. In the second case, the server should reject the ticket\n based on a malformed ticket, incorrect key version, or failed\n decryption. A server implementation based on this update receiving\n an RFC 4507 extension would interpret the first length field as the\n\n\n\n\n\nSalowey, et al. Standards Track [Page 17]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n length of the ticket and include the second two length bytes as the\n first bytes in the ticket, resulting in the ticket being rejected\n based on a malformed ticket, incorrect key version, or failed\n decryption.\n\n Note that the encoding of an empty SessionTicket extension was\n ambiguous in RFC 4507. An RFC 4507 implementation may have encoded\n it as:\n\n 00 23 Extension type 35\n 00 02 Length of extension contents\n 00 00 Length of ticket\n\n or it may have encoded it the same way as this update:\n\n 00 23 Extension type 35\n 00 00 Length of extension contents\n\n A server wishing to support RFC 4507 clients should respond to an\n empty SessionTicket extension encoded the same way as it received it.\n\n A server implementation can construct tickets such that it can detect\n an RFC 4507 implementation, if one existed, by including a cookie at\n the beginning of the tickets that can be differentiated from a valid\n length. For example, if an implementation constructed tickets to\n start with the hex values FF FF, then it could determine where the\n ticket begins and determine the length correctly from the type of\n length fields present.\n\n This document makes a few additional changes to RFC 4507 listed\n below.\n\n o Clarifying that the server can allow session resumption using a\n ticket without issuing a new ticket in Section 3.1.\n\n o Clarifying that the lifetime is relative to when the ticket is\n received in section 3.3.\n\n o Clarifying that the NewSessionTicket handshake message is included\n in the hash generated for the Finished messages in Section 3.3.\n\n o Clarifying the interaction with TLS Session ID in Section 3.4.\n\n o Recommending the use of SHA-256 for the integrity protection of\n the ticket in Section 4.\n\n o Clarifying that additional data can be included in the\n StatePlaintext structure in Section 4.\n\n\n\nSalowey, et al. Standards Track [Page 18]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAuthors' Addresses\n\n Joseph Salowey\n Cisco Systems\n 2901 3rd Ave\n Seattle, WA 98121\n US\n\n EMail: [redacted-email]\n\n\n Hao Zhou\n Cisco Systems\n 4125 Highlander Parkway\n Richfield, OH 44286\n US\n\n EMail: [redacted-email]\n\n\n Pasi Eronen\n Nokia Research Center\n P.O. Box 407\n FIN-00045 Nokia Group\n Finland\n\n EMail: [redacted-email]\n\n\n Hannes Tschofenig\n Nokia Siemens Networks\n Otto-Hahn-Ring 6\n Munich, Bayern 81739\n Germany\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 19]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nFull Copyright Statement\n\n Copyright (C) The IETF Trust (2008).\n\n This document is subject to the rights, licenses and restrictions\n contained in BCP 78, and except as set forth therein, the authors\n retain all their rights.\n\n This document and the information contained herein are provided on an\n \"AS IS\" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\n OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\n THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\n OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\n THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\n WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\n\nIntellectual Property\n\n The IETF takes no position regarding the validity or scope of any\n Intellectual Property Rights or other rights that might be claimed to\n pertain to the implementation or use of the technology described in\n this document or the extent to which any license under such rights\n might or might not be available; nor does it represent that it has\n made any independent effort to identify any such rights. Information\n on the procedures with respect to rights in RFC documents can be\n found in BCP 78 and BCP 79.\n\n Copies of IPR disclosures made to the IETF Secretariat and any\n assurances of licenses to be made available, or the result of an\n attempt made to obtain a general license or permission for the use of\n such proprietary rights by implementers or users of this\n specification can be obtained from the IETF on-line IPR repository at\n http://www.ietf.org/ipr.\n\n The IETF invites any interested party to bring to its attention any\n copyrights, patents or patent applications, or other proprietary\n rights that may cover technology that may be required to implement\n this standard. Please address the information to the IETF at\n ietf-ipr@ietf.org.\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 20]\n\n===== DOCUMENT =====\n\nNetwork Working Group T. Dierks\nRequest for Comments: 5246 Independent\nObsoletes: 3268, 4346, 4366 E. Rescorla\nUpdates: 4492 RTFM, Inc.\nCategory: Standards Track August 2008\n\n\n The Transport Layer Security (TLS) Protocol\n Version 1.2\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document specifies Version 1.2 of the Transport Layer Security\n (TLS) protocol. The TLS protocol provides communications security\n over the Internet. The protocol allows client/server applications to\n communicate in a way that is designed to prevent eavesdropping,\n tampering, or message forgery.\n\nTable of Contents\n\n 1. Introduction ....................................................4\n 1.1. Requirements Terminology ...................................5\n 1.2. Major Differences from TLS 1.1 .............................5\n 2. Goals ...........................................................6\n 3. Goals of This Document ..........................................7\n 4. Presentation Language ...........................................7\n 4.1. Basic Block Size ...........................................7\n 4.2. Miscellaneous ..............................................8\n 4.3. Vectors ....................................................8\n 4.4. Numbers ....................................................9\n 4.5. Enumerateds ................................................9\n 4.6. Constructed Types .........................................10\n 4.6.1. Variants ...........................................10\n 4.7. Cryptographic Attributes ..................................12\n 4.8. Constants .................................................14\n 5. HMAC and the Pseudorandom Function .............................14\n 6. The TLS Record Protocol ........................................15\n 6.1. Connection States .........................................16\n 6.2. Record Layer ..............................................19\n 6.2.1. Fragmentation ......................................19\n\n\n\nDierks & Rescorla Standards Track [Page 1]\n\f\nRFC 5246 TLS August 2008\n\n\n 6.2.2. Record Compression and Decompression ...............20\n 6.2.3. Record Payload Protection ..........................21\n 6.2.3.1. Null or Standard Stream Cipher ............22\n 6.2.3.2. CBC Block Cipher ..........................22\n 6.2.3.3. AEAD Ciphers ..............................24\n 6.3. Key Calculation ...........................................25\n 7. The TLS Handshaking Protocols ..................................26\n 7.1. Change Cipher Spec Protocol ...............................27\n 7.2. Alert Protocol ............................................28\n 7.2.1. Closure Alerts .....................................29\n 7.2.2. Error Alerts .......................................30\n 7.3. Handshake Protocol Overview ...............................33\n 7.4. Handshake Protocol ........................................37\n 7.4.1. Hello Messages .....................................38\n 7.4.1.1. Hello Request .............................38\n 7.4.1.2. Client Hello ..............................39\n 7.4.1.3. Server Hello ..............................42\n 7.4.1.4. Hello Extensions ..........................44\n 7.4.1.4.1. Signature Algorithms ...........45\n 7.4.2. Server Certificate .................................47\n 7.4.3. Server Key Exchange Message ........................50\n 7.4.4. Certificate Request ................................53\n 7.4.5. Server Hello Done ..................................55\n 7.4.6. Client Certificate .................................55\n 7.4.7. Client Key Exchange Message ........................57\n 7.4.7.1. RSA-Encrypted Premaster Secret Message ....58\n 7.4.7.2. Client Diffie-Hellman Public Value ........61\n 7.4.8. Certificate Verify .................................62\n 7.4.9. Finished ...........................................63\n 8. Cryptographic Computations .....................................64\n 8.1. Computing the Master Secret ...............................64\n 8.1.1. RSA ................................................65\n 8.1.2. Diffie-Hellman .....................................65\n 9. Mandatory Cipher Suites ........................................65\n 10. Application Data Protocol .....................................65\n 11. Security Considerations .......................................65\n 12. IANA Considerations ...........................................65\n Appendix A. Protocol Data Structures and Constant Values ..........68\n A.1. Record Layer ..............................................68\n A.2. Change Cipher Specs Message ...............................69\n A.3. Alert Messages ............................................69\n A.4. Handshake Protocol ........................................70\n A.4.1. Hello Messages .....................................71\n A.4.2. Server Authentication and Key Exchange Messages ....72\n A.4.3. Client Authentication and Key Exchange Messages ....74\n A.4.4. Handshake Finalization Message .....................74\n A.5. The Cipher Suite ..........................................75\n A.6. The Security Parameters ...................................77\n\n\n\nDierks & Rescorla Standards Track [Page 2]\n\f\nRFC 5246 TLS August 2008\n\n\n A.7. Changes to RFC 4492 .......................................78\n Appendix B. Glossary ..............................................78\n Appendix C. Cipher Suite Definitions ..............................83\n Appendix D. Implementation Notes ..................................85\n D.1. Random Number Generation and Seeding ......................85\n D.2. Certificates and Authentication ...........................85\n D.3. Cipher Suites .............................................85\n D.4. Implementation Pitfalls ...................................85\n Appendix E. Backward Compatibility ................................87\n E.1. Compatibility with TLS 1.0/1.1 and SSL 3.0 ................87\n E.2. Compatibility with SSL 2.0 ................................88\n E.3. Avoiding Man-in-the-Middle Version Rollback ...............90\n Appendix F. Security Analysis .....................................91\n F.1. Handshake Protocol ........................................91\n F.1.1. Authentication and Key Exchange ....................91\n F.1.1.1. Anonymous Key Exchange ....................91\n F.1.1.2. RSA Key Exchange and Authentication .......92\n F.1.1.3. Diffie-Hellman Key Exchange with\n Authentication ............................92\n F.1.2. Version Rollback Attacks ...........................93\n F.1.3. Detecting Attacks Against the Handshake Protocol ...94\n F.1.4. Resuming Sessions ..................................94\n F.2. Protecting Application Data ...............................94\n F.3. Explicit IVs ..............................................95\n F.4. Security of Composite Cipher Modes ........................95\n F.5. Denial of Service .........................................96\n F.6. Final Notes ...............................................96\n Normative References ..............................................97\n Informative References ............................................98\n Working Group Information ........................................101\n Contributors .....................................................101\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 3]\n\f\nRFC 5246 TLS August 2008\n\n\n1. Introduction\n\n The primary goal of the TLS protocol is to provide privacy and data\n integrity between two communicating applications. The protocol is\n composed of two layers: the TLS Record Protocol and the TLS Handshake\n Protocol. At the lowest level, layered on top of some reliable\n transport protocol (e.g., TCP [TCP]), is the TLS Record Protocol.\n The TLS Record Protocol provides connection security that has two\n basic properties:\n\n - The connection is private. Symmetric cryptography is used for\n data encryption (e.g., AES [AES], RC4 [SCH], etc.). The keys for\n this symmetric encryption are generated uniquely for each\n connection and are based on a secret negotiated by another\n protocol (such as the TLS Handshake Protocol). The Record\n Protocol can also be used without encryption.\n\n - The connection is reliable. Message transport includes a message\n integrity check using a keyed MAC. Secure hash functions (e.g.,\n SHA-1, etc.) are used for MAC computations. The Record Protocol\n can operate without a MAC, but is generally only used in this mode\n while another protocol is using the Record Protocol as a transport\n for negotiating security parameters.\n\n The TLS Record Protocol is used for encapsulation of various higher-\n level protocols. One such encapsulated protocol, the TLS Handshake\n Protocol, allows the server and client to authenticate each other and\n to negotiate an encryption algorithm and cryptographic keys before\n the application protocol transmits or receives its first byte of\n data. The TLS Handshake Protocol provides connection security that\n has three basic properties:\n\n - The peer's identity can be authenticated using asymmetric, or\n public key, cryptography (e.g., RSA [RSA], DSA [DSS], etc.). This\n authentication can be made optional, but is generally required for\n at least one of the peers.\n\n - The negotiation of a shared secret is secure: the negotiated\n secret is unavailable to eavesdroppers, and for any authenticated\n connection the secret cannot be obtained, even by an attacker who\n can place himself in the middle of the connection.\n\n - The negotiation is reliable: no attacker can modify the\n negotiation communication without being detected by the parties to\n the communication.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 4]\n\f\nRFC 5246 TLS August 2008\n\n\n One advantage of TLS is that it is application protocol independent.\n Higher-level protocols can layer on top of the TLS protocol\n transparently. The TLS standard, however, does not specify how\n protocols add security with TLS; the decisions on how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n1.1. Requirements Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [REQ].\n\n1.2. Major Differences from TLS 1.1\n\n This document is a revision of the TLS 1.1 [TLS1.1] protocol which\n contains improved flexibility, particularly for negotiation of\n cryptographic algorithms. The major changes are:\n\n - The MD5/SHA-1 combination in the pseudorandom function (PRF) has\n been replaced with cipher-suite-specified PRFs. All cipher suites\n in this document use P_SHA256.\n\n - The MD5/SHA-1 combination in the digitally-signed element has been\n replaced with a single hash. Signed elements now include a field\n that explicitly specifies the hash algorithm used.\n\n - Substantial cleanup to the client's and server's ability to\n specify which hash and signature algorithms they will accept.\n Note that this also relaxes some of the constraints on signature\n and hash algorithms from previous versions of TLS.\n\n - Addition of support for authenticated encryption with additional\n data modes.\n\n - TLS Extensions definition and AES Cipher Suites were merged in\n from external [TLSEXT] and [TLSAES].\n\n - Tighter checking of EncryptedPreMasterSecret version numbers.\n\n - Tightened up a number of requirements.\n\n - Verify_data length now depends on the cipher suite (default is\n still 12).\n\n - Cleaned up description of Bleichenbacher/Klima attack defenses.\n\n\n\n\nDierks & Rescorla Standards Track [Page 5]\n\f\nRFC 5246 TLS August 2008\n\n\n - Alerts MUST now be sent in many cases.\n\n - After a certificate_request, if no certificates are available,\n clients now MUST send an empty certificate list.\n\n - TLS_RSA_WITH_AES_128_CBC_SHA is now the mandatory to implement\n cipher suite.\n\n - Added HMAC-SHA256 cipher suites.\n\n - Removed IDEA and DES cipher suites. They are now deprecated and\n will be documented in a separate document.\n\n - Support for the SSLv2 backward-compatible hello is now a MAY, not\n a SHOULD, with sending it a SHOULD NOT. Support will probably\n become a SHOULD NOT in the future.\n\n - Added limited \"fall-through\" to the presentation language to allow\n multiple case arms to have the same encoding.\n\n - Added an Implementation Pitfalls sections\n\n - The usual clarifications and editorial work.\n\n2. Goals\n\n The goals of the TLS protocol, in order of priority, are as follows:\n\n 1. Cryptographic security: TLS should be used to establish a secure\n connection between two parties.\n\n 2. Interoperability: Independent programmers should be able to\n develop applications utilizing TLS that can successfully exchange\n cryptographic parameters without knowledge of one another's code.\n\n 3. Extensibility: TLS seeks to provide a framework into which new\n public key and bulk encryption methods can be incorporated as\n necessary. This will also accomplish two sub-goals: preventing\n the need to create a new protocol (and risking the introduction of\n possible new weaknesses) and avoiding the need to implement an\n entire new security library.\n\n 4. Relative efficiency: Cryptographic operations tend to be highly\n CPU intensive, particularly public key operations. For this\n reason, the TLS protocol has incorporated an optional session\n caching scheme to reduce the number of connections that need to be\n established from scratch. Additionally, care has been taken to\n reduce network activity.\n\n\n\nDierks & Rescorla Standards Track [Page 6]\n\f\nRFC 5246 TLS August 2008\n\n\n3. Goals of This Document\n\n This document and the TLS protocol itself are based on the SSL 3.0\n Protocol Specification as published by Netscape. The differences\n between this protocol and SSL 3.0 are not dramatic, but they are\n significant enough that the various versions of TLS and SSL 3.0 do\n not interoperate (although each protocol incorporates a mechanism by\n which an implementation can back down to prior versions). This\n document is intended primarily for readers who will be implementing\n the protocol and for those doing cryptographic analysis of it. The\n specification has been written with this in mind, and it is intended\n to reflect the needs of those two groups. For that reason, many of\n the algorithm-dependent data structures and rules are included in the\n body of the text (as opposed to in an appendix), providing easier\n access to them.\n\n This document is not intended to supply any details of service\n definition or of interface definition, although it does cover select\n areas of policy as they are required for the maintenance of solid\n security.\n\n4. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used. The syntax draws from\n several sources in its structure. Although it resembles the\n programming language \"C\" in its syntax and XDR [XDR] in both its\n syntax and intent, it would be risky to draw too many parallels. The\n purpose of this presentation language is to document TLS only; it has\n no general application beyond that particular goal.\n\n4.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 7]\n\f\nRFC 5246 TLS August 2008\n\n\n4.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" double\n brackets.\n\n Single-byte entities containing uninterpreted data are of type\n opaque.\n\n4.3. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* 3 consecutive 3 byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n In the following example, mandatory is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 4.4). On the\n other hand, longer can represent up to 800 bytes of data, or 400\n uint16 elements, and it may be empty. Its encoding will include a\n\n\n\nDierks & Rescorla Standards Track [Page 8]\n\f\nRFC 5246 TLS August 2008\n\n\n two-byte actual length field prepended to the vector. The length of\n an encoded vector must be an even multiple of the length of a single\n element (for example, a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is 2 bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n4.4. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are formed from fixed-length series of bytes\n concatenated as described in Section 4.1 and are also unsigned. The\n following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are stored in\n network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n Note that in some cases (e.g., DH parameters) it is necessary to\n represent integers as opaque vectors. In such cases, they are\n represented as unsigned integers (i.e., leading zero octets are not\n required even if the most significant bit is set).\n\n4.5. Enumerateds\n\n An additional sparse data type is available called enum. A field of\n type enum can only assume the values declared in the definition.\n Each definition is a different type. Only enumerateds of the same\n type may be assigned or compared. Every element of an enumerated\n must be assigned a value, as demonstrated in the following example.\n Since the elements of the enumerated are not ordered, they can be\n assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n\n\n\nDierks & Rescorla Standards Track [Page 9]\n\f\nRFC 5246 TLS August 2008\n\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n For enumerateds that are never converted to external representation,\n the numerical information may be omitted.\n\n enum { low, medium, high } Amount;\n\n4.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax for definition is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } [[T]];\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n Structure definitions may be embedded.\n\n4.6.1. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. There\n must be a case arm for every element of the enumeration declared in\n the select. Case arms have limited fall-through: if two case arms\n follow in immediate succession with no fields in between, then they\n\n\n\nDierks & Rescorla Standards Track [Page 10]\n\f\nRFC 5246 TLS August 2008\n\n\n both contain the same fields. Thus, in the example below, \"orange\"\n and \"banana\" both contain V2. Note that this is a new piece of\n syntax in TLS 1.2.\n\n The body of the variant structure may be given a label for reference.\n The mechanism by which the variant is selected at runtime is not\n prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1;\n case e2: Te2;\n case e3: case e4: Te3;\n ....\n case en: Ten;\n } [[fv]];\n } [[Tv]];\n\n For example:\n\n enum { apple, orange, banana } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n select (VariantTag) { /* value of selector is implicit */\n case apple:\n V1; /* VariantBody, tag = apple */\n case orange:\n case banana:\n V2; /* VariantBody, tag = orange or banana */\n } variant_body; /* optional label on variant */\n } VariantRecord;\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 11]\n\f\nRFC 5246 TLS August 2008\n\n\n4.7. Cryptographic Attributes\n\n The five cryptographic operations -- digital signing, stream cipher\n encryption, block cipher encryption, authenticated encryption with\n additional data (AEAD) encryption, and public key encryption -- are\n designated digitally-signed, stream-ciphered, block-ciphered, aead-\n ciphered, and public-key-encrypted, respectively. A field's\n cryptographic processing is specified by prepending an appropriate\n key word designation before the field's type specification.\n Cryptographic keys are implied by the current session state (see\n Section 6.1).\n\n A digitally-signed element is encoded as a struct DigitallySigned:\n\n struct {\n SignatureAndHashAlgorithm algorithm;\n opaque signature<0..2^16-1>;\n } DigitallySigned;\n\n The algorithm field specifies the algorithm used (see Section\n 7.4.1.4.1 for the definition of this field). Note that the\n introduction of the algorithm field is a change from previous\n versions. The signature is a digital signature using those\n algorithms over the contents of the element. The contents themselves\n do not appear on the wire but are simply calculated. The length of\n the signature is specified by the signing algorithm and key.\n\n In RSA signing, the opaque vector contains the signature generated\n using the RSASSA-PKCS1-v1_5 signature scheme defined in [PKCS1]. As\n discussed in [PKCS1], the DigestInfo MUST be DER-encoded [X680]\n [X690]. For hash algorithms without parameters (which includes\n SHA-1), the DigestInfo.AlgorithmIdentifier.parameters field MUST be\n NULL, but implementations MUST accept both without parameters and\n with NULL parameters. Note that earlier versions of TLS used a\n different RSA signature scheme that did not include a DigestInfo\n encoding.\n\n In DSA, the 20 bytes of the SHA-1 hash are run directly through the\n Digital Signing Algorithm with no additional hashing. This produces\n two values, r and s. The DSA signature is an opaque vector, as\n above, the contents of which are the DER encoding of:\n\n Dss-Sig-Value ::= SEQUENCE {\n r INTEGER,\n s INTEGER\n }\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 12]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: In current terminology, DSA refers to the Digital Signature\n Algorithm and DSS refers to the NIST standard. In the original SSL\n and TLS specs, \"DSS\" was used universally. This document uses \"DSA\"\n to refer to the algorithm, \"DSS\" to refer to the standard, and it\n uses \"DSS\" in the code point definitions for historical continuity.\n\n In stream cipher encryption, the plaintext is exclusive-ORed with an\n identical amount of output generated from a cryptographically secure\n keyed pseudorandom number generator.\n\n In block cipher encryption, every block of plaintext encrypts to a\n block of ciphertext. All block cipher encryption is done in CBC\n (Cipher Block Chaining) mode, and all items that are block-ciphered\n will be an exact multiple of the cipher block length.\n\n In AEAD encryption, the plaintext is simultaneously encrypted and\n integrity protected. The input may be of any length, and aead-\n ciphered output is generally larger than the input in order to\n accommodate the integrity check value.\n\n In public key encryption, a public key algorithm is used to encrypt\n data in such a way that it can be decrypted only with the matching\n private key. A public-key-encrypted element is encoded as an opaque\n vector <0..2^16-1>, where the length is specified by the encryption\n algorithm and key.\n\n RSA encryption is done using the RSAES-PKCS1-v1_5 encryption scheme\n defined in [PKCS1].\n\n In the following example\n\n stream-ciphered struct {\n uint8 field1;\n uint8 field2;\n digitally-signed opaque {\n uint8 field3<0..255>;\n uint8 field4;\n };\n } UserType;\n\n The contents of the inner struct (field3 and field4) are used as\n input for the signature/hash algorithm, and then the entire structure\n is encrypted with a stream cipher. The length of this structure, in\n bytes, would be equal to two bytes for field1 and field2, plus two\n bytes for the signature and hash algorithm, plus two bytes for the\n length of the signature, plus the length of the output of the signing\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 13]\n\f\nRFC 5246 TLS August 2008\n\n\n algorithm. The length of the signature is known because the\n algorithm and key used for the signing are known prior to encoding or\n decoding this structure.\n\n4.8. Constants\n\n Typed constants can be defined for purposes of specification by\n declaring a symbol of the desired type and assigning values to it.\n\n Under-specified types (opaque, variable-length vectors, and\n structures that contain opaque) cannot be assigned values. No fields\n of a multi-element structure or vector may be elided.\n\n For example:\n\n struct {\n uint8 f1;\n uint8 f2;\n } Example1;\n\n Example1 ex1 = {1, 4}; /* assigns f1 = 1, f2 = 4 */\n\n5. HMAC and the Pseudorandom Function\n\n The TLS record layer uses a keyed Message Authentication Code (MAC)\n to protect message integrity. The cipher suites defined in this\n document use a construction known as HMAC, described in [HMAC], which\n is based on a hash function. Other cipher suites MAY define their\n own MAC constructions, if needed.\n\n In addition, a construction is required to do expansion of secrets\n into blocks of data for the purposes of key generation or validation.\n This pseudorandom function (PRF) takes as input a secret, a seed, and\n an identifying label and produces an output of arbitrary length.\n\n In this section, we define one PRF, based on HMAC. This PRF with the\n SHA-256 hash function is used for all cipher suites defined in this\n document and in TLS documents published prior to this document when\n TLS 1.2 is negotiated. New cipher suites MUST explicitly specify a\n PRF and, in general, SHOULD use the TLS PRF with SHA-256 or a\n stronger standard hash function.\n\n First, we define a data expansion function, P_hash(secret, data),\n that uses a single hash function to expand a secret and seed into an\n arbitrary quantity of output:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 14]\n\f\nRFC 5246 TLS August 2008\n\n\n P_hash(secret, seed) = HMAC_hash(secret, A(1) + seed) +\n HMAC_hash(secret, A(2) + seed) +\n HMAC_hash(secret, A(3) + seed) + ...\n\n where + indicates concatenation.\n\n A() is defined as:\n\n A(0) = seed\n A(i) = HMAC_hash(secret, A(i-1))\n\n P_hash can be iterated as many times as necessary to produce the\n required quantity of data. For example, if P_SHA256 is being used to\n create 80 bytes of data, it will have to be iterated three times\n (through A(3)), creating 96 bytes of output data; the last 16 bytes\n of the final iteration will then be discarded, leaving 80 bytes of\n output data.\n\n TLS's PRF is created by applying P_hash to the secret as:\n\n PRF(secret, label, seed) = P_(secret, label + seed)\n\n The label is an ASCII string. It should be included in the exact\n form it is given without a length byte or trailing null character.\n For example, the label \"slithy toves\" would be processed by hashing\n the following bytes:\n\n 73 6C 69 74 68 79 20 74 6F 76 65 73\n\n6. The TLS Record Protocol\n\n The TLS Record Protocol is a layered protocol. At each layer,\n messages may include fields for length, description, and content.\n The Record Protocol takes messages to be transmitted, fragments the\n data into manageable blocks, optionally compresses the data, applies\n a MAC, encrypts, and transmits the result. Received data is\n decrypted, verified, decompressed, reassembled, and then delivered to\n higher-level clients.\n\n Four protocols that use the record protocol are described in this\n document: the handshake protocol, the alert protocol, the change\n cipher spec protocol, and the application data protocol. In order to\n allow extension of the TLS protocol, additional record content types\n can be supported by the record protocol. New record content type\n values are assigned by IANA in the TLS Content Type Registry as\n described in Section 12.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 15]\n\f\nRFC 5246 TLS August 2008\n\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST send an\n unexpected_message alert.\n\n Any protocol designed for use over TLS must be carefully designed to\n deal with all possible attacks against it. As a practical matter,\n this means that the protocol designer must be aware of what security\n properties TLS does and does not provide and cannot safely rely on\n the latter.\n\n Note in particular that type and length of a record are not protected\n by encryption. If this information is itself sensitive, application\n designers may wish to take steps (padding, cover traffic) to minimize\n information leakage.\n\n6.1. Connection States\n\n\n\n===== DOCUMENT =====\n\n A TLS connection state is the operating environment of the TLS Record\n Protocol. It specifies a compression algorithm, an encryption\n algorithm, and a MAC algorithm. In addition, the parameters for\n these algorithms are known: the MAC key and the bulk encryption keys\n for the connection in both the read and the write directions.\n Logically, there are always four connection states outstanding: the\n current read and write states, and the pending read and write states.\n All records are processed under the current read and write states.\n The security parameters for the pending states can be set by the TLS\n Handshake Protocol, and the ChangeCipherSpec can selectively make\n either of the pending states current, in which case the appropriate\n current state is disposed of and replaced with the pending state; the\n pending state is then reinitialized to an empty state. It is illegal\n to make a state that has not been initialized with security\n parameters a current state. The initial current state always\n specifies that no encryption, compression, or MAC will be used.\n\n The security parameters for a TLS Connection read and write state are\n set by providing the following values:\n\n connection end\n Whether this entity is considered the \"client\" or the \"server\" in\n this connection.\n\n PRF algorithm\n An algorithm used to generate keys from the master secret (see\n Sections 5 and 6.3).\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 16]\n\f\nRFC 5246 TLS August 2008\n\n\n bulk encryption algorithm\n An algorithm to be used for bulk encryption. This specification\n includes the key size of this algorithm, whether it is a block,\n stream, or AEAD cipher, the block size of the cipher (if\n appropriate), and the lengths of explicit and implicit\n initialization vectors (or nonces).\n\n MAC algorithm\n An algorithm to be used for message authentication. This\n specification includes the size of the value returned by the MAC\n algorithm.\n\n compression algorithm\n An algorithm to be used for data compression. This specification\n must include all information the algorithm requires to do\n compression.\n\n master secret\n A 48-byte secret shared between the two peers in the connection.\n\n client random\n A 32-byte value provided by the client.\n\n server random\n A 32-byte value provided by the server.\n\n These parameters are defined in the presentation language as:\n\n enum { server, client } ConnectionEnd;\n\n enum { tls_prf_sha256 } PRFAlgorithm;\n\n enum { null, rc4, 3des, aes }\n BulkCipherAlgorithm;\n\n enum { stream, block, aead } CipherType;\n\n enum { null, hmac_md5, hmac_sha1, hmac_sha256,\n hmac_sha384, hmac_sha512} MACAlgorithm;\n\n enum { null(0), (255) } CompressionMethod;\n\n /* The algorithms specified in CompressionMethod, PRFAlgorithm,\n BulkCipherAlgorithm, and MACAlgorithm may be added to. */\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 17]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ConnectionEnd entity;\n PRFAlgorithm prf_algorithm;\n BulkCipherAlgorithm bulk_cipher_algorithm;\n CipherType cipher_type;\n uint8 enc_key_length;\n uint8 block_length;\n uint8 fixed_iv_length;\n uint8 record_iv_length;\n MACAlgorithm mac_algorithm;\n uint8 mac_length;\n uint8 mac_key_length;\n CompressionMethod compression_algorithm;\n opaque master_secret[48];\n opaque client_random[32];\n opaque server_random[32];\n } SecurityParameters;\n\n The record layer will use the security parameters to generate the\n following six items (some of which are not required by all ciphers,\n and are thus empty):\n\n client write MAC key\n server write MAC key\n client write encryption key\n server write encryption key\n client write IV\n server write IV\n\n The client write parameters are used by the server when receiving and\n processing records and vice versa. The algorithm used for generating\n these items from the security parameters is described in Section 6.3.\n\n Once the security parameters have been set and the keys have been\n generated, the connection states can be instantiated by making them\n the current states. These current states MUST be updated for each\n record processed. Each connection state includes the following\n elements:\n\n compression state\n The current state of the compression algorithm.\n\n cipher state\n The current state of the encryption algorithm. This will consist\n of the scheduled key for that connection. For stream ciphers,\n this will also contain whatever state information is necessary to\n allow the stream to continue to encrypt or decrypt data.\n\n\n\n\nDierks & Rescorla Standards Track [Page 18]\n\f\nRFC 5246 TLS August 2008\n\n\n MAC key\n The MAC key for this connection, as generated above.\n\n sequence number\n Each connection state contains a sequence number, which is\n maintained separately for read and write states. The sequence\n number MUST be set to zero whenever a connection state is made the\n active state. Sequence numbers are of type uint64 and may not\n exceed 2^64-1. Sequence numbers do not wrap. If a TLS\n implementation would need to wrap a sequence number, it must\n renegotiate instead. A sequence number is incremented after each\n record: specifically, the first record transmitted under a\n particular connection state MUST use sequence number 0.\n\n6.2. Record Layer\n\n The TLS record layer receives uninterpreted data from higher layers\n in non-empty blocks of arbitrary size.\n\n6.2.1. Fragmentation\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Client\n message boundaries are not preserved in the record layer (i.e.,\n multiple client messages of the same ContentType MAY be coalesced\n into a single TLSPlaintext record, or a single message MAY be\n fragmented across several records).\n\n struct {\n uint8 major;\n uint8 minor;\n } ProtocolVersion;\n\n enum {\n change_cipher_spec(20), alert(21), handshake(22),\n application_data(23), (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type\n The higher-level protocol used to process the enclosed fragment.\n\n\n\n\nDierks & Rescorla Standards Track [Page 19]\n\f\nRFC 5246 TLS August 2008\n\n\n version\n The version of the protocol being employed. This document\n describes TLS Version 1.2, which uses the version { 3, 3 }. The\n version value 3.3 is historical, deriving from the use of {3, 1}\n for TLS 1.0. (See Appendix A.1.) Note that a client that\n supports multiple versions of TLS may not know what version will\n be employed before it receives the ServerHello. See Appendix E\n for discussion about what record layer version number should be\n employed for ClientHello.\n\n length\n The length (in bytes) of the following TLSPlaintext.fragment. The\n length MUST NOT exceed 2^14.\n\n fragment\n The application data. This data is transparent and treated as an\n independent block to be dealt with by the higher-level protocol\n specified by the type field.\n\n Implementations MUST NOT send zero-length fragments of Handshake,\n Alert, or ChangeCipherSpec content types. Zero-length fragments of\n Application data MAY be sent as they are potentially useful as a\n traffic analysis countermeasure.\n\n Note: Data of different TLS record layer content types MAY be\n interleaved. Application data is generally of lower precedence for\n transmission than other content types. However, records MUST be\n delivered to the network in the same order as they are protected by\n the record layer. Recipients MUST receive and process interleaved\n application layer traffic during handshakes subsequent to the first\n one on a connection.\n\n6.2.2. Record Compression and Decompression\n\n All records are compressed using the compression algorithm defined in\n the current session state. There is always an active compression\n algorithm; however, initially it is defined as\n CompressionMethod.null. The compression algorithm translates a\n TLSPlaintext structure into a TLSCompressed structure. Compression\n functions are initialized with default state information whenever a\n connection state is made active. [RFC3749] describes compression\n algorithms for TLS.\n\n Compression must be lossless and may not increase the content length\n by more than 1024 bytes. If the decompression function encounters a\n TLSCompressed.fragment that would decompress to a length in excess of\n 2^14 bytes, it MUST report a fatal decompression failure error.\n\n\n\n\nDierks & Rescorla Standards Track [Page 20]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ContentType type; /* same as TLSPlaintext.type */\n ProtocolVersion version;/* same as TLSPlaintext.version */\n uint16 length;\n opaque fragment[TLSCompressed.length];\n } TLSCompressed;\n\n length\n The length (in bytes) of the following TLSCompressed.fragment.\n The length MUST NOT exceed 2^14 + 1024.\n\n fragment\n The compressed form of TLSPlaintext.fragment.\n\n Note: A CompressionMethod.null operation is an identity operation;\n no fields are altered.\n\n Implementation note: Decompression functions are responsible for\n ensuring that messages cannot cause internal buffer overflows.\n\n6.2.3. Record Payload Protection\n\n The encryption and MAC functions translate a TLSCompressed\n structure into a TLSCiphertext. The decryption functions reverse\n the process. The MAC of the record also includes a sequence\n number so that missing, extra, or repeated messages are\n detectable.\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n select (SecurityParameters.cipher_type) {\n case stream: GenericStreamCipher;\n case block: GenericBlockCipher;\n case aead: GenericAEADCipher;\n } fragment;\n } TLSCiphertext;\n\n type\n The type field is identical to TLSCompressed.type.\n\n version\n The version field is identical to TLSCompressed.version.\n\n length\n The length (in bytes) of the following TLSCiphertext.fragment.\n The length MUST NOT exceed 2^14 + 2048.\n\n\n\nDierks & Rescorla Standards Track [Page 21]\n\f\nRFC 5246 TLS August 2008\n\n\n fragment\n The encrypted form of TLSCompressed.fragment, with the MAC.\n\n6.2.3.1. Null or Standard Stream Cipher\n\n Stream ciphers (including BulkCipherAlgorithm.null; see Appendix A.6)\n convert TLSCompressed.fragment structures to and from stream\n TLSCiphertext.fragment structures.\n\n stream-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n } GenericStreamCipher;\n\n The MAC is generated as:\n\n MAC(MAC_write_key, seq_num +\n TLSCompressed.type +\n TLSCompressed.version +\n TLSCompressed.length +\n TLSCompressed.fragment);\n\n where \"+\" denotes concatenation.\n\n seq_num\n The sequence number for this record.\n\n MAC\n The MAC algorithm specified by SecurityParameters.mac_algorithm.\n\n Note that the MAC is computed before encryption. The stream cipher\n encrypts the entire block, including the MAC. For stream ciphers\n that do not use a synchronization vector (such as RC4), the stream\n cipher state from the end of one record is simply used on the\n subsequent packet. If the cipher suite is TLS_NULL_WITH_NULL_NULL,\n encryption consists of the identity operation (i.e., the data is not\n encrypted, and the MAC size is zero, implying that no MAC is used).\n For both null and stream ciphers, TLSCiphertext.length is\n TLSCompressed.length plus SecurityParameters.mac_length.\n\n6.2.3.2. CBC Block Cipher\n\n For block ciphers (such as 3DES or AES), the encryption and MAC\n functions convert TLSCompressed.fragment structures to and from block\n TLSCiphertext.fragment structures.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 22]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n opaque IV[SecurityParameters.record_iv_length];\n block-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n uint8 padding[GenericBlockCipher.padding_length];\n uint8 padding_length;\n };\n } GenericBlockCipher;\n\n The MAC is generated as described in Section 6.2.3.1.\n\n IV\n The Initialization Vector (IV) SHOULD be chosen at random, and\n MUST be unpredictable. Note that in versions of TLS prior to 1.1,\n there was no IV field, and the last ciphertext block of the\n previous record (the \"CBC residue\") was used as the IV. This was\n changed to prevent the attacks described in [CBCATT]. For block\n ciphers, the IV length is of length\n SecurityParameters.record_iv_length, which is equal to the\n SecurityParameters.block_size.\n\n padding\n Padding that is added to force the length of the plaintext to be\n an integral multiple of the block cipher's block length. The\n padding MAY be any length up to 255 bytes, as long as it results\n in the TLSCiphertext.length being an integral multiple of the\n block length. Lengths longer than necessary might be desirable to\n frustrate attacks on a protocol that are based on analysis of the\n lengths of exchanged messages. Each uint8 in the padding data\n vector MUST be filled with the padding length value. The receiver\n MUST check this padding and MUST use the bad_record_mac alert to\n indicate padding errors.\n\n padding_length\n The padding length MUST be such that the total size of the\n GenericBlockCipher structure is a multiple of the cipher's block\n length. Legal values range from zero to 255, inclusive. This\n length specifies the length of the padding field exclusive of the\n padding_length field itself.\n\n The encrypted data length (TLSCiphertext.length) is one more than the\n sum of SecurityParameters.block_length, TLSCompressed.length,\n SecurityParameters.mac_length, and padding_length.\n\n Example: If the block length is 8 bytes, the content length\n (TLSCompressed.length) is 61 bytes, and the MAC length is 20 bytes,\n then the length before padding is 82 bytes (this does not include the\n\n\n\nDierks & Rescorla Standards Track [Page 23]\n\f\nRFC 5246 TLS August 2008\n\n\n IV. Thus, the padding length modulo 8 must be equal to 6 in order to\n make the total length an even multiple of 8 bytes (the block length).\n The padding length can be 6, 14, 22, and so on, through 254. If the\n padding length were the minimum necessary, 6, the padding would be 6\n bytes, each containing the value 6. Thus, the last 8 octets of the\n GenericBlockCipher before block encryption would be xx 06 06 06 06 06\n 06 06, where xx is the last octet of the MAC.\n\n Note: With block ciphers in CBC mode (Cipher Block Chaining), it is\n critical that the entire plaintext of the record be known before any\n ciphertext is transmitted. Otherwise, it is possible for the\n attacker to mount the attack described in [CBCATT].\n\n Implementation note: Canvel et al. [CBCTIME] have demonstrated a\n timing attack on CBC padding based on the time required to compute\n the MAC. In order to defend against this attack, implementations\n MUST ensure that record processing time is essentially the same\n whether or not the padding is correct. In general, the best way to\n do this is to compute the MAC even if the padding is incorrect, and\n only then reject the packet. For instance, if the pad appears to be\n incorrect, the implementation might assume a zero-length pad and then\n compute the MAC. This leaves a small timing channel, since MAC\n performance depends to some extent on the size of the data fragment,\n but it is not believed to be large enough to be exploitable, due to\n the large block size of existing MACs and the small size of the\n timing signal.\n\n6.2.3.3. AEAD Ciphers\n\n For AEAD [AEAD] ciphers (such as [CCM] or [GCM]), the AEAD function\n converts TLSCompressed.fragment structures to and from AEAD\n TLSCiphertext.fragment structures.\n\n struct {\n opaque nonce_explicit[SecurityParameters.record_iv_length];\n aead-ciphered struct {\n opaque content[TLSCompressed.length];\n };\n } GenericAEADCipher;\n\n AEAD ciphers take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [AEAD]. The key is either the\n client_write_key or the server_write_key. No MAC key is used.\n\n Each AEAD cipher suite MUST specify how the nonce supplied to the\n AEAD operation is constructed, and what is the length of the\n GenericAEADCipher.nonce_explicit part. In many cases, it is\n\n\n\nDierks & Rescorla Standards Track [Page 24]\n\f\nRFC 5246 TLS August 2008\n\n\n appropriate to use the partially implicit nonce technique described\n in Section 3.2.1 of [AEAD]; with record_iv_length being the length of\n the explicit part. In this case, the implicit part SHOULD be derived\n from key_block as client_write_iv and server_write_iv (as described\n in Section 6.3), and the explicit part is included in\n GenericAEAEDCipher.nonce_explicit.\n\n The plaintext is the TLSCompressed.fragment.\n\n The additional authenticated data, which we denote as\n additional_data, is defined as follows:\n\n additional_data = seq_num + TLSCompressed.type +\n TLSCompressed.version + TLSCompressed.length;\n\n where \"+\" denotes concatenation.\n\n The aead_output consists of the ciphertext output by the AEAD\n encryption operation. The length will generally be larger than\n TLSCompressed.length, but by an amount that varies with the AEAD\n cipher. Since the ciphers might incorporate padding, the amount of\n overhead could vary with different TLSCompressed.length values. Each\n AEAD cipher MUST NOT produce an expansion of greater than 1024 bytes.\n Symbolically,\n\n AEADEncrypted = AEAD-Encrypt(write_key, nonce, plaintext,\n additional_data)\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, the \"additional_data\", and the AEADEncrypted value. The\n output is either the plaintext or an error indicating that the\n decryption failed. There is no separate integrity check. That is:\n\n TLSCompressed.fragment = AEAD-Decrypt(write_key, nonce,\n AEADEncrypted,\n additional_data)\n\n If the decryption fails, a fatal bad_record_mac alert MUST be\n generated.\n\n6.3. Key Calculation\n\n The Record Protocol requires an algorithm to generate keys required\n by the current connection state (see Appendix A.6) from the security\n parameters provided by the handshake protocol.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 25]\n\f\nRFC 5246 TLS August 2008\n\n\n The master secret is expanded into a sequence of secure bytes, which\n is then split to a client write MAC key, a server write MAC key, a\n client write encryption key, and a server write encryption key. Each\n of these is generated from the byte sequence in that order. Unused\n values are empty. Some AEAD ciphers may additionally require a\n client write IV and a server write IV (see Section 6.2.3.3).\n\n When keys and MAC keys are generated, the master secret is used as an\n entropy source.\n\n To generate the key material, compute\n\n key_block = PRF(SecurityParameters.master_secret,\n \"key expansion\",\n SecurityParameters.server_random +\n SecurityParameters.client_random);\n\n until enough output has been generated. Then, the key_block is\n partitioned as follows:\n\n client_write_MAC_key[SecurityParameters.mac_key_length]\n server_write_MAC_key[SecurityParameters.mac_key_length]\n client_write_key[SecurityParameters.enc_key_length]\n server_write_key[SecurityParameters.enc_key_length]\n client_write_IV[SecurityParameters.fixed_iv_length]\n server_write_IV[SecurityParameters.fixed_iv_length]\n\n Currently, the client_write_IV and server_write_IV are only generated\n for implicit nonce techniques as described in Section 3.2.1 of\n [AEAD].\n\n Implementation note: The currently defined cipher suite which\n requires the most material is AES_256_CBC_SHA256. It requires 2 x 32\n byte keys and 2 x 32 byte MAC keys, for a total 128 bytes of key\n material.\n\n7. The TLS Handshaking Protocols\n\n TLS has three subprotocols that are used to allow peers to agree upon\n security parameters for the record layer, to authenticate themselves,\n to instantiate negotiated security parameters, and to report error\n conditions to each other.\n\n The Handshake Protocol is responsible for negotiating a session,\n which consists of the following items:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 26]\n\f\nRFC 5246 TLS August 2008\n\n\n session identifier\n An arbitrary byte sequence chosen by the server to identify an\n active or resumable session state.\n\n peer certificate\n X509v3 [PKIX] certificate of the peer. This element of the state\n may be null.\n\n compression method\n The algorithm used to compress data prior to encryption.\n\n cipher spec\n Specifies the pseudorandom function (PRF) used to generate keying\n material, the bulk data encryption algorithm (such as null, AES,\n etc.) and the MAC algorithm (such as HMAC-SHA1). It also defines\n cryptographic attributes such as the mac_length. (See Appendix\n A.6 for formal definition.)\n\n master secret\n 48-byte secret shared between the client and server.\n\n is resumable\n A flag indicating whether the session can be used to initiate new\n connections.\n\n These items are then used to create security parameters for use by\n the record layer when protecting application data. Many connections\n can be instantiated using the same session through the resumption\n feature of the TLS Handshake Protocol.\n\n7.1. Change Cipher Spec Protocol\n\n The change cipher spec protocol exists to signal transitions in\n ciphering strategies. The protocol consists of a single message,\n which is encrypted and compressed under the current (not the pending)\n connection state. The message consists of a single byte of value 1.\n\n struct {\n enum { change_cipher_spec(1), (255) } type;\n } ChangeCipherSpec;\n\n The ChangeCipherSpec message is sent by both the client and the\n server to notify the receiving party that subsequent records will be\n protected under the newly negotiated CipherSpec and keys. Reception\n of this message causes the receiver to instruct the record layer to\n immediately copy the read pending state into the read current state.\n Immediately after sending this message, the sender MUST instruct the\n record layer to make the write pending state the write active state.\n\n\n\nDierks & Rescorla Standards Track [Page 27]\n\f\nRFC 5246 TLS August 2008\n\n\n (See Section 6.1.) The ChangeCipherSpec message is sent during the\n handshake after the security parameters have been agreed upon, but\n before the verifying Finished message is sent.\n\n Note: If a rehandshake occurs while data is flowing on a connection,\n the communicating parties may continue to send data using the old\n CipherSpec. However, once the ChangeCipherSpec has been sent, the\n new CipherSpec MUST be used. The first side to send the\n ChangeCipherSpec does not know that the other side has finished\n computing the new keying material (e.g., if it has to perform a\n time-consuming public key operation). Thus, a small window of time,\n during which the recipient must buffer the data, MAY exist. In\n practice, with modern machines this interval is likely to be fairly\n short.\n\n7.2. Alert Protocol\n\n One of the content types supported by the TLS record layer is the\n alert type. Alert messages convey the severity of the message\n (warning or fatal) and a description of the alert. Alert messages\n with a level of fatal result in the immediate termination of the\n connection. In this case, other connections corresponding to the\n session may continue, but the session identifier MUST be invalidated,\n preventing the failed session from being used to establish new\n connections. Like other messages, alert messages are encrypted and\n compressed, as specified by the current connection state.\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n\n\n\nDierks & Rescorla Standards Track [Page 28]\n\f\nRFC 5246 TLS August 2008\n\n\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n7.2.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack. Either party may\n initiate the exchange of closing messages.\n\n close_notify\n This message notifies the recipient that the sender will not send\n any more messages on this connection. Note that as of TLS 1.1,\n failure to properly close a connection no longer requires that a\n session not be resumed. This is a change from TLS 1.0 to conform\n with widespread implementation practice.\n\n Either party may initiate a close by sending a close_notify alert.\n Any data received after a closure alert is ignored.\n\n Unless some other fatal alert has been transmitted, each party is\n required to send a close_notify alert before closing the write side\n of the connection. The other party MUST respond with a close_notify\n alert of its own and close down the connection immediately,\n discarding any pending writes. It is not required for the initiator\n of the close to wait for the responding close_notify alert before\n closing the read side of the connection.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation must receive the responding\n close_notify alert before indicating to the application layer that\n the TLS connection has ended. If the application protocol will not\n transfer any additional data, but will only close the underlying\n transport connection, then the implementation MAY choose to close the\n transport without waiting for the responding close_notify. No part\n\n\n\n\nDierks & Rescorla Standards Track [Page 29]\n\f\nRFC 5246 TLS August 2008\n\n\n of this standard should be taken to dictate the manner in which a\n usage profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing a connection reliably delivers\n pending data before destroying the transport.\n\n7.2.2. Error Alerts\n\n Error handling in the TLS Handshake protocol is very simple. When an\n error is detected, the detecting party sends a message to the other\n party. Upon transmission or receipt of a fatal alert message, both\n parties immediately close the connection. Servers and clients MUST\n forget any session-identifiers, keys, and secrets associated with a\n failed connection. Thus, any connection terminated with a fatal\n alert MUST NOT be resumed.\n\n Whenever an implementation encounters a condition which is defined as\n a fatal alert, it MUST send the appropriate alert prior to closing\n the connection. For all errors where an alert level is not\n explicitly specified, the sending party MAY determine at its\n discretion whether to treat this as a fatal error or not. If the\n implementation chooses to send an alert but intends to close the\n connection immediately afterwards, it MUST send that alert at the\n fatal alert level.\n\n If an alert with a level of warning is sent and received, generally\n the connection can continue normally. If the receiving party decides\n not to proceed with the connection (e.g., after having received a\n no_renegotiation alert that it is not willing to accept), it SHOULD\n send a fatal alert to terminate the connection. Given this, the\n sending party cannot, in general, know how the receiving party will\n behave. Therefore, warning alerts are not very useful when the\n sending party wants to continue the connection, and thus are\n sometimes omitted. For example, if a peer decides to accept an\n expired certificate (perhaps after confirming this with the user) and\n wants to continue the connection, it would not generally send a\n certificate_expired alert.\n\n The following error alerts are defined:\n\n unexpected_message\n An inappropriate message was received. This alert is always fatal\n and should never be observed in communication between proper\n implementations.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 30]\n\f\nRFC 5246 TLS August 2008\n\n\n bad_record_mac\n This alert is returned if a record is received with an incorrect\n MAC. This alert also MUST be returned if an alert is sent because\n a TLSCiphertext decrypted in an invalid way: either it wasn't an\n even multiple of the block length, or its padding values, when\n checked, weren't correct. This message is always fatal and should\n never be observed in communication between proper implementations\n (except when messages were corrupted in the network).\n\n decryption_failed_RESERVED\n This alert was used in some earlier versions of TLS, and may have\n permitted certain attacks against the CBC mode [CBCATT]. It MUST\n NOT be sent by compliant implementations.\n\n record_overflow\n A TLSCiphertext record was received that had a length more than\n 2^14+2048 bytes, or a record decrypted to a TLSCompressed record\n with more than 2^14+1024 bytes. This message is always fatal and\n should never be observed in communication between proper\n implementations (except when messages were corrupted in the\n network).\n\n decompression_failure\n The decompression function received improper input (e.g., data\n that would expand to excessive length). This message is always\n fatal and should never be observed in communication between proper\n implementations.\n\n handshake_failure\n Reception of a handshake_failure alert message indicates that the\n sender was unable to negotiate an acceptable set of security\n parameters given the options available. This is a fatal error.\n\n no_certificate_RESERVED\n This alert was used in SSLv3 but not any version of TLS. It MUST\n NOT be sent by compliant implementations.\n\n bad_certificate\n A certificate was corrupt, contained signatures that did not\n verify correctly, etc.\n\n unsupported_certificate\n A certificate was of an unsupported type.\n\n certificate_revoked\n A certificate was revoked by its signer.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 31]\n\f\nRFC 5246 TLS August 2008\n\n\n certificate_expired\n A certificate has expired or is not currently valid.\n\n certificate_unknown\n Some other (unspecified) issue arose in processing the\n certificate, rendering it unacceptable.\n\n illegal_parameter\n A field in the handshake was out of range or inconsistent with\n other fields. This message is always fatal.\n\n unknown_ca\n A valid certificate chain or partial chain was received, but the\n certificate was not accepted because the CA certificate could not\n be located or couldn't be matched with a known, trusted CA. This\n message is always fatal.\n\n access_denied\n A valid certificate was received, but when access control was\n applied, the sender decided not to proceed with negotiation. This\n message is always fatal.\n\n decode_error\n A message could not be decoded because some field was out of the\n specified range or the length of the message was incorrect. This\n message is always fatal and should never be observed in\n communication between proper implementations (except when messages\n were corrupted in the network).\n\n decrypt_error\n A handshake cryptographic operation failed, including being unable\n to correctly verify a signature or validate a Finished message.\n This message is always fatal.\n\n export_restriction_RESERVED\n This alert was used in some earlier versions of TLS. It MUST NOT\n be sent by compliant implementations.\n\n protocol_version\n The protocol version the client has attempted to negotiate is\n recognized but not supported. (For example, old protocol versions\n might be avoided for security reasons.) This message is always\n fatal.\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 32]\n\f\nRFC 5246 TLS August 2008\n\n\n insufficient_security\n Returned instead of handshake_failure when a negotiation has\n failed specifically because the server requires ciphers more\n secure than those supported by the client. This message is always\n fatal.\n\n internal_error\n An internal error unrelated to the peer or the correctness of the\n protocol (such as a memory allocation failure) makes it impossible\n to continue. This message is always fatal.\n\n\n\n===== DOCUMENT =====\n\n user_canceled\n This handshake is being canceled for some reason unrelated to a\n protocol failure. If the user cancels an operation after the\n handshake is complete, just closing the connection by sending a\n close_notify is more appropriate. This alert should be followed\n by a close_notify. This message is generally a warning.\n\n no_renegotiation\n Sent by the client in response to a hello request or by the server\n in response to a client hello after initial handshaking. Either\n of these would normally lead to renegotiation; when that is not\n appropriate, the recipient should respond with this alert. At\n that point, the original requester can decide whether to proceed\n with the connection. One case where this would be appropriate is\n where a server has spawned a process to satisfy a request; the\n process might receive security parameters (key length,\n authentication, etc.) at startup, and it might be difficult to\n communicate changes to these parameters after that point. This\n message is always a warning.\n\n unsupported_extension\n sent by clients that receive an extended server hello containing\n an extension that they did not put in the corresponding client\n hello. This message is always fatal.\n\n New Alert values are assigned by IANA as described in Section 12.\n\n7.3. Handshake Protocol Overview\n\n The cryptographic parameters of the session state are produced by the\n TLS Handshake Protocol, which operates on top of the TLS record\n layer. When a TLS client and server first start communicating, they\n agree on a protocol version, select cryptographic algorithms,\n optionally authenticate each other, and use public-key encryption\n techniques to generate shared secrets.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 33]\n\f\nRFC 5246 TLS August 2008\n\n\n The TLS Handshake Protocol involves the following steps:\n\n - Exchange hello messages to agree on algorithms, exchange random\n values, and check for session resumption.\n\n - Exchange the necessary cryptographic parameters to allow the\n client and server to agree on a premaster secret.\n\n - Exchange certificates and cryptographic information to allow the\n client and server to authenticate themselves.\n\n - Generate a master secret from the premaster secret and exchanged\n random values.\n\n - Provide security parameters to the record layer.\n\n - Allow the client and server to verify that their peer has\n calculated the same security parameters and that the handshake\n occurred without tampering by an attacker.\n\n Note that higher layers should not be overly reliant on whether TLS\n always negotiates the strongest possible connection between two\n peers. There are a number of ways in which a man-in-the-middle\n attacker can attempt to make two entities drop down to the least\n secure method they support. The protocol has been designed to\n minimize this risk, but there are still attacks available: for\n example, an attacker could block access to the port a secure service\n runs on, or attempt to get the peers to negotiate an unauthenticated\n connection. The fundamental rule is that higher levels must be\n cognizant of what their security requirements are and never transmit\n information over a channel less secure than what they require. The\n TLS protocol is secure in that any cipher suite offers its promised\n level of security: if you negotiate 3DES with a 1024-bit RSA key\n exchange with a host whose certificate you have verified, you can\n expect to be that secure.\n\n These goals are achieved by the handshake protocol, which can be\n summarized as follows: The client sends a ClientHello message to\n which the server must respond with a ServerHello message, or else a\n fatal error will occur and the connection will fail. The ClientHello\n and ServerHello are used to establish security enhancement\n capabilities between client and server. The ClientHello and\n ServerHello establish the following attributes: Protocol Version,\n Session ID, Cipher Suite, and Compression Method. Additionally, two\n random values are generated and exchanged: ClientHello.random and\n ServerHello.random.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 34]\n\f\nRFC 5246 TLS August 2008\n\n\n The actual key exchange uses up to four messages: the server\n Certificate, the ServerKeyExchange, the client Certificate, and the\n ClientKeyExchange. New key exchange methods can be created by\n specifying a format for these messages and by defining the use of the\n messages to allow the client and server to agree upon a shared\n secret. This secret MUST be quite long; currently defined key\n exchange methods exchange secrets that range from 46 bytes upwards.\n\n Following the hello messages, the server will send its certificate in\n a Certificate message if it is to be authenticated. Additionally, a\n ServerKeyExchange message may be sent, if it is required (e.g., if\n the server has no certificate, or if its certificate is for signing\n only). If the server is authenticated, it may request a certificate\n from the client, if that is appropriate to the cipher suite selected.\n Next, the server will send the ServerHelloDone message, indicating\n that the hello-message phase of the handshake is complete. The\n server will then wait for a client response. If the server has sent\n a CertificateRequest message, the client MUST send the Certificate\n message. The ClientKeyExchange message is now sent, and the content\n of that message will depend on the public key algorithm selected\n between the ClientHello and the ServerHello. If the client has sent\n a certificate with signing ability, a digitally-signed\n CertificateVerify message is sent to explicitly verify possession of\n the private key in the certificate.\n\n At this point, a ChangeCipherSpec message is sent by the client, and\n the client copies the pending Cipher Spec into the current Cipher\n Spec. The client then immediately sends the Finished message under\n the new algorithms, keys, and secrets. In response, the server will\n send its own ChangeCipherSpec message, transfer the pending to the\n current Cipher Spec, and send its Finished message under the new\n Cipher Spec. At this point, the handshake is complete, and the\n client and server may begin to exchange application layer data. (See\n flow chart below.) Application data MUST NOT be sent prior to the\n completion of the first handshake (before a cipher suite other than\n TLS_NULL_WITH_NULL_NULL is established).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 35]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1. Message flow for a full handshake\n\n * Indicates optional or situation-dependent messages that are not\n always sent.\n\n Note: To help avoid pipeline stalls, ChangeCipherSpec is an\n independent TLS protocol content type, and is not actually a TLS\n handshake message.\n\n When the client and server decide to resume a previous session or\n duplicate an existing session (instead of negotiating new security\n parameters), the message flow is as follows:\n\n The client sends a ClientHello using the Session ID of the session to\n be resumed. The server then checks its session cache for a match.\n If a match is found, and the server is willing to re-establish the\n connection under the specified session state, it will send a\n ServerHello with the same Session ID value. At this point, both\n client and server MUST send ChangeCipherSpec messages and proceed\n directly to Finished messages. Once the re-establishment is\n complete, the client and server MAY begin to exchange application\n layer data. (See flow chart below.) If a Session ID match is not\n found, the server generates a new session ID, and the TLS client and\n server perform a full handshake.\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 36]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2. Message flow for an abbreviated handshake\n\n The contents and significance of each message will be presented in\n detail in the following sections.\n\n7.4. Handshake Protocol\n\n The TLS Handshake Protocol is one of the defined higher-level clients\n of the TLS Record Protocol. This protocol is used to negotiate the\n secure attributes of a session. Handshake messages are supplied to\n the TLS record layer, where they are encapsulated within one or more\n TLSPlaintext structures, which are processed and transmitted as\n specified by the current active session state.\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n } body;\n } Handshake;\n\n\n\nDierks & Rescorla Standards Track [Page 37]\n\f\nRFC 5246 TLS August 2008\n\n\n The handshake protocol messages are presented below in the order they\n MUST be sent; sending handshake messages in an unexpected order\n results in a fatal error. Unneeded handshake messages can be\n omitted, however. Note one exception to the ordering: the\n Certificate message is used twice in the handshake (from server to\n client, then from client to server), but described only in its first\n position. The one message that is not bound by these ordering rules\n is the HelloRequest message, which can be sent at any time, but which\n SHOULD be ignored by the client if it arrives in the middle of a\n handshake.\n\n New handshake message types are assigned by IANA as described in\n Section 12.\n\n7.4.1. Hello Messages\n\n The hello phase messages are used to exchange security enhancement\n capabilities between the client and server. When a new session\n begins, the record layer's connection state encryption, hash, and\n compression algorithms are initialized to null. The current\n connection state is used for renegotiation messages.\n\n7.4.1.1. Hello Request\n\n When this message will be sent:\n\n The HelloRequest message MAY be sent by the server at any time.\n\n Meaning of this message:\n\n HelloRequest is a simple notification that the client should begin\n the negotiation process anew. In response, the client should send\n a ClientHello message when convenient. This message is not\n intended to establish which side is the client or server but\n merely to initiate a new negotiation. Servers SHOULD NOT send a\n HelloRequest immediately upon the client's initial connection. It\n is the client's job to send a ClientHello at that time.\n\n This message will be ignored by the client if the client is\n currently negotiating a session. This message MAY be ignored by\n the client if it does not wish to renegotiate a session, or the\n client may, if it wishes, respond with a no_renegotiation alert.\n Since handshake messages are intended to have transmission\n precedence over application data, it is expected that the\n negotiation will begin before no more than a few records are\n received from the client. If the server sends a HelloRequest but\n does not receive a ClientHello in response, it may close the\n connection with a fatal alert.\n\n\n\nDierks & Rescorla Standards Track [Page 38]\n\f\nRFC 5246 TLS August 2008\n\n\n After sending a HelloRequest, servers SHOULD NOT repeat the\n request until the subsequent handshake negotiation is complete.\n\n Structure of this message:\n\n struct { } HelloRequest;\n\n This message MUST NOT be included in the message hashes that are\n maintained throughout the handshake and used in the Finished messages\n and the certificate verify message.\n\n7.4.1.2. Client Hello\n\n When this message will be sent:\n\n When a client first connects to a server, it is required to send\n the ClientHello as its first message. The client can also send a\n ClientHello in response to a HelloRequest or on its own initiative\n in order to renegotiate the security parameters in an existing\n connection.\n\n Structure of this message:\n\n The ClientHello message includes a random structure, which is used\n later in the protocol.\n\n struct {\n uint32 gmt_unix_time;\n opaque random_bytes[28];\n } Random;\n\n gmt_unix_time\n The current time and date in standard UNIX 32-bit format\n (seconds since the midnight starting Jan 1, 1970, UTC, ignoring\n leap seconds) according to the sender's internal clock. Clocks\n are not required to be set correctly by the basic TLS protocol;\n higher-level or application protocols may define additional\n requirements. Note that, for historical reasons, the data\n element is named using GMT, the predecessor of the current\n worldwide time base, UTC.\n\n random_bytes\n 28 bytes generated by a secure random number generator.\n\n The ClientHello message includes a variable-length session\n identifier. If not empty, the value identifies a session between the\n same client and server whose security parameters the client wishes to\n reuse. The session identifier MAY be from an earlier connection,\n\n\n\nDierks & Rescorla Standards Track [Page 39]\n\f\nRFC 5246 TLS August 2008\n\n\n this connection, or from another currently active connection. The\n second option is useful if the client only wishes to update the\n random structures and derived values of a connection, and the third\n option makes it possible to establish several independent secure\n connections without repeating the full handshake protocol. These\n independent connections may occur sequentially or simultaneously; a\n SessionID becomes valid when the handshake negotiating it completes\n with the exchange of Finished messages and persists until it is\n removed due to aging or because a fatal error was encountered on a\n connection associated with the session. The actual contents of the\n SessionID are defined by the server.\n\n opaque SessionID<0..32>;\n\n Warning: Because the SessionID is transmitted without encryption or\n immediate MAC protection, servers MUST NOT place confidential\n information in session identifiers or let the contents of fake\n session identifiers cause any breach of security. (Note that the\n content of the handshake as a whole, including the SessionID, is\n protected by the Finished messages exchanged at the end of the\n handshake.)\n\n The cipher suite list, passed from the client to the server in the\n ClientHello message, contains the combinations of cryptographic\n algorithms supported by the client in order of the client's\n preference (favorite choice first). Each cipher suite defines a key\n exchange algorithm, a bulk encryption algorithm (including secret key\n length), a MAC algorithm, and a PRF. The server will select a cipher\n suite or, if no acceptable choices are presented, return a handshake\n failure alert and close the connection. If the list contains cipher\n suites the server does not recognize, support, or wish to use, the\n server MUST ignore those cipher suites, and process the remaining\n ones as usual.\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n The ClientHello includes a list of compression algorithms supported\n by the client, ordered according to the client's preference.\n\n enum { null(0), (255) } CompressionMethod;\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 40]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ProtocolVersion client_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suites<2..2^16-2>;\n CompressionMethod compression_methods<1..2^8-1>;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ClientHello;\n\n TLS allows extensions to follow the compression_methods field in an\n extensions block. The presence of extensions can be detected by\n determining whether there are bytes following the compression_methods\n at the end of the ClientHello. Note that this method of detecting\n optional data differs from the normal TLS method of having a\n variable-length field, but it is used for compatibility with TLS\n before extensions were defined.\n\n client_version\n The version of the TLS protocol by which the client wishes to\n communicate during this session. This SHOULD be the latest\n (highest valued) version supported by the client. For this\n version of the specification, the version will be 3.3 (see\n Appendix E for details about backward compatibility).\n\n random\n A client-generated random structure.\n\n session_id\n The ID of a session the client wishes to use for this connection.\n This field is empty if no session_id is available, or if the\n client wishes to generate new security parameters.\n\n cipher_suites\n This is a list of the cryptographic options supported by the\n client, with the client's first preference first. If the\n session_id field is not empty (implying a session resumption\n request), this vector MUST include at least the cipher_suite from\n that session. Values are defined in Appendix A.5.\n\n compression_methods\n This is a list of the compression methods supported by the client,\n sorted by client preference. If the session_id field is not empty\n (implying a session resumption request), it MUST include the\n\n\n\nDierks & Rescorla Standards Track [Page 41]\n\f\nRFC 5246 TLS August 2008\n\n\n compression_method from that session. This vector MUST contain,\n and all implementations MUST support, CompressionMethod.null.\n Thus, a client and server will always be able to agree on a\n compression method.\n\n extensions\n Clients MAY request extended functionality from servers by sending\n data in the extensions field. The actual \"Extension\" format is\n defined in Section 7.4.1.4.\n\n In the event that a client requests additional functionality using\n extensions, and this functionality is not supplied by the server, the\n client MAY abort the handshake. A server MUST accept ClientHello\n messages both with and without the extensions field, and (as for all\n other messages) it MUST check that the amount of data in the message\n precisely matches one of these formats; if not, then it MUST send a\n fatal \"decode_error\" alert.\n\n After sending the ClientHello message, the client waits for a\n ServerHello message. Any handshake message returned by the server,\n except for a HelloRequest, is treated as a fatal error.\n\n7.4.1.3. Server Hello\n\n When this message will be sent:\n\n The server will send this message in response to a ClientHello\n message when it was able to find an acceptable set of algorithms.\n If it cannot find such a match, it will respond with a handshake\n failure alert.\n\n Structure of this message:\n\n struct {\n ProtocolVersion server_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ServerHello;\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 42]\n\f\nRFC 5246 TLS August 2008\n\n\n The presence of extensions can be detected by determining whether\n there are bytes following the compression_method field at the end of\n the ServerHello.\n\n server_version\n This field will contain the lower of that suggested by the client\n in the client hello and the highest supported by the server. For\n this version of the specification, the version is 3.3. (See\n Appendix E for details about backward compatibility.)\n\n random\n This structure is generated by the server and MUST be\n independently generated from the ClientHello.random.\n\n session_id\n This is the identity of the session corresponding to this\n connection. If the ClientHello.session_id was non-empty, the\n server will look in its session cache for a match. If a match is\n found and the server is willing to establish the new connection\n using the specified session state, the server will respond with\n the same value as was supplied by the client. This indicates a\n resumed session and dictates that the parties must proceed\n directly to the Finished messages. Otherwise, this field will\n contain a different value identifying the new session. The server\n may return an empty session_id to indicate that the session will\n not be cached and therefore cannot be resumed. If a session is\n resumed, it must be resumed using the same cipher suite it was\n originally negotiated with. Note that there is no requirement\n that the server resume any session even if it had formerly\n provided a session_id. Clients MUST be prepared to do a full\n negotiation -- including negotiating new cipher suites -- during\n any handshake.\n\n cipher_suite\n The single cipher suite selected by the server from the list in\n ClientHello.cipher_suites. For resumed sessions, this field is\n the value from the state of the session being resumed.\n\n compression_method\n The single compression algorithm selected by the server from the\n list in ClientHello.compression_methods. For resumed sessions,\n this field is the value from the resumed session state.\n\n extensions\n A list of extensions. Note that only extensions offered by the\n client can appear in the server's list.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 43]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.1.4. Hello Extensions\n\n The extension format is:\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n signature_algorithms(13), (65535)\n } ExtensionType;\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The initial set of extensions is defined in a companion document\n [TLSEXT]. The list of extension types is maintained by IANA as\n described in Section 12.\n\n An extension type MUST NOT appear in the ServerHello unless the same\n extension type appeared in the corresponding ClientHello. If a\n client receives an extension type in ServerHello that it did not\n request in the associated ClientHello, it MUST abort the handshake\n with an unsupported_extension fatal alert.\n\n Nonetheless, \"server-oriented\" extensions may be provided in the\n future within this framework. Such an extension (say, of type x)\n would require the client to first send an extension of type x in a\n ClientHello with empty extension_data to indicate that it supports\n the extension type. In this case, the client is offering the\n capability to understand the extension type, and the server is taking\n the client up on its offer.\n\n When multiple extensions of different types are present in the\n ClientHello or ServerHello messages, the extensions MAY appear in any\n order. There MUST NOT be more than one extension of the same type.\n\n Finally, note that extensions can be sent both when starting a new\n session and when requesting session resumption. Indeed, a client\n that requests session resumption does not in general know whether the\n server will accept this request, and therefore it SHOULD send the\n same extensions as it would send if it were not attempting\n resumption.\n\n\n\nDierks & Rescorla Standards Track [Page 44]\n\f\nRFC 5246 TLS August 2008\n\n\n In general, the specification of each extension type needs to\n describe the effect of the extension both during full handshake and\n session resumption. Most current TLS extensions are relevant only\n when a session is initiated: when an older session is resumed, the\n server does not process these extensions in Client Hello, and does\n not include them in Server Hello. However, some extensions may\n specify different behavior during session resumption.\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions, and some are simply refusals to support particular\n features. In general, error alerts should be used for the former,\n and a field in the server extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem.\n\n Often the fact that the extension fields are included in the\n inputs to the Finished message hashes will be sufficient, but\n extreme care is needed when the extension changes the meaning of\n messages sent in the handshake phase. Designers and implementors\n should be aware of the fact that until the handshake has been\n authenticated, active attackers can modify messages and insert,\n remove, or replace extensions.\n\n - It would be technically possible to use extensions to change major\n aspects of the design of TLS; for example the design of cipher\n suite negotiation. This is not recommended; it would be more\n appropriate to define a new version of TLS -- particularly since\n the TLS handshake algorithms have specific protection against\n version rollback attacks based on the version number, and the\n possibility of version rollback should be a significant\n consideration in any major design change.\n\n7.4.1.4.1. Signature Algorithms\n\n The client uses the \"signature_algorithms\" extension to indicate to\n the server which signature/hash algorithm pairs may be used in\n digital signatures. The \"extension_data\" field of this extension\n contains a \"supported_signature_algorithms\" value.\n\n\n\nDierks & Rescorla Standards Track [Page 45]\n\f\nRFC 5246 TLS August 2008\n\n\n enum {\n none(0), md5(1), sha1(2), sha224(3), sha256(4), sha384(5),\n sha512(6), (255)\n } HashAlgorithm;\n\n enum { anonymous(0), rsa(1), dsa(2), ecdsa(3), (255) }\n SignatureAlgorithm;\n\n struct {\n HashAlgorithm hash;\n SignatureAlgorithm signature;\n } SignatureAndHashAlgorithm;\n\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2..2^16-2>;\n\n Each SignatureAndHashAlgorithm value lists a single hash/signature\n pair that the client is willing to verify. The values are indicated\n in descending order of preference.\n\n Note: Because not all signature algorithms and hash algorithms may be\n accepted by an implementation (e.g., DSA with SHA-1, but not\n SHA-256), algorithms here are listed in pairs.\n\n hash\n This field indicates the hash algorithm which may be used. The\n values indicate support for unhashed data, MD5 [MD5], SHA-1,\n SHA-224, SHA-256, SHA-384, and SHA-512 [SHS], respectively. The\n \"none\" value is provided for future extensibility, in case of a\n signature algorithm which does not require hashing before signing.\n\n signature\n This field indicates the signature algorithm that may be used.\n The values indicate anonymous signatures, RSASSA-PKCS1-v1_5\n [PKCS1] and DSA [DSS], and ECDSA [ECDSA], respectively. The\n \"anonymous\" value is meaningless in this context but used in\n Section 7.4.3. It MUST NOT appear in this extension.\n\n The semantics of this extension are somewhat complicated because the\n cipher suite indicates permissible signature algorithms but not hash\n algorithms. Sections 7.4.2 and 7.4.3 describe the appropriate rules.\n\n If the client supports only the default hash and signature algorithms\n (listed in this section), it MAY omit the signature_algorithms\n extension. If the client does not support the default algorithms, or\n supports other hash and signature algorithms (and it is willing to\n use them for verifying messages sent by the server, i.e., server\n certificates and server key exchange), it MUST send the\n\n\n\nDierks & Rescorla Standards Track [Page 46]\n\f\nRFC 5246 TLS August 2008\n\n\n signature_algorithms extension, listing the algorithms it is willing\n to accept.\n\n If the client does not send the signature_algorithms extension, the\n server MUST do the following:\n\n - If the negotiated key exchange algorithm is one of (RSA, DHE_RSA,\n DH_RSA, RSA_PSK, ECDH_RSA, ECDHE_RSA), behave as if client had\n sent the value {sha1,rsa}.\n\n - If the negotiated key exchange algorithm is one of (DHE_DSS,\n DH_DSS), behave as if the client had sent the value {sha1,dsa}.\n\n - If the negotiated key exchange algorithm is one of (ECDH_ECDSA,\n ECDHE_ECDSA), behave as if the client had sent value {sha1,ecdsa}.\n\n Note: this is a change from TLS 1.1 where there are no explicit\n rules, but as a practical matter one can assume that the peer\n supports MD5 and SHA-1.\n\n Note: this extension is not meaningful for TLS versions prior to 1.2.\n Clients MUST NOT offer it if they are offering prior versions.\n However, even if clients do offer it, the rules specified in [TLSEXT]\n require servers to ignore extensions they do not understand.\n\n Servers MUST NOT send this extension. TLS servers MUST support\n receiving this extension.\n\n When performing session resumption, this extension is not included in\n Server Hello, and the server ignores the extension in Client Hello\n (if present).\n\n7.4.2. Server Certificate\n\n When this message will be sent:\n\n The server MUST send a Certificate message whenever the agreed-\n upon key exchange method uses certificates for authentication\n (this includes all key exchange methods defined in this document\n except DH_anon). This message will always immediately follow the\n ServerHello message.\n\n Meaning of this message:\n\n This message conveys the server's certificate chain to the client.\n\n The certificate MUST be appropriate for the negotiated cipher\n suite's key exchange algorithm and any negotiated extensions.\n\n\n\nDierks & Rescorla Standards Track [Page 47]\n\f\nRFC 5246 TLS August 2008\n\n\n Structure of this message:\n\n opaque ASN.1Cert<1..2^24-1>;\n\n struct {\n ASN.1Cert certificate_list<0..2^24-1>;\n } Certificate;\n\n certificate_list\n This is a sequence (chain) of certificates. The sender's\n certificate MUST come first in the list. Each following\n certificate MUST directly certify the one preceding it. Because\n certificate validation requires that root keys be distributed\n independently, the self-signed certificate that specifies the root\n certificate authority MAY be omitted from the chain, under the\n assumption that the remote end must already possess it in order to\n validate it in any case.\n\n The same message type and structure will be used for the client's\n response to a certificate request message. Note that a client MAY\n send no certificates if it does not have an appropriate certificate\n to send in response to the server's authentication request.\n\n Note: PKCS #7 [PKCS7] is not used as the format for the certificate\n vector because PKCS #6 [PKCS6] extended certificates are not used.\n Also, PKCS #7 defines a SET rather than a SEQUENCE, making the task\n of parsing the list more difficult.\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3, unless explicitly negotiated\n otherwise (e.g., [TLSPGP]).\n\n - The end entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected key exchange\n algorithm.\n\n Key Exchange Alg. Certificate Key Type\n\n RSA RSA public key; the certificate MUST allow the\n RSA_PSK key to be used for encryption (the\n keyEncipherment bit MUST be set if the key\n usage extension is present).\n Note: RSA_PSK is defined in [TLSPSK].\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 48]\n\f\nRFC 5246 TLS August 2008\n\n\n DHE_RSA RSA public key; the certificate MUST allow the\n ECDHE_RSA key to be used for signing (the\n digitalSignature bit MUST be set if the key\n usage extension is present) with the signature\n scheme and hash algorithm that will be employed\n in the server key exchange message.\n Note: ECDHE_RSA is defined in [TLSECC].\n\n DHE_DSS DSA public key; the certificate MUST allow the\n key to be used for signing with the hash\n algorithm that will be employed in the server\n key exchange message.\n\n DH_DSS Diffie-Hellman public key; the keyAgreement bit\n DH_RSA MUST be set if the key usage extension is\n present.\n\n ECDH_ECDSA ECDH-capable public key; the public key MUST\n ECDH_RSA use a curve and point format supported by the\n client, as described in [TLSECC].\n\n ECDHE_ECDSA ECDSA-capable public key; the certificate MUST\n allow the key to be used for signing with the\n hash algorithm that will be employed in the\n server key exchange message. The public key\n MUST use a curve and point format supported by\n the client, as described in [TLSECC].\n\n - The \"server_name\" and \"trusted_ca_keys\" extensions [TLSEXT] are\n used to guide certificate selection.\n\n\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 5705 RTFM, Inc.\nCategory: Standards Track March 2010\nISSN: 2070-1721\n\n\n Keying Material Exporters for Transport Layer Security (TLS)\n\nAbstract\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n to perform key establishment but then use some of the keying material\n for their own purposes. This document describes a general mechanism\n for allowing that.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/5705.\n\nCopyright Notice\n\n Copyright (c) 2010 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2\n 2. Conventions Used In This Document . . . . . . . . . . . . . . . 3\n 3. Binding to Application Contexts . . . . . . . . . . . . . . . . 3\n 4. Exporter Definition . . . . . . . . . . . . . . . . . . . . . . 4\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . . 5\n 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 6\n 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . 6\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 7\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 7\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 7\n\n1. Introduction\n\n Note: The mechanism described in this document was previously known\n as \"TLS Extractors\" but was changed to avoid a name conflict\n with the use of the term \"Extractor\" in the cryptographic\n community.\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n [RFC5246] or Datagram TLS (DTLS) [RFC4347] to perform key\n establishment but then use some of the keying material for their own\n purposes. A typical example is DTLS-SRTP [DTLS-SRTP], a key\n management scheme for the Secure Real-time Transport Protocol (SRTP)\n that uses DTLS to perform a key exchange and negotiate the SRTP\n [RFC3711] protection suite and then uses the DTLS master_secret to\n generate the SRTP keys.\n\n These applications imply a need to be able to export keying material\n (later called Exported Keying Material or EKM) from TLS/DTLS to an\n application or protocol residing at an upper layer, and to securely\n agree on the upper-layer context where the keying material will be\n used. The mechanism for exporting the keying material has the\n following requirements:\n\n o Both client and server need to be able to export the same EKM\n value.\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n o EKM values should be indistinguishable from random data to\n attackers who don't know the master_secret.\n\n o It should be possible to export multiple EKM values from the same\n TLS/DTLS association.\n\n o Knowing one EKM value should not reveal any useful information\n about the master_secret or about other EKM values.\n\n The mechanism described in this document is intended to fulfill these\n requirements. This mechanism is compatible with all versions of TLS.\n\n2. Conventions Used In This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Binding to Application Contexts\n\n In addition to using an exporter to obtain keying material, an\n application using the keying material has to securely establish the\n upper-layer context where the keying material will be used. The\n details of this context depend on the application, but it could\n include things such as algorithms and parameters that will be used\n with the keys, identifier(s) for the endpoint(s) who will use the\n keys, identifier(s) for the session(s) where the keys will be used,\n and the lifetime(s) for the context and/or keys. At a minimum, there\n should be some mechanism for signaling that an exporter will be used.\n\n This specification does not mandate a single mechanism for agreeing\n on such context; instead, there are several possibilities that can be\n used (and can complement each other). For example:\n\n o Information about the upper-layer context can be included in the\n optional data after the exporter label (see Section 4).\n\n o Information about the upper-layer context can be exchanged in TLS\n extensions included in the ClientHello and ServerHello messages.\n This approach is used in [DTLS-SRTP]. The handshake messages are\n protected by the Finished messages, so once the handshake\n completes, the peers will have the same view of the information.\n Extensions also allow a limited form of negotiation: for example,\n the TLS client could propose several alternatives for some context\n parameters, and the TLS server could select one of them.\n\n o The upper-layer protocol can include its own handshake, which can\n be protected using the keys exported by TLS.\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n No matter how the context is agreed, it is required that it has one\n part that indicates which application will use the exported keys.\n This part is the disambiguating label string (see Section 4).\n\n It is important to note that just embedding TLS messages in the\n upper-layer protocol may not automatically secure all the important\n context information, since the upper-layer messages are not covered\n by TLS Finished messages.\n\n4. Exporter Definition\n\n The output of the exporter is intended to be used in a single scope,\n which is associated with the TLS session, the label, and the context\n value.\n\n The exporter takes three input values:\n\n o a disambiguating label string,\n\n o a per-association context value provided by the application using\n the exporter, and\n\n o a length value.\n\n If no context is provided, it then computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random\n )[length]\n\n If context is provided, it computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random +\n context_value_length + context_value\n )[length]\n\n Where PRF is the TLS Pseudorandom Function in use for the session.\n The output is a pseudorandom bit string of length bytes generated\n from the master_secret. (This construction allows for\n interoperability with older exporter-type constructions which do not\n use context values, e.g., [RFC5281]).\n\n Labels here have the same definition as in TLS, i.e., an ASCII string\n with no terminating NULL. Label values beginning with \"EXPERIMENTAL\"\n MAY be used for private use without registration. All other label\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n values MUST be registered via Specification Required as described by\n RFC 5226 [RFC5226]. Note that exporter labels have the potential to\n collide with existing PRF labels. In order to prevent this, labels\n SHOULD begin with \"EXPORTER\". This is not a MUST because there are\n existing uses that have labels which do not begin with this prefix.\n\n The context value allows the application using the exporter to mix\n its own data with the TLS PRF for the exporter output. One example\n of where this might be useful is an authentication setting where the\n client credentials are valid for more than one identity; the context\n value could then be used to mix the expected identity into the keying\n material, thus preventing substitution attacks. The context value\n length is encoded as an unsigned, 16-bit quantity (uint16; see\n [RFC5246], Section 4.4) representing the length of the context value.\n The context MAY be zero length. Because the context value is mixed\n with the master_secret via the PRF, it is safe to mix confidential\n information into the exporter, provided that the master_secret will\n not be known to the attacker.\n\n5. Security Considerations\n\n The prime security requirement for exporter outputs is that they be\n independent. More formally, after a particular TLS session, if an\n adversary is allowed to choose multiple (label, context value) pairs\n and is given the output of the PRF for those values, the attacker is\n still unable to distinguish between the output of the PRF for a\n (label, context value) pair (different from the ones that it\n submitted) and a random value of the same length. In particular,\n there may be settings, such as the one described in Section 4, where\n the attacker can control the context value; such an attacker MUST NOT\n be able to predict the output of the exporter. Similarly, an\n attacker who does not know the master secret should not be able to\n distinguish valid exporter outputs from random values. The current\n set of TLS PRFs is believed to meet this objective, provided the\n master secret is randomly generated.\n\n Because an exporter produces the same value if applied twice with the\n same label to the same master_secret, it is critical that two EKM\n values generated with the same label not be used for two different\n purposes -- hence, the requirement for IANA registration. However,\n because exporters depend on the TLS PRF, it is not a threat to the\n use of an EKM value generated from one label to reveal an EKM value\n generated from another label.\n\n With certain TLS cipher suites, the TLS master secret is not\n necessarily unique to a single TLS session. In particular, with RSA\n key exchange, a malicious party acting as TLS server in one session\n and as TLS client in another session can cause those two sessions to\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n have the same TLS master secret (though the sessions must be\n established simultaneously to get adequate control of the Random\n values). Applications using the EKM need to consider this in how\n they use the EKM; in some cases, requiring the use of other cipher\n suites (such as those using a Diffie-Hellman key exchange) may be\n advisable.\n\n Designing a secure mechanism that uses exporters is not necessarily\n straightforward. This document only provides the exporter mechanism,\n but the problem of agreeing on the surrounding context and the\n meaning of the information passed to and from the exporter remains.\n Any new uses of the exporter mechanism should be subject to careful\n review.\n\n6. IANA Considerations\n\n IANA has created a TLS Exporter Label registry for this purpose. The\n initial contents of the registry are given below:\n\n Value Reference Note\n ----------------------------- --------- ----\n client finished [RFC5246] (1)\n server finished [RFC5246] (1)\n master secret [RFC5246] (1)\n key expansion [RFC5246] (1)\n client EAP encryption [RFC5216]\n ttls keying material [RFC5281]\n ttls challenge [RFC5281]\n\n Note: (1) These entries are reserved and MUST NOT be used for the\n purpose described in RFC 5705, in order to avoid confusion with\n similar, but distinct, use in RFC 5246.\n\n Future values are allocated via the RFC 5226 Specification Required\n policy. The label is a string consisting of printable ASCII\n characters. IANA MUST also verify that one label is not a prefix of\n any other label. For example, labels \"key\" or \"master secretary\" are\n forbidden.\n\n7. Acknowledgments\n\n Thanks to Pasi Eronen for valuable comments and for the contents of\n the IANA section and Section 3. Thanks to David McGrew for helpful\n discussion of the security considerations and to Vijay Gurbani and\n Alfred Hoenes for editorial comments.\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246,\n August 2008.\n\n8.2. Informative References\n\n [DTLS-SRTP] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for Secure\n Real-time Transport Protocol (SRTP)\", Work in Progress,\n February 2009.\n\n [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and\n K. Norrman, \"The Secure Real-time Transport Protocol\n (SRTP)\", RFC 3711, March 2004.\n\n [RFC4347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security\", RFC 4347, April 2006.\n\n [RFC5216] Simon, D., Aboba, B., and R. Hurst, \"The EAP-TLS\n Authentication Protocol\", RFC 5216, March 2008.\n\n [RFC5281] Funk, P. and S. Blake-Wilson, \"Extensible Authentication\n Protocol Tunneled Transport Layer Security Authenticated\n Protocol Version 0 (EAP-TTLSv0)\", RFC 5281, August 2008.\n\nAuthor's Address\n\n Eric Rescorla\n RTFM, Inc.\n 2064 Edgewood Drive\n Palo Alto, CA 94303\n USA\n\n EMail: [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) D. Eastlake 3rd\nRequest for Comments: 6066 Huawei\nObsoletes: 4366 January 2011\nCategory: Standards Track\nISSN: 2070-1721\n\n\n Transport Layer Security (TLS) Extensions: Extension Definitions\n\nAbstract\n\n This document provides specifications for existing TLS extensions.\n It is a companion document for RFC 5246, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\". The extensions specified are\n server_name, max_fragment_length, client_certificate_url,\n trusted_ca_keys, truncated_hmac, and status_request.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6066.\n\nCopyright Notice\n\n Copyright (c) 2011 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n\n\n\n\n\nEastlake Standards Track [Page 1]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction ....................................................3\n 1.1. Specific Extensions Covered ................................3\n 1.2. Conventions Used in This Document ..........................5\n 2. Extensions to the Handshake Protocol ............................5\n 3. Server Name Indication ..........................................6\n 4. Maximum Fragment Length Negotiation .............................8\n 5. Client Certificate URLs .........................................9\n 6. Trusted CA Indication ..........................................12\n 7. Truncated HMAC .................................................13\n 8. Certificate Status Request .....................................14\n 9. Error Alerts ...................................................16\n 10. IANA Considerations ...........................................17\n 10.1. pkipath MIME Type Registration ...........................17\n 10.2. Reference for TLS Alerts, TLS HandshakeTypes, and\n ExtensionTypes ...........................................19\n 11. Security Considerations .......................................19\n 11.1. Security Considerations for server_name ..................19\n 11.2. Security Considerations for max_fragment_length ..........20\n 11.3. Security Considerations for client_certificate_url .......20\n 11.4. Security Considerations for trusted_ca_keys ..............21\n 11.5. Security Considerations for truncated_hmac ...............21\n 11.6. Security Considerations for status_request ...............22\n 12. Normative References ..........................................22\n 13. Informative References ........................................23\n Appendix A. Changes from RFC 4366 .................................24\n Appendix B. Acknowledgements ......................................25\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 2]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1. Introduction\n\n The Transport Layer Security (TLS) Protocol Version 1.2 is specified\n in [RFC5246]. That specification includes the framework for\n extensions to TLS, considerations in designing such extensions (see\n Section 7.4.1.4 of [RFC5246]), and IANA Considerations for the\n allocation of new extension code points; however, it does not specify\n any particular extensions other than Signature Algorithms (see\n Section 7.4.1.4.1 of [RFC5246]).\n\n This document provides the specifications for existing TLS\n extensions. It is, for the most part, the adaptation and editing of\n material from RFC 4366, which covered TLS extensions for TLS 1.0 (RFC\n 2246) and TLS 1.1 (RFC 4346).\n\n1.1. Specific Extensions Covered\n\n The extensions described here focus on extending the functionality\n provided by the TLS protocol message formats. Other issues, such as\n the addition of new cipher suites, are deferred.\n\n The extension types defined in this document are:\n\n enum {\n server_name(0), max_fragment_length(1),\n client_certificate_url(2), trusted_ca_keys(3),\n truncated_hmac(4), status_request(5), (65535)\n } ExtensionType;\n\n Specifically, the extensions described in this document:\n\n - Allow TLS clients to provide to the TLS server the name of the\n server they are contacting. This functionality is desirable in\n order to facilitate secure connections to servers that host\n multiple 'virtual' servers at a single underlying network address.\n\n - Allow TLS clients and servers to negotiate the maximum fragment\n length to be sent. This functionality is desirable as a result of\n memory constraints among some clients, and bandwidth constraints\n among some access networks.\n\n - Allow TLS clients and servers to negotiate the use of client\n certificate URLs. This functionality is desirable in order to\n conserve memory on constrained clients.\n\n\n\n\n\n\n\nEastlake Standards Track [Page 3]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - Allow TLS clients to indicate to TLS servers which certification\n authority (CA) root keys they possess. This functionality is\n desirable in order to prevent multiple handshake failures\n involving TLS clients that are only able to store a small number\n of CA root keys due to memory limitations.\n\n - Allow TLS clients and servers to negotiate the use of truncated\n Message Authentication Codes (MACs). This functionality is\n desirable in order to conserve bandwidth in constrained access\n networks.\n\n - Allow TLS clients and servers to negotiate that the server sends\n the client certificate status information (e.g., an Online\n Certificate Status Protocol (OCSP) [RFC2560] response) during a\n TLS handshake. This functionality is desirable in order to avoid\n sending a Certificate Revocation List (CRL) over a constrained\n access network and therefore saving bandwidth.\n\n TLS clients and servers may use the extensions described in this\n document. The extensions are designed to be backwards compatible,\n meaning that TLS clients that support the extensions can talk to TLS\n servers that do not support the extensions, and vice versa.\n\n Note that any messages associated with these extensions that are sent\n during the TLS handshake MUST be included in the hash calculations\n involved in \"Finished\" messages.\n\n Note also that all the extensions defined in this document are\n relevant only when a session is initiated. A client that requests\n session resumption does not in general know whether the server will\n accept this request, and therefore it SHOULD send the same extensions\n as it would send if it were not attempting resumption. When a client\n includes one or more of the defined extension types in an extended\n client hello while requesting session resumption:\n\n - The server name indication extension MAY be used by the server\n when deciding whether or not to resume a session as described in\n Section 3.\n\n - If the resumption request is denied, the use of the extensions is\n negotiated as normal.\n\n - If, on the other hand, the older session is resumed, then the\n server MUST ignore the extensions and send a server hello\n containing none of the extension types. In this case, the\n functionality of these extensions negotiated during the original\n session initiation is applied to the resumed session.\n\n\n\n\nEastlake Standards Track [Page 4]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1.2. Conventions Used in This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n [RFC2119].\n\n2. Extensions to the Handshake Protocol\n\n This document specifies the use of two new handshake messages,\n \"CertificateURL\" and \"CertificateStatus\". These messages are\n described in Sections 5 and 8, respectively. The new handshake\n message structure therefore becomes:\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), certificate_url(21), certificate_status(22),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case certificate_url: CertificateURL;\n case certificate_status: CertificateStatus;\n } body;\n } Handshake;\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 5]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n3. Server Name Indication\n\n TLS does not provide a mechanism for a client to tell a server the\n name of the server it is contacting. It may be desirable for clients\n to provide this information to facilitate secure connections to\n servers that host multiple 'virtual' servers at a single underlying\n network address.\n\n In order to provide any of the server names, clients MAY include an\n extension of type \"server_name\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"ServerNameList\" where:\n\n struct {\n NameType name_type;\n select (name_type) {\n case host_name: HostName;\n } name;\n } ServerName;\n\n enum {\n host_name(0), (255)\n } NameType;\n\n opaque HostName<1..2^16-1>;\n\n struct {\n ServerName server_name_list<1..2^16-1>\n } ServerNameList;\n\n The ServerNameList MUST NOT contain more than one name of the same\n name_type. If the server understood the ClientHello extension but\n does not recognize the server name, the server SHOULD take one of two\n actions: either abort the handshake by sending a fatal-level\n unrecognized_name(112) alert or continue the handshake. It is NOT\n RECOMMENDED to send a warning-level unrecognized_name(112) alert,\n because the client's behavior in response to warning-level alerts is\n unpredictable. If there is a mismatch between the server name used\n by the client application and the server name of the credential\n chosen by the server, this mismatch will become apparent when the\n client application performs the server endpoint identification, at\n which point the client application will have to decide whether to\n proceed with the communication. TLS implementations are encouraged\n to make information available to application callers about warning-\n level alerts that were received or sent during a TLS handshake. Such\n information can be useful for diagnostic purposes.\n\n\n\n\n\nEastlake Standards Track [Page 6]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note: Earlier versions of this specification permitted multiple\n names of the same name_type. In practice, current client\n implementations only send one name, and the client cannot\n necessarily find out which name the server selected. Multiple\n names of the same name_type are therefore now prohibited.\n\n Currently, the only server names supported are DNS hostnames;\n however, this does not imply any dependency of TLS on DNS, and other\n name types may be added in the future (by an RFC that updates this\n document). The data structure associated with the host_name NameType\n is a variable-length vector that begins with a 16-bit length. For\n backward compatibility, all future data structures associated with\n new NameTypes MUST begin with a 16-bit length field. TLS MAY treat\n provided server names as opaque data and pass the names and types to\n the application.\n\n \"HostName\" contains the fully qualified DNS hostname of the server,\n as understood by the client. The hostname is represented as a byte\n string using ASCII encoding without a trailing dot. This allows the\n support of internationalized domain names through the use of A-labels\n defined in [RFC5890]. DNS hostnames are case-insensitive. The\n algorithm to compare hostnames is described in [RFC5890], Section\n 2.3.2.4.\n\n Literal IPv4 and IPv6 addresses are not permitted in \"HostName\".\n\n It is RECOMMENDED that clients include an extension of type\n \"server_name\" in the client hello whenever they locate a server by a\n supported name type.\n\n A server that receives a client hello containing the \"server_name\"\n extension MAY use the information contained in the extension to guide\n its selection of an appropriate certificate to return to the client,\n and/or other aspects of security policy. In this event, the server\n SHALL include an extension of type \"server_name\" in the (extended)\n server hello. The \"extension_data\" field of this extension SHALL be\n empty.\n\n When the server is deciding whether or not to accept a request to\n resume a session, the contents of a server_name extension MAY be used\n in the lookup of the session in the session cache. The client SHOULD\n include the same server_name extension in the session resumption\n request as it did in the full handshake that established the session.\n A server that implements this extension MUST NOT accept the request\n to resume the session if the server_name extension contains a\n different name. Instead, it proceeds with a full handshake to\n establish a new session. When resuming a session, the server MUST\n NOT include a server_name extension in the server hello.\n\n\n\nEastlake Standards Track [Page 7]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n If an application negotiates a server name using an application\n protocol and then upgrades to TLS, and if a server_name extension is\n sent, then the extension SHOULD contain the same name that was\n negotiated in the application protocol. If the server_name is\n established in the TLS session handshake, the client SHOULD NOT\n attempt to request a different server name at the application layer.\n\n4. Maximum Fragment Length Negotiation\n\n Without this extension, TLS specifies a fixed maximum plaintext\n fragment length of 2^14 bytes. It may be desirable for constrained\n clients to negotiate a smaller maximum fragment length due to memory\n limitations or bandwidth limitations.\n\n In order to negotiate smaller maximum fragment lengths, clients MAY\n include an extension of type \"max_fragment_length\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain:\n\n enum{\n 2^9(1), 2^10(2), 2^11(3), 2^12(4), (255)\n } MaxFragmentLength;\n\n whose value is the desired maximum fragment length. The allowed\n values for this field are: 2^9, 2^10, 2^11, and 2^12.\n\n Servers that receive an extended client hello containing a\n \"max_fragment_length\" extension MAY accept the requested maximum\n fragment length by including an extension of type\n \"max_fragment_length\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL contain a\n \"MaxFragmentLength\" whose value is the same as the requested maximum\n fragment length.\n\n If a server receives a maximum fragment length negotiation request\n for a value other than the allowed values, it MUST abort the\n handshake with an \"illegal_parameter\" alert. Similarly, if a client\n receives a maximum fragment length negotiation response that differs\n from the length it requested, it MUST also abort the handshake with\n an \"illegal_parameter\" alert.\n\n Once a maximum fragment length other than 2^14 has been successfully\n negotiated, the client and server MUST immediately begin fragmenting\n messages (including handshake messages) to ensure that no fragment\n larger than the negotiated length is sent. Note that TLS already\n requires clients and servers to support fragmentation of handshake\n messages.\n\n\n\n\nEastlake Standards Track [Page 8]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n The negotiated length applies for the duration of the session\n including session resumptions.\n\n The negotiated length limits the input that the record layer may\n process without fragmentation (that is, the maximum value of\n TLSPlaintext.length; see [RFC5246], Section 6.2.1). Note that the\n output of the record layer may be larger. For example, if the\n negotiated length is 2^9=512, then, when using currently defined\n cipher suites (those defined in [RFC5246] and [RFC2712]) and null\n compression, the record-layer output can be at most 805 bytes: 5\n bytes of headers, 512 bytes of application data, 256 bytes of\n padding, and 32 bytes of MAC. This means that in this event a TLS\n record-layer peer receiving a TLS record-layer message larger than\n 805 bytes MUST discard the message and send a \"record_overflow\"\n alert, without decrypting the message. When this extension is used\n with Datagram Transport Layer Security (DTLS), implementations SHOULD\n NOT generate record_overflow alerts unless the packet passes message\n authentication.\n\n5. Client Certificate URLs\n\n Without this extension, TLS specifies that when client authentication\n is performed, client certificates are sent by clients to servers\n during the TLS handshake. It may be desirable for constrained\n clients to send certificate URLs in place of certificates, so that\n they do not need to store their certificates and can therefore save\n memory.\n\n In order to negotiate sending certificate URLs to a server, clients\n MAY include an extension of type \"client_certificate_url\" in the\n (extended) client hello. The \"extension_data\" field of this\n extension SHALL be empty.\n\n (Note that it is necessary to negotiate the use of client certificate\n URLs in order to avoid \"breaking\" existing TLS servers.)\n\n Servers that receive an extended client hello containing a\n \"client_certificate_url\" extension MAY indicate that they are willing\n to accept certificate URLs by including an extension of type\n \"client_certificate_url\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n After negotiation of the use of client certificate URLs has been\n successfully completed (by exchanging hellos including\n \"client_certificate_url\" extensions), clients MAY send a\n \"CertificateURL\" message in place of a \"Certificate\" message as\n follows (see also Section 2):\n\n\n\n\nEastlake Standards Track [Page 9]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n enum {\n individual_certs(0), pkipath(1), (255)\n } CertChainType;\n\n struct {\n CertChainType type;\n URLAndHash url_and_hash_list<1..2^16-1>;\n } CertificateURL;\n\n struct {\n opaque url<1..2^16-1>;\n unint8 padding;\n opaque SHA1Hash[20];\n } URLAndHash;\n\n Here, \"url_and_hash_list\" contains a sequence of URLs and hashes.\n Each \"url\" MUST be an absolute URI reference according to [RFC3986]\n that can be immediately used to fetch the certificate(s).\n\n When X.509 certificates are used, there are two possibilities:\n\n - If CertificateURL.type is \"individual_certs\", each URL refers to a\n single DER-encoded X.509v3 certificate, with the URL for the\n client's certificate first.\n\n - If CertificateURL.type is \"pkipath\", the list contains a single\n URL referring to a DER-encoded certificate chain, using the type\n PkiPath described in Section 10.1.\n\n When any other certificate format is used, the specification that\n describes use of that format in TLS should define the encoding format\n of certificates or certificate chains, and any constraint on their\n ordering.\n\n The \"padding\" byte MUST be 0x01. It is present to make the structure\n backwards compatible.\n\n The hash corresponding to each URL is the SHA-1 hash of the\n certificate or certificate chain (in the case of X.509 certificates,\n the DER-encoded certificate or the DER-encoded PkiPath).\n\n Note that when a list of URLs for X.509 certificates is used, the\n ordering of URLs is the same as that used in the TLS Certificate\n message (see [RFC5246], Section 7.4.2), but opposite to the order in\n which certificates are encoded in PkiPath. In either case, the self-\n signed root certificate MAY be omitted from the chain, under the\n assumption that the server must already possess it in order to\n validate it.\n\n\n\nEastlake Standards Track [Page 10]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Servers receiving \"CertificateURL\" SHALL attempt to retrieve the\n client's certificate chain from the URLs and then process the\n certificate chain as usual. A cached copy of the content of any URL\n in the chain MAY be used, provided that the SHA-1 hash matches the\n hash of the cached copy.\n\n Servers that support this extension MUST support the 'http' URI\n scheme for certificate URLs and MAY support other schemes. Use of\n other schemes than 'http', 'https', or 'ftp' may create unexpected\n problems.\n\n If the protocol used is HTTP, then the HTTP server can be configured\n to use the Cache-Control and Expires directives described in\n [RFC2616] to specify whether and for how long certificates or\n certificate chains should be cached.\n\n The TLS server MUST NOT follow HTTP redirects when retrieving the\n certificates or certificate chain. The URLs used in this extension\n MUST NOT be chosen to depend on such redirects.\n\n If the protocol used to retrieve certificates or certificate chains\n returns a MIME-formatted response (as HTTP does), then the following\n MIME Content-Types SHALL be used: when a single X.509v3 certificate\n is returned, the Content-Type is \"application/pkix-cert\" [RFC2585],\n and when a chain of X.509v3 certificates is returned, the Content-\n Type is \"application/pkix-pkipath\" (Section 10.1).\n\n The server MUST check that the SHA-1 hash of the contents of the\n object retrieved from that URL (after decoding any MIME Content-\n Transfer-Encoding) matches the given hash. If any retrieved object\n does not have the correct SHA-1 hash, the server MUST abort the\n handshake with a bad_certificate_hash_value(114) alert. This alert\n is always fatal.\n\n Clients may choose to send either \"Certificate\" or \"CertificateURL\"\n after successfully negotiating the option to send certificate URLs.\n The option to send a certificate is included to provide flexibility\n to clients possessing multiple certificates.\n\n If a server is unable to obtain certificates in a given\n CertificateURL, it MUST send a fatal certificate_unobtainable(111)\n alert if it requires the certificates to complete the handshake. If\n the server does not require the certificates, then the server\n continues the handshake. The server MAY send a warning-level alert\n in this case. Clients receiving such an alert SHOULD log the alert\n and continue with the handshake if possible.\n\n\n\n\n\nEastlake Standards Track [Page 11]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n6. Trusted CA Indication\n\n Constrained clients that, due to memory limitations, possess only a\n small number of CA root keys may wish to indicate to servers which\n root keys they possess, in order to avoid repeated handshake\n failures.\n\n In order to indicate which CA root keys they possess, clients MAY\n include an extension of type \"trusted_ca_keys\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain \"TrustedAuthorities\" where:\n\n struct {\n TrustedAuthority trusted_authorities_list<0..2^16-1>;\n } TrustedAuthorities;\n\n struct {\n IdentifierType identifier_type;\n select (identifier_type) {\n case pre_agreed: struct {};\n case key_sha1_hash: SHA1Hash;\n case x509_name: DistinguishedName;\n case cert_sha1_hash: SHA1Hash;\n } identifier;\n } TrustedAuthority;\n\n enum {\n pre_agreed(0), key_sha1_hash(1), x509_name(2),\n cert_sha1_hash(3), (255)\n } IdentifierType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n Here, \"TrustedAuthorities\" provides a list of CA root key identifiers\n that the client possesses. Each CA root key is identified via\n either:\n\n - \"pre_agreed\": no CA root key identity supplied.\n\n - \"key_sha1_hash\": contains the SHA-1 hash of the CA root key. For\n Digital Signature Algorithm (DSA) and Elliptic Curve Digital\n Signature Algorithm (ECDSA) keys, this is the hash of the\n \"subjectPublicKey\" value. For RSA keys, the hash is of the big-\n endian byte string representation of the modulus without any\n initial zero-valued bytes. (This copies the key hash formats\n deployed in other environments.)\n\n\n\n\n\nEastlake Standards Track [Page 12]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - \"x509_name\": contains the DER-encoded X.509 DistinguishedName of\n the CA.\n\n - \"cert_sha1_hash\": contains the SHA-1 hash of a DER-encoded\n Certificate containing the CA root key.\n\n Note that clients may include none, some, or all of the CA root keys\n they possess in this extension.\n\n Note also that it is possible that a key hash or a Distinguished Name\n alone may not uniquely identify a certificate issuer (for example, if\n a particular CA has multiple key pairs). However, here we assume\n this is the case following the use of Distinguished Names to identify\n certificate issuers in TLS.\n\n The option to include no CA root keys is included to allow the client\n to indicate possession of some pre-defined set of CA root keys.\n\n Servers that receive a client hello containing the \"trusted_ca_keys\"\n extension MAY use the information contained in the extension to guide\n their selection of an appropriate certificate chain to return to the\n client. In this event, the server SHALL include an extension of type\n \"trusted_ca_keys\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n7. Truncated HMAC\n\n Currently defined TLS cipher suites use the MAC construction HMAC\n [RFC2104] to authenticate record-layer communications. In TLS, the\n entire output of the hash function is used as the MAC tag. However,\n it may be desirable in constrained environments to save bandwidth by\n truncating the output of the hash function to 80 bits when forming\n MAC tags.\n\n In order to negotiate the use of 80-bit truncated HMAC, clients MAY\n include an extension of type \"truncated_hmac\" in the extended client\n hello. The \"extension_data\" field of this extension SHALL be empty.\n\n Servers that receive an extended hello containing a \"truncated_hmac\"\n extension MAY agree to use a truncated HMAC by including an extension\n of type \"truncated_hmac\", with empty \"extension_data\", in the\n extended server hello.\n\n Note that if new cipher suites are added that do not use HMAC, and\n the session negotiates one of these cipher suites, this extension\n will have no effect. It is strongly recommended that any new cipher\n suites using other MACs consider the MAC size an integral part of the\n\n\n\n\nEastlake Standards Track [Page 13]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n cipher suite definition, taking into account both security and\n bandwidth considerations.\n\n If HMAC truncation has been successfully negotiated during a TLS\n handshake, and the negotiated cipher suite uses HMAC, both the client\n and the server pass this fact to the TLS record layer along with the\n other negotiated security parameters. Subsequently during the\n session, clients and servers MUST use truncated HMACs, calculated as\n specified in [RFC2104]. That is, SecurityParameters.mac_length is 10\n bytes, and only the first 10 bytes of the HMAC output are transmitted\n and checked. Note that this extension does not affect the\n calculation of the pseudo-random function (PRF) as part of\n handshaking or key derivation.\n\n The negotiated HMAC truncation size applies for the duration of the\n session including session resumptions.\n\n8. Certificate Status Request\n\n Constrained clients may wish to use a certificate-status protocol\n such as OCSP [RFC2560] to check the validity of server certificates,\n in order to avoid transmission of CRLs and therefore save bandwidth\n on constrained networks. This extension allows for such information\n to be sent in the TLS handshake, saving roundtrips and resources.\n\n In order to indicate their desire to receive certificate status\n information, clients MAY include an extension of type\n \"status_request\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"CertificateStatusRequest\" where:\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n } request;\n } CertificateStatusRequest;\n\n enum { ocsp(1), (255) } CertificateStatusType;\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n\n\n\nEastlake Standards Track [Page 14]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n In the OCSPStatusRequest, the \"ResponderIDs\" provides a list of OCSP\n responders that the client trusts. A zero-length \"responder_id_list\"\n sequence has the special meaning that the responders are implicitly\n known to the server, e.g., by prior arrangement. \"Extensions\" is a\n DER encoding of OCSP request extensions.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC2560]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a zero-length ASN.1 SEQUENCE, which is not\n valid for the \"Extensions\" type).\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement).\n\n Servers that receive a client hello containing the \"status_request\"\n extension MAY return a suitable certificate status response to the\n client along with their certificate. If OCSP is requested, they\n SHOULD use the information contained in the extension when selecting\n an OCSP responder and SHOULD include request_extensions in the OCSP\n request.\n\n Servers return a certificate response along with their certificate by\n sending a \"CertificateStatus\" message immediately after the\n \"Certificate\" message (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message, then the server MUST have included an\n extension of type \"status_request\" with empty \"extension_data\" in the\n extended server hello. The \"CertificateStatus\" message is conveyed\n using the handshake message type \"certificate_status\" as follows (see\n also Section 2):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<1..2^24-1>;\n\n An \"ocsp_response\" contains a complete, DER-encoded OCSP response\n (using the ASN.1 type OCSPResponse defined in [RFC2560]). Only one\n OCSP response may be sent.\n\n\n\n\nEastlake Standards Track [Page 15]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if has received a \"status_request\" extension in the\n client hello message and has sent a \"status_request\" extension in the\n server hello message.\n\n Note in addition that a server MUST NOT send the \"CertificateStatus\"\n message unless it received a \"status_request\" extension in the client\n hello message and sent a \"status_request\" extension in the server\n hello message.\n\n Clients requesting an OCSP response and receiving an OCSP response in\n a \"CertificateStatus\" message MUST check the OCSP response and abort\n the handshake if the response is not satisfactory with\n bad_certificate_status_response(113) alert. This alert is always\n fatal.\n\n9. Error Alerts\n\n Four new error alerts are defined for use with the TLS extensions\n defined in this document. To avoid \"breaking\" existing clients and\n servers, these alerts MUST NOT be sent unless the sending party has\n received an extended hello message from the party they are\n communicating with. These error alerts are conveyed using the\n following syntax. The new alerts are the last four, as indicated by\n the comments on the same line as the error alert number.\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n /* 41 is not defined, for historical reasons */\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction(60),\n protocol_version(70),\n insufficient_security(71),\n\n\n\nEastlake Standards Track [Page 16]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n certificate_unobtainable(111), /* new */\n unrecognized_name(112), /* new */\n bad_certificate_status_response(113), /* new */\n bad_certificate_hash_value(114), /* new */\n (255)\n } AlertDescription;\n\n \"certificate_unobtainable\" is described in Section 5.\n \"unrecognized_name\" is described in Section 3.\n \"bad_certificate_status_response\" is described in Section 8.\n \"bad_certificate_hash_value\" is described in Section 5.\n\n10. IANA Considerations\n\n IANA Considerations for TLS extensions and the creation of a registry\n are covered in Section 12 of [RFC5246] except for the registration of\n MIME type application/pkix-pkipath, which appears below.\n\n The IANA TLS extensions and MIME type application/pkix-pkipath\n registry entries that reference RFC 4366 have been updated to\n reference this document.\n\n10.1. pkipath MIME Type Registration\n\n MIME media type name: application\n MIME subtype name: pkix-pkipath\n Required parameters: none\n\n Optional parameters: version (default value is \"1\")\n\n Encoding considerations:\n Binary; this MIME type is a DER encoding of the ASN.1 type\n PkiPath, defined as follows:\n PkiPath ::= SEQUENCE OF Certificate\n PkiPath is used to represent a certification path. Within the\n sequence, the order of certificates is such that the subject of\n the first certificate is the issuer of the second certificate,\n etc.\n This is identical to the definition published in [X509-4th-TC1];\n note that it is different from that in [X509-4th].\n\n All Certificates MUST conform to [RFC5280]. (This should be\n interpreted as a requirement to encode only PKIX-conformant\n certificates using this type. It does not necessarily require\n\n\n\nEastlake Standards Track [Page 17]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n that all certificates that are not strictly PKIX-conformant must\n be rejected by relying parties, although the security consequences\n of accepting any such certificates should be considered\n carefully.)\n\n DER (as opposed to BER) encoding MUST be used. If this type is\n sent over a 7-bit transport, base64 encoding SHOULD be used.\n\n Security considerations:\n The security considerations of [X509-4th] and [RFC5280] (or any\n updates to them) apply, as well as those of any protocol that uses\n this type (e.g., TLS).\n\n Note that this type only specifies a certificate chain that can be\n assessed for validity according to the relying party's existing\n configuration of trusted CAs; it is not intended to be used to\n specify any change to that configuration.\n\n Interoperability considerations:\n No specific interoperability problems are known with this type,\n but for recommendations relating to X.509 certificates in general,\n see [RFC5280].\n\n Published specification: This document and [RFC5280].\n\n Applications that use this media type:\n TLS. It may also be used by other protocols or for general\n interchange of PKIX certificate chains.\n\n Additional information:\n Magic number(s): DER-encoded ASN.1 can be easily recognized.\n Further parsing is required to distinguish it from other ASN.1\n types.\n File extension(s): .pkipath\n Macintosh File Type Code(s): not specified\n\n Person & email address to contact for further information:\n Magnus Nystrom <[redacted-email]>\n\n Intended usage: COMMON\n\n Change controller: IESG <[redacted-email]>\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 18]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n10.2. Reference for TLS Alerts, TLS HandshakeTypes, and ExtensionTypes\n\n The following values in the TLS Alert Registry have been updated to\n reference this document:\n\n 111 certificate_unobtainable\n 112 unrecognized_name\n 113 bad_certificate_status_response\n 114 bad_certificate_hash_value\n\n The following values in the TLS HandshakeType Registry have been\n updated to reference this document:\n\n 21 certificate_url\n 22 certificate_status\n\n The following ExtensionType values have been updated to reference\n this document:\n\n 0 server_name\n 1 max_fragment_length\n 2 client_certificate_url\n 3 trusted_ca_keys\n 4 truncated_hmac\n 5 status_request\n\n11. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security Considerations for particular extensions\n specified in this document are given below.\n\n In general, implementers should continue to monitor the state of the\n art and address any weaknesses identified.\n\n11.1. Security Considerations for server_name\n\n If a single server hosts several domains, then clearly it is\n necessary for the owners of each domain to ensure that this satisfies\n their security needs. Apart from this, server_name does not appear\n to introduce significant security issues.\n\n Since it is possible for a client to present a different server_name\n in the application protocol, application server implementations that\n rely upon these names being the same MUST check to make sure the\n client did not present a different name in the application protocol.\n\n\n\n\n\nEastlake Standards Track [Page 19]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Implementations MUST ensure that a buffer overflow does not occur,\n whatever the values of the length fields in server_name.\n\n11.2. Security Considerations for max_fragment_length\n\n The maximum fragment length takes effect immediately, including for\n handshake messages. However, that does not introduce any security\n complications that are not already present in TLS, since TLS requires\n implementations to be able to handle fragmented handshake messages.\n\n Note that, as described in Section 4, once a non-null cipher suite\n has been activated, the effective maximum fragment length depends on\n the cipher suite and compression method, as well as on the negotiated\n max_fragment_length. This must be taken into account when sizing\n buffers and checking for buffer overflow.\n\n11.3. Security Considerations for client_certificate_url\n\n Support for client_certificate_url involves the server's acting as a\n client in another URI-scheme-dependent protocol. The server\n therefore becomes subject to many of the same security concerns that\n clients of the URI scheme are subject to, with the added concern that\n the client can attempt to prompt the server to connect to some\n (possibly weird-looking) URL.\n\n In general, this issue means that an attacker might use the server to\n indirectly attack another host that is vulnerable to some security\n flaw. It also introduces the possibility of denial-of-service\n attacks in which an attacker makes many connections to the server,\n each of which results in the server's attempting a connection to the\n target of the attack.\n\n Note that the server may be behind a firewall or otherwise able to\n access hosts that would not be directly accessible from the public\n Internet. This could exacerbate the potential security and denial-\n of-service problems described above, as well as allow the existence\n of internal hosts to be confirmed when they would otherwise be\n hidden.\n\n The detailed security concerns involved will depend on the URI\n schemes supported by the server. In the case of HTTP, the concerns\n are similar to those that apply to a publicly accessible HTTP proxy\n server. In the case of HTTPS, loops and deadlocks may be created,\n and this should be addressed. In the case of FTP, attacks arise that\n are similar to FTP bounce attacks.\n\n\n\n\n\n\nEastlake Standards Track [Page 20]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n As a result of this issue, it is RECOMMENDED that the\n client_certificate_url extension should have to be specifically\n enabled by a server administrator, rather than be enabled by default.\n It is also RECOMMENDED that URI schemes be enabled by the\n administrator individually, and only a minimal set of schemes be\n enabled. Unusual protocols that offer limited security or whose\n security is not well understood SHOULD be avoided.\n\n As discussed in [RFC3986], URLs that specify ports other than the\n default may cause problems, as may very long URLs (which are more\n likely to be useful in exploiting buffer overflow bugs).\n\n This extension continues to use SHA-1 (as in RFC 4366) and does not\n provide algorithm agility. The property required of SHA-1 in this\n case is second pre-image resistance, not collision resistance.\n Furthermore, even if second pre-image attacks against SHA-1 are found\n in the future, an attack against client_certificate_url would require\n a second pre-image that is accepted as a valid certificate by the\n server and contains the same public key.\n\n Also note that HTTP caching proxies are common on the Internet, and\n some proxies do not check for the latest version of an object\n correctly. If a request using HTTP (or another caching protocol)\n goes through a misconfigured or otherwise broken proxy, the proxy may\n return an out-of-date response.\n\n11.4. Security Considerations for trusted_ca_keys\n\n Potentially, the CA root keys a client possesses could be regarded as\n confidential information. As a result, the CA root key indication\n extension should be used with care.\n\n The use of the SHA-1 certificate hash alternative ensures that each\n certificate is specified unambiguously. This context does not\n require a cryptographic hash function, so the use of SHA-1 is\n considered acceptable, and no algorithm agility is provided.\n\n11.5. Security Considerations for truncated_hmac\n\n It is possible that truncated MACs are weaker than \"un-truncated\"\n MACs. However, no significant weaknesses are currently known or\n expected to exist for HMAC with MD5 or SHA-1, truncated to 80 bits.\n\n Note that the output length of a MAC need not be as long as the\n length of a symmetric cipher key, since forging of MAC values cannot\n be done off-line: in TLS, a single failed MAC guess will cause the\n immediate termination of the TLS session.\n\n\n\n\nEastlake Standards Track [Page 21]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Since the MAC algorithm only takes effect after all handshake\n messages that affect extension parameters have been authenticated by\n the hashes in the Finished messages, it is not possible for an active\n attacker to force negotiation of the truncated HMAC extension where\n it would not otherwise be used (to the extent that the handshake\n authentication is secure). Therefore, in the event that any security\n problems were found with truncated HMAC in the future, if either the\n client or the server for a given session were updated to take the\n problem into account, it would be able to veto use of this extension.\n\n11.6. Security Considerations for status_request\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC2560] for further details.\n\n12. Normative References\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC:\n Keyed-Hashing for Message Authentication\", RFC 2104,\n February 1997.\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and\n C. Adams, \"X.509 Internet Public Key Infrastructure\n Online Certificate Status Protocol - OCSP\", RFC 2560,\n June 1999.\n\n [RFC2585] Housley, R. and P. Hoffman, \"Internet X.509 Public Key\n Infrastructure Operational Protocols: FTP and HTTP\",\n RFC 2585, May 1999.\n\n [RFC2616] Fielding, R., Gettys, J., Mogul, J., Frystyk, H.,\n Masinter, L., Leach, P., and T. Berners-Lee,\n \"Hypertext Transfer Protocol -- HTTP/1.1\", RFC 2616,\n June 1999.\n\n [RFC3986] Berners-Lee, T., Fielding, R., and L. Masinter,\n \"Uniform Resource Identifier (URI): Generic Syntax\",\n STD 66, RFC 3986, January 2005.\n\n\n\nEastlake Standards Track [Page 22]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246, August\n 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation\n List (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC5890] Klensin, J., \"Internationalized Domain Names for\n Applications (IDNA): Definitions and Document\n Framework\", RFC 5890, August 2010.\n\n13. Informative References\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [X509-4th] ITU-T Recommendation X.509 (2000) | ISO/IEC\n 9594-8:2001, \"Information Systems - Open Systems\n Interconnection - The Directory: Public key and\n attribute certificate frameworks\".\n\n [X509-4th-TC1] ITU-T Recommendation X.509(2000) Corrigendum 1(2001) |\n ISO/IEC 9594-8:2001/Cor.1:2002, Technical Corrigendum\n 1 to ISO/IEC 9594:8:2001.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 23]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix A. Changes from RFC 4366\n\n The significant changes between RFC 4366 and this document are\n described below.\n\n RFC 4366 described both general extension mechanisms (for the TLS\n handshake and client and server hellos) as well as specific\n extensions. RFC 4366 was associated with RFC 4346, TLS 1.1. The\n client and server hello extension mechanisms have been moved into RFC\n 5246, TLS 1.2, so this document, which is associated with RFC 5246,\n includes only the handshake extension mechanisms and the specific\n extensions from RFC 4366. RFC 5246 also specifies the unknown\n extension error and new extension specification considerations, so\n that material has been removed from this document.\n\n The Server Name extension now specifies only ASCII representation,\n eliminating UTF-8. It is provided that the ServerNameList can\n contain more than only one name of any particular name_type. If a\n server name is provided but not recognized, the server should either\n continue the handshake without an error or send a fatal error.\n Sending a warning-level message is not recommended because client\n behavior will be unpredictable. Provision was added for the user\n using the server_name extension in deciding whether or not to resume\n a session. Furthermore, this extension should be the same in a\n session resumption request as it was in the full handshake that\n established the session. Such a resumption request must not be\n accepted if the server_name extension is different, but instead a\n full handshake must be done to possibly establish a new session.\n\n The Client Certificate URLs extension has been changed to make the\n presence of a hash mandatory.\n\n For the case of DTLS, the requirement to report an overflow of the\n negotiated maximum fragment length is made conditional on passing\n authentication.\n\n TLS servers are now prohibited from following HTTP redirects when\n retrieving certificates.\n\n The material was also re-organized in minor ways. For example,\n information as to which errors are fatal is moved from the \"Error\n Alerts\" section to the individual extension specifications.\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 24]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix B. Acknowledgements\n\n This document is based on material from RFC 4366 for which the\n authors were S. Blake-Wilson, M. Nystrom, D. Hopwood, J. Mikkelsen,\n and T. Wright. Other contributors include Joseph Salowey, Alexey\n Melnikov, Peter Saint-Andre, and Adrian Farrel.\n\nAuthor's Address\n\n Donald Eastlake 3rd\n Huawei\n 155 Beaver Street\n Milford, MA 01757 USA\n\n Phone: +1-508-333-2270\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 25]\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) Y. Pettersen\nRequest for Comments: 6961 June 2013\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\n\nAbstract\n\n This document defines the Transport Layer Security (TLS) Certificate\n Status Version 2 Extension to allow clients to specify and support\n several certificate status methods. (The use of the Certificate\n Status extension is commonly referred to as \"OCSP stapling\".) Also\n defined is a new method based on the Online Certificate Status\n Protocol (OCSP) that servers can use to provide status information\n about not only the server's own certificate but also the status of\n intermediate certificates in the chain.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6961.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 1]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nCopyright Notice\n\n Copyright (c) 2013 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n1. Introduction\n\n The Transport Layer Security (TLS) Extension [RFC6066] framework\n defines, among other extensions, the Certificate Status extension\n (also referred to as \"OCSP stapling\") that clients can use to request\n the server's copy of the current status of its certificate. The\n benefits of this extension include a reduced number of roundtrips and\n network delays for the client to verify the status of the server's\n certificate and a reduced load on the certificate issuer's status\n response servers, thus solving a problem that can become significant\n when the issued certificate is presented by a frequently visited\n server.\n\n There are two problems with the existing Certificate Status\n extension. First, it does not provide functionality to request the\n status information about intermediate Certification Authority (CA)\n certificates, which means the client has to request status\n information through other methods, such as Certificate Revocation\n Lists (CRLs), introducing further delays. Second, the current format\n of the extension and requirements in the TLS protocol prevent a\n client from offering the server multiple status methods.\n\n\n\nPettersen Standards Track [Page 2]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n Many CAs are now issuing intermediate CA certificates that not only\n specify the publication point for their CRLs in a CRL Distribution\n Point [RFC5280] but also specify a URL for their OCSP [RFC6960]\n server in Authority Information Access [RFC5280]. Given that\n client-cached CRLs are frequently out of date, clients would benefit\n from using OCSP to access up-to-date status information about\n intermediate CA certificates. The benefit to the issuing CA is less\n clear, as providing the bandwidth for the OCSP responder can be\n costly, especially for CAs with many high-traffic subscriber sites,\n and this cost is a concern for many CAs. There are cases where OCSP\n requests for a single high-traffic site caused significant network\n problems for the issuing CA.\n\n Clients will benefit from the TLS server providing certificate status\n information regardless of type, not just for the server certificate\n but also for the intermediate CA certificates. Combining the status\n checks into one extension will reduce the roundtrips needed to\n complete the handshake by the client to just those needed for\n negotiating the TLS connection. Also, for the Certification\n Authorities, the load on their servers will depend on the number of\n certificates they have issued, not on the number of visitors to those\n sites. Additionally, using this extension significantly reduces\n privacy concerns around the clients informing the certificate issuer\n about which sites they are visiting.\n\n For such a new system to be introduced seamlessly, clients need to be\n able to indicate support for the existing OCSP Certificate Status\n method and a new multiple-OCSP mode.\n\n Unfortunately, the definition of the Certificate Status extension\n only allows a single Certificate Status extension to be defined in a\n single extension record in the handshake, and the TLS protocol\n [RFC5246] only allows a single record in the extension list for any\n given extension. This means that it is not possible for clients to\n indicate support for new methods while still supporting older\n methods, which would cause problems for interoperability between\n newer clients and older servers. This will not just be an issue for\n the multiple status request mode proposed above but also for any\n other future status methods that might be introduced. This will be\n true not just for the current PKIX infrastructure [RFC5280] but also\n for alternative PKI structures.\n\n The solution to this problem is to define a new extension,\n \"status_request_v2\", with an extended format that allows the client\n to indicate support for multiple status request methods. This is\n implemented using a list of CertificateStatusRequestItemV2 records in\n the extension record. As the server will select the single status\n\n\n\n\nPettersen Standards Track [Page 3]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n method based on the selected cipher suite and the certificate\n presented, no significant changes are needed in the server's\n extension format.\n\n1.1. Requirements Language\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [RFC2119].\n\n1.2. Presentation Language\n\n This document defines protocol structures using the same conventions\n and presentation language as defined in Section 4 of [RFC5246].\n\n2. Multiple Certificate Status Extension\n\n2.1. New Extension\n\n The extension defined by this document is indicated by\n \"status_request_v2\" in the ExtensionType enum (originally defined by\n [RFC6066]), which uses the following value:\n\n enum {\n status_request_v2(17), (65535)\n } ExtensionType;\n\n2.2. Multiple Certificate Status Request Record\n\n Clients that support a certificate status protocol like OCSP may send\n the \"status_request_v2\" extension to the server in order to use the\n TLS handshake to transfer such data instead of downloading it through\n separate connections. When using this extension, the\n \"extension_data\" field (defined in Section 7.4.1.4 of [RFC5246]) of\n the extension SHALL contain a CertificateStatusRequestListV2 where:\n\n struct {\n CertificateStatusType status_type;\n uint16 request_length; /* Length of request field in bytes */\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n case ocsp_multi: OCSPStatusRequest;\n } request;\n } CertificateStatusRequestItemV2;\n\n enum { ocsp(1), ocsp_multi(2), (255) } CertificateStatusType;\n\n\n\n\n\nPettersen Standards Track [Page 4]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n struct {\n CertificateStatusRequestItemV2\n certificate_status_req_list<1..2^16-1>;\n } CertificateStatusRequestListV2;\n\n In the OCSPStatusRequest (originally defined by [RFC6066]), the\n \"ResponderID\" provides a list of OCSP responders that the client\n trusts. A zero-length \"responder_id_list\" sequence has the special\n meaning that the responders are implicitly known to the server, e.g.,\n by prior arrangement, or are identified by the certificates used by\n the server. \"Extensions\" is a DER encoding [X.690] of the OCSP\n request extensions, and if the server supports the forwarding of OCSP\n request extensions, this value MUST be forwarded without\n modification.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC6960]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a DER-encoded zero-length ASN.1 SEQUENCE,\n which is not valid for the \"Extensions\" type).\n\n Servers that support a client's selection of responders using the\n ResponderID field could implement this selection by matching the\n responder ID values from the client's list with the ResponderIDs of\n known OCSP responders, either by using a binary compare of the values\n or a hash calculation and compare method.\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement). This\n has been clarified in [RFC6960].\n\n The items in the list of CertificateStatusRequestItemV2 entries are\n ordered according to the client's preference (favorite choice first).\n\n A server that receives a client hello containing the\n \"status_request_v2\" extension MAY return a suitable certificate\n status response message to the client along with the server's\n\n\n\nPettersen Standards Track [Page 5]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n certificate message. If OCSP is requested, it SHOULD use the\n information contained in the extension when selecting an OCSP\n responder and SHOULD include request_extensions in the OCSP request.\n\n The server returns a certificate status response along with its\n certificate by sending a \"CertificateStatus\" message (originally\n defined by [RFC6066]) immediately after the \"Certificate\" message\n (Section 7.4.2 of [RFC5246]) (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message in response to a \"status_request_v2\"\n request, then the server MUST have included an extension of type\n \"status_request_v2\" with empty \"extension_data\" in the extended\n server hello.\n\n The \"CertificateStatus\" message is conveyed using the handshake\n message type \"certificate_status\" (defined in [RFC6066]) as follows\n (updated from the definition in [RFC6066]):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n case ocsp_multi: OCSPResponseList;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<0..2^24-1>;\n\n struct {\n OCSPResponse ocsp_response_list<1..2^24-1>;\n } OCSPResponseList;\n\n An \"OCSPResponse\" element (originally defined by [RFC6066]) contains\n a complete, DER-encoded OCSP response (using the ASN.1 [X.680] type\n OCSPResponse defined in [RFC6960]). Only one OCSP response, with a\n length of at least one byte, may be sent for status_type \"ocsp\".\n\n An \"ocsp_response_list\" contains a list of \"OCSPResponse\" elements,\n as specified above, each containing the OCSP response for the\n matching corresponding certificate in the server's Certificate TLS\n handshake message. That is, the first entry is the OCSP response for\n the first certificate in the Certificate list, the second entry is\n the response for the second certificate, and so on. The list MAY\n contain fewer OCSP responses than there were certificates in the\n Certificate handshake message, but there MUST NOT be more responses\n than there were certificates in the list. Individual elements of the\n list MAY have a length of 0 (zero) bytes if the server does not have\n the OCSP response for that particular certificate stored, in which\n\n\n\nPettersen Standards Track [Page 6]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n case the client MUST act as if a response was not received for that\n particular certificate. If the client receives a\n \"ocsp_response_list\" that does not contain a response for one or more\n of the certificates in the completed certificate chain, the client\n SHOULD attempt to validate the certificate using an alternative\n retrieval method, such as downloading the relevant CRL; OCSP SHOULD\n in this situation only be used for the end-entity certificate, not\n intermediate CA certificates, for reasons stated above.\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if it has received a \"status_request_v2\" extension in\n the client hello message and has sent a \"status_request_v2\" extension\n in the server hello message. Additionally, note that a server MUST\n NOT send the \"CertificateStatus\" message unless it received either a\n \"status_request\" or \"status_request_v2\" extension in the client hello\n message and sent a corresponding \"status_request\" or\n \"status_request_v2\" extension in the server hello message.\n\n Clients requesting an OCSP response and receiving one or more OCSP\n responses in a \"CertificateStatus\" message MUST check the OCSP\n response(s) and abort the handshake if the response is a \"revoked\"\n status or other unacceptable responses (as determined by client\n policy) with a bad_certificate_status_response(113) alert. This\n alert is always fatal.\n\n If the OCSP response received from the server does not result in a\n definite \"good\" or \"revoked\" status, it is inconclusive. A TLS\n client in such a case MAY check the validity of the server\n certificate through other means, e.g., by directly querying the\n certificate issuer. If such processing still results in an\n inconclusive response, then the application using the TLS connection\n will have to decide whether to close the connection or not. Note\n that this problem cannot be decided by the generic TLS client code\n without information from the application. If the application doesn't\n provide any such information, then the client MUST abort the\n connection, since the server certificate has not been sufficiently\n validated.\n\n An example of where the application might wish to continue is with\n EAP-TLS (Extensible Authentication Protocol - TLS), where the\n application can use another mechanism to check the status of a\n certificate once it obtains network access. In this case, the\n application could have the client continue with the handshake, but it\n MUST NOT disclose a username and password until it has fully\n validated the server certificate.\n\n\n\n\n\n\nPettersen Standards Track [Page 7]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n3. IANA Considerations\n\n Section 2.1 defines the new TLS extension status_request_v2 (17)\n enum, which has been added to the \"ExtensionType Values\" list in the\n IANA \"Transport Layer Security (TLS) Extensions\" registry.\n\n Section 2.2 describes a TLS CertificateStatusType registry that is\n now maintained by IANA. The \"TLS Certificate Status Types\" registry\n has been created under the \"Transport Layer Security (TLS)\n Extensions\" registry. CertificateStatusType values are to be\n assigned via IETF Review as defined in [RFC5226]. The initial\n registry corresponds to the definition of \"CertificateStatusType\" in\n Section 2.2.\n\n Value Description Reference\n -----------------------------------------\n 0 Reserved [RFC6961]\n 1 ocsp [RFC6066] [RFC6961]\n 2 ocsp_multi [RFC6961]\n 3-255 Unassigned\n\n4. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security considerations for the particular extension\n specified in this document are given below. In general, implementers\n should continue to monitor the state of the art and address any\n weaknesses identified.\n\n4.1. Security Considerations for status_request_v2\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC6960] for further details.\n\n This extension allows the client to send arbitrary data to the\n server. The server implementers need to handle such data carefully\n to avoid introducing security vulnerabilities.\n\n The security considerations of [RFC6960] apply to OCSP requests and\n responses.\n\n\n\n\nPettersen Standards Track [Page 8]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n5. Acknowledgements\n\n This document is based on [RFC6066], authored by Donald Eastlake 3rd.\n\n6. References\n\n6.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246, August 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC6066] Eastlake, D., \"Transport Layer Security (TLS) Extensions:\n Extension Definitions\", RFC 6066, January 2011.\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, June 2013.\n\n [X.680] ITU-T Recommendation X.680 (2008) | ISO/IEC 8824-1:2008,\n \"Abstract Syntax Notation One (ASN.1): Specification of\n basic notation\", November 2008.\n\n [X.690] ITU-T Recommendation X.690 (2008) | ISO/IEC 8825-1:2008,\n \"ASN.1 encoding rules: Specification of Basic Encoding\n Rules (BER), Canonical Encoding Rules (CER) and\n Distinguished Encoding Rules (DER)\", November 2008.\n\n6.2. Informative References\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and C.\n Adams, \"X.509 Internet Public Key Infrastructure Online\n Certificate Status Protocol - OCSP\", RFC 2560, June 1999.\n\n\n\n\n\n\nPettersen Standards Track [Page 9]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nAuthor's Address\n\n Yngve N. Pettersen\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 10]\n\n===== DOCUMENT =====\n\nNetwork Working Group E. Rescorla\nInternet-Draft RTFM, Inc.\nObsoletes: 5077, 5246, 6961 (if March 20, 2018\n approved)\nUpdates: 4492, 5705, 6066 (if approved)\nIntended status: Standards Track\nExpires: September 21, 2018\n\n\n\n===== DOCUMENT =====\n\nInternet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 8446 Mozilla\nObsoletes: 5077, 5246, 6961 August 2018\nUpdates: 5705, 6066\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS) Protocol Version 1.3\n\nAbstract\n\n This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol. TLS allows client/server applications to communicate\n over the Internet in a way that is designed to prevent eavesdropping,\n tampering, and message forgery.\n\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077,\n 5246, and 6961. This document also specifies new requirements for\n TLS 1.2 implementations.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 7841.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n https://www.rfc-editor.org/info/rfc8446.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 8446 TLS August 2018\n\n\nCopyright Notice\n\n Copyright (c) 2018 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (https://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 8446 TLS August 2018\n\n\nTable of Contents\n\n 1. Introduction ....................................................6\n 1.1. Conventions and Terminology ................................7\n 1.2. Major Differences from TLS 1.2 .............................8\n 1.3. Updates Affecting TLS 1.2 ..................................9\n 2. Protocol Overview ..............................................10\n 2.1. Incorrect DHE Share .......................................14\n 2.2. Resumption and Pre-Shared Key (PSK) .......................15\n 2.3. 0-RTT Data ................................................17\n 3. Presentation Language ..........................................19\n 3.1. Basic Block Size ..........................................19\n 3.2. Miscellaneous .............................................20\n 3.3. Numbers ...................................................20\n 3.4. Vectors ...................................................20\n 3.5. Enumerateds ...............................................21\n 3.6. Constructed Types .........................................22\n 3.7. Constants .................................................23\n 3.8. Variants ..................................................23\n 4. Handshake Protocol .............................................24\n 4.1. Key Exchange Messages .....................................25\n 4.1.1. Cryptographic Negotiation ..........................26\n 4.1.2. Client Hello .......................................27\n 4.1.3. Server Hello .......................................31\n 4.1.4. Hello Retry Request ................................33\n 4.2. Extensions ................................................35\n 4.2.1. Supported Versions .................................39\n 4.2.2. Cookie .............................................40\n 4.2.3. Signature Algorithms ...............................41\n 4.2.4. Certificate Authorities ............................45\n 4.2.5. OID Filters ........................................45\n 4.2.6. Post-Handshake Client Authentication ...............47\n 4.2.7. Supported Groups ...................................47\n 4.2.8. Key Share ..........................................48\n 4.2.9. Pre-Shared Key Exchange Modes ......................51\n 4.2.10. Early Data Indication .............................52\n 4.2.11. Pre-Shared Key Extension ..........................55\n 4.3. Server Parameters .........................................59\n 4.3.1. Encrypted Extensions ...............................60\n 4.3.2. Certificate Request ................................60\n 4.4. Authentication Messages ...................................61\n 4.4.1. The Transcript Hash ................................63\n 4.4.2. Certificate ........................................64\n 4.4.3. Certificate Verify .................................69\n 4.4.4. Finished ...........................................71\n 4.5. End of Early Data .........................................72\n\n\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 8446 TLS August 2018\n\n\n 4.6. Post-Handshake Messages ...................................73\n 4.6.1. New Session Ticket Message .........................73\n 4.6.2. Post-Handshake Authentication ......................75\n 4.6.3. Key and Initialization Vector Update ...............76\n 5. Record Protocol ................................................77\n 5.1. Record Layer ..............................................78\n 5.2. Record Payload Protection .................................80\n 5.3. Per-Record Nonce ..........................................82\n 5.4. Record Padding ............................................83\n 5.5. Limits on Key Usage .......................................84\n 6. Alert Protocol .................................................85\n 6.1. Closure Alerts ............................................87\n 6.2. Error Alerts ..............................................88\n 7. Cryptographic Computations .....................................90\n 7.1. Key Schedule ..............................................91\n 7.2. Updating Traffic Secrets ..................................94\n 7.3. Traffic Key Calculation ...................................95\n 7.4. (EC)DHE Shared Secret Calculation .........................95\n 7.4.1. Finite Field Diffie-Hellman ........................95\n 7.4.2. Elliptic Curve Diffie-Hellman ......................96\n 7.5. Exporters .................................................97\n 8. 0-RTT and Anti-Replay ..........................................98\n 8.1. Single-Use Tickets ........................................99\n 8.2. Client Hello Recording ....................................99\n 8.3. Freshness Checks .........................................101\n 9. Compliance Requirements .......................................102\n 9.1. Mandatory-to-Implement Cipher Suites .....................102\n 9.2. Mandatory-to-Implement Extensions ........................103\n 9.3. Protocol Invariants ......................................104\n 10. Security Considerations ......................................106\n 11. IANA Considerations ..........................................106\n 12. References ...................................................109\n 12.1. Normative References ....................................109\n 12.2. Informative References ..................................112\n Appendix A. State Machine ........................................120\n A.1. Client ....................................................120\n A.2. Server ....................................................121\n Appendix B. Protocol Data Structures and Constant Values .........122\n B.1. Record Layer ..............................................122\n B.2. Alert Messages ............................................123\n B.3. Handshake Protocol ........................................124\n B.3.1. Key Exchange Messages .................................125\n B.3.2. Server Parameters Messages ............................131\n B.3.3. Authentication Messages ...............................132\n B.3.4. Ticket Establishment ..................................132\n B.3.5. Updating Keys .........................................133\n B.4. Cipher Suites .............................................133\n\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 8446 TLS August 2018\n\n\n Appendix C. Implementation Notes .................................134\n C.1. Random Number Generation and Seeding ......................134\n C.2. Certificates and Authentication ...........................135\n C.3. Implementation Pitfalls ...................................135\n C.4. Client Tracking Prevention ................................137\n C.5. Unauthenticated Operation .................................137\n Appendix D. Backward Compatibility ...............................138\n D.1. Negotiating with an Older Server ..........................139\n D.2. Negotiating with an Older Client ..........................139\n D.3. 0-RTT Backward Compatibility ..............................140\n D.4. Middlebox Compatibility Mode ..............................140\n D.5. Security Restrictions Related to Backward Compatibility ...141\n Appendix E. Overview of Security Properties ......................142\n E.1. Handshake .................................................142\n E.1.1. Key Derivation and HKDF ...............................145\n E.1.2. Client Authentication .................................146\n E.1.3. 0-RTT .................................................146\n E.1.4. Exporter Independence .................................146\n E.1.5. Post-Compromise Security ..............................146\n E.1.6. External References ...................................147\n E.2. Record Layer ..............................................147\n E.2.1. External References ...................................148\n E.3. Traffic Analysis ..........................................148\n E.4. Side-Channel Attacks ......................................149\n E.5. Replay Attacks on 0-RTT ...................................150\n E.5.1. Replay and Exporters ..................................151\n E.6. PSK Identity Exposure .....................................152\n E.7. Sharing PSKs ..............................................152\n E.8. Attacks on Static RSA .....................................152\n Contributors .....................................................153\n Author's Address .................................................160\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 8446 TLS August 2018\n\n\n1. Introduction\n\n The primary goal of TLS is to provide a secure channel between two\n communicating peers; the only requirement from the underlying\n transport is a reliable, in-order data stream. Specifically, the\n secure channel should provide the following properties:\n\n - Authentication: The server side of the channel is always\n authenticated; the client side is optionally authenticated.\n Authentication can happen via asymmetric cryptography (e.g., RSA\n [RSA], the Elliptic Curve Digital Signature Algorithm (ECDSA)\n [ECDSA], or the Edwards-Curve Digital Signature Algorithm (EdDSA)\n [RFC8032]) or a symmetric pre-shared key (PSK).\n\n - Confidentiality: Data sent over the channel after establishment is\n only visible to the endpoints. TLS does not hide the length of\n the data it transmits, though endpoints are able to pad TLS\n records in order to obscure lengths and improve protection against\n traffic analysis techniques.\n\n - Integrity: Data sent over the channel after establishment cannot\n be modified by attackers without detection.\n\n These properties should be true even in the face of an attacker who\n has complete control of the network, as described in [RFC3552]. See\n Appendix E for a more complete statement of the relevant security\n properties.\n\n TLS consists of two primary components:\n\n - A handshake protocol (Section 4) that authenticates the\n communicating parties, negotiates cryptographic modes and\n parameters, and establishes shared keying material. The handshake\n protocol is designed to resist tampering; an active attacker\n should not be able to force the peers to negotiate different\n parameters than they would if the connection were not under\n attack.\n\n - A record protocol (Section 5) that uses the parameters established\n by the handshake protocol to protect traffic between the\n communicating peers. The record protocol divides traffic up into\n a series of records, each of which is independently protected\n using the traffic keys.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 8446 TLS August 2018\n\n\n TLS is application protocol independent; higher-level protocols can\n layer on top of TLS transparently. The TLS standard, however, does\n not specify how protocols add security with TLS; how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n This document defines TLS version 1.3. While TLS 1.3 is not directly\n compatible with previous versions, all versions of TLS incorporate a\n versioning mechanism which allows clients and servers to\n interoperably negotiate a common version if one is supported by both\n peers.\n\n This document supersedes and obsoletes previous versions of TLS,\n including version 1.2 [RFC5246]. It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2. Because TLS 1.3 changes the way keys are\n derived, it updates [RFC5705] as described in Section 7.5. It also\n changes how Online Certificate Status Protocol (OCSP) messages are\n carried and therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in Section 4.4.2.1.\n\n1.1. Conventions and Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all\n capitals, as shown here.\n\n The following terms are used:\n\n client: The endpoint initiating the TLS connection.\n\n connection: A transport-layer connection between two endpoints.\n\n endpoint: Either the client or server of the connection.\n\n handshake: An initial negotiation between client and server that\n establishes the parameters of their subsequent interactions\n within TLS.\n\n peer: An endpoint. When discussing a particular endpoint, \"peer\"\n refers to the endpoint that is not the primary subject of\n discussion.\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\f\nRFC 8446 TLS August 2018\n\n\n receiver: An endpoint that is receiving records.\n\n sender: An endpoint that is transmitting records.\n\n server: The endpoint that did not initiate the TLS connection.\n\n1.2. Major Differences from TLS 1.2\n\n The following is a list of the major functional differences between\n TLS 1.2 and TLS 1.3. It is not intended to be exhaustive, and there\n are many minor differences.\n\n - The list of supported symmetric encryption algorithms has been\n pruned of all algorithms that are considered legacy. Those that\n remain are all Authenticated Encryption with Associated Data\n (AEAD) algorithms. The cipher suite concept has been changed to\n separate the authentication and key exchange mechanisms from the\n record protection algorithm (including secret key length) and a\n hash to be used with both the key derivation function and\n handshake message authentication code (MAC).\n\n - A zero round-trip time (0-RTT) mode was added, saving a round trip\n at connection setup for some application data, at the cost of\n certain security properties.\n\n - Static RSA and Diffie-Hellman cipher suites have been removed; all\n public-key based key exchange mechanisms now provide forward\n secrecy.\n\n - All handshake messages after the ServerHello are now encrypted.\n The newly introduced EncryptedExtensions message allows various\n extensions previously sent in the clear in the ServerHello to also\n enjoy confidentiality protection.\n\n - The key derivation functions have been redesigned. The new design\n allows easier analysis by cryptographers due to their improved key\n separation properties. The HMAC-based Extract-and-Expand Key\n Derivation Function (HKDF) is used as an underlying primitive.\n\n - The handshake state machine has been significantly restructured to\n be more consistent and to remove superfluous messages such as\n ChangeCipherSpec (except when needed for middlebox compatibility).\n\n - Elliptic curve algorithms are now in the base spec, and new\n signature algorithms, such as EdDSA, are included. TLS 1.3\n removed point format negotiation in favor of a single point format\n for each curve.\n\n\n\n\nRescorla Standards Track [Page 8]\n\f\nRFC 8446 TLS August 2018\n\n\n - Other cryptographic improvements were made, including changing the\n RSA padding to use the RSA Probabilistic Signature Scheme\n (RSASSA-PSS), and the removal of compression, the Digital\n Signature Algorithm (DSA), and custom Ephemeral Diffie-Hellman\n (DHE) groups.\n\n - The TLS 1.2 version negotiation mechanism has been deprecated in\n favor of a version list in an extension. This increases\n compatibility with existing servers that incorrectly implemented\n version negotiation.\n\n - Session resumption with and without server-side state as well as\n the PSK-based cipher suites of earlier TLS versions have been\n replaced by a single new PSK exchange.\n\n - References have been updated to point to the updated versions of\n RFCs, as appropriate (e.g., RFC 5280 rather than RFC 3280).\n\n1.3. Updates Affecting TLS 1.2\n\n This document defines several changes that optionally affect\n implementations of TLS 1.2, including those which do not also support\n TLS 1.3:\n\n - A version downgrade protection mechanism is described in\n Section 4.1.3.\n\n - RSASSA-PSS signature schemes are defined in Section 4.2.3.\n\n - The \"supported_versions\" ClientHello extension can be used to\n negotiate the version of TLS to use, in preference to the\n legacy_version field of the ClientHello.\n\n - The \"signature_algorithms_cert\" extension allows a client to\n indicate which signature algorithms it can validate in X.509\n certificates.\n\n Additionally, this document clarifies some compliance requirements\n for earlier versions of TLS; see Section 9.3.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 9]\n\f\nRFC 8446 TLS August 2018\n\n\n2. Protocol Overview\n\n The cryptographic parameters used by the secure channel are produced\n by the TLS handshake protocol. This sub-protocol of TLS is used by\n the client and server when first communicating with each other. The\n handshake protocol allows peers to negotiate a protocol version,\n select cryptographic algorithms, optionally authenticate each other,\n and establish shared secret keying material. Once the handshake is\n complete, the peers use the established keys to protect the\n application-layer traffic.\n\n A failure of the handshake or other protocol error triggers the\n termination of the connection, optionally preceded by an alert\n message (Section 6).\n\n TLS supports three basic key exchange modes:\n\n - (EC)DHE (Diffie-Hellman over either finite fields or elliptic\n curves)\n\n - PSK-only\n\n - PSK with (EC)DHE\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 10]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 1 below shows the basic full TLS handshake:\n\n Client Server\n\nKey ^ ClientHello\nExch | + key_share*\n | + signature_algorithms*\n | + psk_key_exchange_modes*\n v + pre_shared_key* -------->\n ServerHello ^ Key\n + key_share* | Exch\n + pre_shared_key* v\n {EncryptedExtensions} ^ Server\n {CertificateRequest*} v Params\n {Certificate*} ^\n {CertificateVerify*} | Auth\n {Finished} v\n <-------- [Application Data*]\n ^ {Certificate*}\nAuth | {CertificateVerify*}\n v {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 1: Message Flow for Full TLS Handshake\n\n The handshake can be thought of as having three phases (indicated in\n the diagram above):\n\n - Key Exchange: Establish shared keying material and select the\n cryptographic parameters. Everything after this phase is\n encrypted.\n\n - Server Parameters: Establish other handshake parameters\n (whether the client is authenticated, application-layer protocol\n support, etc.).\n\n\n\n\nRescorla Standards Track [Page 11]\n\f\nRFC 8446 TLS August 2018\n\n\n - Authentication: Authenticate the server (and, optionally, the\n client) and provide key confirmation and handshake integrity.\n\n In the Key Exchange phase, the client sends the ClientHello\n (Section 4.1.2) message, which contains a random nonce\n (ClientHello.random); its offered protocol versions; a list of\n symmetric cipher/HKDF hash pairs; either a set of Diffie-Hellman key\n shares (in the \"key_share\" (Section 4.2.8) extension), a set of\n pre-shared key labels (in the \"pre_shared_key\" (Section 4.2.11)\n extension), or both; and potentially additional extensions.\n Additional fields and/or messages may also be present for middlebox\n compatibility.\n\n The server processes the ClientHello and determines the appropriate\n cryptographic parameters for the connection. It then responds with\n its own ServerHello (Section 4.1.3), which indicates the negotiated\n connection parameters. The combination of the ClientHello and the\n ServerHello determines the shared keys. If (EC)DHE key establishment\n is in use, then the ServerHello contains a \"key_share\" extension with\n the server's ephemeral Diffie-Hellman share; the server's share MUST\n be in the same group as one of the client's shares. If PSK key\n establishment is in use, then the ServerHello contains a\n \"pre_shared_key\" extension indicating which of the client's offered\n PSKs was selected. Note that implementations can use (EC)DHE and PSK\n together, in which case both extensions will be supplied.\n\n The server then sends two messages to establish the Server\n Parameters:\n\n EncryptedExtensions: responses to ClientHello extensions that are\n not required to determine the cryptographic parameters, other than\n those that are specific to individual certificates.\n [Section 4.3.1]\n\n CertificateRequest: if certificate-based client authentication is\n desired, the desired parameters for that certificate. This\n message is omitted if client authentication is not desired.\n [Section 4.3.2]\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 12]\n\f\nRFC 8446 TLS August 2018\n\n\n Finally, the client and server exchange Authentication messages. TLS\n uses the same set of messages every time that certificate-based\n authentication is needed. (PSK-based authentication happens as a\n side effect of key exchange.) Specifically:\n\n Certificate: The certificate of the endpoint and any per-certificate\n extensions. This message is omitted by the server if not\n authenticating with a certificate and by the client if the server\n did not send CertificateRequest (thus indicating that the client\n should not authenticate with a certificate). Note that if raw\n public keys [RFC7250] or the cached information extension\n [RFC7924] are in use, then this message will not contain a\n certificate but rather some other value corresponding to the\n server's long-term key. [Section 4.4.2]\n\n CertificateVerify: A signature over the entire handshake using the\n private key corresponding to the public key in the Certificate\n message. This message is omitted if the endpoint is not\n authenticating via a certificate. [Section 4.4.3]\n\n Finished: A MAC (Message Authentication Code) over the entire\n handshake. This message provides key confirmation, binds the\n endpoint's identity to the exchanged keys, and in PSK mode also\n authenticates the handshake. [Section 4.4.4]\n\n Upon receiving the server's messages, the client responds with its\n Authentication messages, namely Certificate and CertificateVerify (if\n requested), and Finished.\n\n At this point, the handshake is complete, and the client and server\n derive the keying material required by the record layer to exchange\n application-layer data protected through authenticated encryption.\n Application Data MUST NOT be sent prior to sending the Finished\n message, except as specified in Section 2.3. Note that while the\n server may send Application Data prior to receiving the client's\n Authentication messages, any data sent at that point is, of course,\n being sent to an unauthenticated peer.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 13]\n\f\nRFC 8446 TLS August 2018\n\n\n2.1. Incorrect DHE Share\n\n If the client has not provided a sufficient \"key_share\" extension\n (e.g., it includes only DHE or ECDHE groups unacceptable to or\n unsupported by the server), the server corrects the mismatch with a\n HelloRetryRequest and the client needs to restart the handshake with\n an appropriate \"key_share\" extension, as shown in Figure 2. If no\n common cryptographic parameters can be negotiated, the server MUST\n abort the handshake with an appropriate alert.\n\n Client Server\n\n ClientHello\n + key_share -------->\n HelloRetryRequest\n <-------- + key_share\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 2: Message Flow for a Full Handshake with\n Mismatched Parameters\n\n Note: The handshake transcript incorporates the initial\n ClientHello/HelloRetryRequest exchange; it is not reset with the\n new ClientHello.\n\n TLS also allows several optimized variants of the basic handshake, as\n described in the following sections.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 14]\n\f\nRFC 8446 TLS August 2018\n\n\n2.2. Resumption and Pre-Shared Key (PSK)\n\n Although TLS PSKs can be established out of band, PSKs can also be\n established in a previous connection and then used to establish a new\n connection (\"session resumption\" or \"resuming\" with a PSK). Once a\n handshake has completed, the server can send the client a PSK\n identity that corresponds to a unique key derived from the initial\n handshake (see Section 4.6.1). The client can then use that PSK\n identity in future handshakes to negotiate the use of the associated\n PSK. If the server accepts the PSK, then the security context of the\n new connection is cryptographically tied to the original connection\n and the key derived from the initial handshake is used to bootstrap\n the cryptographic state instead of a full handshake. In TLS 1.2 and\n below, this functionality was provided by \"session IDs\" and \"session\n tickets\" [RFC5077]. Both mechanisms are obsoleted in TLS 1.3.\n\n PSKs can be used with (EC)DHE key exchange in order to provide\n forward secrecy in combination with shared keys, or can be used\n alone, at the cost of losing forward secrecy for the application\n data.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 15]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 3 shows a pair of handshakes in which the first handshake\n establishes a PSK and the second handshake uses it:\n\n Client Server\n\n Initial Handshake:\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n <-------- [NewSessionTicket]\n [Application Data] <-------> [Application Data]\n\n\n Subsequent Handshake:\n ClientHello\n + key_share*\n + pre_shared_key -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n {Finished}\n <-------- [Application Data*]\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 3: Message Flow for Resumption and PSK\n\n As the server is authenticating via a PSK, it does not send a\n Certificate or a CertificateVerify message. When a client offers\n resumption via a PSK, it SHOULD also supply a \"key_share\" extension\n to the server to allow the server to decline resumption and fall back\n to a full handshake, if needed. The server responds with a\n \"pre_shared_key\" extension to negotiate the use of PSK key\n establishment and can (as shown here) respond with a \"key_share\"\n extension to do (EC)DHE key establishment, thus providing forward\n secrecy.\n\n\n\n\nRescorla Standards Track [Page 16]\n\f\nRFC 8446 TLS August 2018\n\n\n When PSKs are provisioned out of band, the PSK identity and the KDF\n hash algorithm to be used with the PSK MUST also be provisioned.\n\n Note: When using an out-of-band provisioned pre-shared secret, a\n critical consideration is using sufficient entropy during the key\n generation, as discussed in [RFC4086]. Deriving a shared secret\n from a password or other low-entropy sources is not secure. A\n low-entropy secret, or password, is subject to dictionary attacks\n based on the PSK binder. The specified PSK authentication is not\n a strong password-based authenticated key exchange even when used\n with Diffie-Hellman key establishment. Specifically, it does not\n prevent an attacker that can observe the handshake from performing\n a brute-force attack on the password/pre-shared key.\n\n2.3. 0-RTT Data\n\n When clients and servers share a PSK (either obtained externally or\n via a previous handshake), TLS 1.3 allows clients to send data on the\n first flight (\"early data\"). The client uses the PSK to authenticate\n the server and to encrypt the early data.\n\n As shown in Figure 4, the 0-RTT data is just added to the 1-RTT\n handshake in the first flight. The rest of the handshake uses the\n same messages as for a 1-RTT handshake with PSK resumption.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 17]\n\f\nRFC 8446 TLS August 2018\n\n\n Client Server\n\n ClientHello\n + early_data\n + key_share*\n + psk_key_exchange_modes\n + pre_shared_key\n (Application Data*) -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n + early_data*\n {Finished}\n <-------- [Application Data*]\n (EndOfEarlyData)\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n () Indicates messages protected using keys\n derived from a client_early_traffic_secret.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 4: Message Flow for a 0-RTT Handshake\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 18]\n\f\nRFC 8446 TLS August 2018\n\n\n IMPORTANT NOTE: The security properties for 0-RTT data are weaker\n than those for other kinds of TLS data. Specifically:\n\n 1. This data is not forward secret, as it is encrypted solely under\n keys derived using the offered PSK.\n\n 2. There are no guarantees of non-replay between connections.\n Protection against replay for ordinary TLS 1.3 1-RTT data is\n provided via the server's Random value, but 0-RTT data does not\n depend on the ServerHello and therefore has weaker guarantees.\n This is especially relevant if the data is authenticated either\n with TLS client authentication or inside the application\n protocol. The same warnings apply to any use of the\n early_exporter_master_secret.\n\n\n\n===== DOCUMENT =====\n\n\nRFC 8446 TLS August 2018\n\n\n Sam Scott\n Royal Holloway, University of London\n [redacted-email]\n\n Thomas Shrimpton\n University of Florida\n [redacted-email]\n\n Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n Brian Smith\n Independent\n [redacted-email]\n\n Brian Sniffen\n Akamai Technologies\n [redacted-email]\n\n Nick Sullivan\n Cloudflare Inc.\n [redacted-email]\n\n Bjoern Tackmann\n University of California, San Diego\n [redacted-email]\n\n Tim Taubert\n Mozilla\n [redacted-email]\n\n Martin Thomson\n Mozilla\n [redacted-email]\n\n Hannes Tschofenig\n Arm Limited\n [redacted-email]\n\n Sean Turner\n sn3rd\n [redacted-email]\n\n Steven Valdez\n Google\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 159]\n\f\nRFC 8446 TLS August 2018\n\n\n Filippo Valsorda\n Cloudflare Inc.\n [redacted-email]\n\n Thyla van der Merwe\n Royal Holloway, University of London\n [redacted-email]\n\n Victor Vasiliev\n Google\n [redacted-email]\n\n Hoeteck Wee\n Ecole Normale Superieure, Paris\n [redacted-email]\n\n Tom Weinstein\n\n David Wong\n NCC Group\n [redacted-email]\n\n Christopher A. Wood\n Apple Inc.\n [redacted-email]\n\n Tim Wright\n Vodafone\n [redacted-email]\n\n Peter Wu\n Independent\n [redacted-email]\n\n Kazu Yamamoto\n Internet Initiative Japan Inc.\n [redacted-email]\n\nAuthor's Address\n\n Eric Rescorla\n Mozilla\n\n Email: [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 160]", "domain": "standards", "dossier_id": "a00388a585cd2f4d89b3", "essential_artifact_ids": ["ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"], "event_count": 6, "evidence_distance": 64677, "evidence_span_tokens": 64670, "executable_proof_id": "0c64de0f18afd43b8e63", "generation_integration": "task_replay_promotion_v1", "graph": {"hop_count": 2, "n_essential_artifacts": 7, "n_essential_events": 6, "proof_depth": 2}, "holdout": {"group_id": "af2707ad6cb50ab3", "strategy": "world"}, "hop_count": 2, "hybrid_causal_edges": [], "ietf_requirement_task": {"answer": {"current_protocol": "TLS_1_3_RFC8446", "obsoletes_ocsp_stapling": "RFC6961", "obsoletes_session_tickets": "RFC5077", "obsoletes_tls12": "RFC5246", "updates_exporters": "RFC5705", "updates_extensions": "RFC6066"}, "answer_program_id": "ietf.tls13_handshake_succession.v1", "cutoff": "2026-01-31T00:00:00Z", "essential_evidence_ids": ["current_protocol", "obsoletes_session_tickets", "obsoletes_tls12", "obsoletes_ocsp_stapling", "updates_exporters", "updates_extensions"], "essential_relation_ids": ["ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "ietf:updates:ietf:rfc:8446:ietf:rfc:6066"], "evidence_items": [{"char_end": 447, "char_start": 361, "evidence_id": "current_protocol", "evidence_quote": "This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol.", "quote_sha256": "213f297f8fd7940d76115d7115fd430d489bced81930ca9d9319c2eaddff65ea", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15703, "char_start": 15575, "evidence_id": "obsoletes_session_tickets", "evidence_quote": "It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2.", "quote_sha256": "957220687c3a075f9479d4799d2a2a6bc3065f4b5f968376a62cd724206781ff", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15573, "char_start": 15473, "evidence_id": "obsoletes_tls12", "evidence_quote": "This document supersedes and obsoletes previous versions of TLS,\n including version 1.2 [RFC5246].", "quote_sha256": "0ffd331074a612c946a7f65eb52848b1224f4497f7dad642bd1c34ac255fe340", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15989, "char_start": 15934, "evidence_id": "obsoletes_ocsp_stapling", "evidence_quote": "obsoletes [RFC6961] as\n described in Section 4.4.2.1.", "quote_sha256": "836a51c84442de8cd356808fe29a5bc51913ffc3248fa087461cea3fbe9f26cd", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15807, "char_start": 15705, "evidence_id": "updates_exporters", "evidence_quote": "Because TLS 1.3 changes the way keys are\n derived, it updates [RFC5705] as described in Section 7.5.", "quote_sha256": "e5c566bced7efdd9df1c7483ab71fa5ce896e674e1a35613667bd90d52341b0d", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 15989, "char_start": 15902, "evidence_id": "updates_extensions", "evidence_quote": "therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in Section 4.4.2.1.", "quote_sha256": "46d4d3c9d3c46874270771eceb979e3b19a7f09528933d4e9071941d096f5b9f", "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}], "query_type": "protocol_succession_resolution", "question": "Resolve the effective TLS 1.3 handshake protocol succession at 2026-01-31T00:00:00Z for this scenario (canonical JSON): {\"protocol\":\"tls\",\"publication\":\"rfc8446\",\"succession\":\"obsoletes_and_updates\",\"version\":\"1.3\"}. Return exactly one JSON object with these keys in this order: [\"current_protocol\",\"obsoletes_session_tickets\",\"obsoletes_tls12\",\"obsoletes_ocsp_stapling\",\"updates_exporters\",\"updates_extensions\"]. Use this exact per-field output codebook (canonical JSON): {\"current_protocol\":{\"code\":\"TLS_1_3_RFC8446\",\"meaning\":\"RFC 8446 is TLS protocol version 1.3\"},\"obsoletes_ocsp_stapling\":{\"code\":\"RFC6961\",\"meaning\":\"RFC 8446 obsoletes multiple certificate status requests\"},\"obsoletes_session_tickets\":{\"code\":\"RFC5077\",\"meaning\":\"RFC 8446 obsoletes TLS session resumption tickets\"},\"obsoletes_tls12\":{\"code\":\"RFC5246\",\"meaning\":\"RFC 8446 obsoletes TLS protocol version 1.2\"},\"updates_exporters\":{\"code\":\"RFC5705\",\"meaning\":\"RFC 8446 updates TLS keying material exporters\"},\"updates_extensions\":{\"code\":\"RFC6066\",\"meaning\":\"RFC 8446 updates TLS extension definitions\"}}. Resolve each field independently from the supplied RFC graph; use the string \"UNKNOWN\" only for a field whose required evidence or relation is absent.", "scenario": {"protocol": "tls", "publication": "rfc8446", "succession": "obsoletes_and_updates", "version": "1.3"}, "schema_version": "longworld.ietf-tls13-handshake-succession-task.v1", "source_manifest": {"authorization": {"allowed_actions": ["fetch_datatracker_document", "fetch_datatracker_relation", "fetch_draft_revision", "fetch_rfc"], "basis": "IETF open-records research export for a closed published_as plus Updates/Obsoletes graph. Distinct from HTTP/3/QUIC packing, OAuth, and RFC 9421 revision stacking. RFC 5077 and RFC 6961 are included because RFC 8446 authentically obsoletes them.", "record_id": "p57-ietf-tls13-handshake-graph-v1", "reviewed_at": "2026-09-07T04:00:00Z", "scope": "public IETF Datatracker, Internet-Draft archive, and RFC Editor records for the TLS 1.3 handshake requirement graph over RFC 8446 and its authentic Updates/Obsoletes targets"}, "data_stage": "source_inventory", "fetch_inventory_sha256": "c86bc411fe4e31382a729a63b451fff99251ebacc2b1323213b4fa8cba891c36", "fetch_receipt": {"allowed_actions": ["fetch_datatracker_document", "fetch_datatracker_relation", "fetch_draft_revision", "fetch_rfc"], "completed_at": "2026-09-07T04:14:41.225467Z", "max_retries": 2, "policy_url": "https://www.ietf.org/about/open-records/", "request_file": "ietf_fetch_request.json", "request_sha256": "577344d9c626e814c0e57a20755290a4e9bdaa66c1e7b545370366600ef34c5f", "requests_per_second": 1.0, "retrievals": [{"content_type": "application/json", "final_url": "https://datatracker.ietf.org/api/v1/doc/document/draft-ietf-tls-tls13/", "kind": "datatracker_document", "observed_at": "2026-09-07T04:14:30.932265Z", "redirect_chain": [], "requested_url": "https://datatracker.ietf.org/api/v1/doc/document/draft-ietf-tls-tls13/", "retrieval_file": "datatracker-draft-ietf-tls-tls13.json", "sha256": "a632cb66584c8f5d364d909b76eed121c1a9d00e341908463e019efad0faee0f", "status": 200}, {"content_type": "application/json", "final_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=draft-ietf-tls-tls13&limit=100", "kind": "datatracker_relation", "observed_at": "2026-09-07T04:14:32.295913Z", "redirect_chain": [], "requested_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=draft-ietf-tls-tls13&limit=100", "retrieval_file": "datatracker-draft-ietf-tls-tls13-relations.json", "sha256": "3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "kind": "draft_revision", "observed_at": "2026-09-07T04:14:32.641074Z", "redirect_chain": [], "requested_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "retrieval_file": "draft-ietf-tls-tls13-28.txt", "sha256": "9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:33.763234Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "retrieval_file": "rfc5077.txt", "sha256": "e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:34.735753Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "retrieval_file": "rfc5246.txt", "sha256": "2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:35.724001Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "retrieval_file": "rfc5705.txt", "sha256": "29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:36.718981Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "retrieval_file": "rfc6066.txt", "sha256": "6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:37.734605Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "retrieval_file": "rfc6961.txt", "sha256": "c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db", "status": 200}, {"content_type": "text/plain", "final_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "kind": "rfc", "observed_at": "2026-09-07T04:14:38.737647Z", "redirect_chain": [], "requested_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "retrieval_file": "rfc8446.txt", "sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd", "status": 200}, {"content_type": "application/json", "final_url": "https://datatracker.ietf.org/api/v1/doc/document/rfc8446/", "kind": "datatracker_document", "observed_at": "2026-09-07T04:14:39.917810Z", "redirect_chain": [], "requested_url": "https://datatracker.ietf.org/api/v1/doc/document/rfc8446/", "retrieval_file": "datatracker-rfc8446.json", "sha256": "89b01010d1ff76dc260539d5a453dde0b116d857db43c9e09d4a3dc062c9ce31", "status": 200}, {"content_type": "application/json", "final_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=rfc8446&limit=100", "kind": "datatracker_relation", "observed_at": "2026-09-07T04:14:41.148823Z", "redirect_chain": [], "requested_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=rfc8446&limit=100", "retrieval_file": "datatracker-rfc8446-relations.json", "sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7", "status": 200}], "started_at": "2026-09-07T04:14:30.455103Z", "user_agent_sha256": "1c8887e16ad14ac078ad6e7a52c6cdf9f988eb709fa5671439746106b1265652"}, "generated_at": "2026-09-07T04:16:16.549092Z", "generation_integration": "disabled", "hybrid_train_ready": false, "n": 7, "n_relations": 11, "n_supporting_records": 4, "production_eligible": false, "records": [{"draft_references": [], "excluded_rfc_relation_targets": [{"kind": "obsoletes", "reason": "not_requested", "rfc_number": 4507}], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5077 H. Zhou", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "5077"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 4507 Cisco Systems", "fact_id": "obsoletes:4507", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "4507"}], "kind": "rfc", "occurred_at": "2008-01-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 4, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef", "record_id": "ietf:rfc:5077", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "rfc_number": 5077, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef", "source_url": "https://www.rfc-editor.org/rfc/rfc5077.txt", "temporal_semantics": "source_event_time", "text": "Network Working Group J. Salowey\nRequest for Comments: 5077 H. Zhou\nObsoletes: 4507 Cisco Systems\nCategory: Standards Track P. Eronen\n Nokia\n H. Tschofenig\n Nokia Siemens Networks\n January 2008\n\n\n Transport Layer Security (TLS) Session Resumption without\n Server-Side State\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document describes a mechanism that enables the Transport Layer\n Security (TLS) server to resume sessions and avoid keeping per-client\n session state. The TLS server encapsulates the session state into a\n ticket and forwards it to the client. The client can subsequently\n resume a session using the obtained ticket. This document obsoletes\n RFC 4507.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 1]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3. Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . 3\n 3.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . 4\n 3.2. SessionTicket TLS Extension . . . . . . . . . . . . . . . 7\n 3.3. NewSessionTicket Handshake Message . . . . . . . . . . . . 8\n 3.4. Interaction with TLS Session ID . . . . . . . . . . . . . 9\n 4. Recommended Ticket Construction . . . . . . . . . . . . . . . 10\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . 12\n 5.1. Invalidating Sessions . . . . . . . . . . . . . . . . . . 12\n 5.2. Stolen Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.3. Forged Tickets . . . . . . . . . . . . . . . . . . . . . . 12\n 5.4. Denial of Service Attacks . . . . . . . . . . . . . . . . 12\n 5.5. Ticket Protection Key Management . . . . . . . . . . . . . 13\n 5.6. Ticket Lifetime . . . . . . . . . . . . . . . . . . . . . 13\n 5.7. Alternate Ticket Formats and Distribution Schemes . . . . 13\n 5.8. Identity Privacy, Anonymity, and Unlinkability . . . . . . 14\n 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 14\n 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 15\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 15\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 15\n Appendix A. Discussion of Changes to RFC 4507 . . . . . . . . . . 17\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 2]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n1. Introduction\n\n This document defines a way to resume a Transport Layer Security\n (TLS) session without requiring session-specific state at the TLS\n server. This mechanism may be used with any TLS ciphersuite. This\n document applies to both TLS 1.0 defined in [RFC2246], and TLS 1.1\n defined in [RFC4346]. The mechanism makes use of TLS extensions\n defined in [RFC4366] and defines a new TLS message type.\n\n This mechanism is useful in the following situations:\n\n 1. servers that handle a large number of transactions from different\n users\n\n 2. servers that desire to cache sessions for a long time\n\n 3. ability to load balance requests across servers\n\n 4. embedded servers with little memory\n\n This document obsoletes RFC 4507 [RFC4507] to correct an error in the\n encoding that caused the specification to differ from deployed\n implementations. At the time of this writing, there are no known\n implementations that follow the encoding specified in RFC 4507. This\n update to RFC 4507 aligns the document with currently deployed\n implementations. More details of the change are given in Appendix A.\n\n2. Terminology\n\n Within this document, the term 'ticket' refers to a cryptographically\n protected data structure that is created and consumed by the server\n to rebuild session-specific state.\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Protocol\n\n This specification describes a mechanism to distribute encrypted\n session-state information to the client in the form of a ticket and a\n mechanism to present the ticket back to the server. The ticket is\n created by a TLS server and sent to a TLS client. The TLS client\n presents the ticket to the TLS server to resume a session.\n Implementations of this specification are expected to support both\n mechanisms. Other specifications can take advantage of the session\n tickets, perhaps specifying alternative means for distribution or\n selection. For example, a separate specification may describe an\n\n\n\nSalowey, et al. Standards Track [Page 3]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n alternate way to distribute a ticket and use the TLS extension in\n this document to resume the session. This behavior is beyond the\n scope of the document and would need to be described in a separate\n specification.\n\n3.1. Overview\n\n The client indicates that it supports this mechanism by including a\n SessionTicket TLS extension in the ClientHello message. The\n extension will be empty if the client does not already possess a\n ticket for the server. The server sends an empty SessionTicket\n extension to indicate that it will send a new session ticket using\n the NewSessionTicket handshake message. The extension is described\n in Section 3.2.\n\n If the server wants to use this mechanism, it stores its session\n state (such as ciphersuite and master secret) to a ticket that is\n encrypted and integrity-protected by a key known only to the server.\n The ticket is distributed to the client using the NewSessionTicket\n TLS handshake message described in Section 3.3. This message is sent\n during the TLS handshake before the ChangeCipherSpec message, after\n the server has successfully verified the client's Finished message.\n\n Client Server\n\n ClientHello\n (empty SessionTicket extension)-------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1: Message Flow for Full Handshake Issuing New Session Ticket\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 4]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n The client caches this ticket along with the master secret and other\n parameters associated with the current session. When the client\n wishes to resume the session, it includes the ticket in the\n SessionTicket extension within the ClientHello message. Appendix A\n provides a detailed description of the encoding of the extension and\n changes from RFC 4507. The server then decrypts the received ticket,\n verifies the ticket's validity, retrieves the session state from the\n contents of the ticket, and uses this state to resume the session.\n The interaction with the TLS Session ID is described in Section 3.4.\n If the server successfully verifies the client's ticket, then it may\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello.\n\n Client Server\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2: Message Flow for Abbreviated Handshake Using New Session\n Ticket\n\n A recommended ticket format is given in Section 4.\n\n If the server cannot or does not want to honor the ticket, then it\n can initiate a full handshake with the client.\n\n In the case that the server does not wish to issue a new ticket at\n this time, it just completes the handshake without including a\n SessionTicket extension or NewSessionTicket handshake message. This\n is shown below (this flow is identical to Figure 1 in RFC 4346,\n except for the SessionTicket extension in the first message):\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 5]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 3: Message Flow for Server Completing Full Handshake Without\n Issuing New Session Ticket\n\n It is also permissible to have an exchange similar to Figure 3 using\n the abbreviated handshake defined in Figure 2 of RFC 4346, where the\n client uses the SessionTicket extension to resume the session, but\n the server does not wish to issue a new ticket, and therefore does\n not send a SessionTicket extension.\n\n If the server rejects the ticket, it may still wish to issue a new\n ticket after performing the full handshake as shown below (this flow\n is identical to Figure 1, except the SessionTicket extension in the\n ClientHello is not empty):\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 6]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n Client Server\n\n ClientHello\n (SessionTicket extension) -------->\n ServerHello\n (empty SessionTicket extension)\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n NewSessionTicket\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 4: Message Flow for Server Rejecting Ticket, Performing Full\n Handshake, and Issuing New Session Ticket\n\n3.2. SessionTicket TLS Extension\n\n The SessionTicket TLS extension is based on [RFC4366]. The format of\n the ticket is an opaque structure used to carry session-specific\n state information. This extension may be sent in the ClientHello and\n ServerHello.\n\n If the client possesses a ticket that it wants to use to resume a\n session, then it includes the ticket in the SessionTicket extension\n in the ClientHello. If the client does not have a ticket and is\n prepared to receive one in the NewSessionTicket handshake message,\n then it MUST include a zero-length ticket in the SessionTicket\n extension. If the client is not prepared to receive a ticket in the\n NewSessionTicket handshake message, then it MUST NOT include a\n SessionTicket extension unless it is sending a non-empty ticket it\n received through some other means from the server.\n\n The server uses a zero-length SessionTicket extension to indicate to\n the client that it will send a new session ticket using the\n NewSessionTicket handshake message described in Section 3.3. The\n server MUST send this extension in the ServerHello if it wishes to\n issue a new ticket to the client using the NewSessionTicket handshake\n message. The server MUST NOT send this extension if it does not\n receive one in the ClientHello.\n\n\n\n\nSalowey, et al. Standards Track [Page 7]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n If the server fails to verify the ticket, then it falls back to\n performing a full handshake. If the ticket is accepted by the server\n but the handshake fails, the client SHOULD delete the ticket.\n\n The SessionTicket extension has been assigned the number 35. The\n extension_data field of SessionTicket extension contains the ticket.\n\n3.3. NewSessionTicket Handshake Message\n\n This message is sent by the server during the TLS handshake before\n the ChangeCipherSpec message. This message MUST be sent if the\n server included a SessionTicket extension in the ServerHello. This\n message MUST NOT be sent if the server did not include a\n SessionTicket extension in the ServerHello. This message is included\n in the hash used to create and verify the Finished message. In the\n case of a full handshake, the server MUST verify the client's\n Finished message before sending the ticket. The client MUST NOT\n treat the ticket as valid until it has verified the server's Finished\n message. If the server determines that it does not want to include a\n ticket after it has included the SessionTicket extension in the\n ServerHello, then it sends a zero-length ticket in the\n NewSessionTicket handshake message.\n\n If the server successfully verifies the client's ticket, then it MAY\n renew the ticket by including a NewSessionTicket handshake message\n after the ServerHello in the abbreviated handshake. The client\n should start using the new ticket as soon as possible after it\n verifies the server's Finished message for new connections. Note\n that since the updated ticket is issued before the handshake\n completes, it is possible that the client may not put the new ticket\n into use before it initiates new connections. The server MUST NOT\n assume that the client actually received the updated ticket until it\n successfully verifies the client's Finished message.\n\n The NewSessionTicket handshake message has been assigned the number 4\n and its definition is given at the end of this section. The\n ticket_lifetime_hint field contains a hint from the server about how\n long the ticket should be stored. The value indicates the lifetime\n in seconds as a 32-bit unsigned integer in network byte order\n relative to when the ticket is received. A value of zero is reserved\n to indicate that the lifetime of the ticket is unspecified. A client\n SHOULD delete the ticket and associated state when the time expires.\n It MAY delete the ticket earlier based on local policy. A server MAY\n treat a ticket as valid for a shorter or longer period of time than\n what is stated in the ticket_lifetime_hint.\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 8]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n HandshakeType msg_type;\n uint24 length;\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case session_ticket: NewSessionTicket; /* NEW */\n } body;\n } Handshake;\n\n\n struct {\n uint32 ticket_lifetime_hint;\n opaque ticket<0..2^16-1>;\n } NewSessionTicket;\n\n3.4. Interaction with TLS Session ID\n\n If a server is planning on issuing a session ticket to a client that\n does not present one, it SHOULD include an empty Session ID in the\n ServerHello. If the server rejects the ticket and falls back to the\n full handshake then it may include a non-empty Session ID to indicate\n its support for stateful session resumption. If the client receives\n a session ticket from the server, then it discards any Session ID\n that was sent in the ServerHello.\n\n When presenting a ticket, the client MAY generate and include a\n Session ID in the TLS ClientHello. If the server accepts the ticket\n and the Session ID is not empty, then it MUST respond with the same\n Session ID present in the ClientHello. This allows the client to\n easily differentiate when the server is resuming a session from when\n it is falling back to a full handshake. Since the client generates a\n Session ID, the server MUST NOT rely upon the Session ID having a\n particular value when validating the ticket. If a ticket is\n presented by the client, the server MUST NOT attempt to use the\n Session ID in the ClientHello for stateful session resumption.\n Alternatively, the client MAY include an empty Session ID in the\n ClientHello. In this case, the client ignores the Session ID sent in\n the ServerHello and determines if the server is resuming a session by\n the subsequent handshake messages.\n\n\n\nSalowey, et al. Standards Track [Page 9]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n4. Recommended Ticket Construction\n\n This section describes a recommended format and protection for the\n ticket. Note that the ticket is opaque to the client, so the\n structure is not subject to interoperability concerns, and\n implementations may diverge from this format. If implementations do\n diverge from this format, they must take security concerns seriously.\n Clients MUST NOT examine the ticket under the assumption that it\n complies with this document.\n\n The server uses two different keys: one 128-bit key for Advanced\n Encryption Standard (AES) [AES] in Cipher Block Chaining (CBC) mode\n [CBC] encryption and one 256-bit key for HMAC-SHA-256 [RFC4634].\n\n The ticket is structured as follows:\n\n struct {\n opaque key_name[16];\n opaque iv[16];\n opaque encrypted_state<0..2^16-1>;\n opaque mac[32];\n } ticket;\n\n Here, key_name serves to identify a particular set of keys used to\n protect the ticket. It enables the server to easily recognize\n tickets it has issued. The key_name should be randomly generated to\n avoid collisions between servers. One possibility is to generate new\n random keys and key_name every time the server is started.\n\n The actual state information in encrypted_state is encrypted using\n 128-bit AES in CBC mode with the given IV. The Message\n Authentication Code (MAC) is calculated using HMAC-SHA-256 over\n key_name (16 octets) and IV (16 octets), followed by the length of\n the encrypted_state field (2 octets) and its contents (variable\n length).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 10]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n struct {\n ProtocolVersion protocol_version;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n opaque master_secret[48];\n ClientIdentity client_identity;\n uint32 timestamp;\n } StatePlaintext;\n\n enum {\n anonymous(0),\n certificate_based(1),\n psk(2)\n } ClientAuthenticationType;\n\n struct {\n ClientAuthenticationType client_authentication_type;\n select (ClientAuthenticationType) {\n case anonymous: struct {};\n case certificate_based:\n ASN.1Cert certificate_list<0..2^24-1>;\n case psk:\n opaque psk_identity<0..2^16-1>; /* from [RFC4279] */\n };\n } ClientIdentity;\n\n The structure StatePlaintext stores the TLS session state including\n the master_secret. The timestamp within this structure allows the\n TLS server to expire tickets. To cover the authentication and key\n exchange protocols provided by TLS, the ClientIdentity structure\n contains the authentication type of the client used in the initial\n exchange (see ClientAuthenticationType). To offer the TLS server\n with the same capabilities for authentication and authorization, a\n certificate list is included in case of public-key-based\n authentication. The TLS server is therefore able to inspect a number\n of different attributes within these certificates. A specific\n implementation might choose to store a subset of this information or\n additional information. Other authentication mechanisms, such as\n Kerberos [RFC2712], would require different client identity data.\n Other TLS extensions may require the inclusion of additional data in\n the StatePlaintext structure.\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 11]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5. Security Considerations\n\n This section addresses security issues related to the usage of a\n ticket. Tickets must be authenticated and encrypted to prevent\n modification or eavesdropping by an attacker. Several attacks\n described below will be possible if this is not carefully done.\n\n Implementations should take care to ensure that the processing of\n tickets does not increase the chance of denial of service as\n described below.\n\n5.1. Invalidating Sessions\n\n The TLS specification requires that TLS sessions be invalidated when\n errors occur. [CSSC] discusses the security implications of this in\n detail. In the analysis within this paper, failure to invalidate\n sessions does not pose a security risk. This is because the TLS\n handshake uses a non-reversible function to derive keys for a session\n so information about one session does not provide an advantage to\n attack the master secret or a different session. If a session\n invalidation scheme is used, the implementation should verify the\n integrity of the ticket before using the contents to invalidate a\n session to ensure that an attacker cannot invalidate a chosen\n session.\n\n5.2. Stolen Tickets\n\n An eavesdropper or man-in-the-middle may obtain the ticket and\n attempt to use it to establish a session with the server; however,\n since the ticket is encrypted and the attacker does not know the\n secret key, a stolen ticket does not help an attacker resume a\n session. A TLS server MUST use strong encryption and integrity\n protection for the ticket to prevent an attacker from using a brute\n force mechanism to obtain the ticket's contents.\n\n5.3. Forged Tickets\n\n A malicious user could forge or alter a ticket in order to resume a\n session, to extend its lifetime, to impersonate another user, or to\n gain additional privileges. This attack is not possible if the\n ticket is protected using a strong integrity protection algorithm\n such as a keyed HMAC-SHA-256.\n\n5.4. Denial of Service Attacks\n\n The key_name field defined in the recommended ticket format helps the\n server efficiently reject tickets that it did not issue. However, an\n adversary could store or generate a large number of tickets to send\n\n\n\nSalowey, et al. Standards Track [Page 12]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n to the TLS server for verification. To minimize the possibility of a\n denial of service, the verification of the ticket should be\n lightweight (e.g., using efficient symmetric key cryptographic\n algorithms).\n\n5.5. Ticket Protection Key Management\n\n A full description of the management of the keys used to protect the\n ticket is beyond the scope of this document. A list of RECOMMENDED\n practices is given below.\n\n o The keys should be generated securely following the randomness\n recommendations in [RFC4086].\n\n o The keys and cryptographic protection algorithms should be at\n least 128 bits in strength. Some ciphersuites and applications\n may require cryptographic protection greater than 128 bits in\n strength.\n\n o The keys should not be used for any purpose other than generating\n and verifying tickets.\n\n o The keys should be changed regularly.\n\n o The keys should be changed if the ticket format or cryptographic\n protection algorithms change.\n\n5.6. Ticket Lifetime\n\n The TLS server controls the lifetime of the ticket. Servers\n determine the acceptable lifetime based on the operational and\n security requirements of the environments in which they are deployed.\n The ticket lifetime may be longer than the 24-hour lifetime\n recommended in [RFC4346]. TLS clients may be given a hint of the\n lifetime of the ticket. Since the lifetime of a ticket may be\n unspecified, a client has its own local policy that determines when\n it discards tickets.\n\n5.7. Alternate Ticket Formats and Distribution Schemes\n\n If the ticket format or distribution scheme defined in this document\n is not used, then great care must be taken in analyzing the security\n of the solution. In particular, if confidential information, such as\n a secret key, is transferred to the client, it MUST be done using\n secure communication so as to prevent attackers from obtaining or\n modifying the key. Also, the ticket MUST have its integrity and\n confidentiality protected with strong cryptographic techniques to\n prevent a breach in the security of the system.\n\n\n\nSalowey, et al. Standards Track [Page 13]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n5.8. Identity Privacy, Anonymity, and Unlinkability\n\n This document mandates that the content of the ticket is\n confidentiality protected in order to avoid leakage of its content,\n such as user-relevant information. As such, it prevents disclosure\n of potentially sensitive information carried within the ticket.\n\n The initial handshake exchange, which was used to obtain the ticket,\n might not provide identity confidentiality of the client based on the\n properties of TLS. Another relevant security threat is the ability\n for an on-path adversary to observe multiple TLS handshakes where the\n same ticket is used, therefore concluding they belong to the same\n communication endpoints. Application designers that use the ticket\n mechanism described in this document should consider that\n unlinkability [ANON] is not necessarily provided.\n\n While a full discussion of these topics is beyond the scope of this\n document, it should be noted that it is possible to issue a ticket\n using a TLS renegotiation handshake that occurs after a secure tunnel\n has been established by a previous handshake. This may help address\n some privacy and unlinkability issues in some environments.\n\n6. Acknowledgements\n\n The authors would like to thank the following people for their help\n with preparing and reviewing this document: Eric Rescorla, Mohamad\n Badra, Tim Dierks, Nelson Bolyard, Nancy Cam-Winget, David McGrew,\n Rob Dugal, Russ Housley, Amir Herzberg, Bernard Aboba, and members of\n the TLS working group.\n\n [CSSC] describes a solution that is very similar to the one described\n in this document and gives a detailed analysis of the security\n considerations involved. [RFC2712] describes a mechanism for using\n Kerberos [RFC4120] in TLS ciphersuites, which helped inspire the use\n of tickets to avoid server state. [RFC4851] makes use of a similar\n mechanism to avoid maintaining server state for the cryptographic\n tunnel. [SC97] also investigates the concept of stateless sessions.\n\n The authors would also like to thank Jan Nordqvist, who found the\n encoding error in RFC 4507, corrected by this document. In addition\n Nagendra Modadugu, Wan-Teh Chang, and Michael D'Errico provided\n useful feedback during the review of this document.\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 14]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n7. IANA Considerations\n\n IANA has assigned a TLS extension number of 35 to the SessionTicket\n TLS extension from the TLS registry of ExtensionType values defined\n in [RFC4366].\n\n IANA has assigned a TLS HandshakeType number 4 to the\n NewSessionTicket handshake type from the TLS registry of\n HandshakeType values defined in [RFC4346].\n\n This document does not require any actions or assignments from IANA.\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2246] Dierks, T. and C. Allen, \"The TLS Protocol Version 1.0\",\n RFC 2246, January 1999.\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346, April 2006.\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, April 2006.\n\n [RFC4507] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 4507, May 2006.\n\n8.2. Informative References\n\n [AES] National Institute of Standards and Technology, \"Advanced\n Encryption Standard (AES)\", Federal Information Processing\n Standards (FIPS) Publication 197, November 2001.\n\n [ANON] Pfitzmann, A. and M. Hansen, \"Anonymity, Unlinkability,\n Unobservability, Pseudonymity, and Identity Management - A\n Consolidated Proposal for Terminology\", http://\n dud.inf.tu-dresden.de/literatur/\n Anon_Terminology_v0.26-1.pdf Version 0.26, December 2005.\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 15]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n [CBC] National Institute of Standards and Technology,\n \"Recommendation for Block Cipher Modes of Operation -\n Methods and Techniques\", NIST Special Publication 800-38A,\n December 2001.\n\n [CSSC] Shacham, H., Boneh, D., and E. Rescorla, \"Client-side\n caching for TLS\", Transactions on Information and System\n Security (TISSEC) , Volume 7, Issue 4, November 2004.\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [RFC4086] Eastlake, D., Schiller, J., and S. Crocker, \"Randomness\n Requirements for Security\", BCP 106, RFC 4086, June 2005.\n\n [RFC4120] Neuman, C., Yu, T., Hartman, S., and K. Raeburn, \"The\n Kerberos Network Authentication Service (V5)\", RFC 4120,\n July 2005.\n\n [RFC4279] Eronen, P. and H. Tschofenig, \"Pre-Shared Key Ciphersuites\n for Transport Layer Security (TLS)\", RFC 4279,\n December 2005.\n\n [RFC4634] Eastlake, D. and T. Hansen, \"US Secure Hash Algorithms\n (SHA and HMAC-SHA)\", RFC 4634, July 2006.\n\n [RFC4851] Cam-Winget, N., McGrew, D., Salowey, J., and H. Zhou, \"The\n Flexible Authentication via Secure Tunneling Extensible\n Authentication Protocol Method (EAP-FAST)\", RFC 4851,\n May 2007.\n\n [SC97] Aura, T. and P. Nikander, \"Stateless Connections\",\n Proceedings of the First International Conference on\n Information and Communication Security (ICICS '97) , 1997.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 16]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAppendix A. Discussion of Changes to RFC 4507\n\n RFC 4507 [RFC4507] defines a mechanism to resume a TLS session\n without maintaining server side state by specifying an encrypted\n ticket that is maintained on the client. The client presents this\n ticket to the server in a SessionTicket hello extension. The\n encoding in RFC 4507 used the XDR style encoding specified in TLS\n [RFC4346].\n\n An error in the encoding caused the specification to differ from\n deployed implementations. At the time of this writing there are no\n known implementations that follow the encoding specified in RFC 4507.\n This update to RFC 4507 aligns the document with these currently\n deployed implementations.\n\n Erroneous encoding in RFC 4507 resulted in two length fields; one for\n the extension contents and one for the ticket itself. Hence, for a\n ticket that is 256 bytes long and begins with the hex value FF FF,\n the encoding of the extension would be as follows according to RFC\n 4507:\n\n 00 23 Ticket Extension type 35\n 01 02 Length of extension contents\n 01 00 Length of ticket\n FF FF .. .. Actual ticket\n\n The update proposed in this document reflects what implementations\n actually encode, namely it removes the redundant length field. So,\n for a ticket that is 256 bytes long and begins with the hex value FF\n FF, the encoding of the extension would be as follows according to\n this update:\n\n 00 23 Extension type 35\n 01 00 Length of extension contents (ticket)\n FF FF .. .. Actual ticket\n\n A server implemented according to RFC 4507 receiving a ticket\n extension from a client conforming to this document would interpret\n the first two bytes of the ticket as the length of this ticket. This\n will result in either an inconsistent length field or in the\n processing of a ticket missing the first two bytes. In the first\n case, the server should reject the request based on a malformed\n length. In the second case, the server should reject the ticket\n based on a malformed ticket, incorrect key version, or failed\n decryption. A server implementation based on this update receiving\n an RFC 4507 extension would interpret the first length field as the\n\n\n\n\n\nSalowey, et al. Standards Track [Page 17]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\n length of the ticket and include the second two length bytes as the\n first bytes in the ticket, resulting in the ticket being rejected\n based on a malformed ticket, incorrect key version, or failed\n decryption.\n\n Note that the encoding of an empty SessionTicket extension was\n ambiguous in RFC 4507. An RFC 4507 implementation may have encoded\n it as:\n\n 00 23 Extension type 35\n 00 02 Length of extension contents\n 00 00 Length of ticket\n\n or it may have encoded it the same way as this update:\n\n 00 23 Extension type 35\n 00 00 Length of extension contents\n\n A server wishing to support RFC 4507 clients should respond to an\n empty SessionTicket extension encoded the same way as it received it.\n\n A server implementation can construct tickets such that it can detect\n an RFC 4507 implementation, if one existed, by including a cookie at\n the beginning of the tickets that can be differentiated from a valid\n length. For example, if an implementation constructed tickets to\n start with the hex values FF FF, then it could determine where the\n ticket begins and determine the length correctly from the type of\n length fields present.\n\n This document makes a few additional changes to RFC 4507 listed\n below.\n\n o Clarifying that the server can allow session resumption using a\n ticket without issuing a new ticket in Section 3.1.\n\n o Clarifying that the lifetime is relative to when the ticket is\n received in section 3.3.\n\n o Clarifying that the NewSessionTicket handshake message is included\n in the hash generated for the Finished messages in Section 3.3.\n\n o Clarifying the interaction with TLS Session ID in Section 3.4.\n\n o Recommending the use of SHA-256 for the integrity protection of\n the ticket in Section 4.\n\n o Clarifying that additional data can be included in the\n StatePlaintext structure in Section 4.\n\n\n\nSalowey, et al. Standards Track [Page 18]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nAuthors' Addresses\n\n Joseph Salowey\n Cisco Systems\n 2901 3rd Ave\n Seattle, WA 98121\n US\n\n EMail: [redacted-email]\n\n\n Hao Zhou\n Cisco Systems\n 4125 Highlander Parkway\n Richfield, OH 44286\n US\n\n EMail: [redacted-email]\n\n\n Pasi Eronen\n Nokia Research Center\n P.O. Box 407\n FIN-00045 Nokia Group\n Finland\n\n EMail: [redacted-email]\n\n\n Hannes Tschofenig\n Nokia Siemens Networks\n Otto-Hahn-Ring 6\n Munich, Bayern 81739\n Germany\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 19]\n\f\nRFC 5077 Stateless TLS Session Resumption January 2008\n\n\nFull Copyright Statement\n\n Copyright (C) The IETF Trust (2008).\n\n This document is subject to the rights, licenses and restrictions\n contained in BCP 78, and except as set forth therein, the authors\n retain all their rights.\n\n This document and the information contained herein are provided on an\n \"AS IS\" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\n OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\n THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\n OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\n THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\n WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\n\nIntellectual Property\n\n The IETF takes no position regarding the validity or scope of any\n Intellectual Property Rights or other rights that might be claimed to\n pertain to the implementation or use of the technology described in\n this document or the extent to which any license under such rights\n might or might not be available; nor does it represent that it has\n made any independent effort to identify any such rights. Information\n on the procedures with respect to rights in RFC documents can be\n found in BCP 78 and BCP 79.\n\n Copies of IPR disclosures made to the IETF Secretariat and any\n assurances of licenses to be made available, or the result of an\n attempt made to obtain a general license or permission for the use of\n such proprietary rights by implementers or users of this\n specification can be obtained from the IETF on-line IPR repository at\n http://www.ietf.org/ipr.\n\n The IETF invites any interested party to bring to its attention any\n copyrights, patents or patent applications, or other proprietary\n rights that may cover technology that may be required to implement\n this standard. Please address the information to the IETF at\n ietf-ipr@ietf.org.\n\n\n\n\n\n\n\n\n\n\n\n\nSalowey, et al. Standards Track [Page 20]", "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"draft_references": [], "excluded_rfc_relation_targets": [{"kind": "obsoletes", "reason": "not_requested", "rfc_number": 3268}, {"kind": "obsoletes", "reason": "not_requested", "rfc_number": 4346}, {"kind": "obsoletes", "reason": "not_requested", "rfc_number": 4366}, {"kind": "updates", "reason": "not_requested", "rfc_number": 4492}], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5246 Independent", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "5246"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 3268, 4346, 4366 E. Rescorla", "fact_id": "obsoletes:3268", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "3268"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 3268, 4346, 4366 E. Rescorla", "fact_id": "obsoletes:4346", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "4346"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 3268, 4346, 4366 E. Rescorla", "fact_id": "obsoletes:4366", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "4366"}, {"char_end": 291, "char_start": 219, "evidence_quote": "Updates: 4492 RTFM, Inc.", "fact_id": "updates:4492", "field": "updates", "raw_char_end": 297, "raw_char_start": 225, "value": "4492"}], "kind": "rfc", "occurred_at": "2008-08-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 22, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb", "record_id": "ietf:rfc:5246", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "rfc_number": 5246, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb", "source_url": "https://www.rfc-editor.org/rfc/rfc5246.txt", "temporal_semantics": "source_event_time", "text": "Network Working Group T. Dierks\nRequest for Comments: 5246 Independent\nObsoletes: 3268, 4346, 4366 E. Rescorla\nUpdates: 4492 RTFM, Inc.\nCategory: Standards Track August 2008\n\n\n The Transport Layer Security (TLS) Protocol\n Version 1.2\n\nStatus of This Memo\n\n This document specifies an Internet standards track protocol for the\n Internet community, and requests discussion and suggestions for\n improvements. Please refer to the current edition of the \"Internet\n Official Protocol Standards\" (STD 1) for the standardization state\n and status of this protocol. Distribution of this memo is unlimited.\n\nAbstract\n\n This document specifies Version 1.2 of the Transport Layer Security\n (TLS) protocol. The TLS protocol provides communications security\n over the Internet. The protocol allows client/server applications to\n communicate in a way that is designed to prevent eavesdropping,\n tampering, or message forgery.\n\nTable of Contents\n\n 1. Introduction ....................................................4\n 1.1. Requirements Terminology ...................................5\n 1.2. Major Differences from TLS 1.1 .............................5\n 2. Goals ...........................................................6\n 3. Goals of This Document ..........................................7\n 4. Presentation Language ...........................................7\n 4.1. Basic Block Size ...........................................7\n 4.2. Miscellaneous ..............................................8\n 4.3. Vectors ....................................................8\n 4.4. Numbers ....................................................9\n 4.5. Enumerateds ................................................9\n 4.6. Constructed Types .........................................10\n 4.6.1. Variants ...........................................10\n 4.7. Cryptographic Attributes ..................................12\n 4.8. Constants .................................................14\n 5. HMAC and the Pseudorandom Function .............................14\n 6. The TLS Record Protocol ........................................15\n 6.1. Connection States .........................................16\n 6.2. Record Layer ..............................................19\n 6.2.1. Fragmentation ......................................19\n\n\n\nDierks & Rescorla Standards Track [Page 1]\n\f\nRFC 5246 TLS August 2008\n\n\n 6.2.2. Record Compression and Decompression ...............20\n 6.2.3. Record Payload Protection ..........................21\n 6.2.3.1. Null or Standard Stream Cipher ............22\n 6.2.3.2. CBC Block Cipher ..........................22\n 6.2.3.3. AEAD Ciphers ..............................24\n 6.3. Key Calculation ...........................................25\n 7. The TLS Handshaking Protocols ..................................26\n 7.1. Change Cipher Spec Protocol ...............................27\n 7.2. Alert Protocol ............................................28\n 7.2.1. Closure Alerts .....................................29\n 7.2.2. Error Alerts .......................................30\n 7.3. Handshake Protocol Overview ...............................33\n 7.4. Handshake Protocol ........................................37\n 7.4.1. Hello Messages .....................................38\n 7.4.1.1. Hello Request .............................38\n 7.4.1.2. Client Hello ..............................39\n 7.4.1.3. Server Hello ..............................42\n 7.4.1.4. Hello Extensions ..........................44\n 7.4.1.4.1. Signature Algorithms ...........45\n 7.4.2. Server Certificate .................................47\n 7.4.3. Server Key Exchange Message ........................50\n 7.4.4. Certificate Request ................................53\n 7.4.5. Server Hello Done ..................................55\n 7.4.6. Client Certificate .................................55\n 7.4.7. Client Key Exchange Message ........................57\n 7.4.7.1. RSA-Encrypted Premaster Secret Message ....58\n 7.4.7.2. Client Diffie-Hellman Public Value ........61\n 7.4.8. Certificate Verify .................................62\n 7.4.9. Finished ...........................................63\n 8. Cryptographic Computations .....................................64\n 8.1. Computing the Master Secret ...............................64\n 8.1.1. RSA ................................................65\n 8.1.2. Diffie-Hellman .....................................65\n 9. Mandatory Cipher Suites ........................................65\n 10. Application Data Protocol .....................................65\n 11. Security Considerations .......................................65\n 12. IANA Considerations ...........................................65\n Appendix A. Protocol Data Structures and Constant Values ..........68\n A.1. Record Layer ..............................................68\n A.2. Change Cipher Specs Message ...............................69\n A.3. Alert Messages ............................................69\n A.4. Handshake Protocol ........................................70\n A.4.1. Hello Messages .....................................71\n A.4.2. Server Authentication and Key Exchange Messages ....72\n A.4.3. Client Authentication and Key Exchange Messages ....74\n A.4.4. Handshake Finalization Message .....................74\n A.5. The Cipher Suite ..........................................75\n A.6. The Security Parameters ...................................77\n\n\n\nDierks & Rescorla Standards Track [Page 2]\n\f\nRFC 5246 TLS August 2008\n\n\n A.7. Changes to RFC 4492 .......................................78\n Appendix B. Glossary ..............................................78\n Appendix C. Cipher Suite Definitions ..............................83\n Appendix D. Implementation Notes ..................................85\n D.1. Random Number Generation and Seeding ......................85\n D.2. Certificates and Authentication ...........................85\n D.3. Cipher Suites .............................................85\n D.4. Implementation Pitfalls ...................................85\n Appendix E. Backward Compatibility ................................87\n E.1. Compatibility with TLS 1.0/1.1 and SSL 3.0 ................87\n E.2. Compatibility with SSL 2.0 ................................88\n E.3. Avoiding Man-in-the-Middle Version Rollback ...............90\n Appendix F. Security Analysis .....................................91\n F.1. Handshake Protocol ........................................91\n F.1.1. Authentication and Key Exchange ....................91\n F.1.1.1. Anonymous Key Exchange ....................91\n F.1.1.2. RSA Key Exchange and Authentication .......92\n F.1.1.3. Diffie-Hellman Key Exchange with\n Authentication ............................92\n F.1.2. Version Rollback Attacks ...........................93\n F.1.3. Detecting Attacks Against the Handshake Protocol ...94\n F.1.4. Resuming Sessions ..................................94\n F.2. Protecting Application Data ...............................94\n F.3. Explicit IVs ..............................................95\n F.4. Security of Composite Cipher Modes ........................95\n F.5. Denial of Service .........................................96\n F.6. Final Notes ...............................................96\n Normative References ..............................................97\n Informative References ............................................98\n Working Group Information ........................................101\n Contributors .....................................................101\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 3]\n\f\nRFC 5246 TLS August 2008\n\n\n1. Introduction\n\n The primary goal of the TLS protocol is to provide privacy and data\n integrity between two communicating applications. The protocol is\n composed of two layers: the TLS Record Protocol and the TLS Handshake\n Protocol. At the lowest level, layered on top of some reliable\n transport protocol (e.g., TCP [TCP]), is the TLS Record Protocol.\n The TLS Record Protocol provides connection security that has two\n basic properties:\n\n - The connection is private. Symmetric cryptography is used for\n data encryption (e.g., AES [AES], RC4 [SCH], etc.). The keys for\n this symmetric encryption are generated uniquely for each\n connection and are based on a secret negotiated by another\n protocol (such as the TLS Handshake Protocol). The Record\n Protocol can also be used without encryption.\n\n - The connection is reliable. Message transport includes a message\n integrity check using a keyed MAC. Secure hash functions (e.g.,\n SHA-1, etc.) are used for MAC computations. The Record Protocol\n can operate without a MAC, but is generally only used in this mode\n while another protocol is using the Record Protocol as a transport\n for negotiating security parameters.\n\n The TLS Record Protocol is used for encapsulation of various higher-\n level protocols. One such encapsulated protocol, the TLS Handshake\n Protocol, allows the server and client to authenticate each other and\n to negotiate an encryption algorithm and cryptographic keys before\n the application protocol transmits or receives its first byte of\n data. The TLS Handshake Protocol provides connection security that\n has three basic properties:\n\n - The peer's identity can be authenticated using asymmetric, or\n public key, cryptography (e.g., RSA [RSA], DSA [DSS], etc.). This\n authentication can be made optional, but is generally required for\n at least one of the peers.\n\n - The negotiation of a shared secret is secure: the negotiated\n secret is unavailable to eavesdroppers, and for any authenticated\n connection the secret cannot be obtained, even by an attacker who\n can place himself in the middle of the connection.\n\n - The negotiation is reliable: no attacker can modify the\n negotiation communication without being detected by the parties to\n the communication.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 4]\n\f\nRFC 5246 TLS August 2008\n\n\n One advantage of TLS is that it is application protocol independent.\n Higher-level protocols can layer on top of the TLS protocol\n transparently. The TLS standard, however, does not specify how\n protocols add security with TLS; the decisions on how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n1.1. Requirements Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [REQ].\n\n1.2. Major Differences from TLS 1.1\n\n This document is a revision of the TLS 1.1 [TLS1.1] protocol which\n contains improved flexibility, particularly for negotiation of\n cryptographic algorithms. The major changes are:\n\n - The MD5/SHA-1 combination in the pseudorandom function (PRF) has\n been replaced with cipher-suite-specified PRFs. All cipher suites\n in this document use P_SHA256.\n\n - The MD5/SHA-1 combination in the digitally-signed element has been\n replaced with a single hash. Signed elements now include a field\n that explicitly specifies the hash algorithm used.\n\n - Substantial cleanup to the client's and server's ability to\n specify which hash and signature algorithms they will accept.\n Note that this also relaxes some of the constraints on signature\n and hash algorithms from previous versions of TLS.\n\n - Addition of support for authenticated encryption with additional\n data modes.\n\n - TLS Extensions definition and AES Cipher Suites were merged in\n from external [TLSEXT] and [TLSAES].\n\n - Tighter checking of EncryptedPreMasterSecret version numbers.\n\n - Tightened up a number of requirements.\n\n - Verify_data length now depends on the cipher suite (default is\n still 12).\n\n - Cleaned up description of Bleichenbacher/Klima attack defenses.\n\n\n\n\nDierks & Rescorla Standards Track [Page 5]\n\f\nRFC 5246 TLS August 2008\n\n\n - Alerts MUST now be sent in many cases.\n\n - After a certificate_request, if no certificates are available,\n clients now MUST send an empty certificate list.\n\n - TLS_RSA_WITH_AES_128_CBC_SHA is now the mandatory to implement\n cipher suite.\n\n - Added HMAC-SHA256 cipher suites.\n\n - Removed IDEA and DES cipher suites. They are now deprecated and\n will be documented in a separate document.\n\n - Support for the SSLv2 backward-compatible hello is now a MAY, not\n a SHOULD, with sending it a SHOULD NOT. Support will probably\n become a SHOULD NOT in the future.\n\n - Added limited \"fall-through\" to the presentation language to allow\n multiple case arms to have the same encoding.\n\n - Added an Implementation Pitfalls sections\n\n - The usual clarifications and editorial work.\n\n2. Goals\n\n The goals of the TLS protocol, in order of priority, are as follows:\n\n 1. Cryptographic security: TLS should be used to establish a secure\n connection between two parties.\n\n 2. Interoperability: Independent programmers should be able to\n develop applications utilizing TLS that can successfully exchange\n cryptographic parameters without knowledge of one another's code.\n\n 3. Extensibility: TLS seeks to provide a framework into which new\n public key and bulk encryption methods can be incorporated as\n necessary. This will also accomplish two sub-goals: preventing\n the need to create a new protocol (and risking the introduction of\n possible new weaknesses) and avoiding the need to implement an\n entire new security library.\n\n 4. Relative efficiency: Cryptographic operations tend to be highly\n CPU intensive, particularly public key operations. For this\n reason, the TLS protocol has incorporated an optional session\n caching scheme to reduce the number of connections that need to be\n established from scratch. Additionally, care has been taken to\n reduce network activity.\n\n\n\nDierks & Rescorla Standards Track [Page 6]\n\f\nRFC 5246 TLS August 2008\n\n\n3. Goals of This Document\n\n This document and the TLS protocol itself are based on the SSL 3.0\n Protocol Specification as published by Netscape. The differences\n between this protocol and SSL 3.0 are not dramatic, but they are\n significant enough that the various versions of TLS and SSL 3.0 do\n not interoperate (although each protocol incorporates a mechanism by\n which an implementation can back down to prior versions). This\n document is intended primarily for readers who will be implementing\n the protocol and for those doing cryptographic analysis of it. The\n specification has been written with this in mind, and it is intended\n to reflect the needs of those two groups. For that reason, many of\n the algorithm-dependent data structures and rules are included in the\n body of the text (as opposed to in an appendix), providing easier\n access to them.\n\n This document is not intended to supply any details of service\n definition or of interface definition, although it does cover select\n areas of policy as they are required for the maintenance of solid\n security.\n\n4. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used. The syntax draws from\n several sources in its structure. Although it resembles the\n programming language \"C\" in its syntax and XDR [XDR] in both its\n syntax and intent, it would be risky to draw too many parallels. The\n purpose of this presentation language is to document TLS only; it has\n no general application beyond that particular goal.\n\n4.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 7]\n\f\nRFC 5246 TLS August 2008\n\n\n4.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" double\n brackets.\n\n Single-byte entities containing uninterpreted data are of type\n opaque.\n\n4.3. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* 3 consecutive 3 byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n In the following example, mandatory is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 4.4). On the\n other hand, longer can represent up to 800 bytes of data, or 400\n uint16 elements, and it may be empty. Its encoding will include a\n\n\n\nDierks & Rescorla Standards Track [Page 8]\n\f\nRFC 5246 TLS August 2008\n\n\n two-byte actual length field prepended to the vector. The length of\n an encoded vector must be an even multiple of the length of a single\n element (for example, a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is 2 bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n4.4. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are formed from fixed-length series of bytes\n concatenated as described in Section 4.1 and are also unsigned. The\n following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are stored in\n network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n Note that in some cases (e.g., DH parameters) it is necessary to\n represent integers as opaque vectors. In such cases, they are\n represented as unsigned integers (i.e., leading zero octets are not\n required even if the most significant bit is set).\n\n4.5. Enumerateds\n\n An additional sparse data type is available called enum. A field of\n type enum can only assume the values declared in the definition.\n Each definition is a different type. Only enumerateds of the same\n type may be assigned or compared. Every element of an enumerated\n must be assigned a value, as demonstrated in the following example.\n Since the elements of the enumerated are not ordered, they can be\n assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n\n\n\nDierks & Rescorla Standards Track [Page 9]\n\f\nRFC 5246 TLS August 2008\n\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n For enumerateds that are never converted to external representation,\n the numerical information may be omitted.\n\n enum { low, medium, high } Amount;\n\n4.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax for definition is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } [[T]];\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n Structure definitions may be embedded.\n\n4.6.1. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. There\n must be a case arm for every element of the enumeration declared in\n the select. Case arms have limited fall-through: if two case arms\n follow in immediate succession with no fields in between, then they\n\n\n\nDierks & Rescorla Standards Track [Page 10]\n\f\nRFC 5246 TLS August 2008\n\n\n both contain the same fields. Thus, in the example below, \"orange\"\n and \"banana\" both contain V2. Note that this is a new piece of\n syntax in TLS 1.2.\n\n The body of the variant structure may be given a label for reference.\n The mechanism by which the variant is selected at runtime is not\n prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1;\n case e2: Te2;\n case e3: case e4: Te3;\n ....\n case en: Ten;\n } [[fv]];\n } [[Tv]];\n\n For example:\n\n enum { apple, orange, banana } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n select (VariantTag) { /* value of selector is implicit */\n case apple:\n V1; /* VariantBody, tag = apple */\n case orange:\n case banana:\n V2; /* VariantBody, tag = orange or banana */\n } variant_body; /* optional label on variant */\n } VariantRecord;\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 11]\n\f\nRFC 5246 TLS August 2008\n\n\n4.7. Cryptographic Attributes\n\n The five cryptographic operations -- digital signing, stream cipher\n encryption, block cipher encryption, authenticated encryption with\n additional data (AEAD) encryption, and public key encryption -- are\n designated digitally-signed, stream-ciphered, block-ciphered, aead-\n ciphered, and public-key-encrypted, respectively. A field's\n cryptographic processing is specified by prepending an appropriate\n key word designation before the field's type specification.\n Cryptographic keys are implied by the current session state (see\n Section 6.1).\n\n A digitally-signed element is encoded as a struct DigitallySigned:\n\n struct {\n SignatureAndHashAlgorithm algorithm;\n opaque signature<0..2^16-1>;\n } DigitallySigned;\n\n The algorithm field specifies the algorithm used (see Section\n 7.4.1.4.1 for the definition of this field). Note that the\n introduction of the algorithm field is a change from previous\n versions. The signature is a digital signature using those\n algorithms over the contents of the element. The contents themselves\n do not appear on the wire but are simply calculated. The length of\n the signature is specified by the signing algorithm and key.\n\n In RSA signing, the opaque vector contains the signature generated\n using the RSASSA-PKCS1-v1_5 signature scheme defined in [PKCS1]. As\n discussed in [PKCS1], the DigestInfo MUST be DER-encoded [X680]\n [X690]. For hash algorithms without parameters (which includes\n SHA-1), the DigestInfo.AlgorithmIdentifier.parameters field MUST be\n NULL, but implementations MUST accept both without parameters and\n with NULL parameters. Note that earlier versions of TLS used a\n different RSA signature scheme that did not include a DigestInfo\n encoding.\n\n In DSA, the 20 bytes of the SHA-1 hash are run directly through the\n Digital Signing Algorithm with no additional hashing. This produces\n two values, r and s. The DSA signature is an opaque vector, as\n above, the contents of which are the DER encoding of:\n\n Dss-Sig-Value ::= SEQUENCE {\n r INTEGER,\n s INTEGER\n }\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 12]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: In current terminology, DSA refers to the Digital Signature\n Algorithm and DSS refers to the NIST standard. In the original SSL\n and TLS specs, \"DSS\" was used universally. This document uses \"DSA\"\n to refer to the algorithm, \"DSS\" to refer to the standard, and it\n uses \"DSS\" in the code point definitions for historical continuity.\n\n In stream cipher encryption, the plaintext is exclusive-ORed with an\n identical amount of output generated from a cryptographically secure\n keyed pseudorandom number generator.\n\n In block cipher encryption, every block of plaintext encrypts to a\n block of ciphertext. All block cipher encryption is done in CBC\n (Cipher Block Chaining) mode, and all items that are block-ciphered\n will be an exact multiple of the cipher block length.\n\n In AEAD encryption, the plaintext is simultaneously encrypted and\n integrity protected. The input may be of any length, and aead-\n ciphered output is generally larger than the input in order to\n accommodate the integrity check value.\n\n In public key encryption, a public key algorithm is used to encrypt\n data in such a way that it can be decrypted only with the matching\n private key. A public-key-encrypted element is encoded as an opaque\n vector <0..2^16-1>, where the length is specified by the encryption\n algorithm and key.\n\n RSA encryption is done using the RSAES-PKCS1-v1_5 encryption scheme\n defined in [PKCS1].\n\n In the following example\n\n stream-ciphered struct {\n uint8 field1;\n uint8 field2;\n digitally-signed opaque {\n uint8 field3<0..255>;\n uint8 field4;\n };\n } UserType;\n\n The contents of the inner struct (field3 and field4) are used as\n input for the signature/hash algorithm, and then the entire structure\n is encrypted with a stream cipher. The length of this structure, in\n bytes, would be equal to two bytes for field1 and field2, plus two\n bytes for the signature and hash algorithm, plus two bytes for the\n length of the signature, plus the length of the output of the signing\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 13]\n\f\nRFC 5246 TLS August 2008\n\n\n algorithm. The length of the signature is known because the\n algorithm and key used for the signing are known prior to encoding or\n decoding this structure.\n\n4.8. Constants\n\n Typed constants can be defined for purposes of specification by\n declaring a symbol of the desired type and assigning values to it.\n\n Under-specified types (opaque, variable-length vectors, and\n structures that contain opaque) cannot be assigned values. No fields\n of a multi-element structure or vector may be elided.\n\n For example:\n\n struct {\n uint8 f1;\n uint8 f2;\n } Example1;\n\n Example1 ex1 = {1, 4}; /* assigns f1 = 1, f2 = 4 */\n\n5. HMAC and the Pseudorandom Function\n\n The TLS record layer uses a keyed Message Authentication Code (MAC)\n to protect message integrity. The cipher suites defined in this\n document use a construction known as HMAC, described in [HMAC], which\n is based on a hash function. Other cipher suites MAY define their\n own MAC constructions, if needed.\n\n In addition, a construction is required to do expansion of secrets\n into blocks of data for the purposes of key generation or validation.\n This pseudorandom function (PRF) takes as input a secret, a seed, and\n an identifying label and produces an output of arbitrary length.\n\n In this section, we define one PRF, based on HMAC. This PRF with the\n SHA-256 hash function is used for all cipher suites defined in this\n document and in TLS documents published prior to this document when\n TLS 1.2 is negotiated. New cipher suites MUST explicitly specify a\n PRF and, in general, SHOULD use the TLS PRF with SHA-256 or a\n stronger standard hash function.\n\n First, we define a data expansion function, P_hash(secret, data),\n that uses a single hash function to expand a secret and seed into an\n arbitrary quantity of output:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 14]\n\f\nRFC 5246 TLS August 2008\n\n\n P_hash(secret, seed) = HMAC_hash(secret, A(1) + seed) +\n HMAC_hash(secret, A(2) + seed) +\n HMAC_hash(secret, A(3) + seed) + ...\n\n where + indicates concatenation.\n\n A() is defined as:\n\n A(0) = seed\n A(i) = HMAC_hash(secret, A(i-1))\n\n P_hash can be iterated as many times as necessary to produce the\n required quantity of data. For example, if P_SHA256 is being used to\n create 80 bytes of data, it will have to be iterated three times\n (through A(3)), creating 96 bytes of output data; the last 16 bytes\n of the final iteration will then be discarded, leaving 80 bytes of\n output data.\n\n TLS's PRF is created by applying P_hash to the secret as:\n\n PRF(secret, label, seed) = P_(secret, label + seed)\n\n The label is an ASCII string. It should be included in the exact\n form it is given without a length byte or trailing null character.\n For example, the label \"slithy toves\" would be processed by hashing\n the following bytes:\n\n 73 6C 69 74 68 79 20 74 6F 76 65 73\n\n6. The TLS Record Protocol\n\n The TLS Record Protocol is a layered protocol. At each layer,\n messages may include fields for length, description, and content.\n The Record Protocol takes messages to be transmitted, fragments the\n data into manageable blocks, optionally compresses the data, applies\n a MAC, encrypts, and transmits the result. Received data is\n decrypted, verified, decompressed, reassembled, and then delivered to\n higher-level clients.\n\n Four protocols that use the record protocol are described in this\n document: the handshake protocol, the alert protocol, the change\n cipher spec protocol, and the application data protocol. In order to\n allow extension of the TLS protocol, additional record content types\n can be supported by the record protocol. New record content type\n values are assigned by IANA in the TLS Content Type Registry as\n described in Section 12.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 15]\n\f\nRFC 5246 TLS August 2008\n\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST send an\n unexpected_message alert.\n\n Any protocol designed for use over TLS must be carefully designed to\n deal with all possible attacks against it. As a practical matter,\n this means that the protocol designer must be aware of what security\n properties TLS does and does not provide and cannot safely rely on\n the latter.\n\n Note in particular that type and length of a record are not protected\n by encryption. If this information is itself sensitive, application\n designers may wish to take steps (padding, cover traffic) to minimize\n information leakage.\n\n6.1. Connection States\n\n A TLS connection state is the operating environment of the TLS Record\n Protocol. It specifies a compression algorithm, an encryption\n algorithm, and a MAC algorithm. In addition, the parameters for\n these algorithms are known: the MAC key and the bulk encryption keys\n for the connection in both the read and the write directions.\n Logically, there are always four connection states outstanding: the\n current read and write states, and the pending read and write states.\n All records are processed under the current read and write states.\n The security parameters for the pending states can be set by the TLS\n Handshake Protocol, and the ChangeCipherSpec can selectively make\n either of the pending states current, in which case the appropriate\n current state is disposed of and replaced with the pending state; the\n pending state is then reinitialized to an empty state. It is illegal\n to make a state that has not been initialized with security\n parameters a current state. The initial current state always\n specifies that no encryption, compression, or MAC will be used.\n\n The security parameters for a TLS Connection read and write state are\n set by providing the following values:\n\n connection end\n Whether this entity is considered the \"client\" or the \"server\" in\n this connection.\n\n PRF algorithm\n An algorithm used to generate keys from the master secret (see\n Sections 5 and 6.3).\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 16]\n\f\nRFC 5246 TLS August 2008\n\n\n bulk encryption algorithm\n An algorithm to be used for bulk encryption. This specification\n includes the key size of this algorithm, whether it is a block,\n stream, or AEAD cipher, the block size of the cipher (if\n appropriate), and the lengths of explicit and implicit\n initialization vectors (or nonces).\n\n MAC algorithm\n An algorithm to be used for message authentication. This\n specification includes the size of the value returned by the MAC\n algorithm.\n\n compression algorithm\n An algorithm to be used for data compression. This specification\n must include all information the algorithm requires to do\n compression.\n\n master secret\n A 48-byte secret shared between the two peers in the connection.\n\n client random\n A 32-byte value provided by the client.\n\n server random\n A 32-byte value provided by the server.\n\n These parameters are defined in the presentation language as:\n\n enum { server, client } ConnectionEnd;\n\n enum { tls_prf_sha256 } PRFAlgorithm;\n\n enum { null, rc4, 3des, aes }\n BulkCipherAlgorithm;\n\n enum { stream, block, aead } CipherType;\n\n enum { null, hmac_md5, hmac_sha1, hmac_sha256,\n hmac_sha384, hmac_sha512} MACAlgorithm;\n\n enum { null(0), (255) } CompressionMethod;\n\n /* The algorithms specified in CompressionMethod, PRFAlgorithm,\n BulkCipherAlgorithm, and MACAlgorithm may be added to. */\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 17]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ConnectionEnd entity;\n PRFAlgorithm prf_algorithm;\n BulkCipherAlgorithm bulk_cipher_algorithm;\n CipherType cipher_type;\n uint8 enc_key_length;\n uint8 block_length;\n uint8 fixed_iv_length;\n uint8 record_iv_length;\n MACAlgorithm mac_algorithm;\n uint8 mac_length;\n uint8 mac_key_length;\n CompressionMethod compression_algorithm;\n opaque master_secret[48];\n opaque client_random[32];\n opaque server_random[32];\n } SecurityParameters;\n\n The record layer will use the security parameters to generate the\n following six items (some of which are not required by all ciphers,\n and are thus empty):\n\n client write MAC key\n server write MAC key\n client write encryption key\n server write encryption key\n client write IV\n server write IV\n\n The client write parameters are used by the server when receiving and\n processing records and vice versa. The algorithm used for generating\n these items from the security parameters is described in Section 6.3.\n\n Once the security parameters have been set and the keys have been\n generated, the connection states can be instantiated by making them\n the current states. These current states MUST be updated for each\n record processed. Each connection state includes the following\n elements:\n\n compression state\n The current state of the compression algorithm.\n\n cipher state\n The current state of the encryption algorithm. This will consist\n of the scheduled key for that connection. For stream ciphers,\n this will also contain whatever state information is necessary to\n allow the stream to continue to encrypt or decrypt data.\n\n\n\n\nDierks & Rescorla Standards Track [Page 18]\n\f\nRFC 5246 TLS August 2008\n\n\n MAC key\n The MAC key for this connection, as generated above.\n\n sequence number\n Each connection state contains a sequence number, which is\n maintained separately for read and write states. The sequence\n number MUST be set to zero whenever a connection state is made the\n active state. Sequence numbers are of type uint64 and may not\n exceed 2^64-1. Sequence numbers do not wrap. If a TLS\n implementation would need to wrap a sequence number, it must\n renegotiate instead. A sequence number is incremented after each\n record: specifically, the first record transmitted under a\n particular connection state MUST use sequence number 0.\n\n6.2. Record Layer\n\n The TLS record layer receives uninterpreted data from higher layers\n in non-empty blocks of arbitrary size.\n\n6.2.1. Fragmentation\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Client\n message boundaries are not preserved in the record layer (i.e.,\n multiple client messages of the same ContentType MAY be coalesced\n into a single TLSPlaintext record, or a single message MAY be\n fragmented across several records).\n\n struct {\n uint8 major;\n uint8 minor;\n } ProtocolVersion;\n\n enum {\n change_cipher_spec(20), alert(21), handshake(22),\n application_data(23), (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type\n The higher-level protocol used to process the enclosed fragment.\n\n\n\n\nDierks & Rescorla Standards Track [Page 19]\n\f\nRFC 5246 TLS August 2008\n\n\n version\n The version of the protocol being employed. This document\n describes TLS Version 1.2, which uses the version { 3, 3 }. The\n version value 3.3 is historical, deriving from the use of {3, 1}\n for TLS 1.0. (See Appendix A.1.) Note that a client that\n supports multiple versions of TLS may not know what version will\n be employed before it receives the ServerHello. See Appendix E\n for discussion about what record layer version number should be\n employed for ClientHello.\n\n length\n The length (in bytes) of the following TLSPlaintext.fragment. The\n length MUST NOT exceed 2^14.\n\n fragment\n The application data. This data is transparent and treated as an\n independent block to be dealt with by the higher-level protocol\n specified by the type field.\n\n Implementations MUST NOT send zero-length fragments of Handshake,\n Alert, or ChangeCipherSpec content types. Zero-length fragments of\n Application data MAY be sent as they are potentially useful as a\n traffic analysis countermeasure.\n\n Note: Data of different TLS record layer content types MAY be\n interleaved. Application data is generally of lower precedence for\n transmission than other content types. However, records MUST be\n delivered to the network in the same order as they are protected by\n the record layer. Recipients MUST receive and process interleaved\n application layer traffic during handshakes subsequent to the first\n one on a connection.\n\n6.2.2. Record Compression and Decompression\n\n All records are compressed using the compression algorithm defined in\n the current session state. There is always an active compression\n algorithm; however, initially it is defined as\n CompressionMethod.null. The compression algorithm translates a\n TLSPlaintext structure into a TLSCompressed structure. Compression\n functions are initialized with default state information whenever a\n connection state is made active. [RFC3749] describes compression\n algorithms for TLS.\n\n Compression must be lossless and may not increase the content length\n by more than 1024 bytes. If the decompression function encounters a\n TLSCompressed.fragment that would decompress to a length in excess of\n 2^14 bytes, it MUST report a fatal decompression failure error.\n\n\n\n\nDierks & Rescorla Standards Track [Page 20]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ContentType type; /* same as TLSPlaintext.type */\n ProtocolVersion version;/* same as TLSPlaintext.version */\n uint16 length;\n opaque fragment[TLSCompressed.length];\n } TLSCompressed;\n\n length\n The length (in bytes) of the following TLSCompressed.fragment.\n The length MUST NOT exceed 2^14 + 1024.\n\n fragment\n The compressed form of TLSPlaintext.fragment.\n\n Note: A CompressionMethod.null operation is an identity operation;\n no fields are altered.\n\n Implementation note: Decompression functions are responsible for\n ensuring that messages cannot cause internal buffer overflows.\n\n6.2.3. Record Payload Protection\n\n The encryption and MAC functions translate a TLSCompressed\n structure into a TLSCiphertext. The decryption functions reverse\n the process. The MAC of the record also includes a sequence\n number so that missing, extra, or repeated messages are\n detectable.\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n select (SecurityParameters.cipher_type) {\n case stream: GenericStreamCipher;\n case block: GenericBlockCipher;\n case aead: GenericAEADCipher;\n } fragment;\n } TLSCiphertext;\n\n type\n The type field is identical to TLSCompressed.type.\n\n version\n The version field is identical to TLSCompressed.version.\n\n length\n The length (in bytes) of the following TLSCiphertext.fragment.\n The length MUST NOT exceed 2^14 + 2048.\n\n\n\nDierks & Rescorla Standards Track [Page 21]\n\f\nRFC 5246 TLS August 2008\n\n\n fragment\n The encrypted form of TLSCompressed.fragment, with the MAC.\n\n6.2.3.1. Null or Standard Stream Cipher\n\n Stream ciphers (including BulkCipherAlgorithm.null; see Appendix A.6)\n convert TLSCompressed.fragment structures to and from stream\n TLSCiphertext.fragment structures.\n\n stream-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n } GenericStreamCipher;\n\n The MAC is generated as:\n\n MAC(MAC_write_key, seq_num +\n TLSCompressed.type +\n TLSCompressed.version +\n TLSCompressed.length +\n TLSCompressed.fragment);\n\n where \"+\" denotes concatenation.\n\n seq_num\n The sequence number for this record.\n\n MAC\n The MAC algorithm specified by SecurityParameters.mac_algorithm.\n\n Note that the MAC is computed before encryption. The stream cipher\n encrypts the entire block, including the MAC. For stream ciphers\n that do not use a synchronization vector (such as RC4), the stream\n cipher state from the end of one record is simply used on the\n subsequent packet. If the cipher suite is TLS_NULL_WITH_NULL_NULL,\n encryption consists of the identity operation (i.e., the data is not\n encrypted, and the MAC size is zero, implying that no MAC is used).\n For both null and stream ciphers, TLSCiphertext.length is\n TLSCompressed.length plus SecurityParameters.mac_length.\n\n6.2.3.2. CBC Block Cipher\n\n For block ciphers (such as 3DES or AES), the encryption and MAC\n functions convert TLSCompressed.fragment structures to and from block\n TLSCiphertext.fragment structures.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 22]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n opaque IV[SecurityParameters.record_iv_length];\n block-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n uint8 padding[GenericBlockCipher.padding_length];\n uint8 padding_length;\n };\n } GenericBlockCipher;\n\n The MAC is generated as described in Section 6.2.3.1.\n\n IV\n The Initialization Vector (IV) SHOULD be chosen at random, and\n MUST be unpredictable. Note that in versions of TLS prior to 1.1,\n there was no IV field, and the last ciphertext block of the\n previous record (the \"CBC residue\") was used as the IV. This was\n changed to prevent the attacks described in [CBCATT]. For block\n ciphers, the IV length is of length\n SecurityParameters.record_iv_length, which is equal to the\n SecurityParameters.block_size.\n\n padding\n Padding that is added to force the length of the plaintext to be\n an integral multiple of the block cipher's block length. The\n padding MAY be any length up to 255 bytes, as long as it results\n in the TLSCiphertext.length being an integral multiple of the\n block length. Lengths longer than necessary might be desirable to\n frustrate attacks on a protocol that are based on analysis of the\n lengths of exchanged messages. Each uint8 in the padding data\n vector MUST be filled with the padding length value. The receiver\n MUST check this padding and MUST use the bad_record_mac alert to\n indicate padding errors.\n\n padding_length\n The padding length MUST be such that the total size of the\n GenericBlockCipher structure is a multiple of the cipher's block\n length. Legal values range from zero to 255, inclusive. This\n length specifies the length of the padding field exclusive of the\n padding_length field itself.\n\n The encrypted data length (TLSCiphertext.length) is one more than the\n sum of SecurityParameters.block_length, TLSCompressed.length,\n SecurityParameters.mac_length, and padding_length.\n\n Example: If the block length is 8 bytes, the content length\n (TLSCompressed.length) is 61 bytes, and the MAC length is 20 bytes,\n then the length before padding is 82 bytes (this does not include the\n\n\n\nDierks & Rescorla Standards Track [Page 23]\n\f\nRFC 5246 TLS August 2008\n\n\n IV. Thus, the padding length modulo 8 must be equal to 6 in order to\n make the total length an even multiple of 8 bytes (the block length).\n The padding length can be 6, 14, 22, and so on, through 254. If the\n padding length were the minimum necessary, 6, the padding would be 6\n bytes, each containing the value 6. Thus, the last 8 octets of the\n GenericBlockCipher before block encryption would be xx 06 06 06 06 06\n 06 06, where xx is the last octet of the MAC.\n\n Note: With block ciphers in CBC mode (Cipher Block Chaining), it is\n critical that the entire plaintext of the record be known before any\n ciphertext is transmitted. Otherwise, it is possible for the\n attacker to mount the attack described in [CBCATT].\n\n Implementation note: Canvel et al. [CBCTIME] have demonstrated a\n timing attack on CBC padding based on the time required to compute\n the MAC. In order to defend against this attack, implementations\n MUST ensure that record processing time is essentially the same\n whether or not the padding is correct. In general, the best way to\n do this is to compute the MAC even if the padding is incorrect, and\n only then reject the packet. For instance, if the pad appears to be\n incorrect, the implementation might assume a zero-length pad and then\n compute the MAC. This leaves a small timing channel, since MAC\n performance depends to some extent on the size of the data fragment,\n but it is not believed to be large enough to be exploitable, due to\n the large block size of existing MACs and the small size of the\n timing signal.\n\n6.2.3.3. AEAD Ciphers\n\n For AEAD [AEAD] ciphers (such as [CCM] or [GCM]), the AEAD function\n converts TLSCompressed.fragment structures to and from AEAD\n TLSCiphertext.fragment structures.\n\n struct {\n opaque nonce_explicit[SecurityParameters.record_iv_length];\n aead-ciphered struct {\n opaque content[TLSCompressed.length];\n };\n } GenericAEADCipher;\n\n AEAD ciphers take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [AEAD]. The key is either the\n client_write_key or the server_write_key. No MAC key is used.\n\n Each AEAD cipher suite MUST specify how the nonce supplied to the\n AEAD operation is constructed, and what is the length of the\n GenericAEADCipher.nonce_explicit part. In many cases, it is\n\n\n\nDierks & Rescorla Standards Track [Page 24]\n\f\nRFC 5246 TLS August 2008\n\n\n appropriate to use the partially implicit nonce technique described\n in Section 3.2.1 of [AEAD]; with record_iv_length being the length of\n the explicit part. In this case, the implicit part SHOULD be derived\n from key_block as client_write_iv and server_write_iv (as described\n in Section 6.3), and the explicit part is included in\n GenericAEAEDCipher.nonce_explicit.\n\n The plaintext is the TLSCompressed.fragment.\n\n The additional authenticated data, which we denote as\n additional_data, is defined as follows:\n\n additional_data = seq_num + TLSCompressed.type +\n TLSCompressed.version + TLSCompressed.length;\n\n where \"+\" denotes concatenation.\n\n The aead_output consists of the ciphertext output by the AEAD\n encryption operation. The length will generally be larger than\n TLSCompressed.length, but by an amount that varies with the AEAD\n cipher. Since the ciphers might incorporate padding, the amount of\n overhead could vary with different TLSCompressed.length values. Each\n AEAD cipher MUST NOT produce an expansion of greater than 1024 bytes.\n Symbolically,\n\n AEADEncrypted = AEAD-Encrypt(write_key, nonce, plaintext,\n additional_data)\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, the \"additional_data\", and the AEADEncrypted value. The\n output is either the plaintext or an error indicating that the\n decryption failed. There is no separate integrity check. That is:\n\n TLSCompressed.fragment = AEAD-Decrypt(write_key, nonce,\n AEADEncrypted,\n additional_data)\n\n If the decryption fails, a fatal bad_record_mac alert MUST be\n generated.\n\n6.3. Key Calculation\n\n The Record Protocol requires an algorithm to generate keys required\n by the current connection state (see Appendix A.6) from the security\n parameters provided by the handshake protocol.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 25]\n\f\nRFC 5246 TLS August 2008\n\n\n The master secret is expanded into a sequence of secure bytes, which\n is then split to a client write MAC key, a server write MAC key, a\n client write encryption key, and a server write encryption key. Each\n of these is generated from the byte sequence in that order. Unused\n values are empty. Some AEAD ciphers may additionally require a\n client write IV and a server write IV (see Section 6.2.3.3).\n\n When keys and MAC keys are generated, the master secret is used as an\n entropy source.\n\n To generate the key material, compute\n\n key_block = PRF(SecurityParameters.master_secret,\n \"key expansion\",\n SecurityParameters.server_random +\n SecurityParameters.client_random);\n\n until enough output has been generated. Then, the key_block is\n partitioned as follows:\n\n client_write_MAC_key[SecurityParameters.mac_key_length]\n server_write_MAC_key[SecurityParameters.mac_key_length]\n client_write_key[SecurityParameters.enc_key_length]\n server_write_key[SecurityParameters.enc_key_length]\n client_write_IV[SecurityParameters.fixed_iv_length]\n server_write_IV[SecurityParameters.fixed_iv_length]\n\n Currently, the client_write_IV and server_write_IV are only generated\n for implicit nonce techniques as described in Section 3.2.1 of\n [AEAD].\n\n Implementation note: The currently defined cipher suite which\n requires the most material is AES_256_CBC_SHA256. It requires 2 x 32\n byte keys and 2 x 32 byte MAC keys, for a total 128 bytes of key\n material.\n\n7. The TLS Handshaking Protocols\n\n TLS has three subprotocols that are used to allow peers to agree upon\n security parameters for the record layer, to authenticate themselves,\n to instantiate negotiated security parameters, and to report error\n conditions to each other.\n\n The Handshake Protocol is responsible for negotiating a session,\n which consists of the following items:\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 26]\n\f\nRFC 5246 TLS August 2008\n\n\n session identifier\n An arbitrary byte sequence chosen by the server to identify an\n active or resumable session state.\n\n peer certificate\n X509v3 [PKIX] certificate of the peer. This element of the state\n may be null.\n\n compression method\n The algorithm used to compress data prior to encryption.\n\n cipher spec\n Specifies the pseudorandom function (PRF) used to generate keying\n material, the bulk data encryption algorithm (such as null, AES,\n etc.) and the MAC algorithm (such as HMAC-SHA1). It also defines\n cryptographic attributes such as the mac_length. (See Appendix\n A.6 for formal definition.)\n\n master secret\n 48-byte secret shared between the client and server.\n\n is resumable\n A flag indicating whether the session can be used to initiate new\n connections.\n\n These items are then used to create security parameters for use by\n the record layer when protecting application data. Many connections\n can be instantiated using the same session through the resumption\n feature of the TLS Handshake Protocol.\n\n7.1. Change Cipher Spec Protocol\n\n The change cipher spec protocol exists to signal transitions in\n ciphering strategies. The protocol consists of a single message,\n which is encrypted and compressed under the current (not the pending)\n connection state. The message consists of a single byte of value 1.\n\n struct {\n enum { change_cipher_spec(1), (255) } type;\n } ChangeCipherSpec;\n\n The ChangeCipherSpec message is sent by both the client and the\n server to notify the receiving party that subsequent records will be\n protected under the newly negotiated CipherSpec and keys. Reception\n of this message causes the receiver to instruct the record layer to\n immediately copy the read pending state into the read current state.\n Immediately after sending this message, the sender MUST instruct the\n record layer to make the write pending state the write active state.\n\n\n\nDierks & Rescorla Standards Track [Page 27]\n\f\nRFC 5246 TLS August 2008\n\n\n (See Section 6.1.) The ChangeCipherSpec message is sent during the\n handshake after the security parameters have been agreed upon, but\n before the verifying Finished message is sent.\n\n Note: If a rehandshake occurs while data is flowing on a connection,\n the communicating parties may continue to send data using the old\n CipherSpec. However, once the ChangeCipherSpec has been sent, the\n new CipherSpec MUST be used. The first side to send the\n ChangeCipherSpec does not know that the other side has finished\n computing the new keying material (e.g., if it has to perform a\n time-consuming public key operation). Thus, a small window of time,\n during which the recipient must buffer the data, MAY exist. In\n practice, with modern machines this interval is likely to be fairly\n short.\n\n7.2. Alert Protocol\n\n One of the content types supported by the TLS record layer is the\n alert type. Alert messages convey the severity of the message\n (warning or fatal) and a description of the alert. Alert messages\n with a level of fatal result in the immediate termination of the\n connection. In this case, other connections corresponding to the\n session may continue, but the session identifier MUST be invalidated,\n preventing the failed session from being used to establish new\n connections. Like other messages, alert messages are encrypted and\n compressed, as specified by the current connection state.\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n\n\n\nDierks & Rescorla Standards Track [Page 28]\n\f\nRFC 5246 TLS August 2008\n\n\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n7.2.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack. Either party may\n initiate the exchange of closing messages.\n\n close_notify\n This message notifies the recipient that the sender will not send\n any more messages on this connection. Note that as of TLS 1.1,\n failure to properly close a connection no longer requires that a\n session not be resumed. This is a change from TLS 1.0 to conform\n with widespread implementation practice.\n\n Either party may initiate a close by sending a close_notify alert.\n Any data received after a closure alert is ignored.\n\n Unless some other fatal alert has been transmitted, each party is\n required to send a close_notify alert before closing the write side\n of the connection. The other party MUST respond with a close_notify\n alert of its own and close down the connection immediately,\n discarding any pending writes. It is not required for the initiator\n of the close to wait for the responding close_notify alert before\n closing the read side of the connection.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation must receive the responding\n close_notify alert before indicating to the application layer that\n the TLS connection has ended. If the application protocol will not\n transfer any additional data, but will only close the underlying\n transport connection, then the implementation MAY choose to close the\n transport without waiting for the responding close_notify. No part\n\n\n\n\nDierks & Rescorla Standards Track [Page 29]\n\f\nRFC 5246 TLS August 2008\n\n\n of this standard should be taken to dictate the manner in which a\n usage profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing a connection reliably delivers\n pending data before destroying the transport.\n\n7.2.2. Error Alerts\n\n Error handling in the TLS Handshake protocol is very simple. When an\n error is detected, the detecting party sends a message to the other\n party. Upon transmission or receipt of a fatal alert message, both\n parties immediately close the connection. Servers and clients MUST\n forget any session-identifiers, keys, and secrets associated with a\n failed connection. Thus, any connection terminated with a fatal\n alert MUST NOT be resumed.\n\n Whenever an implementation encounters a condition which is defined as\n a fatal alert, it MUST send the appropriate alert prior to closing\n the connection. For all errors where an alert level is not\n explicitly specified, the sending party MAY determine at its\n discretion whether to treat this as a fatal error or not. If the\n implementation chooses to send an alert but intends to close the\n connection immediately afterwards, it MUST send that alert at the\n fatal alert level.\n\n If an alert with a level of warning is sent and received, generally\n the connection can continue normally. If the receiving party decides\n not to proceed with the connection (e.g., after having received a\n no_renegotiation alert that it is not willing to accept), it SHOULD\n send a fatal alert to terminate the connection. Given this, the\n sending party cannot, in general, know how the receiving party will\n behave. Therefore, warning alerts are not very useful when the\n sending party wants to continue the connection, and thus are\n sometimes omitted. For example, if a peer decides to accept an\n expired certificate (perhaps after confirming this with the user) and\n wants to continue the connection, it would not generally send a\n certificate_expired alert.\n\n The following error alerts are defined:\n\n unexpected_message\n An inappropriate message was received. This alert is always fatal\n and should never be observed in communication between proper\n implementations.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 30]\n\f\nRFC 5246 TLS August 2008\n\n\n bad_record_mac\n This alert is returned if a record is received with an incorrect\n MAC. This alert also MUST be returned if an alert is sent because\n a TLSCiphertext decrypted in an invalid way: either it wasn't an\n even multiple of the block length, or its padding values, when\n checked, weren't correct. This message is always fatal and should\n never be observed in communication between proper implementations\n (except when messages were corrupted in the network).\n\n decryption_failed_RESERVED\n This alert was used in some earlier versions of TLS, and may have\n permitted certain attacks against the CBC mode [CBCATT]. It MUST\n NOT be sent by compliant implementations.\n\n record_overflow\n A TLSCiphertext record was received that had a length more than\n 2^14+2048 bytes, or a record decrypted to a TLSCompressed record\n with more than 2^14+1024 bytes. This message is always fatal and\n should never be observed in communication between proper\n implementations (except when messages were corrupted in the\n network).\n\n decompression_failure\n The decompression function received improper input (e.g., data\n that would expand to excessive length). This message is always\n fatal and should never be observed in communication between proper\n implementations.\n\n handshake_failure\n Reception of a handshake_failure alert message indicates that the\n sender was unable to negotiate an acceptable set of security\n parameters given the options available. This is a fatal error.\n\n no_certificate_RESERVED\n This alert was used in SSLv3 but not any version of TLS. It MUST\n NOT be sent by compliant implementations.\n\n bad_certificate\n A certificate was corrupt, contained signatures that did not\n verify correctly, etc.\n\n unsupported_certificate\n A certificate was of an unsupported type.\n\n certificate_revoked\n A certificate was revoked by its signer.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 31]\n\f\nRFC 5246 TLS August 2008\n\n\n certificate_expired\n A certificate has expired or is not currently valid.\n\n certificate_unknown\n Some other (unspecified) issue arose in processing the\n certificate, rendering it unacceptable.\n\n illegal_parameter\n A field in the handshake was out of range or inconsistent with\n other fields. This message is always fatal.\n\n unknown_ca\n A valid certificate chain or partial chain was received, but the\n certificate was not accepted because the CA certificate could not\n be located or couldn't be matched with a known, trusted CA. This\n message is always fatal.\n\n access_denied\n A valid certificate was received, but when access control was\n applied, the sender decided not to proceed with negotiation. This\n message is always fatal.\n\n decode_error\n A message could not be decoded because some field was out of the\n specified range or the length of the message was incorrect. This\n message is always fatal and should never be observed in\n communication between proper implementations (except when messages\n were corrupted in the network).\n\n decrypt_error\n A handshake cryptographic operation failed, including being unable\n to correctly verify a signature or validate a Finished message.\n This message is always fatal.\n\n export_restriction_RESERVED\n This alert was used in some earlier versions of TLS. It MUST NOT\n be sent by compliant implementations.\n\n protocol_version\n The protocol version the client has attempted to negotiate is\n recognized but not supported. (For example, old protocol versions\n might be avoided for security reasons.) This message is always\n fatal.\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 32]\n\f\nRFC 5246 TLS August 2008\n\n\n insufficient_security\n Returned instead of handshake_failure when a negotiation has\n failed specifically because the server requires ciphers more\n secure than those supported by the client. This message is always\n fatal.\n\n internal_error\n An internal error unrelated to the peer or the correctness of the\n protocol (such as a memory allocation failure) makes it impossible\n to continue. This message is always fatal.\n\n user_canceled\n This handshake is being canceled for some reason unrelated to a\n protocol failure. If the user cancels an operation after the\n handshake is complete, just closing the connection by sending a\n close_notify is more appropriate. This alert should be followed\n by a close_notify. This message is generally a warning.\n\n no_renegotiation\n Sent by the client in response to a hello request or by the server\n in response to a client hello after initial handshaking. Either\n of these would normally lead to renegotiation; when that is not\n appropriate, the recipient should respond with this alert. At\n that point, the original requester can decide whether to proceed\n with the connection. One case where this would be appropriate is\n where a server has spawned a process to satisfy a request; the\n process might receive security parameters (key length,\n authentication, etc.) at startup, and it might be difficult to\n communicate changes to these parameters after that point. This\n message is always a warning.\n\n unsupported_extension\n sent by clients that receive an extended server hello containing\n an extension that they did not put in the corresponding client\n hello. This message is always fatal.\n\n New Alert values are assigned by IANA as described in Section 12.\n\n7.3. Handshake Protocol Overview\n\n The cryptographic parameters of the session state are produced by the\n TLS Handshake Protocol, which operates on top of the TLS record\n layer. When a TLS client and server first start communicating, they\n agree on a protocol version, select cryptographic algorithms,\n optionally authenticate each other, and use public-key encryption\n techniques to generate shared secrets.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 33]\n\f\nRFC 5246 TLS August 2008\n\n\n The TLS Handshake Protocol involves the following steps:\n\n - Exchange hello messages to agree on algorithms, exchange random\n values, and check for session resumption.\n\n - Exchange the necessary cryptographic parameters to allow the\n client and server to agree on a premaster secret.\n\n - Exchange certificates and cryptographic information to allow the\n client and server to authenticate themselves.\n\n - Generate a master secret from the premaster secret and exchanged\n random values.\n\n - Provide security parameters to the record layer.\n\n - Allow the client and server to verify that their peer has\n calculated the same security parameters and that the handshake\n occurred without tampering by an attacker.\n\n Note that higher layers should not be overly reliant on whether TLS\n always negotiates the strongest possible connection between two\n peers. There are a number of ways in which a man-in-the-middle\n attacker can attempt to make two entities drop down to the least\n secure method they support. The protocol has been designed to\n minimize this risk, but there are still attacks available: for\n example, an attacker could block access to the port a secure service\n runs on, or attempt to get the peers to negotiate an unauthenticated\n connection. The fundamental rule is that higher levels must be\n cognizant of what their security requirements are and never transmit\n information over a channel less secure than what they require. The\n TLS protocol is secure in that any cipher suite offers its promised\n level of security: if you negotiate 3DES with a 1024-bit RSA key\n exchange with a host whose certificate you have verified, you can\n expect to be that secure.\n\n These goals are achieved by the handshake protocol, which can be\n summarized as follows: The client sends a ClientHello message to\n which the server must respond with a ServerHello message, or else a\n fatal error will occur and the connection will fail. The ClientHello\n and ServerHello are used to establish security enhancement\n capabilities between client and server. The ClientHello and\n ServerHello establish the following attributes: Protocol Version,\n Session ID, Cipher Suite, and Compression Method. Additionally, two\n random values are generated and exchanged: ClientHello.random and\n ServerHello.random.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 34]\n\f\nRFC 5246 TLS August 2008\n\n\n The actual key exchange uses up to four messages: the server\n Certificate, the ServerKeyExchange, the client Certificate, and the\n ClientKeyExchange. New key exchange methods can be created by\n specifying a format for these messages and by defining the use of the\n messages to allow the client and server to agree upon a shared\n secret. This secret MUST be quite long; currently defined key\n exchange methods exchange secrets that range from 46 bytes upwards.\n\n Following the hello messages, the server will send its certificate in\n a Certificate message if it is to be authenticated. Additionally, a\n ServerKeyExchange message may be sent, if it is required (e.g., if\n the server has no certificate, or if its certificate is for signing\n only). If the server is authenticated, it may request a certificate\n from the client, if that is appropriate to the cipher suite selected.\n Next, the server will send the ServerHelloDone message, indicating\n that the hello-message phase of the handshake is complete. The\n server will then wait for a client response. If the server has sent\n a CertificateRequest message, the client MUST send the Certificate\n message. The ClientKeyExchange message is now sent, and the content\n of that message will depend on the public key algorithm selected\n between the ClientHello and the ServerHello. If the client has sent\n a certificate with signing ability, a digitally-signed\n CertificateVerify message is sent to explicitly verify possession of\n the private key in the certificate.\n\n At this point, a ChangeCipherSpec message is sent by the client, and\n the client copies the pending Cipher Spec into the current Cipher\n Spec. The client then immediately sends the Finished message under\n the new algorithms, keys, and secrets. In response, the server will\n send its own ChangeCipherSpec message, transfer the pending to the\n current Cipher Spec, and send its Finished message under the new\n Cipher Spec. At this point, the handshake is complete, and the\n client and server may begin to exchange application layer data. (See\n flow chart below.) Application data MUST NOT be sent prior to the\n completion of the first handshake (before a cipher suite other than\n TLS_NULL_WITH_NULL_NULL is established).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 35]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n Certificate*\n ServerKeyExchange*\n CertificateRequest*\n <-------- ServerHelloDone\n Certificate*\n ClientKeyExchange\n CertificateVerify*\n [ChangeCipherSpec]\n Finished -------->\n [ChangeCipherSpec]\n <-------- Finished\n Application Data <-------> Application Data\n\n Figure 1. Message flow for a full handshake\n\n * Indicates optional or situation-dependent messages that are not\n always sent.\n\n Note: To help avoid pipeline stalls, ChangeCipherSpec is an\n independent TLS protocol content type, and is not actually a TLS\n handshake message.\n\n When the client and server decide to resume a previous session or\n duplicate an existing session (instead of negotiating new security\n parameters), the message flow is as follows:\n\n The client sends a ClientHello using the Session ID of the session to\n be resumed. The server then checks its session cache for a match.\n If a match is found, and the server is willing to re-establish the\n connection under the specified session state, it will send a\n ServerHello with the same Session ID value. At this point, both\n client and server MUST send ChangeCipherSpec messages and proceed\n directly to Finished messages. Once the re-establishment is\n complete, the client and server MAY begin to exchange application\n layer data. (See flow chart below.) If a Session ID match is not\n found, the server generates a new session ID, and the TLS client and\n server perform a full handshake.\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 36]\n\f\nRFC 5246 TLS August 2008\n\n\n Client Server\n\n ClientHello -------->\n ServerHello\n [ChangeCipherSpec]\n <-------- Finished\n [ChangeCipherSpec]\n Finished -------->\n Application Data <-------> Application Data\n\n Figure 2. Message flow for an abbreviated handshake\n\n The contents and significance of each message will be presented in\n detail in the following sections.\n\n7.4. Handshake Protocol\n\n The TLS Handshake Protocol is one of the defined higher-level clients\n of the TLS Record Protocol. This protocol is used to negotiate the\n secure attributes of a session. Handshake messages are supplied to\n the TLS record layer, where they are encapsulated within one or more\n TLSPlaintext structures, which are processed and transmitted as\n specified by the current active session state.\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n } body;\n } Handshake;\n\n\n\nDierks & Rescorla Standards Track [Page 37]\n\f\nRFC 5246 TLS August 2008\n\n\n The handshake protocol messages are presented below in the order they\n MUST be sent; sending handshake messages in an unexpected order\n results in a fatal error. Unneeded handshake messages can be\n omitted, however. Note one exception to the ordering: the\n Certificate message is used twice in the handshake (from server to\n client, then from client to server), but described only in its first\n position. The one message that is not bound by these ordering rules\n is the HelloRequest message, which can be sent at any time, but which\n SHOULD be ignored by the client if it arrives in the middle of a\n handshake.\n\n New handshake message types are assigned by IANA as described in\n Section 12.\n\n7.4.1. Hello Messages\n\n The hello phase messages are used to exchange security enhancement\n capabilities between the client and server. When a new session\n begins, the record layer's connection state encryption, hash, and\n compression algorithms are initialized to null. The current\n connection state is used for renegotiation messages.\n\n7.4.1.1. Hello Request\n\n When this message will be sent:\n\n The HelloRequest message MAY be sent by the server at any time.\n\n Meaning of this message:\n\n HelloRequest is a simple notification that the client should begin\n the negotiation process anew. In response, the client should send\n a ClientHello message when convenient. This message is not\n intended to establish which side is the client or server but\n merely to initiate a new negotiation. Servers SHOULD NOT send a\n HelloRequest immediately upon the client's initial connection. It\n is the client's job to send a ClientHello at that time.\n\n This message will be ignored by the client if the client is\n currently negotiating a session. This message MAY be ignored by\n the client if it does not wish to renegotiate a session, or the\n client may, if it wishes, respond with a no_renegotiation alert.\n Since handshake messages are intended to have transmission\n precedence over application data, it is expected that the\n negotiation will begin before no more than a few records are\n received from the client. If the server sends a HelloRequest but\n does not receive a ClientHello in response, it may close the\n connection with a fatal alert.\n\n\n\nDierks & Rescorla Standards Track [Page 38]\n\f\nRFC 5246 TLS August 2008\n\n\n After sending a HelloRequest, servers SHOULD NOT repeat the\n request until the subsequent handshake negotiation is complete.\n\n Structure of this message:\n\n struct { } HelloRequest;\n\n This message MUST NOT be included in the message hashes that are\n maintained throughout the handshake and used in the Finished messages\n and the certificate verify message.\n\n7.4.1.2. Client Hello\n\n When this message will be sent:\n\n When a client first connects to a server, it is required to send\n the ClientHello as its first message. The client can also send a\n ClientHello in response to a HelloRequest or on its own initiative\n in order to renegotiate the security parameters in an existing\n connection.\n\n Structure of this message:\n\n The ClientHello message includes a random structure, which is used\n later in the protocol.\n\n struct {\n uint32 gmt_unix_time;\n opaque random_bytes[28];\n } Random;\n\n gmt_unix_time\n The current time and date in standard UNIX 32-bit format\n (seconds since the midnight starting Jan 1, 1970, UTC, ignoring\n leap seconds) according to the sender's internal clock. Clocks\n are not required to be set correctly by the basic TLS protocol;\n higher-level or application protocols may define additional\n requirements. Note that, for historical reasons, the data\n element is named using GMT, the predecessor of the current\n worldwide time base, UTC.\n\n random_bytes\n 28 bytes generated by a secure random number generator.\n\n The ClientHello message includes a variable-length session\n identifier. If not empty, the value identifies a session between the\n same client and server whose security parameters the client wishes to\n reuse. The session identifier MAY be from an earlier connection,\n\n\n\nDierks & Rescorla Standards Track [Page 39]\n\f\nRFC 5246 TLS August 2008\n\n\n this connection, or from another currently active connection. The\n second option is useful if the client only wishes to update the\n random structures and derived values of a connection, and the third\n option makes it possible to establish several independent secure\n connections without repeating the full handshake protocol. These\n independent connections may occur sequentially or simultaneously; a\n SessionID becomes valid when the handshake negotiating it completes\n with the exchange of Finished messages and persists until it is\n removed due to aging or because a fatal error was encountered on a\n connection associated with the session. The actual contents of the\n SessionID are defined by the server.\n\n opaque SessionID<0..32>;\n\n Warning: Because the SessionID is transmitted without encryption or\n immediate MAC protection, servers MUST NOT place confidential\n information in session identifiers or let the contents of fake\n session identifiers cause any breach of security. (Note that the\n content of the handshake as a whole, including the SessionID, is\n protected by the Finished messages exchanged at the end of the\n handshake.)\n\n The cipher suite list, passed from the client to the server in the\n ClientHello message, contains the combinations of cryptographic\n algorithms supported by the client in order of the client's\n preference (favorite choice first). Each cipher suite defines a key\n exchange algorithm, a bulk encryption algorithm (including secret key\n length), a MAC algorithm, and a PRF. The server will select a cipher\n suite or, if no acceptable choices are presented, return a handshake\n failure alert and close the connection. If the list contains cipher\n suites the server does not recognize, support, or wish to use, the\n server MUST ignore those cipher suites, and process the remaining\n ones as usual.\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n The ClientHello includes a list of compression algorithms supported\n by the client, ordered according to the client's preference.\n\n enum { null(0), (255) } CompressionMethod;\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 40]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n ProtocolVersion client_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suites<2..2^16-2>;\n CompressionMethod compression_methods<1..2^8-1>;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ClientHello;\n\n TLS allows extensions to follow the compression_methods field in an\n extensions block. The presence of extensions can be detected by\n determining whether there are bytes following the compression_methods\n at the end of the ClientHello. Note that this method of detecting\n optional data differs from the normal TLS method of having a\n variable-length field, but it is used for compatibility with TLS\n before extensions were defined.\n\n client_version\n The version of the TLS protocol by which the client wishes to\n communicate during this session. This SHOULD be the latest\n (highest valued) version supported by the client. For this\n version of the specification, the version will be 3.3 (see\n Appendix E for details about backward compatibility).\n\n random\n A client-generated random structure.\n\n session_id\n The ID of a session the client wishes to use for this connection.\n This field is empty if no session_id is available, or if the\n client wishes to generate new security parameters.\n\n cipher_suites\n This is a list of the cryptographic options supported by the\n client, with the client's first preference first. If the\n session_id field is not empty (implying a session resumption\n request), this vector MUST include at least the cipher_suite from\n that session. Values are defined in Appendix A.5.\n\n compression_methods\n This is a list of the compression methods supported by the client,\n sorted by client preference. If the session_id field is not empty\n (implying a session resumption request), it MUST include the\n\n\n\nDierks & Rescorla Standards Track [Page 41]\n\f\nRFC 5246 TLS August 2008\n\n\n compression_method from that session. This vector MUST contain,\n and all implementations MUST support, CompressionMethod.null.\n Thus, a client and server will always be able to agree on a\n compression method.\n\n extensions\n Clients MAY request extended functionality from servers by sending\n data in the extensions field. The actual \"Extension\" format is\n defined in Section 7.4.1.4.\n\n In the event that a client requests additional functionality using\n extensions, and this functionality is not supplied by the server, the\n client MAY abort the handshake. A server MUST accept ClientHello\n messages both with and without the extensions field, and (as for all\n other messages) it MUST check that the amount of data in the message\n precisely matches one of these formats; if not, then it MUST send a\n fatal \"decode_error\" alert.\n\n After sending the ClientHello message, the client waits for a\n ServerHello message. Any handshake message returned by the server,\n except for a HelloRequest, is treated as a fatal error.\n\n7.4.1.3. Server Hello\n\n When this message will be sent:\n\n The server will send this message in response to a ClientHello\n message when it was able to find an acceptable set of algorithms.\n If it cannot find such a match, it will respond with a handshake\n failure alert.\n\n Structure of this message:\n\n struct {\n ProtocolVersion server_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ServerHello;\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 42]\n\f\nRFC 5246 TLS August 2008\n\n\n The presence of extensions can be detected by determining whether\n there are bytes following the compression_method field at the end of\n the ServerHello.\n\n server_version\n This field will contain the lower of that suggested by the client\n in the client hello and the highest supported by the server. For\n this version of the specification, the version is 3.3. (See\n Appendix E for details about backward compatibility.)\n\n random\n This structure is generated by the server and MUST be\n independently generated from the ClientHello.random.\n\n session_id\n This is the identity of the session corresponding to this\n connection. If the ClientHello.session_id was non-empty, the\n server will look in its session cache for a match. If a match is\n found and the server is willing to establish the new connection\n using the specified session state, the server will respond with\n the same value as was supplied by the client. This indicates a\n resumed session and dictates that the parties must proceed\n directly to the Finished messages. Otherwise, this field will\n contain a different value identifying the new session. The server\n may return an empty session_id to indicate that the session will\n not be cached and therefore cannot be resumed. If a session is\n resumed, it must be resumed using the same cipher suite it was\n originally negotiated with. Note that there is no requirement\n that the server resume any session even if it had formerly\n provided a session_id. Clients MUST be prepared to do a full\n negotiation -- including negotiating new cipher suites -- during\n any handshake.\n\n cipher_suite\n The single cipher suite selected by the server from the list in\n ClientHello.cipher_suites. For resumed sessions, this field is\n the value from the state of the session being resumed.\n\n compression_method\n The single compression algorithm selected by the server from the\n list in ClientHello.compression_methods. For resumed sessions,\n this field is the value from the resumed session state.\n\n extensions\n A list of extensions. Note that only extensions offered by the\n client can appear in the server's list.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 43]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.1.4. Hello Extensions\n\n The extension format is:\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n signature_algorithms(13), (65535)\n } ExtensionType;\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The initial set of extensions is defined in a companion document\n [TLSEXT]. The list of extension types is maintained by IANA as\n described in Section 12.\n\n An extension type MUST NOT appear in the ServerHello unless the same\n extension type appeared in the corresponding ClientHello. If a\n client receives an extension type in ServerHello that it did not\n request in the associated ClientHello, it MUST abort the handshake\n with an unsupported_extension fatal alert.\n\n Nonetheless, \"server-oriented\" extensions may be provided in the\n future within this framework. Such an extension (say, of type x)\n would require the client to first send an extension of type x in a\n ClientHello with empty extension_data to indicate that it supports\n the extension type. In this case, the client is offering the\n capability to understand the extension type, and the server is taking\n the client up on its offer.\n\n When multiple extensions of different types are present in the\n ClientHello or ServerHello messages, the extensions MAY appear in any\n order. There MUST NOT be more than one extension of the same type.\n\n Finally, note that extensions can be sent both when starting a new\n session and when requesting session resumption. Indeed, a client\n that requests session resumption does not in general know whether the\n server will accept this request, and therefore it SHOULD send the\n same extensions as it would send if it were not attempting\n resumption.\n\n\n\nDierks & Rescorla Standards Track [Page 44]\n\f\nRFC 5246 TLS August 2008\n\n\n In general, the specification of each extension type needs to\n describe the effect of the extension both during full handshake and\n session resumption. Most current TLS extensions are relevant only\n when a session is initiated: when an older session is resumed, the\n server does not process these extensions in Client Hello, and does\n not include them in Server Hello. However, some extensions may\n specify different behavior during session resumption.\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions, and some are simply refusals to support particular\n features. In general, error alerts should be used for the former,\n and a field in the server extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem.\n\n Often the fact that the extension fields are included in the\n inputs to the Finished message hashes will be sufficient, but\n extreme care is needed when the extension changes the meaning of\n messages sent in the handshake phase. Designers and implementors\n should be aware of the fact that until the handshake has been\n authenticated, active attackers can modify messages and insert,\n remove, or replace extensions.\n\n - It would be technically possible to use extensions to change major\n aspects of the design of TLS; for example the design of cipher\n suite negotiation. This is not recommended; it would be more\n appropriate to define a new version of TLS -- particularly since\n the TLS handshake algorithms have specific protection against\n version rollback attacks based on the version number, and the\n possibility of version rollback should be a significant\n consideration in any major design change.\n\n7.4.1.4.1. Signature Algorithms\n\n The client uses the \"signature_algorithms\" extension to indicate to\n the server which signature/hash algorithm pairs may be used in\n digital signatures. The \"extension_data\" field of this extension\n contains a \"supported_signature_algorithms\" value.\n\n\n\nDierks & Rescorla Standards Track [Page 45]\n\f\nRFC 5246 TLS August 2008\n\n\n enum {\n none(0), md5(1), sha1(2), sha224(3), sha256(4), sha384(5),\n sha512(6), (255)\n } HashAlgorithm;\n\n enum { anonymous(0), rsa(1), dsa(2), ecdsa(3), (255) }\n SignatureAlgorithm;\n\n struct {\n HashAlgorithm hash;\n SignatureAlgorithm signature;\n } SignatureAndHashAlgorithm;\n\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2..2^16-2>;\n\n Each SignatureAndHashAlgorithm value lists a single hash/signature\n pair that the client is willing to verify. The values are indicated\n in descending order of preference.\n\n Note: Because not all signature algorithms and hash algorithms may be\n accepted by an implementation (e.g., DSA with SHA-1, but not\n SHA-256), algorithms here are listed in pairs.\n\n hash\n This field indicates the hash algorithm which may be used. The\n values indicate support for unhashed data, MD5 [MD5], SHA-1,\n SHA-224, SHA-256, SHA-384, and SHA-512 [SHS], respectively. The\n \"none\" value is provided for future extensibility, in case of a\n signature algorithm which does not require hashing before signing.\n\n signature\n This field indicates the signature algorithm that may be used.\n The values indicate anonymous signatures, RSASSA-PKCS1-v1_5\n [PKCS1] and DSA [DSS], and ECDSA [ECDSA], respectively. The\n \"anonymous\" value is meaningless in this context but used in\n Section 7.4.3. It MUST NOT appear in this extension.\n\n The semantics of this extension are somewhat complicated because the\n cipher suite indicates permissible signature algorithms but not hash\n algorithms. Sections 7.4.2 and 7.4.3 describe the appropriate rules.\n\n If the client supports only the default hash and signature algorithms\n (listed in this section), it MAY omit the signature_algorithms\n extension. If the client does not support the default algorithms, or\n supports other hash and signature algorithms (and it is willing to\n use them for verifying messages sent by the server, i.e., server\n certificates and server key exchange), it MUST send the\n\n\n\nDierks & Rescorla Standards Track [Page 46]\n\f\nRFC 5246 TLS August 2008\n\n\n signature_algorithms extension, listing the algorithms it is willing\n to accept.\n\n If the client does not send the signature_algorithms extension, the\n server MUST do the following:\n\n - If the negotiated key exchange algorithm is one of (RSA, DHE_RSA,\n DH_RSA, RSA_PSK, ECDH_RSA, ECDHE_RSA), behave as if client had\n sent the value {sha1,rsa}.\n\n - If the negotiated key exchange algorithm is one of (DHE_DSS,\n DH_DSS), behave as if the client had sent the value {sha1,dsa}.\n\n - If the negotiated key exchange algorithm is one of (ECDH_ECDSA,\n ECDHE_ECDSA), behave as if the client had sent value {sha1,ecdsa}.\n\n Note: this is a change from TLS 1.1 where there are no explicit\n rules, but as a practical matter one can assume that the peer\n supports MD5 and SHA-1.\n\n Note: this extension is not meaningful for TLS versions prior to 1.2.\n Clients MUST NOT offer it if they are offering prior versions.\n However, even if clients do offer it, the rules specified in [TLSEXT]\n require servers to ignore extensions they do not understand.\n\n Servers MUST NOT send this extension. TLS servers MUST support\n receiving this extension.\n\n When performing session resumption, this extension is not included in\n Server Hello, and the server ignores the extension in Client Hello\n (if present).\n\n7.4.2. Server Certificate\n\n When this message will be sent:\n\n The server MUST send a Certificate message whenever the agreed-\n upon key exchange method uses certificates for authentication\n (this includes all key exchange methods defined in this document\n except DH_anon). This message will always immediately follow the\n ServerHello message.\n\n Meaning of this message:\n\n This message conveys the server's certificate chain to the client.\n\n The certificate MUST be appropriate for the negotiated cipher\n suite's key exchange algorithm and any negotiated extensions.\n\n\n\nDierks & Rescorla Standards Track [Page 47]\n\f\nRFC 5246 TLS August 2008\n\n\n Structure of this message:\n\n opaque ASN.1Cert<1..2^24-1>;\n\n struct {\n ASN.1Cert certificate_list<0..2^24-1>;\n } Certificate;\n\n certificate_list\n This is a sequence (chain) of certificates. The sender's\n certificate MUST come first in the list. Each following\n certificate MUST directly certify the one preceding it. Because\n certificate validation requires that root keys be distributed\n independently, the self-signed certificate that specifies the root\n certificate authority MAY be omitted from the chain, under the\n assumption that the remote end must already possess it in order to\n validate it in any case.\n\n The same message type and structure will be used for the client's\n response to a certificate request message. Note that a client MAY\n send no certificates if it does not have an appropriate certificate\n to send in response to the server's authentication request.\n\n Note: PKCS #7 [PKCS7] is not used as the format for the certificate\n vector because PKCS #6 [PKCS6] extended certificates are not used.\n Also, PKCS #7 defines a SET rather than a SEQUENCE, making the task\n of parsing the list more difficult.\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3, unless explicitly negotiated\n otherwise (e.g., [TLSPGP]).\n\n - The end entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected key exchange\n algorithm.\n\n Key Exchange Alg. Certificate Key Type\n\n RSA RSA public key; the certificate MUST allow the\n RSA_PSK key to be used for encryption (the\n keyEncipherment bit MUST be set if the key\n usage extension is present).\n Note: RSA_PSK is defined in [TLSPSK].\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 48]\n\f\nRFC 5246 TLS August 2008\n\n\n DHE_RSA RSA public key; the certificate MUST allow the\n ECDHE_RSA key to be used for signing (the\n digitalSignature bit MUST be set if the key\n usage extension is present) with the signature\n scheme and hash algorithm that will be employed\n in the server key exchange message.\n Note: ECDHE_RSA is defined in [TLSECC].\n\n DHE_DSS DSA public key; the certificate MUST allow the\n key to be used for signing with the hash\n algorithm that will be employed in the server\n key exchange message.\n\n DH_DSS Diffie-Hellman public key; the keyAgreement bit\n DH_RSA MUST be set if the key usage extension is\n present.\n\n ECDH_ECDSA ECDH-capable public key; the public key MUST\n ECDH_RSA use a curve and point format supported by the\n client, as described in [TLSECC].\n\n ECDHE_ECDSA ECDSA-capable public key; the certificate MUST\n allow the key to be used for signing with the\n hash algorithm that will be employed in the\n server key exchange message. The public key\n MUST use a curve and point format supported by\n the client, as described in [TLSECC].\n\n - The \"server_name\" and \"trusted_ca_keys\" extensions [TLSEXT] are\n used to guide certificate selection.\n\n If the client provided a \"signature_algorithms\" extension, then all\n certificates provided by the server MUST be signed by a\n hash/signature algorithm pair that appears in that extension. Note\n that this implies that a certificate containing a key for one\n signature algorithm MAY be signed using a different signature\n algorithm (for instance, an RSA key signed with a DSA key). This is\n a departure from TLS 1.1, which required that the algorithms be the\n same. Note that this also implies that the DH_DSS, DH_RSA,\n ECDH_ECDSA, and ECDH_RSA key exchange algorithms do not restrict the\n algorithm used to sign the certificate. Fixed DH certificates MAY be\n signed with any hash/signature algorithm pair appearing in the\n extension. The names DH_DSS, DH_RSA, ECDH_ECDSA, and ECDH_RSA are\n historical.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 49]\n\f\nRFC 5246 TLS August 2008\n\n\n If the server has multiple certificates, it chooses one of them based\n on the above-mentioned criteria (in addition to other criteria, such\n as transport layer endpoint, local configuration and preferences,\n etc.). If the server has a single certificate, it SHOULD attempt to\n validate that it meets these criteria.\n\n Note that there are certificates that use algorithms and/or algorithm\n combinations that cannot be currently used with TLS. For example, a\n certificate with RSASSA-PSS signature key (id-RSASSA-PSS OID in\n SubjectPublicKeyInfo) cannot be used because TLS defines no\n corresponding signature algorithm.\n\n As cipher suites that specify new key exchange methods are specified\n for the TLS protocol, they will imply the certificate format and the\n required encoded keying information.\n\n7.4.3. Server Key Exchange Message\n\n When this message will be sent:\n\n This message will be sent immediately after the server Certificate\n message (or the ServerHello message, if this is an anonymous\n negotiation).\n\n The ServerKeyExchange message is sent by the server only when the\n server Certificate message (if sent) does not contain enough data\n to allow the client to exchange a premaster secret. This is true\n for the following key exchange methods:\n\n DHE_DSS\n DHE_RSA\n DH_anon\n\n It is not legal to send the ServerKeyExchange message for the\n following key exchange methods:\n\n RSA\n DH_DSS\n DH_RSA\n\n Other key exchange algorithms, such as those defined in [TLSECC],\n MUST specify whether the ServerKeyExchange message is sent or not;\n and if the message is sent, its contents.\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 50]\n\f\nRFC 5246 TLS August 2008\n\n\n Meaning of this message:\n\n This message conveys cryptographic information to allow the client\n to communicate the premaster secret: a Diffie-Hellman public key\n with which the client can complete a key exchange (with the result\n being the premaster secret) or a public key for some other\n algorithm.\n\n Structure of this message:\n\n enum { dhe_dss, dhe_rsa, dh_anon, rsa, dh_dss, dh_rsa\n /* may be extended, e.g., for ECDH -- see [TLSECC] */\n } KeyExchangeAlgorithm;\n\n struct {\n opaque dh_p<1..2^16-1>;\n opaque dh_g<1..2^16-1>;\n opaque dh_Ys<1..2^16-1>;\n } ServerDHParams; /* Ephemeral DH parameters */\n\n dh_p\n The prime modulus used for the Diffie-Hellman operation.\n\n dh_g\n The generator used for the Diffie-Hellman operation.\n\n dh_Ys\n The server's Diffie-Hellman public value (g^X mod p).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 51]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n select (KeyExchangeAlgorithm) {\n case dh_anon:\n ServerDHParams params;\n case dhe_dss:\n case dhe_rsa:\n ServerDHParams params;\n digitally-signed struct {\n opaque client_random[32];\n opaque server_random[32];\n ServerDHParams params;\n } signed_params;\n case rsa:\n case dh_dss:\n case dh_rsa:\n struct {} ;\n /* message is omitted for rsa, dh_dss, and dh_rsa */\n /* may be extended, e.g., for ECDH -- see [TLSECC] */\n };\n } ServerKeyExchange;\n\n params\n The server's key exchange parameters.\n\n signed_params\n For non-anonymous key exchanges, a signature over the server's\n key exchange parameters.\n\n If the client has offered the \"signature_algorithms\" extension, the\n signature algorithm and hash algorithm MUST be a pair listed in that\n extension. Note that there is a possibility for inconsistencies\n here. For instance, the client might offer DHE_DSS key exchange but\n omit any DSA pairs from its \"signature_algorithms\" extension. In\n order to negotiate correctly, the server MUST check any candidate\n cipher suites against the \"signature_algorithms\" extension before\n selecting them. This is somewhat inelegant but is a compromise\n designed to minimize changes to the original cipher suite design.\n\n In addition, the hash and signature algorithms MUST be compatible\n with the key in the server's end-entity certificate. RSA keys MAY be\n used with any permitted hash algorithm, subject to restrictions in\n the certificate, if any.\n\n Because DSA signatures do not contain any secure indication of hash\n algorithm, there is a risk of hash substitution if multiple hashes\n may be used with any key. Currently, DSA [DSS] may only be used with\n SHA-1. Future revisions of DSS [DSS-3] are expected to allow the use\n of other digest algorithms with DSA, as well as guidance as to which\n\n\n\nDierks & Rescorla Standards Track [Page 52]\n\f\nRFC 5246 TLS August 2008\n\n\n digest algorithms should be used with each key size. In addition,\n future revisions of [PKIX] may specify mechanisms for certificates to\n indicate which digest algorithms are to be used with DSA.\n\n As additional cipher suites are defined for TLS that include new key\n exchange algorithms, the server key exchange message will be sent if\n and only if the certificate type associated with the key exchange\n algorithm does not provide enough information for the client to\n exchange a premaster secret.\n\n7.4.4. Certificate Request\n\n When this message will be sent:\n\n A non-anonymous server can optionally request a certificate from\n the client, if appropriate for the selected cipher suite. This\n message, if sent, will immediately follow the ServerKeyExchange\n message (if it is sent; otherwise, this message follows the\n server's Certificate message).\n\n Structure of this message:\n\n enum {\n rsa_sign(1), dss_sign(2), rsa_fixed_dh(3), dss_fixed_dh(4),\n rsa_ephemeral_dh_RESERVED(5), dss_ephemeral_dh_RESERVED(6),\n fortezza_dms_RESERVED(20), (255)\n } ClientCertificateType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n ClientCertificateType certificate_types<1..2^8-1>;\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2^16-1>;\n DistinguishedName certificate_authorities<0..2^16-1>;\n } CertificateRequest;\n\n certificate_types\n A list of the types of certificate types that the client may\n offer.\n\n rsa_sign a certificate containing an RSA key\n dss_sign a certificate containing a DSA key\n rsa_fixed_dh a certificate containing a static DH key.\n dss_fixed_dh a certificate containing a static DH key\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 53]\n\f\nRFC 5246 TLS August 2008\n\n\n supported_signature_algorithms\n A list of the hash/signature algorithm pairs that the server is\n able to verify, listed in descending order of preference.\n\n certificate_authorities\n A list of the distinguished names [X501] of acceptable\n certificate_authorities, represented in DER-encoded format. These\n distinguished names may specify a desired distinguished name for a\n root CA or for a subordinate CA; thus, this message can be used to\n describe known roots as well as a desired authorization space. If\n the certificate_authorities list is empty, then the client MAY\n send any certificate of the appropriate ClientCertificateType,\n unless there is some external arrangement to the contrary.\n\n The interaction of the certificate_types and\n supported_signature_algorithms fields is somewhat complicated.\n certificate_types has been present in TLS since SSLv3, but was\n somewhat underspecified. Much of its functionality is superseded by\n supported_signature_algorithms. The following rules apply:\n\n - Any certificates provided by the client MUST be signed using a\n hash/signature algorithm pair found in\n supported_signature_algorithms.\n\n - The end-entity certificate provided by the client MUST contain a\n key that is compatible with certificate_types. If the key is a\n signature key, it MUST be usable with some hash/signature\n algorithm pair in supported_signature_algorithms.\n\n - For historical reasons, the names of some client certificate types\n include the algorithm used to sign the certificate. For example,\n in earlier versions of TLS, rsa_fixed_dh meant a certificate\n signed with RSA and containing a static DH key. In TLS 1.2, this\n functionality has been obsoleted by the\n supported_signature_algorithms, and the certificate type no longer\n restricts the algorithm used to sign the certificate. For\n example, if the server sends dss_fixed_dh certificate type and\n {{sha1, dsa}, {sha1, rsa}} signature types, the client MAY reply\n with a certificate containing a static DH key, signed with RSA-\n SHA1.\n\n New ClientCertificateType values are assigned by IANA as described in\n Section 12.\n\n Note: Values listed as RESERVED may not be used. They were used in\n SSLv3.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 54]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: It is a fatal handshake_failure alert for an anonymous server\n to request client authentication.\n\n7.4.5. Server Hello Done\n\n When this message will be sent:\n\n The ServerHelloDone message is sent by the server to indicate the\n end of the ServerHello and associated messages. After sending\n this message, the server will wait for a client response.\n\n Meaning of this message:\n\n This message means that the server is done sending messages to\n support the key exchange, and the client can proceed with its\n phase of the key exchange.\n\n Upon receipt of the ServerHelloDone message, the client SHOULD\n verify that the server provided a valid certificate, if required,\n and check that the server hello parameters are acceptable.\n\n Structure of this message:\n\n struct { } ServerHelloDone;\n\n7.4.6. Client Certificate\n\n When this message will be sent:\n\n This is the first message the client can send after receiving a\n ServerHelloDone message. This message is only sent if the server\n requests a certificate. If no suitable certificate is available,\n the client MUST send a certificate message containing no\n certificates. That is, the certificate_list structure has a\n length of zero. If the client does not send any certificates, the\n server MAY at its discretion either continue the handshake without\n client authentication, or respond with a fatal handshake_failure\n alert. Also, if some aspect of the certificate chain was\n unacceptable (e.g., it was not signed by a known, trusted CA), the\n server MAY at its discretion either continue the handshake\n (considering the client unauthenticated) or send a fatal alert.\n\n Client certificates are sent using the Certificate structure\n defined in Section 7.4.2.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 55]\n\f\nRFC 5246 TLS August 2008\n\n\n Meaning of this message:\n\n This message conveys the client's certificate chain to the server;\n the server will use it when verifying the CertificateVerify\n message (when the client authentication is based on signing) or\n calculating the premaster secret (for non-ephemeral Diffie-\n Hellman). The certificate MUST be appropriate for the negotiated\n cipher suite's key exchange algorithm, and any negotiated\n extensions.\n\n In particular:\n\n - The certificate type MUST be X.509v3, unless explicitly negotiated\n otherwise (e.g., [TLSPGP]).\n\n - The end-entity certificate's public key (and associated\n restrictions) has to be compatible with the certificate types\n listed in CertificateRequest:\n\n Client Cert. Type Certificate Key Type\n\n rsa_sign RSA public key; the certificate MUST allow the\n key to be used for signing with the signature\n scheme and hash algorithm that will be\n employed in the certificate verify message.\n\n dss_sign DSA public key; the certificate MUST allow the\n key to be used for signing with the hash\n algorithm that will be employed in the\n certificate verify message.\n\n ecdsa_sign ECDSA-capable public key; the certificate MUST\n allow the key to be used for signing with the\n hash algorithm that will be employed in the\n certificate verify message; the public key\n MUST use a curve and point format supported by\n the server.\n\n rsa_fixed_dh Diffie-Hellman public key; MUST use the same\n dss_fixed_dh parameters as server's key.\n\n rsa_fixed_ecdh ECDH-capable public key; MUST use the\n ecdsa_fixed_ecdh same curve as the server's key, and MUST use a\n point format supported by the server.\n\n - If the certificate_authorities list in the certificate request\n message was non-empty, one of the certificates in the certificate\n chain SHOULD be issued by one of the listed CAs.\n\n\n\nDierks & Rescorla Standards Track [Page 56]\n\f\nRFC 5246 TLS August 2008\n\n\n - The certificates MUST be signed using an acceptable hash/\n signature algorithm pair, as described in Section 7.4.4. Note\n that this relaxes the constraints on certificate-signing\n algorithms found in prior versions of TLS.\n\n Note that, as with the server certificate, there are certificates\n that use algorithms/algorithm combinations that cannot be currently\n used with TLS.\n\n7.4.7. Client Key Exchange Message\n\n When this message will be sent:\n\n This message is always sent by the client. It MUST immediately\n follow the client certificate message, if it is sent. Otherwise,\n it MUST be the first message sent by the client after it receives\n the ServerHelloDone message.\n\n Meaning of this message:\n\n With this message, the premaster secret is set, either by direct\n transmission of the RSA-encrypted secret or by the transmission of\n Diffie-Hellman parameters that will allow each side to agree upon\n the same premaster secret.\n\n When the client is using an ephemeral Diffie-Hellman exponent,\n then this message contains the client's Diffie-Hellman public\n value. If the client is sending a certificate containing a static\n DH exponent (i.e., it is doing fixed_dh client authentication),\n then this message MUST be sent but MUST be empty.\n\n Structure of this message:\n\n The choice of messages depends on which key exchange method has\n been selected. See Section 7.4.3 for the KeyExchangeAlgorithm\n definition.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 57]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n select (KeyExchangeAlgorithm) {\n case rsa:\n EncryptedPreMasterSecret;\n case dhe_dss:\n case dhe_rsa:\n case dh_dss:\n case dh_rsa:\n case dh_anon:\n ClientDiffieHellmanPublic;\n } exchange_keys;\n } ClientKeyExchange;\n\n7.4.7.1. RSA-Encrypted Premaster Secret Message\n\n Meaning of this message:\n\n If RSA is being used for key agreement and authentication, the\n client generates a 48-byte premaster secret, encrypts it using the\n public key from the server's certificate, and sends the result in\n an encrypted premaster secret message. This structure is a\n variant of the ClientKeyExchange message and is not a message in\n itself.\n\n Structure of this message:\n\n struct {\n ProtocolVersion client_version;\n opaque random[46];\n } PreMasterSecret;\n\n client_version\n The latest (newest) version supported by the client. This is\n used to detect version rollback attacks.\n\n random\n 46 securely-generated random bytes.\n\n struct {\n public-key-encrypted PreMasterSecret pre_master_secret;\n } EncryptedPreMasterSecret;\n\n pre_master_secret\n This random value is generated by the client and is used to\n generate the master secret, as specified in Section 8.1.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 58]\n\f\nRFC 5246 TLS August 2008\n\n\n Note: The version number in the PreMasterSecret is the version\n offered by the client in the ClientHello.client_version, not the\n version negotiated for the connection. This feature is designed to\n prevent rollback attacks. Unfortunately, some old implementations\n use the negotiated version instead, and therefore checking the\n version number may lead to failure to interoperate with such\n incorrect client implementations.\n\n Client implementations MUST always send the correct version number in\n PreMasterSecret. If ClientHello.client_version is TLS 1.1 or higher,\n server implementations MUST check the version number as described in\n the note below. If the version number is TLS 1.0 or earlier, server\n implementations SHOULD check the version number, but MAY have a\n configuration option to disable the check. Note that if the check\n fails, the PreMasterSecret SHOULD be randomized as described below.\n\n Note: Attacks discovered by Bleichenbacher [BLEI] and Klima et al.\n [KPR03] can be used to attack a TLS server that reveals whether a\n particular message, when decrypted, is properly PKCS#1 formatted,\n contains a valid PreMasterSecret structure, or has the correct\n version number.\n\n As described by Klima [KPR03], these vulnerabilities can be avoided\n by treating incorrectly formatted message blocks and/or mismatched\n version numbers in a manner indistinguishable from correctly\n formatted RSA blocks. In other words:\n\n 1. Generate a string R of 46 random bytes\n\n 2. Decrypt the message to recover the plaintext M\n\n 3. If the PKCS#1 padding is not correct, or the length of message\n M is not exactly 48 bytes:\n pre_master_secret = ClientHello.client_version || R\n else If ClientHello.client_version <= TLS 1.0, and version\n number check is explicitly disabled:\n pre_master_secret = M\n else:\n pre_master_secret = ClientHello.client_version || M[2..47]\n\n Note that explicitly constructing the pre_master_secret with the\n ClientHello.client_version produces an invalid master_secret if the\n client has sent the wrong version in the original pre_master_secret.\n\n An alternative approach is to treat a version number mismatch as a\n PKCS-1 formatting error and randomize the premaster secret\n completely:\n\n\n\n\nDierks & Rescorla Standards Track [Page 59]\n\f\nRFC 5246 TLS August 2008\n\n\n 1. Generate a string R of 48 random bytes\n\n 2. Decrypt the message to recover the plaintext M\n\n 3. If the PKCS#1 padding is not correct, or the length of message\n M is not exactly 48 bytes:\n pre_master_secret = R\n else If ClientHello.client_version <= TLS 1.0, and version\n number check is explicitly disabled:\n premaster secret = M\n else If M[0..1] != ClientHello.client_version:\n premaster secret = R\n else:\n premaster secret = M\n\n Although no practical attacks against this construction are known,\n Klima et al. [KPR03] describe some theoretical attacks, and therefore\n the first construction described is RECOMMENDED.\n\n In any case, a TLS server MUST NOT generate an alert if processing an\n RSA-encrypted premaster secret message fails, or the version number\n is not as expected. Instead, it MUST continue the handshake with a\n randomly generated premaster secret. It may be useful to log the\n real cause of failure for troubleshooting purposes; however, care\n must be taken to avoid leaking the information to an attacker\n (through, e.g., timing, log files, or other channels.)\n\n The RSAES-OAEP encryption scheme defined in [PKCS1] is more secure\n against the Bleichenbacher attack. However, for maximal\n compatibility with earlier versions of TLS, this specification uses\n the RSAES-PKCS1-v1_5 scheme. No variants of the Bleichenbacher\n attack are known to exist provided that the above recommendations are\n followed.\n\n Implementation note: Public-key-encrypted data is represented as an\n opaque vector <0..2^16-1> (see Section 4.7). Thus, the RSA-encrypted\n PreMasterSecret in a ClientKeyExchange is preceded by two length\n bytes. These bytes are redundant in the case of RSA because the\n EncryptedPreMasterSecret is the only data in the ClientKeyExchange\n and its length can therefore be unambiguously determined. The SSLv3\n specification was not clear about the encoding of public-key-\n encrypted data, and therefore many SSLv3 implementations do not\n include the length bytes -- they encode the RSA-encrypted data\n directly in the ClientKeyExchange message.\n\n This specification requires correct encoding of the\n EncryptedPreMasterSecret complete with length bytes. The resulting\n PDU is incompatible with many SSLv3 implementations. Implementors\n\n\n\nDierks & Rescorla Standards Track [Page 60]\n\f\nRFC 5246 TLS August 2008\n\n\n upgrading from SSLv3 MUST modify their implementations to generate\n and accept the correct encoding. Implementors who wish to be\n compatible with both SSLv3 and TLS should make their implementation's\n behavior dependent on the protocol version.\n\n Implementation note: It is now known that remote timing-based attacks\n on TLS are possible, at least when the client and server are on the\n same LAN. Accordingly, implementations that use static RSA keys MUST\n use RSA blinding or some other anti-timing technique, as described in\n [TIMING].\n\n7.4.7.2. Client Diffie-Hellman Public Value\n\n Meaning of this message:\n\n This structure conveys the client's Diffie-Hellman public value\n (Yc) if it was not already included in the client's certificate.\n The encoding used for Yc is determined by the enumerated\n PublicValueEncoding. This structure is a variant of the client\n key exchange message, and not a message in itself.\n\n Structure of this message:\n\n enum { implicit, explicit } PublicValueEncoding;\n\n implicit\n If the client has sent a certificate which contains a suitable\n Diffie-Hellman key (for fixed_dh client authentication), then\n Yc is implicit and does not need to be sent again. In this\n case, the client key exchange message will be sent, but it MUST\n be empty.\n\n explicit\n Yc needs to be sent.\n\n struct {\n select (PublicValueEncoding) {\n case implicit: struct { };\n case explicit: opaque dh_Yc<1..2^16-1>;\n } dh_public;\n } ClientDiffieHellmanPublic;\n\n dh_Yc\n The client's Diffie-Hellman public value (Yc).\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 61]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.8. Certificate Verify\n\n When this message will be sent:\n\n This message is used to provide explicit verification of a client\n certificate. This message is only sent following a client\n certificate that has signing capability (i.e., all certificates\n except those containing fixed Diffie-Hellman parameters). When\n sent, it MUST immediately follow the client key exchange message.\n\n Structure of this message:\n\n struct {\n digitally-signed struct {\n opaque handshake_messages[handshake_messages_length];\n }\n } CertificateVerify;\n\n Here handshake_messages refers to all handshake messages sent or\n received, starting at client hello and up to, but not including,\n this message, including the type and length fields of the\n handshake messages. This is the concatenation of all the\n Handshake structures (as defined in Section 7.4) exchanged thus\n far. Note that this requires both sides to either buffer the\n messages or compute running hashes for all potential hash\n algorithms up to the time of the CertificateVerify computation.\n Servers can minimize this computation cost by offering a\n restricted set of digest algorithms in the CertificateRequest\n message.\n\n The hash and signature algorithms used in the signature MUST be\n one of those present in the supported_signature_algorithms field\n of the CertificateRequest message. In addition, the hash and\n signature algorithms MUST be compatible with the key in the\n client's end-entity certificate. RSA keys MAY be used with any\n permitted hash algorithm, subject to restrictions in the\n certificate, if any.\n\n Because DSA signatures do not contain any secure indication of\n hash algorithm, there is a risk of hash substitution if multiple\n hashes may be used with any key. Currently, DSA [DSS] may only be\n used with SHA-1. Future revisions of DSS [DSS-3] are expected to\n allow the use of other digest algorithms with DSA, as well as\n guidance as to which digest algorithms should be used with each\n key size. In addition, future revisions of [PKIX] may specify\n mechanisms for certificates to indicate which digest algorithms\n are to be used with DSA.\n\n\n\n\nDierks & Rescorla Standards Track [Page 62]\n\f\nRFC 5246 TLS August 2008\n\n\n7.4.9. Finished\n\n When this message will be sent:\n\n A Finished message is always sent immediately after a change\n cipher spec message to verify that the key exchange and\n authentication processes were successful. It is essential that a\n change cipher spec message be received between the other handshake\n messages and the Finished message.\n\n Meaning of this message:\n\n The Finished message is the first one protected with the just\n negotiated algorithms, keys, and secrets. Recipients of Finished\n messages MUST verify that the contents are correct. Once a side\n has sent its Finished message and received and validated the\n Finished message from its peer, it may begin to send and receive\n application data over the connection.\n\n Structure of this message:\n\n struct {\n opaque verify_data[verify_data_length];\n } Finished;\n\n verify_data\n PRF(master_secret, finished_label, Hash(handshake_messages))\n [0..verify_data_length-1];\n\n finished_label\n For Finished messages sent by the client, the string\n \"client finished\". For Finished messages sent by the server,\n the string \"server finished\".\n\n Hash denotes a Hash of the handshake messages. For the PRF\n defined in Section 5, the Hash MUST be the Hash used as the basis\n for the PRF. Any cipher suite which defines a different PRF MUST\n also define the Hash to use in the Finished computation.\n\n In previous versions of TLS, the verify_data was always 12 octets\n long. In the current version of TLS, it depends on the cipher\n suite. Any cipher suite which does not explicitly specify\n verify_data_length has a verify_data_length equal to 12. This\n includes all existing cipher suites. Note that this\n representation has the same encoding as with previous versions.\n Future cipher suites MAY specify other lengths but such length\n MUST be at least 12 bytes.\n\n\n\n\nDierks & Rescorla Standards Track [Page 63]\n\f\nRFC 5246 TLS August 2008\n\n\n handshake_messages\n All of the data from all messages in this handshake (not\n including any HelloRequest messages) up to, but not including,\n this message. This is only data visible at the handshake layer\n and does not include record layer headers. This is the\n concatenation of all the Handshake structures as defined in\n Section 7.4, exchanged thus far.\n\n It is a fatal error if a Finished message is not preceded by a\n ChangeCipherSpec message at the appropriate point in the handshake.\n\n The value handshake_messages includes all handshake messages starting\n at ClientHello up to, but not including, this Finished message. This\n may be different from handshake_messages in Section 7.4.8 because it\n would include the CertificateVerify message (if sent). Also, the\n handshake_messages for the Finished message sent by the client will\n be different from that for the Finished message sent by the server,\n because the one that is sent second will include the prior one.\n\n Note: ChangeCipherSpec messages, alerts, and any other record types\n are not handshake messages and are not included in the hash\n computations. Also, HelloRequest messages are omitted from handshake\n hashes.\n\n8. Cryptographic Computations\n\n In order to begin connection protection, the TLS Record Protocol\n requires specification of a suite of algorithms, a master secret, and\n the client and server random values. The authentication, encryption,\n and MAC algorithms are determined by the cipher_suite selected by the\n server and revealed in the ServerHello message. The compression\n algorithm is negotiated in the hello messages, and the random values\n are exchanged in the hello messages. All that remains is to\n calculate the master secret.\n\n8.1. Computing the Master Secret\n\n For all key exchange methods, the same algorithm is used to convert\n the pre_master_secret into the master_secret. The pre_master_secret\n should be deleted from memory once the master_secret has been\n computed.\n\n master_secret = PRF(pre_master_secret, \"master secret\",\n ClientHello.random + ServerHello.random)\n [0..47];\n\n The master secret is always exactly 48 bytes in length. The length\n of the premaster secret will vary depending on key exchange method.\n\n\n\nDierks & Rescorla Standards Track [Page 64]\n\f\nRFC 5246 TLS August 2008\n\n\n8.1.1. RSA\n\n When RSA is used for server authentication and key exchange, a 48-\n byte pre_master_secret is generated by the client, encrypted under\n the server's public key, and sent to the server. The server uses its\n private key to decrypt the pre_master_secret. Both parties then\n convert the pre_master_secret into the master_secret, as specified\n above.\n\n8.1.2. Diffie-Hellman\n\n A conventional Diffie-Hellman computation is performed. The\n negotiated key (Z) is used as the pre_master_secret, and is converted\n into the master_secret, as specified above. Leading bytes of Z that\n contain all zero bits are stripped before it is used as the\n pre_master_secret.\n\n Note: Diffie-Hellman parameters are specified by the server and may\n be either ephemeral or contained within the server's certificate.\n\n9. Mandatory Cipher Suites\n\n In the absence of an application profile standard specifying\n otherwise, a TLS-compliant application MUST implement the cipher\n suite TLS_RSA_WITH_AES_128_CBC_SHA (see Appendix A.5 for the\n definition).\n\n10. Application Data Protocol\n\n Application data messages are carried by the record layer and are\n fragmented, compressed, and encrypted based on the current connection\n state. The messages are treated as transparent data to the record\n layer.\n\n11. Security Considerations\n\n Security issues are discussed throughout this memo, especially in\n Appendices D, E, and F.\n\n12. IANA Considerations\n\n This document uses several registries that were originally created in\n [TLS1.1]. IANA has updated these to reference this document. The\n registries and their allocation policies (unchanged from [TLS1.1])\n are listed below.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 65]\n\f\nRFC 5246 TLS August 2008\n\n\n - TLS ClientCertificateType Identifiers Registry: Future values in\n the range 0-63 (decimal) inclusive are assigned via Standards\n Action [RFC2434]. Values in the range 64-223 (decimal) inclusive\n are assigned via Specification Required [RFC2434]. Values from\n 224-255 (decimal) inclusive are reserved for Private Use\n [RFC2434].\n\n - TLS Cipher Suite Registry: Future values with the first byte in\n the range 0-191 (decimal) inclusive are assigned via Standards\n Action [RFC2434]. Values with the first byte in the range 192-254\n (decimal) are assigned via Specification Required [RFC2434].\n Values with the first byte 255 (decimal) are reserved for Private\n Use [RFC2434].\n\n - This document defines several new HMAC-SHA256-based cipher suites,\n whose values (in Appendix A.5) have been allocated from the TLS\n Cipher Suite registry.\n\n - TLS ContentType Registry: Future values are allocated via\n Standards Action [RFC2434].\n\n - TLS Alert Registry: Future values are allocated via Standards\n Action [RFC2434].\n\n - TLS HandshakeType Registry: Future values are allocated via\n Standards Action [RFC2434].\n\n This document also uses a registry originally created in [RFC4366].\n IANA has updated it to reference this document. The registry and its\n allocation policy (unchanged from [RFC4366]) is listed below:\n\n - TLS ExtensionType Registry: Future values are allocated via IETF\n Consensus [RFC2434]. IANA has updated this registry to include\n the signature_algorithms extension and its corresponding value\n (see Section 7.4.1.4).\n\n In addition, this document defines two new registries to be\n maintained by IANA:\n\n - TLS SignatureAlgorithm Registry: The registry has been initially\n populated with the values described in Section 7.4.1.4.1. Future\n values in the range 0-63 (decimal) inclusive are assigned via\n Standards Action [RFC2434]. Values in the range 64-223 (decimal)\n inclusive are assigned via Specification Required [RFC2434].\n Values from 224-255 (decimal) inclusive are reserved for Private\n Use [RFC2434].\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 66]\n\f\nRFC 5246 TLS August 2008\n\n\n - TLS HashAlgorithm Registry: The registry has been initially\n populated with the values described in Section 7.4.1.4.1. Future\n values in the range 0-63 (decimal) inclusive are assigned via\n Standards Action [RFC2434]. Values in the range 64-223 (decimal)\n inclusive are assigned via Specification Required [RFC2434].\n Values from 224-255 (decimal) inclusive are reserved for Private\n Use [RFC2434].\n\n This document also uses the TLS Compression Method Identifiers\n Registry, defined in [RFC3749]. IANA has allocated value 0 for\n the \"null\" compression method.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 67]\n\f\nRFC 5246 TLS August 2008\n\n\nAppendix A. Protocol Data Structures and Constant Values\n\n This section describes protocol types and constants.\n\nA.1. Record Layer\n\n struct {\n uint8 major;\n uint8 minor;\n } ProtocolVersion;\n\n ProtocolVersion version = { 3, 3 }; /* TLS v1.2*/\n\n enum {\n change_cipher_spec(20), alert(21), handshake(22),\n application_data(23), (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n opaque fragment[TLSCompressed.length];\n } TLSCompressed;\n\n struct {\n ContentType type;\n ProtocolVersion version;\n uint16 length;\n select (SecurityParameters.cipher_type) {\n case stream: GenericStreamCipher;\n case block: GenericBlockCipher;\n case aead: GenericAEADCipher;\n } fragment;\n } TLSCiphertext;\n\n stream-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n } GenericStreamCipher;\n\n\n\n\nDierks & Rescorla Standards Track [Page 68]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n opaque IV[SecurityParameters.record_iv_length];\n block-ciphered struct {\n opaque content[TLSCompressed.length];\n opaque MAC[SecurityParameters.mac_length];\n uint8 padding[GenericBlockCipher.padding_length];\n uint8 padding_length;\n };\n } GenericBlockCipher;\n\n struct {\n opaque nonce_explicit[SecurityParameters.record_iv_length];\n aead-ciphered struct {\n opaque content[TLSCompressed.length];\n };\n } GenericAEADCipher;\n\nA.2. Change Cipher Specs Message\n\n struct {\n enum { change_cipher_spec(1), (255) } type;\n } ChangeCipherSpec;\n\nA.3. Alert Messages\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction_RESERVED(60),\n protocol_version(70),\n\n\n\nDierks & Rescorla Standards Track [Page 69]\n\f\nRFC 5246 TLS August 2008\n\n\n insufficient_security(71),\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110), /* new */\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\nA.4. Handshake Protocol\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20)\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type;\n uint24 length;\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n } body;\n } Handshake;\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 70]\n\f\nRFC 5246 TLS August 2008\n\n\nA.4.1. Hello Messages\n\n struct { } HelloRequest;\n\n struct {\n uint32 gmt_unix_time;\n opaque random_bytes[28];\n } Random;\n\n opaque SessionID<0..32>;\n\n uint8 CipherSuite[2];\n\n enum { null(0), (255) } CompressionMethod;\n\n struct {\n ProtocolVersion client_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suites<2..2^16-2>;\n CompressionMethod compression_methods<1..2^8-1>;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ClientHello;\n\n struct {\n ProtocolVersion server_version;\n Random random;\n SessionID session_id;\n CipherSuite cipher_suite;\n CompressionMethod compression_method;\n select (extensions_present) {\n case false:\n struct {};\n case true:\n Extension extensions<0..2^16-1>;\n };\n } ServerHello;\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n\n\n\nDierks & Rescorla Standards Track [Page 71]\n\f\nRFC 5246 TLS August 2008\n\n\n enum {\n signature_algorithms(13), (65535)\n } ExtensionType;\n\n enum{\n none(0), md5(1), sha1(2), sha224(3), sha256(4), sha384(5),\n sha512(6), (255)\n } HashAlgorithm;\n enum {\n anonymous(0), rsa(1), dsa(2), ecdsa(3), (255)\n } SignatureAlgorithm;\n\n struct {\n HashAlgorithm hash;\n SignatureAlgorithm signature;\n } SignatureAndHashAlgorithm;\n\n SignatureAndHashAlgorithm\n supported_signature_algorithms<2..2^16-1>;\n\nA.4.2. Server Authentication and Key Exchange Messages\n\n opaque ASN.1Cert<2^24-1>;\n\n struct {\n ASN.1Cert certificate_list<0..2^24-1>;\n } Certificate;\n\n enum { dhe_dss, dhe_rsa, dh_anon, rsa,dh_dss, dh_rsa\n /* may be extended, e.g., for ECDH -- see [TLSECC] */\n } KeyExchangeAlgorithm;\n\n struct {\n opaque dh_p<1..2^16-1>;\n opaque dh_g<1..2^16-1>;\n opaque dh_Ys<1..2^16-1>;\n } ServerDHParams; /* Ephemeral DH parameters */\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 72]\n\f\nRFC 5246 TLS August 2008\n\n\n struct {\n select (KeyExchangeAlgorithm) {\n case dh_anon:\n ServerDHParams params;\n case dhe_dss:\n case dhe_rsa:\n ServerDHParams params;\n digitally-signed struct {\n opaque client_random[32];\n opaque server_random[32];\n ServerDHParams params;\n } signed_params;\n case rsa:\n case dh_dss:\n case dh_rsa:\n struct {} ;\n /* message is omitted for rsa, dh_dss, and dh_rsa */\n /* may be extended, e.g., for ECDH -- see [TLSECC] */\n } ServerKeyExchange;\n\n enum {\n rsa_sign(1), dss_sign(2), rsa_fixed_dh(3), dss_fixed_dh(4),\n rsa_ephemeral_dh_RESERVED(5), dss_ephemeral_dh_RESERVED(6),\n fortezza_dms_RESERVED(20),\n (255)\n } ClientCertificateType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n ClientCertificateType certificate_types<1..2^8-1>;\n DistinguishedName certificate_authorities<0..2^16-1>;\n } CertificateRequest;\n\n struct { } ServerHelloDone;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 73]\n\f\nRFC 5246 TLS August 2008\n\n\nA.4.3. Client Authentication and Key Exchange Messages\n\n struct {\n select (KeyExchangeAlgorithm) {\n case rsa:\n EncryptedPreMasterSecret;\n case dhe_dss:\n case dhe_rsa:\n case dh_dss:\n case dh_rsa:\n case dh_anon:\n ClientDiffieHellmanPublic;\n } exchange_keys;\n } ClientKeyExchange;\n\n struct {\n ProtocolVersion client_version;\n opaque random[46];\n } PreMasterSecret;\n\n struct {\n public-key-encrypted PreMasterSecret pre_master_secret;\n } EncryptedPreMasterSecret;\n\n enum { implicit, explicit } PublicValueEncoding;\n\n struct {\n select (PublicValueEncoding) {\n case implicit: struct {};\n case explicit: opaque DH_Yc<1..2^16-1>;\n } dh_public;\n } ClientDiffieHellmanPublic;\n\n struct {\n digitally-signed struct {\n opaque handshake_messages[handshake_messages_length];\n }\n } CertificateVerify;\n\nA.4.4. Handshake Finalization Message\n\n struct {\n opaque verify_data[verify_data_length];\n } Finished;\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 74]\n\f\nRFC 5246 TLS August 2008\n\n\nA.5. The Cipher Suite\n\n The following values define the cipher suite codes used in the\n ClientHello and ServerHello messages.\n\n A cipher suite defines a cipher specification supported in TLS\n Version 1.2.\n\n TLS_NULL_WITH_NULL_NULL is specified and is the initial state of a\n TLS connection during the first handshake on that channel, but MUST\n NOT be negotiated, as it provides no more protection than an\n unsecured connection.\n\n CipherSuite TLS_NULL_WITH_NULL_NULL = { 0x00,0x00 };\n\n The following CipherSuite definitions require that the server provide\n an RSA certificate that can be used for key exchange. The server may\n request any signature-capable certificate in the certificate request\n message.\n\n CipherSuite TLS_RSA_WITH_NULL_MD5 = { 0x00,0x01 };\n CipherSuite TLS_RSA_WITH_NULL_SHA = { 0x00,0x02 };\n CipherSuite TLS_RSA_WITH_NULL_SHA256 = { 0x00,0x3B };\n CipherSuite TLS_RSA_WITH_RC4_128_MD5 = { 0x00,0x04 };\n CipherSuite TLS_RSA_WITH_RC4_128_SHA = { 0x00,0x05 };\n CipherSuite TLS_RSA_WITH_3DES_EDE_CBC_SHA = { 0x00,0x0A };\n CipherSuite TLS_RSA_WITH_AES_128_CBC_SHA = { 0x00,0x2F };\n CipherSuite TLS_RSA_WITH_AES_256_CBC_SHA = { 0x00,0x35 };\n CipherSuite TLS_RSA_WITH_AES_128_CBC_SHA256 = { 0x00,0x3C };\n CipherSuite TLS_RSA_WITH_AES_256_CBC_SHA256 = { 0x00,0x3D };\n\n The following cipher suite definitions are used for server-\n authenticated (and optionally client-authenticated) Diffie-Hellman.\n DH denotes cipher suites in which the server's certificate contains\n the Diffie-Hellman parameters signed by the certificate authority\n (CA). DHE denotes ephemeral Diffie-Hellman, where the Diffie-Hellman\n parameters are signed by a signature-capable certificate, which has\n been signed by the CA. The signing algorithm used by the server is\n specified after the DHE component of the CipherSuite name. The\n server can request any signature-capable certificate from the client\n for client authentication, or it may request a Diffie-Hellman\n certificate. Any Diffie-Hellman certificate provided by the client\n must use the parameters (group and generator) described by the\n server.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 75]\n\f\nRFC 5246 TLS August 2008\n\n\n CipherSuite TLS_DH_DSS_WITH_3DES_EDE_CBC_SHA = { 0x00,0x0D };\n CipherSuite TLS_DH_RSA_WITH_3DES_EDE_CBC_SHA = { 0x00,0x10 };\n CipherSuite TLS_DHE_DSS_WITH_3DES_EDE_CBC_SHA = { 0x00,0x13 };\n CipherSuite TLS_DHE_RSA_WITH_3DES_EDE_CBC_SHA = { 0x00,0x16 };\n CipherSuite TLS_DH_DSS_WITH_AES_128_CBC_SHA = { 0x00,0x30 };\n CipherSuite TLS_DH_RSA_WITH_AES_128_CBC_SHA = { 0x00,0x31 };\n CipherSuite TLS_DHE_DSS_WITH_AES_128_CBC_SHA = { 0x00,0x32 };\n CipherSuite TLS_DHE_RSA_WITH_AES_128_CBC_SHA = { 0x00,0x33 };\n CipherSuite TLS_DH_DSS_WITH_AES_256_CBC_SHA = { 0x00,0x36 };\n CipherSuite TLS_DH_RSA_WITH_AES_256_CBC_SHA = { 0x00,0x37 };\n CipherSuite TLS_DHE_DSS_WITH_AES_256_CBC_SHA = { 0x00,0x38 };\n CipherSuite TLS_DHE_RSA_WITH_AES_256_CBC_SHA = { 0x00,0x39 };\n CipherSuite TLS_DH_DSS_WITH_AES_128_CBC_SHA256 = { 0x00,0x3E };\n CipherSuite TLS_DH_RSA_WITH_AES_128_CBC_SHA256 = { 0x00,0x3F };\n CipherSuite TLS_DHE_DSS_WITH_AES_128_CBC_SHA256 = { 0x00,0x40 };\n CipherSuite TLS_DHE_RSA_WITH_AES_128_CBC_SHA256 = { 0x00,0x67 };\n CipherSuite TLS_DH_DSS_WITH_AES_256_CBC_SHA256 = { 0x00,0x68 };\n CipherSuite TLS_DH_RSA_WITH_AES_256_CBC_SHA256 = { 0x00,0x69 };\n CipherSuite TLS_DHE_DSS_WITH_AES_256_CBC_SHA256 = { 0x00,0x6A };\n CipherSuite TLS_DHE_RSA_WITH_AES_256_CBC_SHA256 = { 0x00,0x6B };\n\n The following cipher suites are used for completely anonymous\n Diffie-Hellman communications in which neither party is\n authenticated. Note that this mode is vulnerable to man-in-the-\n middle attacks. Using this mode therefore is of limited use: These\n cipher suites MUST NOT be used by TLS 1.2 implementations unless the\n application layer has specifically requested to allow anonymous key\n exchange. (Anonymous key exchange may sometimes be acceptable, for\n example, to support opportunistic encryption when no set-up for\n authentication is in place, or when TLS is used as part of more\n complex security protocols that have other means to ensure\n authentication.)\n\n CipherSuite TLS_DH_anon_WITH_RC4_128_MD5 = { 0x00,0x18 };\n CipherSuite TLS_DH_anon_WITH_3DES_EDE_CBC_SHA = { 0x00,0x1B };\n CipherSuite TLS_DH_anon_WITH_AES_128_CBC_SHA = { 0x00,0x34 };\n CipherSuite TLS_DH_anon_WITH_AES_256_CBC_SHA = { 0x00,0x3A };\n CipherSuite TLS_DH_anon_WITH_AES_128_CBC_SHA256 = { 0x00,0x6C };\n CipherSuite TLS_DH_anon_WITH_AES_256_CBC_SHA256 = { 0x00,0x6D };\n\n Note that using non-anonymous key exchange without actually verifying\n the key exchange is essentially equivalent to anonymous key exchange,\n and the same precautions apply. While non-anonymous key exchange\n will generally involve a higher computational and communicational\n cost than anonymous key exchange, it may be in the interest of\n interoperability not to disable non-anonymous key exchange when the\n application layer is allowing anonymous key exchange.\n\n\n\n\nDierks & Rescorla Standards Track [Page 76]\n\f\nRFC 5246 TLS August 2008\n\n\n New cipher suite values have been assigned by IANA as described in\n Section 12.\n\n Note: The cipher suite values { 0x00, 0x1C } and { 0x00, 0x1D } are\n reserved to avoid collision with Fortezza-based cipher suites in\n SSL 3.\n\nA.6. The Security Parameters\n\n These security parameters are determined by the TLS Handshake\n Protocol and provided as parameters to the TLS record layer in order\n to initialize a connection state. SecurityParameters includes:\n\n enum { null(0), (255) } CompressionMethod;\n\n enum { server, client } ConnectionEnd;\n\n enum { tls_prf_sha256 } PRFAlgorithm;\n\n enum { null, rc4, 3des, aes } BulkCipherAlgorithm;\n\n enum { stream, block, aead } CipherType;\n\n enum { null, hmac_md5, hmac_sha1, hmac_sha256, hmac_sha384,\n hmac_sha512} MACAlgorithm;\n\n /* Other values may be added to the algorithms specified in\n CompressionMethod, PRFAlgorithm, BulkCipherAlgorithm, and\n MACAlgorithm. */\n\n struct {\n ConnectionEnd entity;\n PRFAlgorithm prf_algorithm;\n BulkCipherAlgorithm bulk_cipher_algorithm;\n CipherType cipher_type;\n uint8 enc_key_length;\n uint8 block_length;\n uint8 fixed_iv_length;\n uint8 record_iv_length;\n MACAlgorithm mac_algorithm;\n uint8 mac_length;\n uint8 mac_key_length;\n CompressionMethod compression_algorithm;\n opaque master_secret[48];\n opaque client_random[32];\n opaque server_random[32];\n } SecurityParameters;\n\n\n\n\nDierks & Rescorla Standards Track [Page 77]\n\f\nRFC 5246 TLS August 2008\n\n\nA.7. Changes to RFC 4492\n\n RFC 4492 [TLSECC] adds Elliptic Curve cipher suites to TLS. This\n document changes some of the structures used in that document. This\n section details the required changes for implementors of both RFC\n 4492 and TLS 1.2. Implementors of TLS 1.2 who are not implementing\n RFC 4492 do not need to read this section.\n\n This document adds a \"signature_algorithm\" field to the digitally-\n signed element in order to identify the signature and digest\n algorithms used to create a signature. This change applies to\n digital signatures formed using ECDSA as well, thus allowing ECDSA\n signatures to be used with digest algorithms other than SHA-1,\n provided such use is compatible with the certificate and any\n restrictions imposed by future revisions of [PKIX].\n\n As described in Sections 7.4.2 and 7.4.6, the restrictions on the\n signature algorithms used to sign certificates are no longer tied to\n the cipher suite (when used by the server) or the\n ClientCertificateType (when used by the client). Thus, the\n restrictions on the algorithm used to sign certificates specified in\n Sections 2 and 3 of RFC 4492 are also relaxed. As in this document,\n the restrictions on the keys in the end-entity certificate remain.\n\nAppendix B. Glossary\n\n Advanced Encryption Standard (AES)\n AES [AES] is a widely used symmetric encryption algorithm. AES is\n a block cipher with a 128-, 192-, or 256-bit keys and a 16-byte\n block size. TLS currently only supports the 128- and 256-bit key\n sizes.\n\n application protocol\n An application protocol is a protocol that normally layers\n directly on top of the transport layer (e.g., TCP/IP). Examples\n include HTTP, TELNET, FTP, and SMTP.\n\n asymmetric cipher\n See public key cryptography.\n\n authenticated encryption with additional data (AEAD)\n A symmetric encryption algorithm that simultaneously provides\n confidentiality and message integrity.\n\n authentication\n Authentication is the ability of one entity to determine the\n identity of another entity.\n\n\n\n\nDierks & Rescorla Standards Track [Page 78]\n\f\nRFC 5246 TLS August 2008\n\n\n block cipher\n A block cipher is an algorithm that operates on plaintext in\n groups of bits, called blocks. 64 bits was, and 128 bits is, a\n common block size.\n\n bulk cipher\n A symmetric encryption algorithm used to encrypt large quantities\n of data.\n\n cipher block chaining (CBC)\n CBC is a mode in which every plaintext block encrypted with a\n block cipher is first exclusive-ORed with the previous ciphertext\n block (or, in the case of the first block, with the initialization\n vector). For decryption, every block is first decrypted, then\n exclusive-ORed with the previous ciphertext block (or IV).\n\n certificate\n As part of the X.509 protocol (a.k.a. ISO Authentication\n framework), certificates are assigned by a trusted Certificate\n Authority and provide a strong binding between a party's identity\n or some other attributes and its public key.\n\n client\n The application entity that initiates a TLS connection to a\n server. This may or may not imply that the client initiated the\n underlying transport connection. The primary operational\n difference between the server and client is that the server is\n generally authenticated, while the client is only optionally\n authenticated.\n\n client write key\n The key used to encrypt data written by the client.\n\n client write MAC key\n The secret data used to authenticate data written by the client.\n\n connection\n A connection is a transport (in the OSI layering model definition)\n that provides a suitable type of service. For TLS, such\n connections are peer-to-peer relationships. The connections are\n transient. Every connection is associated with one session.\n\n Data Encryption Standard\n DES [DES] still is a very widely used symmetric encryption\n algorithm although it is considered as rather weak now. DES is a\n block cipher with a 56-bit key and an 8-byte block size. Note\n that in TLS, for key generation purposes, DES is treated as having\n an 8-byte key length (64 bits), but it still only provides 56 bits\n\n\n\nDierks & Rescorla Standards Track [Page 79]\n\f\nRFC 5246 TLS August 2008\n\n\n of protection. (The low bit of each key byte is presumed to be\n set to produce odd parity in that key byte.) DES can also be\n operated in a mode [3DES] where three independent keys and three\n encryptions are used for each block of data; this uses 168 bits of\n key (24 bytes in the TLS key generation method) and provides the\n equivalent of 112 bits of security.\n\n Digital Signature Standard (DSS)\n A standard for digital signing, including the Digital Signing\n Algorithm, approved by the National Institute of Standards and\n Technology, defined in NIST FIPS PUB 186-2, \"Digital Signature\n Standard\", published January 2000 by the U.S. Department of\n Commerce [DSS]. A significant update [DSS-3] has been drafted and\n was published in March 2006.\n\n digital signatures\n Digital signatures utilize public key cryptography and one-way\n hash functions to produce a signature of the data that can be\n authenticated, and is difficult to forge or repudiate.\n\n handshake An initial negotiation between client and server that\n establishes the parameters of their transactions.\n\n Initialization Vector (IV)\n When a block cipher is used in CBC mode, the initialization vector\n is exclusive-ORed with the first plaintext block prior to\n encryption.\n\n Message Authentication Code (MAC)\n A Message Authentication Code is a one-way hash computed from a\n message and some secret data. It is difficult to forge without\n knowing the secret data. Its purpose is to detect if the message\n has been altered.\n\n master secret\n Secure secret data used for generating encryption keys, MAC\n secrets, and IVs.\n\n MD5\n MD5 [MD5] is a hashing function that converts an arbitrarily long\n data stream into a hash of fixed size (16 bytes). Due to\n significant progress in cryptanalysis, at the time of publication\n of this document, MD5 no longer can be considered a 'secure'\n hashing function.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 80]\n\f\nRFC 5246 TLS August 2008\n\n\n public key cryptography\n A class of cryptographic techniques employing two-key ciphers.\n Messages encrypted with the public key can only be decrypted with\n the associated private key. Conversely, messages signed with the\n private key can be verified with the public key.\n\n one-way hash function\n A one-way transformation that converts an arbitrary amount of data\n into a fixed-length hash. It is computationally hard to reverse\n the transformation or to find collisions. MD5 and SHA are\n examples of one-way hash functions.\n\n RC4\n A stream cipher invented by Ron Rivest. A compatible cipher is\n described in [SCH].\n\n RSA\n A very widely used public key algorithm that can be used for\n either encryption or digital signing. [RSA]\n\n server\n The server is the application entity that responds to requests for\n connections from clients. See also \"client\".\n\n session\n A TLS session is an association between a client and a server.\n Sessions are created by the handshake protocol. Sessions define a\n set of cryptographic security parameters that can be shared among\n multiple connections. Sessions are used to avoid the expensive\n negotiation of new security parameters for each connection.\n\n session identifier\n A session identifier is a value generated by a server that\n identifies a particular session.\n\n server write key\n The key used to encrypt data written by the server.\n\n server write MAC key\n The secret data used to authenticate data written by the server.\n\n SHA\n The Secure Hash Algorithm [SHS] is defined in FIPS PUB 180-2. It\n produces a 20-byte output. Note that all references to SHA\n (without a numerical suffix) actually use the modified SHA-1\n algorithm.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 81]\n\f\nRFC 5246 TLS August 2008\n\n\n SHA-256\n The 256-bit Secure Hash Algorithm is defined in FIPS PUB 180-2.\n It produces a 32-byte output.\n\n SSL\n Netscape's Secure Socket Layer protocol [SSL3]. TLS is based on\n SSL Version 3.0.\n\n stream cipher\n An encryption algorithm that converts a key into a\n cryptographically strong keystream, which is then exclusive-ORed\n with the plaintext.\n\n symmetric cipher\n See bulk cipher.\n\n Transport Layer Security (TLS)\n This protocol; also, the Transport Layer Security working group of\n the Internet Engineering Task Force (IETF). See \"Working Group\n Information\" at the end of this document (see page 99).\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 82]\n\f\nRFC 5246 TLS August 2008\n\n\nAppendix C. Cipher Suite Definitions\n\nCipher Suite Key Cipher Mac\n Exchange\n\nTLS_NULL_WITH_NULL_NULL NULL NULL NULL\nTLS_RSA_WITH_NULL_MD5 RSA NULL MD5\nTLS_RSA_WITH_NULL_SHA RSA NULL SHA\nTLS_RSA_WITH_NULL_SHA256 RSA NULL SHA256\nTLS_RSA_WITH_RC4_128_MD5 RSA RC4_128 MD5\nTLS_RSA_WITH_RC4_128_SHA RSA RC4_128 SHA\nTLS_RSA_WITH_3DES_EDE_CBC_SHA RSA 3DES_EDE_CBC SHA\nTLS_RSA_WITH_AES_128_CBC_SHA RSA AES_128_CBC SHA\nTLS_RSA_WITH_AES_256_CBC_SHA RSA AES_256_CBC SHA\nTLS_RSA_WITH_AES_128_CBC_SHA256 RSA AES_128_CBC SHA256\nTLS_RSA_WITH_AES_256_CBC_SHA256 RSA AES_256_CBC SHA256\nTLS_DH_DSS_WITH_3DES_EDE_CBC_SHA DH_DSS 3DES_EDE_CBC SHA\nTLS_DH_RSA_WITH_3DES_EDE_CBC_SHA DH_RSA 3DES_EDE_CBC SHA\nTLS_DHE_DSS_WITH_3DES_EDE_CBC_SHA DHE_DSS 3DES_EDE_CBC SHA\nTLS_DHE_RSA_WITH_3DES_EDE_CBC_SHA DHE_RSA 3DES_EDE_CBC SHA\nTLS_DH_anon_WITH_RC4_128_MD5 DH_anon RC4_128 MD5\nTLS_DH_anon_WITH_3DES_EDE_CBC_SHA DH_anon 3DES_EDE_CBC SHA\nTLS_DH_DSS_WITH_AES_128_CBC_SHA DH_DSS AES_128_CBC SHA\nTLS_DH_RSA_WITH_AES_128_CBC_SHA DH_RSA AES_128_CBC SHA\nTLS_DHE_DSS_WITH_AES_128_CBC_SHA DHE_DSS AES_128_CBC SHA\nTLS_DHE_RSA_WITH_AES_128_CBC_SHA DHE_RSA AES_128_CBC SHA\nTLS_DH_anon_WITH_AES_128_CBC_SHA DH_anon AES_128_CBC SHA\nTLS_DH_DSS_WITH_AES_256_CBC_SHA DH_DSS AES_256_CBC SHA\nTLS_DH_RSA_WITH_AES_256_CBC_SHA DH_RSA AES_256_CBC SHA\nTLS_DHE_DSS_WITH_AES_256_CBC_SHA DHE_DSS AES_256_CBC SHA\nTLS_DHE_RSA_WITH_AES_256_CBC_SHA DHE_RSA AES_256_CBC SHA\nTLS_DH_anon_WITH_AES_256_CBC_SHA DH_anon AES_256_CBC SHA\nTLS_DH_DSS_WITH_AES_128_CBC_SHA256 DH_DSS AES_128_CBC SHA256\nTLS_DH_RSA_WITH_AES_128_CBC_SHA256 DH_RSA AES_128_CBC SHA256\nTLS_DHE_DSS_WITH_AES_128_CBC_SHA256 DHE_DSS AES_128_CBC SHA256\nTLS_DHE_RSA_WITH_AES_128_CBC_SHA256 DHE_RSA AES_128_CBC SHA256\nTLS_DH_anon_WITH_AES_128_CBC_SHA256 DH_anon AES_128_CBC SHA256\nTLS_DH_DSS_WITH_AES_256_CBC_SHA256 DH_DSS AES_256_CBC SHA256\nTLS_DH_RSA_WITH_AES_256_CBC_SHA256 DH_RSA AES_256_CBC SHA256\nTLS_DHE_DSS_WITH_AES_256_CBC_SHA256 DHE_DSS AES_256_CBC SHA256\nTLS_DHE_RSA_WITH_AES_256_CBC_SHA256 DHE_RSA AES_256_CBC SHA256\nTLS_DH_anon_WITH_AES_256_CBC_SHA256 DH_anon AES_256_CBC SHA256\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 83]\n\f\nRFC 5246 TLS August 2008\n\n\n Key IV Block\nCipher Type Material Size Size\n------------ ------ -------- ---- -----\nNULL Stream 0 0 N/A\nRC4_128 Stream 16 0 N/A\n3DES_EDE_CBC Block 24 8 8\nAES_128_CBC Block 16 16 16\nAES_256_CBC Block 32 16 16\n\n\nMAC Algorithm mac_length mac_key_length\n-------- ----------- ---------- --------------\nNULL N/A 0 0\nMD5 HMAC-MD5 16 16\nSHA HMAC-SHA1 20 20\nSHA256 HMAC-SHA256 32 32\n\n Type\n Indicates whether this is a stream cipher or a block cipher\n running in CBC mode.\n\n Key Material\n The number of bytes from the key_block that are used for\n generating the write keys.\n\n IV Size\n The amount of data needed to be generated for the initialization\n vector. Zero for stream ciphers; equal to the block size for\n block ciphers (this is equal to\n SecurityParameters.record_iv_length).\n\n Block Size\n The amount of data a block cipher enciphers in one chunk; a block\n cipher running in CBC mode can only encrypt an even multiple of\n its block size.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 84]\n\f\nRFC 5246 TLS August 2008\n\n\nAppendix D. Implementation Notes\n\n The TLS protocol cannot prevent many common security mistakes. This\n section provides several recommendations to assist implementors.\n\nD.1. Random Number Generation and Seeding\n\n TLS requires a cryptographically secure pseudorandom number generator\n (PRNG). Care must be taken in designing and seeding PRNGs. PRNGs\n based on secure hash operations, most notably SHA-1, are acceptable,\n but cannot provide more security than the size of the random number\n generator state.\n\n To estimate the amount of seed material being produced, add the\n number of bits of unpredictable information in each seed byte. For\n example, keystroke timing values taken from a PC compatible's 18.2 Hz\n timer provide 1 or 2 secure bits each, even though the total size of\n the counter value is 16 bits or more. Seeding a 128-bit PRNG would\n thus require approximately 100 such timer values.\n\n [RANDOM] provides guidance on the generation of random values.\n\nD.2. Certificates and Authentication\n\n Implementations are responsible for verifying the integrity of\n certificates and should generally support certificate revocation\n messages. Certificates should always be verified to ensure proper\n signing by a trusted Certificate Authority (CA). The selection and\n addition of trusted CAs should be done very carefully. Users should\n be able to view information about the certificate and root CA.\n\nD.3. Cipher Suites\n\n TLS supports a range of key sizes and security levels, including some\n that provide no or minimal security. A proper implementation will\n probably not support many cipher suites. For instance, anonymous\n Diffie-Hellman is strongly discouraged because it cannot prevent man-\n in-the-middle attacks. Applications should also enforce minimum and\n maximum key sizes. For example, certificate chains containing 512-\n bit RSA keys or signatures are not appropriate for high-security\n applications.\n\nD.4. Implementation Pitfalls\n\n Implementation experience has shown that certain parts of earlier TLS\n specifications are not easy to understand, and have been a source of\n interoperability and security problems. Many of these areas have\n\n\n\n\nDierks & Rescorla Standards Track [Page 85]\n\f\nRFC 5246 TLS August 2008\n\n\n been clarified in this document, but this appendix contains a short\n list of the most important things that require special attention from\n implementors.\n\n TLS protocol issues:\n\n - Do you correctly handle handshake messages that are fragmented to\n multiple TLS records (see Section 6.2.1)? Including corner cases\n like a ClientHello that is split to several small fragments? Do\n you fragment handshake messages that exceed the maximum fragment\n size? In particular, the certificate and certificate request\n handshake messages can be large enough to require fragmentation.\n\n - Do you ignore the TLS record layer version number in all TLS\n records before ServerHello (see Appendix E.1)?\n\n - Do you handle TLS extensions in ClientHello correctly, including\n omitting the extensions field completely?\n\n - Do you support renegotiation, both client and server initiated?\n While renegotiation is an optional feature, supporting it is\n highly recommended.\n\n - When the server has requested a client certificate, but no\n suitable certificate is available, do you correctly send an empty\n Certificate message, instead of omitting the whole message (see\n Section 7.4.6)?\n\n Cryptographic details:\n\n - In the RSA-encrypted Premaster Secret, do you correctly send and\n verify the version number? When an error is encountered, do you\n continue the handshake to avoid the Bleichenbacher attack (see\n Section 7.4.7.1)?\n\n - What countermeasures do you use to prevent timing attacks against\n RSA decryption and signing operations (see Section 7.4.7.1)?\n\n - When verifying RSA signatures, do you accept both NULL and missing\n parameters (see Section 4.7)? Do you verify that the RSA padding\n doesn't have additional data after the hash value? [FI06]\n\n - When using Diffie-Hellman key exchange, do you correctly strip\n leading zero bytes from the negotiated key (see Section 8.1.2)?\n\n - Does your TLS client check that the Diffie-Hellman parameters sent\n by the server are acceptable (see Section F.1.1.3)?\n\n\n\n\nDierks & Rescorla Standards Track [Page 86]\n\f\nRFC 5246 TLS August 2008\n\n\n - How do you generate unpredictable IVs for CBC mode ciphers (see\n Section 6.2.3.2)?\n\n - Do you accept long CBC mode padding (up to 255 bytes; see Section\n 6.2.3.2)?\n\n - How do you address CBC mode timing attacks (Section 6.2.3.2)?\n\n - Do you use a strong and, most importantly, properly seeded random\n number generator (see Appendix D.1) for generating the premaster\n secret (for RSA key exchange), Diffie-Hellman private values, the\n DSA \"k\" parameter, and other security-critical values?\n\nAppendix E. Backward Compatibility\n\nE.1. Compatibility with TLS 1.0/1.1 and SSL 3.0\n\n Since there are various versions of TLS (1.0, 1.1, 1.2, and any\n future versions) and SSL (2.0 and 3.0), means are needed to negotiate\n the specific protocol version to use. The TLS protocol provides a\n built-in mechanism for version negotiation so as not to bother other\n protocol components with the complexities of version selection.\n\n TLS versions 1.0, 1.1, and 1.2, and SSL 3.0 are very similar, and use\n compatible ClientHello messages; thus, supporting all of them is\n relatively easy. Similarly, servers can easily handle clients trying\n to use future versions of TLS as long as the ClientHello format\n remains compatible, and the client supports the highest protocol\n version available in the server.\n\n A TLS 1.2 client who wishes to negotiate with such older servers will\n send a normal TLS 1.2 ClientHello, containing { 3, 3 } (TLS 1.2) in\n ClientHello.client_version. If the server does not support this\n version, it will respond with a ServerHello containing an older\n version number. If the client agrees to use this version, the\n negotiation will proceed as appropriate for the negotiated protocol.\n\n If the version chosen by the server is not supported by the client\n (or not acceptable), the client MUST send a \"protocol_version\" alert\n message and close the connection.\n\n If a TLS server receives a ClientHello containing a version number\n greater than the highest version supported by the server, it MUST\n reply according to the highest version supported by the server.\n\n A TLS server can also receive a ClientHello containing a version\n number smaller than the highest supported version. If the server\n wishes to negotiate with old clients, it will proceed as appropriate\n\n\n\nDierks & Rescorla Standards Track [Page 87]\n\f\nRFC 5246 TLS August 2008\n\n\n for the highest version supported by the server that is not greater\n than ClientHello.client_version. For example, if the server supports\n TLS 1.0, 1.1, and 1.2, and client_version is TLS 1.0, the server will\n proceed with a TLS 1.0 ServerHello. If server supports (or is\n willing to use) only versions greater than client_version, it MUST\n send a \"protocol_version\" alert message and close the connection.\n\n Whenever a client already knows the highest protocol version known to\n a server (for example, when resuming a session), it SHOULD initiate\n the connection in that native protocol.\n\n Note: some server implementations are known to implement version\n negotiation incorrectly. For example, there are buggy TLS 1.0\n servers that simply close the connection when the client offers a\n version newer than TLS 1.0. Also, it is known that some servers will\n refuse the connection if any TLS extensions are included in\n ClientHello. Interoperability with such buggy servers is a complex\n topic beyond the scope of this document, and may require multiple\n connection attempts by the client.\n\n Earlier versions of the TLS specification were not fully clear on\n what the record layer version number (TLSPlaintext.version) should\n contain when sending ClientHello (i.e., before it is known which\n version of the protocol will be employed). Thus, TLS servers\n compliant with this specification MUST accept any value {03,XX} as\n the record layer version number for ClientHello.\n\n TLS clients that wish to negotiate with older servers MAY send any\n value {03,XX} as the record layer version number. Typical values\n would be {03,00}, the lowest version number supported by the client,\n and the value of ClientHello.client_version. No single value will\n guarantee interoperability with all old servers, but this is a\n complex topic beyond the scope of this document.\n\nE.2. Compatibility with SSL 2.0\n\n TLS 1.2 clients that wish to support SSL 2.0 servers MUST send\n version 2.0 CLIENT-HELLO messages defined in [SSL2]. The message\n MUST contain the same version number as would be used for ordinary\n ClientHello, and MUST encode the supported TLS cipher suites in the\n CIPHER-SPECS-DATA field as described below.\n\n Warning: The ability to send version 2.0 CLIENT-HELLO messages will\n be phased out with all due haste, since the newer ClientHello format\n provides better mechanisms for moving to newer versions and\n negotiating extensions. TLS 1.2 clients SHOULD NOT support SSL 2.0.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 88]\n\f\nRFC 5246 TLS August 2008\n\n\n However, even TLS servers that do not support SSL 2.0 MAY accept\n version 2.0 CLIENT-HELLO messages. The message is presented below in\n sufficient detail for TLS server implementors; the true definition is\n still assumed to be [SSL2].\n\n For negotiation purposes, 2.0 CLIENT-HELLO is interpreted the same\n way as a ClientHello with a \"null\" compression method and no\n extensions. Note that this message MUST be sent directly on the\n wire, not wrapped as a TLS record. For the purposes of calculating\n Finished and CertificateVerify, the msg_length field is not\n considered to be a part of the handshake message.\n\n uint8 V2CipherSpec[3];\n struct {\n uint16 msg_length;\n uint8 msg_type;\n Version version;\n uint16 cipher_spec_length;\n uint16 session_id_length;\n uint16 challenge_length;\n V2CipherSpec cipher_specs[V2ClientHello.cipher_spec_length];\n opaque session_id[V2ClientHello.session_id_length];\n opaque challenge[V2ClientHello.challenge_length;\n } V2ClientHello;\n\n msg_length\n The highest bit MUST be 1; the remaining bits contain the length\n of the following data in bytes.\n\n msg_type\n This field, in conjunction with the version field, identifies a\n version 2 ClientHello message. The value MUST be 1.\n\n version\n Equal to ClientHello.client_version.\n\n cipher_spec_length\n This field is the total length of the field cipher_specs. It\n cannot be zero and MUST be a multiple of the V2CipherSpec length\n (3).\n\n session_id_length\n This field MUST have a value of zero for a client that claims to\n support TLS 1.2.\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 89]\n\f\nRFC 5246 TLS August 2008\n\n\n challenge_length\n The length in bytes of the client's challenge to the server to\n authenticate itself. Historically, permissible values are between\n 16 and 32 bytes inclusive. When using the SSLv2 backward-\n compatible handshake the client SHOULD use a 32-byte challenge.\n\n cipher_specs\n This is a list of all CipherSpecs the client is willing and able\n to use. In addition to the 2.0 cipher specs defined in [SSL2],\n this includes the TLS cipher suites normally sent in\n ClientHello.cipher_suites, with each cipher suite prefixed by a\n zero byte. For example, the TLS cipher suite {0x00,0x0A} would be\n sent as {0x00,0x00,0x0A}.\n\n session_id\n This field MUST be empty.\n\n challenge\n Corresponds to ClientHello.random. If the challenge length is\n less than 32, the TLS server will pad the data with leading (note:\n not trailing) zero bytes to make it 32 bytes long.\n\n Note: Requests to resume a TLS session MUST use a TLS client hello.\n\nE.3. Avoiding Man-in-the-Middle Version Rollback\n\n When TLS clients fall back to Version 2.0 compatibility mode, they\n MUST use special PKCS#1 block formatting. This is done so that TLS\n servers will reject Version 2.0 sessions with TLS-capable clients.\n\n When a client negotiates SSL 2.0 but also supports TLS, it MUST set\n the right-hand (least-significant) 8 random bytes of the PKCS padding\n (not including the terminal null of the padding) for the RSA\n encryption of the ENCRYPTED-KEY-DATA field of the CLIENT-MASTER-KEY\n to 0x03 (the other padding bytes are random).\n\n When a TLS-capable server negotiates SSL 2.0 it SHOULD, after\n decrypting the ENCRYPTED-KEY-DATA field, check that these 8 padding\n bytes are 0x03. If they are not, the server SHOULD generate a random\n value for SECRET-KEY-DATA, and continue the handshake (which will\n eventually fail since the keys will not match). Note that reporting\n the error situation to the client could make the server vulnerable to\n attacks described in [BLEI].\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 90]\n\f\nRFC 5246 TLS August 2008\n\n\nAppendix F. Security Analysis\n\n The TLS protocol is designed to establish a secure connection between\n a client and a server communicating over an insecure channel. This\n document makes several traditional assumptions, including that\n attackers have substantial computational resources and cannot obtain\n secret information from sources outside the protocol. Attackers are\n assumed to have the ability to capture, modify, delete, replay, and\n otherwise tamper with messages sent over the communication channel.\n This appendix outlines how TLS has been designed to resist a variety\n of attacks.\n\nF.1. Handshake Protocol\n\n The handshake protocol is responsible for selecting a cipher spec and\n generating a master secret, which together comprise the primary\n cryptographic parameters associated with a secure session. The\n handshake protocol can also optionally authenticate parties who have\n certificates signed by a trusted certificate authority.\n\nF.1.1. Authentication and Key Exchange\n\n TLS supports three authentication modes: authentication of both\n parties, server authentication with an unauthenticated client, and\n total anonymity. Whenever the server is authenticated, the channel\n is secure against man-in-the-middle attacks, but completely anonymous\n sessions are inherently vulnerable to such attacks. Anonymous\n servers cannot authenticate clients. If the server is authenticated,\n its certificate message must provide a valid certificate chain\n leading to an acceptable certificate authority. Similarly,\n authenticated clients must supply an acceptable certificate to the\n server. Each party is responsible for verifying that the other's\n certificate is valid and has not expired or been revoked.\n\n The general goal of the key exchange process is to create a\n pre_master_secret known to the communicating parties and not to\n attackers. The pre_master_secret will be used to generate the\n master_secret (see Section 8.1). The master_secret is required to\n generate the Finished messages, encryption keys, and MAC keys (see\n Sections 7.4.9 and 6.3). By sending a correct Finished message,\n parties thus prove that they know the correct pre_master_secret.\n\nF.1.1.1. Anonymous Key Exchange\n\n Completely anonymous sessions can be established using Diffie-Hellman\n for key exchange. The server's public parameters are contained in\n the server key exchange message, and the client's are sent in the\n\n\n\n\nDierks & Rescorla Standards Track [Page 91]\n\f\nRFC 5246 TLS August 2008\n\n\n client key exchange message. Eavesdroppers who do not know the\n private values should not be able to find the Diffie-Hellman result\n (i.e., the pre_master_secret).\n\n Warning: Completely anonymous connections only provide protection\n against passive eavesdropping. Unless an independent tamper-proof\n channel is used to verify that the Finished messages were not\n replaced by an attacker, server authentication is required in\n environments where active man-in-the-middle attacks are a concern.\n\nF.1.1.2. RSA Key Exchange and Authentication\n\n With RSA, key exchange and server authentication are combined. The\n public key is contained in the server's certificate. Note that\n compromise of the server's static RSA key results in a loss of\n confidentiality for all sessions protected under that static key.\n TLS users desiring Perfect Forward Secrecy should use DHE cipher\n suites. The damage done by exposure of a private key can be limited\n by changing one's private key (and certificate) frequently.\n\n After verifying the server's certificate, the client encrypts a\n pre_master_secret with the server's public key. By successfully\n decoding the pre_master_secret and producing a correct Finished\n message, the server demonstrates that it knows the private key\n corresponding to the server certificate.\n\n When RSA is used for key exchange, clients are authenticated using\n the certificate verify message (see Section 7.4.8). The client signs\n a value derived from all preceding handshake messages. These\n handshake messages include the server certificate, which binds the\n signature to the server, and ServerHello.random, which binds the\n signature to the current handshake process.\n\nF.1.1.3. Diffie-Hellman Key Exchange with Authentication\n\n When Diffie-Hellman key exchange is used, the server can either\n supply a certificate containing fixed Diffie-Hellman parameters or\n use the server key exchange message to send a set of temporary\n Diffie-Hellman parameters signed with a DSA or RSA certificate.\n Temporary parameters are hashed with the hello.random values before\n signing to ensure that attackers do not replay old parameters. In\n either case, the client can verify the certificate or signature to\n ensure that the parameters belong to the server.\n\n If the client has a certificate containing fixed Diffie-Hellman\n parameters, its certificate contains the information required to\n complete the key exchange. Note that in this case the client and\n server will generate the same Diffie-Hellman result (i.e.,\n\n\n\nDierks & Rescorla Standards Track [Page 92]\n\f\nRFC 5246 TLS August 2008\n\n\n pre_master_secret) every time they communicate. To prevent the\n pre_master_secret from staying in memory any longer than necessary,\n it should be converted into the master_secret as soon as possible.\n Client Diffie-Hellman parameters must be compatible with those\n supplied by the server for the key exchange to work.\n\n If the client has a standard DSA or RSA certificate or is\n unauthenticated, it sends a set of temporary parameters to the server\n in the client key exchange message, then optionally uses a\n certificate verify message to authenticate itself.\n\n If the same DH keypair is to be used for multiple handshakes, either\n because the client or server has a certificate containing a fixed DH\n keypair or because the server is reusing DH keys, care must be taken\n to prevent small subgroup attacks. Implementations SHOULD follow the\n guidelines found in [SUBGROUP].\n\n Small subgroup attacks are most easily avoided by using one of the\n DHE cipher suites and generating a fresh DH private key (X) for each\n handshake. If a suitable base (such as 2) is chosen, g^X mod p can\n be computed very quickly; therefore, the performance cost is\n minimized. Additionally, using a fresh key for each handshake\n provides Perfect Forward Secrecy. Implementations SHOULD generate a\n new X for each handshake when using DHE cipher suites.\n\n Because TLS allows the server to provide arbitrary DH groups, the\n client should verify that the DH group is of suitable size as defined\n by local policy. The client SHOULD also verify that the DH public\n exponent appears to be of adequate size. [KEYSIZ] provides a useful\n guide to the strength of various group sizes. The server MAY choose\n to assist the client by providing a known group, such as those\n defined in [IKEALG] or [MODP]. These can be verified by simple\n comparison.\n\nF.1.2. Version Rollback Attacks\n\n Because TLS includes substantial improvements over SSL Version 2.0,\n attackers may try to make TLS-capable clients and servers fall back\n to Version 2.0. This attack can occur if (and only if) two TLS-\n capable parties use an SSL 2.0 handshake.\n\n Although the solution using non-random PKCS #1 block type 2 message\n padding is inelegant, it provides a reasonably secure way for Version\n 3.0 servers to detect the attack. This solution is not secure\n against attackers who can brute-force the key and substitute a new\n ENCRYPTED-KEY-DATA message containing the same key (but with normal\n padding) before the application-specified wait threshold has expired.\n Altering the padding of the least-significant 8 bytes of the PKCS\n\n\n\nDierks & Rescorla Standards Track [Page 93]\n\f\nRFC 5246 TLS August 2008\n\n\n padding does not impact security for the size of the signed hashes\n and RSA key lengths used in the protocol, since this is essentially\n equivalent to increasing the input block size by 8 bytes.\n\nF.1.3. Detecting Attacks Against the Handshake Protocol\n\n An attacker might try to influence the handshake exchange to make the\n parties select different encryption algorithms than they would\n normally choose.\n\n For this attack, an attacker must actively change one or more\n handshake messages. If this occurs, the client and server will\n compute different values for the handshake message hashes. As a\n result, the parties will not accept each others' Finished messages.\n Without the master_secret, the attacker cannot repair the Finished\n messages, so the attack will be discovered.\n\nF.1.4. Resuming Sessions\n\n When a connection is established by resuming a session, new\n ClientHello.random and ServerHello.random values are hashed with the\n session's master_secret. Provided that the master_secret has not\n been compromised and that the secure hash operations used to produce\n the encryption keys and MAC keys are secure, the connection should be\n secure and effectively independent from previous connections.\n Attackers cannot use known encryption keys or MAC secrets to\n compromise the master_secret without breaking the secure hash\n operations.\n\n Sessions cannot be resumed unless both the client and server agree.\n If either party suspects that the session may have been compromised,\n or that certificates may have expired or been revoked, it should\n force a full handshake. An upper limit of 24 hours is suggested for\n session ID lifetimes, since an attacker who obtains a master_secret\n may be able to impersonate the compromised party until the\n corresponding session ID is retired. Applications that may be run in\n relatively insecure environments should not write session IDs to\n stable storage.\n\nF.2. Protecting Application Data\n\n The master_secret is hashed with the ClientHello.random and\n ServerHello.random to produce unique data encryption keys and MAC\n secrets for each connection.\n\n Outgoing data is protected with a MAC before transmission. To\n prevent message replay or modification attacks, the MAC is computed\n from the MAC key, the sequence number, the message length, the\n\n\n\nDierks & Rescorla Standards Track [Page 94]\n\f\nRFC 5246 TLS August 2008\n\n\n message contents, and two fixed character strings. The message type\n field is necessary to ensure that messages intended for one TLS\n record layer client are not redirected to another. The sequence\n number ensures that attempts to delete or reorder messages will be\n detected. Since sequence numbers are 64 bits long, they should never\n overflow. Messages from one party cannot be inserted into the\n other's output, since they use independent MAC keys. Similarly, the\n server write and client write keys are independent, so stream cipher\n keys are used only once.\n\n If an attacker does break an encryption key, all messages encrypted\n with it can be read. Similarly, compromise of a MAC key can make\n message-modification attacks possible. Because MACs are also\n encrypted, message-alteration attacks generally require breaking the\n encryption algorithm as well as the MAC.\n\n Note: MAC keys may be larger than encryption keys, so messages can\n remain tamper resistant even if encryption keys are broken.\n\nF.3. Explicit IVs\n\n [CBCATT] describes a chosen plaintext attack on TLS that depends on\n knowing the IV for a record. Previous versions of TLS [TLS1.0] used\n the CBC residue of the previous record as the IV and therefore\n enabled this attack. This version uses an explicit IV in order to\n protect against this attack.\n\nF.4. Security of Composite Cipher Modes\n\n TLS secures transmitted application data via the use of symmetric\n encryption and authentication functions defined in the negotiated\n cipher suite. The objective is to protect both the integrity and\n confidentiality of the transmitted data from malicious actions by\n active attackers in the network. It turns out that the order in\n which encryption and authentication functions are applied to the data\n plays an important role for achieving this goal [ENCAUTH].\n\n The most robust method, called encrypt-then-authenticate, first\n applies encryption to the data and then applies a MAC to the\n ciphertext. This method ensures that the integrity and\n confidentiality goals are obtained with ANY pair of encryption and\n MAC functions, provided that the former is secure against chosen\n plaintext attacks and that the MAC is secure against chosen-message\n attacks. TLS uses another method, called authenticate-then-encrypt,\n in which first a MAC is computed on the plaintext and then the\n concatenation of plaintext and MAC is encrypted. This method has\n been proven secure for CERTAIN combinations of encryption functions\n and MAC functions, but it is not guaranteed to be secure in general.\n\n\n\nDierks & Rescorla Standards Track [Page 95]\n\f\nRFC 5246 TLS August 2008\n\n\n In particular, it has been shown that there exist perfectly secure\n encryption functions (secure even in the information-theoretic sense)\n that combined with any secure MAC function, fail to provide the\n confidentiality goal against an active attack. Therefore, new cipher\n suites and operation modes adopted into TLS need to be analyzed under\n the authenticate-then-encrypt method to verify that they achieve the\n stated integrity and confidentiality goals.\n\n Currently, the security of the authenticate-then-encrypt method has\n been proven for some important cases. One is the case of stream\n ciphers in which a computationally unpredictable pad of the length of\n the message, plus the length of the MAC tag, is produced using a\n pseudorandom generator and this pad is exclusive-ORed with the\n concatenation of plaintext and MAC tag. The other is the case of CBC\n mode using a secure block cipher. In this case, security can be\n shown if one applies one CBC encryption pass to the concatenation of\n plaintext and MAC and uses a new, independent, and unpredictable IV\n for each new pair of plaintext and MAC. In versions of TLS prior to\n 1.1, CBC mode was used properly EXCEPT that it used a predictable IV\n in the form of the last block of the previous ciphertext. This made\n TLS open to chosen plaintext attacks. This version of the protocol\n is immune to those attacks. For exact details in the encryption\n modes proven secure, see [ENCAUTH].\n\nF.5. Denial of Service\n\n TLS is susceptible to a number of denial-of-service (DoS) attacks.\n In particular, an attacker who initiates a large number of TCP\n connections can cause a server to consume large amounts of CPU for\n doing RSA decryption. However, because TLS is generally used over\n TCP, it is difficult for the attacker to hide his point of origin if\n proper TCP SYN randomization is used [SEQNUM] by the TCP stack.\n\n Because TLS runs over TCP, it is also susceptible to a number of DoS\n attacks on individual connections. In particular, attackers can\n forge RSTs, thereby terminating connections, or forge partial TLS\n records, thereby causing the connection to stall. These attacks\n cannot in general be defended against by a TCP-using protocol.\n Implementors or users who are concerned with this class of attack\n should use IPsec AH [AH] or ESP [ESP].\n\nF.6. Final Notes\n\n For TLS to be able to provide a secure connection, both the client\n and server systems, keys, and applications must be secure. In\n addition, the implementation must be free of security errors.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 96]\n\f\nRFC 5246 TLS August 2008\n\n\n The system is only as strong as the weakest key exchange and\n authentication algorithm supported, and only trustworthy\n cryptographic functions should be used. Short public keys and\n anonymous servers should be used with great caution. Implementations\n and users must be careful when deciding which certificates and\n certificate authorities are acceptable; a dishonest certificate\n authority can do tremendous damage.\n\nNormative References\n\n [AES] National Institute of Standards and Technology,\n \"Specification for the Advanced Encryption Standard (AES)\"\n FIPS 197. November 26, 2001.\n\n [3DES] National Institute of Standards and Technology,\n \"Recommendation for the Triple Data Encryption Algorithm\n (TDEA) Block Cipher\", NIST Special Publication 800-67, May\n 2004.\n\n [DSS] NIST FIPS PUB 186-2, \"Digital Signature Standard\",\n National Institute of Standards and Technology, U.S.\n Department of Commerce, 2000.\n\n [HMAC] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC: Keyed-\n Hashing for Message Authentication\", RFC 2104, February\n 1997.\n\n [MD5] Rivest, R., \"The MD5 Message-Digest Algorithm\", RFC 1321,\n April 1992.\n\n [PKCS1] Jonsson, J. and B. Kaliski, \"Public-Key Cryptography\n Standards (PKCS) #1: RSA Cryptography Specifications\n Version 2.1\", RFC 3447, February 2003.\n\n [PKIX] Housley, R., Polk, W., Ford, W., and D. Solo, \"Internet\n X.509 Public Key Infrastructure Certificate and\n Certificate Revocation List (CRL) Profile\", RFC 3280,\n April 2002.\n\n [SCH] B. Schneier. \"Applied Cryptography: Protocols, Algorithms,\n and Source Code in C, 2nd ed.\", Published by John Wiley &\n Sons, Inc. 1996.\n\n [SHS] NIST FIPS PUB 180-2, \"Secure Hash Standard\", National\n Institute of Standards and Technology, U.S. Department of\n Commerce, August 2002.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 97]\n\f\nRFC 5246 TLS August 2008\n\n\n [REQ] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2434] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 2434,\n October 1998.\n\n [X680] ITU-T Recommendation X.680 (2002) | ISO/IEC 8824-1:2002,\n Information technology - Abstract Syntax Notation One\n (ASN.1): Specification of basic notation.\n\n [X690] ITU-T Recommendation X.690 (2002) | ISO/IEC 8825-1:2002,\n Information technology - ASN.1 encoding Rules:\n Specification of Basic Encoding Rules (BER), Canonical\n Encoding Rules (CER) and Distinguished Encoding Rules\n (DER).\n\nInformative References\n\n [AEAD] McGrew, D., \"An Interface and Algorithms for Authenticated\n Encryption\", RFC 5116, January 2008.\n\n [AH] Kent, S., \"IP Authentication Header\", RFC 4302, December\n 2005.\n\n [BLEI] Bleichenbacher D., \"Chosen Ciphertext Attacks against\n Protocols Based on RSA Encryption Standard PKCS #1\" in\n Advances in Cryptology -- CRYPTO'98, LNCS vol. 1462,\n pages: 1-12, 1998.\n\n [CBCATT] Moeller, B., \"Security of CBC Ciphersuites in SSL/TLS:\n Problems and Countermeasures\",\n http://www.openssl.org/~bodo/tls-cbc.txt.\n\n [CBCTIME] Canvel, B., Hiltgen, A., Vaudenay, S., and M. Vuagnoux,\n \"Password Interception in a SSL/TLS Channel\", Advances in\n Cryptology -- CRYPTO 2003, LNCS vol. 2729, 2003.\n\n [CCM] \"NIST Special Publication 800-38C: The CCM Mode for\n Authentication and Confidentiality\",\n http://csrc.nist.gov/publications/nistpubs/800-38C/\n SP800-38C.pdf\n\n [DES] National Institute of Standards and Technology, \"Data\n Encryption Standard (DES)\", FIPS PUB 46-3, October 1999.\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 98]\n\f\nRFC 5246 TLS August 2008\n\n\n [DSS-3] NIST FIPS PUB 186-3 Draft, \"Digital Signature Standard\",\n National Institute of Standards and Technology, U.S.\n Department of Commerce, 2006.\n\n [ECDSA] American National Standards Institute, \"Public Key\n Cryptography for the Financial Services Industry: The\n Elliptic Curve Digital Signature Algorithm (ECDSA)\", ANS\n X9.62-2005, November 2005.\n\n [ENCAUTH] Krawczyk, H., \"The Order of Encryption and Authentication\n for Protecting Communications (Or: How Secure is SSL?)\",\n Crypto 2001.\n\n [ESP] Kent, S., \"IP Encapsulating Security Payload (ESP)\", RFC\n 4303, December 2005.\n\n [FI06] Hal Finney, \"Bleichenbacher's RSA signature forgery based\n on implementation error\", [redacted-email] mailing\n list, 27 August 2006, http://www.imc.org/ietf-openpgp/\n mail-archive/msg14307.html.\n\n [GCM] Dworkin, M., NIST Special Publication 800-38D,\n \"Recommendation for Block Cipher Modes of Operation:\n Galois/Counter Mode (GCM) and GMAC\", November 2007.\n\n [IKEALG] Schiller, J., \"Cryptographic Algorithms for Use in the\n Internet Key Exchange Version 2 (IKEv2)\", RFC 4307,\n December 2005.\n\n [KEYSIZ] Orman, H. and P. Hoffman, \"Determining Strengths For\n Public Keys Used For Exchanging Symmetric Keys\", BCP 86,\n RFC 3766, April 2004.\n\n [KPR03] Klima, V., Pokorny, O., Rosa, T., \"Attacking RSA-based\n Sessions in SSL/TLS\", http://eprint.iacr.org/2003/052/,\n March 2003.\n\n [MODP] Kivinen, T. and M. Kojo, \"More Modular Exponential (MODP)\n Diffie-Hellman groups for Internet Key Exchange (IKE)\",\n RFC 3526, May 2003.\n\n [PKCS6] RSA Laboratories, \"PKCS #6: RSA Extended Certificate\n Syntax Standard\", version 1.5, November 1993.\n\n [PKCS7] RSA Laboratories, \"PKCS #7: RSA Cryptographic Message\n Syntax Standard\", version 1.5, November 1993.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 99]\n\f\nRFC 5246 TLS August 2008\n\n\n [RANDOM] Eastlake, D., 3rd, Schiller, J., and S. Crocker,\n \"Randomness Requirements for Security\", BCP 106, RFC 4086,\n June 2005.\n\n [RFC3749] Hollenbeck, S., \"Transport Layer Security Protocol\n Compression Methods\", RFC 3749, May 2004.\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, April 2006.\n\n [RSA] R. Rivest, A. Shamir, and L. M. Adleman, \"A Method for\n Obtaining Digital Signatures and Public-Key\n Cryptosystems\", Communications of the ACM, v. 21, n. 2,\n Feb 1978, pp. 120-126.\n\n [SEQNUM] Bellovin, S., \"Defending Against Sequence Number Attacks\",\n RFC 1948, May 1996.\n\n [SSL2] Hickman, Kipp, \"The SSL Protocol\", Netscape Communications\n Corp., Feb 9, 1995.\n\n [SSL3] A. Freier, P. Karlton, and P. Kocher, \"The SSL 3.0\n Protocol\", Netscape Communications Corp., Nov 18, 1996.\n\n [SUBGROUP] Zuccherato, R., \"Methods for Avoiding the \"Small-Subgroup\"\n Attacks on the Diffie-Hellman Key Agreement Method for\n S/MIME\", RFC 2785, March 2000.\n\n [TCP] Postel, J., \"Transmission Control Protocol\", STD 7, RFC\n 793, September 1981.\n\n [TIMING] Boneh, D., Brumley, D., \"Remote timing attacks are\n practical\", USENIX Security Symposium 2003.\n\n [TLSAES] Chown, P., \"Advanced Encryption Standard (AES)\n Ciphersuites for Transport Layer Security (TLS)\", RFC\n 3268, June 2002.\n\n [TLSECC] Blake-Wilson, S., Bolyard, N., Gupta, V., Hawk, C., and B.\n Moeller, \"Elliptic Curve Cryptography (ECC) Cipher Suites\n for Transport Layer Security (TLS)\", RFC 4492, May 2006.\n\n [TLSEXT] Eastlake, D., 3rd, \"Transport Layer Security (TLS)\n Extensions: Extension Definitions\", Work in Progress,\n February 2008.\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 100]\n\f\nRFC 5246 TLS August 2008\n\n\n [TLSPGP] Mavrogiannopoulos, N., \"Using OpenPGP Keys for Transport\n Layer Security (TLS) Authentication\", RFC 5081, November\n 2007.\n\n [TLSPSK] Eronen, P., Ed., and H. Tschofenig, Ed., \"Pre-Shared Key\n Ciphersuites for Transport Layer Security (TLS)\", RFC\n 4279, December 2005.\n\n [TLS1.0] Dierks, T. and C. Allen, \"The TLS Protocol Version 1.0\",\n RFC 2246, January 1999.\n\n [TLS1.1] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346, April 2006.\n\n [X501] ITU-T Recommendation X.501: Information Technology - Open\n Systems Interconnection - The Directory: Models, 1993.\n\n [XDR] Eisler, M., Ed., \"XDR: External Data Representation\n Standard\", STD 67, RFC 4506, May 2006.\n\nWorking Group Information\n\n The discussion list for the IETF TLS working group is located at the\n e-mail address <[redacted-email]>. Information on the group and\n information on how to subscribe to the list is at\n \n\n Archives of the list can be found at:\n \n\nContributors\n\n Christopher Allen (co-editor of TLS 1.0)\n Alacrity Ventures\n [redacted-email]\n\n Martin Abadi\n University of California, Santa Cruz\n [redacted-email]\n\n Steven M. Bellovin\n Columbia University\n [redacted-email]\n\n Simon Blake-Wilson\n BCI\n [redacted-email]\n\n\n\n\nDierks & Rescorla Standards Track [Page 101]\n\f\nRFC 5246 TLS August 2008\n\n\n Ran Canetti\n IBM\n [redacted-email]\n\n Pete Chown\n Skygate Technology Ltd\n [redacted-email]\n\n Taher Elgamal\n [redacted-email]\n Securify\n\n Pasi Eronen\n [redacted-email]\n Nokia\n\n Anil Gangolli\n [redacted-email]\n\n Kipp Hickman\n\n Alfred Hoenes\n\n David Hopwood\n Independent Consultant\n [redacted-email]\n\n Phil Karlton (co-author of SSLv3)\n\n Paul Kocher (co-author of SSLv3)\n Cryptography Research\n [redacted-email]\n\n Hugo Krawczyk\n IBM\n [redacted-email]\n\n Jan Mikkelsen\n Transactionware\n [redacted-email]\n\n Magnus Nystrom\n RSA Security\n [redacted-email]\n\n Robert Relyea\n Netscape Communications\n [redacted-email]\n\n\n\nDierks & Rescorla Standards Track [Page 102]\n\f\nRFC 5246 TLS August 2008\n\n\n Jim Roskind\n Netscape Communications\n [redacted-email]\n\n Michael Sabin\n\n Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n Tom Weinstein\n\n Tim Wright\n Vodafone\n [redacted-email]\n\nEditors' Addresses\n\n Tim Dierks\n Independent\n EMail: [redacted-email]\n\n Eric Rescorla\n RTFM, Inc.\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 103]\n\f\nRFC 5246 TLS August 2008\n\n\nFull Copyright Statement\n\n Copyright (C) The IETF Trust (2008).\n\n This document is subject to the rights, licenses and restrictions\n contained in BCP 78, and except as set forth therein, the authors\n retain all their rights.\n\n This document and the information contained herein are provided on an\n \"AS IS\" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\n OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\n THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\n OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\n THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\n WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\n\nIntellectual Property\n\n The IETF takes no position regarding the validity or scope of any\n Intellectual Property Rights or other rights that might be claimed to\n pertain to the implementation or use of the technology described in\n this document or the extent to which any license under such rights\n might or might not be available; nor does it represent that it has\n made any independent effort to identify any such rights. Information\n on the procedures with respect to rights in RFC documents can be\n found in BCP 78 and BCP 79.\n\n Copies of IPR disclosures made to the IETF Secretariat and any\n assurances of licenses to be made available, or the result of an\n attempt made to obtain a general license or permission for the use of\n such proprietary rights by implementers or users of this\n specification can be obtained from the IETF on-line IPR repository at\n http://www.ietf.org/ipr.\n\n The IETF invites any interested party to bring to its attention any\n copyrights, patents or patent applications, or other proprietary\n rights that may cover technology that may be required to implement\n this standard. Please address the information to the IETF at\n ietf-ipr@ietf.org.\n\n\n\n\n\n\n\n\n\n\n\n\nDierks & Rescorla Standards Track [Page 104]", "text_sha256": "0ba07ad7c1e04ded76e38500ef83eaa39b002af6cc226329ff75c3e8859fed85"}, {"draft_references": [], "excluded_rfc_relation_targets": [], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5705 RTFM, Inc.", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "5705"}], "kind": "rfc", "occurred_at": "2010-03-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 1, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a", "record_id": "ietf:rfc:5705", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "rfc_number": 5705, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a", "source_url": "https://www.rfc-editor.org/rfc/rfc5705.txt", "temporal_semantics": "source_event_time", "text": "Internet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 5705 RTFM, Inc.\nCategory: Standards Track March 2010\nISSN: 2070-1721\n\n\n Keying Material Exporters for Transport Layer Security (TLS)\n\nAbstract\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n to perform key establishment but then use some of the keying material\n for their own purposes. This document describes a general mechanism\n for allowing that.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/5705.\n\nCopyright Notice\n\n Copyright (c) 2010 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2\n 2. Conventions Used In This Document . . . . . . . . . . . . . . . 3\n 3. Binding to Application Contexts . . . . . . . . . . . . . . . . 3\n 4. Exporter Definition . . . . . . . . . . . . . . . . . . . . . . 4\n 5. Security Considerations . . . . . . . . . . . . . . . . . . . . 5\n 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 6\n 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . 6\n 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 7\n 8.1. Normative References . . . . . . . . . . . . . . . . . . . 7\n 8.2. Informative References . . . . . . . . . . . . . . . . . . 7\n\n1. Introduction\n\n Note: The mechanism described in this document was previously known\n as \"TLS Extractors\" but was changed to avoid a name conflict\n with the use of the term \"Extractor\" in the cryptographic\n community.\n\n A number of protocols wish to leverage Transport Layer Security (TLS)\n [RFC5246] or Datagram TLS (DTLS) [RFC4347] to perform key\n establishment but then use some of the keying material for their own\n purposes. A typical example is DTLS-SRTP [DTLS-SRTP], a key\n management scheme for the Secure Real-time Transport Protocol (SRTP)\n that uses DTLS to perform a key exchange and negotiate the SRTP\n [RFC3711] protection suite and then uses the DTLS master_secret to\n generate the SRTP keys.\n\n These applications imply a need to be able to export keying material\n (later called Exported Keying Material or EKM) from TLS/DTLS to an\n application or protocol residing at an upper layer, and to securely\n agree on the upper-layer context where the keying material will be\n used. The mechanism for exporting the keying material has the\n following requirements:\n\n o Both client and server need to be able to export the same EKM\n value.\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n o EKM values should be indistinguishable from random data to\n attackers who don't know the master_secret.\n\n o It should be possible to export multiple EKM values from the same\n TLS/DTLS association.\n\n o Knowing one EKM value should not reveal any useful information\n about the master_secret or about other EKM values.\n\n The mechanism described in this document is intended to fulfill these\n requirements. This mechanism is compatible with all versions of TLS.\n\n2. Conventions Used In This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in [RFC2119].\n\n3. Binding to Application Contexts\n\n In addition to using an exporter to obtain keying material, an\n application using the keying material has to securely establish the\n upper-layer context where the keying material will be used. The\n details of this context depend on the application, but it could\n include things such as algorithms and parameters that will be used\n with the keys, identifier(s) for the endpoint(s) who will use the\n keys, identifier(s) for the session(s) where the keys will be used,\n and the lifetime(s) for the context and/or keys. At a minimum, there\n should be some mechanism for signaling that an exporter will be used.\n\n This specification does not mandate a single mechanism for agreeing\n on such context; instead, there are several possibilities that can be\n used (and can complement each other). For example:\n\n o Information about the upper-layer context can be included in the\n optional data after the exporter label (see Section 4).\n\n o Information about the upper-layer context can be exchanged in TLS\n extensions included in the ClientHello and ServerHello messages.\n This approach is used in [DTLS-SRTP]. The handshake messages are\n protected by the Finished messages, so once the handshake\n completes, the peers will have the same view of the information.\n Extensions also allow a limited form of negotiation: for example,\n the TLS client could propose several alternatives for some context\n parameters, and the TLS server could select one of them.\n\n o The upper-layer protocol can include its own handshake, which can\n be protected using the keys exported by TLS.\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n No matter how the context is agreed, it is required that it has one\n part that indicates which application will use the exported keys.\n This part is the disambiguating label string (see Section 4).\n\n It is important to note that just embedding TLS messages in the\n upper-layer protocol may not automatically secure all the important\n context information, since the upper-layer messages are not covered\n by TLS Finished messages.\n\n4. Exporter Definition\n\n The output of the exporter is intended to be used in a single scope,\n which is associated with the TLS session, the label, and the context\n value.\n\n The exporter takes three input values:\n\n o a disambiguating label string,\n\n o a per-association context value provided by the application using\n the exporter, and\n\n o a length value.\n\n If no context is provided, it then computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random\n )[length]\n\n If context is provided, it computes:\n\n PRF(SecurityParameters.master_secret, label,\n SecurityParameters.client_random +\n SecurityParameters.server_random +\n context_value_length + context_value\n )[length]\n\n Where PRF is the TLS Pseudorandom Function in use for the session.\n The output is a pseudorandom bit string of length bytes generated\n from the master_secret. (This construction allows for\n interoperability with older exporter-type constructions which do not\n use context values, e.g., [RFC5281]).\n\n Labels here have the same definition as in TLS, i.e., an ASCII string\n with no terminating NULL. Label values beginning with \"EXPERIMENTAL\"\n MAY be used for private use without registration. All other label\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n values MUST be registered via Specification Required as described by\n RFC 5226 [RFC5226]. Note that exporter labels have the potential to\n collide with existing PRF labels. In order to prevent this, labels\n SHOULD begin with \"EXPORTER\". This is not a MUST because there are\n existing uses that have labels which do not begin with this prefix.\n\n The context value allows the application using the exporter to mix\n its own data with the TLS PRF for the exporter output. One example\n of where this might be useful is an authentication setting where the\n client credentials are valid for more than one identity; the context\n value could then be used to mix the expected identity into the keying\n material, thus preventing substitution attacks. The context value\n length is encoded as an unsigned, 16-bit quantity (uint16; see\n [RFC5246], Section 4.4) representing the length of the context value.\n The context MAY be zero length. Because the context value is mixed\n with the master_secret via the PRF, it is safe to mix confidential\n information into the exporter, provided that the master_secret will\n not be known to the attacker.\n\n5. Security Considerations\n\n The prime security requirement for exporter outputs is that they be\n independent. More formally, after a particular TLS session, if an\n adversary is allowed to choose multiple (label, context value) pairs\n and is given the output of the PRF for those values, the attacker is\n still unable to distinguish between the output of the PRF for a\n (label, context value) pair (different from the ones that it\n submitted) and a random value of the same length. In particular,\n there may be settings, such as the one described in Section 4, where\n the attacker can control the context value; such an attacker MUST NOT\n be able to predict the output of the exporter. Similarly, an\n attacker who does not know the master secret should not be able to\n distinguish valid exporter outputs from random values. The current\n set of TLS PRFs is believed to meet this objective, provided the\n master secret is randomly generated.\n\n Because an exporter produces the same value if applied twice with the\n same label to the same master_secret, it is critical that two EKM\n values generated with the same label not be used for two different\n purposes -- hence, the requirement for IANA registration. However,\n because exporters depend on the TLS PRF, it is not a threat to the\n use of an EKM value generated from one label to reveal an EKM value\n generated from another label.\n\n With certain TLS cipher suites, the TLS master secret is not\n necessarily unique to a single TLS session. In particular, with RSA\n key exchange, a malicious party acting as TLS server in one session\n and as TLS client in another session can cause those two sessions to\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n have the same TLS master secret (though the sessions must be\n established simultaneously to get adequate control of the Random\n values). Applications using the EKM need to consider this in how\n they use the EKM; in some cases, requiring the use of other cipher\n suites (such as those using a Diffie-Hellman key exchange) may be\n advisable.\n\n Designing a secure mechanism that uses exporters is not necessarily\n straightforward. This document only provides the exporter mechanism,\n but the problem of agreeing on the surrounding context and the\n meaning of the information passed to and from the exporter remains.\n Any new uses of the exporter mechanism should be subject to careful\n review.\n\n6. IANA Considerations\n\n IANA has created a TLS Exporter Label registry for this purpose. The\n initial contents of the registry are given below:\n\n Value Reference Note\n ----------------------------- --------- ----\n client finished [RFC5246] (1)\n server finished [RFC5246] (1)\n master secret [RFC5246] (1)\n key expansion [RFC5246] (1)\n client EAP encryption [RFC5216]\n ttls keying material [RFC5281]\n ttls challenge [RFC5281]\n\n Note: (1) These entries are reserved and MUST NOT be used for the\n purpose described in RFC 5705, in order to avoid confusion with\n similar, but distinct, use in RFC 5246.\n\n Future values are allocated via the RFC 5226 Specification Required\n policy. The label is a string consisting of printable ASCII\n characters. IANA MUST also verify that one label is not a prefix of\n any other label. For example, labels \"key\" or \"master secretary\" are\n forbidden.\n\n7. Acknowledgments\n\n Thanks to Pasi Eronen for valuable comments and for the contents of\n the IANA section and Section 3. Thanks to David McGrew for helpful\n discussion of the security considerations and to Vijay Gurbani and\n Alfred Hoenes for editorial comments.\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 5705 TLS Exporters March 2010\n\n\n8. References\n\n8.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246,\n August 2008.\n\n8.2. Informative References\n\n [DTLS-SRTP] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for Secure\n Real-time Transport Protocol (SRTP)\", Work in Progress,\n February 2009.\n\n [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and\n K. Norrman, \"The Secure Real-time Transport Protocol\n (SRTP)\", RFC 3711, March 2004.\n\n [RFC4347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security\", RFC 4347, April 2006.\n\n [RFC5216] Simon, D., Aboba, B., and R. Hurst, \"The EAP-TLS\n Authentication Protocol\", RFC 5216, March 2008.\n\n [RFC5281] Funk, P. and S. Blake-Wilson, \"Extensible Authentication\n Protocol Tunneled Transport Layer Security Authenticated\n Protocol Version 0 (EAP-TTLSv0)\", RFC 5281, August 2008.\n\nAuthor's Address\n\n Eric Rescorla\n RTFM, Inc.\n 2064 Edgewood Drive\n Palo Alto, CA 94303\n USA\n\n EMail: [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 7]", "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"draft_references": [], "excluded_rfc_relation_targets": [{"kind": "obsoletes", "reason": "not_requested", "rfc_number": 2011}, {"kind": "obsoletes", "reason": "not_requested", "rfc_number": 4366}], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6066 Huawei", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "6066"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 4366 January 2011", "fact_id": "obsoletes:4366", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "4366"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 4366 January 2011", "fact_id": "obsoletes:2011", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "2011"}], "kind": "rfc", "occurred_at": "2011-01-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 3, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0", "record_id": "ietf:rfc:6066", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "rfc_number": 6066, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0", "source_url": "https://www.rfc-editor.org/rfc/rfc6066.txt", "temporal_semantics": "source_event_time", "text": "Internet Engineering Task Force (IETF) D. Eastlake 3rd\nRequest for Comments: 6066 Huawei\nObsoletes: 4366 January 2011\nCategory: Standards Track\nISSN: 2070-1721\n\n\n Transport Layer Security (TLS) Extensions: Extension Definitions\n\nAbstract\n\n This document provides specifications for existing TLS extensions.\n It is a companion document for RFC 5246, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\". The extensions specified are\n server_name, max_fragment_length, client_certificate_url,\n trusted_ca_keys, truncated_hmac, and status_request.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6066.\n\nCopyright Notice\n\n Copyright (c) 2011 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n\n\n\n\n\nEastlake Standards Track [Page 1]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction ....................................................3\n 1.1. Specific Extensions Covered ................................3\n 1.2. Conventions Used in This Document ..........................5\n 2. Extensions to the Handshake Protocol ............................5\n 3. Server Name Indication ..........................................6\n 4. Maximum Fragment Length Negotiation .............................8\n 5. Client Certificate URLs .........................................9\n 6. Trusted CA Indication ..........................................12\n 7. Truncated HMAC .................................................13\n 8. Certificate Status Request .....................................14\n 9. Error Alerts ...................................................16\n 10. IANA Considerations ...........................................17\n 10.1. pkipath MIME Type Registration ...........................17\n 10.2. Reference for TLS Alerts, TLS HandshakeTypes, and\n ExtensionTypes ...........................................19\n 11. Security Considerations .......................................19\n 11.1. Security Considerations for server_name ..................19\n 11.2. Security Considerations for max_fragment_length ..........20\n 11.3. Security Considerations for client_certificate_url .......20\n 11.4. Security Considerations for trusted_ca_keys ..............21\n 11.5. Security Considerations for truncated_hmac ...............21\n 11.6. Security Considerations for status_request ...............22\n 12. Normative References ..........................................22\n 13. Informative References ........................................23\n Appendix A. Changes from RFC 4366 .................................24\n Appendix B. Acknowledgements ......................................25\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 2]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1. Introduction\n\n The Transport Layer Security (TLS) Protocol Version 1.2 is specified\n in [RFC5246]. That specification includes the framework for\n extensions to TLS, considerations in designing such extensions (see\n Section 7.4.1.4 of [RFC5246]), and IANA Considerations for the\n allocation of new extension code points; however, it does not specify\n any particular extensions other than Signature Algorithms (see\n Section 7.4.1.4.1 of [RFC5246]).\n\n This document provides the specifications for existing TLS\n extensions. It is, for the most part, the adaptation and editing of\n material from RFC 4366, which covered TLS extensions for TLS 1.0 (RFC\n 2246) and TLS 1.1 (RFC 4346).\n\n1.1. Specific Extensions Covered\n\n The extensions described here focus on extending the functionality\n provided by the TLS protocol message formats. Other issues, such as\n the addition of new cipher suites, are deferred.\n\n The extension types defined in this document are:\n\n enum {\n server_name(0), max_fragment_length(1),\n client_certificate_url(2), trusted_ca_keys(3),\n truncated_hmac(4), status_request(5), (65535)\n } ExtensionType;\n\n Specifically, the extensions described in this document:\n\n - Allow TLS clients to provide to the TLS server the name of the\n server they are contacting. This functionality is desirable in\n order to facilitate secure connections to servers that host\n multiple 'virtual' servers at a single underlying network address.\n\n - Allow TLS clients and servers to negotiate the maximum fragment\n length to be sent. This functionality is desirable as a result of\n memory constraints among some clients, and bandwidth constraints\n among some access networks.\n\n - Allow TLS clients and servers to negotiate the use of client\n certificate URLs. This functionality is desirable in order to\n conserve memory on constrained clients.\n\n\n\n\n\n\n\nEastlake Standards Track [Page 3]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - Allow TLS clients to indicate to TLS servers which certification\n authority (CA) root keys they possess. This functionality is\n desirable in order to prevent multiple handshake failures\n involving TLS clients that are only able to store a small number\n of CA root keys due to memory limitations.\n\n - Allow TLS clients and servers to negotiate the use of truncated\n Message Authentication Codes (MACs). This functionality is\n desirable in order to conserve bandwidth in constrained access\n networks.\n\n - Allow TLS clients and servers to negotiate that the server sends\n the client certificate status information (e.g., an Online\n Certificate Status Protocol (OCSP) [RFC2560] response) during a\n TLS handshake. This functionality is desirable in order to avoid\n sending a Certificate Revocation List (CRL) over a constrained\n access network and therefore saving bandwidth.\n\n TLS clients and servers may use the extensions described in this\n document. The extensions are designed to be backwards compatible,\n meaning that TLS clients that support the extensions can talk to TLS\n servers that do not support the extensions, and vice versa.\n\n Note that any messages associated with these extensions that are sent\n during the TLS handshake MUST be included in the hash calculations\n involved in \"Finished\" messages.\n\n Note also that all the extensions defined in this document are\n relevant only when a session is initiated. A client that requests\n session resumption does not in general know whether the server will\n accept this request, and therefore it SHOULD send the same extensions\n as it would send if it were not attempting resumption. When a client\n includes one or more of the defined extension types in an extended\n client hello while requesting session resumption:\n\n - The server name indication extension MAY be used by the server\n when deciding whether or not to resume a session as described in\n Section 3.\n\n - If the resumption request is denied, the use of the extensions is\n negotiated as normal.\n\n - If, on the other hand, the older session is resumed, then the\n server MUST ignore the extensions and send a server hello\n containing none of the extension types. In this case, the\n functionality of these extensions negotiated during the original\n session initiation is applied to the resumed session.\n\n\n\n\nEastlake Standards Track [Page 4]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n1.2. Conventions Used in This Document\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n [RFC2119].\n\n2. Extensions to the Handshake Protocol\n\n This document specifies the use of two new handshake messages,\n \"CertificateURL\" and \"CertificateStatus\". These messages are\n described in Sections 5 and 8, respectively. The new handshake\n message structure therefore becomes:\n\n enum {\n hello_request(0), client_hello(1), server_hello(2),\n certificate(11), server_key_exchange (12),\n certificate_request(13), server_hello_done(14),\n certificate_verify(15), client_key_exchange(16),\n finished(20), certificate_url(21), certificate_status(22),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (HandshakeType) {\n case hello_request: HelloRequest;\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case certificate: Certificate;\n case server_key_exchange: ServerKeyExchange;\n case certificate_request: CertificateRequest;\n case server_hello_done: ServerHelloDone;\n case certificate_verify: CertificateVerify;\n case client_key_exchange: ClientKeyExchange;\n case finished: Finished;\n case certificate_url: CertificateURL;\n case certificate_status: CertificateStatus;\n } body;\n } Handshake;\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 5]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n3. Server Name Indication\n\n TLS does not provide a mechanism for a client to tell a server the\n name of the server it is contacting. It may be desirable for clients\n to provide this information to facilitate secure connections to\n servers that host multiple 'virtual' servers at a single underlying\n network address.\n\n In order to provide any of the server names, clients MAY include an\n extension of type \"server_name\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"ServerNameList\" where:\n\n struct {\n NameType name_type;\n select (name_type) {\n case host_name: HostName;\n } name;\n } ServerName;\n\n enum {\n host_name(0), (255)\n } NameType;\n\n opaque HostName<1..2^16-1>;\n\n struct {\n ServerName server_name_list<1..2^16-1>\n } ServerNameList;\n\n The ServerNameList MUST NOT contain more than one name of the same\n name_type. If the server understood the ClientHello extension but\n does not recognize the server name, the server SHOULD take one of two\n actions: either abort the handshake by sending a fatal-level\n unrecognized_name(112) alert or continue the handshake. It is NOT\n RECOMMENDED to send a warning-level unrecognized_name(112) alert,\n because the client's behavior in response to warning-level alerts is\n unpredictable. If there is a mismatch between the server name used\n by the client application and the server name of the credential\n chosen by the server, this mismatch will become apparent when the\n client application performs the server endpoint identification, at\n which point the client application will have to decide whether to\n proceed with the communication. TLS implementations are encouraged\n to make information available to application callers about warning-\n level alerts that were received or sent during a TLS handshake. Such\n information can be useful for diagnostic purposes.\n\n\n\n\n\nEastlake Standards Track [Page 6]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note: Earlier versions of this specification permitted multiple\n names of the same name_type. In practice, current client\n implementations only send one name, and the client cannot\n necessarily find out which name the server selected. Multiple\n names of the same name_type are therefore now prohibited.\n\n Currently, the only server names supported are DNS hostnames;\n however, this does not imply any dependency of TLS on DNS, and other\n name types may be added in the future (by an RFC that updates this\n document). The data structure associated with the host_name NameType\n is a variable-length vector that begins with a 16-bit length. For\n backward compatibility, all future data structures associated with\n new NameTypes MUST begin with a 16-bit length field. TLS MAY treat\n provided server names as opaque data and pass the names and types to\n the application.\n\n \"HostName\" contains the fully qualified DNS hostname of the server,\n as understood by the client. The hostname is represented as a byte\n string using ASCII encoding without a trailing dot. This allows the\n support of internationalized domain names through the use of A-labels\n defined in [RFC5890]. DNS hostnames are case-insensitive. The\n algorithm to compare hostnames is described in [RFC5890], Section\n 2.3.2.4.\n\n Literal IPv4 and IPv6 addresses are not permitted in \"HostName\".\n\n It is RECOMMENDED that clients include an extension of type\n \"server_name\" in the client hello whenever they locate a server by a\n supported name type.\n\n A server that receives a client hello containing the \"server_name\"\n extension MAY use the information contained in the extension to guide\n its selection of an appropriate certificate to return to the client,\n and/or other aspects of security policy. In this event, the server\n SHALL include an extension of type \"server_name\" in the (extended)\n server hello. The \"extension_data\" field of this extension SHALL be\n empty.\n\n When the server is deciding whether or not to accept a request to\n resume a session, the contents of a server_name extension MAY be used\n in the lookup of the session in the session cache. The client SHOULD\n include the same server_name extension in the session resumption\n request as it did in the full handshake that established the session.\n A server that implements this extension MUST NOT accept the request\n to resume the session if the server_name extension contains a\n different name. Instead, it proceeds with a full handshake to\n establish a new session. When resuming a session, the server MUST\n NOT include a server_name extension in the server hello.\n\n\n\nEastlake Standards Track [Page 7]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n If an application negotiates a server name using an application\n protocol and then upgrades to TLS, and if a server_name extension is\n sent, then the extension SHOULD contain the same name that was\n negotiated in the application protocol. If the server_name is\n established in the TLS session handshake, the client SHOULD NOT\n attempt to request a different server name at the application layer.\n\n4. Maximum Fragment Length Negotiation\n\n Without this extension, TLS specifies a fixed maximum plaintext\n fragment length of 2^14 bytes. It may be desirable for constrained\n clients to negotiate a smaller maximum fragment length due to memory\n limitations or bandwidth limitations.\n\n In order to negotiate smaller maximum fragment lengths, clients MAY\n include an extension of type \"max_fragment_length\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain:\n\n enum{\n 2^9(1), 2^10(2), 2^11(3), 2^12(4), (255)\n } MaxFragmentLength;\n\n whose value is the desired maximum fragment length. The allowed\n values for this field are: 2^9, 2^10, 2^11, and 2^12.\n\n Servers that receive an extended client hello containing a\n \"max_fragment_length\" extension MAY accept the requested maximum\n fragment length by including an extension of type\n \"max_fragment_length\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL contain a\n \"MaxFragmentLength\" whose value is the same as the requested maximum\n fragment length.\n\n If a server receives a maximum fragment length negotiation request\n for a value other than the allowed values, it MUST abort the\n handshake with an \"illegal_parameter\" alert. Similarly, if a client\n receives a maximum fragment length negotiation response that differs\n from the length it requested, it MUST also abort the handshake with\n an \"illegal_parameter\" alert.\n\n Once a maximum fragment length other than 2^14 has been successfully\n negotiated, the client and server MUST immediately begin fragmenting\n messages (including handshake messages) to ensure that no fragment\n larger than the negotiated length is sent. Note that TLS already\n requires clients and servers to support fragmentation of handshake\n messages.\n\n\n\n\nEastlake Standards Track [Page 8]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n The negotiated length applies for the duration of the session\n including session resumptions.\n\n The negotiated length limits the input that the record layer may\n process without fragmentation (that is, the maximum value of\n TLSPlaintext.length; see [RFC5246], Section 6.2.1). Note that the\n output of the record layer may be larger. For example, if the\n negotiated length is 2^9=512, then, when using currently defined\n cipher suites (those defined in [RFC5246] and [RFC2712]) and null\n compression, the record-layer output can be at most 805 bytes: 5\n bytes of headers, 512 bytes of application data, 256 bytes of\n padding, and 32 bytes of MAC. This means that in this event a TLS\n record-layer peer receiving a TLS record-layer message larger than\n 805 bytes MUST discard the message and send a \"record_overflow\"\n alert, without decrypting the message. When this extension is used\n with Datagram Transport Layer Security (DTLS), implementations SHOULD\n NOT generate record_overflow alerts unless the packet passes message\n authentication.\n\n5. Client Certificate URLs\n\n Without this extension, TLS specifies that when client authentication\n is performed, client certificates are sent by clients to servers\n during the TLS handshake. It may be desirable for constrained\n clients to send certificate URLs in place of certificates, so that\n they do not need to store their certificates and can therefore save\n memory.\n\n In order to negotiate sending certificate URLs to a server, clients\n MAY include an extension of type \"client_certificate_url\" in the\n (extended) client hello. The \"extension_data\" field of this\n extension SHALL be empty.\n\n (Note that it is necessary to negotiate the use of client certificate\n URLs in order to avoid \"breaking\" existing TLS servers.)\n\n Servers that receive an extended client hello containing a\n \"client_certificate_url\" extension MAY indicate that they are willing\n to accept certificate URLs by including an extension of type\n \"client_certificate_url\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n After negotiation of the use of client certificate URLs has been\n successfully completed (by exchanging hellos including\n \"client_certificate_url\" extensions), clients MAY send a\n \"CertificateURL\" message in place of a \"Certificate\" message as\n follows (see also Section 2):\n\n\n\n\nEastlake Standards Track [Page 9]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n enum {\n individual_certs(0), pkipath(1), (255)\n } CertChainType;\n\n struct {\n CertChainType type;\n URLAndHash url_and_hash_list<1..2^16-1>;\n } CertificateURL;\n\n struct {\n opaque url<1..2^16-1>;\n unint8 padding;\n opaque SHA1Hash[20];\n } URLAndHash;\n\n Here, \"url_and_hash_list\" contains a sequence of URLs and hashes.\n Each \"url\" MUST be an absolute URI reference according to [RFC3986]\n that can be immediately used to fetch the certificate(s).\n\n When X.509 certificates are used, there are two possibilities:\n\n - If CertificateURL.type is \"individual_certs\", each URL refers to a\n single DER-encoded X.509v3 certificate, with the URL for the\n client's certificate first.\n\n - If CertificateURL.type is \"pkipath\", the list contains a single\n URL referring to a DER-encoded certificate chain, using the type\n PkiPath described in Section 10.1.\n\n When any other certificate format is used, the specification that\n describes use of that format in TLS should define the encoding format\n of certificates or certificate chains, and any constraint on their\n ordering.\n\n The \"padding\" byte MUST be 0x01. It is present to make the structure\n backwards compatible.\n\n The hash corresponding to each URL is the SHA-1 hash of the\n certificate or certificate chain (in the case of X.509 certificates,\n the DER-encoded certificate or the DER-encoded PkiPath).\n\n Note that when a list of URLs for X.509 certificates is used, the\n ordering of URLs is the same as that used in the TLS Certificate\n message (see [RFC5246], Section 7.4.2), but opposite to the order in\n which certificates are encoded in PkiPath. In either case, the self-\n signed root certificate MAY be omitted from the chain, under the\n assumption that the server must already possess it in order to\n validate it.\n\n\n\nEastlake Standards Track [Page 10]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Servers receiving \"CertificateURL\" SHALL attempt to retrieve the\n client's certificate chain from the URLs and then process the\n certificate chain as usual. A cached copy of the content of any URL\n in the chain MAY be used, provided that the SHA-1 hash matches the\n hash of the cached copy.\n\n Servers that support this extension MUST support the 'http' URI\n scheme for certificate URLs and MAY support other schemes. Use of\n other schemes than 'http', 'https', or 'ftp' may create unexpected\n problems.\n\n If the protocol used is HTTP, then the HTTP server can be configured\n to use the Cache-Control and Expires directives described in\n [RFC2616] to specify whether and for how long certificates or\n certificate chains should be cached.\n\n The TLS server MUST NOT follow HTTP redirects when retrieving the\n certificates or certificate chain. The URLs used in this extension\n MUST NOT be chosen to depend on such redirects.\n\n If the protocol used to retrieve certificates or certificate chains\n returns a MIME-formatted response (as HTTP does), then the following\n MIME Content-Types SHALL be used: when a single X.509v3 certificate\n is returned, the Content-Type is \"application/pkix-cert\" [RFC2585],\n and when a chain of X.509v3 certificates is returned, the Content-\n Type is \"application/pkix-pkipath\" (Section 10.1).\n\n The server MUST check that the SHA-1 hash of the contents of the\n object retrieved from that URL (after decoding any MIME Content-\n Transfer-Encoding) matches the given hash. If any retrieved object\n does not have the correct SHA-1 hash, the server MUST abort the\n handshake with a bad_certificate_hash_value(114) alert. This alert\n is always fatal.\n\n Clients may choose to send either \"Certificate\" or \"CertificateURL\"\n after successfully negotiating the option to send certificate URLs.\n The option to send a certificate is included to provide flexibility\n to clients possessing multiple certificates.\n\n If a server is unable to obtain certificates in a given\n CertificateURL, it MUST send a fatal certificate_unobtainable(111)\n alert if it requires the certificates to complete the handshake. If\n the server does not require the certificates, then the server\n continues the handshake. The server MAY send a warning-level alert\n in this case. Clients receiving such an alert SHOULD log the alert\n and continue with the handshake if possible.\n\n\n\n\n\nEastlake Standards Track [Page 11]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n6. Trusted CA Indication\n\n Constrained clients that, due to memory limitations, possess only a\n small number of CA root keys may wish to indicate to servers which\n root keys they possess, in order to avoid repeated handshake\n failures.\n\n In order to indicate which CA root keys they possess, clients MAY\n include an extension of type \"trusted_ca_keys\" in the (extended)\n client hello. The \"extension_data\" field of this extension SHALL\n contain \"TrustedAuthorities\" where:\n\n struct {\n TrustedAuthority trusted_authorities_list<0..2^16-1>;\n } TrustedAuthorities;\n\n struct {\n IdentifierType identifier_type;\n select (identifier_type) {\n case pre_agreed: struct {};\n case key_sha1_hash: SHA1Hash;\n case x509_name: DistinguishedName;\n case cert_sha1_hash: SHA1Hash;\n } identifier;\n } TrustedAuthority;\n\n enum {\n pre_agreed(0), key_sha1_hash(1), x509_name(2),\n cert_sha1_hash(3), (255)\n } IdentifierType;\n\n opaque DistinguishedName<1..2^16-1>;\n\n Here, \"TrustedAuthorities\" provides a list of CA root key identifiers\n that the client possesses. Each CA root key is identified via\n either:\n\n - \"pre_agreed\": no CA root key identity supplied.\n\n - \"key_sha1_hash\": contains the SHA-1 hash of the CA root key. For\n Digital Signature Algorithm (DSA) and Elliptic Curve Digital\n Signature Algorithm (ECDSA) keys, this is the hash of the\n \"subjectPublicKey\" value. For RSA keys, the hash is of the big-\n endian byte string representation of the modulus without any\n initial zero-valued bytes. (This copies the key hash formats\n deployed in other environments.)\n\n\n\n\n\nEastlake Standards Track [Page 12]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n - \"x509_name\": contains the DER-encoded X.509 DistinguishedName of\n the CA.\n\n - \"cert_sha1_hash\": contains the SHA-1 hash of a DER-encoded\n Certificate containing the CA root key.\n\n Note that clients may include none, some, or all of the CA root keys\n they possess in this extension.\n\n Note also that it is possible that a key hash or a Distinguished Name\n alone may not uniquely identify a certificate issuer (for example, if\n a particular CA has multiple key pairs). However, here we assume\n this is the case following the use of Distinguished Names to identify\n certificate issuers in TLS.\n\n The option to include no CA root keys is included to allow the client\n to indicate possession of some pre-defined set of CA root keys.\n\n Servers that receive a client hello containing the \"trusted_ca_keys\"\n extension MAY use the information contained in the extension to guide\n their selection of an appropriate certificate chain to return to the\n client. In this event, the server SHALL include an extension of type\n \"trusted_ca_keys\" in the (extended) server hello. The\n \"extension_data\" field of this extension SHALL be empty.\n\n7. Truncated HMAC\n\n Currently defined TLS cipher suites use the MAC construction HMAC\n [RFC2104] to authenticate record-layer communications. In TLS, the\n entire output of the hash function is used as the MAC tag. However,\n it may be desirable in constrained environments to save bandwidth by\n truncating the output of the hash function to 80 bits when forming\n MAC tags.\n\n In order to negotiate the use of 80-bit truncated HMAC, clients MAY\n include an extension of type \"truncated_hmac\" in the extended client\n hello. The \"extension_data\" field of this extension SHALL be empty.\n\n Servers that receive an extended hello containing a \"truncated_hmac\"\n extension MAY agree to use a truncated HMAC by including an extension\n of type \"truncated_hmac\", with empty \"extension_data\", in the\n extended server hello.\n\n Note that if new cipher suites are added that do not use HMAC, and\n the session negotiates one of these cipher suites, this extension\n will have no effect. It is strongly recommended that any new cipher\n suites using other MACs consider the MAC size an integral part of the\n\n\n\n\nEastlake Standards Track [Page 13]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n cipher suite definition, taking into account both security and\n bandwidth considerations.\n\n If HMAC truncation has been successfully negotiated during a TLS\n handshake, and the negotiated cipher suite uses HMAC, both the client\n and the server pass this fact to the TLS record layer along with the\n other negotiated security parameters. Subsequently during the\n session, clients and servers MUST use truncated HMACs, calculated as\n specified in [RFC2104]. That is, SecurityParameters.mac_length is 10\n bytes, and only the first 10 bytes of the HMAC output are transmitted\n and checked. Note that this extension does not affect the\n calculation of the pseudo-random function (PRF) as part of\n handshaking or key derivation.\n\n The negotiated HMAC truncation size applies for the duration of the\n session including session resumptions.\n\n8. Certificate Status Request\n\n Constrained clients may wish to use a certificate-status protocol\n such as OCSP [RFC2560] to check the validity of server certificates,\n in order to avoid transmission of CRLs and therefore save bandwidth\n on constrained networks. This extension allows for such information\n to be sent in the TLS handshake, saving roundtrips and resources.\n\n In order to indicate their desire to receive certificate status\n information, clients MAY include an extension of type\n \"status_request\" in the (extended) client hello. The\n \"extension_data\" field of this extension SHALL contain\n \"CertificateStatusRequest\" where:\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n } request;\n } CertificateStatusRequest;\n\n enum { ocsp(1), (255) } CertificateStatusType;\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n\n\n\nEastlake Standards Track [Page 14]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n In the OCSPStatusRequest, the \"ResponderIDs\" provides a list of OCSP\n responders that the client trusts. A zero-length \"responder_id_list\"\n sequence has the special meaning that the responders are implicitly\n known to the server, e.g., by prior arrangement. \"Extensions\" is a\n DER encoding of OCSP request extensions.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC2560]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a zero-length ASN.1 SEQUENCE, which is not\n valid for the \"Extensions\" type).\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement).\n\n Servers that receive a client hello containing the \"status_request\"\n extension MAY return a suitable certificate status response to the\n client along with their certificate. If OCSP is requested, they\n SHOULD use the information contained in the extension when selecting\n an OCSP responder and SHOULD include request_extensions in the OCSP\n request.\n\n Servers return a certificate response along with their certificate by\n sending a \"CertificateStatus\" message immediately after the\n \"Certificate\" message (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message, then the server MUST have included an\n extension of type \"status_request\" with empty \"extension_data\" in the\n extended server hello. The \"CertificateStatus\" message is conveyed\n using the handshake message type \"certificate_status\" as follows (see\n also Section 2):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<1..2^24-1>;\n\n An \"ocsp_response\" contains a complete, DER-encoded OCSP response\n (using the ASN.1 type OCSPResponse defined in [RFC2560]). Only one\n OCSP response may be sent.\n\n\n\n\nEastlake Standards Track [Page 15]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if has received a \"status_request\" extension in the\n client hello message and has sent a \"status_request\" extension in the\n server hello message.\n\n Note in addition that a server MUST NOT send the \"CertificateStatus\"\n message unless it received a \"status_request\" extension in the client\n hello message and sent a \"status_request\" extension in the server\n hello message.\n\n Clients requesting an OCSP response and receiving an OCSP response in\n a \"CertificateStatus\" message MUST check the OCSP response and abort\n the handshake if the response is not satisfactory with\n bad_certificate_status_response(113) alert. This alert is always\n fatal.\n\n9. Error Alerts\n\n Four new error alerts are defined for use with the TLS extensions\n defined in this document. To avoid \"breaking\" existing clients and\n servers, these alerts MUST NOT be sent unless the sending party has\n received an extended hello message from the party they are\n communicating with. These error alerts are conveyed using the\n following syntax. The new alerts are the last four, as indicated by\n the comments on the same line as the error alert number.\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed(21),\n record_overflow(22),\n decompression_failure(30),\n handshake_failure(40),\n /* 41 is not defined, for historical reasons */\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction(60),\n protocol_version(70),\n insufficient_security(71),\n\n\n\nEastlake Standards Track [Page 16]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n internal_error(80),\n user_canceled(90),\n no_renegotiation(100),\n unsupported_extension(110),\n certificate_unobtainable(111), /* new */\n unrecognized_name(112), /* new */\n bad_certificate_status_response(113), /* new */\n bad_certificate_hash_value(114), /* new */\n (255)\n } AlertDescription;\n\n \"certificate_unobtainable\" is described in Section 5.\n \"unrecognized_name\" is described in Section 3.\n \"bad_certificate_status_response\" is described in Section 8.\n \"bad_certificate_hash_value\" is described in Section 5.\n\n10. IANA Considerations\n\n IANA Considerations for TLS extensions and the creation of a registry\n are covered in Section 12 of [RFC5246] except for the registration of\n MIME type application/pkix-pkipath, which appears below.\n\n The IANA TLS extensions and MIME type application/pkix-pkipath\n registry entries that reference RFC 4366 have been updated to\n reference this document.\n\n10.1. pkipath MIME Type Registration\n\n MIME media type name: application\n MIME subtype name: pkix-pkipath\n Required parameters: none\n\n Optional parameters: version (default value is \"1\")\n\n Encoding considerations:\n Binary; this MIME type is a DER encoding of the ASN.1 type\n PkiPath, defined as follows:\n PkiPath ::= SEQUENCE OF Certificate\n PkiPath is used to represent a certification path. Within the\n sequence, the order of certificates is such that the subject of\n the first certificate is the issuer of the second certificate,\n etc.\n This is identical to the definition published in [X509-4th-TC1];\n note that it is different from that in [X509-4th].\n\n All Certificates MUST conform to [RFC5280]. (This should be\n interpreted as a requirement to encode only PKIX-conformant\n certificates using this type. It does not necessarily require\n\n\n\nEastlake Standards Track [Page 17]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n that all certificates that are not strictly PKIX-conformant must\n be rejected by relying parties, although the security consequences\n of accepting any such certificates should be considered\n carefully.)\n\n DER (as opposed to BER) encoding MUST be used. If this type is\n sent over a 7-bit transport, base64 encoding SHOULD be used.\n\n Security considerations:\n The security considerations of [X509-4th] and [RFC5280] (or any\n updates to them) apply, as well as those of any protocol that uses\n this type (e.g., TLS).\n\n Note that this type only specifies a certificate chain that can be\n assessed for validity according to the relying party's existing\n configuration of trusted CAs; it is not intended to be used to\n specify any change to that configuration.\n\n Interoperability considerations:\n No specific interoperability problems are known with this type,\n but for recommendations relating to X.509 certificates in general,\n see [RFC5280].\n\n Published specification: This document and [RFC5280].\n\n Applications that use this media type:\n TLS. It may also be used by other protocols or for general\n interchange of PKIX certificate chains.\n\n Additional information:\n Magic number(s): DER-encoded ASN.1 can be easily recognized.\n Further parsing is required to distinguish it from other ASN.1\n types.\n File extension(s): .pkipath\n Macintosh File Type Code(s): not specified\n\n Person & email address to contact for further information:\n Magnus Nystrom <[redacted-email]>\n\n Intended usage: COMMON\n\n Change controller: IESG <[redacted-email]>\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 18]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n10.2. Reference for TLS Alerts, TLS HandshakeTypes, and ExtensionTypes\n\n The following values in the TLS Alert Registry have been updated to\n reference this document:\n\n 111 certificate_unobtainable\n 112 unrecognized_name\n 113 bad_certificate_status_response\n 114 bad_certificate_hash_value\n\n The following values in the TLS HandshakeType Registry have been\n updated to reference this document:\n\n 21 certificate_url\n 22 certificate_status\n\n The following ExtensionType values have been updated to reference\n this document:\n\n 0 server_name\n 1 max_fragment_length\n 2 client_certificate_url\n 3 trusted_ca_keys\n 4 truncated_hmac\n 5 status_request\n\n11. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security Considerations for particular extensions\n specified in this document are given below.\n\n In general, implementers should continue to monitor the state of the\n art and address any weaknesses identified.\n\n11.1. Security Considerations for server_name\n\n If a single server hosts several domains, then clearly it is\n necessary for the owners of each domain to ensure that this satisfies\n their security needs. Apart from this, server_name does not appear\n to introduce significant security issues.\n\n Since it is possible for a client to present a different server_name\n in the application protocol, application server implementations that\n rely upon these names being the same MUST check to make sure the\n client did not present a different name in the application protocol.\n\n\n\n\n\nEastlake Standards Track [Page 19]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Implementations MUST ensure that a buffer overflow does not occur,\n whatever the values of the length fields in server_name.\n\n11.2. Security Considerations for max_fragment_length\n\n The maximum fragment length takes effect immediately, including for\n handshake messages. However, that does not introduce any security\n complications that are not already present in TLS, since TLS requires\n implementations to be able to handle fragmented handshake messages.\n\n Note that, as described in Section 4, once a non-null cipher suite\n has been activated, the effective maximum fragment length depends on\n the cipher suite and compression method, as well as on the negotiated\n max_fragment_length. This must be taken into account when sizing\n buffers and checking for buffer overflow.\n\n11.3. Security Considerations for client_certificate_url\n\n Support for client_certificate_url involves the server's acting as a\n client in another URI-scheme-dependent protocol. The server\n therefore becomes subject to many of the same security concerns that\n clients of the URI scheme are subject to, with the added concern that\n the client can attempt to prompt the server to connect to some\n (possibly weird-looking) URL.\n\n In general, this issue means that an attacker might use the server to\n indirectly attack another host that is vulnerable to some security\n flaw. It also introduces the possibility of denial-of-service\n attacks in which an attacker makes many connections to the server,\n each of which results in the server's attempting a connection to the\n target of the attack.\n\n Note that the server may be behind a firewall or otherwise able to\n access hosts that would not be directly accessible from the public\n Internet. This could exacerbate the potential security and denial-\n of-service problems described above, as well as allow the existence\n of internal hosts to be confirmed when they would otherwise be\n hidden.\n\n The detailed security concerns involved will depend on the URI\n schemes supported by the server. In the case of HTTP, the concerns\n are similar to those that apply to a publicly accessible HTTP proxy\n server. In the case of HTTPS, loops and deadlocks may be created,\n and this should be addressed. In the case of FTP, attacks arise that\n are similar to FTP bounce attacks.\n\n\n\n\n\n\nEastlake Standards Track [Page 20]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n As a result of this issue, it is RECOMMENDED that the\n client_certificate_url extension should have to be specifically\n enabled by a server administrator, rather than be enabled by default.\n It is also RECOMMENDED that URI schemes be enabled by the\n administrator individually, and only a minimal set of schemes be\n enabled. Unusual protocols that offer limited security or whose\n security is not well understood SHOULD be avoided.\n\n As discussed in [RFC3986], URLs that specify ports other than the\n default may cause problems, as may very long URLs (which are more\n likely to be useful in exploiting buffer overflow bugs).\n\n This extension continues to use SHA-1 (as in RFC 4366) and does not\n provide algorithm agility. The property required of SHA-1 in this\n case is second pre-image resistance, not collision resistance.\n Furthermore, even if second pre-image attacks against SHA-1 are found\n in the future, an attack against client_certificate_url would require\n a second pre-image that is accepted as a valid certificate by the\n server and contains the same public key.\n\n Also note that HTTP caching proxies are common on the Internet, and\n some proxies do not check for the latest version of an object\n correctly. If a request using HTTP (or another caching protocol)\n goes through a misconfigured or otherwise broken proxy, the proxy may\n return an out-of-date response.\n\n11.4. Security Considerations for trusted_ca_keys\n\n Potentially, the CA root keys a client possesses could be regarded as\n confidential information. As a result, the CA root key indication\n extension should be used with care.\n\n The use of the SHA-1 certificate hash alternative ensures that each\n certificate is specified unambiguously. This context does not\n require a cryptographic hash function, so the use of SHA-1 is\n considered acceptable, and no algorithm agility is provided.\n\n11.5. Security Considerations for truncated_hmac\n\n It is possible that truncated MACs are weaker than \"un-truncated\"\n MACs. However, no significant weaknesses are currently known or\n expected to exist for HMAC with MD5 or SHA-1, truncated to 80 bits.\n\n Note that the output length of a MAC need not be as long as the\n length of a symmetric cipher key, since forging of MAC values cannot\n be done off-line: in TLS, a single failed MAC guess will cause the\n immediate termination of the TLS session.\n\n\n\n\nEastlake Standards Track [Page 21]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n Since the MAC algorithm only takes effect after all handshake\n messages that affect extension parameters have been authenticated by\n the hashes in the Finished messages, it is not possible for an active\n attacker to force negotiation of the truncated HMAC extension where\n it would not otherwise be used (to the extent that the handshake\n authentication is secure). Therefore, in the event that any security\n problems were found with truncated HMAC in the future, if either the\n client or the server for a given session were updated to take the\n problem into account, it would be able to veto use of this extension.\n\n11.6. Security Considerations for status_request\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC2560] for further details.\n\n12. Normative References\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC:\n Keyed-Hashing for Message Authentication\", RFC 2104,\n February 1997.\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and\n C. Adams, \"X.509 Internet Public Key Infrastructure\n Online Certificate Status Protocol - OCSP\", RFC 2560,\n June 1999.\n\n [RFC2585] Housley, R. and P. Hoffman, \"Internet X.509 Public Key\n Infrastructure Operational Protocols: FTP and HTTP\",\n RFC 2585, May 1999.\n\n [RFC2616] Fielding, R., Gettys, J., Mogul, J., Frystyk, H.,\n Masinter, L., Leach, P., and T. Berners-Lee,\n \"Hypertext Transfer Protocol -- HTTP/1.1\", RFC 2616,\n June 1999.\n\n [RFC3986] Berners-Lee, T., Fielding, R., and L. Masinter,\n \"Uniform Resource Identifier (URI): Generic Syntax\",\n STD 66, RFC 3986, January 2005.\n\n\n\nEastlake Standards Track [Page 22]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer\n Security (TLS) Protocol Version 1.2\", RFC 5246, August\n 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation\n List (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC5890] Klensin, J., \"Internationalized Domain Names for\n Applications (IDNA): Definitions and Document\n Framework\", RFC 5890, August 2010.\n\n13. Informative References\n\n [RFC2712] Medvinsky, A. and M. Hur, \"Addition of Kerberos Cipher\n Suites to Transport Layer Security (TLS)\", RFC 2712,\n October 1999.\n\n [X509-4th] ITU-T Recommendation X.509 (2000) | ISO/IEC\n 9594-8:2001, \"Information Systems - Open Systems\n Interconnection - The Directory: Public key and\n attribute certificate frameworks\".\n\n [X509-4th-TC1] ITU-T Recommendation X.509(2000) Corrigendum 1(2001) |\n ISO/IEC 9594-8:2001/Cor.1:2002, Technical Corrigendum\n 1 to ISO/IEC 9594:8:2001.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 23]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix A. Changes from RFC 4366\n\n The significant changes between RFC 4366 and this document are\n described below.\n\n RFC 4366 described both general extension mechanisms (for the TLS\n handshake and client and server hellos) as well as specific\n extensions. RFC 4366 was associated with RFC 4346, TLS 1.1. The\n client and server hello extension mechanisms have been moved into RFC\n 5246, TLS 1.2, so this document, which is associated with RFC 5246,\n includes only the handshake extension mechanisms and the specific\n extensions from RFC 4366. RFC 5246 also specifies the unknown\n extension error and new extension specification considerations, so\n that material has been removed from this document.\n\n The Server Name extension now specifies only ASCII representation,\n eliminating UTF-8. It is provided that the ServerNameList can\n contain more than only one name of any particular name_type. If a\n server name is provided but not recognized, the server should either\n continue the handshake without an error or send a fatal error.\n Sending a warning-level message is not recommended because client\n behavior will be unpredictable. Provision was added for the user\n using the server_name extension in deciding whether or not to resume\n a session. Furthermore, this extension should be the same in a\n session resumption request as it was in the full handshake that\n established the session. Such a resumption request must not be\n accepted if the server_name extension is different, but instead a\n full handshake must be done to possibly establish a new session.\n\n The Client Certificate URLs extension has been changed to make the\n presence of a hash mandatory.\n\n For the case of DTLS, the requirement to report an overflow of the\n negotiated maximum fragment length is made conditional on passing\n authentication.\n\n TLS servers are now prohibited from following HTTP redirects when\n retrieving certificates.\n\n The material was also re-organized in minor ways. For example,\n information as to which errors are fatal is moved from the \"Error\n Alerts\" section to the individual extension specifications.\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 24]\n\f\nRFC 6066 TLS Extension Definitions January 2011\n\n\nAppendix B. Acknowledgements\n\n This document is based on material from RFC 4366 for which the\n authors were S. Blake-Wilson, M. Nystrom, D. Hopwood, J. Mikkelsen,\n and T. Wright. Other contributors include Joseph Salowey, Alexey\n Melnikov, Peter Saint-Andre, and Adrian Farrel.\n\nAuthor's Address\n\n Donald Eastlake 3rd\n Huawei\n 155 Beaver Street\n Milford, MA 01757 USA\n\n Phone: +1-508-333-2270\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nEastlake Standards Track [Page 25]", "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}, {"draft_references": [], "excluded_rfc_relation_targets": [], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6961 June 2013", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "6961"}], "kind": "rfc", "occurred_at": "2013-06-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 1, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db", "record_id": "ietf:rfc:6961", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "rfc_number": 6961, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db", "source_url": "https://www.rfc-editor.org/rfc/rfc6961.txt", "temporal_semantics": "source_event_time", "text": "Internet Engineering Task Force (IETF) Y. Pettersen\nRequest for Comments: 6961 June 2013\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\n\nAbstract\n\n This document defines the Transport Layer Security (TLS) Certificate\n Status Version 2 Extension to allow clients to specify and support\n several certificate status methods. (The use of the Certificate\n Status extension is commonly referred to as \"OCSP stapling\".) Also\n defined is a new method based on the Online Certificate Status\n Protocol (OCSP) that servers can use to provide status information\n about not only the server's own certificate but also the status of\n intermediate certificates in the chain.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 5741.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n http://www.rfc-editor.org/info/rfc6961.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 1]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nCopyright Notice\n\n Copyright (c) 2013 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n1. Introduction\n\n The Transport Layer Security (TLS) Extension [RFC6066] framework\n defines, among other extensions, the Certificate Status extension\n (also referred to as \"OCSP stapling\") that clients can use to request\n the server's copy of the current status of its certificate. The\n benefits of this extension include a reduced number of roundtrips and\n network delays for the client to verify the status of the server's\n certificate and a reduced load on the certificate issuer's status\n response servers, thus solving a problem that can become significant\n when the issued certificate is presented by a frequently visited\n server.\n\n There are two problems with the existing Certificate Status\n extension. First, it does not provide functionality to request the\n status information about intermediate Certification Authority (CA)\n certificates, which means the client has to request status\n information through other methods, such as Certificate Revocation\n Lists (CRLs), introducing further delays. Second, the current format\n of the extension and requirements in the TLS protocol prevent a\n client from offering the server multiple status methods.\n\n\n\nPettersen Standards Track [Page 2]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n Many CAs are now issuing intermediate CA certificates that not only\n specify the publication point for their CRLs in a CRL Distribution\n Point [RFC5280] but also specify a URL for their OCSP [RFC6960]\n server in Authority Information Access [RFC5280]. Given that\n client-cached CRLs are frequently out of date, clients would benefit\n from using OCSP to access up-to-date status information about\n intermediate CA certificates. The benefit to the issuing CA is less\n clear, as providing the bandwidth for the OCSP responder can be\n costly, especially for CAs with many high-traffic subscriber sites,\n and this cost is a concern for many CAs. There are cases where OCSP\n requests for a single high-traffic site caused significant network\n problems for the issuing CA.\n\n Clients will benefit from the TLS server providing certificate status\n information regardless of type, not just for the server certificate\n but also for the intermediate CA certificates. Combining the status\n checks into one extension will reduce the roundtrips needed to\n complete the handshake by the client to just those needed for\n negotiating the TLS connection. Also, for the Certification\n Authorities, the load on their servers will depend on the number of\n certificates they have issued, not on the number of visitors to those\n sites. Additionally, using this extension significantly reduces\n privacy concerns around the clients informing the certificate issuer\n about which sites they are visiting.\n\n For such a new system to be introduced seamlessly, clients need to be\n able to indicate support for the existing OCSP Certificate Status\n method and a new multiple-OCSP mode.\n\n Unfortunately, the definition of the Certificate Status extension\n only allows a single Certificate Status extension to be defined in a\n single extension record in the handshake, and the TLS protocol\n [RFC5246] only allows a single record in the extension list for any\n given extension. This means that it is not possible for clients to\n indicate support for new methods while still supporting older\n methods, which would cause problems for interoperability between\n newer clients and older servers. This will not just be an issue for\n the multiple status request mode proposed above but also for any\n other future status methods that might be introduced. This will be\n true not just for the current PKIX infrastructure [RFC5280] but also\n for alternative PKI structures.\n\n The solution to this problem is to define a new extension,\n \"status_request_v2\", with an extended format that allows the client\n to indicate support for multiple status request methods. This is\n implemented using a list of CertificateStatusRequestItemV2 records in\n the extension record. As the server will select the single status\n\n\n\n\nPettersen Standards Track [Page 3]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n method based on the selected cipher suite and the certificate\n presented, no significant changes are needed in the server's\n extension format.\n\n1.1. Requirements Language\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"MAY\", and \"OPTIONAL\" in this\n document are to be interpreted as described in RFC 2119 [RFC2119].\n\n1.2. Presentation Language\n\n This document defines protocol structures using the same conventions\n and presentation language as defined in Section 4 of [RFC5246].\n\n2. Multiple Certificate Status Extension\n\n2.1. New Extension\n\n The extension defined by this document is indicated by\n \"status_request_v2\" in the ExtensionType enum (originally defined by\n [RFC6066]), which uses the following value:\n\n enum {\n status_request_v2(17), (65535)\n } ExtensionType;\n\n2.2. Multiple Certificate Status Request Record\n\n Clients that support a certificate status protocol like OCSP may send\n the \"status_request_v2\" extension to the server in order to use the\n TLS handshake to transfer such data instead of downloading it through\n separate connections. When using this extension, the\n \"extension_data\" field (defined in Section 7.4.1.4 of [RFC5246]) of\n the extension SHALL contain a CertificateStatusRequestListV2 where:\n\n struct {\n CertificateStatusType status_type;\n uint16 request_length; /* Length of request field in bytes */\n select (status_type) {\n case ocsp: OCSPStatusRequest;\n case ocsp_multi: OCSPStatusRequest;\n } request;\n } CertificateStatusRequestItemV2;\n\n enum { ocsp(1), ocsp_multi(2), (255) } CertificateStatusType;\n\n\n\n\n\nPettersen Standards Track [Page 4]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n struct {\n ResponderID responder_id_list<0..2^16-1>;\n Extensions request_extensions;\n } OCSPStatusRequest;\n\n opaque ResponderID<1..2^16-1>;\n opaque Extensions<0..2^16-1>;\n\n struct {\n CertificateStatusRequestItemV2\n certificate_status_req_list<1..2^16-1>;\n } CertificateStatusRequestListV2;\n\n In the OCSPStatusRequest (originally defined by [RFC6066]), the\n \"ResponderID\" provides a list of OCSP responders that the client\n trusts. A zero-length \"responder_id_list\" sequence has the special\n meaning that the responders are implicitly known to the server, e.g.,\n by prior arrangement, or are identified by the certificates used by\n the server. \"Extensions\" is a DER encoding [X.690] of the OCSP\n request extensions, and if the server supports the forwarding of OCSP\n request extensions, this value MUST be forwarded without\n modification.\n\n Both \"ResponderID\" and \"Extensions\" are DER-encoded ASN.1 types as\n defined in [RFC6960]. \"Extensions\" is imported from [RFC5280]. A\n zero-length \"request_extensions\" value means that there are no\n extensions (as opposed to a DER-encoded zero-length ASN.1 SEQUENCE,\n which is not valid for the \"Extensions\" type).\n\n Servers that support a client's selection of responders using the\n ResponderID field could implement this selection by matching the\n responder ID values from the client's list with the ResponderIDs of\n known OCSP responders, either by using a binary compare of the values\n or a hash calculation and compare method.\n\n In the case of the \"id-pkix-ocsp-nonce\" OCSP extension, [RFC2560] is\n unclear about its encoding; for clarification, the nonce MUST be a\n DER-encoded OCTET STRING, which is encapsulated as another OCTET\n STRING (note that implementations based on an existing OCSP client\n will need to be checked for conformance to this requirement). This\n has been clarified in [RFC6960].\n\n The items in the list of CertificateStatusRequestItemV2 entries are\n ordered according to the client's preference (favorite choice first).\n\n A server that receives a client hello containing the\n \"status_request_v2\" extension MAY return a suitable certificate\n status response message to the client along with the server's\n\n\n\nPettersen Standards Track [Page 5]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n certificate message. If OCSP is requested, it SHOULD use the\n information contained in the extension when selecting an OCSP\n responder and SHOULD include request_extensions in the OCSP request.\n\n The server returns a certificate status response along with its\n certificate by sending a \"CertificateStatus\" message (originally\n defined by [RFC6066]) immediately after the \"Certificate\" message\n (Section 7.4.2 of [RFC5246]) (and before any \"ServerKeyExchange\" or\n \"CertificateRequest\" messages). If a server returns a\n \"CertificateStatus\" message in response to a \"status_request_v2\"\n request, then the server MUST have included an extension of type\n \"status_request_v2\" with empty \"extension_data\" in the extended\n server hello.\n\n The \"CertificateStatus\" message is conveyed using the handshake\n message type \"certificate_status\" (defined in [RFC6066]) as follows\n (updated from the definition in [RFC6066]):\n\n struct {\n CertificateStatusType status_type;\n select (status_type) {\n case ocsp: OCSPResponse;\n case ocsp_multi: OCSPResponseList;\n } response;\n } CertificateStatus;\n\n opaque OCSPResponse<0..2^24-1>;\n\n struct {\n OCSPResponse ocsp_response_list<1..2^24-1>;\n } OCSPResponseList;\n\n An \"OCSPResponse\" element (originally defined by [RFC6066]) contains\n a complete, DER-encoded OCSP response (using the ASN.1 [X.680] type\n OCSPResponse defined in [RFC6960]). Only one OCSP response, with a\n length of at least one byte, may be sent for status_type \"ocsp\".\n\n An \"ocsp_response_list\" contains a list of \"OCSPResponse\" elements,\n as specified above, each containing the OCSP response for the\n matching corresponding certificate in the server's Certificate TLS\n handshake message. That is, the first entry is the OCSP response for\n the first certificate in the Certificate list, the second entry is\n the response for the second certificate, and so on. The list MAY\n contain fewer OCSP responses than there were certificates in the\n Certificate handshake message, but there MUST NOT be more responses\n than there were certificates in the list. Individual elements of the\n list MAY have a length of 0 (zero) bytes if the server does not have\n the OCSP response for that particular certificate stored, in which\n\n\n\nPettersen Standards Track [Page 6]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n case the client MUST act as if a response was not received for that\n particular certificate. If the client receives a\n \"ocsp_response_list\" that does not contain a response for one or more\n of the certificates in the completed certificate chain, the client\n SHOULD attempt to validate the certificate using an alternative\n retrieval method, such as downloading the relevant CRL; OCSP SHOULD\n in this situation only be used for the end-entity certificate, not\n intermediate CA certificates, for reasons stated above.\n\n Note that a server MAY also choose not to send a \"CertificateStatus\"\n message, even if it has received a \"status_request_v2\" extension in\n the client hello message and has sent a \"status_request_v2\" extension\n in the server hello message. Additionally, note that a server MUST\n NOT send the \"CertificateStatus\" message unless it received either a\n \"status_request\" or \"status_request_v2\" extension in the client hello\n message and sent a corresponding \"status_request\" or\n \"status_request_v2\" extension in the server hello message.\n\n Clients requesting an OCSP response and receiving one or more OCSP\n responses in a \"CertificateStatus\" message MUST check the OCSP\n response(s) and abort the handshake if the response is a \"revoked\"\n status or other unacceptable responses (as determined by client\n policy) with a bad_certificate_status_response(113) alert. This\n alert is always fatal.\n\n If the OCSP response received from the server does not result in a\n definite \"good\" or \"revoked\" status, it is inconclusive. A TLS\n client in such a case MAY check the validity of the server\n certificate through other means, e.g., by directly querying the\n certificate issuer. If such processing still results in an\n inconclusive response, then the application using the TLS connection\n will have to decide whether to close the connection or not. Note\n that this problem cannot be decided by the generic TLS client code\n without information from the application. If the application doesn't\n provide any such information, then the client MUST abort the\n connection, since the server certificate has not been sufficiently\n validated.\n\n An example of where the application might wish to continue is with\n EAP-TLS (Extensible Authentication Protocol - TLS), where the\n application can use another mechanism to check the status of a\n certificate once it obtains network access. In this case, the\n application could have the client continue with the handshake, but it\n MUST NOT disclose a username and password until it has fully\n validated the server certificate.\n\n\n\n\n\n\nPettersen Standards Track [Page 7]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n3. IANA Considerations\n\n Section 2.1 defines the new TLS extension status_request_v2 (17)\n enum, which has been added to the \"ExtensionType Values\" list in the\n IANA \"Transport Layer Security (TLS) Extensions\" registry.\n\n Section 2.2 describes a TLS CertificateStatusType registry that is\n now maintained by IANA. The \"TLS Certificate Status Types\" registry\n has been created under the \"Transport Layer Security (TLS)\n Extensions\" registry. CertificateStatusType values are to be\n assigned via IETF Review as defined in [RFC5226]. The initial\n registry corresponds to the definition of \"CertificateStatusType\" in\n Section 2.2.\n\n Value Description Reference\n -----------------------------------------\n 0 Reserved [RFC6961]\n 1 ocsp [RFC6066] [RFC6961]\n 2 ocsp_multi [RFC6961]\n 3-255 Unassigned\n\n4. Security Considerations\n\n General security considerations for TLS extensions are covered in\n [RFC5246]. Security considerations for the particular extension\n specified in this document are given below. In general, implementers\n should continue to monitor the state of the art and address any\n weaknesses identified.\n\n4.1. Security Considerations for status_request_v2\n\n If a client requests an OCSP response, it must take into account that\n an attacker's server using a compromised key could (and probably\n would) pretend not to support the extension. In this case, a client\n that requires OCSP validation of certificates SHOULD either contact\n the OCSP server directly or abort the handshake.\n\n Use of the OCSP nonce request extension (id-pkix-ocsp-nonce) may\n improve security against attacks that attempt to replay OCSP\n responses; see Section 4.4.1 of [RFC6960] for further details.\n\n This extension allows the client to send arbitrary data to the\n server. The server implementers need to handle such data carefully\n to avoid introducing security vulnerabilities.\n\n The security considerations of [RFC6960] apply to OCSP requests and\n responses.\n\n\n\n\nPettersen Standards Track [Page 8]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\n5. Acknowledgements\n\n This document is based on [RFC6066], authored by Donald Eastlake 3rd.\n\n6. References\n\n6.1. Normative References\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119, March 1997.\n\n [RFC5226] Narten, T. and H. Alvestrand, \"Guidelines for Writing an\n IANA Considerations Section in RFCs\", BCP 26, RFC 5226,\n May 2008.\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246, August 2008.\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, May 2008.\n\n [RFC6066] Eastlake, D., \"Transport Layer Security (TLS) Extensions:\n Extension Definitions\", RFC 6066, January 2011.\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, June 2013.\n\n [X.680] ITU-T Recommendation X.680 (2008) | ISO/IEC 8824-1:2008,\n \"Abstract Syntax Notation One (ASN.1): Specification of\n basic notation\", November 2008.\n\n [X.690] ITU-T Recommendation X.690 (2008) | ISO/IEC 8825-1:2008,\n \"ASN.1 encoding rules: Specification of Basic Encoding\n Rules (BER), Canonical Encoding Rules (CER) and\n Distinguished Encoding Rules (DER)\", November 2008.\n\n6.2. Informative References\n\n [RFC2560] Myers, M., Ankney, R., Malpani, A., Galperin, S., and C.\n Adams, \"X.509 Internet Public Key Infrastructure Online\n Certificate Status Protocol - OCSP\", RFC 2560, June 1999.\n\n\n\n\n\n\nPettersen Standards Track [Page 9]\n\f\nRFC 6961 Multiple Certificate Status Extension June 2013\n\n\nAuthor's Address\n\n Yngve N. Pettersen\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPettersen Standards Track [Page 10]", "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"draft_name": "draft-ietf-tls-tls13", "facts": [{"char_end": 454, "char_start": 431, "evidence_quote": "draft-ietf-tls-tls13-28", "fact_id": "draft_revision", "field": "draft_revision", "raw_char_end": 458, "raw_char_start": 435, "value": "draft-ietf-tls-tls13-28"}], "kind": "draft_revision", "occurred_at": "2018-03-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 85, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6", "record_id": "ietf:draft:draft-ietf-tls-tls13-28", "retrieval_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "revision": "28", "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6", "source_url": "https://www.ietf.org/archive/id/draft-ietf-tls-tls13-28.txt", "temporal_semantics": "source_event_time", "text": "Network Working Group E. Rescorla\nInternet-Draft RTFM, Inc.\nObsoletes: 5077, 5246, 6961 (if March 20, 2018\n approved)\nUpdates: 4492, 5705, 6066 (if approved)\nIntended status: Standards Track\nExpires: September 21, 2018\n\n\n The Transport Layer Security (TLS) Protocol Version 1.3\n draft-ietf-tls-tls13-28\n\nAbstract\n\n This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol. TLS allows client/server applications to communicate\n over the Internet in a way that is designed to prevent eavesdropping,\n tampering, and message forgery.\n\n This document updates RFCs 4492, 5705, and 6066 and it obsoletes RFCs\n 5077, 5246, and 6961. This document also specifies new requirements\n for TLS 1.2 implementations.\n\nStatus of This Memo\n\n This Internet-Draft is submitted in full conformance with the\n provisions of BCP 78 and BCP 79.\n\n Internet-Drafts are working documents of the Internet Engineering\n Task Force (IETF). Note that other groups may also distribute\n working documents as Internet-Drafts. The list of current Internet-\n Drafts is at http://datatracker.ietf.org/drafts/current/.\n\n Internet-Drafts are draft documents valid for a maximum of six months\n and may be updated, replaced, or obsoleted by other documents at any\n time. It is inappropriate to use Internet-Drafts as reference\n material or to cite them other than as \"work in progress.\"\n\n This Internet-Draft will expire on September 21, 2018.\n\nCopyright Notice\n\n Copyright (c) 2018 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (http://trustee.ietf.org/license-info) in effect on the date of\n\n\n\nRescorla Expires September 21, 2018 [Page 1]\n\f\nInternet-Draft TLS March 2018\n\n\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\nTable of Contents\n\n 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 5\n 1.1. Conventions and Terminology . . . . . . . . . . . . . . . 6\n 1.2. Change Log . . . . . . . . . . . . . . . . . . . . . . . 7\n 1.3. Major Differences from TLS 1.2 . . . . . . . . . . . . . 16\n 1.4. Updates Affecting TLS 1.2 . . . . . . . . . . . . . . . . 17\n 2. Protocol Overview . . . . . . . . . . . . . . . . . . . . . . 18\n 2.1. Incorrect DHE Share . . . . . . . . . . . . . . . . . . . 21\n 2.2. Resumption and Pre-Shared Key (PSK) . . . . . . . . . . . 22\n 2.3. 0-RTT Data . . . . . . . . . . . . . . . . . . . . . . . 24\n 3. Presentation Language . . . . . . . . . . . . . . . . . . . . 26\n 3.1. Basic Block Size . . . . . . . . . . . . . . . . . . . . 26\n 3.2. Miscellaneous . . . . . . . . . . . . . . . . . . . . . . 26\n 3.3. Numbers . . . . . . . . . . . . . . . . . . . . . . . . . 27\n 3.4. Vectors . . . . . . . . . . . . . . . . . . . . . . . . . 27\n 3.5. Enumerateds . . . . . . . . . . . . . . . . . . . . . . . 28\n 3.6. Constructed Types . . . . . . . . . . . . . . . . . . . . 29\n 3.7. Constants . . . . . . . . . . . . . . . . . . . . . . . . 29\n 3.8. Variants . . . . . . . . . . . . . . . . . . . . . . . . 30\n 4. Handshake Protocol . . . . . . . . . . . . . . . . . . . . . 31\n 4.1. Key Exchange Messages . . . . . . . . . . . . . . . . . . 32\n 4.1.1. Cryptographic Negotiation . . . . . . . . . . . . . . 32\n 4.1.2. Client Hello . . . . . . . . . . . . . . . . . . . . 33\n 4.1.3. Server Hello . . . . . . . . . . . . . . . . . . . . 36\n 4.1.4. Hello Retry Request . . . . . . . . . . . . . . . . . 38\n 4.2. Extensions . . . . . . . . . . . . . . . . . . . . . . . 40\n 4.2.1. Supported Versions . . . . . . . . . . . . . . . . . 43\n 4.2.2. Cookie . . . . . . . . . . . . . . . . . . . . . . . 45\n\n\n\nRescorla Expires September 21, 2018 [Page 2]\n\f\nInternet-Draft TLS March 2018\n\n\n 4.2.3. Signature Algorithms . . . . . . . . . . . . . . . . 46\n 4.2.4. Certificate Authorities . . . . . . . . . . . . . . . 50\n 4.2.5. OID Filters . . . . . . . . . . . . . . . . . . . . . 50\n 4.2.6. Post-Handshake Client Authentication . . . . . . . . 51\n 4.2.7. Negotiated Groups . . . . . . . . . . . . . . . . . . 52\n 4.2.8. Key Share . . . . . . . . . . . . . . . . . . . . . . 53\n 4.2.9. Pre-Shared Key Exchange Modes . . . . . . . . . . . . 56\n 4.2.10. Early Data Indication . . . . . . . . . . . . . . . . 57\n 4.2.11. Pre-Shared Key Extension . . . . . . . . . . . . . . 60\n 4.3. Server Parameters . . . . . . . . . . . . . . . . . . . . 63\n 4.3.1. Encrypted Extensions . . . . . . . . . . . . . . . . 63\n 4.3.2. Certificate Request . . . . . . . . . . . . . . . . . 64\n 4.4. Authentication Messages . . . . . . . . . . . . . . . . . 65\n 4.4.1. The Transcript Hash . . . . . . . . . . . . . . . . . 66\n 4.4.2. Certificate . . . . . . . . . . . . . . . . . . . . . 67\n 4.4.3. Certificate Verify . . . . . . . . . . . . . . . . . 72\n 4.4.4. Finished . . . . . . . . . . . . . . . . . . . . . . 74\n 4.5. End of Early Data . . . . . . . . . . . . . . . . . . . . 75\n 4.6. Post-Handshake Messages . . . . . . . . . . . . . . . . . 76\n 4.6.1. New Session Ticket Message . . . . . . . . . . . . . 76\n 4.6.2. Post-Handshake Authentication . . . . . . . . . . . . 78\n 4.6.3. Key and IV Update . . . . . . . . . . . . . . . . . . 79\n 5. Record Protocol . . . . . . . . . . . . . . . . . . . . . . . 80\n 5.1. Record Layer . . . . . . . . . . . . . . . . . . . . . . 81\n 5.2. Record Payload Protection . . . . . . . . . . . . . . . . 83\n 5.3. Per-Record Nonce . . . . . . . . . . . . . . . . . . . . 85\n 5.4. Record Padding . . . . . . . . . . . . . . . . . . . . . 86\n 5.5. Limits on Key Usage . . . . . . . . . . . . . . . . . . . 87\n 6. Alert Protocol . . . . . . . . . . . . . . . . . . . . . . . 87\n 6.1. Closure Alerts . . . . . . . . . . . . . . . . . . . . . 89\n 6.2. Error Alerts . . . . . . . . . . . . . . . . . . . . . . 90\n 7. Cryptographic Computations . . . . . . . . . . . . . . . . . 93\n 7.1. Key Schedule . . . . . . . . . . . . . . . . . . . . . . 93\n 7.2. Updating Traffic Secrets . . . . . . . . . . . . . . . . 96\n 7.3. Traffic Key Calculation . . . . . . . . . . . . . . . . . 97\n 7.4. (EC)DHE Shared Secret Calculation . . . . . . . . . . . . 98\n 7.4.1. Finite Field Diffie-Hellman . . . . . . . . . . . . . 98\n 7.4.2. Elliptic Curve Diffie-Hellman . . . . . . . . . . . . 98\n 7.5. Exporters . . . . . . . . . . . . . . . . . . . . . . . . 99\n 8. 0-RTT and Anti-Replay . . . . . . . . . . . . . . . . . . . . 99\n 8.1. Single-Use Tickets . . . . . . . . . . . . . . . . . . . 101\n 8.2. Client Hello Recording . . . . . . . . . . . . . . . . . 101\n 8.3. Freshness Checks . . . . . . . . . . . . . . . . . . . . 102\n 9. Compliance Requirements . . . . . . . . . . . . . . . . . . . 104\n 9.1. Mandatory-to-Implement Cipher Suites . . . . . . . . . . 104\n 9.2. Mandatory-to-Implement Extensions . . . . . . . . . . . . 104\n 9.3. Protocol Invariants . . . . . . . . . . . . . . . . . . . 105\n 10. Security Considerations . . . . . . . . . . . . . . . . . . . 106\n\n\n\nRescorla Expires September 21, 2018 [Page 3]\n\f\nInternet-Draft TLS March 2018\n\n\n 11. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 107\n 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 108\n 12.1. Normative References . . . . . . . . . . . . . . . . . . 108\n 12.2. Informative References . . . . . . . . . . . . . . . . . 111\n Appendix A. State Machine . . . . . . . . . . . . . . . . . . . 119\n A.1. Client . . . . . . . . . . . . . . . . . . . . . . . . . 119\n A.2. Server . . . . . . . . . . . . . . . . . . . . . . . . . 120\n Appendix B. Protocol Data Structures and Constant Values . . . . 120\n B.1. Record Layer . . . . . . . . . . . . . . . . . . . . . . 121\n B.2. Alert Messages . . . . . . . . . . . . . . . . . . . . . 121\n B.3. Handshake Protocol . . . . . . . . . . . . . . . . . . . 123\n B.3.1. Key Exchange Messages . . . . . . . . . . . . . . . . 123\n B.3.2. Server Parameters Messages . . . . . . . . . . . . . 128\n B.3.3. Authentication Messages . . . . . . . . . . . . . . . 129\n B.3.4. Ticket Establishment . . . . . . . . . . . . . . . . 130\n B.3.5. Updating Keys . . . . . . . . . . . . . . . . . . . . 131\n B.4. Cipher Suites . . . . . . . . . . . . . . . . . . . . . . 131\n Appendix C. Implementation Notes . . . . . . . . . . . . . . . . 132\n C.1. Random Number Generation and Seeding . . . . . . . . . . 132\n C.2. Certificates and Authentication . . . . . . . . . . . . . 133\n C.3. Implementation Pitfalls . . . . . . . . . . . . . . . . . 133\n C.4. Client Tracking Prevention . . . . . . . . . . . . . . . 134\n C.5. Unauthenticated Operation . . . . . . . . . . . . . . . . 135\n Appendix D. Backward Compatibility . . . . . . . . . . . . . . . 135\n D.1. Negotiating with an older server . . . . . . . . . . . . 136\n D.2. Negotiating with an older client . . . . . . . . . . . . 137\n D.3. 0-RTT backwards compatibility . . . . . . . . . . . . . . 137\n D.4. Middlebox Compatibility Mode . . . . . . . . . . . . . . 137\n D.5. Backwards Compatibility Security Restrictions . . . . . . 138\n Appendix E. Overview of Security Properties . . . . . . . . . . 139\n E.1. Handshake . . . . . . . . . . . . . . . . . . . . . . . . 139\n E.1.1. Key Derivation and HKDF . . . . . . . . . . . . . . . 142\n E.1.2. Client Authentication . . . . . . . . . . . . . . . . 143\n E.1.3. 0-RTT . . . . . . . . . . . . . . . . . . . . . . . . 143\n E.1.4. Exporter Independence . . . . . . . . . . . . . . . . 143\n E.1.5. Post-Compromise Security . . . . . . . . . . . . . . 144\n E.1.6. External References . . . . . . . . . . . . . . . . . 144\n E.2. Record Layer . . . . . . . . . . . . . . . . . . . . . . 144\n E.2.1. External References . . . . . . . . . . . . . . . . . 145\n E.3. Traffic Analysis . . . . . . . . . . . . . . . . . . . . 145\n E.4. Side Channel Attacks . . . . . . . . . . . . . . . . . . 146\n E.5. Replay Attacks on 0-RTT . . . . . . . . . . . . . . . . . 147\n E.5.1. Replay and Exporters . . . . . . . . . . . . . . . . 148\n E.6. PSK Identity Exposure . . . . . . . . . . . . . . . . . . 149\n E.7. Attacks on Static RSA . . . . . . . . . . . . . . . . . . 149\n Appendix F. Working Group Information . . . . . . . . . . . . . 149\n Appendix G. Contributors . . . . . . . . . . . . . . . . . . . . 149\n Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 156\n\n\n\nRescorla Expires September 21, 2018 [Page 4]\n\f\nInternet-Draft TLS March 2018\n\n\n1. Introduction\n\n RFC EDITOR: PLEASE REMOVE THE FOLLOWING PARAGRAPH The source for this\n draft is maintained in GitHub. Suggested changes should be submitted\n as pull requests at https://github.com/tlswg/tls13-spec.\n Instructions are on that page as well. Editorial changes can be\n managed in GitHub, but any substantive change should be discussed on\n the TLS mailing list.\n\n The primary goal of TLS is to provide a secure channel between two\n communicating peers; the only requirement from the underlying\n transport is a reliable, in-order, data stream. Specifically, the\n secure channel should provide the following properties:\n\n - Authentication: The server side of the channel is always\n authenticated; the client side is optionally authenticated.\n Authentication can happen via asymmetric cryptography (e.g., RSA\n [RSA], ECDSA [ECDSA], EdDSA [RFC8032]) or a pre-shared key (PSK).\n\n - Confidentiality: Data sent over the channel after establishment is\n only visible to the endpoints. TLS does not hide the length of\n the data it transmits, though endpoints are able to pad TLS\n records in order to obscure lengths and improve protection against\n traffic analysis techniques.\n\n - Integrity: Data sent over the channel after establishment cannot\n be modified by attackers.\n\n These properties should be true even in the face of an attacker who\n has complete control of the network, as described in [RFC3552]. See\n Appendix E for a more complete statement of the relevant security\n properties.\n\n TLS consists of two primary components:\n\n - A handshake protocol (Section 4) that authenticates the\n communicating parties, negotiates cryptographic modes and\n parameters, and establishes shared keying material. The handshake\n protocol is designed to resist tampering; an active attacker\n should not be able to force the peers to negotiate different\n parameters than they would if the connection were not under\n attack.\n\n - A record protocol (Section 5) that uses the parameters established\n by the handshake protocol to protect traffic between the\n communicating peers. The record protocol divides traffic up into\n a series of records, each of which is independently protected\n using the traffic keys.\n\n\n\nRescorla Expires September 21, 2018 [Page 5]\n\f\nInternet-Draft TLS March 2018\n\n\n TLS is application protocol independent; higher-level protocols can\n layer on top of TLS transparently. The TLS standard, however, does\n not specify how protocols add security with TLS; how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n This document defines TLS version 1.3. While TLS 1.3 is not directly\n compatible with previous versions, all versions of TLS incorporate a\n versioning mechanism which allows clients and servers to\n interoperably negotiate a common version if one is supported by both\n peers.\n\n This document supersedes and obsoletes previous versions of TLS\n including version 1.2 [RFC5246]. It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2. Section 4.2.7 updates [RFC4492] by modifying\n the protocol attributes used to negotiate Elliptic Curves. Because\n TLS 1.3 changes the way keys are derived, it updates [RFC5705] as\n described in Section 7.5. It also changes how OCSP messages are\n carried and therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in section Section 4.4.2.1.\n\n1.1. Conventions and Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in BCP\n 14 [RFC2119] [RFC8174] when, and only when, they appear in all\n capitals, as shown here.\n\n The following terms are used:\n\n client: The endpoint initiating the TLS connection.\n\n connection: A transport-layer connection between two endpoints.\n\n endpoint: Either the client or server of the connection.\n\n handshake: An initial negotiation between client and server that\n establishes the parameters of their subsequent interactions within\n TLS.\n\n peer: An endpoint. When discussing a particular endpoint, \"peer\"\n refers to the endpoint that is not the primary subject of discussion.\n\n receiver: An endpoint that is receiving records.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 6]\n\f\nInternet-Draft TLS March 2018\n\n\n sender: An endpoint that is transmitting records.\n\n server: The endpoint which did not initiate the TLS connection.\n\n1.2. Change Log\n\n RFC EDITOR PLEASE DELETE THIS SECTION.\n\n (*) indicates changes to the wire protocol which may require\n implementations to update.\n\n draft-28\n\n Add a section on exposure of PSK identities.\n\n draft-27\n\n - SHOULD->MUST for being able to process \"supported_versions\"\n without 0x0304.\n\n - Much editorial cleanup.\n\n draft-26\n\n - Clarify that you can't negotiate pre-TLS 1.3 with\n supported_versions.\n\n draft-25\n\n - Add the header to additional data (*)\n\n - Minor clarifications.\n\n - IANA cleanup.\n\n draft-24\n\n - Require that CH2 have version 0303 (*)\n\n - Some clarifications\n\n draft-23\n\n - Renumber key_share (*)\n\n - Add a new extension and new code points to allow negotiating PSS\n separately for certificates and CertificateVerify (*)\n\n\n\n\nRescorla Expires September 21, 2018 [Page 7]\n\f\nInternet-Draft TLS March 2018\n\n\n - Slightly restrict when CCS must be accepted to make implementation\n easier.\n\n - Document protocol invariants\n\n - Add some text on the security of static RSA.\n\n draft-22\n\n - Implement changes for improved middlebox penetration (*)\n\n - Move server_certificate_type to encrypted extensions (*)\n\n - Allow resumption with a different SNI (*)\n\n - Padding extension can change on HRR (*)\n\n - Allow an empty ticket_nonce (*)\n\n - Remove requirement to immediately respond to close_notify with\n close_notify (allowing half-close)\n\n draft-21\n\n - Add a per-ticket nonce so that each ticket is associated with a\n different PSK (*).\n\n - Clarify that clients should send alerts with the handshake key if\n possible.\n\n - Update state machine to show rekeying events\n\n - Add discussion of 0-RTT and replay. Recommend that\n implementations implement some anti-replay mechanism.\n\n draft-20\n\n - Add \"post_handshake_auth\" extension to negotiate post-handshake\n authentication (*).\n\n - Shorten labels for HKDF-Expand-Label so that we can fit within one\n compression block (*).\n\n - Define how RFC 7250 works (*).\n\n - Re-enable post-handshake client authentication even when you do\n PSK. The previous prohibition was editorial error.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 8]\n\f\nInternet-Draft TLS March 2018\n\n\n - Remove cert_type and user_mapping, which don't work on TLS 1.3\n anyway.\n\n - Added the no_application_protocol alert from [RFC7301] to the list\n of extensions.\n\n - Added discussion of traffic analysis and side channel attacks.\n\n draft-19\n\n - Hash context_value input to Exporters (*)\n\n - Add an additional Derive-Secret stage to Exporters (*).\n\n - Hash ClientHello1 in the transcript when HRR is used. This\n reduces the state that needs to be carried in cookies. (*)\n\n - Restructure CertificateRequest to have the selectors in\n extensions. This also allowed defining a\n \"certificate_authorities\" extension which can be used by the\n client instead of trusted_ca_keys (*).\n\n - Tighten record framing requirements and require checking of them\n (*).\n\n - Consolidate \"ticket_early_data_info\" and \"early_data\" into a\n single extension (*).\n\n - Change end_of_early_data to be a handshake message (*).\n\n - Add pre-extract Derive-Secret stages to key schedule (*).\n\n - Remove spurious requirement to implement \"pre_shared_key\".\n\n - Clarify location of \"early_data\" from server (it goes in EE, as\n indicated by the table in S 10).\n\n - Require peer public key validation\n\n - Add state machine diagram.\n\n draft-18\n\n - Remove unnecessary resumption_psk which is the only thing expanded\n from the resumption master secret. (*).\n\n - Fix signature_algorithms entry in extensions table.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 9]\n\f\nInternet-Draft TLS March 2018\n\n\n - Restate rule from RFC 6066 that you can't resume unless SNI is the\n same.\n\n draft-17\n\n - Remove 0-RTT Finished and resumption_context, and replace with a\n psk_binder field in the PSK itself (*)\n\n - Restructure PSK key exchange negotiation modes (*)\n\n - Add max_early_data_size field to TicketEarlyDataInfo (*)\n\n - Add a 0-RTT exporter and change the transcript for the regular\n exporter (*)\n\n - Merge TicketExtensions and Extensions registry. Changes\n ticket_early_data_info code point (*)\n\n - Replace Client.key_shares in response to HRR (*)\n\n - Remove redundant labels for traffic key derivation (*)\n\n - Harmonize requirements about cipher suite matching: for resumption\n you need to match KDF but for 0-RTT you need whole cipher suite.\n This allows PSKs to actually negotiate cipher suites. (*)\n\n - Move SCT and OCSP into Certificate.extensions (*)\n\n - Explicitly allow non-offered extensions in NewSessionTicket\n\n - Explicitly allow predicting client Finished for NST\n\n - Clarify conditions for allowing 0-RTT with PSK\n\n draft-16\n\n - Revise version negotiation (*)\n\n - Change RSASSA-PSS and EdDSA SignatureScheme codepoints for better\n backwards compatibility (*)\n\n - Move HelloRetryRequest.selected_group to an extension (*)\n\n - Clarify the behavior of no exporter context and make it the same\n as an empty context.(*)\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 10]\n\f\nInternet-Draft TLS March 2018\n\n\n - New KeyUpdate format that allows for requesting/not-requesting an\n answer. This also means changes to the key schedule to support\n independent updates (*)\n\n - New certificate_required alert (*)\n\n - Forbid CertificateRequest with 0-RTT and PSK.\n\n - Relax requirement to check SNI for 0-RTT.\n\n draft-15\n\n - New negotiation syntax as discussed in Berlin (*)\n\n - Require CertificateRequest.context to be empty during handshake\n (*)\n\n - Forbid empty tickets (*)\n\n - Forbid application data messages in between post-handshake\n messages from the same flight (*)\n\n - Clean up alert guidance (*)\n\n - Clearer guidance on what is needed for TLS 1.2.\n\n - Guidance on 0-RTT time windows.\n\n - Rename a bunch of fields.\n\n - Remove old PRNG text.\n\n - Explicitly require checking that handshake records not span key\n changes.\n\n draft-14\n\n - Allow cookies to be longer (*)\n\n - Remove the \"context\" from EarlyDataIndication as it was undefined\n and nobody used it (*)\n\n - Remove 0-RTT EncryptedExtensions and replace the ticket_age\n extension with an obfuscated version. Also necessitates a change\n to NewSessionTicket (*).\n\n - Move the downgrade sentinel to the end of ServerHello.Random to\n accommodate tlsdate (*).\n\n\n\nRescorla Expires September 21, 2018 [Page 11]\n\f\nInternet-Draft TLS March 2018\n\n\n - Define ecdsa_sha1 (*).\n\n - Allow resumption even after fatal alerts. This matches current\n practice.\n\n - Remove non-closure warning alerts. Require treating unknown\n alerts as fatal.\n\n - Make the rules for accepting 0-RTT less restrictive.\n\n - Clarify 0-RTT backward-compatibility rules.\n\n - Clarify how 0-RTT and PSK identities interact.\n\n - Add a section describing the data limits for each cipher.\n\n - Major editorial restructuring.\n\n - Replace the Security Analysis section with a WIP draft.\n\n draft-13\n\n - Allow server to send SupportedGroups.\n\n - Remove 0-RTT client authentication\n\n - Remove (EC)DHE 0-RTT.\n\n - Flesh out 0-RTT PSK mode and shrink EarlyDataIndication\n\n - Turn PSK-resumption response into an index to save room\n\n - Move CertificateStatus to an extension\n\n - Extra fields in NewSessionTicket.\n\n - Restructure key schedule and add a resumption_context value.\n\n - Require DH public keys and secrets to be zero-padded to the size\n of the group.\n\n - Remove the redundant length fields in KeyShareEntry.\n\n - Define a cookie field for HRR.\n\n draft-12\n\n - Provide a list of the PSK cipher suites.\n\n\n\nRescorla Expires September 21, 2018 [Page 12]\n\f\nInternet-Draft TLS March 2018\n\n\n - Remove the ability for the ServerHello to have no extensions (this\n aligns the syntax with the text).\n\n - Clarify that the server can send application data after its first\n flight (0.5 RTT data)\n\n - Revise signature algorithm negotiation to group hash, signature\n algorithm, and curve together. This is backwards compatible.\n\n - Make ticket lifetime mandatory and limit it to a week.\n\n - Make the purpose strings lower-case. This matches how people are\n implementing for interop.\n\n - Define exporters.\n\n - Editorial cleanup\n\n draft-11\n\n - Port the CFRG curves & signatures work from RFC4492bis.\n\n - Remove sequence number and version from additional_data, which is\n now empty.\n\n - Reorder values in HkdfLabel.\n\n - Add support for version anti-downgrade mechanism.\n\n - Update IANA considerations section and relax some of the policies.\n\n - Unify authentication modes. Add post-handshake client\n authentication.\n\n - Remove early_handshake content type. Terminate 0-RTT data with an\n alert.\n\n - Reset sequence number upon key change (as proposed by Fournet et\n al.)\n\n draft-10\n\n - Remove ClientCertificateTypes field from CertificateRequest and\n add extensions.\n\n - Merge client and server key shares into a single extension.\n\n draft-09\n\n\n\nRescorla Expires September 21, 2018 [Page 13]\n\f\nInternet-Draft TLS March 2018\n\n\n - Change to RSA-PSS signatures for handshake messages.\n\n - Remove support for DSA.\n\n - Update key schedule per suggestions by Hugo, Hoeteck, and Bjoern\n Tackmann.\n\n - Add support for per-record padding.\n\n - Switch to encrypted record ContentType.\n\n - Change HKDF labeling to include protocol version and value\n lengths.\n\n - Shift the final decision to abort a handshake due to incompatible\n certificates to the client rather than having servers abort early.\n\n - Deprecate SHA-1 with signatures.\n\n - Add MTI algorithms.\n\n draft-08\n\n - Remove support for weak and lesser used named curves.\n\n - Remove support for MD5 and SHA-224 hashes with signatures.\n\n - Update lists of available AEAD cipher suites and error alerts.\n\n - Reduce maximum permitted record expansion for AEAD from 2048 to\n 256 octets.\n\n - Require digital signatures even when a previous configuration is\n used.\n\n - Merge EarlyDataIndication and KnownConfiguration.\n\n - Change code point for server_configuration to avoid collision with\n server_hello_done.\n\n - Relax certificate_list ordering requirement to match current\n practice.\n\n draft-07\n\n - Integration of semi-ephemeral DH proposal.\n\n - Add initial 0-RTT support.\n\n\n\nRescorla Expires September 21, 2018 [Page 14]\n\f\nInternet-Draft TLS March 2018\n\n\n - Remove resumption and replace with PSK + tickets.\n\n - Move ClientKeyShare into an extension.\n\n - Move to HKDF.\n\n draft-06\n\n - Prohibit RC4 negotiation for backwards compatibility.\n\n - Freeze & deprecate record layer version field.\n\n - Update format of signatures with context.\n\n - Remove explicit IV.\n\n draft-05\n\n - Prohibit SSL negotiation for backwards compatibility.\n\n - Fix which MS is used for exporters.\n\n draft-04\n\n - Modify key computations to include session hash.\n\n - Remove ChangeCipherSpec.\n\n - Renumber the new handshake messages to be somewhat more consistent\n with existing convention and to remove a duplicate registration.\n\n - Remove renegotiation.\n\n - Remove point format negotiation.\n\n draft-03\n\n - Remove GMT time.\n\n - Merge in support for ECC from RFC 4492 but without explicit\n curves.\n\n - Remove the unnecessary length field from the AD input to AEAD\n ciphers.\n\n - Rename {Client,Server}KeyExchange to {Client,Server}KeyShare.\n\n - Add an explicit HelloRetryRequest to reject the client's.\n\n\n\nRescorla Expires September 21, 2018 [Page 15]\n\f\nInternet-Draft TLS March 2018\n\n\n draft-02\n\n - Increment version number.\n\n - Rework handshake to provide 1-RTT mode.\n\n - Remove custom DHE groups.\n\n - Remove support for compression.\n\n - Remove support for static RSA and DH key exchange.\n\n - Remove support for non-AEAD ciphers.\n\n1.3. Major Differences from TLS 1.2\n\n The following is a list of the major functional differences between\n TLS 1.2 and TLS 1.3. It is not intended to be exhaustive and there\n are many minor differences.\n\n - The list of supported symmetric algorithms has been pruned of all\n algorithms that are considered legacy. Those that remain all use\n Authenticated Encryption with Associated Data (AEAD) algorithms.\n The ciphersuite concept has been changed to separate the\n authentication and key exchange mechanisms from the record\n protection algorithm (including secret key length) and a hash to\n be used with the key derivation function and HMAC.\n\n - A 0-RTT mode was added, saving a round-trip at connection setup\n for some application data, at the cost of certain security\n properties.\n\n - Static RSA and Diffie-Hellman cipher suites have been removed; all\n public-key based key exchange mechanisms now provide forward\n secrecy.\n\n - All handshake messages after the ServerHello are now encrypted.\n The newly introduced EncryptedExtension message allows various\n extensions previously sent in clear in the ServerHello to also\n enjoy confidentiality protection from active attackers.\n\n - The key derivation functions have been re-designed. The new\n design allows easier analysis by cryptographers due to their\n improved key separation properties. The HMAC-based Extract-and-\n Expand Key Derivation Function (HKDF) is used as an underlying\n primitive.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 16]\n\f\nInternet-Draft TLS March 2018\n\n\n - The handshake state machine has been significantly restructured to\n be more consistent and to remove superfluous messages such as\n ChangeCipherSpec (except when needed for middlebox compatibility).\n\n - Elliptic curve algorithms are now in the base spec and new\n signature algorithms, such as ed25519 and ed448, are included.\n TLS 1.3 removed point format negotiation in favor of a single\n point format for each curve.\n\n - Other cryptographic improvements including the removal of\n compression and custom DHE groups, changing the RSA padding to use\n RSASSA-PSS, and the removal of DSA.\n\n - The TLS 1.2 version negotiation mechanism has been deprecated in\n favor of a version list in an extension. This increases\n compatibility with existing servers that incorrectly implemented\n version negotiation.\n\n - Session resumption with and without server-side state as well as\n the PSK-based ciphersuites of earlier TLS versions have been\n replaced by a single new PSK exchange.\n\n - Updated references to point to the updated versions of RFCs, as\n appropriate (e.g., RFC 5280 rather than RFC 3280).\n\n1.4. Updates Affecting TLS 1.2\n\n This document defines several changes that optionally affect\n implementations of TLS 1.2, including those which do not also support\n TLS 1.3:\n\n - A version downgrade protection mechanism is described in\n Section 4.1.3.\n\n - RSASSA-PSS signature schemes are defined in Section 4.2.3.\n\n - The \"supported_versions\" ClientHello extension can be used to\n negotiate the version of TLS to use, in preference to the\n legacy_version field of the ClientHello.\n\n - The \"signature_algorithms_cert\" extension allows a client to\n indicate which signature algorithms it can validate in X.509\n certificates\n\n Additionally, this document clarifies some compliance requirements\n for earlier versions of TLS; see Section 9.3.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 17]\n\f\nInternet-Draft TLS March 2018\n\n\n2. Protocol Overview\n\n The cryptographic parameters used by the secure channel are produced\n by the TLS handshake protocol. This sub-protocol of TLS is used by\n the client and server when first communicating with each other. The\n handshake protocol allows peers to negotiate a protocol version,\n select cryptographic algorithms, optionally authenticate each other,\n and establish shared secret keying material. Once the handshake is\n complete, the peers use the established keys to protect the\n application layer traffic.\n\n A failure of the handshake or other protocol error triggers the\n termination of the connection, optionally preceded by an alert\n message (Section 6).\n\n TLS supports three basic key exchange modes:\n\n - (EC)DHE (Diffie-Hellman over either finite fields or elliptic\n curves)\n\n - PSK-only\n\n - PSK with (EC)DHE\n\n Figure 1 below shows the basic full TLS handshake:\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 18]\n\f\nInternet-Draft TLS March 2018\n\n\n Client Server\n\nKey ^ ClientHello\nExch | + key_share*\n | + signature_algorithms*\n | + psk_key_exchange_modes*\n v + pre_shared_key* -------->\n ServerHello ^ Key\n + key_share* | Exch\n + pre_shared_key* v\n {EncryptedExtensions} ^ Server\n {CertificateRequest*} v Params\n {Certificate*} ^\n {CertificateVerify*} | Auth\n {Finished} v\n <-------- [Application Data*]\n ^ {Certificate*}\nAuth | {CertificateVerify*}\n v {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N\n\n Figure 1: Message flow for full TLS Handshake\n\n The handshake can be thought of as having three phases (indicated in\n the diagram above):\n\n - Key Exchange: Establish shared keying material and select the\n cryptographic parameters. Everything after this phase is\n encrypted.\n\n - Server Parameters: Establish other handshake parameters (whether\n the client is authenticated, application layer protocol support,\n etc.).\n\n - Authentication: Authenticate the server (and optionally the\n client) and provide key confirmation and handshake integrity.\n\n\n\nRescorla Expires September 21, 2018 [Page 19]\n\f\nInternet-Draft TLS March 2018\n\n\n In the Key Exchange phase, the client sends the ClientHello\n (Section 4.1.2) message, which contains a random nonce\n (ClientHello.random); its offered protocol versions; a list of\n symmetric cipher/HKDF hash pairs; either a set of Diffie-Hellman key\n shares (in the \"key_share\" extension Section 4.2.8), a set of pre-\n shared key labels (in the \"pre_shared_key\" extension Section 4.2.11)\n or both; and potentially additional extensions. Additional fields\n and/or messages may also be present for middlebox compatibility.\n\n The server processes the ClientHello and determines the appropriate\n cryptographic parameters for the connection. It then responds with\n its own ServerHello (Section 4.1.3), which indicates the negotiated\n connection parameters. The combination of the ClientHello and the\n ServerHello determines the shared keys. If (EC)DHE key establishment\n is in use, then the ServerHello contains a \"key_share\" extension with\n the server's ephemeral Diffie-Hellman share; the server's share MUST\n be in the same group as one of the client's shares. If PSK key\n establishment is in use, then the ServerHello contains a\n \"pre_shared_key\" extension indicating which of the client's offered\n PSKs was selected. Note that implementations can use (EC)DHE and PSK\n together, in which case both extensions will be supplied.\n\n The server then sends two messages to establish the Server\n Parameters:\n\n EncryptedExtensions: responses to ClientHello extensions that are\n not required to determine the cryptographic parameters, other than\n those that are specific to individual certificates.\n [Section 4.3.1]\n\n CertificateRequest: if certificate-based client authentication is\n desired, the desired parameters for that certificate. This\n message is omitted if client authentication is not desired.\n [Section 4.3.2]\n\n Finally, the client and server exchange Authentication messages. TLS\n uses the same set of messages every time that certificate-based\n authentication is needed. (PSK-based authentication happens as a\n side effect of key exchange.) Specifically:\n\n Certificate: the certificate of the endpoint and any per-certificate\n extensions. This message is omitted by the server if not\n authenticating with a certificate and by the client if the server\n did not send CertificateRequest (thus indicating that the client\n should not authenticate with a certificate). Note that if raw\n public keys [RFC7250] or the cached information extension\n [RFC7924] are in use, then this message will not contain a\n\n\n\n\nRescorla Expires September 21, 2018 [Page 20]\n\f\nInternet-Draft TLS March 2018\n\n\n certificate but rather some other value corresponding to the\n server's long-term key. [Section 4.4.2]\n\n CertificateVerify: a signature over the entire handshake using the\n private key corresponding to the public key in the Certificate\n message. This message is omitted if the endpoint is not\n authenticating via a certificate. [Section 4.4.3]\n\n Finished: a MAC (Message Authentication Code) over the entire\n handshake. This message provides key confirmation, binds the\n endpoint's identity to the exchanged keys, and in PSK mode also\n authenticates the handshake. [Section 4.4.4]\n\n Upon receiving the server's messages, the client responds with its\n Authentication messages, namely Certificate and CertificateVerify (if\n requested), and Finished.\n\n At this point, the handshake is complete, and the client and server\n derive the keying material required by the record layer to exchange\n application-layer data protected through authenticated encryption.\n Application data MUST NOT be sent prior to sending the Finished\n message, except as specified in [Section 2.3]. Note that while the\n server may send application data prior to receiving the client's\n Authentication messages, any data sent at that point is, of course,\n being sent to an unauthenticated peer.\n\n2.1. Incorrect DHE Share\n\n If the client has not provided a sufficient \"key_share\" extension\n (e.g., it includes only DHE or ECDHE groups unacceptable to or\n unsupported by the server), the server corrects the mismatch with a\n HelloRetryRequest and the client needs to restart the handshake with\n an appropriate \"key_share\" extension, as shown in Figure 2. If no\n common cryptographic parameters can be negotiated, the server MUST\n abort the handshake with an appropriate alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 21]\n\f\nInternet-Draft TLS March 2018\n\n\n Client Server\n\n ClientHello\n + key_share -------->\n HelloRetryRequest\n <-------- + key_share\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 2: Message flow for a full handshake with mismatched\n parameters\n\n Note: The handshake transcript incorporates the initial ClientHello/\n HelloRetryRequest exchange; it is not reset with the new ClientHello.\n\n TLS also allows several optimized variants of the basic handshake, as\n described in the following sections.\n\n2.2. Resumption and Pre-Shared Key (PSK)\n\n Although TLS PSKs can be established out of band, PSKs can also be\n established in a previous connection and then used to establish a new\n connection (\"session resumption\" or \"resuming\" with a PSK). Once a\n handshake has completed, the server can send to the client a PSK\n identity that corresponds to a unique key derived from the initial\n handshake (see Section 4.6.1). The client can then use that PSK\n identity in future handshakes to negotiate the use of the associated\n PSK. If the server accepts the PSK, then the security context of the\n new connection is cryptographically tied to the original connection\n and the key derived from the initial handshake is used to bootstrap\n the cryptographic state instead of a full handshake. In TLS 1.2 and\n below, this functionality was provided by \"session IDs\" and \"session\n tickets\" [RFC5077]. Both mechanisms are obsoleted in TLS 1.3.\n\n PSKs can be used with (EC)DHE key exchange in order to provide\n forward secrecy in combination with shared keys, or can be used\n\n\n\nRescorla Expires September 21, 2018 [Page 22]\n\f\nInternet-Draft TLS March 2018\n\n\n alone, at the cost of losing forward secrecy for the application\n data.\n\n Figure 3 shows a pair of handshakes in which the first establishes a\n PSK and the second uses it:\n\n Client Server\n\n Initial Handshake:\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n <-------- [NewSessionTicket]\n [Application Data] <-------> [Application Data]\n\n\n Subsequent Handshake:\n ClientHello\n + key_share*\n + pre_shared_key -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n {Finished}\n <-------- [Application Data*]\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 3: Message flow for resumption and PSK\n\n As the server is authenticating via a PSK, it does not send a\n Certificate or a CertificateVerify message. When a client offers\n resumption via PSK, it SHOULD also supply a \"key_share\" extension to\n the server to allow the server to decline resumption and fall back to\n a full handshake, if needed. The server responds with a\n \"pre_shared_key\" extension to negotiate use of PSK key establishment\n\n\n\n\nRescorla Expires September 21, 2018 [Page 23]\n\f\nInternet-Draft TLS March 2018\n\n\n and can (as shown here) respond with a \"key_share\" extension to do\n (EC)DHE key establishment, thus providing forward secrecy.\n\n When PSKs are provisioned out of band, the PSK identity and the KDF\n hash algorithm to be used with the PSK MUST also be provisioned.\n\n Note: When using an out-of-band provisioned pre-shared secret, a\n critical consideration is using sufficient entropy during the key\n generation, as discussed in [RFC4086]. Deriving a shared secret\n from a password or other low-entropy sources is not secure. A\n low-entropy secret, or password, is subject to dictionary attacks\n based on the PSK binder. The specified PSK authentication is not\n a strong password-based authenticated key exchange even when used\n with Diffie-Hellman key establishment. Specifically, it does not\n prevent an attacker that can observe the handshake from performing\n a brute-force attack on the password/pre-shared key.\n\n2.3. 0-RTT Data\n\n When clients and servers share a PSK (either obtained externally or\n via a previous handshake), TLS 1.3 allows clients to send data on the\n first flight (\"early data\"). The client uses the PSK to authenticate\n the server and to encrypt the early data.\n\n As shown in Figure 4, the 0-RTT data is just added to the 1-RTT\n handshake in the first flight. The rest of the handshake uses the\n same messages as for a 1-RTT handshake with PSK resumption.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 24]\n\f\nInternet-Draft TLS March 2018\n\n\n Client Server\n\n ClientHello\n + early_data\n + key_share*\n + psk_key_exchange_modes\n + pre_shared_key\n (Application Data*) -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n + early_data*\n {Finished}\n <-------- [Application Data*]\n (EndOfEarlyData)\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n () Indicates messages protected using keys\n derived from client_early_traffic_secret.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N\n\n Figure 4: Message flow for a zero round trip handshake\n\n IMPORTANT NOTE: The security properties for 0-RTT data are weaker\n than those for other kinds of TLS data. Specifically:\n\n 1. This data is not forward secret, as it is encrypted solely under\n keys derived using the offered PSK.\n\n 2. There are no guarantees of non-replay between connections.\n Protection against replay for ordinary TLS 1.3 1-RTT data is\n provided via the server's Random value, but 0-RTT data does not\n depend on the ServerHello and therefore has weaker guarantees.\n This is especially relevant if the data is authenticated either\n with TLS client authentication or inside the application\n\n\n\nRescorla Expires September 21, 2018 [Page 25]\n\f\nInternet-Draft TLS March 2018\n\n\n protocol. The same warnings apply to any use of the\n early_exporter_master_secret.\n\n 0-RTT data cannot be duplicated within a connection (i.e., the server\n will not process the same data twice for the same connection) and an\n attacker will not be able to make 0-RTT data appear to be 1-RTT data\n (because it is protected with different keys.) Appendix E.5 contains\n a description of potential attacks and Section 8 describes mechanisms\n which the server can use to limit the impact of replay.\n\n3. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used.\n\n3.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n3.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" double\n brackets.\n\n Single-byte entities containing uninterpreted data are of type\n opaque.\n\n A type alias T' for an existing type T is defined by:\n\n T T';\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 26]\n\f\nInternet-Draft TLS March 2018\n\n\n3.3. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are formed from fixed-length series of bytes\n concatenated as described in Section 3.1 and are also unsigned. The\n following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are transmitted\n in network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n3.4. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* 3 consecutive 3-byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n\n\n\nRescorla Expires September 21, 2018 [Page 27]\n\f\nInternet-Draft TLS March 2018\n\n\n In the following example, mandatory is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 3.3). Similarly,\n longer can represent up to 800 bytes of data, or 400 uint16 elements,\n and it may be empty. Its encoding will include a two-byte actual\n length field prepended to the vector. The length of an encoded\n vector must be an exact multiple of the length of a single element\n (e.g., a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is 2 bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n3.5. Enumerateds\n\n An additional sparse data type is available called enum or\n enumerated. Each definition is a different type. Only enumerateds\n of the same type may be assigned or compared. Every element of an\n enumerated must be assigned a value, as demonstrated in the following\n example. Since the elements of the enumerated are not ordered, they\n can be assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n Future extensions or additions to the protocol may define new values.\n Implementations need to be able to parse and ignore unknown values\n unless the definition of the field states otherwise.\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4 in the current\n version of the protocol.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n\n\n\nRescorla Expires September 21, 2018 [Page 28]\n\f\nInternet-Draft TLS March 2018\n\n\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n The names assigned to enumerateds do not need to be unique. The\n numerical value can describe a range over which the same name\n applies. The value includes the minimum and maximum inclusive values\n in that range, separated by two period characters. This is\n principally useful for reserving regions of the space.\n\n enum { sad(0), meh(1..254), happy(255) } Mood;\n\n3.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax for definition is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } T;\n\n Fixed- and variable-length vector fields are allowed using the\n standard vector syntax. Structures V1 and V2 in the variants example\n below demonstrate this.\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n\n3.7. Constants\n\n Fields and variables may be assigned a fixed value using \"=\", as in:\n\n struct {\n T1 f1 = 8; /* T.f1 must always be 8 */\n T2 f2;\n } T;\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 29]\n\f\nInternet-Draft TLS March 2018\n\n\n3.8. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. Each\n arm of the select specifies the type of that variant's field and an\n optional field label. The mechanism by which the variant is selected\n at runtime is not prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1 [[fe1]];\n case e2: Te2 [[fe2]];\n ....\n case en: Ten [[fen]];\n };\n } Tv;\n\n For example:\n\n enum { apple(0), orange(1) } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n VariantTag type;\n select (VariantRecord.type) {\n case apple: V1;\n case orange: V2;\n };\n } VariantRecord;\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 30]\n\f\nInternet-Draft TLS March 2018\n\n\n4. Handshake Protocol\n\n The handshake protocol is used to negotiate the security parameters\n of a connection. Handshake messages are supplied to the TLS record\n layer, where they are encapsulated within one or more TLSPlaintext or\n TLSCiphertext structures, which are processed and transmitted as\n specified by the current active connection state.\n\n enum {\n client_hello(1),\n server_hello(2),\n new_session_ticket(4),\n end_of_early_data(5),\n encrypted_extensions(8),\n certificate(11),\n certificate_request(13),\n certificate_verify(15),\n finished(20),\n key_update(24),\n message_hash(254),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (Handshake.msg_type) {\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case end_of_early_data: EndOfEarlyData;\n case encrypted_extensions: EncryptedExtensions;\n case certificate_request: CertificateRequest;\n case certificate: Certificate;\n case certificate_verify: CertificateVerify;\n case finished: Finished;\n case new_session_ticket: NewSessionTicket;\n case key_update: KeyUpdate;\n };\n } Handshake;\n\n Protocol messages MUST be sent in the order defined in Section 4.4.1\n and shown in the diagrams in Section 2. A peer which receives a\n handshake message in an unexpected order MUST abort the handshake\n with an \"unexpected_message\" alert.\n\n New handshake message types are assigned by IANA as described in\n Section 11.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 31]\n\f\nInternet-Draft TLS March 2018\n\n\n4.1. Key Exchange Messages\n\n The key exchange messages are used to determine the security\n capabilities of the client and the server and to establish shared\n secrets including the traffic keys used to protect the rest of the\n handshake and the data.\n\n4.1.1. Cryptographic Negotiation\n\n In TLS, the cryptographic negotiation proceeds by the client offering\n the following four sets of options in its ClientHello:\n\n - A list of cipher suites which indicates the AEAD algorithm/HKDF\n hash pairs which the client supports.\n\n - A \"supported_groups\" (Section 4.2.7) extension which indicates the\n (EC)DHE groups which the client supports and a \"key_share\"\n (Section 4.2.8) extension which contains (EC)DHE shares for some\n or all of these groups.\n\n - A \"signature_algorithms\" (Section 4.2.3) extension which indicates\n the signature algorithms which the client can accept.\n\n - A \"pre_shared_key\" (Section 4.2.11) extension which contains a\n list of symmetric key identities known to the client and a\n \"psk_key_exchange_modes\" (Section 4.2.9) extension which indicates\n the key exchange modes that may be used with PSKs.\n\n If the server does not select a PSK, then the first three of these\n options are entirely orthogonal: the server independently selects a\n cipher suite, an (EC)DHE group and key share for key establishment,\n and a signature algorithm/certificate pair to authenticate itself to\n the client. If there is no overlap between the received\n \"supported_groups\" and the groups supported by the server then the\n server MUST abort the handshake with a \"handshake_failure\" or an\n \"insufficient_security\" alert.\n\n If the server selects a PSK, then it MUST also select a key\n establishment mode from the set indicated by client's\n \"psk_key_exchange_modes\" extension (at present, PSK alone or with\n (EC)DHE). Note that if the PSK can be used without (EC)DHE then non-\n overlap in the \"supported_groups\" parameters need not be fatal, as it\n is in the non-PSK case discussed in the previous paragraph.\n\n If the server selects an (EC)DHE group and the client did not offer a\n compatible \"key_share\" extension in the initial ClientHello, the\n server MUST respond with a HelloRetryRequest (Section 4.1.4) message.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 32]\n\f\nInternet-Draft TLS March 2018\n\n\n If the server successfully selects parameters and does not require a\n HelloRetryRequest, it indicates the selected parameters in the\n ServerHello as follows:\n\n - If PSK is being used, then the server will send a \"pre_shared_key\"\n extension indicating the selected key.\n\n - If PSK is not being used, then (EC)DHE and certificate-based\n authentication are always used.\n\n - When (EC)DHE is in use, the server will also provide a \"key_share\"\n extension.\n\n - When authenticating via a certificate, the server will send the\n Certificate (Section 4.4.2) and CertificateVerify (Section 4.4.3)\n messages. In TLS 1.3 as defined by this document, either a PSK or\n a certificate is always used, but not both. Future documents may\n define how to use them together.\n\n If the server is unable to negotiate a supported set of parameters\n (i.e., there is no overlap between the client and server parameters),\n it MUST abort the handshake with either a \"handshake_failure\" or\n \"insufficient_security\" fatal alert (see Section 6).\n\n4.1.2. Client Hello\n\n When a client first connects to a server, it is REQUIRED to send the\n ClientHello as its first TLS message. The client will also send a\n ClientHello when the server has responded to its ClientHello with a\n HelloRetryRequest. In that case, the client MUST send the same\n ClientHello without modification, except:\n\n - If a \"key_share\" extension was supplied in the HelloRetryRequest,\n replacing the list of shares with a list containing a single\n KeyShareEntry from the indicated group.\n\n - Removing the \"early_data\" extension (Section 4.2.10) if one was\n present. Early data is not permitted after HelloRetryRequest.\n\n - Including a \"cookie\" extension if one was provided in the\n HelloRetryRequest.\n\n - Updating the \"pre_shared_key\" extension if present by recomputing\n the \"obfuscated_ticket_age\" and binder values and (optionally)\n removing any PSKs which are incompatible with the server's\n indicated cipher suite.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 33]\n\f\nInternet-Draft TLS March 2018\n\n\n - Optionally adding, removing, or changing the length of the\n \"padding\" extension [RFC7685].\n\n - Other modifications that may be allowed by an extension defined in\n the future and present in the HelloRetryRequest.\n\n Because TLS 1.3 forbids renegotiation, if a server has negotiated TLS\n 1.3 and receives a ClientHello at any other time, it MUST terminate\n the connection with an \"unexpected_message\" alert.\n\n If a server established a TLS connection with a previous version of\n TLS and receives a TLS 1.3 ClientHello in a renegotiation, it MUST\n retain the previous protocol version. In particular, it MUST NOT\n negotiate TLS 1.3.\n\n Structure of this message:\n\n uint16 ProtocolVersion;\n opaque Random[32];\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id<0..32>;\n CipherSuite cipher_suites<2..2^16-2>;\n opaque legacy_compression_methods<1..2^8-1>;\n Extension extensions<8..2^16-1>;\n } ClientHello;\n\n legacy_version In previous versions of TLS, this field was used for\n version negotiation and represented the highest version number\n supported by the client. Experience has shown that many servers\n do not properly implement version negotiation, leading to \"version\n intolerance\" in which the server rejects an otherwise acceptable\n ClientHello with a version number higher than it supports. In TLS\n 1.3, the client indicates its version preferences in the\n \"supported_versions\" extension (Section 4.2.1) and the\n legacy_version field MUST be set to 0x0303, which is the version\n number for TLS 1.2. (See Appendix D for details about backward\n compatibility.)\n\n random 32 bytes generated by a secure random number generator. See\n Appendix C for additional information.\n\n legacy_session_id Versions of TLS before TLS 1.3 supported a\n \"session resumption\" feature which has been merged with Pre-Shared\n\n\n\nRescorla Expires September 21, 2018 [Page 34]\n\f\nInternet-Draft TLS March 2018\n\n\n Keys in this version (see Section 2.2). A client which has a\n cached session ID set by a pre-TLS 1.3 server SHOULD set this\n field to that value. In compatibility mode (see Appendix D.4)\n this field MUST be non-empty, so a client not offering a pre-TLS\n 1.3 session MUST generate a new 32-byte value. This value need\n not be random but SHOULD be unpredictable to avoid implementations\n fixating on a specific value (also known as ossification).\n Otherwise, it MUST be set as a zero length vector (i.e., a single\n zero byte length field).\n\n cipher_suites This is a list of the symmetric cipher options\n supported by the client, specifically the record protection\n algorithm (including secret key length) and a hash to be used with\n HKDF, in descending order of client preference. If the list\n contains cipher suites that the server does not recognize, support\n or wish to use, the server MUST ignore those cipher suites and\n process the remaining ones as usual. Values are defined in\n Appendix B.4. If the client is attempting a PSK key\n establishment, it SHOULD advertise at least one cipher suite\n indicating a Hash associated with the PSK.\n\n legacy_compression_methods Versions of TLS before 1.3 supported\n compression with the list of supported compression methods being\n sent in this field. For every TLS 1.3 ClientHello, this vector\n MUST contain exactly one byte, set to zero, which corresponds to\n the \"null\" compression method in prior versions of TLS. If a TLS\n 1.3 ClientHello is received with any other value in this field,\n the server MUST abort the handshake with an \"illegal_parameter\"\n alert. Note that TLS 1.3 servers might receive TLS 1.2 or prior\n ClientHellos which contain other compression methods and (if\n negotiating such a prior version) MUST follow the procedures for\n the appropriate prior version of TLS. TLS 1.3 ClientHellos are\n identified as having a legacy_version of 0x0303 and a\n supported_versions extension present with 0x0304 as the highest\n version indicated therein.\n\n extensions Clients request extended functionality from servers by\n sending data in the extensions field. The actual \"Extension\"\n format is defined in Section 4.2. In TLS 1.3, use of certain\n extensions is mandatory, as functionality is moved into extensions\n to preserve ClientHello compatibility with previous versions of\n TLS. Servers MUST ignore unrecognized extensions.\n\n All versions of TLS allow an extensions field to optionally follow\n the compression_methods field. TLS 1.3 ClientHello messages always\n contain extensions (minimally \"supported_versions\", otherwise they\n will be interpreted as TLS 1.2 ClientHello messages). However, TLS\n 1.3 servers might receive ClientHello messages without an extensions\n\n\n\nRescorla Expires September 21, 2018 [Page 35]\n\f\nInternet-Draft TLS March 2018\n\n\n field from prior versions of TLS. The presence of extensions can be\n detected by determining whether there are bytes following the\n compression_methods field at the end of the ClientHello. Note that\n this method of detecting optional data differs from the normal TLS\n method of having a variable-length field, but it is used for\n compatibility with TLS before extensions were defined. TLS 1.3\n servers will need to perform this check first and only attempt to\n negotiate TLS 1.3 if the \"supported_versions\" extension is present.\n If negotiating a version of TLS prior to 1.3, a server MUST check\n that the message either contains no data after\n legacy_compression_methods or that it contains a valid extensions\n block with no data following. If not, then it MUST abort the\n handshake with a \"decode_error\" alert.\n\n In the event that a client requests additional functionality using\n extensions, and this functionality is not supplied by the server, the\n client MAY abort the handshake.\n\n After sending the ClientHello message, the client waits for a\n ServerHello or HelloRetryRequest message. If early data is in use,\n the client may transmit early application data (Section 2.3) while\n waiting for the next handshake message.\n\n4.1.3. Server Hello\n\n The server will send this message in response to a ClientHello\n message to proceed with the handshake if it is able to negotiate an\n acceptable set of handshake parameters based on the ClientHello.\n\n Structure of this message:\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id_echo<0..32>;\n CipherSuite cipher_suite;\n uint8 legacy_compression_method = 0;\n Extension extensions<6..2^16-1>;\n } ServerHello;\n\n legacy_version In previous versions of TLS, this field was used for\n version negotiation and represented the selected version number\n for the connection. Unfortunately, some middleboxes fail when\n presented with new values. In TLS 1.3, the TLS server indicates\n its version using the \"supported_versions\" extension\n (Section 4.2.1), and the legacy_version field MUST be set to\n 0x0303, which is the version number for TLS 1.2. (See Appendix D\n for details about backward compatibility.)\n\n\n\nRescorla Expires September 21, 2018 [Page 36]\n\f\nInternet-Draft TLS March 2018\n\n\n random 32 bytes generated by a secure random number generator. See\n Appendix C for additional information. The last eight bytes MUST\n be overwritten as described below if negotiating TLS 1.2 or TLS\n 1.1, but the remaining bytes MUST be random. This structure is\n generated by the server and MUST be generated independently of the\n ClientHello.random.\n\n legacy_session_id_echo The contents of the client's\n legacy_session_id field. Note that this field is echoed even if\n the client's value corresponded to a cached pre-TLS 1.3 session\n which the server has chosen not to resume. A client which\n receives a legacy_session_id_echo field that does not match what\n it sent in the ClientHello MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n cipher_suite The single cipher suite selected by the server from the\n list in ClientHello.cipher_suites. A client which receives a\n cipher suite that was not offered MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n legacy_compression_method A single byte which MUST have the value 0.\n\n extensions A list of extensions. The ServerHello MUST only include\n extensions which are required to establish the cryptographic\n context and negotiate the protocol version. All TLS 1.3\n ServerHello messages MUST contain the \"supported_versions\"\n extension. Current ServerHello messages additionally contain\n either the \"pre_shared_key\" or \"key_share\" extensions, or both\n when using a PSK with (EC)DHE key establishment. Other extensions\n are sent separately in the EncryptedExtensions message.\n\n For reasons of backward compatibility with middleboxes (see\n Appendix D.4) the HelloRetryRequest message uses the same structure\n as the ServerHello, but with Random set to the special value of the\n SHA-256 of \"HelloRetryRequest\":\n\n CF 21 AD 74 E5 9A 61 11 BE 1D 8C 02 1E 65 B8 91\n C2 A2 11 16 7A BB 8C 5E 07 9E 09 E2 C8 A8 33 9C\n\n Upon receiving a message with type server_hello, implementations MUST\n first examine the Random value and if it matches this value, process\n it as described in Section 4.1.4).\n\n TLS 1.3 has a downgrade protection mechanism embedded in the server's\n random value. TLS 1.3 servers which negotiate TLS 1.2 or below in\n response to a ClientHello MUST set the last eight bytes of their\n Random value specially.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 37]\n\f\nInternet-Draft TLS March 2018\n\n\n If negotiating TLS 1.2, TLS 1.3 servers MUST set the last eight bytes\n of their Random value to the bytes:\n\n 44 4F 57 4E 47 52 44 01\n\n If negotiating TLS 1.1 or below, TLS 1.3 servers MUST and TLS 1.2\n servers SHOULD set the last eight bytes of their Random value to the\n bytes:\n\n 44 4F 57 4E 47 52 44 00\n\n TLS 1.3 clients receiving a ServerHello indicating TLS 1.2 or below\n MUST check that the last eight bytes are not equal to either of these\n values. TLS 1.2 clients SHOULD also check that the last eight bytes\n are not equal to the second value if the ServerHello indicates TLS\n 1.1 or below. If a match is found, the client MUST abort the\n handshake with an \"illegal_parameter\" alert. This mechanism provides\n limited protection against downgrade attacks over and above what is\n provided by the Finished exchange: because the ServerKeyExchange, a\n message present in TLS 1.2 and below, includes a signature over both\n random values, it is not possible for an active attacker to modify\n the random values without detection as long as ephemeral ciphers are\n used. It does not provide downgrade protection when static RSA is\n used.\n\n Note: This is a change from [RFC5246], so in practice many TLS 1.2\n clients and servers will not behave as specified above.\n\n A legacy TLS client performing renegotiation with TLS 1.2 or prior\n and which receives a TLS 1.3 ServerHello during renegotiation MUST\n abort the handshake with a \"protocol_version\" alert. Note that\n renegotiation is not possible when TLS 1.3 has been negotiated.\n\n RFC EDITOR: PLEASE REMOVE THE FOLLOWING PARAGRAPH Implementations of\n draft versions (see Section 4.2.1.1) of this specification SHOULD NOT\n implement this mechanism on either client and server. A pre-RFC\n client connecting to RFC servers, or vice versa, will appear to\n downgrade to TLS 1.2. With the mechanism enabled, this will cause an\n interoperability failure.\n\n4.1.4. Hello Retry Request\n\n The server will send this message in response to a ClientHello\n message if it is able to find an acceptable set of parameters but the\n ClientHello does not contain sufficient information to proceed with\n the handshake. As discussed in Section 4.1.3, the HelloRetryRequest\n has the same format as a ServerHello message, and the legacy_version,\n legacy_session_id_echo, cipher_suite, and legacy_compression methods\n\n\n\nRescorla Expires September 21, 2018 [Page 38]\n\f\nInternet-Draft TLS March 2018\n\n\n fields have the same meaning. However, for convenience we discuss\n HelloRetryRequest throughout this document as if it were a distinct\n message.\n\n The server's extensions MUST contain \"supported_versions\" and\n otherwise the server SHOULD send only the extensions necessary for\n the client to generate a correct ClientHello pair. As with\n ServerHello, a HelloRetryRequest MUST NOT contain any extensions that\n were not first offered by the client in its ClientHello, with the\n exception of optionally the \"cookie\" (see Section 4.2.2) extension.\n\n Upon receipt of a HelloRetryRequest, the client MUST check the\n legacy_version, legacy_session_id_echo, cipher_suite, and\n legacy_compression_method as specified in Section 4.1.3 and then\n process the extensions, starting with determining the version using\n \"supported_versions\". Clients MUST abort the handshake with an\n \"illegal_parameter\" alert if the HelloRetryRequest would not result\n in any change in the ClientHello. If a client receives a second\n HelloRetryRequest in the same connection (i.e., where the ClientHello\n was itself in response to a HelloRetryRequest), it MUST abort the\n handshake with an \"unexpected_message\" alert.\n\n Otherwise, the client MUST process all extensions in the\n HelloRetryRequest and send a second updated ClientHello. The\n HelloRetryRequest extensions defined in this specification are:\n\n - supported_versions (see Section 4.2.1)\n\n - cookie (see Section 4.2.2)\n\n - key_share (see Section 4.2.8)\n\n In addition, in its updated ClientHello, the client SHOULD NOT offer\n any pre-shared keys associated with a hash other than that of the\n selected cipher suite. This allows the client to avoid having to\n compute partial hash transcripts for multiple hashes in the second\n ClientHello. A client which receives a cipher suite that was not\n offered MUST abort the handshake. Servers MUST ensure that they\n negotiate the same cipher suite when receiving a conformant updated\n ClientHello (if the server selects the cipher suite as the first step\n in the negotiation, then this will happen automatically). Upon\n receiving the ServerHello, clients MUST check that the cipher suite\n supplied in the ServerHello is the same as that in the\n HelloRetryRequest and otherwise abort the handshake with an\n \"illegal_parameter\" alert.\n\n The value of selected_version in the HelloRetryRequest\n \"supported_versions\" extension MUST be retained in the ServerHello,\n\n\n\nRescorla Expires September 21, 2018 [Page 39]\n\f\nInternet-Draft TLS March 2018\n\n\n and a client MUST abort the handshake with an \"illegal_parameter\"\n alert if the value changes.\n\n4.2. Extensions\n\n A number of TLS messages contain tag-length-value encoded extensions\n structures.\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n server_name(0), /* RFC 6066 */\n max_fragment_length(1), /* RFC 6066 */\n status_request(5), /* RFC 6066 */\n supported_groups(10), /* RFC 4492, 7919 */\n signature_algorithms(13), /* [[this document]] */\n use_srtp(14), /* RFC 5764 */\n heartbeat(15), /* RFC 6520 */\n application_layer_protocol_negotiation(16), /* RFC 7301 */\n signed_certificate_timestamp(18), /* RFC 6962 */\n client_certificate_type(19), /* RFC 7250 */\n server_certificate_type(20), /* RFC 7250 */\n padding(21), /* RFC 7685 */\n pre_shared_key(41), /* [[this document]] */\n early_data(42), /* [[this document]] */\n supported_versions(43), /* [[this document]] */\n cookie(44), /* [[this document]] */\n psk_key_exchange_modes(45), /* [[this document]] */\n certificate_authorities(47), /* [[this document]] */\n oid_filters(48), /* [[this document]] */\n post_handshake_auth(49), /* [[this document]] */\n signature_algorithms_cert(50), /* [[this document]] */\n key_share(51), /* [[this document]] */\n (65535)\n } ExtensionType;\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The list of extension types is maintained by IANA as described in\n Section 11.\n\n\n\nRescorla Expires September 21, 2018 [Page 40]\n\f\nInternet-Draft TLS March 2018\n\n\n Extensions are generally structured in a request/response fashion,\n though some extensions are just indications with no corresponding\n response. The client sends its extension requests in the ClientHello\n message and the server sends its extension responses in the\n ServerHello, EncryptedExtensions, HelloRetryRequest and Certificate\n messages. The server sends extension requests in the\n CertificateRequest message which a client MAY respond to with a\n Certificate message. The server MAY also send unsolicited extensions\n in the NewSessionTicket, though the client does not respond directly\n to these.\n\n Implementations MUST NOT send extension responses if the remote\n endpoint did not send the corresponding extension requests, with the\n exception of the \"cookie\" extension in HelloRetryRequest. Upon\n receiving such an extension, an endpoint MUST abort the handshake\n with an \"unsupported_extension\" alert.\n\n The table below indicates the messages where a given extension may\n appear, using the following notation: CH (ClientHello), SH\n (ServerHello), EE (EncryptedExtensions), CT (Certificate), CR\n (CertificateRequest), NST (NewSessionTicket) and HRR\n (HelloRetryRequest). If an implementation receives an extension\n which it recognizes and which is not specified for the message in\n which it appears it MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 41]\n\f\nInternet-Draft TLS March 2018\n\n\n +--------------------------------------------------+-------------+\n | Extension | TLS 1.3 |\n +--------------------------------------------------+-------------+\n | server_name [RFC6066] | CH, EE |\n | | |\n | max_fragment_length [RFC6066] | CH, EE |\n | | |\n | status_request [RFC6066] | CH, CR, CT |\n | | |\n | supported_groups [RFC7919] | CH, EE |\n | | |\n | signature_algorithms [RFC5246] | CH, CR |\n | | |\n | use_srtp [RFC5764] | CH, EE |\n | | |\n | heartbeat [RFC6520] | CH, EE |\n | | |\n | application_layer_protocol_negotiation [RFC7301] | CH, EE |\n | | |\n | signed_certificate_timestamp [RFC6962] | CH, CR, CT |\n | | |\n | client_certificate_type [RFC7250] | CH, EE |\n | | |\n | server_certificate_type [RFC7250] | CH, EE |\n | | |\n | padding [RFC7685] | CH |\n | | |\n | key_share [[this document]] | CH, SH, HRR |\n | | |\n | pre_shared_key [[this document]] | CH, SH |\n | | |\n | psk_key_exchange_modes [[this document]] | CH |\n | | |\n | early_data [[this document]] | CH, EE, NST |\n | | |\n | cookie [[this document]] | CH, HRR |\n | | |\n | supported_versions [[this document]] | CH, SH, HRR |\n | | |\n | certificate_authorities [[this document]] | CH, CR |\n | | |\n | oid_filters [[this document]] | CR |\n | | |\n | post_handshake_auth [[this document]] | CH |\n | | |\n | signature_algorithms_cert [[this document]] | CH, CR |\n +--------------------------------------------------+-------------+\n\n\n\n\nRescorla Expires September 21, 2018 [Page 42]\n\f\nInternet-Draft TLS March 2018\n\n\n When multiple extensions of different types are present, the\n extensions MAY appear in any order, with the exception of\n \"pre_shared_key\" Section 4.2.11 which MUST be the last extension in\n the ClientHello. There MUST NOT be more than one extension of the\n same type in a given extension block.\n\n In TLS 1.3, unlike TLS 1.2, extensions are negotiated for each\n handshake even when in resumption-PSK mode. However, 0-RTT\n parameters are those negotiated in the previous handshake; mismatches\n may require rejecting 0-RTT (see Section 4.2.10).\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions (e.g., the handshake cannot continue), and some are\n simply refusals to support particular features. In general, error\n alerts should be used for the former and a field in the server\n extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem. Often the fact that the extension fields are\n included in the inputs to the Finished message hashes will be\n sufficient, but extreme care is needed when the extension changes\n the meaning of messages sent in the handshake phase. Designers\n and implementors should be aware of the fact that until the\n handshake has been authenticated, active attackers can modify\n messages and insert, remove, or replace extensions.\n\n4.2.1. Supported Versions\n\n struct {\n select (Handshake.msg_type) {\n case client_hello:\n ProtocolVersion versions<2..254>;\n\n case server_hello: /* and HelloRetryRequest */\n ProtocolVersion selected_version;\n };\n } SupportedVersions;\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 43]\n\f\nInternet-Draft TLS March 2018\n\n\n The \"supported_versions\" extension is used by the client to indicate\n which versions of TLS it supports and by the server to indicate which\n version it is using. The extension contains a list of supported\n versions in preference order, with the most preferred version first.\n Implementations of this specification MUST send this extension in the\n ClientHello containing all versions of TLS which they are prepared to\n negotiate (for this specification, that means minimally 0x0304, but\n if previous versions of TLS are allowed to be negotiated, they MUST\n be present as well).\n\n If this extension is not present, servers which are compliant with\n this specification, and which also support TLS 1.2, MUST negotiate\n TLS 1.2 or prior as specified in [RFC5246], even if\n ClientHello.legacy_version is 0x0304 or later. Servers MAY abort the\n handshake upon receiving a ClientHello with legacy_version 0x0304 or\n later.\n\n If this extension is present in the ClientHello, servers MUST NOT use\n the ClientHello.legacy_version value for version negotiation and MUST\n use only the \"supported_versions\" extension to determine client\n preferences. Servers MUST only select a version of TLS present in\n that extension and MUST ignore any unknown versions that are present\n in that extension. Note that this mechanism makes it possible to\n negotiate a version prior to TLS 1.2 if one side supports a sparse\n range. Implementations of TLS 1.3 which choose to support prior\n versions of TLS SHOULD support TLS 1.2. Servers MUST be prepared to\n receive ClientHellos that include this extension but do not include\n 0x0304 in the list of versions.\n\n A server which negotiates a version of TLS prior to TLS 1.3 MUST set\n ServerHello.version and MUST NOT send the \"supported_versions\"\n extension. A server which negotiates TLS 1.3 MUST respond by sending\n a \"supported_versions\" extension containing the selected version\n value (0x0304). It MUST set the ServerHello.legacy_version field to\n 0x0303 (TLS 1.2). Clients MUST check for this extension prior to\n processing the rest of the ServerHello (although they will have to\n parse the ServerHello in order to read the extension). If this\n extension is present, clients MUST ignore the\n ServerHello.legacy_version value and MUST use only the\n \"supported_versions\" extension to determine the selected version. If\n the \"supported_versions\" extension in the ServerHello contains a\n version not offered by the client or contains a version prior to TLS\n 1.3, the client MUST abort the handshake with an \"illegal_parameter\"\n alert.\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 44]\n\f\nInternet-Draft TLS March 2018\n\n\n4.2.1.1. Draft Version Indicator\n\n RFC EDITOR: PLEASE REMOVE THIS SECTION\n\n While the eventual version indicator for the RFC version of TLS 1.3\n will be 0x0304, implementations of draft versions of this\n specification SHOULD instead advertise 0x7f00 | draft_version in the\n ServerHello and HelloRetryRequest \"supported_versions\" extension.\n For instance, draft-17 would be encoded as the 0x7f11. This allows\n pre-RFC implementations to safely negotiate with each other, even if\n they would otherwise be incompatible.\n\n4.2.2. Cookie\n\n struct {\n opaque cookie<1..2^16-1>;\n } Cookie;\n\n Cookies serve two primary purposes:\n\n - Allowing the server to force the client to demonstrate\n reachability at their apparent network address (thus providing a\n measure of DoS protection). This is primarily useful for non-\n connection-oriented transports (see [RFC6347] for an example of\n this).\n\n - Allowing the server to offload state to the client, thus allowing\n it to send a HelloRetryRequest without storing any state. The\n server can do this by storing the hash of the ClientHello in the\n HelloRetryRequest cookie (protected with some suitable integrity\n algorithm).\n\n When sending a HelloRetryRequest, the server MAY provide a \"cookie\"\n extension to the client (this is an exception to the usual rule that\n the only extensions that may be sent are those that appear in the\n ClientHello). When sending the new ClientHello, the client MUST copy\n the contents of the extension received in the HelloRetryRequest into\n a \"cookie\" extension in the new ClientHello. Clients MUST NOT use\n cookies in their initial ClientHello in subsequent connections.\n\n When a server is operating statelessly it may receive an unprotected\n record of type change_cipher_spec between the first and second\n ClientHello (see Section 5). Since the server is not storing any\n state this will appear as if it were the first message to be\n received. Servers operating statelessly MUST ignore these records.\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 45]\n\f\nInternet-Draft TLS March 2018\n\n\n4.2.3. Signature Algorithms\n\n TLS 1.3 provides two extensions for indicating which signature\n algorithms may be used in digital signatures. The\n \"signature_algorithms_cert\" extension applies to signatures in\n certificates and the \"signature_algorithms\" extension, which\n originally appeared in TLS 1.2, applies to signatures in\n CertificateVerify messages. The keys found in certificates MUST also\n be of appropriate type for the signature algorithms they are used\n with. This is a particular issue for RSA keys and PSS signatures, as\n described below. If no \"signature_algorithms_cert\" extension is\n present, then the \"signature_algorithms\" extension also applies to\n signatures appearing in certificates. Clients which desire the\n server to authenticate itself via a certificate MUST send\n \"signature_algorithms\". If a server is authenticating via a\n certificate and the client has not sent a \"signature_algorithms\"\n extension, then the server MUST abort the handshake with a\n \"missing_extension\" alert (see Section 9.2).\n\n The \"signature_algorithms_cert\" extension was added to allow\n implementations which supported different sets of algorithms for\n certificates and in TLS itself to clearly signal their capabilities.\n TLS 1.2 implementations SHOULD also process this extension.\n Implementations which have the same policy in both cases MAY omit the\n \"signature_algorithms_cert\" extension.\n\n The \"extension_data\" field of these extensions contains a\n SignatureSchemeList value:\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 46]\n\f\nInternet-Draft TLS March 2018\n\n\n enum {\n /* RSASSA-PKCS1-v1_5 algorithms */\n rsa_pkcs1_sha256(0x0401),\n rsa_pkcs1_sha384(0x0501),\n rsa_pkcs1_sha512(0x0601),\n\n /* ECDSA algorithms */\n ecdsa_secp256r1_sha256(0x0403),\n ecdsa_secp384r1_sha384(0x0503),\n ecdsa_secp521r1_sha512(0x0603),\n\n /* RSASSA-PSS algorithms with public key OID rsaEncryption */\n rsa_pss_rsae_sha256(0x0804),\n rsa_pss_rsae_sha384(0x0805),\n rsa_pss_rsae_sha512(0x0806),\n\n /* EdDSA algorithms */\n ed25519(0x0807),\n ed448(0x0808),\n\n /* RSASSA-PSS algorithms with public key OID RSASSA-PSS */\n rsa_pss_pss_sha256(0x0809),\n rsa_pss_pss_sha384(0x080a),\n rsa_pss_pss_sha512(0x080b),\n\n /* Legacy algorithms */\n rsa_pkcs1_sha1(0x0201),\n ecdsa_sha1(0x0203),\n\n /* Reserved Code Points */\n private_use(0xFE00..0xFFFF),\n (0xFFFF)\n } SignatureScheme;\n\n struct {\n SignatureScheme supported_signature_algorithms<2..2^16-2>;\n } SignatureSchemeList;\n\n Note: This enum is named \"SignatureScheme\" because there is already a\n \"SignatureAlgorithm\" type in TLS 1.2, which this replaces. We use\n the term \"signature algorithm\" throughout the text.\n\n Each SignatureScheme value lists a single signature algorithm that\n the client is willing to verify. The values are indicated in\n descending order of preference. Note that a signature algorithm\n takes as input an arbitrary-length message, rather than a digest.\n Algorithms which traditionally act on a digest should be defined in\n TLS to first hash the input with a specified hash algorithm and then\n\n\n\nRescorla Expires September 21, 2018 [Page 47]\n\f\nInternet-Draft TLS March 2018\n\n\n proceed as usual. The code point groups listed above have the\n following meanings:\n\n RSASSA-PKCS1-v1_5 algorithms Indicates a signature algorithm using\n RSASSA-PKCS1-v1_5 [RFC8017] with the corresponding hash algorithm\n as defined in [SHS]. These values refer solely to signatures\n which appear in certificates (see Section 4.4.2.2) and are not\n defined for use in signed TLS handshake messages, although they\n MAY appear in \"signature_algorithms\" and\n \"signature_algorithms_cert\" for backward compatibility with TLS\n 1.2,\n\n ECDSA algorithms Indicates a signature algorithm using ECDSA\n [ECDSA], the corresponding curve as defined in ANSI X9.62 [X962]\n and FIPS 186-4 [DSS], and the corresponding hash algorithm as\n defined in [SHS]. The signature is represented as a DER-encoded\n [X690] ECDSA-Sig-Value structure.\n\n RSASSA-PSS RSAE algorithms Indicates a signature algorithm using\n RSASSA-PSS [RFC8017] with mask generation function 1. The digest\n used in the mask generation function and the digest being signed\n are both the corresponding hash algorithm as defined in [SHS].\n The length of the salt MUST be equal to the length of the output\n of the digest algorithm. If the public key is carried in an X.509\n certificate, it MUST use the rsaEncryption OID [RFC5280].\n\n EdDSA algorithms Indicates a signature algorithm using EdDSA as\n defined in [RFC8032] or its successors. Note that these\n correspond to the \"PureEdDSA\" algorithms and not the \"prehash\"\n variants.\n\n RSASSA-PSS PSS algorithms Indicates a signature algorithm using\n RSASSA-PSS [RFC8017] with mask generation function 1. The digest\n used in the mask generation function and the digest being signed\n are both the corresponding hash algorithm as defined in [SHS].\n The length of the salt MUST be equal to the length of the digest\n algorithm. If the public key is carried in an X.509 certificate,\n it MUST use the RSASSA-PSS OID [RFC5756]. When used in\n certificate signatures, the algorithm parameters MUST be DER\n encoded. If the corresponding public key's parameters are\n present, then the parameters in the signature MUST be identical to\n those in the public key.\n\n Legacy algorithms Indicates algorithms which are being deprecated\n because they use algorithms with known weaknesses, specifically\n SHA-1 which is used in this context with either with RSA using\n RSASSA-PKCS1-v1_5 or ECDSA. These values refer solely to\n signatures which appear in certificates (see Section 4.4.2.2) and\n\n\n\nRescorla Expires September 21, 2018 [Page 48]\n\f\nInternet-Draft TLS March 2018\n\n\n are not defined for use in signed TLS handshake messages, although\n they MAY appear in \"signature_algorithms\" and\n \"signature_algorithms_cert\" for backward compatibility with TLS\n 1.2, Endpoints SHOULD NOT negotiate these algorithms but are\n permitted to do so solely for backward compatibility. Clients\n offering these values MUST list them as the lowest priority\n (listed after all other algorithms in SignatureSchemeList). TLS\n 1.3 servers MUST NOT offer a SHA-1 signed certificate unless no\n valid certificate chain can be produced without it (see\n Section 4.4.2.2).\n\n The signatures on certificates that are self-signed or certificates\n that are trust anchors are not validated since they begin a\n certification path (see [RFC5280], Section 3.2). A certificate that\n begins a certification path MAY use a signature algorithm that is not\n advertised as being supported in the \"signature_algorithms\"\n extension.\n\n Note that TLS 1.2 defines this extension differently. TLS 1.3\n implementations willing to negotiate TLS 1.2 MUST behave in\n accordance with the requirements of [RFC5246] when negotiating that\n version. In particular:\n\n - TLS 1.2 ClientHellos MAY omit this extension.\n\n - In TLS 1.2, the extension contained hash/signature pairs. The\n pairs are encoded in two octets, so SignatureScheme values have\n been allocated to align with TLS 1.2's encoding. Some legacy\n pairs are left unallocated. These algorithms are deprecated as of\n TLS 1.3. They MUST NOT be offered or negotiated by any\n implementation. In particular, MD5 [SLOTH], SHA-224, and DSA MUST\n NOT be used.\n\n - ECDSA signature schemes align with TLS 1.2's ECDSA hash/signature\n pairs. However, the old semantics did not constrain the signing\n curve. If TLS 1.2 is negotiated, implementations MUST be prepared\n to accept a signature that uses any curve that they advertised in\n the \"supported_groups\" extension.\n\n - Implementations that advertise support for RSASSA-PSS (which is\n mandatory in TLS 1.3), MUST be prepared to accept a signature\n using that scheme even when TLS 1.2 is negotiated. In TLS 1.2,\n RSASSA-PSS is used with RSA cipher suites.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 49]\n\f\nInternet-Draft TLS March 2018\n\n\n4.2.4. Certificate Authorities\n\n The \"certificate_authorities\" extension is used to indicate the\n certificate authorities which an endpoint supports and which SHOULD\n be used by the receiving endpoint to guide certificate selection.\n\n The body of the \"certificate_authorities\" extension consists of a\n CertificateAuthoritiesExtension structure.\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n DistinguishedName authorities<3..2^16-1>;\n } CertificateAuthoritiesExtension;\n\n authorities A list of the distinguished names [X501] of acceptable\n certificate authorities, represented in DER-encoded [X690] format.\n These distinguished names specify a desired distinguished name for\n trust anchor or subordinate CA; thus, this message can be used to\n describe known trust anchors as well as a desired authorization\n space.\n\n The client MAY send the \"certificate_authorities\" extension in the\n ClientHello message. The server MAY send it in the\n CertificateRequest message.\n\n The \"trusted_ca_keys\" extension, which serves a similar purpose\n [RFC6066], but is more complicated, is not used in TLS 1.3 (although\n it may appear in ClientHello messages from clients which are offering\n prior versions of TLS).\n\n4.2.5. OID Filters\n\n The \"oid_filters\" extension allows servers to provide a set of OID/\n value pairs which it would like the client's certificate to match.\n This extension, if provided by the server, MUST only be sent in the\n CertificateRequest message.\n\n struct {\n opaque certificate_extension_oid<1..2^8-1>;\n opaque certificate_extension_values<0..2^16-1>;\n } OIDFilter;\n\n struct {\n OIDFilter filters<0..2^16-1>;\n } OIDFilterExtension;\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 50]\n\f\nInternet-Draft TLS March 2018\n\n\n filters A list of certificate extension OIDs [RFC5280] with their\n allowed value(s) and represented in DER-encoded [X690] format.\n Some certificate extension OIDs allow multiple values (e.g.,\n Extended Key Usage). If the server has included a non-empty\n filters list, the client certificate included in the response MUST\n contain all of the specified extension OIDs that the client\n recognizes. For each extension OID recognized by the client, all\n of the specified values MUST be present in the client certificate\n (but the certificate MAY have other values as well). However, the\n client MUST ignore and skip any unrecognized certificate extension\n OIDs. If the client ignored some of the required certificate\n extension OIDs and supplied a certificate that does not satisfy\n the request, the server MAY at its discretion either continue the\n connection without client authentication, or abort the handshake\n with an \"unsupported_certificate\" alert. Any given OID MUST NOT\n appear more than once in the filters list.\n\n PKIX RFCs define a variety of certificate extension OIDs and their\n corresponding value types. Depending on the type, matching\n certificate extension values are not necessarily bitwise-equal. It\n is expected that TLS implementations will rely on their PKI libraries\n to perform certificate selection using certificate extension OIDs.\n\n This document defines matching rules for two standard certificate\n extensions defined in [RFC5280]:\n\n - The Key Usage extension in a certificate matches the request when\n all key usage bits asserted in the request are also asserted in\n the Key Usage certificate extension.\n\n - The Extended Key Usage extension in a certificate matches the\n request when all key purpose OIDs present in the request are also\n found in the Extended Key Usage certificate extension. The\n special anyExtendedKeyUsage OID MUST NOT be used in the request.\n\n Separate specifications may define matching rules for other\n certificate extensions.\n\n4.2.6. Post-Handshake Client Authentication\n\n The \"post_handshake_auth\" extension is used to indicate that a client\n is willing to perform post-handshake authentication (Section 4.6.2).\n Servers MUST NOT send a post-handshake CertificateRequest to clients\n which do not offer this extension. Servers MUST NOT send this\n extension.\n\n struct {} PostHandshakeAuth;\n\n\n\n\nRescorla Expires September 21, 2018 [Page 51]\n\f\nInternet-Draft TLS March 2018\n\n\n The \"extension_data\" field of the \"post_handshake_auth\" extension is\n zero length.\n\n4.2.7. Negotiated Groups\n\n When sent by the client, the \"supported_groups\" extension indicates\n the named groups which the client supports for key exchange, ordered\n from most preferred to least preferred.\n\n Note: In versions of TLS prior to TLS 1.3, this extension was named\n \"elliptic_curves\" and only contained elliptic curve groups. See\n [RFC4492] and [RFC7919]. This extension was also used to negotiate\n ECDSA curves. Signature algorithms are now negotiated independently\n (see Section 4.2.3).\n\n The \"extension_data\" field of this extension contains a\n \"NamedGroupList\" value:\n\n enum {\n\n /* Elliptic Curve Groups (ECDHE) */\n secp256r1(0x0017), secp384r1(0x0018), secp521r1(0x0019),\n x25519(0x001D), x448(0x001E),\n\n /* Finite Field Groups (DHE) */\n ffdhe2048(0x0100), ffdhe3072(0x0101), ffdhe4096(0x0102),\n ffdhe6144(0x0103), ffdhe8192(0x0104),\n\n /* Reserved Code Points */\n ffdhe_private_use(0x01FC..0x01FF),\n ecdhe_private_use(0xFE00..0xFEFF),\n (0xFFFF)\n } NamedGroup;\n\n struct {\n NamedGroup named_group_list<2..2^16-1>;\n } NamedGroupList;\n\n Elliptic Curve Groups (ECDHE) Indicates support for the\n corresponding named curve, defined either in FIPS 186-4 [DSS] or\n in [RFC7748]. Values 0xFE00 through 0xFEFF are reserved for\n private use.\n\n Finite Field Groups (DHE) Indicates support of the corresponding\n finite field group, defined in [RFC7919]. Values 0x01FC through\n 0x01FF are reserved for private use.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 52]\n\f\nInternet-Draft TLS March 2018\n\n\n Items in named_group_list are ordered according to the client's\n preferences (most preferred choice first).\n\n As of TLS 1.3, servers are permitted to send the \"supported_groups\"\n extension to the client. Clients MUST NOT act upon any information\n found in \"supported_groups\" prior to successful completion of the\n handshake but MAY use the information learned from a successfully\n completed handshake to change what groups they use in their\n \"key_share\" extension in subsequent connections. If the server has a\n group it prefers to the ones in the \"key_share\" extension but is\n still willing to accept the ClientHello, it SHOULD send\n \"supported_groups\" to update the client's view of its preferences;\n this extension SHOULD contain all groups the server supports,\n regardless of whether they are currently supported by the client.\n\n4.2.8. Key Share\n\n The \"key_share\" extension contains the endpoint's cryptographic\n parameters.\n\n Clients MAY send an empty client_shares vector in order to request\n group selection from the server at the cost of an additional round\n trip. (see Section 4.1.4)\n\n struct {\n NamedGroup group;\n opaque key_exchange<1..2^16-1>;\n } KeyShareEntry;\n\n group The named group for the key being exchanged.\n\n key_exchange Key exchange information. The contents of this field\n are determined by the specified group and its corresponding\n definition. Finite Field Diffie-Hellman [DH] parameters are\n described in Section 4.2.8.1; Elliptic Curve Diffie-Hellman\n parameters are described in Section 4.2.8.2.\n\n In the ClientHello message, the \"extension_data\" field of this\n extension contains a \"KeyShareClientHello\" value:\n\n struct {\n KeyShareEntry client_shares<0..2^16-1>;\n } KeyShareClientHello;\n\n client_shares A list of offered KeyShareEntry values in descending\n order of client preference.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 53]\n\f\nInternet-Draft TLS March 2018\n\n\n This vector MAY be empty if the client is requesting a\n HelloRetryRequest. Each KeyShareEntry value MUST correspond to a\n group offered in the \"supported_groups\" extension and MUST appear in\n the same order. However, the values MAY be a non-contiguous subset\n of the \"supported_groups\" extension and MAY omit the most preferred\n groups. Such a situation could arise if the most preferred groups\n are new and unlikely to be supported in enough places to make\n pregenerating key shares for them efficient.\n\n Clients can offer as many KeyShareEntry values as the number of\n supported groups it is offering, each representing a single set of\n key exchange parameters. For instance, a client might offer shares\n for several elliptic curves or multiple FFDHE groups. The\n key_exchange values for each KeyShareEntry MUST be generated\n independently. Clients MUST NOT offer multiple KeyShareEntry values\n for the same group. Clients MUST NOT offer any KeyShareEntry values\n for groups not listed in the client's \"supported_groups\" extension.\n Servers MAY check for violations of these rules and abort the\n handshake with an \"illegal_parameter\" alert if one is violated.\n\n In a HelloRetryRequest message, the \"extension_data\" field of this\n extension contains a KeyShareHelloRetryRequest value:\n\n struct {\n NamedGroup selected_group;\n } KeyShareHelloRetryRequest;\n\n selected_group The mutually supported group the server intends to\n negotiate and is requesting a retried ClientHello/KeyShare for.\n\n Upon receipt of this extension in a HelloRetryRequest, the client\n MUST verify that (1) the selected_group field corresponds to a group\n which was provided in the \"supported_groups\" extension in the\n original ClientHello; and (2) the selected_group field does not\n correspond to a group which was provided in the \"key_share\" extension\n in the original ClientHello. If either of these checks fails, then\n the client MUST abort the handshake with an \"illegal_parameter\"\n alert. Otherwise, when sending the new ClientHello, the client MUST\n replace the original \"key_share\" extension with one containing only a\n new KeyShareEntry for the group indicated in the selected_group field\n of the triggering HelloRetryRequest.\n\n In a ServerHello message, the \"extension_data\" field of this\n extension contains a KeyShareServerHello value:\n\n struct {\n KeyShareEntry server_share;\n } KeyShareServerHello;\n\n\n\nRescorla Expires September 21, 2018 [Page 54]\n\f\nInternet-Draft TLS March 2018\n\n\n server_share A single KeyShareEntry value that is in the same group\n as one of the client's shares.\n\n If using (EC)DHE key establishment, servers offer exactly one\n KeyShareEntry in the ServerHello. This value MUST be in the same\n group as the KeyShareEntry value offered by the client that the\n server has selected for the negotiated key exchange. Servers MUST\n NOT send a KeyShareEntry for any group not indicated in the\n \"supported_groups\" extension and MUST NOT send a KeyShareEntry when\n using the \"psk_ke\" PskKeyExchangeMode. If using (EC)DHE key\n establishment, and a HelloRetryRequest containing a \"key_share\"\n extension was received by the client, the client MUST verify that the\n selected NamedGroup in the ServerHello is the same as that in the\n HelloRetryRequest. If this check fails, the client MUST abort the\n handshake with an \"illegal_parameter\" alert.\n\n4.2.8.1. Diffie-Hellman Parameters\n\n Diffie-Hellman [DH] parameters for both clients and servers are\n encoded in the opaque key_exchange field of a KeyShareEntry in a\n KeyShare structure. The opaque value contains the Diffie-Hellman\n public value (Y = g^X mod p) for the specified group (see [RFC7919]\n for group definitions) encoded as a big-endian integer and padded to\n the left with zeros to the size of p in bytes.\n\n Note: For a given Diffie-Hellman group, the padding results in all\n public keys having the same length.\n\n Peers MUST validate each other's public key Y by ensuring that 1 < Y\n < p-1. This check ensures that the remote peer is properly behaved\n and isn't forcing the local system into a small subgroup.\n\n4.2.8.2. ECDHE Parameters\n\n ECDHE parameters for both clients and servers are encoded in the\n opaque key_exchange field of a KeyShareEntry in a KeyShare structure.\n\n For secp256r1, secp384r1 and secp521r1, the contents are the\n serialized value of the following struct:\n\n struct {\n uint8 legacy_form = 4;\n opaque X[coordinate_length];\n opaque Y[coordinate_length];\n } UncompressedPointRepresentation;\n\n X and Y respectively are the binary representations of the x and y\n values in network byte order. There are no internal length markers,\n\n\n\nRescorla Expires September 21, 2018 [Page 55]\n\f\nInternet-Draft TLS March 2018\n\n\n so each number representation occupies as many octets as implied by\n the curve parameters. For P-256 this means that each of X and Y use\n 32 octets, padded on the left by zeros if necessary. For P-384 they\n take 48 octets each, and for P-521 they take 66 octets each.\n\n For the curves secp256r1, secp384r1 and secp521r1, peers MUST\n validate each other's public value Q by ensuring that the point is a\n valid point on the elliptic curve. The appropriate validation\n procedures are defined in Section 4.3.7 of [X962] and alternatively\n in Section 5.6.2.3 of [KEYAGREEMENT]. This process consists of three\n steps: (1) verify that Q is not the point at infinity (O), (2) verify\n that for Q = (x, y) both integers x and y are in the correct\n interval, (3) ensure that (x, y) is a correct solution to the\n elliptic curve equation. For these curves, implementers do not need\n to verify membership in the correct subgroup.\n\n For X25519 and X448, the contents of the public value are the byte\n string inputs and outputs of the corresponding functions defined in\n [RFC7748], 32 bytes for X25519 and 56 bytes for X448.\n\n Note: Versions of TLS prior to 1.3 permitted point format\n negotiation; TLS 1.3 removes this feature in favor of a single point\n format for each curve.\n\n4.2.9. Pre-Shared Key Exchange Modes\n\n In order to use PSKs, clients MUST also send a\n \"psk_key_exchange_modes\" extension. The semantics of this extension\n are that the client only supports the use of PSKs with these modes,\n which restricts both the use of PSKs offered in this ClientHello and\n those which the server might supply via NewSessionTicket.\n\n A client MUST provide a \"psk_key_exchange_modes\" extension if it\n offers a \"pre_shared_key\" extension. If clients offer\n \"pre_shared_key\" without a \"psk_key_exchange_modes\" extension,\n servers MUST abort the handshake. Servers MUST NOT select a key\n exchange mode that is not listed by the client. This extension also\n restricts the modes for use with PSK resumption; servers SHOULD NOT\n send NewSessionTicket with tickets that are not compatible with the\n advertised modes; however, if a server does so, the impact will just\n be that the client's attempts at resumption fail.\n\n The server MUST NOT send a \"psk_key_exchange_modes\" extension.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 56]\n\f\nInternet-Draft TLS March 2018\n\n\n enum { psk_ke(0), psk_dhe_ke(1), (255) } PskKeyExchangeMode;\n\n struct {\n PskKeyExchangeMode ke_modes<1..255>;\n } PskKeyExchangeModes;\n\n psk_ke PSK-only key establishment. In this mode, the server MUST\n NOT supply a \"key_share\" value.\n\n psk_dhe_ke PSK with (EC)DHE key establishment. In this mode, the\n client and server MUST supply \"key_share\" values as described in\n Section 4.2.8.\n\n Any future values that are allocated must ensure that the transmitted\n protocol messages unambiguously identify which mode was selected by\n the server; at present, this is indicated by the presence of the\n \"key_share\" in the ServerHello.\n\n4.2.10. Early Data Indication\n\n When a PSK is used and early data is allowed for that PSK, the client\n can send application data in its first flight of messages. If the\n client opts to do so, it MUST supply both the \"early_data\" extension\n as well as the \"pre_shared_key\" extension.\n\n The \"extension_data\" field of this extension contains an\n \"EarlyDataIndication\" value.\n\n struct {} Empty;\n\n struct {\n select (Handshake.msg_type) {\n case new_session_ticket: uint32 max_early_data_size;\n case client_hello: Empty;\n case encrypted_extensions: Empty;\n };\n } EarlyDataIndication;\n\n See Section 4.6.1 for the use of the max_early_data_size field.\n\n The parameters for the 0-RTT data (version, symmetric cipher suite,\n ALPN protocol, etc.) are those associated with the PSK in use. For\n externally provisioned PSKs, the associated values are those\n provisioned along with the key. For PSKs established via a\n NewSessionTicket message, the associated values are those which were\n negotiated in the connection which established the PSK. The PSK used\n to encrypt the early data MUST be the first PSK listed in the\n client's \"pre_shared_key\" extension.\n\n\n\nRescorla Expires September 21, 2018 [Page 57]\n\f\nInternet-Draft TLS March 2018\n\n\n For PSKs provisioned via NewSessionTicket, a server MUST validate\n that the ticket age for the selected PSK identity (computed by\n subtracting ticket_age_add from PskIdentity.obfuscated_ticket_age\n modulo 2^32) is within a small tolerance of the time since the ticket\n was issued (see Section 8). If it is not, the server SHOULD proceed\n with the handshake but reject 0-RTT, and SHOULD NOT take any other\n action that assumes that this ClientHello is fresh.\n\n 0-RTT messages sent in the first flight have the same (encrypted)\n content types as messages of the same type sent in other flights\n (handshake and application_data) but are protected under different\n keys. After receiving the server's Finished message, if the server\n has accepted early data, an EndOfEarlyData message will be sent to\n indicate the key change. This message will be encrypted with the\n 0-RTT traffic keys.\n\n A server which receives an \"early_data\" extension MUST behave in one\n of three ways:\n\n - Ignore the extension and return a regular 1-RTT response. The\n server then skips past early data by attempting to deprotect\n received records using the handshake traffic key, discarding\n records which fail deprotection (up to the configured\n max_early_data_size). Once a record is deprotected successfully,\n it is treated as the start of the client's second flight and the\n the server proceeds as with an ordinary 1-RTT handshake.\n\n - Request that the client send another ClientHello by responding\n with a HelloRetryRequest. A client MUST NOT include the\n \"early_data\" extension in its followup ClientHello. The server\n then ignores early data by skipping all records with external\n content type of \"application_data\" (indicating that they are\n encrypted), up to the configured max_early_data_size.\n\n - Return its own \"early_data\" extension in EncryptedExtensions,\n indicating that it intends to process the early data. It is not\n possible for the server to accept only a subset of the early data\n messages. Even though the server sends a message accepting early\n data, the actual early data itself may already be in flight by the\n time the server generates this message.\n\n In order to accept early data, the server MUST have accepted a PSK\n cipher suite and selected the first key offered in the client's\n \"pre_shared_key\" extension. In addition, it MUST verify that the\n following values are the same as those associated with the selected\n PSK:\n\n - The TLS version number\n\n\n\nRescorla Expires September 21, 2018 [Page 58]\n\f\nInternet-Draft TLS March 2018\n\n\n - The selected cipher suite\n\n - The selected ALPN [RFC7301] protocol, if any\n\n These requirements are a superset of those needed to perform a 1-RTT\n handshake using the PSK in question. For externally established\n PSKs, the associated values are those provisioned along with the key.\n For PSKs established via a NewSessionTicket message, the associated\n values are those negotiated in the connection during which the ticket\n was established.\n\n Future extensions MUST define their interaction with 0-RTT.\n\n If any of these checks fail, the server MUST NOT respond with the\n extension and must discard all the first flight data using one of the\n first two mechanisms listed above (thus falling back to 1-RTT or\n 2-RTT). If the client attempts a 0-RTT handshake but the server\n rejects it, the server will generally not have the 0-RTT record\n protection keys and must instead use trial decryption (either with\n the 1-RTT handshake keys or by looking for a cleartext ClientHello in\n the case of HelloRetryRequest) to find the first non-0-RTT message.\n\n If the server chooses to accept the \"early_data\" extension, then it\n MUST comply with the same error handling requirements specified for\n all records when processing early data records. Specifically, if the\n server fails to decrypt a 0-RTT record following an accepted\n \"early_data\" extension it MUST terminate the connection with a\n \"bad_record_mac\" alert as per Section 5.2.\n\n If the server rejects the \"early_data\" extension, the client\n application MAY opt to retransmit the application data previously\n sent in early data once the handshake has been completed. Note that\n automatic re-transmission of early data could result in assumptions\n about the status of the connection being incorrect. For instance,\n when the negotiated connection selects a different ALPN protocol from\n what was used for the early data, an application might need to\n construct different messages. Similarly, if early data assumes\n anything about the connection state, it might be sent in error after\n the handshake completes.\n\n A TLS implementation SHOULD NOT automatically re-send early data;\n applications are in a better position to decide when re-transmission\n is appropriate. A TLS implementation MUST NOT automatically re-send\n early data unless the negotiated connection selects the same ALPN\n protocol.\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 59]\n\f\nInternet-Draft TLS March 2018\n\n\n4.2.11. Pre-Shared Key Extension\n\n The \"pre_shared_key\" extension is used to negotiate the identity of\n the pre-shared key to be used with a given handshake in association\n with PSK key establishment.\n\n The \"extension_data\" field of this extension contains a\n \"PreSharedKeyExtension\" value:\n\n struct {\n opaque identity<1..2^16-1>;\n uint32 obfuscated_ticket_age;\n } PskIdentity;\n\n opaque PskBinderEntry<32..255>;\n\n struct {\n PskIdentity identities<7..2^16-1>;\n PskBinderEntry binders<33..2^16-1>;\n } OfferedPsks;\n\n struct {\n select (Handshake.msg_type) {\n case client_hello: OfferedPsks;\n case server_hello: uint16 selected_identity;\n };\n } PreSharedKeyExtension;\n\n identity A label for a key. For instance, a ticket defined in\n Appendix B.3.4 or a label for a pre-shared key established\n externally.\n\n obfuscated_ticket_age An obfuscated version of the age of the key.\n Section 4.2.11.1 describes how to form this value for identities\n established via the NewSessionTicket message. For identities\n established externally an obfuscated_ticket_age of 0 SHOULD be\n used, and servers MUST ignore the value.\n\n identities A list of the identities that the client is willing to\n negotiate with the server. If sent alongside the \"early_data\"\n extension (see Section 4.2.10), the first identity is the one used\n for 0-RTT data.\n\n binders A series of HMAC values, one for each PSK offered in the\n \"pre_shared_keys\" extension and in the same order, computed as\n described below.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 60]\n\f\nInternet-Draft TLS March 2018\n\n\n selected_identity The server's chosen identity expressed as a\n (0-based) index into the identities in the client's list.\n\n Each PSK is associated with a single Hash algorithm. For PSKs\n established via the ticket mechanism (Section 4.6.1), this is the KDF\n Hash algorithm on the connection where the ticket was established.\n For externally established PSKs, the Hash algorithm MUST be set when\n the PSK is established, or default to SHA-256 if no such algorithm is\n defined. The server MUST ensure that it selects a compatible PSK (if\n any) and cipher suite.\n\n In TLS versions prior to TLS 1.3, the Server Name Identification\n (SNI) value was intended to be associated with the session (Section 3\n of [RFC6066]), with the server being required to enforce that the SNI\n value associated with the session matches the one specified in the\n resumption handshake. However, in reality the implementations were\n not consistent on which of two supplied SNI values they would use,\n leading to the consistency requirement being de-facto enforced by the\n clients. In TLS 1.3, the SNI value is always explicitly specified in\n the resumption handshake, and there is no need for the server to\n associate an SNI value with the ticket. Clients, however, SHOULD\n store the SNI with the PSK to fulfill the requirements of\n Section 4.6.1.\n\n Implementor's note: when session resumption is the primary use case\n of PSKs the most straightforward way to implement the PSK/cipher\n suite matching requirements is to negotiate the cipher suite first\n and then exclude any incompatible PSKs. Any unknown PSKs (e.g., they\n are not in the PSK database or are encrypted with an unknown key)\n SHOULD simply be ignored. If no acceptable PSKs are found, the\n server SHOULD perform a non-PSK handshake if possible. If backwards\n compatibility is important, client provided, externally established\n PSKs SHOULD influence cipher suite selection.\n\n Prior to accepting PSK key establishment, the server MUST validate\n the corresponding binder value (see Section 4.2.11.2 below). If this\n value is not present or does not validate, the server MUST abort the\n handshake. Servers SHOULD NOT attempt to validate multiple binders;\n rather they SHOULD select a single PSK and validate solely the binder\n that corresponds to that PSK. See [Section 8.2] and [Appendix E.6]\n for the security rationale for this requirement. In order to accept\n PSK key establishment, the server sends a \"pre_shared_key\" extension\n indicating the selected identity.\n\n Clients MUST verify that the server's selected_identity is within the\n range supplied by the client, that the server selected a cipher suite\n indicating a Hash associated with the PSK and that a server\n \"key_share\" extension is present if required by the ClientHello\n\n\n\nRescorla Expires September 21, 2018 [Page 61]\n\f\nInternet-Draft TLS March 2018\n\n\n \"psk_key_exchange_modes\". If these values are not consistent the\n client MUST abort the handshake with an \"illegal_parameter\" alert.\n\n If the server supplies an \"early_data\" extension, the client MUST\n verify that the server's selected_identity is 0. If any other value\n is returned, the client MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n The \"pre_shared_key\" extension MUST be the last extension in the\n ClientHello (this facilitates implementation as described below).\n Servers MUST check that it is the last extension and otherwise fail\n the handshake with an \"illegal_parameter\" alert.\n\n4.2.11.1. Ticket Age\n\n The client's view of the age of a ticket is the time since the\n receipt of the NewSessionTicket message. Clients MUST NOT attempt to\n use tickets which have ages greater than the \"ticket_lifetime\" value\n which was provided with the ticket. The \"obfuscated_ticket_age\"\n field of each PskIdentity contains an obfuscated version of the\n ticket age formed by taking the age in milliseconds and adding the\n \"ticket_age_add\" value that was included with the ticket (see\n Section 4.6.1), modulo 2^32. This addition prevents passive\n observers from correlating connections unless tickets are reused.\n Note that the \"ticket_lifetime\" field in the NewSessionTicket message\n is in seconds but the \"obfuscated_ticket_age\" is in milliseconds.\n Because ticket lifetimes are restricted to a week, 32 bits is enough\n to represent any plausible age, even in milliseconds.\n\n4.2.11.2. PSK Binder\n\n The PSK binder value forms a binding between a PSK and the current\n handshake, as well as a binding between the handshake in which the\n PSK was generated (if via a NewSessionTicket message) and the current\n handshake. Each entry in the binders list is computed as an HMAC\n over a transcript hash (see Section 4.4.1) containing a partial\n ClientHello up to and including the PreSharedKeyExtension.identities\n field. That is, it includes all of the ClientHello but not the\n binders list itself. The length fields for the message (including\n the overall length, the length of the extensions block, and the\n length of the \"pre_shared_key\" extension) are all set as if binders\n of the correct lengths were present.\n\n The PskBinderEntry is computed in the same way as the Finished\n message (Section 4.4.4) but with the BaseKey being the binder_key\n derived via the key schedule from the corresponding PSK which is\n being offered (see Section 7.1).\n\n\n\n\nRescorla Expires September 21, 2018 [Page 62]\n\f\nInternet-Draft TLS March 2018\n\n\n If the handshake includes a HelloRetryRequest, the initial\n ClientHello and HelloRetryRequest are included in the transcript\n along with the new ClientHello. For instance, if the client sends\n ClientHello1, its binder will be computed over:\n\n Transcript-Hash(Truncate(ClientHello1))\n\n Where Truncate() removes the binders list from the ClientHello.\n\n If the server responds with HelloRetryRequest, and the client then\n sends ClientHello2, its binder will be computed over:\n\n Transcript-Hash(ClientHello1,\n HelloRetryRequest,\n Truncate(ClientHello2))\n\n The full ClientHello1/ClientHello2 is included in all other handshake\n hash computations. Note that in the first flight,\n Truncate(ClientHello1) is hashed directly, but in the second flight,\n ClientHello1 is hashed and then reinjected as a \"message_hash\"\n message, as described in Section 4.4.1.\n\n4.2.11.3. Processing Order\n\n Clients are permitted to \"stream\" 0-RTT data until they receive the\n server's Finished, only then sending the EndOfEarlyData message,\n followed by the rest of the handshake. In order to avoid deadlocks,\n when accepting \"early_data\", servers MUST process the client's\n ClientHello and then immediately send their flight of messages,\n rather than waiting for the client's EndOfEarlyData message before\n sending its ServerHello.\n\n4.3. Server Parameters\n\n The next two messages from the server, EncryptedExtensions and\n CertificateRequest, contain information from the server that\n determines the rest of the handshake. These messages are encrypted\n with keys derived from the server_handshake_traffic_secret.\n\n4.3.1. Encrypted Extensions\n\n In all handshakes, the server MUST send the EncryptedExtensions\n message immediately after the ServerHello message. This is the first\n message that is encrypted under keys derived from the\n server_handshake_traffic_secret.\n\n The EncryptedExtensions message contains extensions that can be\n protected, i.e., any which are not needed to establish the\n\n\n\nRescorla Expires September 21, 2018 [Page 63]\n\f\nInternet-Draft TLS March 2018\n\n\n cryptographic context, but which are not associated with individual\n certificates. The client MUST check EncryptedExtensions for the\n presence of any forbidden extensions and if any are found MUST abort\n the handshake with an \"illegal_parameter\" alert.\n\n Structure of this message:\n\n struct {\n Extension extensions<0..2^16-1>;\n } EncryptedExtensions;\n\n extensions A list of extensions. For more information, see the\n table in Section 4.2.\n\n4.3.2. Certificate Request\n\n A server which is authenticating with a certificate MAY optionally\n request a certificate from the client. This message, if sent, MUST\n follow EncryptedExtensions.\n\n Structure of this message:\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n Extension extensions<2..2^16-1>;\n } CertificateRequest;\n\n certificate_request_context An opaque string which identifies the\n certificate request and which will be echoed in the client's\n Certificate message. The certificate_request_context MUST be\n unique within the scope of this connection (thus preventing replay\n of client CertificateVerify messages). This field SHALL be zero\n length unless used for the post-handshake authentication exchanges\n described in Section 4.6.2. When requesting post-handshake\n authentication, the server SHOULD make the context unpredictable\n to the client (e.g., by randomly generating it) in order to\n prevent an attacker who has temporary access to the client's\n private key from pre-computing valid CertificateVerify messages.\n\n extensions A set of extensions describing the parameters of the\n certificate being requested. The \"signature_algorithms\" extension\n MUST be specified, and other extensions may optionally be included\n if defined for this message. Clients MUST ignore unrecognized\n extensions.\n\n In prior versions of TLS, the CertificateRequest message carried a\n list of signature algorithms and certificate authorities which the\n server would accept. In TLS 1.3 the former is expressed by sending\n\n\n\nRescorla Expires September 21, 2018 [Page 64]\n\f\nInternet-Draft TLS March 2018\n\n\n the \"signature_algorithms\" and optionally \"signature_algorithms_cert\"\n extensions. The latter is expressed by sending the\n \"certificate_authorities\" extension (see Section 4.2.4).\n\n Servers which are authenticating with a PSK MUST NOT send the\n CertificateRequest message in the main handshake, though they MAY\n send it in post-handshake authentication (see Section 4.6.2) provided\n that the client has sent the \"post_handshake_auth\" extension (see\n Section 4.2.6).\n\n4.4. Authentication Messages\n\n As discussed in Section 2, TLS generally uses a common set of\n messages for authentication, key confirmation, and handshake\n integrity: Certificate, CertificateVerify, and Finished. (The\n PreSharedKey binders also perform key confirmation, in a similar\n fashion.) These three messages are always sent as the last messages\n in their handshake flight. The Certificate and CertificateVerify\n messages are only sent under certain circumstances, as defined below.\n The Finished message is always sent as part of the Authentication\n block. These messages are encrypted under keys derived from\n [sender]_handshake_traffic_secret.\n\n The computations for the Authentication messages all uniformly take\n the following inputs:\n\n - The certificate and signing key to be used.\n\n - A Handshake Context consisting of the set of messages to be\n included in the transcript hash.\n\n - A base key to be used to compute a MAC key.\n\n Based on these inputs, the messages then contain:\n\n Certificate The certificate to be used for authentication, and any\n supporting certificates in the chain. Note that certificate-based\n client authentication is not available in PSK (including 0-RTT)\n flows.\n\n CertificateVerify A signature over the value Transcript-\n Hash(Handshake Context, Certificate)\n\n Finished A MAC over the value Transcript-Hash(Handshake Context,\n Certificate, CertificateVerify) using a MAC key derived from the\n base key.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 65]\n\f\nInternet-Draft TLS March 2018\n\n\n The following table defines the Handshake Context and MAC Base Key\n for each scenario:\n\n +-----------+----------------------------+--------------------------+\n | Mode | Handshake Context | Base Key |\n +-----------+----------------------------+--------------------------+\n | Server | ClientHello ... later of E | server_handshake_traffic |\n | | ncryptedExtensions/Certifi | _secret |\n | | cateRequest | |\n | | | |\n | Client | ClientHello ... later of | client_handshake_traffic |\n | | server | _secret |\n | | Finished/EndOfEarlyData | |\n | | | |\n | Post- | ClientHello ... client | client_application_traff |\n | Handshake | Finished + | ic_secret_N |\n | | CertificateRequest | |\n +-----------+----------------------------+--------------------------+\n\n4.4.1. The Transcript Hash\n\n Many of the cryptographic computations in TLS make use of a\n transcript hash. This value is computed by hashing the concatenation\n of each included handshake message, including the handshake message\n header carrying the handshake message type and length fields, but not\n including record layer headers. I.e.,\n\n Transcript-Hash(M1, M2, ... Mn) = Hash(M1 || M2 || ... || Mn)\n\n As an exception to this general rule, when the server responds to a\n ClientHello with a HelloRetryRequest, the value of ClientHello1 is\n replaced with a special synthetic handshake message of handshake type\n \"message_hash\" containing Hash(ClientHello1). I.e.,\n\n Transcript-Hash(ClientHello1, HelloRetryRequest, ... Mn) =\n Hash(message_hash || /* Handshake type */\n 00 00 Hash.length || /* Handshake message length (bytes) */\n Hash(ClientHello1) || /* Hash of ClientHello1 */\n HelloRetryRequest || ... || Mn)\n\n The reason for this construction is to allow the server to do a\n stateless HelloRetryRequest by storing just the hash of ClientHello1\n in the cookie, rather than requiring it to export the entire\n intermediate hash state (see Section 4.2.2).\n\n For concreteness, the transcript hash is always taken from the\n following sequence of handshake messages, starting at the first\n ClientHello and including only those messages that were sent:\n\n\n\nRescorla Expires September 21, 2018 [Page 66]\n\f\nInternet-Draft TLS March 2018\n\n\n ClientHello, HelloRetryRequest, ClientHello, ServerHello,\n EncryptedExtensions, server CertificateRequest, server Certificate,\n server CertificateVerify, server Finished, EndOfEarlyData, client\n Certificate, client CertificateVerify, client Finished.\n\n In general, implementations can implement the transcript by keeping a\n running transcript hash value based on the negotiated hash. Note,\n however, that subsequent post-handshake authentications do not\n include each other, just the messages through the end of the main\n handshake.\n\n4.4.2. Certificate\n\n This message conveys the endpoint's certificate chain to the peer.\n\n The server MUST send a Certificate message whenever the agreed-upon\n key exchange method uses certificates for authentication (this\n includes all key exchange methods defined in this document except\n PSK).\n\n The client MUST send a Certificate message if and only if the server\n has requested client authentication via a CertificateRequest message\n (Section 4.3.2). If the server requests client authentication but no\n suitable certificate is available, the client MUST send a Certificate\n message containing no certificates (i.e., with the \"certificate_list\"\n field having length 0). A Finished message MUST be sent regardless\n of whether the Certificate message is empty.\n\n Structure of this message:\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 67]\n\f\nInternet-Draft TLS March 2018\n\n\n /* Managed by IANA */\n enum {\n X509(0),\n RawPublicKey(2),\n (255)\n } CertificateType;\n\n struct {\n select (certificate_type) {\n case RawPublicKey:\n /* From RFC 7250 ASN.1_subjectPublicKeyInfo */\n opaque ASN1_subjectPublicKeyInfo<1..2^24-1>;\n\n case X509:\n opaque cert_data<1..2^24-1>;\n };\n Extension extensions<0..2^16-1>;\n } CertificateEntry;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n CertificateEntry certificate_list<0..2^24-1>;\n } Certificate;\n\n certificate_request_context If this message is in response to a\n CertificateRequest, the value of certificate_request_context in\n that message. Otherwise (in the case of server authentication),\n this field SHALL be zero length.\n\n certificate_list This is a sequence (chain) of CertificateEntry\n structures, each containing a single certificate and set of\n extensions.\n\n extensions: A set of extension values for the CertificateEntry. The\n \"Extension\" format is defined in Section 4.2. Valid extensions\n for server certificates at present include OCSP Status extension\n ([RFC6066]) and SignedCertificateTimestamps ([RFC6962]); future\n extensions may be defined for this message as well. Extensions in\n the Certificate message from the server MUST correspond to ones\n from the ClientHello message. Extensions in the Certificate from\n the client MUST correspond with extensions in the\n CertificateRequest message from the server. If an extension\n applies to the entire chain, it SHOULD be included in the first\n CertificateEntry.\n\n If the corresponding certificate type extension\n (\"server_certificate_type\" or \"client_certificate_type\") was not\n negotiated in Encrypted Extensions, or the X.509 certificate type was\n\n\n\nRescorla Expires September 21, 2018 [Page 68]\n\f\nInternet-Draft TLS March 2018\n\n\n negotiated, then each CertificateEntry contains a DER-encoded X.509\n certificate. The sender's certificate MUST come in the first\n CertificateEntry in the list. Each following certificate SHOULD\n directly certify the one immediately preceding it. Because\n certificate validation requires that trust anchors be distributed\n independently, a certificate that specifies a trust anchor MAY be\n omitted from the chain, provided that supported peers are known to\n possess any omitted certificates.\n\n Note: Prior to TLS 1.3, \"certificate_list\" ordering required each\n certificate to certify the one immediately preceding it; however,\n some implementations allowed some flexibility. Servers sometimes\n send both a current and deprecated intermediate for transitional\n purposes, and others are simply configured incorrectly, but these\n cases can nonetheless be validated properly. For maximum\n compatibility, all implementations SHOULD be prepared to handle\n potentially extraneous certificates and arbitrary orderings from any\n TLS version, with the exception of the end-entity certificate which\n MUST be first.\n\n If the RawPublicKey certificate type was negotiated, then the\n certificate_list MUST contain no more than one CertificateEntry,\n which contains an ASN1_subjectPublicKeyInfo value as defined in\n [RFC7250], Section 3.\n\n The OpenPGP certificate type [RFC6091] MUST NOT be used with TLS 1.3.\n\n The server's certificate_list MUST always be non-empty. A client\n will send an empty certificate_list if it does not have an\n appropriate certificate to send in response to the server's\n authentication request.\n\n4.4.2.1. OCSP Status and SCT Extensions\n\n [RFC6066] and [RFC6961] provide extensions to negotiate the server\n sending OCSP responses to the client. In TLS 1.2 and below, the\n server replies with an empty extension to indicate negotiation of\n this extension and the OCSP information is carried in a\n CertificateStatus message. In TLS 1.3, the server's OCSP information\n is carried in an extension in the CertificateEntry containing the\n associated certificate. Specifically: The body of the\n \"status_request\" extension from the server MUST be a\n CertificateStatus structure as defined in [RFC6066], which is\n interpreted as defined in [RFC6960].\n\n Note: status_request_v2 extension ([RFC6961]) is deprecated. TLS 1.3\n servers MUST NOT act upon its presence or information in it when\n processing Client Hello, in particular they MUST NOT send the\n\n\n\nRescorla Expires September 21, 2018 [Page 69]\n\f\nInternet-Draft TLS March 2018\n\n\n status_request_v2 extension in the Encrypted Extensions, Certificate\n Request or the Certificate messages. TLS 1.3 servers MUST be able to\n process Client Hello messages that include it, as it MAY be sent by\n clients that wish to use it in earlier protocol versions.\n\n A server MAY request that a client present an OCSP response with its\n certificate by sending an empty \"status_request\" extension in its\n CertificateRequest message. If the client opts to send an OCSP\n response, the body of its \"status_request\" extension MUST be a\n CertificateStatus structure as defined in [RFC6066].\n\n Similarly, [RFC6962] provides a mechanism for a server to send a\n Signed Certificate Timestamp (SCT) as an extension in the ServerHello\n in TLS 1.2 and below. In TLS 1.3, the server's SCT information is\n carried in an extension in CertificateEntry.\n\n4.4.2.2. Server Certificate Selection\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3 [RFC5280], unless explicitly\n negotiated otherwise (e.g., [RFC7250]).\n\n - The server's end-entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected authentication\n algorithm from the client's \"signature_algorithms\" extension\n (currently RSA, ECDSA, or EdDSA).\n\n - The certificate MUST allow the key to be used for signing (i.e.,\n the digitalSignature bit MUST be set if the Key Usage extension is\n present) with a signature scheme indicated in the client's\n \"signature_algorithms\"/\"signature_algorithms_cert\" extensions (see\n Section 4.2.3).\n\n - The \"server_name\" [RFC6066] and \"certificate_authorities\"\n extensions are used to guide certificate selection. As servers\n MAY require the presence of the \"server_name\" extension, clients\n SHOULD send this extension, when applicable.\n\n All certificates provided by the server MUST be signed by a signature\n algorithm advertised by the client, if it is able to provide such a\n chain (see Section 4.2.3). Certificates that are self-signed or\n certificates that are expected to be trust anchors are not validated\n as part of the chain and therefore MAY be signed with any algorithm.\n\n If the server cannot produce a certificate chain that is signed only\n via the indicated supported algorithms, then it SHOULD continue the\n handshake by sending the client a certificate chain of its choice\n\n\n\nRescorla Expires September 21, 2018 [Page 70]\n\f\nInternet-Draft TLS March 2018\n\n\n that may include algorithms that are not known to be supported by the\n client. This fallback chain SHOULD NOT use the deprecated SHA-1 hash\n algorithm in general, but MAY do so if the client's advertisement\n permits it, and MUST NOT do so otherwise.\n\n If the client cannot construct an acceptable chain using the provided\n certificates and decides to abort the handshake, then it MUST abort\n the handshake with an appropriate certificate-related alert (by\n default, \"unsupported_certificate\"; see Section 6.2 for more).\n\n If the server has multiple certificates, it chooses one of them based\n on the above-mentioned criteria (in addition to other criteria, such\n as transport layer endpoint, local configuration and preferences).\n\n4.4.2.3. Client Certificate Selection\n\n The following rules apply to certificates sent by the client:\n\n - The certificate type MUST be X.509v3 [RFC5280], unless explicitly\n negotiated otherwise (e.g., [RFC7250]).\n\n - If the \"certificate_authorities\" extension in the\n CertificateRequest message was present, at least one of the\n certificates in the certificate chain SHOULD be issued by one of\n the listed CAs.\n\n - The certificates MUST be signed using an acceptable signature\n algorithm, as described in Section 4.3.2. Note that this relaxes\n the constraints on certificate-signing algorithms found in prior\n versions of TLS.\n\n - If the CertificateRequest message contained a non-empty\n \"oid_filters\" extension, the end-entity certificate MUST match the\n extension OIDs that are recognized by the client, as described in\n Section 4.2.5.\n\n Note that, as with the server certificate, there are certificates\n that use algorithm combinations that cannot be currently used with\n TLS.\n\n4.4.2.4. Receiving a Certificate Message\n\n In general, detailed certificate validation procedures are out of\n scope for TLS (see [RFC5280]). This section provides TLS-specific\n requirements.\n\n If the server supplies an empty Certificate message, the client MUST\n abort the handshake with a \"decode_error\" alert.\n\n\n\nRescorla Expires September 21, 2018 [Page 71]\n\f\nInternet-Draft TLS March 2018\n\n\n If the client does not send any certificates (i.e., it sends an empty\n Certificate message), the server MAY at its discretion either\n continue the handshake without client authentication, or abort the\n handshake with a \"certificate_required\" alert. Also, if some aspect\n of the certificate chain was unacceptable (e.g., it was not signed by\n a known, trusted CA), the server MAY at its discretion either\n continue the handshake (considering the client unauthenticated) or\n abort the handshake.\n\n Any endpoint receiving any certificate which it would need to\n validate using any signature algorithm using an MD5 hash MUST abort\n the handshake with a \"bad_certificate\" alert. SHA-1 is deprecated\n and it is RECOMMENDED that any endpoint receiving any certificate\n which it would need to validate using any signature algorithm using a\n SHA-1 hash abort the handshake with a \"bad_certificate\" alert. For\n clarity, this means that endpoints MAY accept these algorithms for\n certificates that are self-signed or are trust anchors.\n\n All endpoints are RECOMMENDED to transition to SHA-256 or better as\n soon as possible to maintain interoperability with implementations\n currently in the process of phasing out SHA-1 support.\n\n Note that a certificate containing a key for one signature algorithm\n MAY be signed using a different signature algorithm (for instance, an\n RSA key signed with an ECDSA key).\n\n4.4.3. Certificate Verify\n\n This message is used to provide explicit proof that an endpoint\n possesses the private key corresponding to its certificate. The\n CertificateVerify message also provides integrity for the handshake\n up to this point. Servers MUST send this message when authenticating\n via a certificate. Clients MUST send this message whenever\n authenticating via a certificate (i.e., when the Certificate message\n is non-empty). When sent, this message MUST appear immediately after\n the Certificate message and immediately prior to the Finished\n message.\n\n Structure of this message:\n\n struct {\n SignatureScheme algorithm;\n opaque signature<0..2^16-1>;\n } CertificateVerify;\n\n The algorithm field specifies the signature algorithm used (see\n Section 4.2.3 for the definition of this field). The signature is a\n digital signature using that algorithm. The content that is covered\n\n\n\nRescorla Expires September 21, 2018 [Page 72]\n\f\nInternet-Draft TLS March 2018\n\n\n under the signature is the hash output as described in Section 4.4.1,\n namely:\n\n Transcript-Hash(Handshake Context, Certificate)\n\n The digital signature is then computed over the concatenation of:\n\n - A string that consists of octet 32 (0x20) repeated 64 times\n\n - The context string\n\n - A single 0 byte which serves as the separator\n\n - The content to be signed\n\n This structure is intended to prevent an attack on previous versions\n of TLS in which the ServerKeyExchange format meant that attackers\n could obtain a signature of a message with a chosen 32-byte prefix\n (ClientHello.random). The initial 64-byte pad clears that prefix\n along with the server-controlled ServerHello.random.\n\n The context string for a server signature is: \"TLS 1.3, server\n CertificateVerify\" The context string for a client signature is: \"TLS\n 1.3, client CertificateVerify\" It is used to provide separation\n between signatures made in different contexts, helping against\n potential cross-protocol attacks.\n\n For example, if the transcript hash was 32 bytes of 01 (this length\n would make sense for SHA-256), the content covered by the digital\n signature for a server CertificateVerify would be:\n\n 2020202020202020202020202020202020202020202020202020202020202020\n 2020202020202020202020202020202020202020202020202020202020202020\n 544c5320312e332c207365727665722043657274696669636174655665726966\n 79\n 00\n 0101010101010101010101010101010101010101010101010101010101010101\n\n On the sender side the process for computing the signature field of\n the CertificateVerify message takes as input:\n\n - The content covered by the digital signature\n\n - The private signing key corresponding to the certificate sent in\n the previous message\n\n If the CertificateVerify message is sent by a server, the signature\n algorithm MUST be one offered in the client's \"signature_algorithms\"\n\n\n\nRescorla Expires September 21, 2018 [Page 73]\n\f\nInternet-Draft TLS March 2018\n\n\n extension unless no valid certificate chain can be produced without\n unsupported algorithms (see Section 4.2.3).\n\n If sent by a client, the signature algorithm used in the signature\n MUST be one of those present in the supported_signature_algorithms\n field of the \"signature_algorithms\" extension in the\n CertificateRequest message.\n\n In addition, the signature algorithm MUST be compatible with the key\n in the sender's end-entity certificate. RSA signatures MUST use an\n RSASSA-PSS algorithm, regardless of whether RSASSA-PKCS1-v1_5\n algorithms appear in \"signature_algorithms\". The SHA-1 algorithm\n MUST NOT be used in any signatures of CertificateVerify messages.\n All SHA-1 signature algorithms in this specification are defined\n solely for use in legacy certificates and are not valid for\n CertificateVerify signatures.\n\n The receiver of a CertificateVerify message MUST verify the signature\n field. The verification process takes as input:\n\n - The content covered by the digital signature\n\n - The public key contained in the end-entity certificate found in\n the associated Certificate message.\n\n - The digital signature received in the signature field of the\n CertificateVerify message\n\n If the verification fails, the receiver MUST terminate the handshake\n with a \"decrypt_error\" alert.\n\n4.4.4. Finished\n\n The Finished message is the final message in the authentication\n block. It is essential for providing authentication of the handshake\n and of the computed keys.\n\n Recipients of Finished messages MUST verify that the contents are\n correct and if incorrect MUST terminate the connection with a\n \"decrypt_error\" alert.\n\n Once a side has sent its Finished message and received and validated\n the Finished message from its peer, it may begin to send and receive\n application data over the connection. There are two settings in\n which it is permitted to send data prior to receiving the peer's\n Finished:\n\n 1. Clients sending 0-RTT data as described in Section 4.2.10.\n\n\n\nRescorla Expires September 21, 2018 [Page 74]\n\f\nInternet-Draft TLS March 2018\n\n\n 2. Servers MAY send data after sending their first flight, but\n because the handshake is not yet complete, they have no assurance\n of either the peer's identity or of its liveness (i.e., the\n ClientHello might have been replayed).\n\n The key used to compute the Finished message is computed from the\n Base key defined in Section 4.4 using HKDF (see Section 7.1).\n Specifically:\n\n finished_key =\n HKDF-Expand-Label(BaseKey, \"finished\", \"\", Hash.length)\n\n Structure of this message:\n\n struct {\n opaque verify_data[Hash.length];\n } Finished;\n\n The verify_data value is computed as follows:\n\n verify_data =\n HMAC(finished_key,\n Transcript-Hash(Handshake Context,\n Certificate*, CertificateVerify*))\n\n * Only included if present.\n\n HMAC [RFC2104] uses the Hash algorithm for the handshake. As noted\n above, the HMAC input can generally be implemented by a running hash,\n i.e., just the handshake hash at this point.\n\n In previous versions of TLS, the verify_data was always 12 octets\n long. In TLS 1.3, it is the size of the HMAC output for the Hash\n used for the handshake.\n\n Note: Alerts and any other record types are not handshake messages\n and are not included in the hash computations.\n\n Any records following a Finished message MUST be encrypted under the\n appropriate application traffic key as described in Section 7.2. In\n particular, this includes any alerts sent by the server in response\n to client Certificate and CertificateVerify messages.\n\n4.5. End of Early Data\n\n struct {} EndOfEarlyData;\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 75]\n\f\nInternet-Draft TLS March 2018\n\n\n If the server sent an \"early_data\" extension, the client MUST send an\n EndOfEarlyData message after receiving the server Finished. If the\n server does not send an \"early_data\" extension, then the client MUST\n NOT send an EndOfEarlyData message. This message indicates that all\n 0-RTT application_data messages, if any, have been transmitted and\n that the following records are protected under handshake traffic\n keys. Servers MUST NOT send this message and clients receiving it\n MUST terminate the connection with an \"unexpected_message\" alert.\n This message is encrypted under keys derived from the\n client_early_traffic_secret.\n\n4.6. Post-Handshake Messages\n\n TLS also allows other messages to be sent after the main handshake.\n These messages use a handshake content type and are encrypted under\n the appropriate application traffic key.\n\n4.6.1. New Session Ticket Message\n\n At any time after the server has received the client Finished\n message, it MAY send a NewSessionTicket message. This message\n creates a unique association between the ticket value and a secret\n PSK derived from the resumption master secret (see Section 7.\n\n The client MAY use this PSK for future handshakes by including the\n ticket value in the \"pre_shared_key\" extension in its ClientHello\n (Section 4.2.11). Servers MAY send multiple tickets on a single\n connection, either immediately after each other or after specific\n events (see Appendix C.4). For instance, the server might send a new\n ticket after post-handshake authentication in order to encapsulate\n the additional client authentication state. Multiple tickets are\n useful for clients for a variety of purposes, including:\n\n - Opening multiple parallel HTTP connections.\n\n - Performing connection racing across interfaces and address\n families via, e.g., Happy Eyeballs [RFC8305] or related\n techniques.\n\n Any ticket MUST only be resumed with a cipher suite that has the same\n KDF hash algorithm as that used to establish the original connection.\n\n Clients MUST only resume if the new SNI value is valid for the server\n certificate presented in the original session, and SHOULD only resume\n if the SNI value matches the one used in the original session. The\n latter is a performance optimization: normally, there is no reason to\n expect that different servers covered by a single certificate would\n be able to accept each other's tickets, hence attempting resumption\n\n\n\nRescorla Expires September 21, 2018 [Page 76]\n\f\nInternet-Draft TLS March 2018\n\n\n in that case would waste a single-use ticket. If such an indication\n is provided (externally or by any other means), clients MAY resume\n with a different SNI value.\n\n On resumption, if reporting an SNI value to the calling application,\n implementations MUST use the value sent in the resumption ClientHello\n rather than the value sent in the previous session. Note that if a\n server implementation declines all PSK identities with different SNI\n values, these two values are always the same.\n\n Note: Although the resumption master secret depends on the client's\n second flight, servers which do not request client authentication MAY\n compute the remainder of the transcript independently and then send a\n NewSessionTicket immediately upon sending its Finished rather than\n waiting for the client Finished. This might be appropriate in cases\n where the client is expected to open multiple TLS connections in\n parallel and would benefit from the reduced overhead of a resumption\n handshake, for example.\n\n struct {\n uint32 ticket_lifetime;\n uint32 ticket_age_add;\n opaque ticket_nonce<0..255>;\n opaque ticket<1..2^16-1>;\n Extension extensions<0..2^16-2>;\n } NewSessionTicket;\n\n ticket_lifetime Indicates the lifetime in seconds as a 32-bit\n unsigned integer in network byte order from the time of ticket\n issuance. Servers MUST NOT use any value greater than 604800\n seconds (7 days). The value of zero indicates that the ticket\n should be discarded immediately. Clients MUST NOT cache tickets\n for longer than 7 days, regardless of the ticket_lifetime, and MAY\n delete tickets earlier based on local policy. A server MAY treat\n a ticket as valid for a shorter period of time than what is stated\n in the ticket_lifetime.\n\n ticket_age_add A securely generated, random 32-bit value that is\n used to obscure the age of the ticket that the client includes in\n the \"pre_shared_key\" extension. The client-side ticket age is\n added to this value modulo 2^32 to obtain the value that is\n transmitted by the client. The server MUST generate a fresh value\n for each ticket it sends.\n\n ticket_nonce A per-ticket value that is unique across all tickets\n issued on this connection.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 77]\n\f\nInternet-Draft TLS March 2018\n\n\n ticket The value of the ticket to be used as the PSK identity. The\n ticket itself is an opaque label. It MAY either be a database\n lookup key or a self-encrypted and self-authenticated value.\n Section 4 of [RFC5077] describes a recommended ticket construction\n mechanism.\n\n extensions A set of extension values for the ticket. The\n \"Extension\" format is defined in Section 4.2. Clients MUST ignore\n unrecognized extensions.\n\n The sole extension currently defined for NewSessionTicket is\n \"early_data\", indicating that the ticket may be used to send 0-RTT\n data (Section 4.2.10)). It contains the following value:\n\n max_early_data_size The maximum amount of 0-RTT data that the client\n is allowed to send when using this ticket, in bytes. Only\n Application Data payload (i.e., plaintext but not padding or the\n inner content type byte) is counted. A server receiving more than\n max_early_data_size bytes of 0-RTT data SHOULD terminate the\n connection with an \"unexpected_message\" alert. Note that servers\n that reject early data due to lack of cryptographic material will\n be unable to differentiate padding from content, so clients SHOULD\n NOT depend on being able to send large quantities of padding in\n early data records.\n\n The PSK associated with the ticket is computed as:\n\n HKDF-Expand-Label(resumption_master_secret,\n \"resumption\", ticket_nonce, Hash.length)\n\n Because the ticket_nonce value is distinct for each NewSessionTicket\n message, a different PSK will be derived for each ticket.\n\n Note that in principle it is possible to continue issuing new tickets\n which indefinitely extend the lifetime of the keying material\n originally derived from an initial non-PSK handshake (which was most\n likely tied to the peer's certificate). It is RECOMMENDED that\n implementations place limits on the total lifetime of such keying\n material; these limits should take into account the lifetime of the\n peer's certificate, the likelihood of intervening revocation, and the\n time since the peer's online CertificateVerify signature.\n\n4.6.2. Post-Handshake Authentication\n\n When the client has sent the \"post_handshake_auth\" extension (see\n Section 4.2.6), a server MAY request client authentication at any\n time after the handshake has completed by sending a\n CertificateRequest message. The client MUST respond with the\n\n\n\nRescorla Expires September 21, 2018 [Page 78]\n\f\nInternet-Draft TLS March 2018\n\n\n appropriate Authentication messages (see Section 4.4). If the client\n chooses to authenticate, it MUST send Certificate, CertificateVerify,\n and Finished. If it declines, it MUST send a Certificate message\n containing no certificates followed by Finished. All of the client's\n messages for a given response MUST appear consecutively on the wire\n with no intervening messages of other types.\n\n A client that receives a CertificateRequest message without having\n sent the \"post_handshake_auth\" extension MUST send an\n \"unexpected_message\" fatal alert.\n\n Note: Because client authentication could involve prompting the user,\n servers MUST be prepared for some delay, including receiving an\n arbitrary number of other messages between sending the\n CertificateRequest and receiving a response. In addition, clients\n which receive multiple CertificateRequests in close succession MAY\n respond to them in a different order than they were received (the\n certificate_request_context value allows the server to disambiguate\n the responses).\n\n4.6.3. Key and IV Update\n\n enum {\n update_not_requested(0), update_requested(1), (255)\n } KeyUpdateRequest;\n\n struct {\n KeyUpdateRequest request_update;\n } KeyUpdate;\n\n request_update Indicates whether the recipient of the KeyUpdate\n should respond with its own KeyUpdate. If an implementation\n receives any other value, it MUST terminate the connection with an\n \"illegal_parameter\" alert.\n\n The KeyUpdate handshake message is used to indicate that the sender\n is updating its sending cryptographic keys. This message can be sent\n by either peer after it has sent a Finished message. Implementations\n that receive a KeyUpdate message prior to receiving a Finished\n message MUST terminate the connection with an \"unexpected_message\"\n alert. After sending a KeyUpdate message, the sender SHALL send all\n its traffic using the next generation of keys, computed as described\n in Section 7.2. Upon receiving a KeyUpdate, the receiver MUST update\n its receiving keys.\n\n If the request_update field is set to \"update_requested\" then the\n receiver MUST send a KeyUpdate of its own with request_update set to\n \"update_not_requested\" prior to sending its next application data\n\n\n\nRescorla Expires September 21, 2018 [Page 79]\n\f\nInternet-Draft TLS March 2018\n\n\n record. This mechanism allows either side to force an update to the\n entire connection, but causes an implementation which receives\n multiple KeyUpdates while it is silent to respond with a single\n update. Note that implementations may receive an arbitrary number of\n messages between sending a KeyUpdate with request_update set to\n update_requested and receiving the peer's KeyUpdate, because those\n messages may already be in flight. However, because send and receive\n keys are derived from independent traffic secrets, retaining the\n receive traffic secret does not threaten the forward secrecy of data\n sent before the sender changed keys.\n\n If implementations independently send their own KeyUpdates with\n request_update set to \"update_requested\", and they cross in flight,\n then each side will also send a response, with the result that each\n side increments by two generations.\n\n Both sender and receiver MUST encrypt their KeyUpdate messages with\n the old keys. Additionally, both sides MUST enforce that a KeyUpdate\n with the old key is received before accepting any messages encrypted\n with the new key. Failure to do so may allow message truncation\n attacks.\n\n5. Record Protocol\n\n The TLS record protocol takes messages to be transmitted, fragments\n the data into manageable blocks, protects the records, and transmits\n the result. Received data is verified, decrypted, reassembled, and\n then delivered to higher-level clients.\n\n TLS records are typed, which allows multiple higher-level protocols\n to be multiplexed over the same record layer. This document\n specifies four content types: handshake, application data, alert, and\n change_cipher_spec. The change_cipher_spec record is used only for\n compatibility purposes (see Appendix D.4).\n\n An implementation may receive an unencrypted record of type\n change_cipher_spec consisting of the single byte value 0x01 at any\n time after the first ClientHello message has been sent or received\n and before the peer's Finished message has been received and MUST\n simply drop it without further processing. Note that this record may\n appear at a point at the handshake where the implementation is\n expecting protected records and so it is necessary to detect this\n condition prior to attempting to deprotect the record. An\n implementation which receives any other change_cipher_spec value or\n which receives a protected change_cipher_spec record MUST abort the\n handshake with an \"unexpected_message\" alert. A change_cipher_spec\n record received before the first ClientHello message or after the\n peer's Finished message MUST be treated as an unexpected record type\n\n\n\nRescorla Expires September 21, 2018 [Page 80]\n\f\nInternet-Draft TLS March 2018\n\n\n (though stateless servers may not be able to distinguish these cases\n from allowed cases).\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST terminate\n the connection with an \"unexpected_message\" alert. New record\n content type values are assigned by IANA in the TLS Content Type\n Registry as described in Section 11.\n\n5.1. Record Layer\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Message\n boundaries are handled differently depending on the underlying\n ContentType. Any future content types MUST specify appropriate\n rules. Note that these rules are stricter than what was enforced in\n TLS 1.2.\n\n Handshake messages MAY be coalesced into a single TLSPlaintext record\n or fragmented across several records, provided that:\n\n - Handshake messages MUST NOT be interleaved with other record\n types. That is, if a handshake message is split over two or more\n records, there MUST NOT be any other records between them.\n\n - Handshake messages MUST NOT span key changes. Implementations\n MUST verify that all messages immediately preceding a key change\n align with a record boundary; if not, then they MUST terminate the\n connection with an \"unexpected_message\" alert. Because the\n ClientHello, EndOfEarlyData, ServerHello, Finished, and KeyUpdate\n messages can immediately precede a key change, implementations\n MUST send these messages in alignment with a record boundary.\n\n Implementations MUST NOT send zero-length fragments of Handshake\n types, even if those fragments contain padding.\n\n Alert messages (Section 6) MUST NOT be fragmented across records and\n multiple Alert messages MUST NOT be coalesced into a single\n TLSPlaintext record. In other words, a record with an Alert type\n MUST contain exactly one message.\n\n Application Data messages contain data that is opaque to TLS.\n Application Data messages are always protected. Zero-length\n fragments of Application Data MAY be sent as they are potentially\n useful as a traffic analysis countermeasure. Application Data\n fragments MAY be split across multiple records or coalesced into a\n single record.\n\n\n\nRescorla Expires September 21, 2018 [Page 81]\n\f\nInternet-Draft TLS March 2018\n\n\n enum {\n invalid(0),\n change_cipher_spec(20),\n alert(21),\n handshake(22),\n application_data(23),\n (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion legacy_record_version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type The higher-level protocol used to process the enclosed\n fragment.\n\n legacy_record_version This value MUST be set to 0x0303 for all\n records generated by a TLS 1.3 implementation other than an\n initial ClientHello (i.e., one not generated after a\n HelloRetryRequest), where it MAY also be 0x0301 for compatibility\n purposes. This field is deprecated and MUST be ignored for all\n purposes. Previous versions of TLS would use other values in this\n field under some circumstances.\n\n length The length (in bytes) of the following TLSPlaintext.fragment.\n The length MUST NOT exceed 2^14 bytes. An endpoint that receives\n a record that exceeds this length MUST terminate the connection\n with a \"record_overflow\" alert.\n\n fragment The data being transmitted. This value is transparent and\n is treated as an independent block to be dealt with by the higher-\n level protocol specified by the type field.\n\n This document describes TLS 1.3, which uses the version 0x0304. This\n version value is historical, deriving from the use of 0x0301 for TLS\n 1.0 and 0x0300 for SSL 3.0. In order to maximize backwards\n compatibility, records containing an initial ClientHello SHOULD have\n version 0x0301 and a record containing a second ClientHello or a\n ServerHello MUST have version 0x0303, reflecting TLS 1.0 and TLS 1.2\n respectively. When negotiating prior versions of TLS, endpoints\n follow the procedure and requirements in Appendix D.\n\n When record protection has not yet been engaged, TLSPlaintext\n structures are written directly onto the wire. Once record\n protection has started, TLSPlaintext records are protected and sent\n\n\n\nRescorla Expires September 21, 2018 [Page 82]\n\f\nInternet-Draft TLS March 2018\n\n\n as described in the following section. Note that application data\n records MUST NOT be written to the wire unprotected (see Section 2\n for details).\n\n5.2. Record Payload Protection\n\n The record protection functions translate a TLSPlaintext structure\n into a TLSCiphertext. The deprotection functions reverse the\n process. In TLS 1.3, as opposed to previous versions of TLS, all\n ciphers are modeled as \"Authenticated Encryption with Additional\n Data\" (AEAD) [RFC5116]. AEAD functions provide an unified encryption\n and authentication operation which turns plaintext into authenticated\n ciphertext and back again. Each encrypted record consists of a\n plaintext header followed by an encrypted body, which itself contains\n a type and optional padding.\n\n struct {\n opaque content[TLSPlaintext.length];\n ContentType type;\n uint8 zeros[length_of_padding];\n } TLSInnerPlaintext;\n\n struct {\n ContentType opaque_type = application_data; /* 23 */\n ProtocolVersion legacy_record_version = 0x0303; /* TLS v1.2 */\n uint16 length;\n opaque encrypted_record[TLSCiphertext.length];\n } TLSCiphertext;\n\n content The TLSPLaintext.fragment value, containing the byte\n encoding of a handshake or an alert message, or the raw bytes of\n the application's data to send.\n\n type The TLSPlaintext.type value containing the content type of the\n record.\n\n zeros An arbitrary-length run of zero-valued bytes may appear in the\n cleartext after the type field. This provides an opportunity for\n senders to pad any TLS record by a chosen amount as long as the\n total stays within record size limits. See Section 5.4 for more\n details.\n\n opaque_type The outer opaque_type field of a TLSCiphertext record is\n always set to the value 23 (application_data) for outward\n compatibility with middleboxes accustomed to parsing previous\n versions of TLS. The actual content type of the record is found\n in TLSInnerPlaintext.type after decryption.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 83]\n\f\nInternet-Draft TLS March 2018\n\n\n legacy_record_version The legacy_record_version field is always\n 0x0303. TLS 1.3 TLSCiphertexts are not generated until after TLS\n 1.3 has been negotiated, so there are no historical compatibility\n concerns where other values might be received. Note that the\n handshake protocol including the ClientHello and ServerHello\n messages authenticates the protocol version, so this value is\n redundant.\n\n length The length (in bytes) of the following\n TLSCiphertext.encrypted_record, which is the sum of the lengths of\n the content and the padding, plus one for the inner content type,\n plus any expansion added by the AEAD algorithm. The length MUST\n NOT exceed 2^14 + 256 bytes. An endpoint that receives a record\n that exceeds this length MUST terminate the connection with a\n \"record_overflow\" alert.\n\n encrypted_record The AEAD-encrypted form of the serialized\n TLSInnerPlaintext structure.\n\n AEAD algorithms take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [RFC5116]. The key is either the\n client_write_key or the server_write_key, the nonce is derived from\n the sequence number and the client_write_iv or server_write_iv (see\n Section 5.3), and the additional data input is the record header.\n I.e.,\n\n additional_data = TLSCiphertext.opaque_type ||\n TLSCiphertext.legacy_record_version ||\n TLSCiphertext.length\n\n The plaintext input to the AEAD algorithm is the encoded\n TLSInnerPlaintext structure. Derivation of traffic keys is defined\n in Section 7.3.\n\n The AEAD output consists of the ciphertext output from the AEAD\n encryption operation. The length of the plaintext is greater than\n the corresponding TLSPlaintext.length due to the inclusion of\n TLSInnerPlaintext.type and any padding supplied by the sender. The\n length of the AEAD output will generally be larger than the\n plaintext, but by an amount that varies with the AEAD algorithm.\n Since the ciphers might incorporate padding, the amount of overhead\n could vary with different lengths of plaintext. Symbolically,\n\n AEADEncrypted =\n AEAD-Encrypt(write_key, nonce, additional_data, plaintext)\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 84]\n\f\nInternet-Draft TLS March 2018\n\n\n Then the encrypted_record field of TLSCiphertext is set to\n AEADEncrypted.\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, additional data, and the AEADEncrypted value. The output is\n either the plaintext or an error indicating that the decryption\n failed. There is no separate integrity check. That is:\n\n plaintext of encrypted_record =\n AEAD-Decrypt(peer_write_key, nonce, additional_data, AEADEncrypted)\n\n If the decryption fails, the receiver MUST terminate the connection\n with a \"bad_record_mac\" alert.\n\n An AEAD algorithm used in TLS 1.3 MUST NOT produce an expansion\n greater than 255 octets. An endpoint that receives a record from its\n peer with TLSCiphertext.length larger than 2^14 + 256 octets MUST\n terminate the connection with a \"record_overflow\" alert. This limit\n is derived from the maximum TLSInnerPlaintext length of 2^14 octets +\n 1 octet for ContentType + the maximum AEAD expansion of 255 octets.\n\n5.3. Per-Record Nonce\n\n A 64-bit sequence number is maintained separately for reading and\n writing records. The appropriate sequence number is incremented by\n one after reading or writing each record. Each sequence number is\n set to zero at the beginning of a connection and whenever the key is\n changed; the first record transmitted under a particular traffic key\n MUST use sequence number 0.\n\n Because the size of sequence numbers is 64-bit, they should not wrap.\n If a TLS implementation would need to wrap a sequence number, it MUST\n either re-key (Section 4.6.3) or terminate the connection.\n\n Each AEAD algorithm will specify a range of possible lengths for the\n per-record nonce, from N_MIN bytes to N_MAX bytes of input\n ([RFC5116]). The length of the TLS per-record nonce (iv_length) is\n set to the larger of 8 bytes and N_MIN for the AEAD algorithm (see\n [RFC5116] Section 4). An AEAD algorithm where N_MAX is less than 8\n bytes MUST NOT be used with TLS. The per-record nonce for the AEAD\n construction is formed as follows:\n\n 1. The 64-bit record sequence number is encoded in network byte\n order and padded to the left with zeros to iv_length.\n\n 2. The padded sequence number is XORed with the static\n client_write_iv or server_write_iv, depending on the role.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 85]\n\f\nInternet-Draft TLS March 2018\n\n\n The resulting quantity (of length iv_length) is used as the per-\n record nonce.\n\n Note: This is a different construction from that in TLS 1.2, which\n specified a partially explicit nonce.\n\n5.4. Record Padding\n\n All encrypted TLS records can be padded to inflate the size of the\n TLSCiphertext. This allows the sender to hide the size of the\n traffic from an observer.\n\n When generating a TLSCiphertext record, implementations MAY choose to\n pad. An unpadded record is just a record with a padding length of\n zero. Padding is a string of zero-valued bytes appended to the\n ContentType field before encryption. Implementations MUST set the\n padding octets to all zeros before encrypting.\n\n Application Data records may contain a zero-length\n TLSInnerPlaintext.content if the sender desires. This permits\n generation of plausibly-sized cover traffic in contexts where the\n presence or absence of activity may be sensitive. Implementations\n MUST NOT send Handshake or Alert records that have a zero-length\n TLSInnerPlaintext.content; if such a message is received, the\n receiving implementation MUST terminate the connection with an\n \"unexpected_message\" alert.\n\n The padding sent is automatically verified by the record protection\n mechanism; upon successful decryption of a\n TLSCiphertext.encrypted_record, the receiving implementation scans\n the field from the end toward the beginning until it finds a non-zero\n octet. This non-zero octet is the content type of the message. This\n padding scheme was selected because it allows padding of any\n encrypted TLS record by an arbitrary size (from zero up to TLS record\n size limits) without introducing new content types. The design also\n enforces all-zero padding octets, which allows for quick detection of\n padding errors.\n\n Implementations MUST limit their scanning to the cleartext returned\n from the AEAD decryption. If a receiving implementation does not\n find a non-zero octet in the cleartext, it MUST terminate the\n connection with an \"unexpected_message\" alert.\n\n The presence of padding does not change the overall record size\n limitations - the full encoded TLSInnerPlaintext MUST NOT exceed 2^14\n + 1 octets. If the maximum fragment length is reduced, as for\n example by the max_fragment_length extension from [RFC6066], then the\n\n\n\n\nRescorla Expires September 21, 2018 [Page 86]\n\f\nInternet-Draft TLS March 2018\n\n\n reduced limit applies to the full plaintext, including the content\n type and padding.\n\n Selecting a padding policy that suggests when and how much to pad is\n a complex topic and is beyond the scope of this specification. If\n the application layer protocol on top of TLS has its own padding, it\n may be preferable to pad application_data TLS records within the\n application layer. Padding for encrypted handshake and alert TLS\n records must still be handled at the TLS layer, though. Later\n documents may define padding selection algorithms or define a padding\n policy request mechanism through TLS extensions or some other means.\n\n5.5. Limits on Key Usage\n\n There are cryptographic limits on the amount of plaintext which can\n be safely encrypted under a given set of keys. [AEAD-LIMITS]\n provides an analysis of these limits under the assumption that the\n underlying primitive (AES or ChaCha20) has no weaknesses.\n Implementations SHOULD do a key update as described in Section 4.6.3\n prior to reaching these limits.\n\n For AES-GCM, up to 2^24.5 full-size records (about 24 million) may be\n encrypted on a given connection while keeping a safety margin of\n approximately 2^-57 for Authenticated Encryption (AE) security. For\n ChaCha20/Poly1305, the record sequence number would wrap before the\n safety limit is reached.\n\n6. Alert Protocol\n\n One of the content types supported by the TLS record layer is the\n alert type. Like other messages, alert messages are encrypted as\n specified by the current connection state.\n\n Alert messages convey a description of the alert and a legacy field\n that conveyed the severity of the message in previous versions of\n TLS. Alerts are divided into two classes: closure alerts and error\n alerts. In TLS 1.3, the severity is implicit in the type of alert\n being sent, and the 'level' field can safely be ignored. The\n \"close_notify\" alert is used to indicate orderly closure of one\n direction of the connection. Upon receiving such an alert, the TLS\n implementation SHOULD indicate end-of-data to the application.\n\n Error alerts indicate abortive closure of the connection (see\n Section 6.2). Upon receiving an error alert, the TLS implementation\n SHOULD indicate an error to the application and MUST NOT allow any\n further data to be sent or received on the connection. Servers and\n clients MUST forget the secret values and keys established in failed\n\n\n\n\nRescorla Expires September 21, 2018 [Page 87]\n\f\nInternet-Draft TLS March 2018\n\n\n connections, with the exception of the PSKs associated with session\n tickets, which SHOULD be discarded if possible.\n\n All the alerts listed in Section 6.2 MUST be sent with\n AlertLevel=fatal and MUST be treated as error alerts regardless of\n the AlertLevel in the message. Unknown alert types MUST be treated\n as error alerts.\n\n Note: TLS defines two generic alerts (see Section 6) to use upon\n failure to parse a message. Peers which receive a message which\n cannot be parsed according to the syntax (e.g., have a length\n extending beyond the message boundary or contain an out-of-range\n length) MUST terminate the connection with a \"decode_error\" alert.\n Peers which receive a message which is syntactically correct but\n semantically invalid (e.g., a DHE share of p - 1, or an invalid enum)\n MUST terminate the connection with an \"illegal_parameter\" alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 88]\n\f\nInternet-Draft TLS March 2018\n\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n record_overflow(22),\n handshake_failure(40),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n inappropriate_fallback(86),\n user_canceled(90),\n missing_extension(109),\n unsupported_extension(110),\n unrecognized_name(112),\n bad_certificate_status_response(113),\n unknown_psk_identity(115),\n certificate_required(116),\n no_application_protocol(120),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n6.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack.\n\n close_notify This alert notifies the recipient that the sender will\n not send any more messages on this connection. Any data received\n after a closure alert has been received MUST be ignored.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 89]\n\f\nInternet-Draft TLS March 2018\n\n\n user_canceled This alert notifies the recipient that the sender is\n canceling the handshake for some reason unrelated to a protocol\n failure. If a user cancels an operation after the handshake is\n complete, just closing the connection by sending a \"close_notify\"\n is more appropriate. This alert SHOULD be followed by a\n \"close_notify\". This alert generally has AlertLevel=warning.\n\n Either party MAY initiate a close of its write side of the connection\n by sending a \"close_notify\" alert. Any data received after a closure\n alert has been received MUST be ignored. If a transport-level close\n is received prior to a \"close_notify\", the receiver cannot know that\n all the data that was sent has been received.\n\n Each party MUST send a \"close_notify\" alert before closing its write\n side of the connection, unless it has already sent some error alert.\n This does not have any effect on its read side of the connection.\n Note that this is a change from versions of TLS prior to TLS 1.3 in\n which implementations were required to react to a \"close_notify\" by\n discarding pending writes and sending an immediate \"close_notify\"\n alert of their own. That previous requirement could cause truncation\n in the read side. Both parties need not wait to receive a\n \"close_notify\" alert before closing their read side of the\n connection, though doing so would introduce the possibility of\n truncation.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation MUST receive a \"close_notify\" alert\n before indicating end-of-data to the application-layer. No part of\n this standard should be taken to dictate the manner in which a usage\n profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing the write side of a connection\n reliably delivers pending data before destroying the transport.\n\n6.2. Error Alerts\n\n Error handling in the TLS Handshake Protocol is very simple. When an\n error is detected, the detecting party sends a message to its peer.\n Upon transmission or receipt of a fatal alert message, both parties\n MUST immediately close the connection.\n\n Whenever an implementation encounters a fatal error condition, it\n SHOULD send an appropriate fatal alert and MUST close the connection\n without sending or receiving any additional data. In the rest of\n this specification, when the phrases \"terminate the connection\" and\n \"abort the handshake\" are used without a specific alert it means that\n\n\n\nRescorla Expires September 21, 2018 [Page 90]\n\f\nInternet-Draft TLS March 2018\n\n\n the implementation SHOULD send the alert indicated by the\n descriptions below. The phrases \"terminate the connection with a X\n alert\" and \"abort the handshake with a X alert\" mean that the\n implementation MUST send alert X if it sends any alert. All alerts\n defined in this section below, as well as all unknown alerts, are\n universally considered fatal as of TLS 1.3 (see Section 6). The\n implementation SHOULD provide a way to facilitate logging the sending\n and receiving of alerts.\n\n The following error alerts are defined:\n\n unexpected_message An inappropriate message (e.g., the wrong\n handshake message, premature application data, etc.) was received.\n This alert should never be observed in communication between\n proper implementations.\n\n bad_record_mac This alert is returned if a record is received which\n cannot be deprotected. Because AEAD algorithms combine decryption\n and verification, and also to avoid side channel attacks, this\n alert is used for all deprotection failures. This alert should\n never be observed in communication between proper implementations,\n except when messages were corrupted in the network.\n\n record_overflow A TLSCiphertext record was received that had a\n length more than 2^14 + 256 bytes, or a record decrypted to a\n TLSPlaintext record with more than 2^14 bytes (or some other\n negotiated limit). This alert should never be observed in\n communication between proper implementations, except when messages\n were corrupted in the network.\n\n handshake_failure Receipt of a \"handshake_failure\" alert message\n indicates that the sender was unable to negotiate an acceptable\n set of security parameters given the options available.\n\n bad_certificate A certificate was corrupt, contained signatures that\n did not verify correctly, etc.\n\n unsupported_certificate A certificate was of an unsupported type.\n\n certificate_revoked A certificate was revoked by its signer.\n\n certificate_expired A certificate has expired or is not currently\n valid.\n\n certificate_unknown Some other (unspecified) issue arose in\n processing the certificate, rendering it unacceptable.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 91]\n\f\nInternet-Draft TLS March 2018\n\n\n illegal_parameter A field in the handshake was incorrect or\n inconsistent with other fields. This alert is used for errors\n which conform to the formal protocol syntax but are otherwise\n incorrect.\n\n unknown_ca A valid certificate chain or partial chain was received,\n but the certificate was not accepted because the CA certificate\n could not be located or could not be matched with a known trust\n anchor.\n\n access_denied A valid certificate or PSK was received, but when\n access control was applied, the sender decided not to proceed with\n negotiation.\n\n decode_error A message could not be decoded because some field was\n out of the specified range or the length of the message was\n incorrect. This alert is used for errors where the message does\n not conform to the formal protocol syntax. This alert should\n never be observed in communication between proper implementations,\n except when messages were corrupted in the network.\n\n decrypt_error A handshake (not record-layer) cryptographic operation\n failed, including being unable to correctly verify a signature or\n validate a Finished message or a PSK binder.\n\n protocol_version The protocol version the peer has attempted to\n negotiate is recognized but not supported. (see Appendix D)\n\n insufficient_security Returned instead of \"handshake_failure\" when a\n negotiation has failed specifically because the server requires\n parameters more secure than those supported by the client.\n\n internal_error An internal error unrelated to the peer or the\n correctness of the protocol (such as a memory allocation failure)\n makes it impossible to continue.\n\n inappropriate_fallback Sent by a server in response to an invalid\n connection retry attempt from a client (see [RFC7507]).\n\n missing_extension Sent by endpoints that receive a handshake message\n not containing an extension that is mandatory to send for the\n offered TLS version or other negotiated parameters.\n\n unsupported_extension Sent by endpoints receiving any handshake\n message containing an extension known to be prohibited for\n inclusion in the given handshake message, or including any\n extensions in a ServerHello or Certificate not first offered in\n the corresponding ClientHello.\n\n\n\nRescorla Expires September 21, 2018 [Page 92]\n\f\nInternet-Draft TLS March 2018\n\n\n unrecognized_name Sent by servers when no server exists identified\n by the name provided by the client via the \"server_name\" extension\n (see [RFC6066]).\n\n bad_certificate_status_response Sent by clients when an invalid or\n unacceptable OCSP response is provided by the server via the\n \"status_request\" extension (see [RFC6066]).\n\n unknown_psk_identity Sent by servers when PSK key establishment is\n desired but no acceptable PSK identity is provided by the client.\n Sending this alert is OPTIONAL; servers MAY instead choose to send\n a \"decrypt_error\" alert to merely indicate an invalid PSK\n identity.\n\n certificate_required Sent by servers when a client certificate is\n desired but none was provided by the client.\n\n no_application_protocol Sent by servers when a client\n \"application_layer_protocol_negotiation\" extension advertises only\n protocols that the server does not support (see [RFC7301]).\n\n New Alert values are assigned by IANA as described in Section 11.\n\n7. Cryptographic Computations\n\n The TLS handshake establishes one or more input secrets which are\n combined to create the actual working keying material, as detailed\n below. The key derivation process incorporates both the input\n secrets and the handshake transcript. Note that because the\n handshake transcript includes the random values from the Hello\n messages, any given handshake will have different traffic secrets,\n even if the same input secrets are used, as is the case when the same\n PSK is used for multiple connections.\n\n7.1. Key Schedule\n\n The key derivation process makes use of the HKDF-Extract and HKDF-\n Expand functions as defined for HKDF [RFC5869], as well as the\n functions defined below:\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 93]\n\f\nInternet-Draft TLS March 2018\n\n\n HKDF-Expand-Label(Secret, Label, Context, Length) =\n HKDF-Expand(Secret, HkdfLabel, Length)\n\n Where HkdfLabel is specified as:\n\n struct {\n uint16 length = Length;\n opaque label<7..255> = \"tls13 \" + Label;\n opaque context<0..255> = Context;\n } HkdfLabel;\n\n Derive-Secret(Secret, Label, Messages) =\n HKDF-Expand-Label(Secret, Label,\n Transcript-Hash(Messages), Hash.length)\n\n The Hash function used by Transcript-Hash and HKDF is the cipher\n suite hash algorithm. Hash.length is its output length in bytes.\n Messages is the concatenation of the indicated handshake messages,\n including the handshake message type and length fields, but not\n including record layer headers. Note that in some cases a zero-\n length Context (indicated by \"\") is passed to HKDF-Expand-Label. The\n Labels specified in this document are all ASCII strings, and do not\n include a trailing NUL byte.\n\n Note: with common hash functions, any label longer than 12 characters\n requires an additional iteration of the hash function to compute.\n The labels in this specification have all been chosen to fit within\n this limit.\n\n Keys are derived from two input secrets using the HKDF-Extract and\n Derive-Secret functions. The general pattern for adding a new secret\n is to use HKDF-Extract with the salt being the current secret state\n and the IKM being the new secret to be added. In this version of TLS\n 1.3, the two input secrets are:\n\n - PSK (a pre-shared key established externally or derived from the\n resumption_master_secret value from a previous connection)\n\n - (EC)DHE shared secret (Section 7.4)\n\n This produces a full key derivation schedule shown in the diagram\n below. In this diagram, the following formatting conventions apply:\n\n - HKDF-Extract is drawn as taking the Salt argument from the top and\n the IKM argument from the left, with its output to the bottom and\n the name of the output on the right.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 94]\n\f\nInternet-Draft TLS March 2018\n\n\n - Derive-Secret's Secret argument is indicated by the incoming\n arrow. For instance, the Early Secret is the Secret for\n generating the client_early_traffic_secret.\n\n - \"0\" indicates a string of Hash-lengths bytes set to 0.\n\n 0\n |\n v\n PSK -> HKDF-Extract = Early Secret\n |\n +-----> Derive-Secret(.,\n | \"ext binder\" |\n | \"res binder\",\n | \"\")\n | = binder_key\n |\n +-----> Derive-Secret(., \"c e traffic\",\n | ClientHello)\n | = client_early_traffic_secret\n |\n +-----> Derive-Secret(., \"e exp master\",\n | ClientHello)\n | = early_exporter_master_secret\n v\n Derive-Secret(., \"derived\", \"\")\n |\n v\n(EC)DHE -> HKDF-Extract = Handshake Secret\n |\n +-----> Derive-Secret(., \"c hs traffic\",\n | ClientHello...ServerHello)\n | = client_handshake_traffic_secret\n |\n +-----> Derive-Secret(., \"s hs traffic\",\n | ClientHello...ServerHello)\n | = server_handshake_traffic_secret\n v\n Derive-Secret(., \"derived\", \"\")\n |\n v\n 0 -> HKDF-Extract = Master Secret\n |\n +-----> Derive-Secret(., \"c ap traffic\",\n | ClientHello...server Finished)\n | = client_application_traffic_secret_0\n |\n +-----> Derive-Secret(., \"s ap traffic\",\n\n\n\nRescorla Expires September 21, 2018 [Page 95]\n\f\nInternet-Draft TLS March 2018\n\n\n | ClientHello...server Finished)\n | = server_application_traffic_secret_0\n |\n +-----> Derive-Secret(., \"exp master\",\n | ClientHello...server Finished)\n | = exporter_master_secret\n |\n +-----> Derive-Secret(., \"res master\",\n ClientHello...client Finished)\n = resumption_master_secret\n\n The general pattern here is that the secrets shown down the left side\n of the diagram are just raw entropy without context, whereas the\n secrets down the right side include handshake context and therefore\n can be used to derive working keys without additional context. Note\n that the different calls to Derive-Secret may take different Messages\n arguments, even with the same secret. In a 0-RTT exchange, Derive-\n Secret is called with four distinct transcripts; in a 1-RTT-only\n exchange with three distinct transcripts.\n\n If a given secret is not available, then the 0-value consisting of a\n string of Hash.length bytes set to zeros is used. Note that this\n does not mean skipping rounds, so if PSK is not in use Early Secret\n will still be HKDF-Extract(0, 0). For the computation of the\n binder_secret, the label is \"ext binder\" for external PSKs (those\n provisioned outside of TLS) and \"res binder\" for resumption PSKs\n (those provisioned as the resumption master secret of a previous\n handshake). The different labels prevent the substitution of one\n type of PSK for the other.\n\n There are multiple potential Early Secret values depending on which\n PSK the server ultimately selects. The client will need to compute\n one for each potential PSK; if no PSK is selected, it will then need\n to compute the early secret corresponding to the zero PSK.\n\n Once all the values which are to be derived from a given secret have\n been computed, that secret SHOULD be erased.\n\n7.2. Updating Traffic Secrets\n\n Once the handshake is complete, it is possible for either side to\n update its sending traffic keys using the KeyUpdate handshake message\n defined in Section 4.6.3. The next generation of traffic keys is\n computed by generating client_/server_application_traffic_secret_N+1\n from client_/server_application_traffic_secret_N as described in this\n section then re-deriving the traffic keys as described in\n Section 7.3.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 96]\n\f\nInternet-Draft TLS March 2018\n\n\n The next-generation application_traffic_secret is computed as:\n\n application_traffic_secret_N+1 =\n HKDF-Expand-Label(application_traffic_secret_N,\n \"traffic upd\", \"\", Hash.length)\n\n Once client/server_application_traffic_secret_N+1 and its associated\n traffic keys have been computed, implementations SHOULD delete\n client_/server_application_traffic_secret_N and its associated\n traffic keys.\n\n7.3. Traffic Key Calculation\n\n The traffic keying material is generated from the following input\n values:\n\n - A secret value\n\n - A purpose value indicating the specific value being generated\n\n - The length of the key being generated\n\n The traffic keying material is generated from an input traffic secret\n value using:\n\n [sender]_write_key = HKDF-Expand-Label(Secret, \"key\", \"\", key_length)\n [sender]_write_iv = HKDF-Expand-Label(Secret, \"iv\" , \"\", iv_length)\n\n [sender] denotes the sending side. The Secret value for each record\n type is shown in the table below.\n\n +-------------------+---------------------------------------+\n | Record Type | Secret |\n +-------------------+---------------------------------------+\n | 0-RTT Application | client_early_traffic_secret |\n | | |\n | Handshake | [sender]_handshake_traffic_secret |\n | | |\n | Application Data | [sender]_application_traffic_secret_N |\n +-------------------+---------------------------------------+\n\n All the traffic keying material is recomputed whenever the underlying\n Secret changes (e.g., when changing from the handshake to application\n data keys or upon a key update).\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 97]\n\f\nInternet-Draft TLS March 2018\n\n\n7.4. (EC)DHE Shared Secret Calculation\n\n7.4.1. Finite Field Diffie-Hellman\n\n For finite field groups, a conventional Diffie-Hellman [DH76]\n computation is performed. The negotiated key (Z) is converted to a\n byte string by encoding in big-endian and left padded with zeros up\n to the size of the prime. This byte string is used as the shared\n secret in the key schedule as specified above.\n\n Note that this construction differs from previous versions of TLS\n which remove leading zeros.\n\n7.4.2. Elliptic Curve Diffie-Hellman\n\n For secp256r1, secp384r1 and secp521r1, ECDH calculations (including\n parameter and key generation as well as the shared secret\n calculation) are performed according to [IEEE1363] using the ECKAS-\n DH1 scheme with the identity map as key derivation function (KDF), so\n that the shared secret is the x-coordinate of the ECDH shared secret\n elliptic curve point represented as an octet string. Note that this\n octet string (Z in IEEE 1363 terminology) as output by FE2OSP, the\n Field Element to Octet String Conversion Primitive, has constant\n length for any given field; leading zeros found in this octet string\n MUST NOT be truncated.\n\n (Note that this use of the identity KDF is a technicality. The\n complete picture is that ECDH is employed with a non-trivial KDF\n because TLS does not directly use this secret for anything other than\n for computing other secrets.)\n\n ECDH functions are used as follows:\n\n - The public key to put into the KeyShareEntry.key_exchange\n structure is the result of applying the ECDH scalar multiplication\n function to the secret key of appropriate length (into scalar\n input) and the standard public basepoint (into u-coordinate point\n input).\n\n - The ECDH shared secret is the result of applying the ECDH scalar\n multiplication function to the secret key (into scalar input) and\n the peer's public key (into u-coordinate point input). The output\n is used raw, with no processing.\n\n For X25519 and X448, implementations SHOULD use the approach\n specified in [RFC7748] to calculate the Diffie-Hellman shared secret.\n Implementations MUST check whether the computed Diffie-Hellman shared\n secret is the all-zero value and abort if so, as described in\n\n\n\nRescorla Expires September 21, 2018 [Page 98]\n\f\nInternet-Draft TLS March 2018\n\n\n Section 6 of [RFC7748]. If implementors use an alternative\n implementation of these elliptic curves, they SHOULD perform the\n additional checks specified in Section 7 of [RFC7748].\n\n7.5. Exporters\n\n [RFC5705] defines keying material exporters for TLS in terms of the\n TLS pseudorandom function (PRF). This document replaces the PRF with\n HKDF, thus requiring a new construction. The exporter interface\n remains the same.\n\n The exporter value is computed as:\n\n TLS-Exporter(label, context_value, key_length) =\n HKDF-Expand-Label(Derive-Secret(Secret, label, \"\"),\n \"exporter\", Hash(context_value), key_length)\n\n Where Secret is either the early_exporter_master_secret or the\n exporter_master_secret. Implementations MUST use the\n exporter_master_secret unless explicitly specified by the\n application. The early_exporter_master_secret is defined for use in\n settings where an exporter is needed for 0-RTT data. A separate\n interface for the early exporter is RECOMMENDED; this avoids the\n exporter user accidentally using an early exporter when a regular one\n is desired or vice versa.\n\n If no context is provided, the context_value is zero-length.\n Consequently, providing no context computes the same value as\n providing an empty context. This is a change from previous versions\n of TLS where an empty context produced a different output to an\n absent context. As of this document's publication, no allocated\n exporter label is used both with and without a context. Future\n specifications MUST NOT define a use of exporters that permit both an\n empty context and no context with the same label. New uses of\n exporters SHOULD provide a context in all exporter computations,\n though the value could be empty.\n\n Requirements for the format of exporter labels are defined in section\n 4 of [RFC5705].\n\n8. 0-RTT and Anti-Replay\n\n As noted in Section 2.3 and Appendix E.5, TLS does not provide\n inherent replay protections for 0-RTT data. There are two potential\n threats to be concerned with:\n\n - Network attackers who mount a replay attack by simply duplicating\n a flight of 0-RTT data.\n\n\n\nRescorla Expires September 21, 2018 [Page 99]\n\f\nInternet-Draft TLS March 2018\n\n\n - Network attackers who take advantage of client retry behavior to\n arrange for the server to receive multiple copies of an\n application message. This threat already exists to some extent\n because clients that value robustness respond to network errors by\n attempting to retry requests. However, 0-RTT adds an additional\n dimension for any server system which does not maintain globally\n consistent server state. Specifically, if a server system has\n multiple zones where tickets from zone A will not be accepted in\n zone B, then an attacker can duplicate a ClientHello and early\n data intended for A to both A and B. At A, the data will be\n accepted in 0-RTT, but at B the server will reject 0-RTT data and\n instead force a full handshake. If the attacker blocks the\n ServerHello from A, then the client will complete the handshake\n with B and probably retry the request, leading to duplication on\n the server system as a whole.\n\n The first class of attack can be prevented by sharing state to\n guarantee that the 0-RTT data is accepted at most once. Servers\n SHOULD provide that level of replay safety, by implementing one of\n the methods described in this section or by equivalent means. It is\n understood, however, that due to operational concerns not all\n deployments will maintain state at that level. Therefore, in normal\n operation, clients will not know which, if any, of these mechanisms\n servers actually implement and hence MUST only send early data which\n they deem safe to be replayed.\n\n In addition to the direct effects of replays, there is a class of\n attacks where even operations normally considered idempotent could be\n exploited by a large number of replays (timing attacks, resource\n limit exhaustion and others described in Appendix E.5). Those can be\n mitigated by ensuring that every 0-RTT payload can be replayed only a\n limited number of times. The server MUST ensure that any instance of\n it (be it a machine, a thread or any other entity within the relevant\n serving infrastructure) would accept 0-RTT for the same 0-RTT\n handshake at most once; this limits the number of replays to the\n number of server instances in the deployment. Such a guarantee can\n be accomplished by locally recording data from recently-received\n ClientHellos and rejecting repeats, or by any other method that\n provides the same or a stronger guarantee. The \"at most once per\n server instance\" guarantee is a minimum requirement; servers SHOULD\n limit 0-RTT replays further when feasible.\n\n The second class of attack cannot be prevented at the TLS layer and\n MUST be dealt with by any application. Note that any application\n whose clients implement any kind of retry behavior already needs to\n implement some sort of anti-replay defense.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 100]\n\f\nInternet-Draft TLS March 2018\n\n\n8.1. Single-Use Tickets\n\n The simplest form of anti-replay defense is for the server to only\n allow each session ticket to be used once. For instance, the server\n can maintain a database of all outstanding valid tickets; deleting\n each ticket from the database as it is used. If an unknown ticket is\n provided, the server would then fall back to a full handshake.\n\n If the tickets are not self-contained but rather are database keys,\n and the corresponding PSKs are deleted upon use, then connections\n established using PSKs enjoy forward secrecy. This improves security\n for all 0-RTT data and PSK usage when PSK is used without (EC)DHE.\n\n Because this mechanism requires sharing the session database between\n server nodes in environments with multiple distributed servers, it\n may be hard to achieve high rates of successful PSK 0-RTT connections\n when compared to self-encrypted tickets. Unlike session databases,\n session tickets can successfully do PSK-based session establishment\n even without consistent storage, though when 0-RTT is allowed they\n still require consistent storage for anti-replay of 0-RTT data, as\n detailed in the following section.\n\n8.2. Client Hello Recording\n\n An alternative form of anti-replay is to record a unique value\n derived from the ClientHello (generally either the random value or\n the PSK binder) and reject duplicates. Recording all ClientHellos\n causes state to grow without bound, but a server can instead record\n ClientHellos within a given time window and use the\n \"obfuscated_ticket_age\" to ensure that tickets aren't reused outside\n that window.\n\n In order to implement this, when a ClientHello is received, the\n server first verifies the PSK binder as described Section 4.2.11. It\n then computes the expected_arrival_time as described in the next\n section and rejects 0-RTT if it is outside the recording window,\n falling back to the 1-RTT handshake.\n\n If the expected arrival time is in the window, then the server checks\n to see if it has recorded a matching ClientHello. If one is found,\n it either aborts the handshake with an \"illegal_parameter\" alert or\n accepts the PSK but reject 0-RTT. If no matching ClientHello is\n found, then it accepts 0-RTT and then stores the ClientHello for as\n long as the expected_arrival_time is inside the window. Servers MAY\n also implement data stores with false positives, such as Bloom\n filters, in which case they MUST respond to apparent replay by\n rejecting 0-RTT but MUST NOT abort the handshake.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 101]\n\f\nInternet-Draft TLS March 2018\n\n\n The server MUST derive the storage key only from validated sections\n of the ClientHello. If the ClientHello contains multiple PSK\n identities, then an attacker can create multiple ClientHellos with\n different binder values for the less-preferred identity on the\n assumption that the server will not verify it, as recommended by\n Section 4.2.11. I.e., if the client sends PSKs A and B but the\n server prefers A, then the attacker can change the binder for B\n without affecting the binder for A. If the binder for B is part of\n the storage key, then this ClientHello will not appear as a\n duplicate, which will cause the ClientHello to be accepted, and may\n cause side effects such as replay cache pollution, although any 0-RTT\n data will not be decryptable because it will use different keys. If\n the validated binder or the ClientHello.random are used as the\n storage key, then this attack is not possible.\n\n Because this mechanism does not require storing all outstanding\n tickets, it may be easier to implement in distributed systems with\n high rates of resumption and 0-RTT, at the cost of potentially weaker\n anti-replay defense because of the difficulty of reliably storing and\n retrieving the received ClientHello messages. In many such systems,\n it is impractical to have globally consistent storage of all the\n received ClientHellos. In this case, the best anti-replay protection\n is provided by having a single storage zone be authoritative for a\n given ticket and refusing 0-RTT for that ticket in any other zone.\n This approach prevents simple replay by the attacker because only one\n zone will accept 0-RTT data. A weaker design is to implement\n separate storage for each zone but allow 0-RTT in any zone. This\n approach limits the number of replays to once per zone. Application\n message duplication of course remains possible with either design.\n\n When implementations are freshly started, they SHOULD reject 0-RTT as\n long as any portion of their recording window overlaps the startup\n time. Otherwise, they run the risk of accepting replays which were\n originally sent during that period.\n\n Note: If the client's clock is running much faster than the server's\n then a ClientHello may be received that is outside the window in the\n future, in which case it might be accepted for 1-RTT, causing a\n client retry, and then acceptable later for 0-RTT. This is another\n variant of the second form of attack described above.\n\n8.3. Freshness Checks\n\n Because the ClientHello indicates the time at which the client sent\n it, it is possible to efficiently determine whether a ClientHello was\n likely sent reasonably recently and only accept 0-RTT for such a\n ClientHello, otherwise falling back to a 1-RTT handshake. This is\n necessary for the ClientHello storage mechanism described in\n\n\n\nRescorla Expires September 21, 2018 [Page 102]\n\f\nInternet-Draft TLS March 2018\n\n\n Section 8.2 because otherwise the server needs to store an unlimited\n number of ClientHellos and is a useful optimization for self-\n contained single-use tickets because it allows efficient rejection of\n ClientHellos which cannot be used for 0-RTT.\n\n In order to implement this mechanism, a server needs to store the\n time that the server generated the session ticket, offset by an\n estimate of the round trip time between client and server. I.e.,\n\n adjusted_creation_time = creation_time + estimated_RTT\n\n This value can be encoded in the ticket, thus avoiding the need to\n keep state for each outstanding ticket. The server can determine the\n client's view of the age of the ticket by subtracting the ticket's\n \"ticket_age_add value\" from the \"obfuscated_ticket_age\" parameter in\n the client's \"pre_shared_key\" extension. The server can determine\n the \"expected arrival time\" of the ClientHello as:\n\n expected_arrival_time = adjusted_creation_time + clients_ticket_age\n\n When a new ClientHello is received, the expected_arrival_time is then\n compared against the current server wall clock time and if they\n differ by more than a certain amount, 0-RTT is rejected, though the\n 1-RTT handshake can be allowed to complete.\n\n There are several potential sources of error that might cause\n mismatches between the expected arrival time and the measured time.\n Variations in client and server clock rates are likely to be minimal,\n though potentially the absolute times may be off by large values.\n Network propagation delays are the most likely causes of a mismatch\n in legitimate values for elapsed time. Both the NewSessionTicket and\n ClientHello messages might be retransmitted and therefore delayed,\n which might be hidden by TCP. For clients on the Internet, this\n implies windows on the order of ten seconds to account for errors in\n clocks and variations in measurements; other deployment scenarios may\n have different needs. Clock skew distributions are not symmetric, so\n the optimal tradeoff may involve an asymmetric range of permissible\n mismatch values.\n\n Note that freshness checking alone is not sufficient to prevent\n replays because it does not detect them during the error window,\n which, depending on bandwidth and system capacity could include\n billions of replays in real-world settings. In addition, this\n freshness checking is only done at the time the ClientHello is\n received, and not when later early application data records are\n received. After early data is accepted, records may continue to be\n streamed to the server over a longer time period.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 103]\n\f\nInternet-Draft TLS March 2018\n\n\n9. Compliance Requirements\n\n9.1. Mandatory-to-Implement Cipher Suites\n\n In the absence of an application profile standard specifying\n otherwise, a TLS-compliant application MUST implement the\n TLS_AES_128_GCM_SHA256 [GCM] cipher suite and SHOULD implement the\n TLS_AES_256_GCM_SHA384 [GCM] and TLS_CHACHA20_POLY1305_SHA256\n [RFC7539] cipher suites. (see Appendix B.4)\n\n A TLS-compliant application MUST support digital signatures with\n rsa_pkcs1_sha256 (for certificates), rsa_pss_rsae_sha256 (for\n CertificateVerify and certificates), and ecdsa_secp256r1_sha256. A\n TLS-compliant application MUST support key exchange with secp256r1\n (NIST P-256) and SHOULD support key exchange with X25519 [RFC7748].\n\n9.2. Mandatory-to-Implement Extensions\n\n In the absence of an application profile standard specifying\n otherwise, a TLS-compliant application MUST implement the following\n TLS extensions:\n\n - Supported Versions (\"supported_versions\"; Section 4.2.1)\n\n - Cookie (\"cookie\"; Section 4.2.2)\n\n - Signature Algorithms (\"signature_algorithms\"; Section 4.2.3)\n\n - Signature Algorithms Certificate (\"signature_algorithms_cert\";\n Section 4.2.3)\n\n - Negotiated Groups (\"supported_groups\"; Section 4.2.7)\n\n - Key Share (\"key_share\"; Section 4.2.8)\n\n - Server Name Indication (\"server_name\"; Section 3 of [RFC6066])\n\n All implementations MUST send and use these extensions when offering\n applicable features:\n\n - \"supported_versions\" is REQUIRED for all ClientHello, ServerHello\n and HelloRetryRequest messages.\n\n - \"signature_algorithms\" is REQUIRED for certificate authentication.\n\n - \"supported_groups\" is REQUIRED for ClientHello messages using DHE\n or ECDHE key exchange.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 104]\n\f\nInternet-Draft TLS March 2018\n\n\n - \"key_share\" is REQUIRED for DHE or ECDHE key exchange.\n\n - \"pre_shared_key\" is REQUIRED for PSK key agreement.\n\n - \"psk_key_exchange_modes\" is REQUIRED for PSK key agreement.\n\n A client is considered to be attempting to negotiate using this\n specification if the ClientHello contains a \"supported_versions\"\n extension with 0x0304 contained in its body. Such a ClientHello\n message MUST meet the following requirements:\n\n - If not containing a \"pre_shared_key\" extension, it MUST contain\n both a \"signature_algorithms\" extension and a \"supported_groups\"\n extension.\n\n - If containing a \"supported_groups\" extension, it MUST also contain\n a \"key_share\" extension, and vice versa. An empty\n KeyShare.client_shares vector is permitted.\n\n Servers receiving a ClientHello which does not conform to these\n requirements MUST abort the handshake with a \"missing_extension\"\n alert.\n\n Additionally, all implementations MUST support use of the\n \"server_name\" extension with applications capable of using it.\n Servers MAY require clients to send a valid \"server_name\" extension.\n Servers requiring this extension SHOULD respond to a ClientHello\n lacking a \"server_name\" extension by terminating the connection with\n a \"missing_extension\" alert.\n\n9.3. Protocol Invariants\n\n This section describes invariants that TLS endpoints and middleboxes\n MUST follow. It also applies to earlier versions of TLS.\n\n TLS is designed to be securely and compatibly extensible. Newer\n clients or servers, when communicating with newer peers, should\n negotiate the most preferred common parameters. The TLS handshake\n provides downgrade protection: Middleboxes passing traffic between a\n newer client and newer server without terminating TLS should be\n unable to influence the handshake (see Appendix E.1). At the same\n time, deployments update at different rates, so a newer client or\n server MAY continue to support older parameters, which would allow it\n to interoperate with older endpoints.\n\n For this to work, implementations MUST correctly handle extensible\n fields:\n\n\n\n\nRescorla Expires September 21, 2018 [Page 105]\n\f\nInternet-Draft TLS March 2018\n\n\n - A client sending a ClientHello MUST support all parameters\n advertised in it. Otherwise, the server may fail to interoperate\n by selecting one of those parameters.\n\n - A server receiving a ClientHello MUST correctly ignore all\n unrecognized cipher suites, extensions, and other parameters.\n Otherwise, it may fail to interoperate with newer clients. In TLS\n 1.3, a client receiving a CertificateRequest or NewSessionTicket\n MUST also ignore all unrecognized extensions.\n\n - A middlebox which terminates a TLS connection MUST behave as a\n compliant TLS server (to the original client), including having a\n certificate which the client is willing to accept, and as a\n compliant TLS client (to the original server), including verifying\n the original server's certificate. In particular, it MUST\n generate its own ClientHello containing only parameters it\n understands, and it MUST generate a fresh ServerHello random\n value, rather than forwarding the endpoint's value.\n\n Note that TLS's protocol requirements and security analysis only\n apply to the two connections separately. Safely deploying a TLS\n terminator requires additional security considerations which are\n beyond the scope of this document.\n\n - An middlebox which forwards ClientHello parameters it does not\n understand MUST NOT process any messages beyond that ClientHello.\n It MUST forward all subsequent traffic unmodified. Otherwise, it\n may fail to interoperate with newer clients and servers.\n\n Forwarded ClientHellos may contain advertisements for features not\n supported by the middlebox, so the response may include future TLS\n additions the middlebox does not recognize. These additions MAY\n change any message beyond the ClientHello arbitrarily. In\n particular, the values sent in the ServerHello might change, the\n ServerHello format might change, and the TLSCiphertext format\n might change.\n\n The design of TLS 1.3 was constrained by widely-deployed non-\n compliant TLS middleboxes (see Appendix D.4), however it does not\n relax the invariants. Those middleboxes continue to be non-\n compliant.\n\n10. Security Considerations\n\n Security issues are discussed throughout this memo, especially in\n Appendix C, Appendix D, and Appendix E.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 106]\n\f\nInternet-Draft TLS March 2018\n\n\n11. IANA Considerations\n\n This document uses several registries that were originally created in\n [RFC4346]. IANA [SHALL update/has updated] these to reference this\n document. The registries and their allocation policies are below:\n\n - TLS Cipher Suite Registry: values with the first byte in the range\n 0-254 (decimal) are assigned via Specification Required [RFC8126].\n Values with the first byte 255 (decimal) are reserved for Private\n Use [RFC8126].\n\n IANA [SHALL add/has added] the cipher suites listed in\n Appendix B.4 to the registry. The \"Value\" and \"Description\"\n columns are taken from the table. The \"DTLS-OK\" and \"Recommended\"\n columns are both marked as \"Yes\" for each new cipher suite.\n [[This assumes [I-D.ietf-tls-iana-registry-updates] has been\n applied.]]\n\n - TLS ContentType Registry: Future values are allocated via\n Standards Action [RFC8126].\n\n - TLS Alert Registry: Future values are allocated via Standards\n Action [RFC8126]. IANA [SHALL update/has updated] this registry\n to include values for \"missing_extension\" and\n \"certificate_required\". The \"DTLS-OK\" column is marked as \"Yes\"\n for each new alert.\n\n - TLS HandshakeType Registry: Future values are allocated via\n Standards Action [RFC8126]. IANA [SHALL update/has updated] this\n registry to rename item 4 from \"NewSessionTicket\" to\n \"new_session_ticket\" and to add the\n \"hello_retry_request_RESERVED\", \"encrypted_extensions\",\n \"end_of_early_data\", \"key_update\", and \"message_hash\" values. The\n \"DTLS-OK\" are marked as \"Yes\" for each of these additions.\n\n This document also uses the TLS ExtensionType Registry originally\n created in [RFC4366]. IANA has updated it to reference this\n document. Changes to the registry follow:\n\n - IANA [SHALL update/has updated] the registration policy as\n follows:\n\n Values with the first byte in the range 0-254 (decimal) are\n assigned via Specification Required [RFC8126]. Values with the\n first byte 255 (decimal) are reserved for Private Use [RFC8126].\n\n - IANA [SHALL update/has updated] this registry to include the\n \"key_share\", \"pre_shared_key\", \"psk_key_exchange_modes\",\n\n\n\nRescorla Expires September 21, 2018 [Page 107]\n\f\nInternet-Draft TLS March 2018\n\n\n \"early_data\", \"cookie\", \"supported_versions\",\n \"certificate_authorities\", \"oid_filters\", \"post_handshake_auth\",\n and \"signature_algorithms_cert\", extensions with the values\n defined in this document and the Recommended value of \"Yes\".\n\n - IANA [SHALL update/has updated] this registry to include a \"TLS\n 1.3\" column which lists the messages in which the extension may\n appear. This column [SHALL be/has been] initially populated from\n the table in Section 4.2 with any extension not listed there\n marked as \"-\" to indicate that it is not used by TLS 1.3.\n\n In addition, this document defines two new registries to be\n maintained by IANA:\n\n - TLS SignatureScheme Registry: Values with the first byte in the\n range 0-253 (decimal) are assigned via Specification Required\n [RFC8126]. Values with the first byte 254 or 255 (decimal) are\n reserved for Private Use [RFC8126]. Values with the first byte in\n the range 0-6 or with the second byte in the range 0-3 that are\n not currently allocated are reserved for backwards compatibility.\n This registry SHALL have a \"Recommended\" column. The registry\n [shall be/ has been] initially populated with the values described\n in Section 4.2.3. The following values SHALL be marked as\n \"Recommended\": ecdsa_secp256r1_sha256, ecdsa_secp384r1_sha384,\n rsa_pss_rsae_sha256, rsa_pss_rsae_sha384, rsa_pss_rsae_sha512,\n rsa_pss_pss_sha256, rsa_pss_pss_sha384, rsa_pss_pss_sha512, and\n ed25519.\n\n - TLS PskKeyExchangeMode Registry: Values in the range 0-253\n (decimal) are assigned via Specification Required [RFC8126].\n Values with the first byte 254 or 255 (decimal) are reserved for\n Private Use [RFC8126]. This registry SHALL have a \"Recommended\"\n column. The registry [shall be/ has been] initially populated\n psk_ke (0) and psk_dhe_ke (1). Both SHALL be marked as\n \"Recommended\".\n\n12. References\n\n12.1. Normative References\n\n [DH] Diffie, W. and M. Hellman, \"New Directions in\n Cryptography\", IEEE Transactions on Information Theory,\n V.IT-22 n.6 , June 1977.\n\n [DH76] Diffie, W. and M. Hellman, \"New directions in\n cryptography\", IEEE Transactions on Information\n Theory Vol. 22, pp. 644-654, DOI 10.1109/tit.1976.1055638,\n November 1976.\n\n\n\nRescorla Expires September 21, 2018 [Page 108]\n\f\nInternet-Draft TLS March 2018\n\n\n [GCM] Dworkin, M., \"Recommendation for Block Cipher Modes of\n Operation: Galois/Counter Mode (GCM) and GMAC\",\n NIST Special Publication 800-38D, November 2007.\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC: Keyed-\n Hashing for Message Authentication\", RFC 2104,\n DOI 10.17487/RFC2104, February 1997, .\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119,\n DOI 10.17487/RFC2119, March 1997, .\n\n [RFC5116] McGrew, D., \"An Interface and Algorithms for Authenticated\n Encryption\", RFC 5116, DOI 10.17487/RFC5116, January 2008,\n .\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, DOI 10.17487/RFC5280, May 2008,\n .\n\n [RFC5705] Rescorla, E., \"Keying Material Exporters for Transport\n Layer Security (TLS)\", RFC 5705, DOI 10.17487/RFC5705,\n March 2010, .\n\n [RFC5756] Turner, S., Brown, D., Yiu, K., Housley, R., and T. Polk,\n \"Updates for RSAES-OAEP and RSASSA-PSS Algorithm\n Parameters\", RFC 5756, DOI 10.17487/RFC5756, January 2010,\n .\n\n [RFC5869] Krawczyk, H. and P. Eronen, \"HMAC-based Extract-and-Expand\n Key Derivation Function (HKDF)\", RFC 5869,\n DOI 10.17487/RFC5869, May 2010, .\n\n [RFC6066] Eastlake 3rd, D., \"Transport Layer Security (TLS)\n Extensions: Extension Definitions\", RFC 6066,\n DOI 10.17487/RFC6066, January 2011, .\n\n [RFC6655] McGrew, D. and D. Bailey, \"AES-CCM Cipher Suites for\n Transport Layer Security (TLS)\", RFC 6655,\n DOI 10.17487/RFC6655, July 2012, .\n\n\n\n\nRescorla Expires September 21, 2018 [Page 109]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, DOI 10.17487/RFC6960, June 2013,\n .\n\n [RFC6961] Pettersen, Y., \"The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\", RFC 6961,\n DOI 10.17487/RFC6961, June 2013, .\n\n [RFC6962] Laurie, B., Langley, A., and E. Kasper, \"Certificate\n Transparency\", RFC 6962, DOI 10.17487/RFC6962, June 2013,\n .\n\n [RFC6979] Pornin, T., \"Deterministic Usage of the Digital Signature\n Algorithm (DSA) and Elliptic Curve Digital Signature\n Algorithm (ECDSA)\", RFC 6979, DOI 10.17487/RFC6979, August\n 2013, .\n\n [RFC7301] Friedl, S., Popov, A., Langley, A., and E. Stephan,\n \"Transport Layer Security (TLS) Application-Layer Protocol\n Negotiation Extension\", RFC 7301, DOI 10.17487/RFC7301,\n July 2014, .\n\n [RFC7507] Moeller, B. and A. Langley, \"TLS Fallback Signaling Cipher\n Suite Value (SCSV) for Preventing Protocol Downgrade\n Attacks\", RFC 7507, DOI 10.17487/RFC7507, April 2015,\n .\n\n [RFC7539] Nir, Y. and A. Langley, \"ChaCha20 and Poly1305 for IETF\n Protocols\", RFC 7539, DOI 10.17487/RFC7539, May 2015,\n .\n\n [RFC7748] Langley, A., Hamburg, M., and S. Turner, \"Elliptic Curves\n for Security\", RFC 7748, DOI 10.17487/RFC7748, January\n 2016, .\n\n [RFC7919] Gillmor, D., \"Negotiated Finite Field Diffie-Hellman\n Ephemeral Parameters for Transport Layer Security (TLS)\",\n RFC 7919, DOI 10.17487/RFC7919, August 2016,\n .\n\n [RFC8017] Moriarty, K., Ed., Kaliski, B., Jonsson, J., and A. Rusch,\n \"PKCS #1: RSA Cryptography Specifications Version 2.2\",\n RFC 8017, DOI 10.17487/RFC8017, November 2016,\n .\n\n\n\n\nRescorla Expires September 21, 2018 [Page 110]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC8032] Josefsson, S. and I. Liusvaara, \"Edwards-Curve Digital\n Signature Algorithm (EdDSA)\", RFC 8032,\n DOI 10.17487/RFC8032, January 2017, .\n\n [RFC8126] Cotton, M., Leiba, B., and T. Narten, \"Guidelines for\n Writing an IANA Considerations Section in RFCs\", BCP 26,\n RFC 8126, DOI 10.17487/RFC8126, June 2017,\n .\n\n [RFC8174] Leiba, B., \"Ambiguity of Uppercase vs Lowercase in RFC\n 2119 Key Words\", BCP 14, RFC 8174, DOI 10.17487/RFC8174,\n May 2017, .\n\n [SHS] Dang, Q., \"Secure Hash Standard\", National Institute of\n Standards and Technology report,\n DOI 10.6028/nist.fips.180-4, July 2015.\n\n [X690] ITU-T, \"Information technology - ASN.1 encoding Rules:\n Specification of Basic Encoding Rules (BER), Canonical\n Encoding Rules (CER) and Distinguished Encoding Rules\n (DER)\", ISO/IEC 8825-1:2002, 2002.\n\n [X962] ANSI, \"Public Key Cryptography For The Financial Services\n Industry: The Elliptic Curve Digital Signature Algorithm\n (ECDSA)\", ANSI X9.62, 1998.\n\n12.2. Informative References\n\n [AEAD-LIMITS]\n Luykx, A. and K. Paterson, \"Limits on Authenticated\n Encryption Use in TLS\", 2016,\n .\n\n [Anon18] Anonymous, A., \"Secure Channels for Multiplexed Data\n Streams: Analyzing the TLS 1.3 Record Layer Without\n Elision\", In submission to CRYPTO 2018. RFC EDITOR: PLEASE\n UPDATE THIS REFERENCE AFTER FINAL NOTIFICATION\n (2018-4-29). , 2018.\n\n [BBFKZG16]\n Bhargavan, K., Brzuska, C., Fournet, C., Kohlweiss, M.,\n Zanella-Beguelin, S., and M. Green, \"Downgrade Resilience\n in Key-Exchange Protocols\", Proceedings of IEEE Symposium\n on Security and Privacy (Oakland) 2016 , 2016.\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 111]\n\f\nInternet-Draft TLS March 2018\n\n\n [BBK17] Bhargavan, K., Blanchet, B., and N. Kobeissi, \"Verified\n Models and Reference Implementations for the TLS 1.3\n Standard Candidate\", Proceedings of IEEE Symposium on\n Security and Privacy (Oakland) 2017 , 2017.\n\n [BDFKPPRSZZ16]\n Bhargavan, K., Delignat-Lavaud, A., Fournet, C.,\n Kohlweiss, M., Pan, J., Protzenko, J., Rastogi, A., Swamy,\n N., Zanella-Beguelin, S., and J. Zinzindohoue,\n \"Implementing and Proving the TLS 1.3 Record Layer\",\n Proceedings of IEEE Symposium on Security and Privacy\n (Oakland) 2017 , December 2016,\n .\n\n [Ben17a] Benjamin, D., \"Presentation before the TLS WG at IETF\n 100\", 2017,\n .\n\n [Ben17b] Benjamin, D., \"Additional TLS 1.3 results from Chrome\",\n 2017, .\n\n [BMMT15] Badertscher, C., Matt, C., Maurer, U., and B. Tackmann,\n \"Augmented Secure Channels and the Goal of the TLS 1.3\n Record Layer\", ProvSec 2015 , September 2015,\n .\n\n [BT16] Bellare, M. and B. Tackmann, \"The Multi-User Security of\n Authenticated Encryption: AES-GCM in TLS 1.3\", Proceedings\n of CRYPTO 2016 , 2016, .\n\n [CCG16] Cohn-Gordon, K., Cremers, C., and L. Garratt, \"On Post-\n Compromise Security\", IEEE Computer Security Foundations\n Symposium , 2015.\n\n [CHECKOWAY]\n Checkoway, S., Shacham, H., Maskiewicz, J., Garman, C.,\n Fried, J., Cohney, S., Green, M., Heninger, N., Weinmann,\n R., and E. Rescorla, \"A Systematic Analysis of the Juniper\n Dual EC Incident\", Proceedings of the 2016 ACM SIGSAC\n Conference on Computer and Communications Security\n - CCS'16, DOI 10.1145/2976749.2978395, 2016.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 112]\n\f\nInternet-Draft TLS March 2018\n\n\n [CHHSV17] Cremers, C., Horvat, M., Hoyland, J., van der Merwe, T.,\n and S. Scott, \"Awkward Handshake: Possible mismatch of\n client/server view on client authentication in post-\n handshake mode in Revision 18\", 2017,\n .\n\n [CHSV16] Cremers, C., Horvat, M., Scott, S., and T. van der Merwe,\n \"Automated Analysis and Verification of TLS 1.3: 0-RTT,\n Resumption and Delayed Authentication\", Proceedings of\n IEEE Symposium on Security and Privacy (Oakland) 2016 ,\n 2016, .\n\n [CK01] Canetti, R. and H. Krawczyk, \"Analysis of Key-Exchange\n Protocols and Their Use for Building Secure Channels\",\n Proceedings of Eurocrypt 2001 , 2001.\n\n [CLINIC] Miller, B., Huang, L., Joseph, A., and J. Tygar, \"I Know\n Why You Went to the Clinic: Risks and Realization of HTTPS\n Traffic Analysis\", Privacy Enhancing Technologies pp.\n 143-163, DOI 10.1007/978-3-319-08506-7_8, 2014.\n\n [DFGS15] Dowling, B., Fischlin, M., Guenther, F., and D. Stebila,\n \"A Cryptographic Analysis of the TLS 1.3 draft-10 Full and\n Pre-shared Key Handshake Protocol\", Proceedings of ACM CCS\n 2015 , 2015, .\n\n [DFGS16] Dowling, B., Fischlin, M., Guenther, F., and D. Stebila,\n \"A Cryptographic Analysis of the TLS 1.3 draft-10 Full and\n Pre-shared Key Handshake Protocol\", TRON 2016 , 2016,\n .\n\n [DOW92] Diffie, W., van Oorschot, P., and M. Wiener,\n \"Authentication and authenticated key exchanges\", Designs,\n Codes and Cryptography , 1992.\n\n [DSS] National Institute of Standards and Technology, U.S.\n Department of Commerce, \"Digital Signature Standard,\n version 4\", NIST FIPS PUB 186-4, 2013.\n\n [ECDSA] American National Standards Institute, \"Public Key\n Cryptography for the Financial Services Industry: The\n Elliptic Curve Digital Signature Algorithm (ECDSA)\",\n ANSI ANS X9.62-2005, November 2005.\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 113]\n\f\nInternet-Draft TLS March 2018\n\n\n [FG17] Fischlin, M. and F. Guenther, \"Replay Attacks on Zero\n Round-Trip Time: The Case of the TLS 1.3 Handshake\n Candidates\", Proceedings of Euro S\"P 2017 , 2017,\n .\n\n [FGSW16] Fischlin, M., Guenther, F., Schmidt, B., and B. Warinschi,\n \"Key Confirmation in Key Exchange: A Formal Treatment and\n Implications for TLS 1.3\", Proceedings of IEEE Symposium\n on Security and Privacy (Oakland) 2016 , 2016,\n .\n\n [FW15] Florian Weimer, ., \"Factoring RSA Keys With TLS Perfect\n Forward Secrecy\", September 2015.\n\n [HCJ16] Husak, M., Čermak, M., Jirsik, T., and P.\n Čeleda, \"HTTPS traffic analysis and client\n identification using passive SSL/TLS fingerprinting\",\n EURASIP Journal on Information Security Vol. 2016,\n DOI 10.1186/s13635-016-0030-7, February 2016.\n\n [HGFS15] Hlauschek, C., Gruber, M., Fankhauser, F., and C. Schanes,\n \"Prying Open Pandora's Box: KCI Attacks against TLS\",\n Proceedings of USENIX Workshop on Offensive Technologies ,\n 2015.\n\n [I-D.ietf-tls-iana-registry-updates]\n Salowey, J. and S. Turner, \"IANA Registry Updates for TLS\n and DTLS\", draft-ietf-tls-iana-registry-updates-04 (work\n in progress), February 2018.\n\n [I-D.ietf-tls-tls13-vectors]\n Thomson, M., \"Example Handshake Traces for TLS 1.3\",\n draft-ietf-tls-tls13-vectors-03 (work in progress),\n December 2017.\n\n [IEEE1363]\n IEEE, \"Standard Specifications for Public Key\n Cryptography\", IEEE 1363 , 2000.\n\n [JSS15] Jager, T., Schwenk, J., and J. Somorovsky, \"On the\n Security of TLS 1.3 and QUIC Against Weaknesses in PKCS#1\n v1.5 Encryption\", Proceedings of ACM CCS 2015 , 2015,\n .\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 114]\n\f\nInternet-Draft TLS March 2018\n\n\n [KEYAGREEMENT]\n Barker, E., Chen, L., Roginsky, A., and M. Smid,\n \"Recommendation for Pair-Wise Key Establishment Schemes\n Using Discrete Logarithm Cryptography\", National Institute\n of Standards and Technology report,\n DOI 10.6028/nist.sp.800-56ar2, May 2013.\n\n [Kraw10] Krawczyk, H., \"Cryptographic Extraction and Key\n Derivation: The HKDF Scheme\", Proceedings of CRYPTO 2010 ,\n 2010, .\n\n [Kraw16] Krawczyk, H., \"A Unilateral-to-Mutual Authentication\n Compiler for Key Exchange (with Applications to Client\n Authentication in TLS 1.3\", Proceedings of ACM CCS 2016 ,\n 2016, .\n\n [KW16] Krawczyk, H. and H. Wee, \"The OPTLS Protocol and TLS 1.3\",\n Proceedings of Euro S\"P 2016 , 2016,\n .\n\n [LXZFH16] Li, X., Xu, J., Feng, D., Zhang, Z., and H. Hu, \"Multiple\n Handshakes Security of TLS 1.3 Candidates\", Proceedings of\n IEEE Symposium on Security and Privacy (Oakland) 2016 ,\n 2016, .\n\n [Mac17] MacCarthaigh, C., \"Security Review of TLS1.3 0-RTT\", 2017,\n .\n\n [PSK-FINISHED]\n Cremers, C., Horvat, M., van der Merwe, T., and S. Scott,\n \"Revision 10: possible attack if client authentication is\n allowed during PSK\", 2015, .\n\n [REKEY] Abdalla, M. and M. Bellare, \"Increasing the Lifetime of a\n Key: A Comparative Analysis of the Security of Re-keying\n Techniques\", ASIACRYPT2000 , October 2000.\n\n [Res17a] Rescorla, E., \"Preliminary data on Firefox TLS 1.3\n Middlebox experiment\", 2017, .\n\n [Res17b] Rescorla, E., \"More compatibility measurement results\",\n 2017, .\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 115]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC3552] Rescorla, E. and B. Korver, \"Guidelines for Writing RFC\n Text on Security Considerations\", BCP 72, RFC 3552,\n DOI 10.17487/RFC3552, July 2003, .\n\n [RFC4086] Eastlake 3rd, D., Schiller, J., and S. Crocker,\n \"Randomness Requirements for Security\", BCP 106, RFC 4086,\n DOI 10.17487/RFC4086, June 2005, .\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346,\n DOI 10.17487/RFC4346, April 2006, .\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, DOI 10.17487/RFC4366, April 2006,\n .\n\n [RFC4492] Blake-Wilson, S., Bolyard, N., Gupta, V., Hawk, C., and B.\n Moeller, \"Elliptic Curve Cryptography (ECC) Cipher Suites\n for Transport Layer Security (TLS)\", RFC 4492,\n DOI 10.17487/RFC4492, May 2006, .\n\n [RFC5077] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 5077, DOI 10.17487/RFC5077,\n January 2008, .\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246,\n DOI 10.17487/RFC5246, August 2008, .\n\n [RFC5764] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for the Secure\n Real-time Transport Protocol (SRTP)\", RFC 5764,\n DOI 10.17487/RFC5764, May 2010, .\n\n [RFC5929] Altman, J., Williams, N., and L. Zhu, \"Channel Bindings\n for TLS\", RFC 5929, DOI 10.17487/RFC5929, July 2010,\n .\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 116]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC6091] Mavrogiannopoulos, N. and D. Gillmor, \"Using OpenPGP Keys\n for Transport Layer Security (TLS) Authentication\",\n RFC 6091, DOI 10.17487/RFC6091, February 2011,\n .\n\n [RFC6176] Turner, S. and T. Polk, \"Prohibiting Secure Sockets Layer\n (SSL) Version 2.0\", RFC 6176, DOI 10.17487/RFC6176, March\n 2011, .\n\n [RFC6347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security Version 1.2\", RFC 6347, DOI 10.17487/RFC6347,\n January 2012, .\n\n [RFC6520] Seggelmann, R., Tuexen, M., and M. Williams, \"Transport\n Layer Security (TLS) and Datagram Transport Layer Security\n (DTLS) Heartbeat Extension\", RFC 6520,\n DOI 10.17487/RFC6520, February 2012, .\n\n [RFC7230] Fielding, R., Ed. and J. Reschke, Ed., \"Hypertext Transfer\n Protocol (HTTP/1.1): Message Syntax and Routing\",\n RFC 7230, DOI 10.17487/RFC7230, June 2014,\n .\n\n [RFC7250] Wouters, P., Ed., Tschofenig, H., Ed., Gilmore, J.,\n Weiler, S., and T. Kivinen, \"Using Raw Public Keys in\n Transport Layer Security (TLS) and Datagram Transport\n Layer Security (DTLS)\", RFC 7250, DOI 10.17487/RFC7250,\n June 2014, .\n\n [RFC7465] Popov, A., \"Prohibiting RC4 Cipher Suites\", RFC 7465,\n DOI 10.17487/RFC7465, February 2015, .\n\n [RFC7568] Barnes, R., Thomson, M., Pironti, A., and A. Langley,\n \"Deprecating Secure Sockets Layer Version 3.0\", RFC 7568,\n DOI 10.17487/RFC7568, June 2015, .\n\n [RFC7627] Bhargavan, K., Ed., Delignat-Lavaud, A., Pironti, A.,\n Langley, A., and M. Ray, \"Transport Layer Security (TLS)\n Session Hash and Extended Master Secret Extension\",\n RFC 7627, DOI 10.17487/RFC7627, September 2015,\n .\n\n [RFC7685] Langley, A., \"A Transport Layer Security (TLS) ClientHello\n Padding Extension\", RFC 7685, DOI 10.17487/RFC7685,\n October 2015, .\n\n\n\nRescorla Expires September 21, 2018 [Page 117]\n\f\nInternet-Draft TLS March 2018\n\n\n [RFC7924] Santesson, S. and H. Tschofenig, \"Transport Layer Security\n (TLS) Cached Information Extension\", RFC 7924,\n DOI 10.17487/RFC7924, July 2016, .\n\n [RFC8305] Schinazi, D. and T. Pauly, \"Happy Eyeballs Version 2:\n Better Connectivity Using Concurrency\", RFC 8305,\n DOI 10.17487/RFC8305, December 2017, .\n\n [RSA] Rivest, R., Shamir, A., and L. Adleman, \"A Method for\n Obtaining Digital Signatures and Public-Key\n Cryptosystems\", Communications of the ACM v. 21, n. 2, pp.\n 120-126., February 1978.\n\n [SIGMA] Krawczyk, H., \"SIGMA: the 'SIGn-and-MAc' approach to\n authenticated Diffie-Hellman and its use in the IKE\n protocols\", Proceedings of CRYPTO 2003 , 2003.\n\n [SLOTH] Bhargavan, K. and G. Leurent, \"Transcript Collision\n Attacks: Breaking Authentication in TLS, IKE, and SSH\",\n Network and Distributed System Security Symposium (NDSS\n 2016) , 2016.\n\n [SSL2] Hickman, K., \"The SSL Protocol\", February 1995.\n\n [SSL3] Freier, A., Karlton, P., and P. Kocher, \"The SSL 3.0\n Protocol\", November 1996.\n\n [TIMING] Boneh, D. and D. Brumley, \"Remote timing attacks are\n practical\", USENIX Security Symposium, 2003.\n\n [X501] \"Information Technology - Open Systems Interconnection -\n The Directory: Models\", ITU-T X.501, 1993.\n\n12.3. URIs\n\n [1] mailto:[redacted-email]\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 118]\n\f\nInternet-Draft TLS March 2018\n\n\nAppendix A. State Machine\n\n This section provides a summary of the legal state transitions for\n the client and server handshakes. State names (in all capitals,\n e.g., START) have no formal meaning but are provided for ease of\n comprehension. Actions which are taken only in certain circumstances\n are indicated in []. The notation \"K_{send,recv} = foo\" means \"set\n the send/recv key to the given key\".\n\nA.1. Client\n\n START <----+\n Send ClientHello | | Recv HelloRetryRequest\n [K_send = early data] | |\n v |\n / WAIT_SH ----+\n | | Recv ServerHello\n | | K_recv = handshake\n Can | V\n send | WAIT_EE\n early | | Recv EncryptedExtensions\n data | +--------+--------+\n | Using | | Using certificate\n | PSK | v\n | | WAIT_CERT_CR\n | | Recv | | Recv CertificateRequest\n | | Certificate | v\n | | | WAIT_CERT\n | | | | Recv Certificate\n | | v v\n | | WAIT_CV\n | | | Recv CertificateVerify\n | +> WAIT_FINISHED <+\n | | Recv Finished\n \\ | [Send EndOfEarlyData]\n | K_send = handshake\n | [Send Certificate [+ CertificateVerify]]\n Can send | Send Finished\n app data --> | K_send = K_recv = application\n after here v\n CONNECTED\n\n Note that with the transitions as shown above, clients may send\n alerts that derive from post-ServerHello messages in the clear or\n with the early data keys. If clients need to send such alerts, they\n SHOULD first rekey to the handshake keys if possible.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 119]\n\f\nInternet-Draft TLS March 2018\n\n\nA.2. Server\n\n START <-----+\n Recv ClientHello | | Send HelloRetryRequest\n v |\n RECVD_CH ----+\n | Select parameters\n v\n NEGOTIATED\n | Send ServerHello\n | K_send = handshake\n | Send EncryptedExtensions\n | [Send CertificateRequest]\n Can send | [Send Certificate + CertificateVerify]\n app data | Send Finished\n after --> | K_send = application\n here +--------+--------+\n No 0-RTT | | 0-RTT\n | |\n K_recv = handshake | | K_recv = early data\n [Skip decrypt errors] | +------> WAIT_EOED -+\n | | Recv | | Recv EndOfEarlyData\n | | early data | | K_recv = handshake\n | +------------+ |\n | |\n +> WAIT_FLIGHT2 <--------+\n |\n +--------+--------+\n No auth | | Client auth\n | |\n | v\n | WAIT_CERT\n | Recv | | Recv Certificate\n | empty | v\n | Certificate | WAIT_CV\n | | | Recv\n | v | CertificateVerify\n +-> WAIT_FINISHED <---+\n | Recv Finished\n | K_recv = application\n v\n CONNECTED\n\nAppendix B. Protocol Data Structures and Constant Values\n\n This section provides the normative protocol types and constants\n definitions. Values listed as _RESERVED were used in previous\n versions of TLS and are listed here for completeness. TLS 1.3\n\n\n\nRescorla Expires September 21, 2018 [Page 120]\n\f\nInternet-Draft TLS March 2018\n\n\n implementations MUST NOT send them but might receive them from older\n TLS implementations.\n\nB.1. Record Layer\n\n enum {\n invalid(0),\n change_cipher_spec(20),\n alert(21),\n handshake(22),\n application_data(23),\n (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion legacy_record_version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n struct {\n opaque content[TLSPlaintext.length];\n ContentType type;\n uint8 zeros[length_of_padding];\n } TLSInnerPlaintext;\n\n struct {\n ContentType opaque_type = application_data; /* 23 */\n ProtocolVersion legacy_record_version = 0x0303; /* TLS v1.2 */\n uint16 length;\n opaque encrypted_record[TLSCiphertext.length];\n } TLSCiphertext;\n\nB.2. Alert Messages\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 121]\n\f\nInternet-Draft TLS March 2018\n\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure_RESERVED(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n inappropriate_fallback(86),\n user_canceled(90),\n no_renegotiation_RESERVED(100),\n missing_extension(109),\n unsupported_extension(110),\n certificate_unobtainable_RESERVED(111),\n unrecognized_name(112),\n bad_certificate_status_response(113),\n bad_certificate_hash_value_RESERVED(114),\n unknown_psk_identity(115),\n certificate_required(116),\n no_application_protocol(120),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 122]\n\f\nInternet-Draft TLS March 2018\n\n\nB.3. Handshake Protocol\n\n enum {\n hello_request_RESERVED(0),\n client_hello(1),\n server_hello(2),\n hello_verify_request_RESERVED(3),\n new_session_ticket(4),\n end_of_early_data(5),\n hello_retry_request_RESERVED(6),\n encrypted_extensions(8),\n certificate(11),\n server_key_exchange_RESERVED(12),\n certificate_request(13),\n server_hello_done_RESERVED(14),\n certificate_verify(15),\n client_key_exchange_RESERVED(16),\n finished(20),\n key_update(24),\n message_hash(254),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (Handshake.msg_type) {\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case end_of_early_data: EndOfEarlyData;\n case encrypted_extensions: EncryptedExtensions;\n case certificate_request: CertificateRequest;\n case certificate: Certificate;\n case certificate_verify: CertificateVerify;\n case finished: Finished;\n case new_session_ticket: NewSessionTicket;\n case key_update: KeyUpdate;\n };\n } Handshake;\n\nB.3.1. Key Exchange Messages\n\n uint16 ProtocolVersion;\n opaque Random[32];\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n struct {\n\n\n\nRescorla Expires September 21, 2018 [Page 123]\n\f\nInternet-Draft TLS March 2018\n\n\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id<0..32>;\n CipherSuite cipher_suites<2..2^16-2>;\n opaque legacy_compression_methods<1..2^8-1>;\n Extension extensions<8..2^16-1>;\n } ClientHello;\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id_echo<0..32>;\n CipherSuite cipher_suite;\n uint8 legacy_compression_method = 0;\n Extension extensions<6..2^16-1>;\n } ServerHello;\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n server_name(0), /* RFC 6066 */\n max_fragment_length(1), /* RFC 6066 */\n status_request(5), /* RFC 6066 */\n supported_groups(10), /* RFC 4492, 7919 */\n signature_algorithms(13), /* [[this document]] */\n use_srtp(14), /* RFC 5764 */\n heartbeat(15), /* RFC 6520 */\n application_layer_protocol_negotiation(16), /* RFC 7301 */\n signed_certificate_timestamp(18), /* RFC 6962 */\n client_certificate_type(19), /* RFC 7250 */\n server_certificate_type(20), /* RFC 7250 */\n padding(21), /* RFC 7685 */\n RESERVED(40), /* Used but never assigned */\n pre_shared_key(41), /* [[this document]] */\n early_data(42), /* [[this document]] */\n supported_versions(43), /* [[this document]] */\n cookie(44), /* [[this document]] */\n psk_key_exchange_modes(45), /* [[this document]] */\n RESERVED(46), /* Used but never assigned */\n certificate_authorities(47), /* [[this document]] */\n oid_filters(48), /* [[this document]] */\n post_handshake_auth(49), /* [[this document]] */\n signature_algorithms_cert(50), /* [[this document]] */\n key_share(51), /* [[this document]] */\n (65535)\n\n\n\nRescorla Expires September 21, 2018 [Page 124]\n\f\nInternet-Draft TLS March 2018\n\n\n } ExtensionType;\n\n struct {\n NamedGroup group;\n opaque key_exchange<1..2^16-1>;\n } KeyShareEntry;\n\n struct {\n KeyShareEntry client_shares<0..2^16-1>;\n } KeyShareClientHello;\n\n struct {\n NamedGroup selected_group;\n } KeyShareHelloRetryRequest;\n\n struct {\n KeyShareEntry server_share;\n } KeyShareServerHello;\n\n struct {\n uint8 legacy_form = 4;\n opaque X[coordinate_length];\n opaque Y[coordinate_length];\n } UncompressedPointRepresentation;\n\n enum { psk_ke(0), psk_dhe_ke(1), (255) } PskKeyExchangeMode;\n\n struct {\n PskKeyExchangeMode ke_modes<1..255>;\n } PskKeyExchangeModes;\n\n struct {} Empty;\n\n struct {\n select (Handshake.msg_type) {\n case new_session_ticket: uint32 max_early_data_size;\n case client_hello: Empty;\n case encrypted_extensions: Empty;\n };\n } EarlyDataIndication;\n\n struct {\n opaque identity<1..2^16-1>;\n uint32 obfuscated_ticket_age;\n } PskIdentity;\n\n opaque PskBinderEntry<32..255>;\n\n\n\n\nRescorla Expires September 21, 2018 [Page 125]\n\f\nInternet-Draft TLS March 2018\n\n\n struct {\n PskIdentity identities<7..2^16-1>;\n PskBinderEntry binders<33..2^16-1>;\n } OfferedPsks;\n\n struct {\n select (Handshake.msg_type) {\n case client_hello: OfferedPsks;\n case server_hello: uint16 selected_identity;\n };\n } PreSharedKeyExtension;\n\nB.3.1.1. Version Extension\n\n struct {\n select (Handshake.msg_type) {\n case client_hello:\n ProtocolVersion versions<2..254>;\n\n case server_hello: /* and HelloRetryRequest */\n ProtocolVersion selected_version;\n };\n } SupportedVersions;\n\nB.3.1.2. Cookie Extension\n\n struct {\n opaque cookie<1..2^16-1>;\n } Cookie;\n\nB.3.1.3. Signature Algorithm Extension\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 126]\n\f\nInternet-Draft TLS March 2018\n\n\n enum {\n /* RSASSA-PKCS1-v1_5 algorithms */\n rsa_pkcs1_sha256(0x0401),\n rsa_pkcs1_sha384(0x0501),\n rsa_pkcs1_sha512(0x0601),\n\n /* ECDSA algorithms */\n ecdsa_secp256r1_sha256(0x0403),\n ecdsa_secp384r1_sha384(0x0503),\n ecdsa_secp521r1_sha512(0x0603),\n\n /* RSASSA-PSS algorithms with public key OID rsaEncryption */\n rsa_pss_rsae_sha256(0x0804),\n rsa_pss_rsae_sha384(0x0805),\n rsa_pss_rsae_sha512(0x0806),\n\n /* EdDSA algorithms */\n ed25519(0x0807),\n ed448(0x0808),\n\n /* RSASSA-PSS algorithms with public key OID RSASSA-PSS */\n rsa_pss_pss_sha256(0x0809),\n rsa_pss_pss_sha384(0x080a),\n rsa_pss_pss_sha512(0x080b),\n\n /* Legacy algorithms */\n rsa_pkcs1_sha1(0x0201),\n ecdsa_sha1(0x0203),\n\n /* Reserved Code Points */\n obsolete_RESERVED(0x0000..0x0200),\n dsa_sha1_RESERVED(0x0202),\n obsolete_RESERVED(0x0204..0x0400),\n dsa_sha256_RESERVED(0x0402),\n obsolete_RESERVED(0x0404..0x0500),\n dsa_sha384_RESERVED(0x0502),\n obsolete_RESERVED(0x0504..0x0600),\n dsa_sha512_RESERVED(0x0602),\n obsolete_RESERVED(0x0604..0x06FF),\n private_use(0xFE00..0xFFFF),\n (0xFFFF)\n } SignatureScheme;\n\n struct {\n SignatureScheme supported_signature_algorithms<2..2^16-2>;\n } SignatureSchemeList;\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 127]\n\f\nInternet-Draft TLS March 2018\n\n\nB.3.1.4. Supported Groups Extension\n\n enum {\n unallocated_RESERVED(0x0000),\n\n /* Elliptic Curve Groups (ECDHE) */\n obsolete_RESERVED(0x0001..0x0016),\n secp256r1(0x0017), secp384r1(0x0018), secp521r1(0x0019),\n obsolete_RESERVED(0x001A..0x001C),\n x25519(0x001D), x448(0x001E),\n\n /* Finite Field Groups (DHE) */\n ffdhe2048(0x0100), ffdhe3072(0x0101), ffdhe4096(0x0102),\n ffdhe6144(0x0103), ffdhe8192(0x0104),\n\n /* Reserved Code Points */\n ffdhe_private_use(0x01FC..0x01FF),\n ecdhe_private_use(0xFE00..0xFEFF),\n obsolete_RESERVED(0xFF01..0xFF02),\n (0xFFFF)\n } NamedGroup;\n\n struct {\n NamedGroup named_group_list<2..2^16-1>;\n } NamedGroupList;\n\n Values within \"obsolete_RESERVED\" ranges are used in previous\n versions of TLS and MUST NOT be offered or negotiated by TLS 1.3\n implementations. The obsolete curves have various known/theoretical\n weaknesses or have had very little usage, in some cases only due to\n unintentional server configuration issues. They are no longer\n considered appropriate for general use and should be assumed to be\n potentially unsafe. The set of curves specified here is sufficient\n for interoperability with all currently deployed and properly\n configured TLS implementations.\n\nB.3.2. Server Parameters Messages\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 128]\n\f\nInternet-Draft TLS March 2018\n\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n DistinguishedName authorities<3..2^16-1>;\n } CertificateAuthoritiesExtension;\n\n struct {\n opaque certificate_extension_oid<1..2^8-1>;\n opaque certificate_extension_values<0..2^16-1>;\n } OIDFilter;\n\n struct {\n OIDFilter filters<0..2^16-1>;\n } OIDFilterExtension;\n\n struct {} PostHandshakeAuth;\n\n struct {\n Extension extensions<0..2^16-1>;\n } EncryptedExtensions;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n Extension extensions<2..2^16-1>;\n } CertificateRequest;\n\nB.3.3. Authentication Messages\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 129]\n\f\nInternet-Draft TLS March 2018\n\n\n /* Managed by IANA */\n enum {\n X509(0),\n OpenPGP_RESERVED(1),\n RawPublicKey(2),\n (255)\n } CertificateType;\n\n struct {\n select (certificate_type) {\n case RawPublicKey:\n /* From RFC 7250 ASN.1_subjectPublicKeyInfo */\n opaque ASN1_subjectPublicKeyInfo<1..2^24-1>;\n\n case X509:\n opaque cert_data<1..2^24-1>;\n };\n Extension extensions<0..2^16-1>;\n } CertificateEntry;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n CertificateEntry certificate_list<0..2^24-1>;\n } Certificate;\n\n struct {\n SignatureScheme algorithm;\n opaque signature<0..2^16-1>;\n } CertificateVerify;\n\n struct {\n opaque verify_data[Hash.length];\n } Finished;\n\nB.3.4. Ticket Establishment\n\n struct {\n uint32 ticket_lifetime;\n uint32 ticket_age_add;\n opaque ticket_nonce<0..255>;\n opaque ticket<1..2^16-1>;\n Extension extensions<0..2^16-2>;\n } NewSessionTicket;\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 130]\n\f\nInternet-Draft TLS March 2018\n\n\nB.3.5. Updating Keys\n\n struct {} EndOfEarlyData;\n\n enum {\n update_not_requested(0), update_requested(1), (255)\n } KeyUpdateRequest;\n\n struct {\n KeyUpdateRequest request_update;\n } KeyUpdate;\n\nB.4. Cipher Suites\n\n A symmetric cipher suite defines the pair of the AEAD algorithm and\n hash algorithm to be used with HKDF. Cipher suite names follow the\n naming convention:\n\n CipherSuite TLS_AEAD_HASH = VALUE;\n\n +-----------+------------------------------------------------+\n | Component | Contents |\n +-----------+------------------------------------------------+\n | TLS | The string \"TLS\" |\n | | |\n | AEAD | The AEAD algorithm used for record protection |\n | | |\n | HASH | The hash algorithm used with HKDF |\n | | |\n | VALUE | The two byte ID assigned for this cipher suite |\n +-----------+------------------------------------------------+\n\n This specification defines the following cipher suites for use with\n TLS 1.3.\n\n +------------------------------+-------------+\n | Description | Value |\n +------------------------------+-------------+\n | TLS_AES_128_GCM_SHA256 | {0x13,0x01} |\n | | |\n | TLS_AES_256_GCM_SHA384 | {0x13,0x02} |\n | | |\n | TLS_CHACHA20_POLY1305_SHA256 | {0x13,0x03} |\n | | |\n | TLS_AES_128_CCM_SHA256 | {0x13,0x04} |\n | | |\n | TLS_AES_128_CCM_8_SHA256 | {0x13,0x05} |\n +------------------------------+-------------+\n\n\n\nRescorla Expires September 21, 2018 [Page 131]\n\f\nInternet-Draft TLS March 2018\n\n\n The corresponding AEAD algorithms AEAD_AES_128_GCM, AEAD_AES_256_GCM,\n and AEAD_AES_128_CCM are defined in [RFC5116].\n AEAD_CHACHA20_POLY1305 is defined in [RFC7539]. AEAD_AES_128_CCM_8\n is defined in [RFC6655]. The corresponding hash algorithms are\n defined in [SHS].\n\n Although TLS 1.3 uses the same cipher suite space as previous\n versions of TLS, TLS 1.3 cipher suites are defined differently, only\n specifying the symmetric ciphers, and cannot be used for TLS 1.2.\n Similarly, TLS 1.2 and lower cipher suites cannot be used with TLS\n 1.3.\n\n New cipher suite values are assigned by IANA as described in\n Section 11.\n\nAppendix C. Implementation Notes\n\n The TLS protocol cannot prevent many common security mistakes. This\n section provides several recommendations to assist implementors.\n [I-D.ietf-tls-tls13-vectors] provides test vectors for TLS 1.3\n handshakes.\n\nC.1. Random Number Generation and Seeding\n\n TLS requires a cryptographically secure pseudorandom number generator\n (CSPRNG). In most cases, the operating system provides an\n appropriate facility such as /dev/urandom, which should be used\n absent other (performance) concerns. It is RECOMMENDED to use an\n existing CSPRNG implementation in preference to crafting a new one.\n Many adequate cryptographic libraries are already available under\n favorable license terms. Should those prove unsatisfactory,\n [RFC4086] provides guidance on the generation of random values.\n\n TLS uses random values both in public protocol fields such as the\n public Random values in the ClientHello and ServerHello and to\n generate keying material. With a properly functioning CSPRNG, this\n does not present a security problem as it is not feasible to\n determine the CSPRNG state from its output. However, with a broken\n CSPRNG, it may be possible for an attacker to use the public output\n to determine the CSPRNG internal state and thereby predict the keying\n material, as documented in [CHECKOWAY]. Implementations can provide\n extra security against this form of attack by using separate CSPRNGs\n to generate public and private values.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 132]\n\f\nInternet-Draft TLS March 2018\n\n\nC.2. Certificates and Authentication\n\n Implementations are responsible for verifying the integrity of\n certificates and should generally support certificate revocation\n messages. Absent a specific indication from an application profile,\n Certificates should always be verified to ensure proper signing by a\n trusted Certificate Authority (CA). The selection and addition of\n trust anchors should be done very carefully. Users should be able to\n view information about the certificate and trust anchor.\n Applications SHOULD also enforce minimum and maximum key sizes. For\n example, certification paths containing keys or signatures weaker\n than 2048-bit RSA or 224-bit ECDSA are not appropriate for secure\n applications.\n\nC.3. Implementation Pitfalls\n\n Implementation experience has shown that certain parts of earlier TLS\n specifications are not easy to understand and have been a source of\n interoperability and security problems. Many of these areas have\n been clarified in this document but this appendix contains a short\n list of the most important things that require special attention from\n implementors.\n\n TLS protocol issues:\n\n - Do you correctly handle handshake messages that are fragmented to\n multiple TLS records (see Section 5.1)? Including corner cases\n like a ClientHello that is split to several small fragments? Do\n you fragment handshake messages that exceed the maximum fragment\n size? In particular, the Certificate and CertificateRequest\n handshake messages can be large enough to require fragmentation.\n\n - Do you ignore the TLS record layer version number in all\n unencrypted TLS records? (see Appendix D)\n\n - Have you ensured that all support for SSL, RC4, EXPORT ciphers,\n and MD5 (via the \"signature_algorithms\" extension) is completely\n removed from all possible configurations that support TLS 1.3 or\n later, and that attempts to use these obsolete capabilities fail\n correctly? (see Appendix D)\n\n - Do you handle TLS extensions in ClientHello correctly, including\n unknown extensions?\n\n - When the server has requested a client certificate, but no\n suitable certificate is available, do you correctly send an empty\n Certificate message, instead of omitting the whole message (see\n Section 4.4.2.3)?\n\n\n\nRescorla Expires September 21, 2018 [Page 133]\n\f\nInternet-Draft TLS March 2018\n\n\n - When processing the plaintext fragment produced by AEAD-Decrypt\n and scanning from the end for the ContentType, do you avoid\n scanning past the start of the cleartext in the event that the\n peer has sent a malformed plaintext of all-zeros?\n\n - Do you properly ignore unrecognized cipher suites (Section 4.1.2),\n hello extensions (Section 4.2), named groups (Section 4.2.7), key\n shares (Section 4.2.8), supported versions (Section 4.2.1), and\n signature algorithms (Section 4.2.3) in the ClientHello?\n\n - As a server, do you send a HelloRetryRequest to clients which\n support a compatible (EC)DHE group but do not predict it in the\n \"key_share\" extension? As a client, do you correctly handle a\n HelloRetryRequest from the server?\n\n Cryptographic details:\n\n - What countermeasures do you use to prevent timing attacks\n [TIMING]?\n\n - When using Diffie-Hellman key exchange, do you correctly preserve\n leading zero bytes in the negotiated key (see Section 7.4.1)?\n\n - Does your TLS client check that the Diffie-Hellman parameters sent\n by the server are acceptable, (see Section 4.2.8.1)?\n\n - Do you use a strong and, most importantly, properly seeded random\n number generator (see Appendix C.1) when generating Diffie-Hellman\n private values, the ECDSA \"k\" parameter, and other security-\n critical values? It is RECOMMENDED that implementations implement\n \"deterministic ECDSA\" as specified in [RFC6979].\n\n - Do you zero-pad Diffie-Hellman public key values to the group size\n (see Section 4.2.8.1)?\n\n - Do you verify signatures after making them to protect against RSA-\n CRT key leaks? [FW15]\n\nC.4. Client Tracking Prevention\n\n Clients SHOULD NOT reuse a ticket for multiple connections. Reuse of\n a ticket allows passive observers to correlate different connections.\n Servers that issue tickets SHOULD offer at least as many tickets as\n the number of connections that a client might use; for example, a web\n browser using HTTP/1.1 [RFC7230] might open six connections to a\n server. Servers SHOULD issue new tickets with every connection.\n This ensures that clients are always able to use a new ticket when\n creating a new connection.\n\n\n\nRescorla Expires September 21, 2018 [Page 134]\n\f\nInternet-Draft TLS March 2018\n\n\nC.5. Unauthenticated Operation\n\n Previous versions of TLS offered explicitly unauthenticated cipher\n suites based on anonymous Diffie-Hellman. These modes have been\n deprecated in TLS 1.3. However, it is still possible to negotiate\n parameters that do not provide verifiable server authentication by\n several methods, including:\n\n - Raw public keys [RFC7250].\n\n - Using a public key contained in a certificate but without\n validation of the certificate chain or any of its contents.\n\n Either technique used alone is vulnerable to man-in-the-middle\n attacks and therefore unsafe for general use. However, it is also\n possible to bind such connections to an external authentication\n mechanism via out-of-band validation of the server's public key,\n trust on first use, or a mechanism such as channel bindings (though\n the channel bindings described in [RFC5929] are not defined for TLS\n 1.3). If no such mechanism is used, then the connection has no\n protection against active man-in-the-middle attack; applications MUST\n NOT use TLS in such a way absent explicit configuration or a specific\n application profile.\n\nAppendix D. Backward Compatibility\n\n The TLS protocol provides a built-in mechanism for version\n negotiation between endpoints potentially supporting different\n versions of TLS.\n\n TLS 1.x and SSL 3.0 use compatible ClientHello messages. Servers can\n also handle clients trying to use future versions of TLS as long as\n the ClientHello format remains compatible and there is at least one\n protocol version supported by both the client and the server.\n\n Prior versions of TLS used the record layer version number\n (TLSPlaintext.legacy_record_version and\n TLSCiphertext.legacy_record_version) for various purposes. As of TLS\n 1.3, this field is deprecated. The value of\n TLSPlaintext.legacy_record_version MUST be ignored by all\n implementations. The value of TLSCiphertext.legacy_record_version is\n included in the additional data for deprotection but MAY otherwise be\n ignored or MAY be validated to match the fixed constant value.\n Version negotiation is performed using only the handshake versions\n (ClientHello.legacy_version, ServerHello.legacy_version, as well as\n the ClientHello, HelloRetryRequest and ServerHello\n \"supported_versions\" extensions). In order to maximize\n interoperability with older endpoints, implementations that negotiate\n\n\n\nRescorla Expires September 21, 2018 [Page 135]\n\f\nInternet-Draft TLS March 2018\n\n\n the use of TLS 1.0-1.2 SHOULD set the record layer version number to\n the negotiated version for the ServerHello and all records\n thereafter.\n\n For maximum compatibility with previously non-standard behavior and\n misconfigured deployments, all implementations SHOULD support\n validation of certification paths based on the expectations in this\n document, even when handling prior TLS versions' handshakes. (see\n Section 4.4.2.2)\n\n TLS 1.2 and prior supported an \"Extended Master Secret\" [RFC7627]\n extension which digested large parts of the handshake transcript into\n the master secret. Because TLS 1.3 always hashes in the transcript\n up to the server CertificateVerify, implementations which support\n both TLS 1.3 and earlier versions SHOULD indicate the use of the\n Extended Master Secret extension in their APIs whenever TLS 1.3 is\n used.\n\nD.1. Negotiating with an older server\n\n A TLS 1.3 client who wishes to negotiate with servers that do not\n support TLS 1.3 will send a normal TLS 1.3 ClientHello containing\n 0x0303 (TLS 1.2) in ClientHello.legacy_version but with the correct\n version(s) in the \"supported_versions\" extension. If the server does\n not support TLS 1.3 it will respond with a ServerHello containing an\n older version number. If the client agrees to use this version, the\n negotiation will proceed as appropriate for the negotiated protocol.\n A client using a ticket for resumption SHOULD initiate the connection\n using the version that was previously negotiated.\n\n Note that 0-RTT data is not compatible with older servers and SHOULD\n NOT be sent absent knowledge that the server supports TLS 1.3. See\n Appendix D.3.\n\n If the version chosen by the server is not supported by the client\n (or not acceptable), the client MUST abort the handshake with a\n \"protocol_version\" alert.\n\n Some legacy server implementations are known to not implement the TLS\n specification properly and might abort connections upon encountering\n TLS extensions or versions which they are not aware of.\n Interoperability with buggy servers is a complex topic beyond the\n scope of this document. Multiple connection attempts may be required\n in order to negotiate a backwards compatible connection; however,\n this practice is vulnerable to downgrade attacks and is NOT\n RECOMMENDED.\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 136]\n\f\nInternet-Draft TLS March 2018\n\n\nD.2. Negotiating with an older client\n\n A TLS server can also receive a ClientHello indicating a version\n number smaller than its highest supported version. If the\n \"supported_versions\" extension is present, the server MUST negotiate\n using that extension as described in Section 4.2.1. If the\n \"supported_versions\" extension is not present, the server MUST\n negotiate the minimum of ClientHello.legacy_version and TLS 1.2. For\n example, if the server supports TLS 1.0, 1.1, and 1.2, and\n legacy_version is TLS 1.0, the server will proceed with a TLS 1.0\n ServerHello. If the \"supported_versions\" extension is absent and the\n server only supports versions greater than\n ClientHello.legacy_version, the server MUST abort the handshake with\n a \"protocol_version\" alert.\n\n Note that earlier versions of TLS did not clearly specify the record\n layer version number value in all cases\n (TLSPlaintext.legacy_record_version). Servers will receive various\n TLS 1.x versions in this field, but its value MUST always be ignored.\n\nD.3. 0-RTT backwards compatibility\n\n 0-RTT data is not compatible with older servers. An older server\n will respond to the ClientHello with an older ServerHello, but it\n will not correctly skip the 0-RTT data and will fail to complete the\n handshake. This can cause issues when a client attempts to use\n 0-RTT, particularly against multi-server deployments. For example, a\n deployment could deploy TLS 1.3 gradually with some servers\n implementing TLS 1.3 and some implementing TLS 1.2, or a TLS 1.3\n deployment could be downgraded to TLS 1.2.\n\n A client that attempts to send 0-RTT data MUST fail a connection if\n it receives a ServerHello with TLS 1.2 or older. A client that\n attempts to repair this error SHOULD NOT send a TLS 1.2 ClientHello,\n but instead send a TLS 1.3 ClientHello without 0-RTT data.\n\n To avoid this error condition, multi-server deployments SHOULD ensure\n a uniform and stable deployment of TLS 1.3 without 0-RTT prior to\n enabling 0-RTT.\n\nD.4. Middlebox Compatibility Mode\n\n Field measurements [Ben17a], [Ben17b], [Res17a], [Res17b] have found\n that a significant number of middleboxes misbehave when a TLS client/\n server pair negotiates TLS 1.3. Implementations can increase the\n chance of making connections through those middleboxes by making the\n TLS 1.3 handshake look more like a TLS 1.2 handshake:\n\n\n\n\nRescorla Expires September 21, 2018 [Page 137]\n\f\nInternet-Draft TLS March 2018\n\n\n - The client always provides a non-empty session ID in the\n ClientHello, as described in the legacy_session_id section of\n Section 4.1.2.\n\n - If not offering early data, the client sends a dummy\n change_cipher_spec record (see the third paragraph of Section 5.1)\n immediately before its second flight. This may either be before\n its second ClientHello or before its encrypted handshake flight.\n If offering early data, the record is placed immediately after the\n first ClientHello.\n\n - The server sends a dummy change_cipher_spec record immediately\n after its first handshake message. This may either be after a\n ServerHello or a HelloRetryRequest.\n\n When put together, these changes make the TLS 1.3 handshake resemble\n TLS 1.2 session resumption, which improves the chance of successfully\n connecting through middleboxes. This \"compatibility mode\" is\n partially negotiated: The client can opt to provide a session ID or\n not and the server has to echo it. Either side can send\n change_cipher_spec at any time during the handshake, as they must be\n ignored by the peer, but if the client sends a non-empty session ID,\n the server MUST send the change_cipher_spec as described in this\n section.\n\nD.5. Backwards Compatibility Security Restrictions\n\n Implementations negotiating use of older versions of TLS SHOULD\n prefer forward secret and AEAD cipher suites, when available.\n\n The security of RC4 cipher suites is considered insufficient for the\n reasons cited in [RFC7465]. Implementations MUST NOT offer or\n negotiate RC4 cipher suites for any version of TLS for any reason.\n\n Old versions of TLS permitted the use of very low strength ciphers.\n Ciphers with a strength less than 112 bits MUST NOT be offered or\n negotiated for any version of TLS for any reason.\n\n The security of SSL 3.0 [SSL3] is considered insufficient for the\n reasons enumerated in [RFC7568], and it MUST NOT be negotiated for\n any reason.\n\n The security of SSL 2.0 [SSL2] is considered insufficient for the\n reasons enumerated in [RFC6176], and it MUST NOT be negotiated for\n any reason.\n\n Implementations MUST NOT send an SSL version 2.0 compatible CLIENT-\n HELLO. Implementations MUST NOT negotiate TLS 1.3 or later using an\n\n\n\nRescorla Expires September 21, 2018 [Page 138]\n\f\nInternet-Draft TLS March 2018\n\n\n SSL version 2.0 compatible CLIENT-HELLO. Implementations are NOT\n RECOMMENDED to accept an SSL version 2.0 compatible CLIENT-HELLO in\n order to negotiate older versions of TLS.\n\n Implementations MUST NOT send a ClientHello.legacy_version or\n ServerHello.legacy_version set to 0x0300 or less. Any endpoint\n receiving a Hello message with ClientHello.legacy_version or\n ServerHello.legacy_version set to 0x0300 MUST abort the handshake\n with a \"protocol_version\" alert.\n\n Implementations MUST NOT send any records with a version less than\n 0x0300. Implementations SHOULD NOT accept any records with a version\n less than 0x0300 (but may inadvertently do so if the record version\n number is ignored completely).\n\n Implementations MUST NOT use the Truncated HMAC extension, defined in\n Section 7 of [RFC6066], as it is not applicable to AEAD algorithms\n and has been shown to be insecure in some scenarios.\n\nAppendix E. Overview of Security Properties\n\n A complete security analysis of TLS is outside the scope of this\n document. In this section, we provide an informal description the\n desired properties as well as references to more detailed work in the\n research literature which provides more formal definitions.\n\n We cover properties of the handshake separately from those of the\n record layer.\n\nE.1. Handshake\n\n The TLS handshake is an Authenticated Key Exchange (AKE) protocol\n which is intended to provide both one-way authenticated (server-only)\n and mutually authenticated (client and server) functionality. At the\n completion of the handshake, each side outputs its view of the\n following values:\n\n - A set of \"session keys\" (the various secrets derived from the\n master secret) from which can be derived a set of working keys.\n\n - A set of cryptographic parameters (algorithms, etc.)\n\n - The identities of the communicating parties.\n\n We assume the attacker to be an active network attacker, which means\n it has complete control over the network used to communicate between\n the parties [RFC3552]. Even under these conditions, the handshake\n should provide the properties listed below. Note that these\n\n\n\nRescorla Expires September 21, 2018 [Page 139]\n\f\nInternet-Draft TLS March 2018\n\n\n properties are not necessarily independent, but reflect the protocol\n consumers' needs.\n\n Establishing the same session keys. The handshake needs to output\n the same set of session keys on both sides of the handshake,\n provided that it completes successfully on each endpoint (See\n [CK01]; defn 1, part 1).\n\n Secrecy of the session keys. The shared session keys should be known\n only to the communicating parties and not to the attacker (See\n [CK01]; defn 1, part 2). Note that in a unilaterally\n authenticated connection, the attacker can establish its own\n session keys with the server, but those session keys are distinct\n from those established by the client.\n\n Peer Authentication. The client's view of the peer identity should\n reflect the server's identity. If the client is authenticated,\n the server's view of the peer identity should match the client's\n identity.\n\n Uniqueness of the session keys: Any two distinct handshakes should\n produce distinct, unrelated session keys. Individual session keys\n produced by a handshake should also be distinct and independent.\n\n Downgrade protection. The cryptographic parameters should be the\n same on both sides and should be the same as if the peers had been\n communicating in the absence of an attack (See [BBFKZG16]; defns 8\n and 9}).\n\n Forward secret with respect to long-term keys If the long-term\n keying material (in this case the signature keys in certificate-\n based authentication modes or the external/resumption PSK in PSK\n with (EC)DHE modes) is compromised after the handshake is\n complete, this does not compromise the security of the session key\n (See [DOW92]), as long as the session key itself has been erased.\n The forward secrecy property is not satisfied when PSK is used in\n the \"psk_ke\" PskKeyExchangeMode.\n\n Key Compromise Impersonation (KCI) resistance In a mutually-\n authenticated connection with certificates, compromising the long-\n term secret of one actor should not break that actor's\n authentication of their peer in the given connection (see\n [HGFS15]). For example, if a client's signature key is\n compromised, it should not be possible to impersonate arbitrary\n servers to that client in subsequent handshakes.\n\n Protection of endpoint identities. The server's identity\n (certificate) should be protected against passive attackers. The\n\n\n\nRescorla Expires September 21, 2018 [Page 140]\n\f\nInternet-Draft TLS March 2018\n\n\n client's identity should be protected against both passive and\n active attackers.\n\n Informally, the signature-based modes of TLS 1.3 provide for the\n establishment of a unique, secret, shared key established by an\n (EC)DHE key exchange and authenticated by the server's signature over\n the handshake transcript, as well as tied to the server's identity by\n a MAC. If the client is authenticated by a certificate, it also\n signs over the handshake transcript and provides a MAC tied to both\n identities. [SIGMA] describes the design and analysis of this type\n of key exchange protocol. If fresh (EC)DHE keys are used for each\n connection, then the output keys are forward secret.\n\n The external PSK and resumption PSK bootstrap from a long-term shared\n secret into a unique per-connection set of short-term session keys.\n This secret may have been established in a previous handshake. If\n PSK with (EC)DHE key establishment is used, these session keys will\n also be forward secret. The resumption PSK has been designed so that\n the resumption master secret computed by connection N and needed to\n form connection N+1 is separate from the traffic keys used by\n connection N, thus providing forward secrecy between the connections.\n In addition, if multiple tickets are established on the same\n connection, they are associated with different keys, so compromise of\n the PSK associated with one ticket does not lead to the compromise of\n connections established with PSKs associated with other tickets.\n This property is most interesting if tickets are stored in a database\n (and so can be deleted) rather than if they are self-encrypted.\n\n The PSK binder value forms a binding between a PSK and the current\n handshake, as well as between the session where the PSK was\n established and the current session. This binding transitively\n includes the original handshake transcript, because that transcript\n is digested into the values which produce the Resumption Master\n Secret. This requires that both the KDF used to produce the\n resumption master secret and the MAC used to compute the binder be\n collision resistant. See Appendix E.1.1 for more on this. Note: The\n binder does not cover the binder values from other PSKs, though they\n are included in the Finished MAC.\n\n Note: TLS does not currently permit the server to send a\n certificate_request message in non-certificate-based handshakes\n (e.g., PSK). If this restriction were to be relaxed in future, the\n client's signature would not cover the server's certificate directly.\n However, if the PSK was established through a NewSessionTicket, the\n client's signature would transitively cover the server's certificate\n through the PSK binder. [PSK-FINISHED] describes a concrete attack\n on constructions that do not bind to the server's certificate (see\n also [Kraw16]). It is unsafe to use certificate-based client\n\n\n\nRescorla Expires September 21, 2018 [Page 141]\n\f\nInternet-Draft TLS March 2018\n\n\n authentication when the client might potentially share the same PSK/\n key-id pair with two different endpoints. Implementations MUST NOT\n combine external PSKs with certificate-based authentication of either\n the client or the server unless negotiated by some extension.\n\n If an exporter is used, then it produces values which are unique and\n secret (because they are generated from a unique session key).\n Exporters computed with different labels and contexts are\n computationally independent, so it is not feasible to compute one\n from another or the session secret from the exported value. Note:\n exporters can produce arbitrary-length values. If exporters are to\n be used as channel bindings, the exported value MUST be large enough\n to provide collision resistance. The exporters provided in TLS 1.3\n are derived from the same handshake contexts as the early traffic\n keys and the application traffic keys respectively, and thus have\n similar security properties. Note that they do not include the\n client's certificate; future applications which wish to bind to the\n client's certificate may need to define a new exporter that includes\n the full handshake transcript.\n\n For all handshake modes, the Finished MAC (and where present, the\n signature), prevents downgrade attacks. In addition, the use of\n certain bytes in the random nonces as described in Section 4.1.3\n allows the detection of downgrade to previous TLS versions. See\n [BBFKZG16] for more detail on TLS 1.3 and downgrade.\n\n As soon as the client and the server have exchanged enough\n information to establish shared keys, the remainder of the handshake\n is encrypted, thus providing protection against passive attackers,\n even if the computed shared key is not authenticated. Because the\n server authenticates before the client, the client can ensure that if\n it authenticates to the server, it only reveals its identity to an\n authenticated server. Note that implementations must use the\n provided record padding mechanism during the handshake to avoid\n leaking information about the identities due to length. The client's\n proposed PSK identities are not encrypted, nor is the one that the\n server selects.\n\nE.1.1. Key Derivation and HKDF\n\n Key derivation in TLS 1.3 uses the HKDF function defined in [RFC5869]\n and its two components, HKDF-Extract and HKDF-Expand. The full\n rationale for the HKDF construction can be found in [Kraw10] and the\n rationale for the way it is used in TLS 1.3 in [KW16]. Throughout\n this document, each application of HKDF-Extract is followed by one or\n more invocations of HKDF-Expand. This ordering should always be\n followed (including in future revisions of this document), in\n particular, one SHOULD NOT use an output of HKDF-Extract as an input\n\n\n\nRescorla Expires September 21, 2018 [Page 142]\n\f\nInternet-Draft TLS March 2018\n\n\n to another application of HKDF-Extract without an HKDF-Expand in\n between. Consecutive applications of HKDF-Expand are allowed as long\n as these are differentiated via the key and/or the labels.\n\n Note that HKDF-Expand implements a pseudorandom function (PRF) with\n both inputs and outputs of variable length. In some of the uses of\n HKDF in this document (e.g., for generating exporters and the\n resumption_master_secret), it is necessary that the application of\n HKDF-Expand be collision-resistant, namely, it should be infeasible\n to find two different inputs to HKDF-Expand that output the same\n value. This requires the underlying hash function to be collision\n resistant and the output length from HKDF-Expand to be of size at\n least 256 bits (or as much as needed for the hash function to prevent\n finding collisions).\n\nE.1.2. Client Authentication\n\n A client that has sent authentication data to a server, either during\n the handshake or in post-handshake authentication, cannot be sure if\n the server afterwards considers the client to be authenticated or\n not. If the client needs to determine if the server considers the\n connection to be unilaterally or mutually authenticated, this has to\n be provisioned by the application layer. See [CHHSV17] for details.\n In addition, the analysis of post-handshake authentication from\n [Kraw16] shows that the client identified by the certificate sent in\n the post-handshake phase possesses the traffic key. This party is\n therefore the client that participated in the original handshake or\n one to whom the original client delegated the traffic key (assuming\n that the traffic key has not been compromised).\n\nE.1.3. 0-RTT\n\n The 0-RTT mode of operation generally provides similar security\n properties as 1-RTT data, with the two exceptions that the 0-RTT\n encryption keys do not provide full forward secrecy and that the\n server is not able to guarantee uniqueness of the handshake (non-\n replayability) without keeping potentially undue amounts of state.\n See Section 8 for mechanisms to limit the exposure to replay.\n\nE.1.4. Exporter Independence\n\n The exporter_master_secret and early_exporter_master_secret are\n derived to be independent of the traffic keys and therefore do not\n represent a threat to the security of traffic encrypted with those\n keys. However, because these secrets can be used to compute any\n exporter value, they SHOULD be erased as soon as possible. If the\n total set of exporter labels is known, then implementations SHOULD\n pre-compute the inner Derive-Secret stage of the exporter computation\n\n\n\nRescorla Expires September 21, 2018 [Page 143]\n\f\nInternet-Draft TLS March 2018\n\n\n for all those labels, then erase the [early_]exporter_master_secret,\n followed by each inner values as soon as it is known that it will not\n be needed again.\n\nE.1.5. Post-Compromise Security\n\n TLS does not provide security for handshakes which take place after\n the peer's long-term secret (signature key or external PSK) is\n compromised. It therefore does not provide post-compromise security\n [CCG16], sometimes also referred to as backwards or future secrecy.\n This is in contrast to KCI resistance, which describes the security\n guarantees that a party has after its own long-term secret has been\n compromised.\n\nE.1.6. External References\n\n The reader should refer to the following references for analysis of\n the TLS handshake: [DFGS15] [CHSV16] [DFGS16] [KW16] [Kraw16]\n [FGSW16] [LXZFH16] [FG17] [BBK17].\n\nE.2. Record Layer\n\n The record layer depends on the handshake producing strong traffic\n secrets which can be used to derive bidirectional encryption keys and\n nonces. Assuming that is true, and the keys are used for no more\n data than indicated in Section 5.5 then the record layer should\n provide the following guarantees:\n\n Confidentiality. An attacker should not be able to determine the\n plaintext contents of a given record.\n\n Integrity. An attacker should not be able to craft a new record\n which is different from an existing record which will be accepted\n by the receiver.\n\n Order protection/non-replayability An attacker should not be able to\n cause the receiver to accept a record which it has already\n accepted or cause the receiver to accept record N+1 without having\n first processed record N.\n\n Length concealment. Given a record with a given external length, the\n attacker should not be able to determine the amount of the record\n that is content versus padding.\n\n Forward secrecy after key change. If the traffic key update\n mechanism described in Section 4.6.3 has been used and the\n previous generation key is deleted, an attacker who compromises\n\n\n\n\nRescorla Expires September 21, 2018 [Page 144]\n\f\nInternet-Draft TLS March 2018\n\n\n the endpoint should not be able to decrypt traffic encrypted with\n the old key.\n\n Informally, TLS 1.3 provides these properties by AEAD-protecting the\n plaintext with a strong key. AEAD encryption [RFC5116] provides\n confidentiality and integrity for the data. Non-replayability is\n provided by using a separate nonce for each record, with the nonce\n being derived from the record sequence number (Section 5.3), with the\n sequence number being maintained independently at both sides thus\n records which are delivered out of order result in AEAD deprotection\n failures. In order to prevent mass cryptanalysis when the same\n plaintext is repeatedly encrypted by different users under the same\n key (as is commonly the case for HTTP), the nonce is formed by mixing\n the sequence number with a secret per-connection initialization\n vector derived along with the traffic keys. See [BT16] for analysis\n of this construction.\n\n The re-keying technique in TLS 1.3 (see Section 7.2) follows the\n construction of the serial generator in [REKEY], which shows that re-\n keying can allow keys to be used for a larger number of encryptions\n than without re-keying. This relies on the security of the HKDF-\n Expand-Label function as a pseudorandom function (PRF). In addition,\n as long as this function is truly one way, it is not possible to\n compute traffic keys from prior to a key change (forward secrecy).\n\n TLS does not provide security for data which is communicated on a\n connection after a traffic secret of that connection is compromised.\n That is, TLS does not provide post-compromise security/future\n secrecy/backward secrecy with respect to the traffic secret. Indeed,\n an attacker who learns a traffic secret can compute all future\n traffic secrets on that connection. Systems which want such\n guarantees need to do a fresh handshake and establish a new\n connection with an (EC)DHE exchange.\n\nE.2.1. External References\n\n The reader should refer to the following references for analysis of\n the TLS record layer: [BMMT15] [BT16] [BDFKPPRSZZ16] [BBK17]\n [Anon18].\n\nE.3. Traffic Analysis\n\n TLS is susceptible to a variety of traffic analysis attacks based on\n observing the length and timing of encrypted packets [CLINIC]\n [HCJ16]. This is particularly easy when there is a small set of\n possible messages to be distinguished, such as for a video server\n hosting a fixed corpus of content, but still provides usable\n information even in more complicated scenarios.\n\n\n\nRescorla Expires September 21, 2018 [Page 145]\n\f\nInternet-Draft TLS March 2018\n\n\n TLS does not provide any specific defenses against this form of\n attack but does include a padding mechanism for use by applications:\n The plaintext protected by the AEAD function consists of content plus\n variable-length padding, which allows the application to produce\n arbitrary length encrypted records as well as padding-only cover\n traffic to conceal the difference between periods of transmission and\n periods of silence. Because the padding is encrypted alongside the\n actual content, an attacker cannot directly determine the length of\n the padding, but may be able to measure it indirectly by the use of\n timing channels exposed during record processing (i.e., seeing how\n long it takes to process a record or trickling in records to see\n which ones elicit a response from the server). In general, it is not\n known how to remove all of these channels because even a constant\n time padding removal function will likely feed the content into data-\n dependent functions. At minimum, a fully constant time server or\n client would require close cooperation with the application layer\n protocol implementation, including making that higher level protocol\n constant time.\n\n Note: Robust traffic analysis defences will likely lead to inferior\n performance due to delay in transmitting packets and increased\n traffic volume.\n\nE.4. Side Channel Attacks\n\n In general, TLS does not have specific defenses against side-channel\n attacks (i.e., those which attack the communications via secondary\n channels such as timing) leaving those to the implementation of the\n relevant cryptographic primitives. However, certain features of TLS\n are designed to make it easier to write side-channel resistant code:\n\n - Unlike previous versions of TLS which used a composite MAC-then-\n encrypt structure, TLS 1.3 only uses AEAD algorithms, allowing\n implementations to use self-contained constant-time\n implementations of those primitives.\n\n - TLS uses a uniform \"bad_record_mac\" alert for all decryption\n errors, which is intended to prevent an attacker from gaining\n piecewise insight into portions of the message. Additional\n resistance is provided by terminating the connection on such\n errors; a new connection will have different cryptographic\n material, preventing attacks against the cryptographic primitives\n that require multiple trials.\n\n Information leakage through side channels can occur at layers above\n TLS, in application protocols and the applications that use them.\n Resistance to side-channel attacks depends on applications and\n\n\n\n\nRescorla Expires September 21, 2018 [Page 146]\n\f\nInternet-Draft TLS March 2018\n\n\n application protocols separately ensuring that confidential\n information is not inadvertently leaked.\n\nE.5. Replay Attacks on 0-RTT\n\n Replayable 0-RTT data presents a number of security threats to TLS-\n using applications, unless those applications are specifically\n engineered to be safe under replay (minimally, this means idempotent,\n but in many cases may also require other stronger conditions, such as\n constant-time response). Potential attacks include:\n\n - Duplication of actions which cause side effects (e.g., purchasing\n an item or transferring money) to be duplicated, thus harming the\n site or the user.\n\n - Attackers can store and replay 0-RTT messages in order to re-order\n them with respect to other messages (e.g., moving a delete to\n after a create).\n\n - Exploiting cache timing behavior to discover the content of 0-RTT\n messages by replaying a 0-RTT message to a different cache node\n and then using a separate connection to measure request latency,\n to see if the two requests address the same resource.\n\n If data can be replayed a large number of times, additional attacks\n become possible, such as making repeated measurements of the speed of\n cryptographic operations. In addition, they may be able to overload\n rate-limiting systems. For further description of these attacks, see\n [Mac17].\n\n Ultimately, servers have the responsibility to protect themselves\n against attacks employing 0-RTT data replication. The mechanisms\n described in Section 8 are intended to prevent replay at the TLS\n layer but do not provide complete protection against receiving\n multiple copies of client data. TLS 1.3 falls back to the 1-RTT\n handshake when the server does not have any information about the\n client, e.g., because it is in a different cluster which does not\n share state or because the ticket has been deleted as described in\n Section 8.1. If the application layer protocol retransmits data in\n this setting, then it is possible for an attacker to induce message\n duplication by sending the ClientHello to both the original cluster\n (which processes the data immediately) and another cluster which will\n fall back to 1-RTT and process the data upon application layer\n replay. The scale of this attack is limited by the client's\n willingness to retry transactions and therefore only allows a limited\n amount of duplication, with each copy appearing as a new connection\n at the server.\n\n\n\n\nRescorla Expires September 21, 2018 [Page 147]\n\f\nInternet-Draft TLS March 2018\n\n\n If implemented correctly, the mechanisms described in Section 8.1 and\n Section 8.2 prevent a replayed ClientHello and its associated 0-RTT\n data from being accepted multiple times by any cluster with\n consistent state; for servers which limit the use of 0-RTT to one\n cluster for a single ticket, then a given ClientHello and its\n associated 0-RTT data will only be accepted once. However, if state\n is not completely consistent, then an attacker might be able to have\n multiple copies of the data be accepted during the replication\n window. Because clients do not know the exact details of server\n behavior, they MUST NOT send messages in early data which are not\n safe to have replayed and which they would not be willing to retry\n across multiple 1-RTT connections.\n\n Application protocols MUST NOT use 0-RTT data without a profile that\n defines its use. That profile needs to identify which messages or\n interactions are safe to use with 0-RTT and how to handle the\n situation when the server rejects 0-RTT and falls back to 1-RTT.\n\n In addition, to avoid accidental misuse, TLS implementations MUST NOT\n enable 0-RTT (either sending or accepting) unless specifically\n requested by the application and MUST NOT automatically resend 0-RTT\n data if it is rejected by the server unless instructed by the\n application. Server-side applications may wish to implement special\n processing for 0-RTT data for some kinds of application traffic\n (e.g., abort the connection, request that data be resent at the\n application layer, or delay processing until the handshake\n completes). In order to allow applications to implement this kind of\n processing, TLS implementations MUST provide a way for the\n application to determine if the handshake has completed.\n\nE.5.1. Replay and Exporters\n\n Replays of the ClientHello produce the same early exporter, thus\n requiring additional care by applications which use these exporters.\n In particular, if these exporters are used as an authentication\n channel binding (e.g., by signing the output of the exporter) an\n attacker who compromises the PSK can transplant authenticators\n between connections without compromising the authentication key.\n\n In addition, the early exporter SHOULD NOT be used to generate\n server-to-client encryption keys because that would entail the reuse\n of those keys. This parallels the use of the early application\n traffic keys only in the client-to-server direction.\n\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 148]\n\f\nInternet-Draft TLS March 2018\n\n\nE.6. PSK Identity Exposure\n\n Because implementations respond to an invalid PSK binder by aborting\n the handshake, it may be possible for an attacker to verify whether a\n given PSK identity is valid. Specifically, if a server accepts both\n external PSK and certificate-based handshakes, a valid PSK identity\n will result in a failed handshake, whereas an invalid identity will\n just be skipped and result in a successful certificate handshake.\n Servers which solely support PSK handshakes may be able to resist\n this form of attack by treating the cases where there is no valid PSK\n identity and where there is an identity but it has an invalid binder\n identically.\n\nE.7. Attacks on Static RSA\n\n Although TLS 1.3 does not use RSA key transport and so is not\n directly susceptible to Bleichenbacher-type attacks, if TLS 1.3\n servers also support static RSA in the context of previous versions\n of TLS, then it may be possible to impersonate the server for TLS 1.3\n connections [JSS15]. TLS 1.3 implementations can prevent this attack\n by disabling support for static RSA across all versions of TLS. In\n principle, implementations might also be able to separate\n certificates with different keyUsage bits for static RSA decryption\n and RSA signature, but this technique relies on clients refusing to\n accept signatures using keys in certificates that do not have the\n digitalSignature bit set, and many clients do not enforce this\n restriction.\n\nAppendix F. Working Group Information\n\n The discussion list for the IETF TLS working group is located at the\n e-mail address [redacted-email] [1]. Information on the group and\n information on how to subscribe to the list is at\n https://www.ietf.org/mailman/listinfo/tls\n\n Archives of the list can be found at: https://www.ietf.org/mail-\n archive/web/tls/current/index.html\n\nAppendix G. Contributors\n\n - Martin Abadi\n University of California, Santa Cruz\n [redacted-email]\n\n - Christopher Allen (co-editor of TLS 1.0)\n Alacrity Ventures\n [redacted-email]\n\n\n\n\nRescorla Expires September 21, 2018 [Page 149]\n\f\nInternet-Draft TLS March 2018\n\n\n - Richard Barnes\n Cisco\n [redacted-email]\n\n - Steven M. Bellovin\n Columbia University\n [redacted-email]\n\n - David Benjamin\n Google\n [redacted-email]\n\n - Benjamin Beurdouche\n INRIA & Microsoft Research\n [redacted-email]\n\n - Karthikeyan Bhargavan (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n - Simon Blake-Wilson (co-author of [RFC4492])\n BCI\n [redacted-email]\n\n - Nelson Bolyard (co-author of [RFC4492])\n Sun Microsystems, Inc.\n [redacted-email]\n\n - Ran Canetti\n IBM\n [redacted-email]\n\n - Matt Caswell\n OpenSSL\n [redacted-email]\n\n - Stephen Checkoway\n University of Illinois at Chicago\n [redacted-email]\n\n - Pete Chown\n Skygate Technology Ltd\n [redacted-email]\n\n - Katriel Cohn-Gordon\n University of Oxford\n [redacted-email]\n\n\n\n\nRescorla Expires September 21, 2018 [Page 150]\n\f\nInternet-Draft TLS March 2018\n\n\n - Cas Cremers\n University of Oxford\n [redacted-email]\n\n - Antoine Delignat-Lavaud (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n - Tim Dierks (co-editor of TLS 1.0, 1.1, and 1.2)\n Independent\n [redacted-email]\n\n - Roelof DuToit\n Symantec Corporation\n [redacted-email]\n\n - Taher Elgamal\n Securify\n [redacted-email]\n\n - Pasi Eronen\n Nokia\n [redacted-email]\n\n - Cedric Fournet\n Microsoft\n [redacted-email]\n\n - Anil Gangolli\n [redacted-email]\n\n - David M. Garrett\n [redacted-email]\n\n - Illya Gerasymchuk\n Independent\n [redacted-email]\n\n - Alessandro Ghedini\n Cloudflare Inc.\n [redacted-email]\n\n - Daniel Kahn Gillmor\n ACLU\n [redacted-email]\n\n - Matthew Green\n Johns Hopkins University\n\n\n\nRescorla Expires September 21, 2018 [Page 151]\n\f\nInternet-Draft TLS March 2018\n\n\n [redacted-email]\n\n - Jens Guballa\n ETAS\n [redacted-email]\n\n - Felix Guenther\n TU Darmstadt\n [redacted-email]\n\n - Vipul Gupta (co-author of [RFC4492])\n Sun Microsystems Laboratories\n [redacted-email]\n\n - Chris Hawk (co-author of [RFC4492])\n Corriente Networks LLC\n [redacted-email]\n\n - Kipp Hickman\n\n - Alfred Hoenes\n\n - David Hopwood\n Independent Consultant\n [redacted-email]\n\n - Marko Horvat\n MPI-SWS\n [redacted-email]\n\n - Jonathan Hoyland\n Royal Holloway, University of London [redacted-email]\n\n - Subodh Iyengar\n Facebook\n [redacted-email]\n\n - Benjamin Kaduk\n Akamai\n [redacted-email]\n\n - Hubert Kario\n Red Hat Inc.\n [redacted-email]\n\n - Phil Karlton (co-author of SSL 3.0)\n\n - Leon Klingele\n\n\n\nRescorla Expires September 21, 2018 [Page 152]\n\f\nInternet-Draft TLS March 2018\n\n\n Independent\n [redacted-email]\n\n - Paul Kocher (co-author of SSL 3.0)\n Cryptography Research\n [redacted-email]\n\n - Hugo Krawczyk\n IBM\n [redacted-email]\n\n - Adam Langley (co-author of [RFC7627])\n Google\n [redacted-email]\n\n - Olivier Levillain\n ANSSI\n [redacted-email]\n\n - Xiaoyin Liu\n University of North Carolina at Chapel Hill\n [redacted-email]\n\n - Ilari Liusvaara\n Independent\n [redacted-email]\n\n - Atul Luykx\n K.U. Leuven\n [redacted-email]\n\n - Colm MacCarthaigh\n Amazon Web Services\n [redacted-email]\n\n - Carl Mehner\n USAA\n [redacted-email]\n\n - Jan Mikkelsen\n Transactionware\n [redacted-email]\n\n - Bodo Moeller (co-author of [RFC4492])\n Google\n [redacted-email]\n\n - Kyle Nekritz\n\n\n\nRescorla Expires September 21, 2018 [Page 153]\n\f\nInternet-Draft TLS March 2018\n\n\n Facebook\n [redacted-email]\n\n - Erik Nygren\n Akamai Technologies\n [redacted-email]\n\n - Magnus Nystrom\n Microsoft\n [redacted-email]\n\n - Kazuho Oku\n DeNA Co., Ltd.\n [redacted-email]\n\n - Kenny Paterson\n Royal Holloway, University of London\n [redacted-email]\n\n - Alfredo Pironti (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n - Andrei Popov\n Microsoft\n [redacted-email]\n\n - Marsh Ray (co-author of [RFC7627])\n Microsoft\n [redacted-email]\n\n - Robert Relyea\n Netscape Communications\n [redacted-email]\n\n - Kyle Rose\n Akamai Technologies\n [redacted-email]\n\n - Jim Roskind\n Amazon\n [redacted-email]\n\n - Michael Sabin\n\n - Joe Salowey\n Tableau Software\n [redacted-email]\n\n\n\nRescorla Expires September 21, 2018 [Page 154]\n\f\nInternet-Draft TLS March 2018\n\n\n - Rich Salz\n Akamai\n [redacted-email]\n\n - David Schinazi\n Apple Inc.\n [redacted-email]\n\n - Sam Scott\n Royal Holloway, University of London\n [redacted-email]\n\n - Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n - Brian Smith\n Independent\n [redacted-email]\n\n - Brian Sniffen\n Akamai Technologies\n [redacted-email]\n\n - Nick Sullivan\n Cloudflare Inc.\n [redacted-email]\n\n - Bjoern Tackmann\n University of California, San Diego\n [redacted-email]\n\n - Tim Taubert\n Mozilla\n [redacted-email]\n\n - Martin Thomson\n Mozilla\n [redacted-email]\n\n - Sean Turner\n sn3rd\n [redacted-email]\n\n - Steven Valdez\n Google\n [redacted-email]\n\n\n\n\nRescorla Expires September 21, 2018 [Page 155]\n\f\nInternet-Draft TLS March 2018\n\n\n - Filippo Valsorda\n Cloudflare Inc.\n [redacted-email]\n\n - Thyla van der Merwe\n Royal Holloway, University of London\n [redacted-email]\n\n - Victor Vasiliev\n Google\n [redacted-email]\n\n - Tom Weinstein\n\n - Hoeteck Wee\n Ecole Normale Superieure, Paris\n [redacted-email]\n\n - David Wong\n NCC Group\n [redacted-email]\n\n - Christopher A. Wood\n Apple Inc.\n [redacted-email]\n\n - Tim Wright\n Vodafone\n [redacted-email]\n\n - Peter Wu\n Independent\n [redacted-email]\n\n - Kazu Yamamoto\n Internet Initiative Japan Inc.\n [redacted-email]\n\nAuthor's Address\n\n Eric Rescorla\n RTFM, Inc.\n\n EMail: [redacted-email]\n\n\n\n\n\n\n\nRescorla Expires September 21, 2018 [Page 156]", "text_sha256": "a7cda7524edf08714e498aaf4b2a684ee1b5fdf0c9648505eb54649b7cfb0b92"}, {"draft_references": ["draft-ietf-tls-tls13-vectors-06"], "excluded_rfc_relation_targets": [{"kind": "obsoletes", "reason": "not_requested", "rfc_number": 2018}], "facts": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_id": "rfc_number", "field": "rfc_number", "raw_char_end": 151, "raw_char_start": 79, "value": "8446"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_id": "obsoletes:5077", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "5077"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_id": "obsoletes:5246", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "5246"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_id": "obsoletes:6961", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "6961"}, {"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_id": "obsoletes:2018", "field": "obsoletes", "raw_char_end": 224, "raw_char_start": 152, "value": "2018"}, {"char_end": 238, "char_start": 219, "evidence_quote": "Updates: 5705, 6066", "fact_id": "updates:5705", "field": "updates", "raw_char_end": 244, "raw_char_start": 225, "value": "5705"}, {"char_end": 238, "char_start": 219, "evidence_quote": "Updates: 5705, 6066", "fact_id": "updates:6066", "field": "updates", "raw_char_end": 244, "raw_char_start": 225, "value": "6066"}, {"char_end": 261161, "char_start": 261130, "evidence_quote": "draft-ietf-tls-tls13-vectors-06", "fact_id": "draft_reference:draft-ietf-tls-tls13-vectors-06", "field": "draft_reference", "raw_char_end": 261167, "raw_char_start": 261136, "value": "draft-ietf-tls-tls13-vectors-06"}], "kind": "rfc", "occurred_at": "2018-08-01T00:00:00Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 86, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd", "record_id": "ietf:rfc:8446", "retrieval_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "rfc_number": 8446, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd", "source_url": "https://www.rfc-editor.org/rfc/rfc8446.txt", "temporal_semantics": "source_event_time", "text": "Internet Engineering Task Force (IETF) E. Rescorla\nRequest for Comments: 8446 Mozilla\nObsoletes: 5077, 5246, 6961 August 2018\nUpdates: 5705, 6066\nCategory: Standards Track\nISSN: 2070-1721\n\n\n The Transport Layer Security (TLS) Protocol Version 1.3\n\nAbstract\n\n This document specifies version 1.3 of the Transport Layer Security\n (TLS) protocol. TLS allows client/server applications to communicate\n over the Internet in a way that is designed to prevent eavesdropping,\n tampering, and message forgery.\n\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077,\n 5246, and 6961. This document also specifies new requirements for\n TLS 1.2 implementations.\n\nStatus of This Memo\n\n This is an Internet Standards Track document.\n\n This document is a product of the Internet Engineering Task Force\n (IETF). It represents the consensus of the IETF community. It has\n received public review and has been approved for publication by the\n Internet Engineering Steering Group (IESG). Further information on\n Internet Standards is available in Section 2 of RFC 7841.\n\n Information about the current status of this document, any errata,\n and how to provide feedback on it may be obtained at\n https://www.rfc-editor.org/info/rfc8446.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 1]\n\f\nRFC 8446 TLS August 2018\n\n\nCopyright Notice\n\n Copyright (c) 2018 IETF Trust and the persons identified as the\n document authors. All rights reserved.\n\n This document is subject to BCP 78 and the IETF Trust's Legal\n Provisions Relating to IETF Documents\n (https://trustee.ietf.org/license-info) in effect on the date of\n publication of this document. Please review these documents\n carefully, as they describe your rights and restrictions with respect\n to this document. Code Components extracted from this document must\n include Simplified BSD License text as described in Section 4.e of\n the Trust Legal Provisions and are provided without warranty as\n described in the Simplified BSD License.\n\n This document may contain material from IETF Documents or IETF\n Contributions published or made publicly available before November\n 10, 2008. The person(s) controlling the copyright in some of this\n material may not have granted the IETF Trust the right to allow\n modifications of such material outside the IETF Standards Process.\n Without obtaining an adequate license from the person(s) controlling\n the copyright in such materials, this document may not be modified\n outside the IETF Standards Process, and derivative works of it may\n not be created outside the IETF Standards Process, except to format\n it for publication as an RFC or to translate it into languages other\n than English.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 2]\n\f\nRFC 8446 TLS August 2018\n\n\nTable of Contents\n\n 1. Introduction ....................................................6\n 1.1. Conventions and Terminology ................................7\n 1.2. Major Differences from TLS 1.2 .............................8\n 1.3. Updates Affecting TLS 1.2 ..................................9\n 2. Protocol Overview ..............................................10\n 2.1. Incorrect DHE Share .......................................14\n 2.2. Resumption and Pre-Shared Key (PSK) .......................15\n 2.3. 0-RTT Data ................................................17\n 3. Presentation Language ..........................................19\n 3.1. Basic Block Size ..........................................19\n 3.2. Miscellaneous .............................................20\n 3.3. Numbers ...................................................20\n 3.4. Vectors ...................................................20\n 3.5. Enumerateds ...............................................21\n 3.6. Constructed Types .........................................22\n 3.7. Constants .................................................23\n 3.8. Variants ..................................................23\n 4. Handshake Protocol .............................................24\n 4.1. Key Exchange Messages .....................................25\n 4.1.1. Cryptographic Negotiation ..........................26\n 4.1.2. Client Hello .......................................27\n 4.1.3. Server Hello .......................................31\n 4.1.4. Hello Retry Request ................................33\n 4.2. Extensions ................................................35\n 4.2.1. Supported Versions .................................39\n 4.2.2. Cookie .............................................40\n 4.2.3. Signature Algorithms ...............................41\n 4.2.4. Certificate Authorities ............................45\n 4.2.5. OID Filters ........................................45\n 4.2.6. Post-Handshake Client Authentication ...............47\n 4.2.7. Supported Groups ...................................47\n 4.2.8. Key Share ..........................................48\n 4.2.9. Pre-Shared Key Exchange Modes ......................51\n 4.2.10. Early Data Indication .............................52\n 4.2.11. Pre-Shared Key Extension ..........................55\n 4.3. Server Parameters .........................................59\n 4.3.1. Encrypted Extensions ...............................60\n 4.3.2. Certificate Request ................................60\n 4.4. Authentication Messages ...................................61\n 4.4.1. The Transcript Hash ................................63\n 4.4.2. Certificate ........................................64\n 4.4.3. Certificate Verify .................................69\n 4.4.4. Finished ...........................................71\n 4.5. End of Early Data .........................................72\n\n\n\n\n\nRescorla Standards Track [Page 3]\n\f\nRFC 8446 TLS August 2018\n\n\n 4.6. Post-Handshake Messages ...................................73\n 4.6.1. New Session Ticket Message .........................73\n 4.6.2. Post-Handshake Authentication ......................75\n 4.6.3. Key and Initialization Vector Update ...............76\n 5. Record Protocol ................................................77\n 5.1. Record Layer ..............................................78\n 5.2. Record Payload Protection .................................80\n 5.3. Per-Record Nonce ..........................................82\n 5.4. Record Padding ............................................83\n 5.5. Limits on Key Usage .......................................84\n 6. Alert Protocol .................................................85\n 6.1. Closure Alerts ............................................87\n 6.2. Error Alerts ..............................................88\n 7. Cryptographic Computations .....................................90\n 7.1. Key Schedule ..............................................91\n 7.2. Updating Traffic Secrets ..................................94\n 7.3. Traffic Key Calculation ...................................95\n 7.4. (EC)DHE Shared Secret Calculation .........................95\n 7.4.1. Finite Field Diffie-Hellman ........................95\n 7.4.2. Elliptic Curve Diffie-Hellman ......................96\n 7.5. Exporters .................................................97\n 8. 0-RTT and Anti-Replay ..........................................98\n 8.1. Single-Use Tickets ........................................99\n 8.2. Client Hello Recording ....................................99\n 8.3. Freshness Checks .........................................101\n 9. Compliance Requirements .......................................102\n 9.1. Mandatory-to-Implement Cipher Suites .....................102\n 9.2. Mandatory-to-Implement Extensions ........................103\n 9.3. Protocol Invariants ......................................104\n 10. Security Considerations ......................................106\n 11. IANA Considerations ..........................................106\n 12. References ...................................................109\n 12.1. Normative References ....................................109\n 12.2. Informative References ..................................112\n Appendix A. State Machine ........................................120\n A.1. Client ....................................................120\n A.2. Server ....................................................121\n Appendix B. Protocol Data Structures and Constant Values .........122\n B.1. Record Layer ..............................................122\n B.2. Alert Messages ............................................123\n B.3. Handshake Protocol ........................................124\n B.3.1. Key Exchange Messages .................................125\n B.3.2. Server Parameters Messages ............................131\n B.3.3. Authentication Messages ...............................132\n B.3.4. Ticket Establishment ..................................132\n B.3.5. Updating Keys .........................................133\n B.4. Cipher Suites .............................................133\n\n\n\n\nRescorla Standards Track [Page 4]\n\f\nRFC 8446 TLS August 2018\n\n\n Appendix C. Implementation Notes .................................134\n C.1. Random Number Generation and Seeding ......................134\n C.2. Certificates and Authentication ...........................135\n C.3. Implementation Pitfalls ...................................135\n C.4. Client Tracking Prevention ................................137\n C.5. Unauthenticated Operation .................................137\n Appendix D. Backward Compatibility ...............................138\n D.1. Negotiating with an Older Server ..........................139\n D.2. Negotiating with an Older Client ..........................139\n D.3. 0-RTT Backward Compatibility ..............................140\n D.4. Middlebox Compatibility Mode ..............................140\n D.5. Security Restrictions Related to Backward Compatibility ...141\n Appendix E. Overview of Security Properties ......................142\n E.1. Handshake .................................................142\n E.1.1. Key Derivation and HKDF ...............................145\n E.1.2. Client Authentication .................................146\n E.1.3. 0-RTT .................................................146\n E.1.4. Exporter Independence .................................146\n E.1.5. Post-Compromise Security ..............................146\n E.1.6. External References ...................................147\n E.2. Record Layer ..............................................147\n E.2.1. External References ...................................148\n E.3. Traffic Analysis ..........................................148\n E.4. Side-Channel Attacks ......................................149\n E.5. Replay Attacks on 0-RTT ...................................150\n E.5.1. Replay and Exporters ..................................151\n E.6. PSK Identity Exposure .....................................152\n E.7. Sharing PSKs ..............................................152\n E.8. Attacks on Static RSA .....................................152\n Contributors .....................................................153\n Author's Address .................................................160\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 5]\n\f\nRFC 8446 TLS August 2018\n\n\n1. Introduction\n\n The primary goal of TLS is to provide a secure channel between two\n communicating peers; the only requirement from the underlying\n transport is a reliable, in-order data stream. Specifically, the\n secure channel should provide the following properties:\n\n - Authentication: The server side of the channel is always\n authenticated; the client side is optionally authenticated.\n Authentication can happen via asymmetric cryptography (e.g., RSA\n [RSA], the Elliptic Curve Digital Signature Algorithm (ECDSA)\n [ECDSA], or the Edwards-Curve Digital Signature Algorithm (EdDSA)\n [RFC8032]) or a symmetric pre-shared key (PSK).\n\n - Confidentiality: Data sent over the channel after establishment is\n only visible to the endpoints. TLS does not hide the length of\n the data it transmits, though endpoints are able to pad TLS\n records in order to obscure lengths and improve protection against\n traffic analysis techniques.\n\n - Integrity: Data sent over the channel after establishment cannot\n be modified by attackers without detection.\n\n These properties should be true even in the face of an attacker who\n has complete control of the network, as described in [RFC3552]. See\n Appendix E for a more complete statement of the relevant security\n properties.\n\n TLS consists of two primary components:\n\n - A handshake protocol (Section 4) that authenticates the\n communicating parties, negotiates cryptographic modes and\n parameters, and establishes shared keying material. The handshake\n protocol is designed to resist tampering; an active attacker\n should not be able to force the peers to negotiate different\n parameters than they would if the connection were not under\n attack.\n\n - A record protocol (Section 5) that uses the parameters established\n by the handshake protocol to protect traffic between the\n communicating peers. The record protocol divides traffic up into\n a series of records, each of which is independently protected\n using the traffic keys.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 6]\n\f\nRFC 8446 TLS August 2018\n\n\n TLS is application protocol independent; higher-level protocols can\n layer on top of TLS transparently. The TLS standard, however, does\n not specify how protocols add security with TLS; how to initiate TLS\n handshaking and how to interpret the authentication certificates\n exchanged are left to the judgment of the designers and implementors\n of protocols that run on top of TLS.\n\n This document defines TLS version 1.3. While TLS 1.3 is not directly\n compatible with previous versions, all versions of TLS incorporate a\n versioning mechanism which allows clients and servers to\n interoperably negotiate a common version if one is supported by both\n peers.\n\n This document supersedes and obsoletes previous versions of TLS,\n including version 1.2 [RFC5246]. It also obsoletes the TLS ticket\n mechanism defined in [RFC5077] and replaces it with the mechanism\n defined in Section 2.2. Because TLS 1.3 changes the way keys are\n derived, it updates [RFC5705] as described in Section 7.5. It also\n changes how Online Certificate Status Protocol (OCSP) messages are\n carried and therefore updates [RFC6066] and obsoletes [RFC6961] as\n described in Section 4.4.2.1.\n\n1.1. Conventions and Terminology\n\n The key words \"MUST\", \"MUST NOT\", \"REQUIRED\", \"SHALL\", \"SHALL NOT\",\n \"SHOULD\", \"SHOULD NOT\", \"RECOMMENDED\", \"NOT RECOMMENDED\", \"MAY\", and\n \"OPTIONAL\" in this document are to be interpreted as described in\n BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all\n capitals, as shown here.\n\n The following terms are used:\n\n client: The endpoint initiating the TLS connection.\n\n connection: A transport-layer connection between two endpoints.\n\n endpoint: Either the client or server of the connection.\n\n handshake: An initial negotiation between client and server that\n establishes the parameters of their subsequent interactions\n within TLS.\n\n peer: An endpoint. When discussing a particular endpoint, \"peer\"\n refers to the endpoint that is not the primary subject of\n discussion.\n\n\n\n\n\n\nRescorla Standards Track [Page 7]\n\f\nRFC 8446 TLS August 2018\n\n\n receiver: An endpoint that is receiving records.\n\n sender: An endpoint that is transmitting records.\n\n server: The endpoint that did not initiate the TLS connection.\n\n1.2. Major Differences from TLS 1.2\n\n The following is a list of the major functional differences between\n TLS 1.2 and TLS 1.3. It is not intended to be exhaustive, and there\n are many minor differences.\n\n - The list of supported symmetric encryption algorithms has been\n pruned of all algorithms that are considered legacy. Those that\n remain are all Authenticated Encryption with Associated Data\n (AEAD) algorithms. The cipher suite concept has been changed to\n separate the authentication and key exchange mechanisms from the\n record protection algorithm (including secret key length) and a\n hash to be used with both the key derivation function and\n handshake message authentication code (MAC).\n\n - A zero round-trip time (0-RTT) mode was added, saving a round trip\n at connection setup for some application data, at the cost of\n certain security properties.\n\n - Static RSA and Diffie-Hellman cipher suites have been removed; all\n public-key based key exchange mechanisms now provide forward\n secrecy.\n\n - All handshake messages after the ServerHello are now encrypted.\n The newly introduced EncryptedExtensions message allows various\n extensions previously sent in the clear in the ServerHello to also\n enjoy confidentiality protection.\n\n - The key derivation functions have been redesigned. The new design\n allows easier analysis by cryptographers due to their improved key\n separation properties. The HMAC-based Extract-and-Expand Key\n Derivation Function (HKDF) is used as an underlying primitive.\n\n - The handshake state machine has been significantly restructured to\n be more consistent and to remove superfluous messages such as\n ChangeCipherSpec (except when needed for middlebox compatibility).\n\n - Elliptic curve algorithms are now in the base spec, and new\n signature algorithms, such as EdDSA, are included. TLS 1.3\n removed point format negotiation in favor of a single point format\n for each curve.\n\n\n\n\nRescorla Standards Track [Page 8]\n\f\nRFC 8446 TLS August 2018\n\n\n - Other cryptographic improvements were made, including changing the\n RSA padding to use the RSA Probabilistic Signature Scheme\n (RSASSA-PSS), and the removal of compression, the Digital\n Signature Algorithm (DSA), and custom Ephemeral Diffie-Hellman\n (DHE) groups.\n\n - The TLS 1.2 version negotiation mechanism has been deprecated in\n favor of a version list in an extension. This increases\n compatibility with existing servers that incorrectly implemented\n version negotiation.\n\n - Session resumption with and without server-side state as well as\n the PSK-based cipher suites of earlier TLS versions have been\n replaced by a single new PSK exchange.\n\n - References have been updated to point to the updated versions of\n RFCs, as appropriate (e.g., RFC 5280 rather than RFC 3280).\n\n1.3. Updates Affecting TLS 1.2\n\n This document defines several changes that optionally affect\n implementations of TLS 1.2, including those which do not also support\n TLS 1.3:\n\n - A version downgrade protection mechanism is described in\n Section 4.1.3.\n\n - RSASSA-PSS signature schemes are defined in Section 4.2.3.\n\n - The \"supported_versions\" ClientHello extension can be used to\n negotiate the version of TLS to use, in preference to the\n legacy_version field of the ClientHello.\n\n - The \"signature_algorithms_cert\" extension allows a client to\n indicate which signature algorithms it can validate in X.509\n certificates.\n\n Additionally, this document clarifies some compliance requirements\n for earlier versions of TLS; see Section 9.3.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 9]\n\f\nRFC 8446 TLS August 2018\n\n\n2. Protocol Overview\n\n The cryptographic parameters used by the secure channel are produced\n by the TLS handshake protocol. This sub-protocol of TLS is used by\n the client and server when first communicating with each other. The\n handshake protocol allows peers to negotiate a protocol version,\n select cryptographic algorithms, optionally authenticate each other,\n and establish shared secret keying material. Once the handshake is\n complete, the peers use the established keys to protect the\n application-layer traffic.\n\n A failure of the handshake or other protocol error triggers the\n termination of the connection, optionally preceded by an alert\n message (Section 6).\n\n TLS supports three basic key exchange modes:\n\n - (EC)DHE (Diffie-Hellman over either finite fields or elliptic\n curves)\n\n - PSK-only\n\n - PSK with (EC)DHE\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 10]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 1 below shows the basic full TLS handshake:\n\n Client Server\n\nKey ^ ClientHello\nExch | + key_share*\n | + signature_algorithms*\n | + psk_key_exchange_modes*\n v + pre_shared_key* -------->\n ServerHello ^ Key\n + key_share* | Exch\n + pre_shared_key* v\n {EncryptedExtensions} ^ Server\n {CertificateRequest*} v Params\n {Certificate*} ^\n {CertificateVerify*} | Auth\n {Finished} v\n <-------- [Application Data*]\n ^ {Certificate*}\nAuth | {CertificateVerify*}\n v {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 1: Message Flow for Full TLS Handshake\n\n The handshake can be thought of as having three phases (indicated in\n the diagram above):\n\n - Key Exchange: Establish shared keying material and select the\n cryptographic parameters. Everything after this phase is\n encrypted.\n\n - Server Parameters: Establish other handshake parameters\n (whether the client is authenticated, application-layer protocol\n support, etc.).\n\n\n\n\nRescorla Standards Track [Page 11]\n\f\nRFC 8446 TLS August 2018\n\n\n - Authentication: Authenticate the server (and, optionally, the\n client) and provide key confirmation and handshake integrity.\n\n In the Key Exchange phase, the client sends the ClientHello\n (Section 4.1.2) message, which contains a random nonce\n (ClientHello.random); its offered protocol versions; a list of\n symmetric cipher/HKDF hash pairs; either a set of Diffie-Hellman key\n shares (in the \"key_share\" (Section 4.2.8) extension), a set of\n pre-shared key labels (in the \"pre_shared_key\" (Section 4.2.11)\n extension), or both; and potentially additional extensions.\n Additional fields and/or messages may also be present for middlebox\n compatibility.\n\n The server processes the ClientHello and determines the appropriate\n cryptographic parameters for the connection. It then responds with\n its own ServerHello (Section 4.1.3), which indicates the negotiated\n connection parameters. The combination of the ClientHello and the\n ServerHello determines the shared keys. If (EC)DHE key establishment\n is in use, then the ServerHello contains a \"key_share\" extension with\n the server's ephemeral Diffie-Hellman share; the server's share MUST\n be in the same group as one of the client's shares. If PSK key\n establishment is in use, then the ServerHello contains a\n \"pre_shared_key\" extension indicating which of the client's offered\n PSKs was selected. Note that implementations can use (EC)DHE and PSK\n together, in which case both extensions will be supplied.\n\n The server then sends two messages to establish the Server\n Parameters:\n\n EncryptedExtensions: responses to ClientHello extensions that are\n not required to determine the cryptographic parameters, other than\n those that are specific to individual certificates.\n [Section 4.3.1]\n\n CertificateRequest: if certificate-based client authentication is\n desired, the desired parameters for that certificate. This\n message is omitted if client authentication is not desired.\n [Section 4.3.2]\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 12]\n\f\nRFC 8446 TLS August 2018\n\n\n Finally, the client and server exchange Authentication messages. TLS\n uses the same set of messages every time that certificate-based\n authentication is needed. (PSK-based authentication happens as a\n side effect of key exchange.) Specifically:\n\n Certificate: The certificate of the endpoint and any per-certificate\n extensions. This message is omitted by the server if not\n authenticating with a certificate and by the client if the server\n did not send CertificateRequest (thus indicating that the client\n should not authenticate with a certificate). Note that if raw\n public keys [RFC7250] or the cached information extension\n [RFC7924] are in use, then this message will not contain a\n certificate but rather some other value corresponding to the\n server's long-term key. [Section 4.4.2]\n\n CertificateVerify: A signature over the entire handshake using the\n private key corresponding to the public key in the Certificate\n message. This message is omitted if the endpoint is not\n authenticating via a certificate. [Section 4.4.3]\n\n Finished: A MAC (Message Authentication Code) over the entire\n handshake. This message provides key confirmation, binds the\n endpoint's identity to the exchanged keys, and in PSK mode also\n authenticates the handshake. [Section 4.4.4]\n\n Upon receiving the server's messages, the client responds with its\n Authentication messages, namely Certificate and CertificateVerify (if\n requested), and Finished.\n\n At this point, the handshake is complete, and the client and server\n derive the keying material required by the record layer to exchange\n application-layer data protected through authenticated encryption.\n Application Data MUST NOT be sent prior to sending the Finished\n message, except as specified in Section 2.3. Note that while the\n server may send Application Data prior to receiving the client's\n Authentication messages, any data sent at that point is, of course,\n being sent to an unauthenticated peer.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 13]\n\f\nRFC 8446 TLS August 2018\n\n\n2.1. Incorrect DHE Share\n\n If the client has not provided a sufficient \"key_share\" extension\n (e.g., it includes only DHE or ECDHE groups unacceptable to or\n unsupported by the server), the server corrects the mismatch with a\n HelloRetryRequest and the client needs to restart the handshake with\n an appropriate \"key_share\" extension, as shown in Figure 2. If no\n common cryptographic parameters can be negotiated, the server MUST\n abort the handshake with an appropriate alert.\n\n Client Server\n\n ClientHello\n + key_share -------->\n HelloRetryRequest\n <-------- + key_share\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 2: Message Flow for a Full Handshake with\n Mismatched Parameters\n\n Note: The handshake transcript incorporates the initial\n ClientHello/HelloRetryRequest exchange; it is not reset with the\n new ClientHello.\n\n TLS also allows several optimized variants of the basic handshake, as\n described in the following sections.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 14]\n\f\nRFC 8446 TLS August 2018\n\n\n2.2. Resumption and Pre-Shared Key (PSK)\n\n Although TLS PSKs can be established out of band, PSKs can also be\n established in a previous connection and then used to establish a new\n connection (\"session resumption\" or \"resuming\" with a PSK). Once a\n handshake has completed, the server can send the client a PSK\n identity that corresponds to a unique key derived from the initial\n handshake (see Section 4.6.1). The client can then use that PSK\n identity in future handshakes to negotiate the use of the associated\n PSK. If the server accepts the PSK, then the security context of the\n new connection is cryptographically tied to the original connection\n and the key derived from the initial handshake is used to bootstrap\n the cryptographic state instead of a full handshake. In TLS 1.2 and\n below, this functionality was provided by \"session IDs\" and \"session\n tickets\" [RFC5077]. Both mechanisms are obsoleted in TLS 1.3.\n\n PSKs can be used with (EC)DHE key exchange in order to provide\n forward secrecy in combination with shared keys, or can be used\n alone, at the cost of losing forward secrecy for the application\n data.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 15]\n\f\nRFC 8446 TLS August 2018\n\n\n Figure 3 shows a pair of handshakes in which the first handshake\n establishes a PSK and the second handshake uses it:\n\n Client Server\n\n Initial Handshake:\n ClientHello\n + key_share -------->\n ServerHello\n + key_share\n {EncryptedExtensions}\n {CertificateRequest*}\n {Certificate*}\n {CertificateVerify*}\n {Finished}\n <-------- [Application Data*]\n {Certificate*}\n {CertificateVerify*}\n {Finished} -------->\n <-------- [NewSessionTicket]\n [Application Data] <-------> [Application Data]\n\n\n Subsequent Handshake:\n ClientHello\n + key_share*\n + pre_shared_key -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n {Finished}\n <-------- [Application Data*]\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n Figure 3: Message Flow for Resumption and PSK\n\n As the server is authenticating via a PSK, it does not send a\n Certificate or a CertificateVerify message. When a client offers\n resumption via a PSK, it SHOULD also supply a \"key_share\" extension\n to the server to allow the server to decline resumption and fall back\n to a full handshake, if needed. The server responds with a\n \"pre_shared_key\" extension to negotiate the use of PSK key\n establishment and can (as shown here) respond with a \"key_share\"\n extension to do (EC)DHE key establishment, thus providing forward\n secrecy.\n\n\n\n\nRescorla Standards Track [Page 16]\n\f\nRFC 8446 TLS August 2018\n\n\n When PSKs are provisioned out of band, the PSK identity and the KDF\n hash algorithm to be used with the PSK MUST also be provisioned.\n\n Note: When using an out-of-band provisioned pre-shared secret, a\n critical consideration is using sufficient entropy during the key\n generation, as discussed in [RFC4086]. Deriving a shared secret\n from a password or other low-entropy sources is not secure. A\n low-entropy secret, or password, is subject to dictionary attacks\n based on the PSK binder. The specified PSK authentication is not\n a strong password-based authenticated key exchange even when used\n with Diffie-Hellman key establishment. Specifically, it does not\n prevent an attacker that can observe the handshake from performing\n a brute-force attack on the password/pre-shared key.\n\n2.3. 0-RTT Data\n\n When clients and servers share a PSK (either obtained externally or\n via a previous handshake), TLS 1.3 allows clients to send data on the\n first flight (\"early data\"). The client uses the PSK to authenticate\n the server and to encrypt the early data.\n\n As shown in Figure 4, the 0-RTT data is just added to the 1-RTT\n handshake in the first flight. The rest of the handshake uses the\n same messages as for a 1-RTT handshake with PSK resumption.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 17]\n\f\nRFC 8446 TLS August 2018\n\n\n Client Server\n\n ClientHello\n + early_data\n + key_share*\n + psk_key_exchange_modes\n + pre_shared_key\n (Application Data*) -------->\n ServerHello\n + pre_shared_key\n + key_share*\n {EncryptedExtensions}\n + early_data*\n {Finished}\n <-------- [Application Data*]\n (EndOfEarlyData)\n {Finished} -------->\n [Application Data] <-------> [Application Data]\n\n + Indicates noteworthy extensions sent in the\n previously noted message.\n\n * Indicates optional or situation-dependent\n messages/extensions that are not always sent.\n\n () Indicates messages protected using keys\n derived from a client_early_traffic_secret.\n\n {} Indicates messages protected using keys\n derived from a [sender]_handshake_traffic_secret.\n\n [] Indicates messages protected using keys\n derived from [sender]_application_traffic_secret_N.\n\n Figure 4: Message Flow for a 0-RTT Handshake\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 18]\n\f\nRFC 8446 TLS August 2018\n\n\n IMPORTANT NOTE: The security properties for 0-RTT data are weaker\n than those for other kinds of TLS data. Specifically:\n\n 1. This data is not forward secret, as it is encrypted solely under\n keys derived using the offered PSK.\n\n 2. There are no guarantees of non-replay between connections.\n Protection against replay for ordinary TLS 1.3 1-RTT data is\n provided via the server's Random value, but 0-RTT data does not\n depend on the ServerHello and therefore has weaker guarantees.\n This is especially relevant if the data is authenticated either\n with TLS client authentication or inside the application\n protocol. The same warnings apply to any use of the\n early_exporter_master_secret.\n\n 0-RTT data cannot be duplicated within a connection (i.e., the server\n will not process the same data twice for the same connection), and an\n attacker will not be able to make 0-RTT data appear to be 1-RTT data\n (because it is protected with different keys). Appendix E.5 contains\n a description of potential attacks, and Section 8 describes\n mechanisms which the server can use to limit the impact of replay.\n\n3. Presentation Language\n\n This document deals with the formatting of data in an external\n representation. The following very basic and somewhat casually\n defined presentation syntax will be used.\n\n3.1. Basic Block Size\n\n The representation of all data items is explicitly specified. The\n basic data block size is one byte (i.e., 8 bits). Multiple-byte data\n items are concatenations of bytes, from left to right, from top to\n bottom. From the byte stream, a multi-byte item (a numeric in the\n following example) is formed (using C notation) by:\n\n value = (byte[0] << 8*(n-1)) | (byte[1] << 8*(n-2)) |\n ... | byte[n-1];\n\n This byte ordering for multi-byte values is the commonplace network\n byte order or big-endian format.\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 19]\n\f\nRFC 8446 TLS August 2018\n\n\n3.2. Miscellaneous\n\n Comments begin with \"/*\" and end with \"*/\".\n\n Optional components are denoted by enclosing them in \"[[ ]]\" (double\n brackets).\n\n Single-byte entities containing uninterpreted data are of\n type opaque.\n\n A type alias T' for an existing type T is defined by:\n\n T T';\n\n3.3. Numbers\n\n The basic numeric data type is an unsigned byte (uint8). All larger\n numeric data types are constructed from a fixed-length series of\n bytes concatenated as described in Section 3.1 and are also unsigned.\n The following numeric types are predefined.\n\n uint8 uint16[2];\n uint8 uint24[3];\n uint8 uint32[4];\n uint8 uint64[8];\n\n All values, here and elsewhere in the specification, are transmitted\n in network byte (big-endian) order; the uint32 represented by the hex\n bytes 01 02 03 04 is equivalent to the decimal value 16909060.\n\n3.4. Vectors\n\n A vector (single-dimensioned array) is a stream of homogeneous data\n elements. The size of the vector may be specified at documentation\n time or left unspecified until runtime. In either case, the length\n declares the number of bytes, not the number of elements, in the\n vector. The syntax for specifying a new type, T', that is a fixed-\n length vector of type T is\n\n T T'[n];\n\n Here, T' occupies n bytes in the data stream, where n is a multiple\n of the size of T. The length of the vector is not included in the\n encoded stream.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 20]\n\f\nRFC 8446 TLS August 2018\n\n\n In the following example, Datum is defined to be three consecutive\n bytes that the protocol does not interpret, while Data is three\n consecutive Datum, consuming a total of nine bytes.\n\n opaque Datum[3]; /* three uninterpreted bytes */\n Datum Data[9]; /* three consecutive 3-byte vectors */\n\n Variable-length vectors are defined by specifying a subrange of legal\n lengths, inclusively, using the notation . When\n these are encoded, the actual length precedes the vector's contents\n in the byte stream. The length will be in the form of a number\n consuming as many bytes as required to hold the vector's specified\n maximum (ceiling) length. A variable-length vector with an actual\n length field of zero is referred to as an empty vector.\n\n T T';\n\n In the following example, \"mandatory\" is a vector that must contain\n between 300 and 400 bytes of type opaque. It can never be empty.\n The actual length field consumes two bytes, a uint16, which is\n sufficient to represent the value 400 (see Section 3.3). Similarly,\n \"longer\" can represent up to 800 bytes of data, or 400 uint16\n elements, and it may be empty. Its encoding will include a two-byte\n actual length field prepended to the vector. The length of an\n encoded vector must be an exact multiple of the length of a single\n element (e.g., a 17-byte vector of uint16 would be illegal).\n\n opaque mandatory<300..400>;\n /* length field is two bytes, cannot be empty */\n uint16 longer<0..800>;\n /* zero to 400 16-bit unsigned integers */\n\n3.5. Enumerateds\n\n An additional sparse data type, called \"enum\" or \"enumerated\", is\n available. Each definition is a different type. Only enumerateds of\n the same type may be assigned or compared. Every element of an\n enumerated must be assigned a value, as demonstrated in the following\n example. Since the elements of the enumerated are not ordered, they\n can be assigned any unique value, in any order.\n\n enum { e1(v1), e2(v2), ... , en(vn) [[, (n)]] } Te;\n\n Future extensions or additions to the protocol may define new values.\n Implementations need to be able to parse and ignore unknown values\n unless the definition of the field states otherwise.\n\n\n\n\n\nRescorla Standards Track [Page 21]\n\f\nRFC 8446 TLS August 2018\n\n\n An enumerated occupies as much space in the byte stream as would its\n maximal defined ordinal value. The following definition would cause\n one byte to be used to carry fields of type Color.\n\n enum { red(3), blue(5), white(7) } Color;\n\n One may optionally specify a value without its associated tag to\n force the width definition without defining a superfluous element.\n\n In the following example, Taste will consume two bytes in the data\n stream but can only assume the values 1, 2, or 4 in the current\n version of the protocol.\n\n enum { sweet(1), sour(2), bitter(4), (32000) } Taste;\n\n The names of the elements of an enumeration are scoped within the\n defined type. In the first example, a fully qualified reference to\n the second element of the enumeration would be Color.blue. Such\n qualification is not required if the target of the assignment is well\n specified.\n\n Color color = Color.blue; /* overspecified, legal */\n Color color = blue; /* correct, type implicit */\n\n The names assigned to enumerateds do not need to be unique. The\n numerical value can describe a range over which the same name\n applies. The value includes the minimum and maximum inclusive values\n in that range, separated by two period characters. This is\n principally useful for reserving regions of the space.\n\n enum { sad(0), meh(1..254), happy(255) } Mood;\n\n3.6. Constructed Types\n\n Structure types may be constructed from primitive types for\n convenience. Each specification declares a new, unique type. The\n syntax used for definitions is much like that of C.\n\n struct {\n T1 f1;\n T2 f2;\n ...\n Tn fn;\n } T;\n\n Fixed- and variable-length vector fields are allowed using the\n standard vector syntax. Structures V1 and V2 in the variants example\n (Section 3.8) demonstrate this.\n\n\n\nRescorla Standards Track [Page 22]\n\f\nRFC 8446 TLS August 2018\n\n\n The fields within a structure may be qualified using the type's name,\n with a syntax much like that available for enumerateds. For example,\n T.f2 refers to the second field of the previous declaration.\n\n3.7. Constants\n\n Fields and variables may be assigned a fixed value using \"=\", as in:\n\n struct {\n T1 f1 = 8; /* T.f1 must always be 8 */\n T2 f2;\n } T;\n\n3.8. Variants\n\n Defined structures may have variants based on some knowledge that is\n available within the environment. The selector must be an enumerated\n type that defines the possible variants the structure defines. Each\n arm of the select (below) specifies the type of that variant's field\n and an optional field label. The mechanism by which the variant is\n selected at runtime is not prescribed by the presentation language.\n\n struct {\n T1 f1;\n T2 f2;\n ....\n Tn fn;\n select (E) {\n case e1: Te1 [[fe1]];\n case e2: Te2 [[fe2]];\n ....\n case en: Ten [[fen]];\n };\n } Tv;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 23]\n\f\nRFC 8446 TLS August 2018\n\n\n For example:\n\n enum { apple(0), orange(1) } VariantTag;\n\n struct {\n uint16 number;\n opaque string<0..10>; /* variable length */\n } V1;\n\n struct {\n uint32 number;\n opaque string[10]; /* fixed length */\n } V2;\n\n struct {\n VariantTag type;\n select (VariantRecord.type) {\n case apple: V1;\n case orange: V2;\n };\n } VariantRecord;\n\n4. Handshake Protocol\n\n The handshake protocol is used to negotiate the security parameters\n of a connection. Handshake messages are supplied to the TLS record\n layer, where they are encapsulated within one or more TLSPlaintext or\n TLSCiphertext structures which are processed and transmitted as\n specified by the current active connection state.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 24]\n\f\nRFC 8446 TLS August 2018\n\n\n enum {\n client_hello(1),\n server_hello(2),\n new_session_ticket(4),\n end_of_early_data(5),\n encrypted_extensions(8),\n certificate(11),\n certificate_request(13),\n certificate_verify(15),\n finished(20),\n key_update(24),\n message_hash(254),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* remaining bytes in message */\n select (Handshake.msg_type) {\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case end_of_early_data: EndOfEarlyData;\n case encrypted_extensions: EncryptedExtensions;\n case certificate_request: CertificateRequest;\n case certificate: Certificate;\n case certificate_verify: CertificateVerify;\n case finished: Finished;\n case new_session_ticket: NewSessionTicket;\n case key_update: KeyUpdate;\n };\n } Handshake;\n\n Protocol messages MUST be sent in the order defined in Section 4.4.1\n and shown in the diagrams in Section 2. A peer which receives a\n handshake message in an unexpected order MUST abort the handshake\n with an \"unexpected_message\" alert.\n\n New handshake message types are assigned by IANA as described in\n Section 11.\n\n4.1. Key Exchange Messages\n\n The key exchange messages are used to determine the security\n capabilities of the client and the server and to establish shared\n secrets, including the traffic keys used to protect the rest of the\n handshake and the data.\n\n\n\n\n\nRescorla Standards Track [Page 25]\n\f\nRFC 8446 TLS August 2018\n\n\n4.1.1. Cryptographic Negotiation\n\n In TLS, the cryptographic negotiation proceeds by the client offering\n the following four sets of options in its ClientHello:\n\n - A list of cipher suites which indicates the AEAD algorithm/HKDF\n hash pairs which the client supports.\n\n - A \"supported_groups\" (Section 4.2.7) extension which indicates the\n (EC)DHE groups which the client supports and a \"key_share\"\n (Section 4.2.8) extension which contains (EC)DHE shares for some\n or all of these groups.\n\n - A \"signature_algorithms\" (Section 4.2.3) extension which indicates\n the signature algorithms which the client can accept. A\n \"signature_algorithms_cert\" extension (Section 4.2.3) may also be\n added to indicate certificate-specific signature algorithms.\n\n - A \"pre_shared_key\" (Section 4.2.11) extension which contains a\n list of symmetric key identities known to the client and a\n \"psk_key_exchange_modes\" (Section 4.2.9) extension which indicates\n the key exchange modes that may be used with PSKs.\n\n If the server does not select a PSK, then the first three of these\n options are entirely orthogonal: the server independently selects a\n cipher suite, an (EC)DHE group and key share for key establishment,\n and a signature algorithm/certificate pair to authenticate itself to\n the client. If there is no overlap between the received\n \"supported_groups\" and the groups supported by the server, then the\n server MUST abort the handshake with a \"handshake_failure\" or an\n \"insufficient_security\" alert.\n\n If the server selects a PSK, then it MUST also select a key\n establishment mode from the set indicated by the client's\n \"psk_key_exchange_modes\" extension (at present, PSK alone or with\n (EC)DHE). Note that if the PSK can be used without (EC)DHE, then\n non-overlap in the \"supported_groups\" parameters need not be fatal,\n as it is in the non-PSK case discussed in the previous paragraph.\n\n If the server selects an (EC)DHE group and the client did not offer a\n compatible \"key_share\" extension in the initial ClientHello, the\n server MUST respond with a HelloRetryRequest (Section 4.1.4) message.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 26]\n\f\nRFC 8446 TLS August 2018\n\n\n If the server successfully selects parameters and does not require a\n HelloRetryRequest, it indicates the selected parameters in the\n ServerHello as follows:\n\n - If PSK is being used, then the server will send a \"pre_shared_key\"\n extension indicating the selected key.\n\n - When (EC)DHE is in use, the server will also provide a \"key_share\"\n extension. If PSK is not being used, then (EC)DHE and\n certificate-based authentication are always used.\n\n - When authenticating via a certificate, the server will send the\n Certificate (Section 4.4.2) and CertificateVerify (Section 4.4.3)\n messages. In TLS 1.3 as defined by this document, either a PSK or\n a certificate is always used, but not both. Future documents may\n define how to use them together.\n\n If the server is unable to negotiate a supported set of parameters\n (i.e., there is no overlap between the client and server parameters),\n it MUST abort the handshake with either a \"handshake_failure\" or\n \"insufficient_security\" fatal alert (see Section 6).\n\n4.1.2. Client Hello\n\n When a client first connects to a server, it is REQUIRED to send the\n ClientHello as its first TLS message. The client will also send a\n ClientHello when the server has responded to its ClientHello with a\n HelloRetryRequest. In that case, the client MUST send the same\n ClientHello without modification, except as follows:\n\n - If a \"key_share\" extension was supplied in the HelloRetryRequest,\n replacing the list of shares with a list containing a single\n KeyShareEntry from the indicated group.\n\n - Removing the \"early_data\" extension (Section 4.2.10) if one was\n present. Early data is not permitted after a HelloRetryRequest.\n\n - Including a \"cookie\" extension if one was provided in the\n HelloRetryRequest.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 27]\n\f\nRFC 8446 TLS August 2018\n\n\n - Updating the \"pre_shared_key\" extension if present by recomputing\n the \"obfuscated_ticket_age\" and binder values and (optionally)\n removing any PSKs which are incompatible with the server's\n indicated cipher suite.\n\n - Optionally adding, removing, or changing the length of the\n \"padding\" extension [RFC7685].\n\n - Other modifications that may be allowed by an extension defined in\n the future and present in the HelloRetryRequest.\n\n Because TLS 1.3 forbids renegotiation, if a server has negotiated\n TLS 1.3 and receives a ClientHello at any other time, it MUST\n terminate the connection with an \"unexpected_message\" alert.\n\n If a server established a TLS connection with a previous version of\n TLS and receives a TLS 1.3 ClientHello in a renegotiation, it MUST\n retain the previous protocol version. In particular, it MUST NOT\n negotiate TLS 1.3.\n\n Structure of this message:\n\n uint16 ProtocolVersion;\n opaque Random[32];\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id<0..32>;\n CipherSuite cipher_suites<2..2^16-2>;\n opaque legacy_compression_methods<1..2^8-1>;\n Extension extensions<8..2^16-1>;\n } ClientHello;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 28]\n\f\nRFC 8446 TLS August 2018\n\n\n legacy_version: In previous versions of TLS, this field was used for\n version negotiation and represented the highest version number\n supported by the client. Experience has shown that many servers\n do not properly implement version negotiation, leading to \"version\n intolerance\" in which the server rejects an otherwise acceptable\n ClientHello with a version number higher than it supports. In\n TLS 1.3, the client indicates its version preferences in the\n \"supported_versions\" extension (Section 4.2.1) and the\n legacy_version field MUST be set to 0x0303, which is the version\n number for TLS 1.2. TLS 1.3 ClientHellos are identified as having\n a legacy_version of 0x0303 and a supported_versions extension\n present with 0x0304 as the highest version indicated therein.\n (See Appendix D for details about backward compatibility.)\n\n random: 32 bytes generated by a secure random number generator. See\n Appendix C for additional information.\n\n legacy_session_id: Versions of TLS before TLS 1.3 supported a\n \"session resumption\" feature which has been merged with pre-shared\n keys in this version (see Section 2.2). A client which has a\n cached session ID set by a pre-TLS 1.3 server SHOULD set this\n field to that value. In compatibility mode (see Appendix D.4),\n this field MUST be non-empty, so a client not offering a\n pre-TLS 1.3 session MUST generate a new 32-byte value. This value\n need not be random but SHOULD be unpredictable to avoid\n implementations fixating on a specific value (also known as\n ossification). Otherwise, it MUST be set as a zero-length vector\n (i.e., a zero-valued single byte length field).\n\n cipher_suites: A list of the symmetric cipher options supported by\n the client, specifically the record protection algorithm\n (including secret key length) and a hash to be used with HKDF, in\n descending order of client preference. Values are defined in\n Appendix B.4. If the list contains cipher suites that the server\n does not recognize, support, or wish to use, the server MUST\n ignore those cipher suites and process the remaining ones as\n usual. If the client is attempting a PSK key establishment, it\n SHOULD advertise at least one cipher suite indicating a Hash\n associated with the PSK.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 29]\n\f\nRFC 8446 TLS August 2018\n\n\n legacy_compression_methods: Versions of TLS before 1.3 supported\n compression with the list of supported compression methods being\n sent in this field. For every TLS 1.3 ClientHello, this vector\n MUST contain exactly one byte, set to zero, which corresponds to\n the \"null\" compression method in prior versions of TLS. If a\n TLS 1.3 ClientHello is received with any other value in this\n field, the server MUST abort the handshake with an\n \"illegal_parameter\" alert. Note that TLS 1.3 servers might\n receive TLS 1.2 or prior ClientHellos which contain other\n compression methods and (if negotiating such a prior version) MUST\n follow the procedures for the appropriate prior version of TLS.\n\n extensions: Clients request extended functionality from servers by\n sending data in the extensions field. The actual \"Extension\"\n format is defined in Section 4.2. In TLS 1.3, the use of certain\n extensions is mandatory, as functionality has moved into\n extensions to preserve ClientHello compatibility with previous\n versions of TLS. Servers MUST ignore unrecognized extensions.\n\n All versions of TLS allow an extensions field to optionally follow\n the compression_methods field. TLS 1.3 ClientHello messages always\n contain extensions (minimally \"supported_versions\", otherwise, they\n will be interpreted as TLS 1.2 ClientHello messages). However,\n TLS 1.3 servers might receive ClientHello messages without an\n extensions field from prior versions of TLS. The presence of\n extensions can be detected by determining whether there are bytes\n following the compression_methods field at the end of the\n ClientHello. Note that this method of detecting optional data\n differs from the normal TLS method of having a variable-length field,\n but it is used for compatibility with TLS before extensions were\n defined. TLS 1.3 servers will need to perform this check first and\n only attempt to negotiate TLS 1.3 if the \"supported_versions\"\n extension is present. If negotiating a version of TLS prior to 1.3,\n a server MUST check that the message either contains no data after\n legacy_compression_methods or that it contains a valid extensions\n block with no data following. If not, then it MUST abort the\n handshake with a \"decode_error\" alert.\n\n In the event that a client requests additional functionality using\n extensions and this functionality is not supplied by the server, the\n client MAY abort the handshake.\n\n After sending the ClientHello message, the client waits for a\n ServerHello or HelloRetryRequest message. If early data is in use,\n the client may transmit early Application Data (Section 2.3) while\n waiting for the next handshake message.\n\n\n\n\n\nRescorla Standards Track [Page 30]\n\f\nRFC 8446 TLS August 2018\n\n\n4.1.3. Server Hello\n\n The server will send this message in response to a ClientHello\n message to proceed with the handshake if it is able to negotiate an\n acceptable set of handshake parameters based on the ClientHello.\n\n Structure of this message:\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id_echo<0..32>;\n CipherSuite cipher_suite;\n uint8 legacy_compression_method = 0;\n Extension extensions<6..2^16-1>;\n } ServerHello;\n\n legacy_version: In previous versions of TLS, this field was used for\n version negotiation and represented the selected version number\n for the connection. Unfortunately, some middleboxes fail when\n presented with new values. In TLS 1.3, the TLS server indicates\n its version using the \"supported_versions\" extension\n (Section 4.2.1), and the legacy_version field MUST be set to\n 0x0303, which is the version number for TLS 1.2. (See Appendix D\n for details about backward compatibility.)\n\n random: 32 bytes generated by a secure random number generator. See\n Appendix C for additional information. The last 8 bytes MUST be\n overwritten as described below if negotiating TLS 1.2 or TLS 1.1,\n but the remaining bytes MUST be random. This structure is\n generated by the server and MUST be generated independently of the\n ClientHello.random.\n\n legacy_session_id_echo: The contents of the client's\n legacy_session_id field. Note that this field is echoed even if\n the client's value corresponded to a cached pre-TLS 1.3 session\n which the server has chosen not to resume. A client which\n receives a legacy_session_id_echo field that does not match what\n it sent in the ClientHello MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n cipher_suite: The single cipher suite selected by the server from\n the list in ClientHello.cipher_suites. A client which receives a\n cipher suite that was not offered MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n legacy_compression_method: A single byte which MUST have the\n value 0.\n\n\n\nRescorla Standards Track [Page 31]\n\f\nRFC 8446 TLS August 2018\n\n\n extensions: A list of extensions. The ServerHello MUST only include\n extensions which are required to establish the cryptographic\n context and negotiate the protocol version. All TLS 1.3\n ServerHello messages MUST contain the \"supported_versions\"\n extension. Current ServerHello messages additionally contain\n either the \"pre_shared_key\" extension or the \"key_share\"\n extension, or both (when using a PSK with (EC)DHE key\n establishment). Other extensions (see Section 4.2) are sent\n separately in the EncryptedExtensions message.\n\n For reasons of backward compatibility with middleboxes (see\n Appendix D.4), the HelloRetryRequest message uses the same structure\n as the ServerHello, but with Random set to the special value of the\n SHA-256 of \"HelloRetryRequest\":\n\n CF 21 AD 74 E5 9A 61 11 BE 1D 8C 02 1E 65 B8 91\n C2 A2 11 16 7A BB 8C 5E 07 9E 09 E2 C8 A8 33 9C\n\n Upon receiving a message with type server_hello, implementations MUST\n first examine the Random value and, if it matches this value, process\n it as described in Section 4.1.4).\n\n TLS 1.3 has a downgrade protection mechanism embedded in the server's\n random value. TLS 1.3 servers which negotiate TLS 1.2 or below in\n response to a ClientHello MUST set the last 8 bytes of their Random\n value specially in their ServerHello.\n\n If negotiating TLS 1.2, TLS 1.3 servers MUST set the last 8 bytes of\n their Random value to the bytes:\n\n 44 4F 57 4E 47 52 44 01\n\n If negotiating TLS 1.1 or below, TLS 1.3 servers MUST, and TLS 1.2\n servers SHOULD, set the last 8 bytes of their ServerHello.Random\n value to the bytes:\n\n 44 4F 57 4E 47 52 44 00\n\n TLS 1.3 clients receiving a ServerHello indicating TLS 1.2 or below\n MUST check that the last 8 bytes are not equal to either of these\n values. TLS 1.2 clients SHOULD also check that the last 8 bytes are\n not equal to the second value if the ServerHello indicates TLS 1.1 or\n below. If a match is found, the client MUST abort the handshake with\n an \"illegal_parameter\" alert. This mechanism provides limited\n protection against downgrade attacks over and above what is provided\n by the Finished exchange: because the ServerKeyExchange, a message\n present in TLS 1.2 and below, includes a signature over both random\n values, it is not possible for an active attacker to modify the\n\n\n\nRescorla Standards Track [Page 32]\n\f\nRFC 8446 TLS August 2018\n\n\n random values without detection as long as ephemeral ciphers are\n used. It does not provide downgrade protection when static RSA\n is used.\n\n Note: This is a change from [RFC5246], so in practice many TLS 1.2\n clients and servers will not behave as specified above.\n\n A legacy TLS client performing renegotiation with TLS 1.2 or prior\n and which receives a TLS 1.3 ServerHello during renegotiation MUST\n abort the handshake with a \"protocol_version\" alert. Note that\n renegotiation is not possible when TLS 1.3 has been negotiated.\n\n4.1.4. Hello Retry Request\n\n The server will send this message in response to a ClientHello\n message if it is able to find an acceptable set of parameters but the\n ClientHello does not contain sufficient information to proceed with\n the handshake. As discussed in Section 4.1.3, the HelloRetryRequest\n has the same format as a ServerHello message, and the legacy_version,\n legacy_session_id_echo, cipher_suite, and legacy_compression_method\n fields have the same meaning. However, for convenience we discuss\n \"HelloRetryRequest\" throughout this document as if it were a distinct\n message.\n\n The server's extensions MUST contain \"supported_versions\".\n Additionally, it SHOULD contain the minimal set of extensions\n necessary for the client to generate a correct ClientHello pair. As\n with the ServerHello, a HelloRetryRequest MUST NOT contain any\n extensions that were not first offered by the client in its\n ClientHello, with the exception of optionally the \"cookie\" (see\n Section 4.2.2) extension.\n\n Upon receipt of a HelloRetryRequest, the client MUST check the\n legacy_version, legacy_session_id_echo, cipher_suite, and\n legacy_compression_method as specified in Section 4.1.3 and then\n process the extensions, starting with determining the version using\n \"supported_versions\". Clients MUST abort the handshake with an\n \"illegal_parameter\" alert if the HelloRetryRequest would not result\n in any change in the ClientHello. If a client receives a second\n HelloRetryRequest in the same connection (i.e., where the ClientHello\n was itself in response to a HelloRetryRequest), it MUST abort the\n handshake with an \"unexpected_message\" alert.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 33]\n\f\nRFC 8446 TLS August 2018\n\n\n Otherwise, the client MUST process all extensions in the\n HelloRetryRequest and send a second updated ClientHello. The\n HelloRetryRequest extensions defined in this specification are:\n\n - supported_versions (see Section 4.2.1)\n\n - cookie (see Section 4.2.2)\n\n - key_share (see Section 4.2.8)\n\n A client which receives a cipher suite that was not offered MUST\n abort the handshake. Servers MUST ensure that they negotiate the\n same cipher suite when receiving a conformant updated ClientHello (if\n the server selects the cipher suite as the first step in the\n negotiation, then this will happen automatically). Upon receiving\n the ServerHello, clients MUST check that the cipher suite supplied in\n the ServerHello is the same as that in the HelloRetryRequest and\n otherwise abort the handshake with an \"illegal_parameter\" alert.\n\n In addition, in its updated ClientHello, the client SHOULD NOT offer\n any pre-shared keys associated with a hash other than that of the\n selected cipher suite. This allows the client to avoid having to\n compute partial hash transcripts for multiple hashes in the second\n ClientHello.\n\n The value of selected_version in the HelloRetryRequest\n \"supported_versions\" extension MUST be retained in the ServerHello,\n and a client MUST abort the handshake with an \"illegal_parameter\"\n alert if the value changes.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 34]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2. Extensions\n\n A number of TLS messages contain tag-length-value encoded extensions\n structures.\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n server_name(0), /* RFC 6066 */\n max_fragment_length(1), /* RFC 6066 */\n status_request(5), /* RFC 6066 */\n supported_groups(10), /* RFC 8422, 7919 */\n signature_algorithms(13), /* RFC 8446 */\n use_srtp(14), /* RFC 5764 */\n heartbeat(15), /* RFC 6520 */\n application_layer_protocol_negotiation(16), /* RFC 7301 */\n signed_certificate_timestamp(18), /* RFC 6962 */\n client_certificate_type(19), /* RFC 7250 */\n server_certificate_type(20), /* RFC 7250 */\n padding(21), /* RFC 7685 */\n pre_shared_key(41), /* RFC 8446 */\n early_data(42), /* RFC 8446 */\n supported_versions(43), /* RFC 8446 */\n cookie(44), /* RFC 8446 */\n psk_key_exchange_modes(45), /* RFC 8446 */\n certificate_authorities(47), /* RFC 8446 */\n oid_filters(48), /* RFC 8446 */\n post_handshake_auth(49), /* RFC 8446 */\n signature_algorithms_cert(50), /* RFC 8446 */\n key_share(51), /* RFC 8446 */\n (65535)\n } ExtensionType;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 35]\n\f\nRFC 8446 TLS August 2018\n\n\n Here:\n\n - \"extension_type\" identifies the particular extension type.\n\n - \"extension_data\" contains information specific to the particular\n extension type.\n\n The list of extension types is maintained by IANA as described in\n Section 11.\n\n Extensions are generally structured in a request/response fashion,\n though some extensions are just indications with no corresponding\n response. The client sends its extension requests in the ClientHello\n message, and the server sends its extension responses in the\n ServerHello, EncryptedExtensions, HelloRetryRequest, and Certificate\n messages. The server sends extension requests in the\n CertificateRequest message which a client MAY respond to with a\n Certificate message. The server MAY also send unsolicited extensions\n in the NewSessionTicket, though the client does not respond directly\n to these.\n\n Implementations MUST NOT send extension responses if the remote\n endpoint did not send the corresponding extension requests, with the\n exception of the \"cookie\" extension in the HelloRetryRequest. Upon\n receiving such an extension, an endpoint MUST abort the handshake\n with an \"unsupported_extension\" alert.\n\n The table below indicates the messages where a given extension may\n appear, using the following notation: CH (ClientHello),\n SH (ServerHello), EE (EncryptedExtensions), CT (Certificate),\n CR (CertificateRequest), NST (NewSessionTicket), and\n HRR (HelloRetryRequest). If an implementation receives an extension\n which it recognizes and which is not specified for the message in\n which it appears, it MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 36]\n\f\nRFC 8446 TLS August 2018\n\n\n +--------------------------------------------------+-------------+\n | Extension | TLS 1.3 |\n +--------------------------------------------------+-------------+\n | server_name [RFC6066] | CH, EE |\n | | |\n | max_fragment_length [RFC6066] | CH, EE |\n | | |\n | status_request [RFC6066] | CH, CR, CT |\n | | |\n | supported_groups [RFC7919] | CH, EE |\n | | |\n | signature_algorithms (RFC 8446) | CH, CR |\n | | |\n | use_srtp [RFC5764] | CH, EE |\n | | |\n | heartbeat [RFC6520] | CH, EE |\n | | |\n | application_layer_protocol_negotiation [RFC7301] | CH, EE |\n | | |\n | signed_certificate_timestamp [RFC6962] | CH, CR, CT |\n | | |\n | client_certificate_type [RFC7250] | CH, EE |\n | | |\n | server_certificate_type [RFC7250] | CH, EE |\n | | |\n | padding [RFC7685] | CH |\n | | |\n | key_share (RFC 8446) | CH, SH, HRR |\n | | |\n | pre_shared_key (RFC 8446) | CH, SH |\n | | |\n | psk_key_exchange_modes (RFC 8446) | CH |\n | | |\n | early_data (RFC 8446) | CH, EE, NST |\n | | |\n | cookie (RFC 8446) | CH, HRR |\n | | |\n | supported_versions (RFC 8446) | CH, SH, HRR |\n | | |\n | certificate_authorities (RFC 8446) | CH, CR |\n | | |\n | oid_filters (RFC 8446) | CR |\n | | |\n | post_handshake_auth (RFC 8446) | CH |\n | | |\n | signature_algorithms_cert (RFC 8446) | CH, CR |\n +--------------------------------------------------+-------------+\n\n\n\n\nRescorla Standards Track [Page 37]\n\f\nRFC 8446 TLS August 2018\n\n\n When multiple extensions of different types are present, the\n extensions MAY appear in any order, with the exception of\n \"pre_shared_key\" (Section 4.2.11) which MUST be the last extension in\n the ClientHello (but can appear anywhere in the ServerHello\n extensions block). There MUST NOT be more than one extension of the\n same type in a given extension block.\n\n In TLS 1.3, unlike TLS 1.2, extensions are negotiated for each\n handshake even when in resumption-PSK mode. However, 0-RTT\n parameters are those negotiated in the previous handshake; mismatches\n may require rejecting 0-RTT (see Section 4.2.10).\n\n There are subtle (and not so subtle) interactions that may occur in\n this protocol between new features and existing features which may\n result in a significant reduction in overall security. The following\n considerations should be taken into account when designing new\n extensions:\n\n - Some cases where a server does not agree to an extension are error\n conditions (e.g., the handshake cannot continue), and some are\n simply refusals to support particular features. In general, error\n alerts should be used for the former and a field in the server\n extension response for the latter.\n\n - Extensions should, as far as possible, be designed to prevent any\n attack that forces use (or non-use) of a particular feature by\n manipulation of handshake messages. This principle should be\n followed regardless of whether the feature is believed to cause a\n security problem. Often the fact that the extension fields are\n included in the inputs to the Finished message hashes will be\n sufficient, but extreme care is needed when the extension changes\n the meaning of messages sent in the handshake phase. Designers\n and implementors should be aware of the fact that until the\n handshake has been authenticated, active attackers can modify\n messages and insert, remove, or replace extensions.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 38]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.1. Supported Versions\n\n struct {\n select (Handshake.msg_type) {\n case client_hello:\n ProtocolVersion versions<2..254>;\n\n case server_hello: /* and HelloRetryRequest */\n ProtocolVersion selected_version;\n };\n } SupportedVersions;\n\n The \"supported_versions\" extension is used by the client to indicate\n which versions of TLS it supports and by the server to indicate which\n version it is using. The extension contains a list of supported\n versions in preference order, with the most preferred version first.\n Implementations of this specification MUST send this extension in the\n ClientHello containing all versions of TLS which they are prepared to\n negotiate (for this specification, that means minimally 0x0304, but\n if previous versions of TLS are allowed to be negotiated, they MUST\n be present as well).\n\n If this extension is not present, servers which are compliant with\n this specification and which also support TLS 1.2 MUST negotiate\n TLS 1.2 or prior as specified in [RFC5246], even if\n ClientHello.legacy_version is 0x0304 or later. Servers MAY abort the\n handshake upon receiving a ClientHello with legacy_version 0x0304 or\n later.\n\n If this extension is present in the ClientHello, servers MUST NOT use\n the ClientHello.legacy_version value for version negotiation and MUST\n use only the \"supported_versions\" extension to determine client\n preferences. Servers MUST only select a version of TLS present in\n that extension and MUST ignore any unknown versions that are present\n in that extension. Note that this mechanism makes it possible to\n negotiate a version prior to TLS 1.2 if one side supports a sparse\n range. Implementations of TLS 1.3 which choose to support prior\n versions of TLS SHOULD support TLS 1.2. Servers MUST be prepared to\n receive ClientHellos that include this extension but do not include\n 0x0304 in the list of versions.\n\n A server which negotiates a version of TLS prior to TLS 1.3 MUST set\n ServerHello.version and MUST NOT send the \"supported_versions\"\n extension. A server which negotiates TLS 1.3 MUST respond by sending\n a \"supported_versions\" extension containing the selected version\n value (0x0304). It MUST set the ServerHello.legacy_version field to\n 0x0303 (TLS 1.2). Clients MUST check for this extension prior to\n processing the rest of the ServerHello (although they will have to\n\n\n\nRescorla Standards Track [Page 39]\n\f\nRFC 8446 TLS August 2018\n\n\n parse the ServerHello in order to read the extension). If this\n extension is present, clients MUST ignore the\n ServerHello.legacy_version value and MUST use only the\n \"supported_versions\" extension to determine the selected version. If\n the \"supported_versions\" extension in the ServerHello contains a\n version not offered by the client or contains a version prior to\n TLS 1.3, the client MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n4.2.2. Cookie\n\n struct {\n opaque cookie<1..2^16-1>;\n } Cookie;\n\n Cookies serve two primary purposes:\n\n - Allowing the server to force the client to demonstrate\n reachability at their apparent network address (thus providing a\n measure of DoS protection). This is primarily useful for\n non-connection-oriented transports (see [RFC6347] for an example\n of this).\n\n - Allowing the server to offload state to the client, thus allowing\n it to send a HelloRetryRequest without storing any state. The\n server can do this by storing the hash of the ClientHello in the\n HelloRetryRequest cookie (protected with some suitable integrity\n protection algorithm).\n\n When sending a HelloRetryRequest, the server MAY provide a \"cookie\"\n extension to the client (this is an exception to the usual rule that\n the only extensions that may be sent are those that appear in the\n ClientHello). When sending the new ClientHello, the client MUST copy\n the contents of the extension received in the HelloRetryRequest into\n a \"cookie\" extension in the new ClientHello. Clients MUST NOT use\n cookies in their initial ClientHello in subsequent connections.\n\n When a server is operating statelessly, it may receive an unprotected\n record of type change_cipher_spec between the first and second\n ClientHello (see Section 5). Since the server is not storing any\n state, this will appear as if it were the first message to be\n received. Servers operating statelessly MUST ignore these records.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 40]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.3. Signature Algorithms\n\n TLS 1.3 provides two extensions for indicating which signature\n algorithms may be used in digital signatures. The\n \"signature_algorithms_cert\" extension applies to signatures in\n certificates, and the \"signature_algorithms\" extension, which\n originally appeared in TLS 1.2, applies to signatures in\n CertificateVerify messages. The keys found in certificates MUST also\n be of appropriate type for the signature algorithms they are used\n with. This is a particular issue for RSA keys and PSS signatures, as\n described below. If no \"signature_algorithms_cert\" extension is\n present, then the \"signature_algorithms\" extension also applies to\n signatures appearing in certificates. Clients which desire the\n server to authenticate itself via a certificate MUST send the\n \"signature_algorithms\" extension. If a server is authenticating via\n a certificate and the client has not sent a \"signature_algorithms\"\n extension, then the server MUST abort the handshake with a\n \"missing_extension\" alert (see Section 9.2).\n\n The \"signature_algorithms_cert\" extension was added to allow\n implementations which supported different sets of algorithms for\n certificates and in TLS itself to clearly signal their capabilities.\n TLS 1.2 implementations SHOULD also process this extension.\n Implementations which have the same policy in both cases MAY omit the\n \"signature_algorithms_cert\" extension.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 41]\n\f\nRFC 8446 TLS August 2018\n\n\n The \"extension_data\" field of these extensions contains a\n SignatureSchemeList value:\n\n enum {\n /* RSASSA-PKCS1-v1_5 algorithms */\n rsa_pkcs1_sha256(0x0401),\n rsa_pkcs1_sha384(0x0501),\n rsa_pkcs1_sha512(0x0601),\n\n /* ECDSA algorithms */\n ecdsa_secp256r1_sha256(0x0403),\n ecdsa_secp384r1_sha384(0x0503),\n ecdsa_secp521r1_sha512(0x0603),\n\n /* RSASSA-PSS algorithms with public key OID rsaEncryption */\n rsa_pss_rsae_sha256(0x0804),\n rsa_pss_rsae_sha384(0x0805),\n rsa_pss_rsae_sha512(0x0806),\n\n /* EdDSA algorithms */\n ed25519(0x0807),\n ed448(0x0808),\n\n /* RSASSA-PSS algorithms with public key OID RSASSA-PSS */\n rsa_pss_pss_sha256(0x0809),\n rsa_pss_pss_sha384(0x080a),\n rsa_pss_pss_sha512(0x080b),\n\n /* Legacy algorithms */\n rsa_pkcs1_sha1(0x0201),\n ecdsa_sha1(0x0203),\n\n /* Reserved Code Points */\n private_use(0xFE00..0xFFFF),\n (0xFFFF)\n } SignatureScheme;\n\n struct {\n SignatureScheme supported_signature_algorithms<2..2^16-2>;\n } SignatureSchemeList;\n\n Note: This enum is named \"SignatureScheme\" because there is already a\n \"SignatureAlgorithm\" type in TLS 1.2, which this replaces. We use\n the term \"signature algorithm\" throughout the text.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 42]\n\f\nRFC 8446 TLS August 2018\n\n\n Each SignatureScheme value lists a single signature algorithm that\n the client is willing to verify. The values are indicated in\n descending order of preference. Note that a signature algorithm\n takes as input an arbitrary-length message, rather than a digest.\n Algorithms which traditionally act on a digest should be defined in\n TLS to first hash the input with a specified hash algorithm and then\n proceed as usual. The code point groups listed above have the\n following meanings:\n\n RSASSA-PKCS1-v1_5 algorithms: Indicates a signature algorithm using\n RSASSA-PKCS1-v1_5 [RFC8017] with the corresponding hash algorithm\n as defined in [SHS]. These values refer solely to signatures\n which appear in certificates (see Section 4.4.2.2) and are not\n defined for use in signed TLS handshake messages, although they\n MAY appear in \"signature_algorithms\" and\n \"signature_algorithms_cert\" for backward compatibility with\n TLS 1.2.\n\n ECDSA algorithms: Indicates a signature algorithm using ECDSA\n [ECDSA], the corresponding curve as defined in ANSI X9.62 [ECDSA]\n and FIPS 186-4 [DSS], and the corresponding hash algorithm as\n defined in [SHS]. The signature is represented as a DER-encoded\n [X690] ECDSA-Sig-Value structure.\n\n RSASSA-PSS RSAE algorithms: Indicates a signature algorithm using\n RSASSA-PSS [RFC8017] with mask generation function 1. The digest\n used in the mask generation function and the digest being signed\n are both the corresponding hash algorithm as defined in [SHS].\n The length of the Salt MUST be equal to the length of the output\n of the digest algorithm. If the public key is carried in an X.509\n certificate, it MUST use the rsaEncryption OID [RFC5280].\n\n EdDSA algorithms: Indicates a signature algorithm using EdDSA as\n defined in [RFC8032] or its successors. Note that these\n correspond to the \"PureEdDSA\" algorithms and not the \"prehash\"\n variants.\n\n RSASSA-PSS PSS algorithms: Indicates a signature algorithm using\n RSASSA-PSS [RFC8017] with mask generation function 1. The digest\n used in the mask generation function and the digest being signed\n are both the corresponding hash algorithm as defined in [SHS].\n The length of the Salt MUST be equal to the length of the digest\n algorithm. If the public key is carried in an X.509 certificate,\n it MUST use the RSASSA-PSS OID [RFC5756]. When used in\n certificate signatures, the algorithm parameters MUST be DER\n encoded. If the corresponding public key's parameters are\n present, then the parameters in the signature MUST be identical to\n those in the public key.\n\n\n\nRescorla Standards Track [Page 43]\n\f\nRFC 8446 TLS August 2018\n\n\n Legacy algorithms: Indicates algorithms which are being deprecated\n because they use algorithms with known weaknesses, specifically\n SHA-1 which is used in this context with either (1) RSA using\n RSASSA-PKCS1-v1_5 or (2) ECDSA. These values refer solely to\n signatures which appear in certificates (see Section 4.4.2.2) and\n are not defined for use in signed TLS handshake messages, although\n they MAY appear in \"signature_algorithms\" and\n \"signature_algorithms_cert\" for backward compatibility with\n TLS 1.2. Endpoints SHOULD NOT negotiate these algorithms but are\n permitted to do so solely for backward compatibility. Clients\n offering these values MUST list them as the lowest priority\n (listed after all other algorithms in SignatureSchemeList).\n TLS 1.3 servers MUST NOT offer a SHA-1 signed certificate unless\n no valid certificate chain can be produced without it (see\n Section 4.4.2.2).\n\n The signatures on certificates that are self-signed or certificates\n that are trust anchors are not validated, since they begin a\n certification path (see [RFC5280], Section 3.2). A certificate that\n begins a certification path MAY use a signature algorithm that is not\n advertised as being supported in the \"signature_algorithms\"\n extension.\n\n Note that TLS 1.2 defines this extension differently. TLS 1.3\n implementations willing to negotiate TLS 1.2 MUST behave in\n accordance with the requirements of [RFC5246] when negotiating that\n version. In particular:\n\n - TLS 1.2 ClientHellos MAY omit this extension.\n\n - In TLS 1.2, the extension contained hash/signature pairs. The\n pairs are encoded in two octets, so SignatureScheme values have\n been allocated to align with TLS 1.2's encoding. Some legacy\n pairs are left unallocated. These algorithms are deprecated as of\n TLS 1.3. They MUST NOT be offered or negotiated by any\n implementation. In particular, MD5 [SLOTH], SHA-224, and DSA\n MUST NOT be used.\n\n - ECDSA signature schemes align with TLS 1.2's ECDSA hash/signature\n pairs. However, the old semantics did not constrain the signing\n curve. If TLS 1.2 is negotiated, implementations MUST be prepared\n to accept a signature that uses any curve that they advertised in\n the \"supported_groups\" extension.\n\n - Implementations that advertise support for RSASSA-PSS (which is\n mandatory in TLS 1.3) MUST be prepared to accept a signature using\n that scheme even when TLS 1.2 is negotiated. In TLS 1.2,\n RSASSA-PSS is used with RSA cipher suites.\n\n\n\nRescorla Standards Track [Page 44]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.4. Certificate Authorities\n\n The \"certificate_authorities\" extension is used to indicate the\n certificate authorities (CAs) which an endpoint supports and which\n SHOULD be used by the receiving endpoint to guide certificate\n selection.\n\n The body of the \"certificate_authorities\" extension consists of a\n CertificateAuthoritiesExtension structure.\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n DistinguishedName authorities<3..2^16-1>;\n } CertificateAuthoritiesExtension;\n\n authorities: A list of the distinguished names [X501] of acceptable\n certificate authorities, represented in DER-encoded [X690] format.\n These distinguished names specify a desired distinguished name for\n a trust anchor or subordinate CA; thus, this message can be used\n to describe known trust anchors as well as a desired authorization\n space.\n\n The client MAY send the \"certificate_authorities\" extension in the\n ClientHello message. The server MAY send it in the\n CertificateRequest message.\n\n The \"trusted_ca_keys\" extension [RFC6066], which serves a similar\n purpose but is more complicated, is not used in TLS 1.3 (although it\n may appear in ClientHello messages from clients which are offering\n prior versions of TLS).\n\n4.2.5. OID Filters\n\n The \"oid_filters\" extension allows servers to provide a set of\n OID/value pairs which it would like the client's certificate to\n match. This extension, if provided by the server, MUST only be sent\n in the CertificateRequest message.\n\n struct {\n opaque certificate_extension_oid<1..2^8-1>;\n opaque certificate_extension_values<0..2^16-1>;\n } OIDFilter;\n\n struct {\n OIDFilter filters<0..2^16-1>;\n } OIDFilterExtension;\n\n\n\n\nRescorla Standards Track [Page 45]\n\f\nRFC 8446 TLS August 2018\n\n\n filters: A list of certificate extension OIDs [RFC5280] with their\n allowed value(s) and represented in DER-encoded [X690] format.\n Some certificate extension OIDs allow multiple values (e.g.,\n Extended Key Usage). If the server has included a non-empty\n filters list, the client certificate included in the response MUST\n contain all of the specified extension OIDs that the client\n recognizes. For each extension OID recognized by the client, all\n of the specified values MUST be present in the client certificate\n (but the certificate MAY have other values as well). However, the\n client MUST ignore and skip any unrecognized certificate extension\n OIDs. If the client ignored some of the required certificate\n extension OIDs and supplied a certificate that does not satisfy\n the request, the server MAY at its discretion either continue the\n connection without client authentication or abort the handshake\n with an \"unsupported_certificate\" alert. Any given OID MUST NOT\n appear more than once in the filters list.\n\n PKIX RFCs define a variety of certificate extension OIDs and their\n corresponding value types. Depending on the type, matching\n certificate extension values are not necessarily bitwise-equal. It\n is expected that TLS implementations will rely on their PKI libraries\n to perform certificate selection using certificate extension OIDs.\n\n This document defines matching rules for two standard certificate\n extensions defined in [RFC5280]:\n\n - The Key Usage extension in a certificate matches the request when\n all key usage bits asserted in the request are also asserted in\n the Key Usage certificate extension.\n\n - The Extended Key Usage extension in a certificate matches the\n request when all key purpose OIDs present in the request are also\n found in the Extended Key Usage certificate extension. The\n special anyExtendedKeyUsage OID MUST NOT be used in the request.\n\n Separate specifications may define matching rules for other\n certificate extensions.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 46]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.6. Post-Handshake Client Authentication\n\n The \"post_handshake_auth\" extension is used to indicate that a client\n is willing to perform post-handshake authentication (Section 4.6.2).\n Servers MUST NOT send a post-handshake CertificateRequest to clients\n which do not offer this extension. Servers MUST NOT send this\n extension.\n\n struct {} PostHandshakeAuth;\n\n The \"extension_data\" field of the \"post_handshake_auth\" extension is\n zero length.\n\n4.2.7. Supported Groups\n\n When sent by the client, the \"supported_groups\" extension indicates\n the named groups which the client supports for key exchange, ordered\n from most preferred to least preferred.\n\n Note: In versions of TLS prior to TLS 1.3, this extension was named\n \"elliptic_curves\" and only contained elliptic curve groups. See\n [RFC8422] and [RFC7919]. This extension was also used to negotiate\n ECDSA curves. Signature algorithms are now negotiated independently\n (see Section 4.2.3).\n\n The \"extension_data\" field of this extension contains a\n \"NamedGroupList\" value:\n\n enum {\n\n /* Elliptic Curve Groups (ECDHE) */\n secp256r1(0x0017), secp384r1(0x0018), secp521r1(0x0019),\n x25519(0x001D), x448(0x001E),\n\n /* Finite Field Groups (DHE) */\n ffdhe2048(0x0100), ffdhe3072(0x0101), ffdhe4096(0x0102),\n ffdhe6144(0x0103), ffdhe8192(0x0104),\n\n /* Reserved Code Points */\n ffdhe_private_use(0x01FC..0x01FF),\n ecdhe_private_use(0xFE00..0xFEFF),\n (0xFFFF)\n } NamedGroup;\n\n struct {\n NamedGroup named_group_list<2..2^16-1>;\n } NamedGroupList;\n\n\n\n\nRescorla Standards Track [Page 47]\n\f\nRFC 8446 TLS August 2018\n\n\n Elliptic Curve Groups (ECDHE): Indicates support for the\n corresponding named curve, defined in either FIPS 186-4 [DSS] or\n [RFC7748]. Values 0xFE00 through 0xFEFF are reserved for\n Private Use [RFC8126].\n\n Finite Field Groups (DHE): Indicates support for the corresponding\n finite field group, defined in [RFC7919]. Values 0x01FC through\n 0x01FF are reserved for Private Use.\n\n Items in named_group_list are ordered according to the sender's\n preferences (most preferred choice first).\n\n As of TLS 1.3, servers are permitted to send the \"supported_groups\"\n extension to the client. Clients MUST NOT act upon any information\n found in \"supported_groups\" prior to successful completion of the\n handshake but MAY use the information learned from a successfully\n completed handshake to change what groups they use in their\n \"key_share\" extension in subsequent connections. If the server has a\n group it prefers to the ones in the \"key_share\" extension but is\n still willing to accept the ClientHello, it SHOULD send\n \"supported_groups\" to update the client's view of its preferences;\n this extension SHOULD contain all groups the server supports,\n regardless of whether they are currently supported by the client.\n\n4.2.8. Key Share\n\n The \"key_share\" extension contains the endpoint's cryptographic\n parameters.\n\n Clients MAY send an empty client_shares vector in order to request\n group selection from the server, at the cost of an additional round\n trip (see Section 4.1.4).\n\n struct {\n NamedGroup group;\n opaque key_exchange<1..2^16-1>;\n } KeyShareEntry;\n\n group: The named group for the key being exchanged.\n\n key_exchange: Key exchange information. The contents of this field\n are determined by the specified group and its corresponding\n definition. Finite Field Diffie-Hellman [DH76] parameters are\n described in Section 4.2.8.1; Elliptic Curve Diffie-Hellman\n parameters are described in Section 4.2.8.2.\n\n\n\n\n\n\nRescorla Standards Track [Page 48]\n\f\nRFC 8446 TLS August 2018\n\n\n In the ClientHello message, the \"extension_data\" field of this\n extension contains a \"KeyShareClientHello\" value:\n\n struct {\n KeyShareEntry client_shares<0..2^16-1>;\n } KeyShareClientHello;\n\n client_shares: A list of offered KeyShareEntry values in descending\n order of client preference.\n\n This vector MAY be empty if the client is requesting a\n HelloRetryRequest. Each KeyShareEntry value MUST correspond to a\n group offered in the \"supported_groups\" extension and MUST appear in\n the same order. However, the values MAY be a non-contiguous subset\n of the \"supported_groups\" extension and MAY omit the most preferred\n groups. Such a situation could arise if the most preferred groups\n are new and unlikely to be supported in enough places to make\n pregenerating key shares for them efficient.\n\n Clients can offer as many KeyShareEntry values as the number of\n supported groups it is offering, each representing a single set of\n key exchange parameters. For instance, a client might offer shares\n for several elliptic curves or multiple FFDHE groups. The\n key_exchange values for each KeyShareEntry MUST be generated\n independently. Clients MUST NOT offer multiple KeyShareEntry values\n for the same group. Clients MUST NOT offer any KeyShareEntry values\n for groups not listed in the client's \"supported_groups\" extension.\n Servers MAY check for violations of these rules and abort the\n handshake with an \"illegal_parameter\" alert if one is violated.\n\n In a HelloRetryRequest message, the \"extension_data\" field of this\n extension contains a KeyShareHelloRetryRequest value:\n\n struct {\n NamedGroup selected_group;\n } KeyShareHelloRetryRequest;\n\n selected_group: The mutually supported group the server intends to\n negotiate and is requesting a retried ClientHello/KeyShare for.\n\n Upon receipt of this extension in a HelloRetryRequest, the client\n MUST verify that (1) the selected_group field corresponds to a group\n which was provided in the \"supported_groups\" extension in the\n original ClientHello and (2) the selected_group field does not\n correspond to a group which was provided in the \"key_share\" extension\n in the original ClientHello. If either of these checks fails, then\n the client MUST abort the handshake with an \"illegal_parameter\"\n alert. Otherwise, when sending the new ClientHello, the client MUST\n\n\n\nRescorla Standards Track [Page 49]\n\f\nRFC 8446 TLS August 2018\n\n\n replace the original \"key_share\" extension with one containing only a\n new KeyShareEntry for the group indicated in the selected_group field\n of the triggering HelloRetryRequest.\n\n In a ServerHello message, the \"extension_data\" field of this\n extension contains a KeyShareServerHello value:\n\n struct {\n KeyShareEntry server_share;\n } KeyShareServerHello;\n\n server_share: A single KeyShareEntry value that is in the same group\n as one of the client's shares.\n\n If using (EC)DHE key establishment, servers offer exactly one\n KeyShareEntry in the ServerHello. This value MUST be in the same\n group as the KeyShareEntry value offered by the client that the\n server has selected for the negotiated key exchange. Servers\n MUST NOT send a KeyShareEntry for any group not indicated in the\n client's \"supported_groups\" extension and MUST NOT send a\n KeyShareEntry when using the \"psk_ke\" PskKeyExchangeMode. If using\n (EC)DHE key establishment and a HelloRetryRequest containing a\n \"key_share\" extension was received by the client, the client MUST\n verify that the selected NamedGroup in the ServerHello is the same as\n that in the HelloRetryRequest. If this check fails, the client MUST\n abort the handshake with an \"illegal_parameter\" alert.\n\n4.2.8.1. Diffie-Hellman Parameters\n\n Diffie-Hellman [DH76] parameters for both clients and servers are\n encoded in the opaque key_exchange field of a KeyShareEntry in a\n KeyShare structure. The opaque value contains the Diffie-Hellman\n public value (Y = g^X mod p) for the specified group (see [RFC7919]\n for group definitions) encoded as a big-endian integer and padded to\n the left with zeros to the size of p in bytes.\n\n Note: For a given Diffie-Hellman group, the padding results in all\n public keys having the same length.\n\n Peers MUST validate each other's public key Y by ensuring that 1 < Y\n < p-1. This check ensures that the remote peer is properly behaved\n and isn't forcing the local system into a small subgroup.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 50]\n\f\nRFC 8446 TLS August 2018\n\n\n4.2.8.2. ECDHE Parameters\n\n ECDHE parameters for both clients and servers are encoded in the\n opaque key_exchange field of a KeyShareEntry in a KeyShare structure.\n\n For secp256r1, secp384r1, and secp521r1, the contents are the\n serialized value of the following struct:\n\n struct {\n uint8 legacy_form = 4;\n opaque X[coordinate_length];\n opaque Y[coordinate_length];\n } UncompressedPointRepresentation;\n\n X and Y, respectively, are the binary representations of the x and y\n values in network byte order. There are no internal length markers,\n so each number representation occupies as many octets as implied by\n the curve parameters. For P-256, this means that each of X and Y use\n 32 octets, padded on the left by zeros if necessary. For P-384, they\n take 48 octets each. For P-521, they take 66 octets each.\n\n For the curves secp256r1, secp384r1, and secp521r1, peers MUST\n validate each other's public value Q by ensuring that the point is a\n valid point on the elliptic curve. The appropriate validation\n procedures are defined in Section 4.3.7 of [ECDSA] and alternatively\n in Section 5.6.2.3 of [KEYAGREEMENT]. This process consists of three\n steps: (1) verify that Q is not the point at infinity (O), (2) verify\n that for Q = (x, y) both integers x and y are in the correct\n interval, and (3) ensure that (x, y) is a correct solution to the\n elliptic curve equation. For these curves, implementors do not need\n to verify membership in the correct subgroup.\n\n For X25519 and X448, the contents of the public value are the byte\n string inputs and outputs of the corresponding functions defined in\n [RFC7748]: 32 bytes for X25519 and 56 bytes for X448.\n\n Note: Versions of TLS prior to 1.3 permitted point format\n negotiation; TLS 1.3 removes this feature in favor of a single point\n format for each curve.\n\n4.2.9. Pre-Shared Key Exchange Modes\n\n In order to use PSKs, clients MUST also send a\n \"psk_key_exchange_modes\" extension. The semantics of this extension\n are that the client only supports the use of PSKs with these modes,\n which restricts both the use of PSKs offered in this ClientHello and\n those which the server might supply via NewSessionTicket.\n\n\n\n\nRescorla Standards Track [Page 51]\n\f\nRFC 8446 TLS August 2018\n\n\n A client MUST provide a \"psk_key_exchange_modes\" extension if it\n offers a \"pre_shared_key\" extension. If clients offer\n \"pre_shared_key\" without a \"psk_key_exchange_modes\" extension,\n servers MUST abort the handshake. Servers MUST NOT select a key\n exchange mode that is not listed by the client. This extension also\n restricts the modes for use with PSK resumption. Servers SHOULD NOT\n send NewSessionTicket with tickets that are not compatible with the\n advertised modes; however, if a server does so, the impact will just\n be that the client's attempts at resumption fail.\n\n The server MUST NOT send a \"psk_key_exchange_modes\" extension.\n\n enum { psk_ke(0), psk_dhe_ke(1), (255) } PskKeyExchangeMode;\n\n struct {\n PskKeyExchangeMode ke_modes<1..255>;\n } PskKeyExchangeModes;\n\n psk_ke: PSK-only key establishment. In this mode, the server\n MUST NOT supply a \"key_share\" value.\n\n psk_dhe_ke: PSK with (EC)DHE key establishment. In this mode, the\n client and server MUST supply \"key_share\" values as described in\n Section 4.2.8.\n\n Any future values that are allocated must ensure that the transmitted\n protocol messages unambiguously identify which mode was selected by\n the server; at present, this is indicated by the presence of the\n \"key_share\" in the ServerHello.\n\n4.2.10. Early Data Indication\n\n When a PSK is used and early data is allowed for that PSK, the client\n can send Application Data in its first flight of messages. If the\n client opts to do so, it MUST supply both the \"pre_shared_key\" and\n \"early_data\" extensions.\n\n The \"extension_data\" field of this extension contains an\n \"EarlyDataIndication\" value.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 52]\n\f\nRFC 8446 TLS August 2018\n\n\n struct {} Empty;\n\n struct {\n select (Handshake.msg_type) {\n case new_session_ticket: uint32 max_early_data_size;\n case client_hello: Empty;\n case encrypted_extensions: Empty;\n };\n } EarlyDataIndication;\n\n See Section 4.6.1 for details regarding the use of the\n max_early_data_size field.\n\n The parameters for the 0-RTT data (version, symmetric cipher suite,\n Application-Layer Protocol Negotiation (ALPN) [RFC7301] protocol,\n etc.) are those associated with the PSK in use. For externally\n provisioned PSKs, the associated values are those provisioned along\n with the key. For PSKs established via a NewSessionTicket message,\n the associated values are those which were negotiated in the\n connection which established the PSK. The PSK used to encrypt the\n early data MUST be the first PSK listed in the client's\n \"pre_shared_key\" extension.\n\n For PSKs provisioned via NewSessionTicket, a server MUST validate\n that the ticket age for the selected PSK identity (computed by\n subtracting ticket_age_add from PskIdentity.obfuscated_ticket_age\n modulo 2^32) is within a small tolerance of the time since the ticket\n was issued (see Section 8). If it is not, the server SHOULD proceed\n with the handshake but reject 0-RTT, and SHOULD NOT take any other\n action that assumes that this ClientHello is fresh.\n\n 0-RTT messages sent in the first flight have the same (encrypted)\n content types as messages of the same type sent in other flights\n (handshake and application_data) but are protected under different\n keys. After receiving the server's Finished message, if the server\n has accepted early data, an EndOfEarlyData message will be sent to\n indicate the key change. This message will be encrypted with the\n 0-RTT traffic keys.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 53]\n\f\nRFC 8446 TLS August 2018\n\n\n A server which receives an \"early_data\" extension MUST behave in one\n of three ways:\n\n - Ignore the extension and return a regular 1-RTT response. The\n server then skips past early data by attempting to deprotect\n received records using the handshake traffic key, discarding\n records which fail deprotection (up to the configured\n max_early_data_size). Once a record is deprotected successfully,\n it is treated as the start of the client's second flight and the\n server proceeds as with an ordinary 1-RTT handshake.\n\n - Request that the client send another ClientHello by responding\n with a HelloRetryRequest. A client MUST NOT include the\n \"early_data\" extension in its followup ClientHello. The server\n then ignores early data by skipping all records with an external\n content type of \"application_data\" (indicating that they are\n encrypted), up to the configured max_early_data_size.\n\n - Return its own \"early_data\" extension in EncryptedExtensions,\n indicating that it intends to process the early data. It is not\n possible for the server to accept only a subset of the early data\n messages. Even though the server sends a message accepting early\n data, the actual early data itself may already be in flight by the\n time the server generates this message.\n\n In order to accept early data, the server MUST have accepted a PSK\n cipher suite and selected the first key offered in the client's\n \"pre_shared_key\" extension. In addition, it MUST verify that the\n following values are the same as those associated with the\n selected PSK:\n\n - The TLS version number\n\n - The selected cipher suite\n\n - The selected ALPN [RFC7301] protocol, if any\n\n These requirements are a superset of those needed to perform a 1-RTT\n handshake using the PSK in question. For externally established\n PSKs, the associated values are those provisioned along with the key.\n For PSKs established via a NewSessionTicket message, the associated\n values are those negotiated in the connection during which the ticket\n was established.\n\n Future extensions MUST define their interaction with 0-RTT.\n\n\n\n\n\n\nRescorla Standards Track [Page 54]\n\f\nRFC 8446 TLS August 2018\n\n\n If any of these checks fail, the server MUST NOT respond with the\n extension and must discard all the first-flight data using one of the\n first two mechanisms listed above (thus falling back to 1-RTT or\n 2-RTT). If the client attempts a 0-RTT handshake but the server\n rejects it, the server will generally not have the 0-RTT record\n protection keys and must instead use trial decryption (either with\n the 1-RTT handshake keys or by looking for a cleartext ClientHello in\n the case of a HelloRetryRequest) to find the first non-0-RTT message.\n\n If the server chooses to accept the \"early_data\" extension, then it\n MUST comply with the same error-handling requirements specified for\n all records when processing early data records. Specifically, if the\n server fails to decrypt a 0-RTT record following an accepted\n \"early_data\" extension, it MUST terminate the connection with a\n \"bad_record_mac\" alert as per Section 5.2.\n\n If the server rejects the \"early_data\" extension, the client\n application MAY opt to retransmit the Application Data previously\n sent in early data once the handshake has been completed. Note that\n automatic retransmission of early data could result in incorrect\n assumptions regarding the status of the connection. For instance,\n when the negotiated connection selects a different ALPN protocol from\n what was used for the early data, an application might need to\n construct different messages. Similarly, if early data assumes\n anything about the connection state, it might be sent in error after\n the handshake completes.\n\n A TLS implementation SHOULD NOT automatically resend early data;\n applications are in a better position to decide when retransmission\n is appropriate. A TLS implementation MUST NOT automatically resend\n early data unless the negotiated connection selects the same ALPN\n protocol.\n\n4.2.11. Pre-Shared Key Extension\n\n The \"pre_shared_key\" extension is used to negotiate the identity of\n the pre-shared key to be used with a given handshake in association\n with PSK key establishment.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 55]\n\f\nRFC 8446 TLS August 2018\n\n\n The \"extension_data\" field of this extension contains a\n \"PreSharedKeyExtension\" value:\n\n struct {\n opaque identity<1..2^16-1>;\n uint32 obfuscated_ticket_age;\n } PskIdentity;\n\n opaque PskBinderEntry<32..255>;\n\n struct {\n PskIdentity identities<7..2^16-1>;\n PskBinderEntry binders<33..2^16-1>;\n } OfferedPsks;\n\n struct {\n select (Handshake.msg_type) {\n case client_hello: OfferedPsks;\n case server_hello: uint16 selected_identity;\n };\n } PreSharedKeyExtension;\n\n identity: A label for a key. For instance, a ticket (as defined in\n Appendix B.3.4) or a label for a pre-shared key established\n externally.\n\n obfuscated_ticket_age: An obfuscated version of the age of the key.\n Section 4.2.11.1 describes how to form this value for identities\n established via the NewSessionTicket message. For identities\n established externally, an obfuscated_ticket_age of 0 SHOULD be\n used, and servers MUST ignore the value.\n\n identities: A list of the identities that the client is willing to\n negotiate with the server. If sent alongside the \"early_data\"\n extension (see Section 4.2.10), the first identity is the one used\n for 0-RTT data.\n\n binders: A series of HMAC values, one for each value in the\n identities list and in the same order, computed as described\n below.\n\n selected_identity: The server's chosen identity expressed as a\n (0-based) index into the identities in the client's list.\n\n Each PSK is associated with a single Hash algorithm. For PSKs\n established via the ticket mechanism (Section 4.6.1), this is the KDF\n Hash algorithm on the connection where the ticket was established.\n For externally established PSKs, the Hash algorithm MUST be set when\n\n\n\nRescorla Standards Track [Page 56]\n\f\nRFC 8446 TLS August 2018\n\n\n the PSK is established or default to SHA-256 if no such algorithm is\n defined. The server MUST ensure that it selects a compatible PSK\n (if any) and cipher suite.\n\n In TLS versions prior to TLS 1.3, the Server Name Identification\n (SNI) value was intended to be associated with the session (Section 3\n of [RFC6066]), with the server being required to enforce that the SNI\n value associated with the session matches the one specified in the\n resumption handshake. However, in reality the implementations were\n not consistent on which of two supplied SNI values they would use,\n leading to the consistency requirement being de facto enforced by the\n clients. In TLS 1.3, the SNI value is always explicitly specified in\n the resumption handshake, and there is no need for the server to\n associate an SNI value with the ticket. Clients, however, SHOULD\n store the SNI with the PSK to fulfill the requirements of\n Section 4.6.1.\n\n Implementor's note: When session resumption is the primary use case\n of PSKs, the most straightforward way to implement the PSK/cipher\n suite matching requirements is to negotiate the cipher suite first\n and then exclude any incompatible PSKs. Any unknown PSKs (e.g., ones\n not in the PSK database or encrypted with an unknown key) SHOULD\n simply be ignored. If no acceptable PSKs are found, the server\n SHOULD perform a non-PSK handshake if possible. If backward\n compatibility is important, client-provided, externally established\n PSKs SHOULD influence cipher suite selection.\n\n Prior to accepting PSK key establishment, the server MUST validate\n the corresponding binder value (see Section 4.2.11.2 below). If this\n value is not present or does not validate, the server MUST abort the\n handshake. Servers SHOULD NOT attempt to validate multiple binders;\n rather, they SHOULD select a single PSK and validate solely the\n binder that corresponds to that PSK. See Section 8.2 and\n Appendix E.6 for the security rationale for this requirement. In\n order to accept PSK key establishment, the server sends a\n \"pre_shared_key\" extension indicating the selected identity.\n\n Clients MUST verify that the server's selected_identity is within the\n range supplied by the client, that the server selected a cipher suite\n indicating a Hash associated with the PSK, and that a server\n \"key_share\" extension is present if required by the ClientHello\n \"psk_key_exchange_modes\" extension. If these values are not\n consistent, the client MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 57]\n\f\nRFC 8446 TLS August 2018\n\n\n If the server supplies an \"early_data\" extension, the client MUST\n verify that the server's selected_identity is 0. If any other value\n is returned, the client MUST abort the handshake with an\n \"illegal_parameter\" alert.\n\n The \"pre_shared_key\" extension MUST be the last extension in the\n ClientHello (this facilitates implementation as described below).\n Servers MUST check that it is the last extension and otherwise fail\n the handshake with an \"illegal_parameter\" alert.\n\n4.2.11.1. Ticket Age\n\n The client's view of the age of a ticket is the time since the\n receipt of the NewSessionTicket message. Clients MUST NOT attempt to\n use tickets which have ages greater than the \"ticket_lifetime\" value\n which was provided with the ticket. The \"obfuscated_ticket_age\"\n field of each PskIdentity contains an obfuscated version of the\n ticket age formed by taking the age in milliseconds and adding the\n \"ticket_age_add\" value that was included with the ticket (see\n Section 4.6.1), modulo 2^32. This addition prevents passive\n observers from correlating connections unless tickets are reused.\n Note that the \"ticket_lifetime\" field in the NewSessionTicket message\n is in seconds but the \"obfuscated_ticket_age\" is in milliseconds.\n Because ticket lifetimes are restricted to a week, 32 bits is enough\n to represent any plausible age, even in milliseconds.\n\n4.2.11.2. PSK Binder\n\n The PSK binder value forms a binding between a PSK and the current\n handshake, as well as a binding between the handshake in which the\n PSK was generated (if via a NewSessionTicket message) and the current\n handshake. Each entry in the binders list is computed as an HMAC\n over a transcript hash (see Section 4.4.1) containing a partial\n ClientHello up to and including the PreSharedKeyExtension.identities\n field. That is, it includes all of the ClientHello but not the\n binders list itself. The length fields for the message (including\n the overall length, the length of the extensions block, and the\n length of the \"pre_shared_key\" extension) are all set as if binders\n of the correct lengths were present.\n\n The PskBinderEntry is computed in the same way as the Finished\n message (Section 4.4.4) but with the BaseKey being the binder_key\n derived via the key schedule from the corresponding PSK which is\n being offered (see Section 7.1).\n\n\n\n\n\n\n\nRescorla Standards Track [Page 58]\n\f\nRFC 8446 TLS August 2018\n\n\n If the handshake includes a HelloRetryRequest, the initial\n ClientHello and HelloRetryRequest are included in the transcript\n along with the new ClientHello. For instance, if the client sends\n ClientHello1, its binder will be computed over:\n\n Transcript-Hash(Truncate(ClientHello1))\n\n Where Truncate() removes the binders list from the ClientHello.\n\n If the server responds with a HelloRetryRequest and the client then\n sends ClientHello2, its binder will be computed over:\n\n Transcript-Hash(ClientHello1,\n HelloRetryRequest,\n Truncate(ClientHello2))\n\n The full ClientHello1/ClientHello2 is included in all other handshake\n hash computations. Note that in the first flight,\n Truncate(ClientHello1) is hashed directly, but in the second flight,\n ClientHello1 is hashed and then reinjected as a \"message_hash\"\n message, as described in Section 4.4.1.\n\n4.2.11.3. Processing Order\n\n Clients are permitted to \"stream\" 0-RTT data until they receive the\n server's Finished, only then sending the EndOfEarlyData message,\n followed by the rest of the handshake. In order to avoid deadlocks,\n when accepting \"early_data\", servers MUST process the client's\n ClientHello and then immediately send their flight of messages,\n rather than waiting for the client's EndOfEarlyData message before\n sending its ServerHello.\n\n4.3. Server Parameters\n\n The next two messages from the server, EncryptedExtensions and\n CertificateRequest, contain information from the server that\n determines the rest of the handshake. These messages are encrypted\n with keys derived from the server_handshake_traffic_secret.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 59]\n\f\nRFC 8446 TLS August 2018\n\n\n4.3.1. Encrypted Extensions\n\n In all handshakes, the server MUST send the EncryptedExtensions\n message immediately after the ServerHello message. This is the first\n message that is encrypted under keys derived from the\n server_handshake_traffic_secret.\n\n The EncryptedExtensions message contains extensions that can be\n protected, i.e., any which are not needed to establish the\n cryptographic context but which are not associated with individual\n certificates. The client MUST check EncryptedExtensions for the\n presence of any forbidden extensions and if any are found MUST abort\n the handshake with an \"illegal_parameter\" alert.\n\n Structure of this message:\n\n struct {\n Extension extensions<0..2^16-1>;\n } EncryptedExtensions;\n\n extensions: A list of extensions. For more information, see the\n table in Section 4.2.\n\n4.3.2. Certificate Request\n\n A server which is authenticating with a certificate MAY optionally\n request a certificate from the client. This message, if sent, MUST\n follow EncryptedExtensions.\n\n Structure of this message:\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n Extension extensions<2..2^16-1>;\n } CertificateRequest;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 60]\n\f\nRFC 8446 TLS August 2018\n\n\n certificate_request_context: An opaque string which identifies the\n certificate request and which will be echoed in the client's\n Certificate message. The certificate_request_context MUST be\n unique within the scope of this connection (thus preventing replay\n of client CertificateVerify messages). This field SHALL be zero\n length unless used for the post-handshake authentication exchanges\n described in Section 4.6.2. When requesting post-handshake\n authentication, the server SHOULD make the context unpredictable\n to the client (e.g., by randomly generating it) in order to\n prevent an attacker who has temporary access to the client's\n private key from pre-computing valid CertificateVerify messages.\n\n extensions: A set of extensions describing the parameters of the\n certificate being requested. The \"signature_algorithms\" extension\n MUST be specified, and other extensions may optionally be included\n if defined for this message. Clients MUST ignore unrecognized\n extensions.\n\n In prior versions of TLS, the CertificateRequest message carried a\n list of signature algorithms and certificate authorities which the\n server would accept. In TLS 1.3, the former is expressed by sending\n the \"signature_algorithms\" and optionally \"signature_algorithms_cert\"\n extensions. The latter is expressed by sending the\n \"certificate_authorities\" extension (see Section 4.2.4).\n\n Servers which are authenticating with a PSK MUST NOT send the\n CertificateRequest message in the main handshake, though they MAY\n send it in post-handshake authentication (see Section 4.6.2) provided\n that the client has sent the \"post_handshake_auth\" extension (see\n Section 4.2.6).\n\n4.4. Authentication Messages\n\n As discussed in Section 2, TLS generally uses a common set of\n messages for authentication, key confirmation, and handshake\n integrity: Certificate, CertificateVerify, and Finished. (The PSK\n binders also perform key confirmation, in a similar fashion.) These\n three messages are always sent as the last messages in their\n handshake flight. The Certificate and CertificateVerify messages are\n only sent under certain circumstances, as defined below. The\n Finished message is always sent as part of the Authentication Block.\n These messages are encrypted under keys derived from the\n [sender]_handshake_traffic_secret.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 61]\n\f\nRFC 8446 TLS August 2018\n\n\n The computations for the Authentication messages all uniformly take\n the following inputs:\n\n - The certificate and signing key to be used.\n\n - A Handshake Context consisting of the set of messages to be\n included in the transcript hash.\n\n - A Base Key to be used to compute a MAC key.\n\n Based on these inputs, the messages then contain:\n\n Certificate: The certificate to be used for authentication, and any\n supporting certificates in the chain. Note that certificate-based\n client authentication is not available in PSK handshake flows\n (including 0-RTT).\n\n CertificateVerify: A signature over the value\n Transcript-Hash(Handshake Context, Certificate).\n\n Finished: A MAC over the value Transcript-Hash(Handshake Context,\n Certificate, CertificateVerify) using a MAC key derived from the\n Base Key.\n\n The following table defines the Handshake Context and MAC Base Key\n for each scenario:\n\n +-----------+-------------------------+-----------------------------+\n | Mode | Handshake Context | Base Key |\n +-----------+-------------------------+-----------------------------+\n | Server | ClientHello ... later | server_handshake_traffic_ |\n | | of EncryptedExtensions/ | secret |\n | | CertificateRequest | |\n | | | |\n | Client | ClientHello ... later | client_handshake_traffic_ |\n | | of server | secret |\n | | Finished/EndOfEarlyData | |\n | | | |\n | Post- | ClientHello ... client | client_application_traffic_ |\n | Handshake | Finished + | secret_N |\n | | CertificateRequest | |\n +-----------+-------------------------+-----------------------------+\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 62]\n\f\nRFC 8446 TLS August 2018\n\n\n4.4.1. The Transcript Hash\n\n Many of the cryptographic computations in TLS make use of a\n transcript hash. This value is computed by hashing the concatenation\n of each included handshake message, including the handshake message\n header carrying the handshake message type and length fields, but not\n including record layer headers. I.e.,\n\n Transcript-Hash(M1, M2, ... Mn) = Hash(M1 || M2 || ... || Mn)\n\n As an exception to this general rule, when the server responds to a\n ClientHello with a HelloRetryRequest, the value of ClientHello1 is\n replaced with a special synthetic handshake message of handshake type\n \"message_hash\" containing Hash(ClientHello1). I.e.,\n\n Transcript-Hash(ClientHello1, HelloRetryRequest, ... Mn) =\n Hash(message_hash || /* Handshake type */\n 00 00 Hash.length || /* Handshake message length (bytes) */\n Hash(ClientHello1) || /* Hash of ClientHello1 */\n HelloRetryRequest || ... || Mn)\n\n The reason for this construction is to allow the server to do a\n stateless HelloRetryRequest by storing just the hash of ClientHello1\n in the cookie, rather than requiring it to export the entire\n intermediate hash state (see Section 4.2.2).\n\n For concreteness, the transcript hash is always taken from the\n following sequence of handshake messages, starting at the first\n ClientHello and including only those messages that were sent:\n ClientHello, HelloRetryRequest, ClientHello, ServerHello,\n EncryptedExtensions, server CertificateRequest, server Certificate,\n server CertificateVerify, server Finished, EndOfEarlyData, client\n Certificate, client CertificateVerify, client Finished.\n\n In general, implementations can implement the transcript by keeping a\n running transcript hash value based on the negotiated hash. Note,\n however, that subsequent post-handshake authentications do not\n include each other, just the messages through the end of the main\n handshake.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 63]\n\f\nRFC 8446 TLS August 2018\n\n\n4.4.2. Certificate\n\n This message conveys the endpoint's certificate chain to the peer.\n\n The server MUST send a Certificate message whenever the agreed-upon\n key exchange method uses certificates for authentication (this\n includes all key exchange methods defined in this document\n except PSK).\n\n The client MUST send a Certificate message if and only if the server\n has requested client authentication via a CertificateRequest message\n (Section 4.3.2). If the server requests client authentication but no\n suitable certificate is available, the client MUST send a Certificate\n message containing no certificates (i.e., with the \"certificate_list\"\n field having length 0). A Finished message MUST be sent regardless\n of whether the Certificate message is empty.\n\n Structure of this message:\n\n enum {\n X509(0),\n RawPublicKey(2),\n (255)\n } CertificateType;\n\n struct {\n select (certificate_type) {\n case RawPublicKey:\n /* From RFC 7250 ASN.1_subjectPublicKeyInfo */\n opaque ASN1_subjectPublicKeyInfo<1..2^24-1>;\n\n case X509:\n opaque cert_data<1..2^24-1>;\n };\n Extension extensions<0..2^16-1>;\n } CertificateEntry;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n CertificateEntry certificate_list<0..2^24-1>;\n } Certificate;\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 64]\n\f\nRFC 8446 TLS August 2018\n\n\n certificate_request_context: If this message is in response to a\n CertificateRequest, the value of certificate_request_context in\n that message. Otherwise (in the case of server authentication),\n this field SHALL be zero length.\n\n certificate_list: A sequence (chain) of CertificateEntry structures,\n each containing a single certificate and set of extensions.\n\n extensions: A set of extension values for the CertificateEntry. The\n \"Extension\" format is defined in Section 4.2. Valid extensions\n for server certificates at present include the OCSP Status\n extension [RFC6066] and the SignedCertificateTimestamp extension\n [RFC6962]; future extensions may be defined for this message as\n well. Extensions in the Certificate message from the server MUST\n correspond to ones from the ClientHello message. Extensions in\n the Certificate message from the client MUST correspond to\n extensions in the CertificateRequest message from the server. If\n an extension applies to the entire chain, it SHOULD be included in\n the first CertificateEntry.\n\n If the corresponding certificate type extension\n (\"server_certificate_type\" or \"client_certificate_type\") was not\n negotiated in EncryptedExtensions, or the X.509 certificate type was\n negotiated, then each CertificateEntry contains a DER-encoded X.509\n certificate. The sender's certificate MUST come in the first\n CertificateEntry in the list. Each following certificate SHOULD\n directly certify the one immediately preceding it. Because\n certificate validation requires that trust anchors be distributed\n independently, a certificate that specifies a trust anchor MAY be\n omitted from the chain, provided that supported peers are known to\n possess any omitted certificates.\n\n Note: Prior to TLS 1.3, \"certificate_list\" ordering required each\n certificate to certify the one immediately preceding it; however,\n some implementations allowed some flexibility. Servers sometimes\n send both a current and deprecated intermediate for transitional\n purposes, and others are simply configured incorrectly, but these\n cases can nonetheless be validated properly. For maximum\n compatibility, all implementations SHOULD be prepared to handle\n potentially extraneous certificates and arbitrary orderings from any\n TLS version, with the exception of the end-entity certificate which\n MUST be first.\n\n If the RawPublicKey certificate type was negotiated, then the\n certificate_list MUST contain no more than one CertificateEntry,\n which contains an ASN1_subjectPublicKeyInfo value as defined in\n [RFC7250], Section 3.\n\n\n\n\nRescorla Standards Track [Page 65]\n\f\nRFC 8446 TLS August 2018\n\n\n The OpenPGP certificate type [RFC6091] MUST NOT be used with TLS 1.3.\n\n The server's certificate_list MUST always be non-empty. A client\n will send an empty certificate_list if it does not have an\n appropriate certificate to send in response to the server's\n authentication request.\n\n4.4.2.1. OCSP Status and SCT Extensions\n\n [RFC6066] and [RFC6961] provide extensions to negotiate the server\n sending OCSP responses to the client. In TLS 1.2 and below, the\n server replies with an empty extension to indicate negotiation of\n this extension and the OCSP information is carried in a\n CertificateStatus message. In TLS 1.3, the server's OCSP information\n is carried in an extension in the CertificateEntry containing the\n associated certificate. Specifically, the body of the\n \"status_request\" extension from the server MUST be a\n CertificateStatus structure as defined in [RFC6066], which is\n interpreted as defined in [RFC6960].\n\n Note: The status_request_v2 extension [RFC6961] is deprecated.\n TLS 1.3 servers MUST NOT act upon its presence or information in it\n when processing ClientHello messages; in particular, they MUST NOT\n send the status_request_v2 extension in the EncryptedExtensions,\n CertificateRequest, or Certificate messages. TLS 1.3 servers MUST be\n able to process ClientHello messages that include it, as it MAY be\n sent by clients that wish to use it in earlier protocol versions.\n\n A server MAY request that a client present an OCSP response with its\n certificate by sending an empty \"status_request\" extension in its\n CertificateRequest message. If the client opts to send an OCSP\n response, the body of its \"status_request\" extension MUST be a\n CertificateStatus structure as defined in [RFC6066].\n\n Similarly, [RFC6962] provides a mechanism for a server to send a\n Signed Certificate Timestamp (SCT) as an extension in the ServerHello\n in TLS 1.2 and below. In TLS 1.3, the server's SCT information is\n carried in an extension in the CertificateEntry.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 66]\n\f\nRFC 8446 TLS August 2018\n\n\n4.4.2.2. Server Certificate Selection\n\n The following rules apply to the certificates sent by the server:\n\n - The certificate type MUST be X.509v3 [RFC5280], unless explicitly\n negotiated otherwise (e.g., [RFC7250]).\n\n - The server's end-entity certificate's public key (and associated\n restrictions) MUST be compatible with the selected authentication\n algorithm from the client's \"signature_algorithms\" extension\n (currently RSA, ECDSA, or EdDSA).\n\n - The certificate MUST allow the key to be used for signing (i.e.,\n the digitalSignature bit MUST be set if the Key Usage extension is\n present) with a signature scheme indicated in the client's\n \"signature_algorithms\"/\"signature_algorithms_cert\" extensions (see\n Section 4.2.3).\n\n - The \"server_name\" [RFC6066] and \"certificate_authorities\"\n extensions are used to guide certificate selection. As servers\n MAY require the presence of the \"server_name\" extension, clients\n SHOULD send this extension, when applicable.\n\n All certificates provided by the server MUST be signed by a signature\n algorithm advertised by the client if it is able to provide such a\n chain (see Section 4.2.3). Certificates that are self-signed or\n certificates that are expected to be trust anchors are not validated\n as part of the chain and therefore MAY be signed with any algorithm.\n\n If the server cannot produce a certificate chain that is signed only\n via the indicated supported algorithms, then it SHOULD continue the\n handshake by sending the client a certificate chain of its choice\n that may include algorithms that are not known to be supported by the\n client. This fallback chain SHOULD NOT use the deprecated SHA-1 hash\n algorithm in general, but MAY do so if the client's advertisement\n permits it, and MUST NOT do so otherwise.\n\n If the client cannot construct an acceptable chain using the provided\n certificates and decides to abort the handshake, then it MUST abort\n the handshake with an appropriate certificate-related alert (by\n default, \"unsupported_certificate\"; see Section 6.2 for more\n information).\n\n If the server has multiple certificates, it chooses one of them based\n on the above-mentioned criteria (in addition to other criteria, such\n as transport-layer endpoint, local configuration, and preferences).\n\n\n\n\n\nRescorla Standards Track [Page 67]\n\f\nRFC 8446 TLS August 2018\n\n\n4.4.2.3. Client Certificate Selection\n\n The following rules apply to certificates sent by the client:\n\n - The certificate type MUST be X.509v3 [RFC5280], unless explicitly\n negotiated otherwise (e.g., [RFC7250]).\n\n - If the \"certificate_authorities\" extension in the\n CertificateRequest message was present, at least one of the\n certificates in the certificate chain SHOULD be issued by one of\n the listed CAs.\n\n - The certificates MUST be signed using an acceptable signature\n algorithm, as described in Section 4.3.2. Note that this relaxes\n the constraints on certificate-signing algorithms found in prior\n versions of TLS.\n\n - If the CertificateRequest message contained a non-empty\n \"oid_filters\" extension, the end-entity certificate MUST match the\n extension OIDs that are recognized by the client, as described in\n Section 4.2.5.\n\n4.4.2.4. Receiving a Certificate Message\n\n In general, detailed certificate validation procedures are out of\n scope for TLS (see [RFC5280]). This section provides TLS-specific\n requirements.\n\n If the server supplies an empty Certificate message, the client MUST\n abort the handshake with a \"decode_error\" alert.\n\n If the client does not send any certificates (i.e., it sends an empty\n Certificate message), the server MAY at its discretion either\n continue the handshake without client authentication or abort the\n handshake with a \"certificate_required\" alert. Also, if some aspect\n of the certificate chain was unacceptable (e.g., it was not signed by\n a known, trusted CA), the server MAY at its discretion either\n continue the handshake (considering the client unauthenticated) or\n abort the handshake.\n\n Any endpoint receiving any certificate which it would need to\n validate using any signature algorithm using an MD5 hash MUST abort\n the handshake with a \"bad_certificate\" alert. SHA-1 is deprecated,\n and it is RECOMMENDED that any endpoint receiving any certificate\n which it would need to validate using any signature algorithm using a\n SHA-1 hash abort the handshake with a \"bad_certificate\" alert. For\n clarity, this means that endpoints can accept these algorithms for\n certificates that are self-signed or are trust anchors.\n\n\n\nRescorla Standards Track [Page 68]\n\f\nRFC 8446 TLS August 2018\n\n\n All endpoints are RECOMMENDED to transition to SHA-256 or better as\n soon as possible to maintain interoperability with implementations\n currently in the process of phasing out SHA-1 support.\n\n Note that a certificate containing a key for one signature algorithm\n MAY be signed using a different signature algorithm (for instance, an\n RSA key signed with an ECDSA key).\n\n4.4.3. Certificate Verify\n\n This message is used to provide explicit proof that an endpoint\n possesses the private key corresponding to its certificate. The\n CertificateVerify message also provides integrity for the handshake\n up to this point. Servers MUST send this message when authenticating\n via a certificate. Clients MUST send this message whenever\n authenticating via a certificate (i.e., when the Certificate message\n is non-empty). When sent, this message MUST appear immediately after\n the Certificate message and immediately prior to the Finished\n message.\n\n Structure of this message:\n\n struct {\n SignatureScheme algorithm;\n opaque signature<0..2^16-1>;\n } CertificateVerify;\n\n The algorithm field specifies the signature algorithm used (see\n Section 4.2.3 for the definition of this type). The signature is a\n digital signature using that algorithm. The content that is covered\n under the signature is the hash output as described in Section 4.4.1,\n namely:\n\n Transcript-Hash(Handshake Context, Certificate)\n\n The digital signature is then computed over the concatenation of:\n\n - A string that consists of octet 32 (0x20) repeated 64 times\n\n - The context string\n\n - A single 0 byte which serves as the separator\n\n - The content to be signed\n\n\n\n\n\n\n\nRescorla Standards Track [Page 69]\n\f\nRFC 8446 TLS August 2018\n\n\n This structure is intended to prevent an attack on previous versions\n of TLS in which the ServerKeyExchange format meant that attackers\n could obtain a signature of a message with a chosen 32-byte prefix\n (ClientHello.random). The initial 64-byte pad clears that prefix\n along with the server-controlled ServerHello.random.\n\n The context string for a server signature is\n \"TLS 1.3, server CertificateVerify\". The context string for a\n client signature is \"TLS 1.3, client CertificateVerify\". It is\n used to provide separation between signatures made in different\n contexts, helping against potential cross-protocol attacks.\n\n For example, if the transcript hash was 32 bytes of 01 (this length\n would make sense for SHA-256), the content covered by the digital\n signature for a server CertificateVerify would be:\n\n 2020202020202020202020202020202020202020202020202020202020202020\n 2020202020202020202020202020202020202020202020202020202020202020\n 544c5320312e332c207365727665722043657274696669636174655665726966\n 79\n 00\n 0101010101010101010101010101010101010101010101010101010101010101\n\n On the sender side, the process for computing the signature field of\n the CertificateVerify message takes as input:\n\n - The content covered by the digital signature\n\n - The private signing key corresponding to the certificate sent in\n the previous message\n\n If the CertificateVerify message is sent by a server, the signature\n algorithm MUST be one offered in the client's \"signature_algorithms\"\n extension unless no valid certificate chain can be produced without\n unsupported algorithms (see Section 4.2.3).\n\n If sent by a client, the signature algorithm used in the signature\n MUST be one of those present in the supported_signature_algorithms\n field of the \"signature_algorithms\" extension in the\n CertificateRequest message.\n\n In addition, the signature algorithm MUST be compatible with the key\n in the sender's end-entity certificate. RSA signatures MUST use an\n RSASSA-PSS algorithm, regardless of whether RSASSA-PKCS1-v1_5\n algorithms appear in \"signature_algorithms\". The SHA-1 algorithm\n MUST NOT be used in any signatures of CertificateVerify messages.\n\n\n\n\n\nRescorla Standards Track [Page 70]\n\f\nRFC 8446 TLS August 2018\n\n\n All SHA-1 signature algorithms in this specification are defined\n solely for use in legacy certificates and are not valid for\n CertificateVerify signatures.\n\n The receiver of a CertificateVerify message MUST verify the signature\n field. The verification process takes as input:\n\n - The content covered by the digital signature\n\n - The public key contained in the end-entity certificate found in\n the associated Certificate message\n\n - The digital signature received in the signature field of the\n CertificateVerify message\n\n If the verification fails, the receiver MUST terminate the handshake\n with a \"decrypt_error\" alert.\n\n4.4.4. Finished\n\n The Finished message is the final message in the Authentication\n Block. It is essential for providing authentication of the handshake\n and of the computed keys.\n\n Recipients of Finished messages MUST verify that the contents are\n correct and if incorrect MUST terminate the connection with a\n \"decrypt_error\" alert.\n\n Once a side has sent its Finished message and has received and\n validated the Finished message from its peer, it may begin to send\n and receive Application Data over the connection. There are two\n settings in which it is permitted to send data prior to receiving the\n peer's Finished:\n\n 1. Clients sending 0-RTT data as described in Section 4.2.10.\n\n 2. Servers MAY send data after sending their first flight, but\n because the handshake is not yet complete, they have no assurance\n of either the peer's identity or its liveness (i.e., the\n ClientHello might have been replayed).\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 71]\n\f\nRFC 8446 TLS August 2018\n\n\n The key used to compute the Finished message is computed from the\n Base Key defined in Section 4.4 using HKDF (see Section 7.1).\n Specifically:\n\n finished_key =\n HKDF-Expand-Label(BaseKey, \"finished\", \"\", Hash.length)\n\n Structure of this message:\n\n struct {\n opaque verify_data[Hash.length];\n } Finished;\n\n The verify_data value is computed as follows:\n\n verify_data =\n HMAC(finished_key,\n Transcript-Hash(Handshake Context,\n Certificate*, CertificateVerify*))\n\n * Only included if present.\n\n HMAC [RFC2104] uses the Hash algorithm for the handshake. As noted\n above, the HMAC input can generally be implemented by a running hash,\n i.e., just the handshake hash at this point.\n\n In previous versions of TLS, the verify_data was always 12 octets\n long. In TLS 1.3, it is the size of the HMAC output for the Hash\n used for the handshake.\n\n Note: Alerts and any other non-handshake record types are not\n handshake messages and are not included in the hash computations.\n\n Any records following a Finished message MUST be encrypted under the\n appropriate application traffic key as described in Section 7.2. In\n particular, this includes any alerts sent by the server in response\n to client Certificate and CertificateVerify messages.\n\n4.5. End of Early Data\n\n struct {} EndOfEarlyData;\n\n If the server sent an \"early_data\" extension in EncryptedExtensions,\n the client MUST send an EndOfEarlyData message after receiving the\n server Finished. If the server does not send an \"early_data\"\n extension in EncryptedExtensions, then the client MUST NOT send an\n EndOfEarlyData message. This message indicates that all 0-RTT\n application_data messages, if any, have been transmitted and that the\n\n\n\nRescorla Standards Track [Page 72]\n\f\nRFC 8446 TLS August 2018\n\n\n following records are protected under handshake traffic keys.\n Servers MUST NOT send this message, and clients receiving it MUST\n terminate the connection with an \"unexpected_message\" alert. This\n message is encrypted under keys derived from the\n client_early_traffic_secret.\n\n4.6. Post-Handshake Messages\n\n TLS also allows other messages to be sent after the main handshake.\n These messages use a handshake content type and are encrypted under\n the appropriate application traffic key.\n\n4.6.1. New Session Ticket Message\n\n At any time after the server has received the client Finished\n message, it MAY send a NewSessionTicket message. This message\n creates a unique association between the ticket value and a secret\n PSK derived from the resumption master secret (see Section 7).\n\n The client MAY use this PSK for future handshakes by including the\n ticket value in the \"pre_shared_key\" extension in its ClientHello\n (Section 4.2.11). Servers MAY send multiple tickets on a single\n connection, either immediately after each other or after specific\n events (see Appendix C.4). For instance, the server might send a new\n ticket after post-handshake authentication in order to encapsulate\n the additional client authentication state. Multiple tickets are\n useful for clients for a variety of purposes, including:\n\n - Opening multiple parallel HTTP connections.\n\n - Performing connection racing across interfaces and address\n families via (for example) Happy Eyeballs [RFC8305] or related\n techniques.\n\n Any ticket MUST only be resumed with a cipher suite that has the same\n KDF hash algorithm as that used to establish the original connection.\n\n Clients MUST only resume if the new SNI value is valid for the server\n certificate presented in the original session and SHOULD only resume\n if the SNI value matches the one used in the original session. The\n latter is a performance optimization: normally, there is no reason to\n expect that different servers covered by a single certificate would\n be able to accept each other's tickets; hence, attempting resumption\n in that case would waste a single-use ticket. If such an indication\n is provided (externally or by any other means), clients MAY resume\n with a different SNI value.\n\n\n\n\n\nRescorla Standards Track [Page 73]\n\f\nRFC 8446 TLS August 2018\n\n\n On resumption, if reporting an SNI value to the calling application,\n implementations MUST use the value sent in the resumption ClientHello\n rather than the value sent in the previous session. Note that if a\n server implementation declines all PSK identities with different SNI\n values, these two values are always the same.\n\n Note: Although the resumption master secret depends on the client's\n second flight, a server which does not request client authentication\n MAY compute the remainder of the transcript independently and then\n send a NewSessionTicket immediately upon sending its Finished rather\n than waiting for the client Finished. This might be appropriate in\n cases where the client is expected to open multiple TLS connections\n in parallel and would benefit from the reduced overhead of a\n resumption handshake, for example.\n\n struct {\n uint32 ticket_lifetime;\n uint32 ticket_age_add;\n opaque ticket_nonce<0..255>;\n opaque ticket<1..2^16-1>;\n Extension extensions<0..2^16-2>;\n } NewSessionTicket;\n\n ticket_lifetime: Indicates the lifetime in seconds as a 32-bit\n unsigned integer in network byte order from the time of ticket\n issuance. Servers MUST NOT use any value greater than\n 604800 seconds (7 days). The value of zero indicates that the\n ticket should be discarded immediately. Clients MUST NOT cache\n tickets for longer than 7 days, regardless of the ticket_lifetime,\n and MAY delete tickets earlier based on local policy. A server\n MAY treat a ticket as valid for a shorter period of time than what\n is stated in the ticket_lifetime.\n\n ticket_age_add: A securely generated, random 32-bit value that is\n used to obscure the age of the ticket that the client includes in\n the \"pre_shared_key\" extension. The client-side ticket age is\n added to this value modulo 2^32 to obtain the value that is\n transmitted by the client. The server MUST generate a fresh value\n for each ticket it sends.\n\n ticket_nonce: A per-ticket value that is unique across all tickets\n issued on this connection.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 74]\n\f\nRFC 8446 TLS August 2018\n\n\n ticket: The value of the ticket to be used as the PSK identity. The\n ticket itself is an opaque label. It MAY be either a database\n lookup key or a self-encrypted and self-authenticated value.\n\n extensions: A set of extension values for the ticket. The\n \"Extension\" format is defined in Section 4.2. Clients MUST ignore\n unrecognized extensions.\n\n The sole extension currently defined for NewSessionTicket is\n \"early_data\", indicating that the ticket may be used to send 0-RTT\n data (Section 4.2.10). It contains the following value:\n\n max_early_data_size: The maximum amount of 0-RTT data that the\n client is allowed to send when using this ticket, in bytes. Only\n Application Data payload (i.e., plaintext but not padding or the\n inner content type byte) is counted. A server receiving more than\n max_early_data_size bytes of 0-RTT data SHOULD terminate the\n connection with an \"unexpected_message\" alert. Note that servers\n that reject early data due to lack of cryptographic material will\n be unable to differentiate padding from content, so clients\n SHOULD NOT depend on being able to send large quantities of\n padding in early data records.\n\n The PSK associated with the ticket is computed as:\n\n HKDF-Expand-Label(resumption_master_secret,\n \"resumption\", ticket_nonce, Hash.length)\n\n Because the ticket_nonce value is distinct for each NewSessionTicket\n message, a different PSK will be derived for each ticket.\n\n Note that in principle it is possible to continue issuing new tickets\n which indefinitely extend the lifetime of the keying material\n originally derived from an initial non-PSK handshake (which was most\n likely tied to the peer's certificate). It is RECOMMENDED that\n implementations place limits on the total lifetime of such keying\n material; these limits should take into account the lifetime of the\n peer's certificate, the likelihood of intervening revocation, and the\n time since the peer's online CertificateVerify signature.\n\n4.6.2. Post-Handshake Authentication\n\n When the client has sent the \"post_handshake_auth\" extension (see\n Section 4.2.6), a server MAY request client authentication at any\n time after the handshake has completed by sending a\n CertificateRequest message. The client MUST respond with the\n appropriate Authentication messages (see Section 4.4). If the client\n chooses to authenticate, it MUST send Certificate, CertificateVerify,\n\n\n\nRescorla Standards Track [Page 75]\n\f\nRFC 8446 TLS August 2018\n\n\n and Finished. If it declines, it MUST send a Certificate message\n containing no certificates followed by Finished. All of the client's\n messages for a given response MUST appear consecutively on the wire\n with no intervening messages of other types.\n\n A client that receives a CertificateRequest message without having\n sent the \"post_handshake_auth\" extension MUST send an\n \"unexpected_message\" fatal alert.\n\n Note: Because client authentication could involve prompting the user,\n servers MUST be prepared for some delay, including receiving an\n arbitrary number of other messages between sending the\n CertificateRequest and receiving a response. In addition, clients\n which receive multiple CertificateRequests in close succession MAY\n respond to them in a different order than they were received (the\n certificate_request_context value allows the server to disambiguate\n the responses).\n\n4.6.3. Key and Initialization Vector Update\n\n The KeyUpdate handshake message is used to indicate that the sender\n is updating its sending cryptographic keys. This message can be sent\n by either peer after it has sent a Finished message. Implementations\n that receive a KeyUpdate message prior to receiving a Finished\n message MUST terminate the connection with an \"unexpected_message\"\n alert. After sending a KeyUpdate message, the sender SHALL send all\n its traffic using the next generation of keys, computed as described\n in Section 7.2. Upon receiving a KeyUpdate, the receiver MUST update\n its receiving keys.\n\n enum {\n update_not_requested(0), update_requested(1), (255)\n } KeyUpdateRequest;\n\n struct {\n KeyUpdateRequest request_update;\n } KeyUpdate;\n\n request_update: Indicates whether the recipient of the KeyUpdate\n should respond with its own KeyUpdate. If an implementation\n receives any other value, it MUST terminate the connection with an\n \"illegal_parameter\" alert.\n\n If the request_update field is set to \"update_requested\", then the\n receiver MUST send a KeyUpdate of its own with request_update set to\n \"update_not_requested\" prior to sending its next Application Data\n record. This mechanism allows either side to force an update to the\n entire connection, but causes an implementation which receives\n\n\n\nRescorla Standards Track [Page 76]\n\f\nRFC 8446 TLS August 2018\n\n\n multiple KeyUpdates while it is silent to respond with a single\n update. Note that implementations may receive an arbitrary number of\n messages between sending a KeyUpdate with request_update set to\n \"update_requested\" and receiving the peer's KeyUpdate, because those\n messages may already be in flight. However, because send and receive\n keys are derived from independent traffic secrets, retaining the\n receive traffic secret does not threaten the forward secrecy of data\n sent before the sender changed keys.\n\n If implementations independently send their own KeyUpdates with\n request_update set to \"update_requested\" and they cross in flight,\n then each side will also send a response, with the result that each\n side increments by two generations.\n\n Both sender and receiver MUST encrypt their KeyUpdate messages with\n the old keys. Additionally, both sides MUST enforce that a KeyUpdate\n with the old key is received before accepting any messages encrypted\n with the new key. Failure to do so may allow message truncation\n attacks.\n\n5. Record Protocol\n\n The TLS record protocol takes messages to be transmitted, fragments\n the data into manageable blocks, protects the records, and transmits\n the result. Received data is verified, decrypted, reassembled, and\n then delivered to higher-level clients.\n\n TLS records are typed, which allows multiple higher-level protocols\n to be multiplexed over the same record layer. This document\n specifies four content types: handshake, application_data, alert, and\n change_cipher_spec. The change_cipher_spec record is used only for\n compatibility purposes (see Appendix D.4).\n\n An implementation may receive an unencrypted record of type\n change_cipher_spec consisting of the single byte value 0x01 at any\n time after the first ClientHello message has been sent or received\n and before the peer's Finished message has been received and MUST\n simply drop it without further processing. Note that this record may\n appear at a point at the handshake where the implementation is\n expecting protected records, and so it is necessary to detect this\n condition prior to attempting to deprotect the record. An\n implementation which receives any other change_cipher_spec value or\n which receives a protected change_cipher_spec record MUST abort the\n handshake with an \"unexpected_message\" alert. If an implementation\n detects a change_cipher_spec record received before the first\n ClientHello message or after the peer's Finished message, it MUST be\n treated as an unexpected record type (though stateless servers may\n not be able to distinguish these cases from allowed cases).\n\n\n\nRescorla Standards Track [Page 77]\n\f\nRFC 8446 TLS August 2018\n\n\n Implementations MUST NOT send record types not defined in this\n document unless negotiated by some extension. If a TLS\n implementation receives an unexpected record type, it MUST terminate\n the connection with an \"unexpected_message\" alert. New record\n content type values are assigned by IANA in the TLS ContentType\n registry as described in Section 11.\n\n5.1. Record Layer\n\n The record layer fragments information blocks into TLSPlaintext\n records carrying data in chunks of 2^14 bytes or less. Message\n boundaries are handled differently depending on the underlying\n ContentType. Any future content types MUST specify appropriate\n rules. Note that these rules are stricter than what was enforced in\n TLS 1.2.\n\n Handshake messages MAY be coalesced into a single TLSPlaintext record\n or fragmented across several records, provided that:\n\n - Handshake messages MUST NOT be interleaved with other record\n types. That is, if a handshake message is split over two or more\n records, there MUST NOT be any other records between them.\n\n - Handshake messages MUST NOT span key changes. Implementations\n MUST verify that all messages immediately preceding a key change\n align with a record boundary; if not, then they MUST terminate the\n connection with an \"unexpected_message\" alert. Because the\n ClientHello, EndOfEarlyData, ServerHello, Finished, and KeyUpdate\n messages can immediately precede a key change, implementations\n MUST send these messages in alignment with a record boundary.\n\n Implementations MUST NOT send zero-length fragments of Handshake\n types, even if those fragments contain padding.\n\n Alert messages (Section 6) MUST NOT be fragmented across records, and\n multiple alert messages MUST NOT be coalesced into a single\n TLSPlaintext record. In other words, a record with an Alert type\n MUST contain exactly one message.\n\n Application Data messages contain data that is opaque to TLS.\n Application Data messages are always protected. Zero-length\n fragments of Application Data MAY be sent, as they are potentially\n useful as a traffic analysis countermeasure. Application Data\n fragments MAY be split across multiple records or coalesced into a\n single record.\n\n\n\n\n\n\nRescorla Standards Track [Page 78]\n\f\nRFC 8446 TLS August 2018\n\n\n enum {\n invalid(0),\n change_cipher_spec(20),\n alert(21),\n handshake(22),\n application_data(23),\n (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion legacy_record_version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n type: The higher-level protocol used to process the enclosed\n fragment.\n\n legacy_record_version: MUST be set to 0x0303 for all records\n generated by a TLS 1.3 implementation other than an initial\n ClientHello (i.e., one not generated after a HelloRetryRequest),\n where it MAY also be 0x0301 for compatibility purposes. This\n field is deprecated and MUST be ignored for all purposes.\n Previous versions of TLS would use other values in this field\n under some circumstances.\n\n length: The length (in bytes) of the following\n TLSPlaintext.fragment. The length MUST NOT exceed 2^14 bytes. An\n endpoint that receives a record that exceeds this length MUST\n terminate the connection with a \"record_overflow\" alert.\n\n fragment: The data being transmitted. This value is transparent and\n is treated as an independent block to be dealt with by the higher-\n level protocol specified by the type field.\n\n This document describes TLS 1.3, which uses the version 0x0304. This\n version value is historical, deriving from the use of 0x0301 for\n TLS 1.0 and 0x0300 for SSL 3.0. In order to maximize backward\n compatibility, a record containing an initial ClientHello SHOULD have\n version 0x0301 (reflecting TLS 1.0) and a record containing a second\n ClientHello or a ServerHello MUST have version 0x0303 (reflecting\n TLS 1.2). When negotiating prior versions of TLS, endpoints follow\n the procedure and requirements provided in Appendix D.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 79]\n\f\nRFC 8446 TLS August 2018\n\n\n When record protection has not yet been engaged, TLSPlaintext\n structures are written directly onto the wire. Once record\n protection has started, TLSPlaintext records are protected and sent\n as described in the following section. Note that Application Data\n records MUST NOT be written to the wire unprotected (see Section 2\n for details).\n\n5.2. Record Payload Protection\n\n The record protection functions translate a TLSPlaintext structure\n into a TLSCiphertext structure. The deprotection functions reverse\n the process. In TLS 1.3, as opposed to previous versions of TLS, all\n ciphers are modeled as \"Authenticated Encryption with Associated\n Data\" (AEAD) [RFC5116]. AEAD functions provide a unified encryption\n and authentication operation which turns plaintext into authenticated\n ciphertext and back again. Each encrypted record consists of a\n plaintext header followed by an encrypted body, which itself contains\n a type and optional padding.\n\n struct {\n opaque content[TLSPlaintext.length];\n ContentType type;\n uint8 zeros[length_of_padding];\n } TLSInnerPlaintext;\n\n struct {\n ContentType opaque_type = application_data; /* 23 */\n ProtocolVersion legacy_record_version = 0x0303; /* TLS v1.2 */\n uint16 length;\n opaque encrypted_record[TLSCiphertext.length];\n } TLSCiphertext;\n\n content: The TLSPlaintext.fragment value, containing the byte\n encoding of a handshake or an alert message, or the raw bytes of\n the application's data to send.\n\n type: The TLSPlaintext.type value containing the content type of the\n record.\n\n zeros: An arbitrary-length run of zero-valued bytes may appear in\n the cleartext after the type field. This provides an opportunity\n for senders to pad any TLS record by a chosen amount as long as\n the total stays within record size limits. See Section 5.4 for\n more details.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 80]\n\f\nRFC 8446 TLS August 2018\n\n\n opaque_type: The outer opaque_type field of a TLSCiphertext record\n is always set to the value 23 (application_data) for outward\n compatibility with middleboxes accustomed to parsing previous\n versions of TLS. The actual content type of the record is found\n in TLSInnerPlaintext.type after decryption.\n\n legacy_record_version: The legacy_record_version field is always\n 0x0303. TLS 1.3 TLSCiphertexts are not generated until after\n TLS 1.3 has been negotiated, so there are no historical\n compatibility concerns where other values might be received. Note\n that the handshake protocol, including the ClientHello and\n ServerHello messages, authenticates the protocol version, so this\n value is redundant.\n\n length: The length (in bytes) of the following\n TLSCiphertext.encrypted_record, which is the sum of the lengths of\n the content and the padding, plus one for the inner content type,\n plus any expansion added by the AEAD algorithm. The length\n MUST NOT exceed 2^14 + 256 bytes. An endpoint that receives a\n record that exceeds this length MUST terminate the connection with\n a \"record_overflow\" alert.\n\n encrypted_record: The AEAD-encrypted form of the serialized\n TLSInnerPlaintext structure.\n\n AEAD algorithms take as input a single key, a nonce, a plaintext, and\n \"additional data\" to be included in the authentication check, as\n described in Section 2.1 of [RFC5116]. The key is either the\n client_write_key or the server_write_key, the nonce is derived from\n the sequence number and the client_write_iv or server_write_iv (see\n Section 5.3), and the additional data input is the record header.\n\n I.e.,\n\n additional_data = TLSCiphertext.opaque_type ||\n TLSCiphertext.legacy_record_version ||\n TLSCiphertext.length\n\n The plaintext input to the AEAD algorithm is the encoded\n TLSInnerPlaintext structure. Derivation of traffic keys is defined\n in Section 7.3.\n\n The AEAD output consists of the ciphertext output from the AEAD\n encryption operation. The length of the plaintext is greater than\n the corresponding TLSPlaintext.length due to the inclusion of\n TLSInnerPlaintext.type and any padding supplied by the sender. The\n length of the AEAD output will generally be larger than the\n plaintext, but by an amount that varies with the AEAD algorithm.\n\n\n\nRescorla Standards Track [Page 81]\n\f\nRFC 8446 TLS August 2018\n\n\n Since the ciphers might incorporate padding, the amount of overhead\n could vary with different lengths of plaintext. Symbolically,\n\n AEADEncrypted =\n AEAD-Encrypt(write_key, nonce, additional_data, plaintext)\n\n The encrypted_record field of TLSCiphertext is set to AEADEncrypted.\n\n In order to decrypt and verify, the cipher takes as input the key,\n nonce, additional data, and the AEADEncrypted value. The output is\n either the plaintext or an error indicating that the decryption\n failed. There is no separate integrity check. Symbolically,\n\n plaintext of encrypted_record =\n AEAD-Decrypt(peer_write_key, nonce,\n additional_data, AEADEncrypted)\n\n If the decryption fails, the receiver MUST terminate the connection\n with a \"bad_record_mac\" alert.\n\n An AEAD algorithm used in TLS 1.3 MUST NOT produce an expansion\n greater than 255 octets. An endpoint that receives a record from its\n peer with TLSCiphertext.length larger than 2^14 + 256 octets MUST\n terminate the connection with a \"record_overflow\" alert. This limit\n is derived from the maximum TLSInnerPlaintext length of 2^14 octets +\n 1 octet for ContentType + the maximum AEAD expansion of 255 octets.\n\n5.3. Per-Record Nonce\n\n A 64-bit sequence number is maintained separately for reading and\n writing records. The appropriate sequence number is incremented by\n one after reading or writing each record. Each sequence number is\n set to zero at the beginning of a connection and whenever the key is\n changed; the first record transmitted under a particular traffic key\n MUST use sequence number 0.\n\n Because the size of sequence numbers is 64-bit, they should not wrap.\n If a TLS implementation would need to wrap a sequence number, it MUST\n either rekey (Section 4.6.3) or terminate the connection.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 82]\n\f\nRFC 8446 TLS August 2018\n\n\n Each AEAD algorithm will specify a range of possible lengths for the\n per-record nonce, from N_MIN bytes to N_MAX bytes of input [RFC5116].\n The length of the TLS per-record nonce (iv_length) is set to the\n larger of 8 bytes and N_MIN for the AEAD algorithm (see [RFC5116],\n Section 4). An AEAD algorithm where N_MAX is less than 8 bytes\n MUST NOT be used with TLS. The per-record nonce for the AEAD\n construction is formed as follows:\n\n 1. The 64-bit record sequence number is encoded in network byte\n order and padded to the left with zeros to iv_length.\n\n 2. The padded sequence number is XORed with either the static\n client_write_iv or server_write_iv (depending on the role).\n\n The resulting quantity (of length iv_length) is used as the\n per-record nonce.\n\n Note: This is a different construction from that in TLS 1.2, which\n specified a partially explicit nonce.\n\n5.4. Record Padding\n\n All encrypted TLS records can be padded to inflate the size of the\n TLSCiphertext. This allows the sender to hide the size of the\n traffic from an observer.\n\n When generating a TLSCiphertext record, implementations MAY choose to\n pad. An unpadded record is just a record with a padding length of\n zero. Padding is a string of zero-valued bytes appended to the\n ContentType field before encryption. Implementations MUST set the\n padding octets to all zeros before encrypting.\n\n Application Data records may contain a zero-length\n TLSInnerPlaintext.content if the sender desires. This permits\n generation of plausibly sized cover traffic in contexts where the\n presence or absence of activity may be sensitive. Implementations\n MUST NOT send Handshake and Alert records that have a zero-length\n TLSInnerPlaintext.content; if such a message is received, the\n receiving implementation MUST terminate the connection with an\n \"unexpected_message\" alert.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 83]\n\f\nRFC 8446 TLS August 2018\n\n\n The padding sent is automatically verified by the record protection\n mechanism; upon successful decryption of a\n TLSCiphertext.encrypted_record, the receiving implementation scans\n the field from the end toward the beginning until it finds a non-zero\n octet. This non-zero octet is the content type of the message. This\n padding scheme was selected because it allows padding of any\n encrypted TLS record by an arbitrary size (from zero up to TLS record\n size limits) without introducing new content types. The design also\n enforces all-zero padding octets, which allows for quick detection of\n padding errors.\n\n Implementations MUST limit their scanning to the cleartext returned\n from the AEAD decryption. If a receiving implementation does not\n find a non-zero octet in the cleartext, it MUST terminate the\n connection with an \"unexpected_message\" alert.\n\n The presence of padding does not change the overall record size\n limitations: the full encoded TLSInnerPlaintext MUST NOT exceed 2^14\n + 1 octets. If the maximum fragment length is reduced -- as, for\n example, by the record_size_limit extension from [RFC8449] -- then\n the reduced limit applies to the full plaintext, including the\n content type and padding.\n\n Selecting a padding policy that suggests when and how much to pad is\n a complex topic and is beyond the scope of this specification. If\n the application-layer protocol on top of TLS has its own padding, it\n may be preferable to pad Application Data TLS records within the\n application layer. Padding for encrypted Handshake or Alert records\n must still be handled at the TLS layer, though. Later documents may\n define padding selection algorithms or define a padding policy\n request mechanism through TLS extensions or some other means.\n\n5.5. Limits on Key Usage\n\n There are cryptographic limits on the amount of plaintext which can\n be safely encrypted under a given set of keys. [AEAD-LIMITS]\n provides an analysis of these limits under the assumption that the\n underlying primitive (AES or ChaCha20) has no weaknesses.\n Implementations SHOULD do a key update as described in Section 4.6.3\n prior to reaching these limits.\n\n For AES-GCM, up to 2^24.5 full-size records (about 24 million) may be\n encrypted on a given connection while keeping a safety margin of\n approximately 2^-57 for Authenticated Encryption (AE) security. For\n ChaCha20/Poly1305, the record sequence number would wrap before the\n safety limit is reached.\n\n\n\n\n\nRescorla Standards Track [Page 84]\n\f\nRFC 8446 TLS August 2018\n\n\n6. Alert Protocol\n\n TLS provides an Alert content type to indicate closure information\n and errors. Like other messages, alert messages are encrypted as\n specified by the current connection state.\n\n Alert messages convey a description of the alert and a legacy field\n that conveyed the severity level of the message in previous versions\n of TLS. Alerts are divided into two classes: closure alerts and\n error alerts. In TLS 1.3, the severity is implicit in the type of\n alert being sent, and the \"level\" field can safely be ignored. The\n \"close_notify\" alert is used to indicate orderly closure of one\n direction of the connection. Upon receiving such an alert, the TLS\n implementation SHOULD indicate end-of-data to the application.\n\n Error alerts indicate abortive closure of the connection (see\n Section 6.2). Upon receiving an error alert, the TLS implementation\n SHOULD indicate an error to the application and MUST NOT allow any\n further data to be sent or received on the connection. Servers and\n clients MUST forget the secret values and keys established in failed\n connections, with the exception of the PSKs associated with session\n tickets, which SHOULD be discarded if possible.\n\n All the alerts listed in Section 6.2 MUST be sent with\n AlertLevel=fatal and MUST be treated as error alerts when received\n regardless of the AlertLevel in the message. Unknown Alert types\n MUST be treated as error alerts.\n\n Note: TLS defines two generic alerts (see Section 6) to use upon\n failure to parse a message. Peers which receive a message which\n cannot be parsed according to the syntax (e.g., have a length\n extending beyond the message boundary or contain an out-of-range\n length) MUST terminate the connection with a \"decode_error\" alert.\n Peers which receive a message which is syntactically correct but\n semantically invalid (e.g., a DHE share of p - 1, or an invalid enum)\n MUST terminate the connection with an \"illegal_parameter\" alert.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 85]\n\f\nRFC 8446 TLS August 2018\n\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n record_overflow(22),\n handshake_failure(40),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n inappropriate_fallback(86),\n user_canceled(90),\n missing_extension(109),\n unsupported_extension(110),\n unrecognized_name(112),\n bad_certificate_status_response(113),\n unknown_psk_identity(115),\n certificate_required(116),\n no_application_protocol(120),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 86]\n\f\nRFC 8446 TLS August 2018\n\n\n6.1. Closure Alerts\n\n The client and the server must share knowledge that the connection is\n ending in order to avoid a truncation attack.\n\n close_notify: This alert notifies the recipient that the sender will\n not send any more messages on this connection. Any data received\n after a closure alert has been received MUST be ignored.\n\n user_canceled: This alert notifies the recipient that the sender is\n canceling the handshake for some reason unrelated to a protocol\n failure. If a user cancels an operation after the handshake is\n complete, just closing the connection by sending a \"close_notify\"\n is more appropriate. This alert SHOULD be followed by a\n \"close_notify\". This alert generally has AlertLevel=warning.\n\n Either party MAY initiate a close of its write side of the connection\n by sending a \"close_notify\" alert. Any data received after a closure\n alert has been received MUST be ignored. If a transport-level close\n is received prior to a \"close_notify\", the receiver cannot know that\n all the data that was sent has been received.\n\n Each party MUST send a \"close_notify\" alert before closing its write\n side of the connection, unless it has already sent some error alert.\n This does not have any effect on its read side of the connection.\n Note that this is a change from versions of TLS prior to TLS 1.3 in\n which implementations were required to react to a \"close_notify\" by\n discarding pending writes and sending an immediate \"close_notify\"\n alert of their own. That previous requirement could cause truncation\n in the read side. Both parties need not wait to receive a\n \"close_notify\" alert before closing their read side of the\n connection, though doing so would introduce the possibility of\n truncation.\n\n If the application protocol using TLS provides that any data may be\n carried over the underlying transport after the TLS connection is\n closed, the TLS implementation MUST receive a \"close_notify\" alert\n before indicating end-of-data to the application layer. No part of\n this standard should be taken to dictate the manner in which a usage\n profile for TLS manages its data transport, including when\n connections are opened or closed.\n\n Note: It is assumed that closing the write side of a connection\n reliably delivers pending data before destroying the transport.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 87]\n\f\nRFC 8446 TLS August 2018\n\n\n6.2. Error Alerts\n\n Error handling in TLS is very simple. When an error is detected, the\n detecting party sends a message to its peer. Upon transmission or\n receipt of a fatal alert message, both parties MUST immediately close\n the connection.\n\n Whenever an implementation encounters a fatal error condition, it\n SHOULD send an appropriate fatal alert and MUST close the connection\n without sending or receiving any additional data. In the rest of\n this specification, when the phrases \"terminate the connection\" and\n \"abort the handshake\" are used without a specific alert it means that\n the implementation SHOULD send the alert indicated by the\n descriptions below. The phrases \"terminate the connection with an X\n alert\" and \"abort the handshake with an X alert\" mean that the\n implementation MUST send alert X if it sends any alert. All alerts\n defined below in this section, as well as all unknown alerts, are\n universally considered fatal as of TLS 1.3 (see Section 6). The\n implementation SHOULD provide a way to facilitate logging the sending\n and receiving of alerts.\n\n The following error alerts are defined:\n\n unexpected_message: An inappropriate message (e.g., the wrong\n handshake message, premature Application Data, etc.) was received.\n This alert should never be observed in communication between\n proper implementations.\n\n bad_record_mac: This alert is returned if a record is received which\n cannot be deprotected. Because AEAD algorithms combine decryption\n and verification, and also to avoid side-channel attacks, this\n alert is used for all deprotection failures. This alert should\n never be observed in communication between proper implementations,\n except when messages were corrupted in the network.\n\n record_overflow: A TLSCiphertext record was received that had a\n length more than 2^14 + 256 bytes, or a record decrypted to a\n TLSPlaintext record with more than 2^14 bytes (or some other\n negotiated limit). This alert should never be observed in\n communication between proper implementations, except when messages\n were corrupted in the network.\n\n handshake_failure: Receipt of a \"handshake_failure\" alert message\n indicates that the sender was unable to negotiate an acceptable\n set of security parameters given the options available.\n\n bad_certificate: A certificate was corrupt, contained signatures\n that did not verify correctly, etc.\n\n\n\nRescorla Standards Track [Page 88]\n\f\nRFC 8446 TLS August 2018\n\n\n unsupported_certificate: A certificate was of an unsupported type.\n\n certificate_revoked: A certificate was revoked by its signer.\n\n certificate_expired: A certificate has expired or is not currently\n valid.\n\n certificate_unknown: Some other (unspecified) issue arose in\n processing the certificate, rendering it unacceptable.\n\n illegal_parameter: A field in the handshake was incorrect or\n inconsistent with other fields. This alert is used for errors\n which conform to the formal protocol syntax but are otherwise\n incorrect.\n\n unknown_ca: A valid certificate chain or partial chain was received,\n but the certificate was not accepted because the CA certificate\n could not be located or could not be matched with a known trust\n anchor.\n\n access_denied: A valid certificate or PSK was received, but when\n access control was applied, the sender decided not to proceed with\n negotiation.\n\n decode_error: A message could not be decoded because some field was\n out of the specified range or the length of the message was\n incorrect. This alert is used for errors where the message does\n not conform to the formal protocol syntax. This alert should\n never be observed in communication between proper implementations,\n except when messages were corrupted in the network.\n\n decrypt_error: A handshake (not record layer) cryptographic\n operation failed, including being unable to correctly verify a\n signature or validate a Finished message or a PSK binder.\n\n protocol_version: The protocol version the peer has attempted to\n negotiate is recognized but not supported (see Appendix D).\n\n insufficient_security: Returned instead of \"handshake_failure\" when\n a negotiation has failed specifically because the server requires\n parameters more secure than those supported by the client.\n\n internal_error: An internal error unrelated to the peer or the\n correctness of the protocol (such as a memory allocation failure)\n makes it impossible to continue.\n\n inappropriate_fallback: Sent by a server in response to an invalid\n connection retry attempt from a client (see [RFC7507]).\n\n\n\nRescorla Standards Track [Page 89]\n\f\nRFC 8446 TLS August 2018\n\n\n missing_extension: Sent by endpoints that receive a handshake\n message not containing an extension that is mandatory to send for\n the offered TLS version or other negotiated parameters.\n\n unsupported_extension: Sent by endpoints receiving any handshake\n message containing an extension known to be prohibited for\n inclusion in the given handshake message, or including any\n extensions in a ServerHello or Certificate not first offered in\n the corresponding ClientHello or CertificateRequest.\n\n unrecognized_name: Sent by servers when no server exists identified\n by the name provided by the client via the \"server_name\" extension\n (see [RFC6066]).\n\n bad_certificate_status_response: Sent by clients when an invalid or\n unacceptable OCSP response is provided by the server via the\n \"status_request\" extension (see [RFC6066]).\n\n unknown_psk_identity: Sent by servers when PSK key establishment is\n desired but no acceptable PSK identity is provided by the client.\n Sending this alert is OPTIONAL; servers MAY instead choose to send\n a \"decrypt_error\" alert to merely indicate an invalid PSK\n identity.\n\n certificate_required: Sent by servers when a client certificate is\n desired but none was provided by the client.\n\n no_application_protocol: Sent by servers when a client\n \"application_layer_protocol_negotiation\" extension advertises only\n protocols that the server does not support (see [RFC7301]).\n\n New Alert values are assigned by IANA as described in Section 11.\n\n7. Cryptographic Computations\n\n The TLS handshake establishes one or more input secrets which are\n combined to create the actual working keying material, as detailed\n below. The key derivation process incorporates both the input\n secrets and the handshake transcript. Note that because the\n handshake transcript includes the random values from the Hello\n messages, any given handshake will have different traffic secrets,\n even if the same input secrets are used, as is the case when the same\n PSK is used for multiple connections.\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 90]\n\f\nRFC 8446 TLS August 2018\n\n\n7.1. Key Schedule\n\n The key derivation process makes use of the HKDF-Extract and\n HKDF-Expand functions as defined for HKDF [RFC5869], as well as the\n functions defined below:\n\n HKDF-Expand-Label(Secret, Label, Context, Length) =\n HKDF-Expand(Secret, HkdfLabel, Length)\n\n Where HkdfLabel is specified as:\n\n struct {\n uint16 length = Length;\n opaque label<7..255> = \"tls13 \" + Label;\n opaque context<0..255> = Context;\n } HkdfLabel;\n\n Derive-Secret(Secret, Label, Messages) =\n HKDF-Expand-Label(Secret, Label,\n Transcript-Hash(Messages), Hash.length)\n\n The Hash function used by Transcript-Hash and HKDF is the cipher\n suite hash algorithm. Hash.length is its output length in bytes.\n Messages is the concatenation of the indicated handshake messages,\n including the handshake message type and length fields, but not\n including record layer headers. Note that in some cases a zero-\n length Context (indicated by \"\") is passed to HKDF-Expand-Label. The\n labels specified in this document are all ASCII strings and do not\n include a trailing NUL byte.\n\n Note: With common hash functions, any label longer than 12 characters\n requires an additional iteration of the hash function to compute.\n The labels in this specification have all been chosen to fit within\n this limit.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 91]\n\f\nRFC 8446 TLS August 2018\n\n\n Keys are derived from two input secrets using the HKDF-Extract and\n Derive-Secret functions. The general pattern for adding a new secret\n is to use HKDF-Extract with the Salt being the current secret state\n and the Input Keying Material (IKM) being the new secret to be added.\n In this version of TLS 1.3, the two input secrets are:\n\n - PSK (a pre-shared key established externally or derived from the\n resumption_master_secret value from a previous connection)\n\n - (EC)DHE shared secret (Section 7.4)\n\n This produces a full key derivation schedule shown in the diagram\n below. In this diagram, the following formatting conventions apply:\n\n - HKDF-Extract is drawn as taking the Salt argument from the top and\n the IKM argument from the left, with its output to the bottom and\n the name of the output on the right.\n\n - Derive-Secret's Secret argument is indicated by the incoming\n arrow. For instance, the Early Secret is the Secret for\n generating the client_early_traffic_secret.\n\n - \"0\" indicates a string of Hash.length bytes set to zero.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 92]\n\f\nRFC 8446 TLS August 2018\n\n\n 0\n |\n v\n PSK -> HKDF-Extract = Early Secret\n |\n +-----> Derive-Secret(., \"ext binder\" | \"res binder\", \"\")\n | = binder_key\n |\n +-----> Derive-Secret(., \"c e traffic\", ClientHello)\n | = client_early_traffic_secret\n |\n +-----> Derive-Secret(., \"e exp master\", ClientHello)\n | = early_exporter_master_secret\n v\n Derive-Secret(., \"derived\", \"\")\n |\n v\n (EC)DHE -> HKDF-Extract = Handshake Secret\n |\n +-----> Derive-Secret(., \"c hs traffic\",\n | ClientHello...ServerHello)\n | = client_handshake_traffic_secret\n |\n +-----> Derive-Secret(., \"s hs traffic\",\n | ClientHello...ServerHello)\n | = server_handshake_traffic_secret\n v\n Derive-Secret(., \"derived\", \"\")\n |\n v\n 0 -> HKDF-Extract = Master Secret\n |\n +-----> Derive-Secret(., \"c ap traffic\",\n | ClientHello...server Finished)\n | = client_application_traffic_secret_0\n |\n +-----> Derive-Secret(., \"s ap traffic\",\n | ClientHello...server Finished)\n | = server_application_traffic_secret_0\n |\n +-----> Derive-Secret(., \"exp master\",\n | ClientHello...server Finished)\n | = exporter_master_secret\n |\n +-----> Derive-Secret(., \"res master\",\n ClientHello...client Finished)\n = resumption_master_secret\n\n\n\n\nRescorla Standards Track [Page 93]\n\f\nRFC 8446 TLS August 2018\n\n\n The general pattern here is that the secrets shown down the left side\n of the diagram are just raw entropy without context, whereas the\n secrets down the right side include Handshake Context and therefore\n can be used to derive working keys without additional context. Note\n that the different calls to Derive-Secret may take different Messages\n arguments, even with the same secret. In a 0-RTT exchange,\n Derive-Secret is called with four distinct transcripts; in a\n 1-RTT-only exchange, it is called with three distinct transcripts.\n\n If a given secret is not available, then the 0-value consisting of a\n string of Hash.length bytes set to zeros is used. Note that this\n does not mean skipping rounds, so if PSK is not in use, Early Secret\n will still be HKDF-Extract(0, 0). For the computation of the\n binder_key, the label is \"ext binder\" for external PSKs (those\n provisioned outside of TLS) and \"res binder\" for resumption PSKs\n (those provisioned as the resumption master secret of a previous\n handshake). The different labels prevent the substitution of one\n type of PSK for the other.\n\n There are multiple potential Early Secret values, depending on which\n PSK the server ultimately selects. The client will need to compute\n one for each potential PSK; if no PSK is selected, it will then need\n to compute the Early Secret corresponding to the zero PSK.\n\n Once all the values which are to be derived from a given secret have\n been computed, that secret SHOULD be erased.\n\n7.2. Updating Traffic Secrets\n\n Once the handshake is complete, it is possible for either side to\n update its sending traffic keys using the KeyUpdate handshake message\n defined in Section 4.6.3. The next generation of traffic keys is\n computed by generating client_/server_application_traffic_secret_N+1\n from client_/server_application_traffic_secret_N as described in this\n section and then re-deriving the traffic keys as described in\n Section 7.3.\n\n The next-generation application_traffic_secret is computed as:\n\n application_traffic_secret_N+1 =\n HKDF-Expand-Label(application_traffic_secret_N,\n \"traffic upd\", \"\", Hash.length)\n\n Once client_/server_application_traffic_secret_N+1 and its associated\n traffic keys have been computed, implementations SHOULD delete\n client_/server_application_traffic_secret_N and its associated\n traffic keys.\n\n\n\n\nRescorla Standards Track [Page 94]\n\f\nRFC 8446 TLS August 2018\n\n\n7.3. Traffic Key Calculation\n\n The traffic keying material is generated from the following input\n values:\n\n - A secret value\n\n - A purpose value indicating the specific value being generated\n\n - The length of the key being generated\n\n The traffic keying material is generated from an input traffic secret\n value using:\n\n [sender]_write_key = HKDF-Expand-Label(Secret, \"key\", \"\", key_length)\n [sender]_write_iv = HKDF-Expand-Label(Secret, \"iv\", \"\", iv_length)\n\n [sender] denotes the sending side. The value of Secret for each\n record type is shown in the table below.\n\n +-------------------+---------------------------------------+\n | Record Type | Secret |\n +-------------------+---------------------------------------+\n | 0-RTT Application | client_early_traffic_secret |\n | | |\n | Handshake | [sender]_handshake_traffic_secret |\n | | |\n | Application Data | [sender]_application_traffic_secret_N |\n +-------------------+---------------------------------------+\n\n All the traffic keying material is recomputed whenever the underlying\n Secret changes (e.g., when changing from the handshake to Application\n Data keys or upon a key update).\n\n7.4. (EC)DHE Shared Secret Calculation\n\n7.4.1. Finite Field Diffie-Hellman\n\n For finite field groups, a conventional Diffie-Hellman [DH76]\n computation is performed. The negotiated key (Z) is converted to a\n byte string by encoding in big-endian form and left-padded with zeros\n up to the size of the prime. This byte string is used as the shared\n secret in the key schedule as specified above.\n\n Note that this construction differs from previous versions of TLS\n which removed leading zeros.\n\n\n\n\n\nRescorla Standards Track [Page 95]\n\f\nRFC 8446 TLS August 2018\n\n\n7.4.2. Elliptic Curve Diffie-Hellman\n\n For secp256r1, secp384r1, and secp521r1, ECDH calculations (including\n parameter and key generation as well as the shared secret\n calculation) are performed according to [IEEE1363] using the\n ECKAS-DH1 scheme with the identity map as the key derivation function\n (KDF), so that the shared secret is the x-coordinate of the ECDH\n shared secret elliptic curve point represented as an octet string.\n Note that this octet string (\"Z\" in IEEE 1363 terminology) as output\n by FE2OSP (the Field Element to Octet String Conversion Primitive)\n has constant length for any given field; leading zeros found in this\n octet string MUST NOT be truncated.\n\n (Note that this use of the identity KDF is a technicality. The\n complete picture is that ECDH is employed with a non-trivial KDF\n because TLS does not directly use this secret for anything other than\n for computing other secrets.)\n\n For X25519 and X448, the ECDH calculations are as follows:\n\n - The public key to put into the KeyShareEntry.key_exchange\n structure is the result of applying the ECDH scalar multiplication\n function to the secret key of appropriate length (into scalar\n input) and the standard public basepoint (into u-coordinate point\n input).\n\n - The ECDH shared secret is the result of applying the ECDH scalar\n multiplication function to the secret key (into scalar input) and\n the peer's public key (into u-coordinate point input). The output\n is used raw, with no processing.\n\n For these curves, implementations SHOULD use the approach specified\n in [RFC7748] to calculate the Diffie-Hellman shared secret.\n Implementations MUST check whether the computed Diffie-Hellman shared\n secret is the all-zero value and abort if so, as described in\n Section 6 of [RFC7748]. If implementors use an alternative\n implementation of these elliptic curves, they SHOULD perform the\n additional checks specified in Section 7 of [RFC7748].\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 96]\n\f\nRFC 8446 TLS August 2018\n\n\n7.5. Exporters\n\n [RFC5705] defines keying material exporters for TLS in terms of the\n TLS pseudorandom function (PRF). This document replaces the PRF with\n HKDF, thus requiring a new construction. The exporter interface\n remains the same.\n\n The exporter value is computed as:\n\n TLS-Exporter(label, context_value, key_length) =\n HKDF-Expand-Label(Derive-Secret(Secret, label, \"\"),\n \"exporter\", Hash(context_value), key_length)\n\n Where Secret is either the early_exporter_master_secret or the\n exporter_master_secret. Implementations MUST use the\n exporter_master_secret unless explicitly specified by the\n application. The early_exporter_master_secret is defined for use in\n settings where an exporter is needed for 0-RTT data. A separate\n interface for the early exporter is RECOMMENDED; this avoids the\n exporter user accidentally using an early exporter when a regular one\n is desired or vice versa.\n\n If no context is provided, the context_value is zero length.\n Consequently, providing no context computes the same value as\n providing an empty context. This is a change from previous versions\n of TLS where an empty context produced a different output than an\n absent context. As of this document's publication, no allocated\n exporter label is used both with and without a context. Future\n specifications MUST NOT define a use of exporters that permit both an\n empty context and no context with the same label. New uses of\n exporters SHOULD provide a context in all exporter computations,\n though the value could be empty.\n\n Requirements for the format of exporter labels are defined in\n Section 4 of [RFC5705].\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 97]\n\f\nRFC 8446 TLS August 2018\n\n\n8. 0-RTT and Anti-Replay\n\n As noted in Section 2.3 and Appendix E.5, TLS does not provide\n inherent replay protections for 0-RTT data. There are two potential\n threats to be concerned with:\n\n - Network attackers who mount a replay attack by simply duplicating\n a flight of 0-RTT data.\n\n - Network attackers who take advantage of client retry behavior to\n arrange for the server to receive multiple copies of an\n application message. This threat already exists to some extent\n because clients that value robustness respond to network errors by\n attempting to retry requests. However, 0-RTT adds an additional\n dimension for any server system which does not maintain globally\n consistent server state. Specifically, if a server system has\n multiple zones where tickets from zone A will not be accepted in\n zone B, then an attacker can duplicate a ClientHello and early\n data intended for A to both A and B. At A, the data will be\n accepted in 0-RTT, but at B the server will reject 0-RTT data and\n instead force a full handshake. If the attacker blocks the\n ServerHello from A, then the client will complete the handshake\n with B and probably retry the request, leading to duplication on\n the server system as a whole.\n\n The first class of attack can be prevented by sharing state to\n guarantee that the 0-RTT data is accepted at most once. Servers\n SHOULD provide that level of replay safety by implementing one of the\n methods described in this section or by equivalent means. It is\n understood, however, that due to operational concerns not all\n deployments will maintain state at that level. Therefore, in normal\n operation, clients will not know which, if any, of these mechanisms\n servers actually implement and hence MUST only send early data which\n they deem safe to be replayed.\n\n In addition to the direct effects of replays, there is a class of\n attacks where even operations normally considered idempotent could be\n exploited by a large number of replays (timing attacks, resource\n limit exhaustion and others, as described in Appendix E.5). Those\n can be mitigated by ensuring that every 0-RTT payload can be replayed\n only a limited number of times. The server MUST ensure that any\n instance of it (be it a machine, a thread, or any other entity within\n the relevant serving infrastructure) would accept 0-RTT for the same\n 0-RTT handshake at most once; this limits the number of replays to\n the number of server instances in the deployment. Such a guarantee\n can be accomplished by locally recording data from recently received\n ClientHellos and rejecting repeats, or by any other method that\n\n\n\n\nRescorla Standards Track [Page 98]\n\f\nRFC 8446 TLS August 2018\n\n\n provides the same or a stronger guarantee. The \"at most once per\n server instance\" guarantee is a minimum requirement; servers SHOULD\n limit 0-RTT replays further when feasible.\n\n The second class of attack cannot be prevented at the TLS layer and\n MUST be dealt with by any application. Note that any application\n whose clients implement any kind of retry behavior already needs to\n implement some sort of anti-replay defense.\n\n8.1. Single-Use Tickets\n\n The simplest form of anti-replay defense is for the server to only\n allow each session ticket to be used once. For instance, the server\n can maintain a database of all outstanding valid tickets, deleting\n each ticket from the database as it is used. If an unknown ticket is\n provided, the server would then fall back to a full handshake.\n\n If the tickets are not self-contained but rather are database keys,\n and the corresponding PSKs are deleted upon use, then connections\n established using PSKs enjoy forward secrecy. This improves security\n for all 0-RTT data and PSK usage when PSK is used without (EC)DHE.\n\n Because this mechanism requires sharing the session database between\n server nodes in environments with multiple distributed servers, it\n may be hard to achieve high rates of successful PSK 0-RTT connections\n when compared to self-encrypted tickets. Unlike session databases,\n session tickets can successfully do PSK-based session establishment\n even without consistent storage, though when 0-RTT is allowed they\n still require consistent storage for anti-replay of 0-RTT data, as\n detailed in the following section.\n\n8.2. Client Hello Recording\n\n An alternative form of anti-replay is to record a unique value\n derived from the ClientHello (generally either the random value or\n the PSK binder) and reject duplicates. Recording all ClientHellos\n causes state to grow without bound, but a server can instead record\n ClientHellos within a given time window and use the\n \"obfuscated_ticket_age\" to ensure that tickets aren't reused outside\n that window.\n\n In order to implement this, when a ClientHello is received, the\n server first verifies the PSK binder as described in Section 4.2.11.\n It then computes the expected_arrival_time as described in the next\n section and rejects 0-RTT if it is outside the recording window,\n falling back to the 1-RTT handshake.\n\n\n\n\n\nRescorla Standards Track [Page 99]\n\f\nRFC 8446 TLS August 2018\n\n\n If the expected_arrival_time is in the window, then the server checks\n to see if it has recorded a matching ClientHello. If one is found,\n it either aborts the handshake with an \"illegal_parameter\" alert or\n accepts the PSK but rejects 0-RTT. If no matching ClientHello is\n found, then it accepts 0-RTT and then stores the ClientHello for as\n long as the expected_arrival_time is inside the window. Servers MAY\n also implement data stores with false positives, such as Bloom\n filters, in which case they MUST respond to apparent replay by\n rejecting 0-RTT but MUST NOT abort the handshake.\n\n The server MUST derive the storage key only from validated sections\n of the ClientHello. If the ClientHello contains multiple PSK\n identities, then an attacker can create multiple ClientHellos with\n different binder values for the less-preferred identity on the\n assumption that the server will not verify it (as recommended by\n Section 4.2.11). I.e., if the client sends PSKs A and B but the\n server prefers A, then the attacker can change the binder for B\n without affecting the binder for A. If the binder for B is part of\n the storage key, then this ClientHello will not appear as a\n duplicate, which will cause the ClientHello to be accepted, and may\n cause side effects such as replay cache pollution, although any 0-RTT\n data will not be decryptable because it will use different keys. If\n the validated binder or the ClientHello.random is used as the storage\n key, then this attack is not possible.\n\n Because this mechanism does not require storing all outstanding\n tickets, it may be easier to implement in distributed systems with\n high rates of resumption and 0-RTT, at the cost of potentially weaker\n anti-replay defense because of the difficulty of reliably storing and\n retrieving the received ClientHello messages. In many such systems,\n it is impractical to have globally consistent storage of all the\n received ClientHellos. In this case, the best anti-replay protection\n is provided by having a single storage zone be authoritative for a\n given ticket and refusing 0-RTT for that ticket in any other zone.\n This approach prevents simple replay by the attacker because only one\n zone will accept 0-RTT data. A weaker design is to implement\n separate storage for each zone but allow 0-RTT in any zone. This\n approach limits the number of replays to once per zone. Application\n message duplication of course remains possible with either design.\n\n When implementations are freshly started, they SHOULD reject 0-RTT as\n long as any portion of their recording window overlaps the startup\n time. Otherwise, they run the risk of accepting replays which were\n originally sent during that period.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 100]\n\f\nRFC 8446 TLS August 2018\n\n\n Note: If the client's clock is running much faster than the server's,\n then a ClientHello may be received that is outside the window in the\n future, in which case it might be accepted for 1-RTT, causing a\n client retry, and then acceptable later for 0-RTT. This is another\n variant of the second form of attack described in Section 8.\n\n8.3. Freshness Checks\n\n Because the ClientHello indicates the time at which the client sent\n it, it is possible to efficiently determine whether a ClientHello was\n likely sent reasonably recently and only accept 0-RTT for such a\n ClientHello, otherwise falling back to a 1-RTT handshake. This is\n necessary for the ClientHello storage mechanism described in\n Section 8.2 because otherwise the server needs to store an unlimited\n number of ClientHellos, and is a useful optimization for self-\n contained single-use tickets because it allows efficient rejection of\n ClientHellos which cannot be used for 0-RTT.\n\n In order to implement this mechanism, a server needs to store the\n time that the server generated the session ticket, offset by an\n estimate of the round-trip time between client and server. I.e.,\n\n adjusted_creation_time = creation_time + estimated_RTT\n\n This value can be encoded in the ticket, thus avoiding the need to\n keep state for each outstanding ticket. The server can determine the\n client's view of the age of the ticket by subtracting the ticket's\n \"ticket_age_add\" value from the \"obfuscated_ticket_age\" parameter in\n the client's \"pre_shared_key\" extension. The server can determine\n the expected_arrival_time of the ClientHello as:\n\n expected_arrival_time = adjusted_creation_time + clients_ticket_age\n\n When a new ClientHello is received, the expected_arrival_time is then\n compared against the current server wall clock time and if they\n differ by more than a certain amount, 0-RTT is rejected, though the\n 1-RTT handshake can be allowed to complete.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 101]\n\f\nRFC 8446 TLS August 2018\n\n\n There are several potential sources of error that might cause\n mismatches between the expected_arrival_time and the measured time.\n Variations in client and server clock rates are likely to be minimal,\n though potentially the absolute times may be off by large values.\n Network propagation delays are the most likely causes of a mismatch\n in legitimate values for elapsed time. Both the NewSessionTicket and\n ClientHello messages might be retransmitted and therefore delayed,\n which might be hidden by TCP. For clients on the Internet, this\n implies windows on the order of ten seconds to account for errors in\n clocks and variations in measurements; other deployment scenarios may\n have different needs. Clock skew distributions are not symmetric, so\n the optimal tradeoff may involve an asymmetric range of permissible\n mismatch values.\n\n Note that freshness checking alone is not sufficient to prevent\n replays because it does not detect them during the error window,\n which -- depending on bandwidth and system capacity -- could include\n billions of replays in real-world settings. In addition, this\n freshness checking is only done at the time the ClientHello is\n received and not when subsequent early Application Data records are\n received. After early data is accepted, records may continue to be\n streamed to the server over a longer time period.\n\n9. Compliance Requirements\n\n9.1. Mandatory-to-Implement Cipher Suites\n\n In the absence of an application profile standard specifying\n otherwise:\n\n A TLS-compliant application MUST implement the TLS_AES_128_GCM_SHA256\n [GCM] cipher suite and SHOULD implement the TLS_AES_256_GCM_SHA384\n [GCM] and TLS_CHACHA20_POLY1305_SHA256 [RFC8439] cipher suites (see\n Appendix B.4).\n\n A TLS-compliant application MUST support digital signatures with\n rsa_pkcs1_sha256 (for certificates), rsa_pss_rsae_sha256 (for\n CertificateVerify and certificates), and ecdsa_secp256r1_sha256. A\n TLS-compliant application MUST support key exchange with secp256r1\n (NIST P-256) and SHOULD support key exchange with X25519 [RFC7748].\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 102]\n\f\nRFC 8446 TLS August 2018\n\n\n9.2. Mandatory-to-Implement Extensions\n\n In the absence of an application profile standard specifying\n otherwise, a TLS-compliant application MUST implement the following\n TLS extensions:\n\n - Supported Versions (\"supported_versions\"; Section 4.2.1)\n\n - Cookie (\"cookie\"; Section 4.2.2)\n\n - Signature Algorithms (\"signature_algorithms\"; Section 4.2.3)\n\n - Signature Algorithms Certificate (\"signature_algorithms_cert\";\n Section 4.2.3)\n\n - Negotiated Groups (\"supported_groups\"; Section 4.2.7)\n\n - Key Share (\"key_share\"; Section 4.2.8)\n\n - Server Name Indication (\"server_name\"; Section 3 of [RFC6066])\n\n All implementations MUST send and use these extensions when offering\n applicable features:\n\n - \"supported_versions\" is REQUIRED for all ClientHello, ServerHello,\n and HelloRetryRequest messages.\n\n - \"signature_algorithms\" is REQUIRED for certificate authentication.\n\n - \"supported_groups\" is REQUIRED for ClientHello messages using DHE\n or ECDHE key exchange.\n\n - \"key_share\" is REQUIRED for DHE or ECDHE key exchange.\n\n - \"pre_shared_key\" is REQUIRED for PSK key agreement.\n\n - \"psk_key_exchange_modes\" is REQUIRED for PSK key agreement.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 103]\n\f\nRFC 8446 TLS August 2018\n\n\n A client is considered to be attempting to negotiate using this\n specification if the ClientHello contains a \"supported_versions\"\n extension with 0x0304 contained in its body. Such a ClientHello\n message MUST meet the following requirements:\n\n - If not containing a \"pre_shared_key\" extension, it MUST contain\n both a \"signature_algorithms\" extension and a \"supported_groups\"\n extension.\n\n - If containing a \"supported_groups\" extension, it MUST also contain\n a \"key_share\" extension, and vice versa. An empty\n KeyShare.client_shares vector is permitted.\n\n Servers receiving a ClientHello which does not conform to these\n requirements MUST abort the handshake with a \"missing_extension\"\n alert.\n\n Additionally, all implementations MUST support the use of the\n \"server_name\" extension with applications capable of using it.\n Servers MAY require clients to send a valid \"server_name\" extension.\n Servers requiring this extension SHOULD respond to a ClientHello\n lacking a \"server_name\" extension by terminating the connection with\n a \"missing_extension\" alert.\n\n9.3. Protocol Invariants\n\n This section describes invariants that TLS endpoints and middleboxes\n MUST follow. It also applies to earlier versions of TLS.\n\n TLS is designed to be securely and compatibly extensible. Newer\n clients or servers, when communicating with newer peers, should\n negotiate the most preferred common parameters. The TLS handshake\n provides downgrade protection: Middleboxes passing traffic between a\n newer client and newer server without terminating TLS should be\n unable to influence the handshake (see Appendix E.1). At the same\n time, deployments update at different rates, so a newer client or\n server MAY continue to support older parameters, which would allow it\n to interoperate with older endpoints.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 104]\n\f\nRFC 8446 TLS August 2018\n\n\n For this to work, implementations MUST correctly handle extensible\n fields:\n\n - A client sending a ClientHello MUST support all parameters\n advertised in it. Otherwise, the server may fail to interoperate\n by selecting one of those parameters.\n\n - A server receiving a ClientHello MUST correctly ignore all\n unrecognized cipher suites, extensions, and other parameters.\n Otherwise, it may fail to interoperate with newer clients. In\n TLS 1.3, a client receiving a CertificateRequest or\n NewSessionTicket MUST also ignore all unrecognized extensions.\n\n - A middlebox which terminates a TLS connection MUST behave as a\n compliant TLS server (to the original client), including having a\n certificate which the client is willing to accept, and also as a\n compliant TLS client (to the original server), including verifying\n the original server's certificate. In particular, it MUST\n generate its own ClientHello containing only parameters it\n understands, and it MUST generate a fresh ServerHello random\n value, rather than forwarding the endpoint's value.\n\n Note that TLS's protocol requirements and security analysis only\n apply to the two connections separately. Safely deploying a TLS\n terminator requires additional security considerations which are\n beyond the scope of this document.\n\n - A middlebox which forwards ClientHello parameters it does not\n understand MUST NOT process any messages beyond that ClientHello.\n It MUST forward all subsequent traffic unmodified. Otherwise, it\n may fail to interoperate with newer clients and servers.\n\n Forwarded ClientHellos may contain advertisements for features not\n supported by the middlebox, so the response may include future TLS\n additions the middlebox does not recognize. These additions MAY\n change any message beyond the ClientHello arbitrarily. In\n particular, the values sent in the ServerHello might change, the\n ServerHello format might change, and the TLSCiphertext format\n might change.\n\n The design of TLS 1.3 was constrained by widely deployed\n non-compliant TLS middleboxes (see Appendix D.4); however, it does\n not relax the invariants. Those middleboxes continue to be\n non-compliant.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 105]\n\f\nRFC 8446 TLS August 2018\n\n\n10. Security Considerations\n\n Security issues are discussed throughout this memo, especially in\n Appendices C, D, and E.\n\n11. IANA Considerations\n\n This document uses several registries that were originally created in\n [RFC4346] and updated in [RFC8447]. IANA has updated these to\n reference this document. The registries and their allocation\n policies are below:\n\n - TLS Cipher Suites registry: values with the first byte in the\n range 0-254 (decimal) are assigned via Specification Required\n [RFC8126]. Values with the first byte 255 (decimal) are reserved\n for Private Use [RFC8126].\n\n IANA has added the cipher suites listed in Appendix B.4 to the\n registry. The \"Value\" and \"Description\" columns are taken from\n the table. The \"DTLS-OK\" and \"Recommended\" columns are both\n marked as \"Y\" for each new cipher suite.\n\n - TLS ContentType registry: Future values are allocated via\n Standards Action [RFC8126].\n\n - TLS Alerts registry: Future values are allocated via Standards\n Action [RFC8126]. IANA has populated this registry with the\n values from Appendix B.2. The \"DTLS-OK\" column is marked as \"Y\"\n for all such values. Values marked as \"_RESERVED\" have comments\n describing their previous usage.\n\n - TLS HandshakeType registry: Future values are allocated via\n Standards Action [RFC8126]. IANA has updated this registry to\n rename item 4 from \"NewSessionTicket\" to \"new_session_ticket\" and\n populated this registry with the values from Appendix B.3. The\n \"DTLS-OK\" column is marked as \"Y\" for all such values. Values\n marked \"_RESERVED\" have comments describing their previous or\n temporary usage.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 106]\n\f\nRFC 8446 TLS August 2018\n\n\n This document also uses the TLS ExtensionType Values registry\n originally created in [RFC4366]. IANA has updated it to reference\n this document. Changes to the registry follow:\n\n - IANA has updated the registration policy as follows:\n\n Values with the first byte in the range 0-254 (decimal) are\n assigned via Specification Required [RFC8126]. Values with the\n first byte 255 (decimal) are reserved for Private Use [RFC8126].\n\n - IANA has updated this registry to include the \"key_share\",\n \"pre_shared_key\", \"psk_key_exchange_modes\", \"early_data\",\n \"cookie\", \"supported_versions\", \"certificate_authorities\",\n \"oid_filters\", \"post_handshake_auth\", and\n \"signature_algorithms_cert\" extensions with the values defined in\n this document and the \"Recommended\" value of \"Y\".\n\n - IANA has updated this registry to include a \"TLS 1.3\" column which\n lists the messages in which the extension may appear. This column\n has been initially populated from the table in Section 4.2, with\n any extension not listed there marked as \"-\" to indicate that it\n is not used by TLS 1.3.\n\n This document updates an entry in the TLS Certificate Types registry\n originally created in [RFC6091] and updated in [RFC8447]. IANA has\n updated the entry for value 1 to have the name \"OpenPGP_RESERVED\",\n \"Recommended\" value \"N\", and comment \"Used in TLS versions prior\n to 1.3.\"\n\n This document updates an entry in the TLS Certificate Status Types\n registry originally created in [RFC6961]. IANA has updated the entry\n for value 2 to have the name \"ocsp_multi_RESERVED\" and comment \"Used\n in TLS versions prior to 1.3\".\n\n This document updates two entries in the TLS Supported Groups\n registry (created under a different name by [RFC4492]; now maintained\n by [RFC8422]) and updated by [RFC7919] and [RFC8447]. The entries\n for values 29 and 30 (x25519 and x448) have been updated to also\n refer to this document.\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 107]\n\f\nRFC 8446 TLS August 2018\n\n\n In addition, this document defines two new registries that are\n maintained by IANA:\n\n - TLS SignatureScheme registry: Values with the first byte in the\n range 0-253 (decimal) are assigned via Specification Required\n [RFC8126]. Values with the first byte 254 or 255 (decimal) are\n reserved for Private Use [RFC8126]. Values with the first byte in\n the range 0-6 or with the second byte in the range 0-3 that are\n not currently allocated are reserved for backward compatibility.\n This registry has a \"Recommended\" column. The registry has been\n initially populated with the values described in Section 4.2.3.\n The following values are marked as \"Recommended\":\n ecdsa_secp256r1_sha256, ecdsa_secp384r1_sha384,\n rsa_pss_rsae_sha256, rsa_pss_rsae_sha384, rsa_pss_rsae_sha512,\n rsa_pss_pss_sha256, rsa_pss_pss_sha384, rsa_pss_pss_sha512, and\n ed25519. The \"Recommended\" column is assigned a value of \"N\"\n unless explicitly requested, and adding a value with a\n \"Recommended\" value of \"Y\" requires Standards Action [RFC8126].\n IESG Approval is REQUIRED for a Y->N transition.\n\n - TLS PskKeyExchangeMode registry: Values in the range 0-253\n (decimal) are assigned via Specification Required [RFC8126].\n The values 254 and 255 (decimal) are reserved for Private Use\n [RFC8126]. This registry has a \"Recommended\" column. The\n registry has been initially populated with psk_ke (0) and\n psk_dhe_ke (1). Both are marked as \"Recommended\". The\n \"Recommended\" column is assigned a value of \"N\" unless explicitly\n requested, and adding a value with a \"Recommended\" value of \"Y\"\n requires Standards Action [RFC8126]. IESG Approval is REQUIRED\n for a Y->N transition.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 108]\n\f\nRFC 8446 TLS August 2018\n\n\n12. References\n\n12.1. Normative References\n\n [DH76] Diffie, W. and M. Hellman, \"New directions in\n cryptography\", IEEE Transactions on Information\n Theory, Vol. 22 No. 6, pp. 644-654,\n DOI 10.1109/TIT.1976.1055638, November 1976.\n\n [ECDSA] American National Standards Institute, \"Public Key\n Cryptography for the Financial Services Industry: The\n Elliptic Curve Digital Signature Algorithm (ECDSA)\",\n ANSI ANS X9.62-2005, November 2005.\n\n [GCM] Dworkin, M., \"Recommendation for Block Cipher Modes of\n Operation: Galois/Counter Mode (GCM) and GMAC\",\n NIST Special Publication 800-38D,\n DOI 10.6028/NIST.SP.800-38D, November 2007.\n\n [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, \"HMAC: Keyed-\n Hashing for Message Authentication\", RFC 2104,\n DOI 10.17487/RFC2104, February 1997,\n .\n\n [RFC2119] Bradner, S., \"Key words for use in RFCs to Indicate\n Requirement Levels\", BCP 14, RFC 2119,\n DOI 10.17487/RFC2119, March 1997,\n .\n\n [RFC5116] McGrew, D., \"An Interface and Algorithms for Authenticated\n Encryption\", RFC 5116, DOI 10.17487/RFC5116, January 2008,\n .\n\n [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,\n Housley, R., and W. Polk, \"Internet X.509 Public Key\n Infrastructure Certificate and Certificate Revocation List\n (CRL) Profile\", RFC 5280, DOI 10.17487/RFC5280, May 2008,\n .\n\n [RFC5705] Rescorla, E., \"Keying Material Exporters for Transport\n Layer Security (TLS)\", RFC 5705, DOI 10.17487/RFC5705,\n March 2010, .\n\n [RFC5756] Turner, S., Brown, D., Yiu, K., Housley, R., and T. Polk,\n \"Updates for RSAES-OAEP and RSASSA-PSS Algorithm\n Parameters\", RFC 5756, DOI 10.17487/RFC5756, January 2010,\n .\n\n\n\n\nRescorla Standards Track [Page 109]\n\f\nRFC 8446 TLS August 2018\n\n\n [RFC5869] Krawczyk, H. and P. Eronen, \"HMAC-based Extract-and-Expand\n Key Derivation Function (HKDF)\", RFC 5869,\n DOI 10.17487/RFC5869, May 2010,\n .\n\n [RFC6066] Eastlake 3rd, D., \"Transport Layer Security (TLS)\n Extensions: Extension Definitions\", RFC 6066,\n DOI 10.17487/RFC6066, January 2011,\n .\n\n [RFC6655] McGrew, D. and D. Bailey, \"AES-CCM Cipher Suites for\n Transport Layer Security (TLS)\", RFC 6655,\n DOI 10.17487/RFC6655, July 2012,\n .\n\n [RFC6960] Santesson, S., Myers, M., Ankney, R., Malpani, A.,\n Galperin, S., and C. Adams, \"X.509 Internet Public Key\n Infrastructure Online Certificate Status Protocol - OCSP\",\n RFC 6960, DOI 10.17487/RFC6960, June 2013,\n .\n\n [RFC6961] Pettersen, Y., \"The Transport Layer Security (TLS)\n Multiple Certificate Status Request Extension\", RFC 6961,\n DOI 10.17487/RFC6961, June 2013,\n .\n\n [RFC6962] Laurie, B., Langley, A., and E. Kasper, \"Certificate\n Transparency\", RFC 6962, DOI 10.17487/RFC6962, June 2013,\n .\n\n [RFC6979] Pornin, T., \"Deterministic Usage of the Digital Signature\n Algorithm (DSA) and Elliptic Curve Digital Signature\n Algorithm (ECDSA)\", RFC 6979, DOI 10.17487/RFC6979,\n August 2013, .\n\n [RFC7301] Friedl, S., Popov, A., Langley, A., and E. Stephan,\n \"Transport Layer Security (TLS) Application-Layer Protocol\n Negotiation Extension\", RFC 7301, DOI 10.17487/RFC7301,\n July 2014, .\n\n [RFC7507] Moeller, B. and A. Langley, \"TLS Fallback Signaling Cipher\n Suite Value (SCSV) for Preventing Protocol Downgrade\n Attacks\", RFC 7507, DOI 10.17487/RFC7507, April 2015,\n .\n\n [RFC7748] Langley, A., Hamburg, M., and S. Turner, \"Elliptic Curves\n for Security\", RFC 7748, DOI 10.17487/RFC7748,\n January 2016, .\n\n\n\nRescorla Standards Track [Page 110]\n\f\nRFC 8446 TLS August 2018\n\n\n [RFC7919] Gillmor, D., \"Negotiated Finite Field Diffie-Hellman\n Ephemeral Parameters for Transport Layer Security (TLS)\",\n RFC 7919, DOI 10.17487/RFC7919, August 2016,\n .\n\n [RFC8017] Moriarty, K., Ed., Kaliski, B., Jonsson, J., and A. Rusch,\n \"PKCS #1: RSA Cryptography Specifications Version 2.2\",\n RFC 8017, DOI 10.17487/RFC8017, November 2016,\n .\n\n [RFC8032] Josefsson, S. and I. Liusvaara, \"Edwards-Curve Digital\n Signature Algorithm (EdDSA)\", RFC 8032,\n DOI 10.17487/RFC8032, January 2017,\n .\n\n [RFC8126] Cotton, M., Leiba, B., and T. Narten, \"Guidelines for\n Writing an IANA Considerations Section in RFCs\", BCP 26,\n RFC 8126, DOI 10.17487/RFC8126, June 2017,\n .\n\n [RFC8174] Leiba, B., \"Ambiguity of Uppercase vs Lowercase in\n RFC 2119 Key Words\", BCP 14, RFC 8174,\n DOI 10.17487/RFC8174, May 2017,\n .\n\n [RFC8439] Nir, Y. and A. Langley, \"ChaCha20 and Poly1305 for IETF\n Protocols\", RFC 8439, DOI 10.17487/RFC8439, June 2018,\n .\n\n [SHS] Dang, Q., \"Secure Hash Standard (SHS)\", National Institute\n of Standards and Technology report,\n DOI 10.6028/NIST.FIPS.180-4, August 2015.\n\n [X690] ITU-T, \"Information technology -- ASN.1 encoding rules:\n Specification of Basic Encoding Rules (BER), Canonical\n Encoding Rules (CER) and Distinguished Encoding Rules\n (DER)\", ISO/IEC 8825-1:2015, November 2015.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 111]\n\f\nRFC 8446 TLS August 2018\n\n\n12.2. Informative References\n\n [AEAD-LIMITS]\n Luykx, A. and K. Paterson, \"Limits on Authenticated\n Encryption Use in TLS\", August 2017,\n .\n\n [BBFGKZ16]\n Bhargavan, K., Brzuska, C., Fournet, C., Green, M.,\n Kohlweiss, M., and S. Zanella-Beguelin, \"Downgrade\n Resilience in Key-Exchange Protocols\", Proceedings of IEEE\n Symposium on Security and Privacy (San Jose),\n DOI 10.1109/SP.2016.37, May 2016.\n\n [BBK17] Bhargavan, K., Blanchet, B., and N. Kobeissi, \"Verified\n Models and Reference Implementations for the TLS 1.3\n Standard Candidate\", Proceedings of IEEE Symposium on\n Security and Privacy (San Jose), DOI 10.1109/SP.2017.26,\n May 2017.\n\n [BDFKPPRSZZ16]\n Bhargavan, K., Delignat-Lavaud, A., Fournet, C.,\n Kohlweiss, M., Pan, J., Protzenko, J., Rastogi, A., Swamy,\n N., Zanella-Beguelin, S., and J. Zinzindohoue,\n \"Implementing and Proving the TLS 1.3 Record Layer\",\n Proceedings of IEEE Symposium on Security and Privacy (San\n Jose), May 2017, .\n\n [Ben17a] Benjamin, D., \"Presentation before the TLS WG at\n IETF 100\", November 2017,\n .\n\n [Ben17b] Benjamin, D., \"Additional TLS 1.3 results from Chrome\",\n message to the TLS mailing list, 18 December 2017,\n .\n\n [Blei98] Bleichenbacher, D., \"Chosen Ciphertext Attacks against\n Protocols Based on RSA Encryption Standard PKCS #1\",\n Proceedings of CRYPTO '98, 1998.\n\n [BMMRT15] Badertscher, C., Matt, C., Maurer, U., Rogaway, P., and B.\n Tackmann, \"Augmented Secure Channels and the Goal of the\n TLS 1.3 Record Layer\", ProvSec 2015, September 2015,\n .\n\n\n\n\n\nRescorla Standards Track [Page 112]\n\f\nRFC 8446 TLS August 2018\n\n\n [BT16] Bellare, M. and B. Tackmann, \"The Multi-User Security of\n Authenticated Encryption: AES-GCM in TLS 1.3\", Proceedings\n of CRYPTO 2016, July 2016,\n .\n\n [CCG16] Cohn-Gordon, K., Cremers, C., and L. Garratt, \"On\n Post-compromise Security\", IEEE Computer Security\n Foundations Symposium, DOI 10.1109/CSF.2016.19, July 2015.\n\n [CHECKOWAY]\n Checkoway, S., Maskiewicz, J., Garman, C., Fried, J.,\n Cohney, S., Green, M., Heninger, N., Weinmann, R.,\n Rescorla, E., and H. Shacham, \"A Systematic Analysis of\n the Juniper Dual EC Incident\", Proceedings of the 2016 ACM\n SIGSAC Conference on Computer and Communications Security\n - CCS '16, DOI 10.1145/2976749.2978395, October 2016.\n\n [CHHSV17] Cremers, C., Horvat, M., Hoyland, J., Scott, S., and T.\n van der Merwe, \"Awkward Handshake: Possible mismatch of\n client/server view on client authentication in\n post-handshake mode in Revision 18\", message to the TLS\n mailing list, 10 February 2017, .\n\n [CHSV16] Cremers, C., Horvat, M., Scott, S., and T. van der Merwe,\n \"Automated Analysis and Verification of TLS 1.3: 0-RTT,\n Resumption and Delayed Authentication\", Proceedings of\n IEEE Symposium on Security and Privacy (San Jose),\n DOI 10.1109/SP.2016.35, May 2016,\n .\n\n [CK01] Canetti, R. and H. Krawczyk, \"Analysis of Key-Exchange\n Protocols and Their Use for Building Secure Channels\",\n Proceedings of Eurocrypt 2001,\n DOI 10.1007/3-540-44987-6_28, April 2001.\n\n [CLINIC] Miller, B., Huang, L., Joseph, A., and J. Tygar, \"I Know\n Why You Went to the Clinic: Risks and Realization of HTTPS\n Traffic Analysis\", Privacy Enhancing Technologies, pp.\n 143-163, DOI 10.1007/978-3-319-08506-7_8, 2014.\n\n [DFGS15] Dowling, B., Fischlin, M., Guenther, F., and D. Stebila,\n \"A Cryptographic Analysis of the TLS 1.3 Handshake\n Protocol Candidates\", Proceedings of ACM CCS 2015,\n October 2015, .\n\n\n\n\n\n\nRescorla Standards Track [Page 113]\n\f\nRFC 8446 TLS August 2018\n\n\n [DFGS16] Dowling, B., Fischlin, M., Guenther, F., and D. Stebila,\n \"A Cryptographic Analysis of the TLS 1.3 Full and\n Pre-shared Key Handshake Protocol\", TRON 2016,\n February 2016, .\n\n [DOW92] Diffie, W., van Oorschot, P., and M. Wiener,\n \"Authentication and authenticated key exchanges\", Designs,\n Codes and Cryptography, DOI 10.1007/BF00124891, June 1992.\n\n [DSS] National Institute of Standards and Technology, U.S.\n Department of Commerce, \"Digital Signature Standard\n (DSS)\", NIST FIPS PUB 186-4, DOI 10.6028/NIST.FIPS.186-4,\n July 2013.\n\n [FG17] Fischlin, M. and F. Guenther, \"Replay Attacks on Zero\n Round-Trip Time: The Case of the TLS 1.3 Handshake\n Candidates\", Proceedings of EuroS&P 2017, April 2017,\n .\n\n [FGSW16] Fischlin, M., Guenther, F., Schmidt, B., and B. Warinschi,\n \"Key Confirmation in Key Exchange: A Formal Treatment and\n Implications for TLS 1.3\", Proceedings of IEEE Symposium\n on Security and Privacy (San Jose),\n DOI 10.1109/SP.2016.34, May 2016,\n .\n\n [FW15] Weimer, F., \"Factoring RSA Keys With TLS Perfect Forward\n Secrecy\", September 2015.\n\n [HCJC16] Husak, M., Cermak, M., Jirsik, T., and P. Celeda, \"HTTPS\n traffic analysis and client identification using passive\n SSL/TLS fingerprinting\", EURASIP Journal on Information\n Security, Vol. 2016, DOI 10.1186/s13635-016-0030-7,\n February 2016.\n\n [HGFS15] Hlauschek, C., Gruber, M., Fankhauser, F., and C. Schanes,\n \"Prying Open Pandora's Box: KCI Attacks against TLS\",\n Proceedings of USENIX Workshop on Offensive Technologies,\n August 2015.\n\n [IEEE1363]\n IEEE, \"IEEE Standard Specifications for Public Key\n Cryptography\", IEEE Std. 1363-2000,\n DOI 10.1109/IEEESTD.2000.92292.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 114]\n\f\nRFC 8446 TLS August 2018\n\n\n [JSS15] Jager, T., Schwenk, J., and J. Somorovsky, \"On the\n Security of TLS 1.3 and QUIC Against Weaknesses in PKCS#1\n v1.5 Encryption\", Proceedings of ACM CCS 2015,\n DOI 10.1145/2810103.2813657, October 2015,\n .\n\n [KEYAGREEMENT]\n Barker, E., Chen, L., Roginsky, A., Vassilev, A., and R.\n Davis, \"Recommendation for Pair-Wise Key Establishment\n Schemes Using Discrete Logarithm Cryptography\", National\n Institute of Standards and Technology,\n DOI 10.6028/NIST.SP.800-56Ar3, April 2018.\n\n [Kraw10] Krawczyk, H., \"Cryptographic Extraction and Key\n Derivation: The HKDF Scheme\", Proceedings of CRYPTO 2010,\n August 2010, .\n\n [Kraw16] Krawczyk, H., \"A Unilateral-to-Mutual Authentication\n Compiler for Key Exchange (with Applications to Client\n Authentication in TLS 1.3\", Proceedings of ACM CCS 2016,\n October 2016, .\n\n [KW16] Krawczyk, H. and H. Wee, \"The OPTLS Protocol and TLS 1.3\",\n Proceedings of EuroS&P 2016, March 2016,\n .\n\n [LXZFH16] Li, X., Xu, J., Zhang, Z., Feng, D., and H. Hu, \"Multiple\n Handshakes Security of TLS 1.3 Candidates\", Proceedings of\n IEEE Symposium on Security and Privacy (San Jose),\n DOI 10.1109/SP.2016.36, May 2016,\n .\n\n [Mac17] MacCarthaigh, C., \"Security Review of TLS1.3 0-RTT\",\n March 2017, .\n\n [PS18] Patton, C. and T. Shrimpton, \"Partially specified\n channels: The TLS 1.3 record layer without elision\", 2018,\n .\n\n [PSK-FINISHED]\n Scott, S., Cremers, C., Horvat, M., and T. van der Merwe,\n \"Revision 10: possible attack if client authentication is\n allowed during PSK\", message to the TLS mailing list,\n 31 October 2015, .\n\n\n\n\nRescorla Standards Track [Page 115]\n\f\nRFC 8446 TLS August 2018\n\n\n [REKEY] Abdalla, M. and M. Bellare, \"Increasing the Lifetime of a\n Key: A Comparative Analysis of the Security of Re-keying\n Techniques\", ASIACRYPT 2000, DOI 10.1007/3-540-44448-3_42,\n October 2000.\n\n [Res17a] Rescorla, E., \"Preliminary data on Firefox TLS 1.3\n Middlebox experiment\", message to the TLS mailing list,\n 5 December 2017, .\n\n [Res17b] Rescorla, E., \"More compatibility measurement results\",\n message to the TLS mailing list, 22 December 2017,\n .\n\n [RFC3552] Rescorla, E. and B. Korver, \"Guidelines for Writing RFC\n Text on Security Considerations\", BCP 72, RFC 3552,\n DOI 10.17487/RFC3552, July 2003,\n .\n\n [RFC4086] Eastlake 3rd, D., Schiller, J., and S. Crocker,\n \"Randomness Requirements for Security\", BCP 106, RFC 4086,\n DOI 10.17487/RFC4086, June 2005,\n .\n\n [RFC4346] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.1\", RFC 4346,\n DOI 10.17487/RFC4346, April 2006,\n .\n\n [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J.,\n and T. Wright, \"Transport Layer Security (TLS)\n Extensions\", RFC 4366, DOI 10.17487/RFC4366, April 2006,\n .\n\n [RFC4492] Blake-Wilson, S., Bolyard, N., Gupta, V., Hawk, C., and B.\n Moeller, \"Elliptic Curve Cryptography (ECC) Cipher Suites\n for Transport Layer Security (TLS)\", RFC 4492,\n DOI 10.17487/RFC4492, May 2006,\n .\n\n [RFC5077] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,\n \"Transport Layer Security (TLS) Session Resumption without\n Server-Side State\", RFC 5077, DOI 10.17487/RFC5077,\n January 2008, .\n\n\n\n\n\n\nRescorla Standards Track [Page 116]\n\f\nRFC 8446 TLS August 2018\n\n\n [RFC5246] Dierks, T. and E. Rescorla, \"The Transport Layer Security\n (TLS) Protocol Version 1.2\", RFC 5246,\n DOI 10.17487/RFC5246, August 2008,\n .\n\n [RFC5764] McGrew, D. and E. Rescorla, \"Datagram Transport Layer\n Security (DTLS) Extension to Establish Keys for the Secure\n Real-time Transport Protocol (SRTP)\", RFC 5764,\n DOI 10.17487/RFC5764, May 2010,\n .\n\n [RFC5929] Altman, J., Williams, N., and L. Zhu, \"Channel Bindings\n for TLS\", RFC 5929, DOI 10.17487/RFC5929, July 2010,\n .\n\n [RFC6091] Mavrogiannopoulos, N. and D. Gillmor, \"Using OpenPGP Keys\n for Transport Layer Security (TLS) Authentication\",\n RFC 6091, DOI 10.17487/RFC6091, February 2011,\n .\n\n [RFC6101] Freier, A., Karlton, P., and P. Kocher, \"The Secure\n Sockets Layer (SSL) Protocol Version 3.0\", RFC 6101,\n DOI 10.17487/RFC6101, August 2011,\n .\n\n [RFC6176] Turner, S. and T. Polk, \"Prohibiting Secure Sockets Layer\n (SSL) Version 2.0\", RFC 6176, DOI 10.17487/RFC6176,\n March 2011, .\n\n [RFC6347] Rescorla, E. and N. Modadugu, \"Datagram Transport Layer\n Security Version 1.2\", RFC 6347, DOI 10.17487/RFC6347,\n January 2012, .\n\n [RFC6520] Seggelmann, R., Tuexen, M., and M. Williams, \"Transport\n Layer Security (TLS) and Datagram Transport Layer Security\n (DTLS) Heartbeat Extension\", RFC 6520,\n DOI 10.17487/RFC6520, February 2012,\n .\n\n [RFC7230] Fielding, R., Ed. and J. Reschke, Ed., \"Hypertext Transfer\n Protocol (HTTP/1.1): Message Syntax and Routing\",\n RFC 7230, DOI 10.17487/RFC7230, June 2014,\n .\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 117]\n\f\nRFC 8446 TLS August 2018\n\n\n [RFC7250] Wouters, P., Ed., Tschofenig, H., Ed., Gilmore, J.,\n Weiler, S., and T. Kivinen, \"Using Raw Public Keys in\n Transport Layer Security (TLS) and Datagram Transport\n Layer Security (DTLS)\", RFC 7250, DOI 10.17487/RFC7250,\n June 2014, .\n\n [RFC7465] Popov, A., \"Prohibiting RC4 Cipher Suites\", RFC 7465,\n DOI 10.17487/RFC7465, February 2015,\n .\n\n [RFC7568] Barnes, R., Thomson, M., Pironti, A., and A. Langley,\n \"Deprecating Secure Sockets Layer Version 3.0\", RFC 7568,\n DOI 10.17487/RFC7568, June 2015,\n .\n\n [RFC7627] Bhargavan, K., Ed., Delignat-Lavaud, A., Pironti, A.,\n Langley, A., and M. Ray, \"Transport Layer Security (TLS)\n Session Hash and Extended Master Secret Extension\",\n RFC 7627, DOI 10.17487/RFC7627, September 2015,\n .\n\n [RFC7685] Langley, A., \"A Transport Layer Security (TLS) ClientHello\n Padding Extension\", RFC 7685, DOI 10.17487/RFC7685,\n October 2015, .\n\n [RFC7924] Santesson, S. and H. Tschofenig, \"Transport Layer Security\n (TLS) Cached Information Extension\", RFC 7924,\n DOI 10.17487/RFC7924, July 2016,\n .\n\n [RFC8305] Schinazi, D. and T. Pauly, \"Happy Eyeballs Version 2:\n Better Connectivity Using Concurrency\", RFC 8305,\n DOI 10.17487/RFC8305, December 2017,\n .\n\n [RFC8422] Nir, Y., Josefsson, S., and M. Pegourie-Gonnard, \"Elliptic\n Curve Cryptography (ECC) Cipher Suites for Transport Layer\n Security (TLS) Versions 1.2 and Earlier\", RFC 8422,\n DOI 10.17487/RFC8422, August 2018,\n .\n\n [RFC8447] Salowey, J. and S. Turner, \"IANA Registry Updates for TLS\n and DTLS\", RFC 8447, DOI 10.17487/RFC8447, August 2018,\n .\n\n [RFC8449] Thomson, M., \"Record Size Limit Extension for TLS\",\n RFC 8449, DOI 10.17487/RFC8449, August 2018,\n .\n\n\n\nRescorla Standards Track [Page 118]\n\f\nRFC 8446 TLS August 2018\n\n\n [RSA] Rivest, R., Shamir, A., and L. Adleman, \"A Method for\n Obtaining Digital Signatures and Public-Key\n Cryptosystems\", Communications of the ACM, Vol. 21 No. 2,\n pp. 120-126, DOI 10.1145/359340.359342, February 1978.\n\n [SIGMA] Krawczyk, H., \"SIGMA: The 'SIGn-and-MAc' Approach to\n Authenticated Diffie-Hellman and its Use in the IKE\n Protocols\", Proceedings of CRYPTO 2003,\n DOI 10.1007/978-3-540-45146-4_24, August 2003.\n\n [SLOTH] Bhargavan, K. and G. Leurent, \"Transcript Collision\n Attacks: Breaking Authentication in TLS, IKE, and SSH\",\n Network and Distributed System Security\n Symposium (NDSS 2016), DOI 10.14722/ndss.2016.23418,\n February 2016.\n\n [SSL2] Hickman, K., \"The SSL Protocol\", February 1995.\n\n [TIMING] Boneh, D. and D. Brumley, \"Remote Timing Attacks Are\n Practical\", USENIX Security Symposium, August 2003.\n\n [TLS13-TRACES]\n Thomson, M., \"Example Handshake Traces for TLS 1.3\", Work\n in Progress, draft-ietf-tls-tls13-vectors-06, July 2018.\n\n [X501] ITU-T, \"Information Technology - Open Systems\n Interconnection - The Directory: Models\", ITU-T X.501,\n October 2016, .\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 119]\n\f\nRFC 8446 TLS August 2018\n\n\nAppendix A. State Machine\n\n This appendix provides a summary of the legal state transitions for\n the client and server handshakes. State names (in all capitals,\n e.g., START) have no formal meaning but are provided for ease of\n comprehension. Actions which are taken only in certain circumstances\n are indicated in []. The notation \"K_{send,recv} = foo\" means \"set\n the send/recv key to the given key\".\n\nA.1. Client\n\n START <----+\n Send ClientHello | | Recv HelloRetryRequest\n [K_send = early data] | |\n v |\n / WAIT_SH ----+\n | | Recv ServerHello\n | | K_recv = handshake\n Can | V\n send | WAIT_EE\n early | | Recv EncryptedExtensions\n data | +--------+--------+\n | Using | | Using certificate\n | PSK | v\n | | WAIT_CERT_CR\n | | Recv | | Recv CertificateRequest\n | | Certificate | v\n | | | WAIT_CERT\n | | | | Recv Certificate\n | | v v\n | | WAIT_CV\n | | | Recv CertificateVerify\n | +> WAIT_FINISHED <+\n | | Recv Finished\n \\ | [Send EndOfEarlyData]\n | K_send = handshake\n | [Send Certificate [+ CertificateVerify]]\n Can send | Send Finished\n app data --> | K_send = K_recv = application\n after here v\n CONNECTED\n\n Note that with the transitions as shown above, clients may send\n alerts that derive from post-ServerHello messages in the clear or\n with the early data keys. If clients need to send such alerts, they\n SHOULD first rekey to the handshake keys if possible.\n\n\n\n\n\nRescorla Standards Track [Page 120]\n\f\nRFC 8446 TLS August 2018\n\n\nA.2. Server\n\n START <-----+\n Recv ClientHello | | Send HelloRetryRequest\n v |\n RECVD_CH ----+\n | Select parameters\n v\n NEGOTIATED\n | Send ServerHello\n | K_send = handshake\n | Send EncryptedExtensions\n | [Send CertificateRequest]\n Can send | [Send Certificate + CertificateVerify]\n app data | Send Finished\n after --> | K_send = application\n here +--------+--------+\n No 0-RTT | | 0-RTT\n | |\n K_recv = handshake | | K_recv = early data\n [Skip decrypt errors] | +------> WAIT_EOED -+\n | | Recv | | Recv EndOfEarlyData\n | | early data | | K_recv = handshake\n | +------------+ |\n | |\n +> WAIT_FLIGHT2 <--------+\n |\n +--------+--------+\n No auth | | Client auth\n | |\n | v\n | WAIT_CERT\n | Recv | | Recv Certificate\n | empty | v\n | Certificate | WAIT_CV\n | | | Recv\n | v | CertificateVerify\n +-> WAIT_FINISHED <---+\n | Recv Finished\n | K_recv = application\n v\n CONNECTED\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 121]\n\f\nRFC 8446 TLS August 2018\n\n\nAppendix B. Protocol Data Structures and Constant Values\n\n This appendix provides the normative protocol types and the\n definitions for constants. Values listed as \"_RESERVED\" were used in\n previous versions of TLS and are listed here for completeness.\n TLS 1.3 implementations MUST NOT send them but might receive them\n from older TLS implementations.\n\nB.1. Record Layer\n\n enum {\n invalid(0),\n change_cipher_spec(20),\n alert(21),\n handshake(22),\n application_data(23),\n heartbeat(24), /* RFC 6520 */\n (255)\n } ContentType;\n\n struct {\n ContentType type;\n ProtocolVersion legacy_record_version;\n uint16 length;\n opaque fragment[TLSPlaintext.length];\n } TLSPlaintext;\n\n struct {\n opaque content[TLSPlaintext.length];\n ContentType type;\n uint8 zeros[length_of_padding];\n } TLSInnerPlaintext;\n\n struct {\n ContentType opaque_type = application_data; /* 23 */\n ProtocolVersion legacy_record_version = 0x0303; /* TLS v1.2 */\n uint16 length;\n opaque encrypted_record[TLSCiphertext.length];\n } TLSCiphertext;\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 122]\n\f\nRFC 8446 TLS August 2018\n\n\nB.2. Alert Messages\n\n enum { warning(1), fatal(2), (255) } AlertLevel;\n\n enum {\n close_notify(0),\n unexpected_message(10),\n bad_record_mac(20),\n decryption_failed_RESERVED(21),\n record_overflow(22),\n decompression_failure_RESERVED(30),\n handshake_failure(40),\n no_certificate_RESERVED(41),\n bad_certificate(42),\n unsupported_certificate(43),\n certificate_revoked(44),\n certificate_expired(45),\n certificate_unknown(46),\n illegal_parameter(47),\n unknown_ca(48),\n access_denied(49),\n decode_error(50),\n decrypt_error(51),\n export_restriction_RESERVED(60),\n protocol_version(70),\n insufficient_security(71),\n internal_error(80),\n inappropriate_fallback(86),\n user_canceled(90),\n no_renegotiation_RESERVED(100),\n missing_extension(109),\n unsupported_extension(110),\n certificate_unobtainable_RESERVED(111),\n unrecognized_name(112),\n bad_certificate_status_response(113),\n bad_certificate_hash_value_RESERVED(114),\n unknown_psk_identity(115),\n certificate_required(116),\n no_application_protocol(120),\n (255)\n } AlertDescription;\n\n struct {\n AlertLevel level;\n AlertDescription description;\n } Alert;\n\n\n\n\n\nRescorla Standards Track [Page 123]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3. Handshake Protocol\n\n enum {\n hello_request_RESERVED(0),\n client_hello(1),\n server_hello(2),\n hello_verify_request_RESERVED(3),\n new_session_ticket(4),\n end_of_early_data(5),\n hello_retry_request_RESERVED(6),\n encrypted_extensions(8),\n certificate(11),\n server_key_exchange_RESERVED(12),\n certificate_request(13),\n server_hello_done_RESERVED(14),\n certificate_verify(15),\n client_key_exchange_RESERVED(16),\n finished(20),\n certificate_url_RESERVED(21),\n certificate_status_RESERVED(22),\n supplemental_data_RESERVED(23),\n key_update(24),\n message_hash(254),\n (255)\n } HandshakeType;\n\n struct {\n HandshakeType msg_type; /* handshake type */\n uint24 length; /* bytes in message */\n select (Handshake.msg_type) {\n case client_hello: ClientHello;\n case server_hello: ServerHello;\n case end_of_early_data: EndOfEarlyData;\n case encrypted_extensions: EncryptedExtensions;\n case certificate_request: CertificateRequest;\n case certificate: Certificate;\n case certificate_verify: CertificateVerify;\n case finished: Finished;\n case new_session_ticket: NewSessionTicket;\n case key_update: KeyUpdate;\n };\n } Handshake;\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 124]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.1. Key Exchange Messages\n\n uint16 ProtocolVersion;\n opaque Random[32];\n\n uint8 CipherSuite[2]; /* Cryptographic suite selector */\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id<0..32>;\n CipherSuite cipher_suites<2..2^16-2>;\n opaque legacy_compression_methods<1..2^8-1>;\n Extension extensions<8..2^16-1>;\n } ClientHello;\n\n struct {\n ProtocolVersion legacy_version = 0x0303; /* TLS v1.2 */\n Random random;\n opaque legacy_session_id_echo<0..32>;\n CipherSuite cipher_suite;\n uint8 legacy_compression_method = 0;\n Extension extensions<6..2^16-1>;\n } ServerHello;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 125]\n\f\nRFC 8446 TLS August 2018\n\n\n struct {\n ExtensionType extension_type;\n opaque extension_data<0..2^16-1>;\n } Extension;\n\n enum {\n server_name(0), /* RFC 6066 */\n max_fragment_length(1), /* RFC 6066 */\n status_request(5), /* RFC 6066 */\n supported_groups(10), /* RFC 8422, 7919 */\n signature_algorithms(13), /* RFC 8446 */\n use_srtp(14), /* RFC 5764 */\n heartbeat(15), /* RFC 6520 */\n application_layer_protocol_negotiation(16), /* RFC 7301 */\n signed_certificate_timestamp(18), /* RFC 6962 */\n client_certificate_type(19), /* RFC 7250 */\n server_certificate_type(20), /* RFC 7250 */\n padding(21), /* RFC 7685 */\n RESERVED(40), /* Used but never\n assigned */\n pre_shared_key(41), /* RFC 8446 */\n early_data(42), /* RFC 8446 */\n supported_versions(43), /* RFC 8446 */\n cookie(44), /* RFC 8446 */\n psk_key_exchange_modes(45), /* RFC 8446 */\n RESERVED(46), /* Used but never\n assigned */\n certificate_authorities(47), /* RFC 8446 */\n oid_filters(48), /* RFC 8446 */\n post_handshake_auth(49), /* RFC 8446 */\n signature_algorithms_cert(50), /* RFC 8446 */\n key_share(51), /* RFC 8446 */\n (65535)\n } ExtensionType;\n\n struct {\n NamedGroup group;\n opaque key_exchange<1..2^16-1>;\n } KeyShareEntry;\n\n struct {\n KeyShareEntry client_shares<0..2^16-1>;\n } KeyShareClientHello;\n\n struct {\n NamedGroup selected_group;\n } KeyShareHelloRetryRequest;\n\n\n\n\nRescorla Standards Track [Page 126]\n\f\nRFC 8446 TLS August 2018\n\n\n struct {\n KeyShareEntry server_share;\n } KeyShareServerHello;\n\n struct {\n uint8 legacy_form = 4;\n opaque X[coordinate_length];\n opaque Y[coordinate_length];\n } UncompressedPointRepresentation;\n\n enum { psk_ke(0), psk_dhe_ke(1), (255) } PskKeyExchangeMode;\n\n struct {\n PskKeyExchangeMode ke_modes<1..255>;\n } PskKeyExchangeModes;\n\n struct {} Empty;\n\n struct {\n select (Handshake.msg_type) {\n case new_session_ticket: uint32 max_early_data_size;\n case client_hello: Empty;\n case encrypted_extensions: Empty;\n };\n } EarlyDataIndication;\n\n struct {\n opaque identity<1..2^16-1>;\n uint32 obfuscated_ticket_age;\n } PskIdentity;\n\n opaque PskBinderEntry<32..255>;\n\n struct {\n PskIdentity identities<7..2^16-1>;\n PskBinderEntry binders<33..2^16-1>;\n } OfferedPsks;\n\n struct {\n select (Handshake.msg_type) {\n case client_hello: OfferedPsks;\n case server_hello: uint16 selected_identity;\n };\n } PreSharedKeyExtension;\n\n\n\n\n\n\n\nRescorla Standards Track [Page 127]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.1.1. Version Extension\n\n struct {\n select (Handshake.msg_type) {\n case client_hello:\n ProtocolVersion versions<2..254>;\n\n case server_hello: /* and HelloRetryRequest */\n ProtocolVersion selected_version;\n };\n } SupportedVersions;\n\nB.3.1.2. Cookie Extension\n\n struct {\n opaque cookie<1..2^16-1>;\n } Cookie;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 128]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.1.3. Signature Algorithm Extension\n\n enum {\n /* RSASSA-PKCS1-v1_5 algorithms */\n rsa_pkcs1_sha256(0x0401),\n rsa_pkcs1_sha384(0x0501),\n rsa_pkcs1_sha512(0x0601),\n\n /* ECDSA algorithms */\n ecdsa_secp256r1_sha256(0x0403),\n ecdsa_secp384r1_sha384(0x0503),\n ecdsa_secp521r1_sha512(0x0603),\n\n /* RSASSA-PSS algorithms with public key OID rsaEncryption */\n rsa_pss_rsae_sha256(0x0804),\n rsa_pss_rsae_sha384(0x0805),\n rsa_pss_rsae_sha512(0x0806),\n\n /* EdDSA algorithms */\n ed25519(0x0807),\n ed448(0x0808),\n\n /* RSASSA-PSS algorithms with public key OID RSASSA-PSS */\n rsa_pss_pss_sha256(0x0809),\n rsa_pss_pss_sha384(0x080a),\n rsa_pss_pss_sha512(0x080b),\n\n /* Legacy algorithms */\n rsa_pkcs1_sha1(0x0201),\n ecdsa_sha1(0x0203),\n\n /* Reserved Code Points */\n obsolete_RESERVED(0x0000..0x0200),\n dsa_sha1_RESERVED(0x0202),\n obsolete_RESERVED(0x0204..0x0400),\n dsa_sha256_RESERVED(0x0402),\n obsolete_RESERVED(0x0404..0x0500),\n dsa_sha384_RESERVED(0x0502),\n obsolete_RESERVED(0x0504..0x0600),\n dsa_sha512_RESERVED(0x0602),\n obsolete_RESERVED(0x0604..0x06FF),\n private_use(0xFE00..0xFFFF),\n (0xFFFF)\n } SignatureScheme;\n\n struct {\n SignatureScheme supported_signature_algorithms<2..2^16-2>;\n } SignatureSchemeList;\n\n\n\nRescorla Standards Track [Page 129]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.1.4. Supported Groups Extension\n\n enum {\n unallocated_RESERVED(0x0000),\n\n /* Elliptic Curve Groups (ECDHE) */\n obsolete_RESERVED(0x0001..0x0016),\n secp256r1(0x0017), secp384r1(0x0018), secp521r1(0x0019),\n obsolete_RESERVED(0x001A..0x001C),\n x25519(0x001D), x448(0x001E),\n\n /* Finite Field Groups (DHE) */\n ffdhe2048(0x0100), ffdhe3072(0x0101), ffdhe4096(0x0102),\n ffdhe6144(0x0103), ffdhe8192(0x0104),\n\n /* Reserved Code Points */\n ffdhe_private_use(0x01FC..0x01FF),\n ecdhe_private_use(0xFE00..0xFEFF),\n obsolete_RESERVED(0xFF01..0xFF02),\n (0xFFFF)\n } NamedGroup;\n\n struct {\n NamedGroup named_group_list<2..2^16-1>;\n } NamedGroupList;\n\n Values within \"obsolete_RESERVED\" ranges are used in previous\n versions of TLS and MUST NOT be offered or negotiated by TLS 1.3\n implementations. The obsolete curves have various known/theoretical\n weaknesses or have had very little usage, in some cases only due to\n unintentional server configuration issues. They are no longer\n considered appropriate for general use and should be assumed to be\n potentially unsafe. The set of curves specified here is sufficient\n for interoperability with all currently deployed and properly\n configured TLS implementations.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 130]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.2. Server Parameters Messages\n\n opaque DistinguishedName<1..2^16-1>;\n\n struct {\n DistinguishedName authorities<3..2^16-1>;\n } CertificateAuthoritiesExtension;\n\n struct {\n opaque certificate_extension_oid<1..2^8-1>;\n opaque certificate_extension_values<0..2^16-1>;\n } OIDFilter;\n\n struct {\n OIDFilter filters<0..2^16-1>;\n } OIDFilterExtension;\n\n struct {} PostHandshakeAuth;\n\n struct {\n Extension extensions<0..2^16-1>;\n } EncryptedExtensions;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n Extension extensions<2..2^16-1>;\n } CertificateRequest;\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 131]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.3. Authentication Messages\n\n enum {\n X509(0),\n OpenPGP_RESERVED(1),\n RawPublicKey(2),\n (255)\n } CertificateType;\n\n struct {\n select (certificate_type) {\n case RawPublicKey:\n /* From RFC 7250 ASN.1_subjectPublicKeyInfo */\n opaque ASN1_subjectPublicKeyInfo<1..2^24-1>;\n\n case X509:\n opaque cert_data<1..2^24-1>;\n };\n Extension extensions<0..2^16-1>;\n } CertificateEntry;\n\n struct {\n opaque certificate_request_context<0..2^8-1>;\n CertificateEntry certificate_list<0..2^24-1>;\n } Certificate;\n\n struct {\n SignatureScheme algorithm;\n opaque signature<0..2^16-1>;\n } CertificateVerify;\n\n struct {\n opaque verify_data[Hash.length];\n } Finished;\n\nB.3.4. Ticket Establishment\n\n struct {\n uint32 ticket_lifetime;\n uint32 ticket_age_add;\n opaque ticket_nonce<0..255>;\n opaque ticket<1..2^16-1>;\n Extension extensions<0..2^16-2>;\n } NewSessionTicket;\n\n\n\n\n\n\n\nRescorla Standards Track [Page 132]\n\f\nRFC 8446 TLS August 2018\n\n\nB.3.5. Updating Keys\n\n struct {} EndOfEarlyData;\n\n enum {\n update_not_requested(0), update_requested(1), (255)\n } KeyUpdateRequest;\n\n struct {\n KeyUpdateRequest request_update;\n } KeyUpdate;\n\nB.4. Cipher Suites\n\n A symmetric cipher suite defines the pair of the AEAD algorithm and\n hash algorithm to be used with HKDF. Cipher suite names follow the\n naming convention:\n\n CipherSuite TLS_AEAD_HASH = VALUE;\n\n +-----------+------------------------------------------------+\n | Component | Contents |\n +-----------+------------------------------------------------+\n | TLS | The string \"TLS\" |\n | | |\n | AEAD | The AEAD algorithm used for record protection |\n | | |\n | HASH | The hash algorithm used with HKDF |\n | | |\n | VALUE | The two-byte ID assigned for this cipher suite |\n +-----------+------------------------------------------------+\n\n This specification defines the following cipher suites for use with\n TLS 1.3.\n\n +------------------------------+-------------+\n | Description | Value |\n +------------------------------+-------------+\n | TLS_AES_128_GCM_SHA256 | {0x13,0x01} |\n | | |\n | TLS_AES_256_GCM_SHA384 | {0x13,0x02} |\n | | |\n | TLS_CHACHA20_POLY1305_SHA256 | {0x13,0x03} |\n | | |\n | TLS_AES_128_CCM_SHA256 | {0x13,0x04} |\n | | |\n | TLS_AES_128_CCM_8_SHA256 | {0x13,0x05} |\n +------------------------------+-------------+\n\n\n\nRescorla Standards Track [Page 133]\n\f\nRFC 8446 TLS August 2018\n\n\n The corresponding AEAD algorithms AEAD_AES_128_GCM, AEAD_AES_256_GCM,\n and AEAD_AES_128_CCM are defined in [RFC5116].\n AEAD_CHACHA20_POLY1305 is defined in [RFC8439]. AEAD_AES_128_CCM_8\n is defined in [RFC6655]. The corresponding hash algorithms are\n defined in [SHS].\n\n Although TLS 1.3 uses the same cipher suite space as previous\n versions of TLS, TLS 1.3 cipher suites are defined differently, only\n specifying the symmetric ciphers, and cannot be used for TLS 1.2.\n Similarly, cipher suites for TLS 1.2 and lower cannot be used with\n TLS 1.3.\n\n New cipher suite values are assigned by IANA as described in\n Section 11.\n\nAppendix C. Implementation Notes\n\n The TLS protocol cannot prevent many common security mistakes. This\n appendix provides several recommendations to assist implementors.\n [TLS13-TRACES] provides test vectors for TLS 1.3 handshakes.\n\nC.1. Random Number Generation and Seeding\n\n TLS requires a cryptographically secure pseudorandom number generator\n (CSPRNG). In most cases, the operating system provides an\n appropriate facility such as /dev/urandom, which should be used\n absent other (e.g., performance) concerns. It is RECOMMENDED to use\n an existing CSPRNG implementation in preference to crafting a new\n one. Many adequate cryptographic libraries are already available\n under favorable license terms. Should those prove unsatisfactory,\n [RFC4086] provides guidance on the generation of random values.\n\n TLS uses random values (1) in public protocol fields such as the\n public Random values in the ClientHello and ServerHello and (2) to\n generate keying material. With a properly functioning CSPRNG, this\n does not present a security problem, as it is not feasible to\n determine the CSPRNG state from its output. However, with a broken\n CSPRNG, it may be possible for an attacker to use the public output\n to determine the CSPRNG internal state and thereby predict the keying\n material, as documented in [CHECKOWAY]. Implementations can provide\n extra security against this form of attack by using separate CSPRNGs\n to generate public and private values.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 134]\n\f\nRFC 8446 TLS August 2018\n\n\nC.2. Certificates and Authentication\n\n Implementations are responsible for verifying the integrity of\n certificates and should generally support certificate revocation\n messages. Absent a specific indication from an application profile,\n certificates should always be verified to ensure proper signing by a\n trusted certificate authority (CA). The selection and addition of\n trust anchors should be done very carefully. Users should be able to\n view information about the certificate and trust anchor.\n Applications SHOULD also enforce minimum and maximum key sizes. For\n example, certification paths containing keys or signatures weaker\n than 2048-bit RSA or 224-bit ECDSA are not appropriate for secure\n applications.\n\nC.3. Implementation Pitfalls\n\n Implementation experience has shown that certain parts of earlier TLS\n specifications are not easy to understand and have been a source of\n interoperability and security problems. Many of these areas have\n been clarified in this document, but this appendix contains a short\n list of the most important things that require special attention from\n implementors.\n\n TLS protocol issues:\n\n - Do you correctly handle handshake messages that are fragmented to\n multiple TLS records (see Section 5.1)? Do you correctly handle\n corner cases like a ClientHello that is split into several small\n fragments? Do you fragment handshake messages that exceed the\n maximum fragment size? In particular, the Certificate and\n CertificateRequest handshake messages can be large enough to\n require fragmentation.\n\n - Do you ignore the TLS record layer version number in all\n unencrypted TLS records (see Appendix D)?\n\n - Have you ensured that all support for SSL, RC4, EXPORT ciphers,\n and MD5 (via the \"signature_algorithms\" extension) is completely\n removed from all possible configurations that support TLS 1.3 or\n later, and that attempts to use these obsolete capabilities fail\n correctly (see Appendix D)?\n\n - Do you handle TLS extensions in ClientHellos correctly, including\n unknown extensions?\n\n\n\n\n\n\n\nRescorla Standards Track [Page 135]\n\f\nRFC 8446 TLS August 2018\n\n\n - When the server has requested a client certificate but no suitable\n certificate is available, do you correctly send an empty\n Certificate message, instead of omitting the whole message (see\n Section 4.4.2)?\n\n - When processing the plaintext fragment produced by AEAD-Decrypt\n and scanning from the end for the ContentType, do you avoid\n scanning past the start of the cleartext in the event that the\n peer has sent a malformed plaintext of all zeros?\n\n - Do you properly ignore unrecognized cipher suites (Section 4.1.2),\n hello extensions (Section 4.2), named groups (Section 4.2.7), key\n shares (Section 4.2.8), supported versions (Section 4.2.1), and\n signature algorithms (Section 4.2.3) in the ClientHello?\n\n - As a server, do you send a HelloRetryRequest to clients which\n support a compatible (EC)DHE group but do not predict it in the\n \"key_share\" extension? As a client, do you correctly handle a\n HelloRetryRequest from the server?\n\n Cryptographic details:\n\n - What countermeasures do you use to prevent timing attacks\n [TIMING]?\n\n - When using Diffie-Hellman key exchange, do you correctly preserve\n leading zero bytes in the negotiated key (see Section 7.4.1)?\n\n - Does your TLS client check that the Diffie-Hellman parameters sent\n by the server are acceptable (see Section 4.2.8.1)?\n\n - Do you use a strong and, most importantly, properly seeded random\n number generator (see Appendix C.1) when generating Diffie-Hellman\n private values, the ECDSA \"k\" parameter, and other security-\n critical values? It is RECOMMENDED that implementations implement\n \"deterministic ECDSA\" as specified in [RFC6979].\n\n - Do you zero-pad Diffie-Hellman public key values and shared\n secrets to the group size (see Section 4.2.8.1 and Section 7.4.1)?\n\n - Do you verify signatures after making them, to protect against\n RSA-CRT key leaks [FW15]?\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 136]\n\f\nRFC 8446 TLS August 2018\n\n\nC.4. Client Tracking Prevention\n\n Clients SHOULD NOT reuse a ticket for multiple connections. Reuse of\n a ticket allows passive observers to correlate different connections.\n Servers that issue tickets SHOULD offer at least as many tickets as\n the number of connections that a client might use; for example, a web\n browser using HTTP/1.1 [RFC7230] might open six connections to a\n server. Servers SHOULD issue new tickets with every connection.\n This ensures that clients are always able to use a new ticket when\n creating a new connection.\n\nC.5. Unauthenticated Operation\n\n Previous versions of TLS offered explicitly unauthenticated cipher\n suites based on anonymous Diffie-Hellman. These modes have been\n deprecated in TLS 1.3. However, it is still possible to negotiate\n parameters that do not provide verifiable server authentication by\n several methods, including:\n\n - Raw public keys [RFC7250].\n\n - Using a public key contained in a certificate but without\n validation of the certificate chain or any of its contents.\n\n Either technique used alone is vulnerable to man-in-the-middle\n attacks and therefore unsafe for general use. However, it is also\n possible to bind such connections to an external authentication\n mechanism via out-of-band validation of the server's public key,\n trust on first use, or a mechanism such as channel bindings (though\n the channel bindings described in [RFC5929] are not defined for\n TLS 1.3). If no such mechanism is used, then the connection has no\n protection against active man-in-the-middle attack; applications\n MUST NOT use TLS in such a way absent explicit configuration or a\n specific application profile.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 137]\n\f\nRFC 8446 TLS August 2018\n\n\nAppendix D. Backward Compatibility\n\n The TLS protocol provides a built-in mechanism for version\n negotiation between endpoints potentially supporting different\n versions of TLS.\n\n TLS 1.x and SSL 3.0 use compatible ClientHello messages. Servers can\n also handle clients trying to use future versions of TLS as long as\n the ClientHello format remains compatible and there is at least one\n protocol version supported by both the client and the server.\n\n Prior versions of TLS used the record layer version number\n (TLSPlaintext.legacy_record_version and\n TLSCiphertext.legacy_record_version) for various purposes. As of\n TLS 1.3, this field is deprecated. The value of\n TLSPlaintext.legacy_record_version MUST be ignored by all\n implementations. The value of TLSCiphertext.legacy_record_version is\n included in the additional data for deprotection but MAY otherwise be\n ignored or MAY be validated to match the fixed constant value.\n Version negotiation is performed using only the handshake versions\n (ClientHello.legacy_version and ServerHello.legacy_version, as well\n as the ClientHello, HelloRetryRequest, and ServerHello\n \"supported_versions\" extensions). In order to maximize\n interoperability with older endpoints, implementations that negotiate\n the use of TLS 1.0-1.2 SHOULD set the record layer version number to\n the negotiated version for the ServerHello and all records\n thereafter.\n\n For maximum compatibility with previously non-standard behavior and\n misconfigured deployments, all implementations SHOULD support\n validation of certification paths based on the expectations in this\n document, even when handling prior TLS versions' handshakes (see\n Section 4.4.2.2).\n\n TLS 1.2 and prior supported an \"Extended Master Secret\" [RFC7627]\n extension which digested large parts of the handshake transcript into\n the master secret. Because TLS 1.3 always hashes in the transcript\n up to the server Finished, implementations which support both TLS 1.3\n and earlier versions SHOULD indicate the use of the Extended Master\n Secret extension in their APIs whenever TLS 1.3 is used.\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 138]\n\f\nRFC 8446 TLS August 2018\n\n\nD.1. Negotiating with an Older Server\n\n A TLS 1.3 client who wishes to negotiate with servers that do not\n support TLS 1.3 will send a normal TLS 1.3 ClientHello containing\n 0x0303 (TLS 1.2) in ClientHello.legacy_version but with the correct\n version(s) in the \"supported_versions\" extension. If the server does\n not support TLS 1.3, it will respond with a ServerHello containing an\n older version number. If the client agrees to use this version, the\n negotiation will proceed as appropriate for the negotiated protocol.\n A client using a ticket for resumption SHOULD initiate the connection\n using the version that was previously negotiated.\n\n Note that 0-RTT data is not compatible with older servers and\n SHOULD NOT be sent absent knowledge that the server supports TLS 1.3.\n See Appendix D.3.\n\n If the version chosen by the server is not supported by the client\n (or is not acceptable), the client MUST abort the handshake with a\n \"protocol_version\" alert.\n\n Some legacy server implementations are known to not implement the TLS\n specification properly and might abort connections upon encountering\n TLS extensions or versions which they are not aware of.\n Interoperability with buggy servers is a complex topic beyond the\n scope of this document. Multiple connection attempts may be required\n in order to negotiate a backward-compatible connection; however, this\n practice is vulnerable to downgrade attacks and is NOT RECOMMENDED.\n\nD.2. Negotiating with an Older Client\n\n A TLS server can also receive a ClientHello indicating a version\n number smaller than its highest supported version. If the\n \"supported_versions\" extension is present, the server MUST negotiate\n using that extension as described in Section 4.2.1. If the\n \"supported_versions\" extension is not present, the server MUST\n negotiate the minimum of ClientHello.legacy_version and TLS 1.2. For\n example, if the server supports TLS 1.0, 1.1, and 1.2, and\n legacy_version is TLS 1.0, the server will proceed with a TLS 1.0\n ServerHello. If the \"supported_versions\" extension is absent and the\n server only supports versions greater than\n ClientHello.legacy_version, the server MUST abort the handshake with\n a \"protocol_version\" alert.\n\n Note that earlier versions of TLS did not clearly specify the record\n layer version number value in all cases\n (TLSPlaintext.legacy_record_version). Servers will receive various\n TLS 1.x versions in this field, but its value MUST always be ignored.\n\n\n\n\nRescorla Standards Track [Page 139]\n\f\nRFC 8446 TLS August 2018\n\n\nD.3. 0-RTT Backward Compatibility\n\n 0-RTT data is not compatible with older servers. An older server\n will respond to the ClientHello with an older ServerHello, but it\n will not correctly skip the 0-RTT data and will fail to complete the\n handshake. This can cause issues when a client attempts to use\n 0-RTT, particularly against multi-server deployments. For example, a\n deployment could deploy TLS 1.3 gradually with some servers\n implementing TLS 1.3 and some implementing TLS 1.2, or a TLS 1.3\n deployment could be downgraded to TLS 1.2.\n\n A client that attempts to send 0-RTT data MUST fail a connection if\n it receives a ServerHello with TLS 1.2 or older. It can then retry\n the connection with 0-RTT disabled. To avoid a downgrade attack, the\n client SHOULD NOT disable TLS 1.3, only 0-RTT.\n\n To avoid this error condition, multi-server deployments SHOULD ensure\n a uniform and stable deployment of TLS 1.3 without 0-RTT prior to\n enabling 0-RTT.\n\nD.4. Middlebox Compatibility Mode\n\n Field measurements [Ben17a] [Ben17b] [Res17a] [Res17b] have found\n that a significant number of middleboxes misbehave when a TLS\n client/server pair negotiates TLS 1.3. Implementations can increase\n the chance of making connections through those middleboxes by making\n the TLS 1.3 handshake look more like a TLS 1.2 handshake:\n\n - The client always provides a non-empty session ID in the\n ClientHello, as described in the legacy_session_id section of\n Section 4.1.2.\n\n - If not offering early data, the client sends a dummy\n change_cipher_spec record (see the third paragraph of Section 5)\n immediately before its second flight. This may either be before\n its second ClientHello or before its encrypted handshake flight.\n If offering early data, the record is placed immediately after the\n first ClientHello.\n\n - The server sends a dummy change_cipher_spec record immediately\n after its first handshake message. This may either be after a\n ServerHello or a HelloRetryRequest.\n\n When put together, these changes make the TLS 1.3 handshake resemble\n TLS 1.2 session resumption, which improves the chance of successfully\n connecting through middleboxes. This \"compatibility mode\" is\n partially negotiated: the client can opt to provide a session ID or\n not, and the server has to echo it. Either side can send\n\n\n\nRescorla Standards Track [Page 140]\n\f\nRFC 8446 TLS August 2018\n\n\n change_cipher_spec at any time during the handshake, as they must be\n ignored by the peer, but if the client sends a non-empty session ID,\n the server MUST send the change_cipher_spec as described in this\n appendix.\n\nD.5. Security Restrictions Related to Backward Compatibility\n\n Implementations negotiating the use of older versions of TLS SHOULD\n prefer forward secret and AEAD cipher suites, when available.\n\n The security of RC4 cipher suites is considered insufficient for the\n reasons cited in [RFC7465]. Implementations MUST NOT offer or\n negotiate RC4 cipher suites for any version of TLS for any reason.\n\n Old versions of TLS permitted the use of very low strength ciphers.\n Ciphers with a strength less than 112 bits MUST NOT be offered or\n negotiated for any version of TLS for any reason.\n\n The security of SSL 3.0 [RFC6101] is considered insufficient for the\n reasons enumerated in [RFC7568], and it MUST NOT be negotiated for\n any reason.\n\n The security of SSL 2.0 [SSL2] is considered insufficient for the\n reasons enumerated in [RFC6176], and it MUST NOT be negotiated for\n any reason.\n\n Implementations MUST NOT send an SSL version 2.0 compatible\n CLIENT-HELLO. Implementations MUST NOT negotiate TLS 1.3 or later\n using an SSL version 2.0 compatible CLIENT-HELLO. Implementations\n are NOT RECOMMENDED to accept an SSL version 2.0 compatible\n CLIENT-HELLO in order to negotiate older versions of TLS.\n\n Implementations MUST NOT send a ClientHello.legacy_version or\n ServerHello.legacy_version set to 0x0300 or less. Any endpoint\n receiving a Hello message with ClientHello.legacy_version or\n ServerHello.legacy_version set to 0x0300 MUST abort the handshake\n with a \"protocol_version\" alert.\n\n Implementations MUST NOT send any records with a version less than\n 0x0300. Implementations SHOULD NOT accept any records with a version\n less than 0x0300 (but may inadvertently do so if the record version\n number is ignored completely).\n\n Implementations MUST NOT use the Truncated HMAC extension, defined in\n Section 7 of [RFC6066], as it is not applicable to AEAD algorithms\n and has been shown to be insecure in some scenarios.\n\n\n\n\n\nRescorla Standards Track [Page 141]\n\f\nRFC 8446 TLS August 2018\n\n\nAppendix E. Overview of Security Properties\n\n A complete security analysis of TLS is outside the scope of this\n document. In this appendix, we provide an informal description of\n the desired properties as well as references to more detailed work in\n the research literature which provides more formal definitions.\n\n We cover properties of the handshake separately from those of the\n record layer.\n\nE.1. Handshake\n\n The TLS handshake is an Authenticated Key Exchange (AKE) protocol\n which is intended to provide both one-way authenticated (server-only)\n and mutually authenticated (client and server) functionality. At the\n completion of the handshake, each side outputs its view of the\n following values:\n\n - A set of \"session keys\" (the various secrets derived from the\n master secret) from which can be derived a set of working keys.\n\n - A set of cryptographic parameters (algorithms, etc.).\n\n - The identities of the communicating parties.\n\n We assume the attacker to be an active network attacker, which means\n it has complete control over the network used to communicate between\n the parties [RFC3552]. Even under these conditions, the handshake\n should provide the properties listed below. Note that these\n properties are not necessarily independent, but reflect the protocol\n consumers' needs.\n\n Establishing the same session keys: The handshake needs to output\n the same set of session keys on both sides of the handshake,\n provided that it completes successfully on each endpoint (see\n [CK01], Definition 1, part 1).\n\n Secrecy of the session keys: The shared session keys should be known\n only to the communicating parties and not to the attacker (see\n [CK01], Definition 1, part 2). Note that in a unilaterally\n authenticated connection, the attacker can establish its own\n session keys with the server, but those session keys are distinct\n from those established by the client.\n\n Peer authentication: The client's view of the peer identity should\n reflect the server's identity. If the client is authenticated,\n the server's view of the peer identity should match the client's\n identity.\n\n\n\nRescorla Standards Track [Page 142]\n\f\nRFC 8446 TLS August 2018\n\n\n Uniqueness of the session keys: Any two distinct handshakes should\n produce distinct, unrelated session keys. Individual session keys\n produced by a handshake should also be distinct and independent.\n\n Downgrade protection: The cryptographic parameters should be the\n same on both sides and should be the same as if the peers had been\n communicating in the absence of an attack (see [BBFGKZ16],\n Definitions 8 and 9).\n\n Forward secret with respect to long-term keys: If the long-term\n keying material (in this case the signature keys in certificate-\n based authentication modes or the external/resumption PSK in PSK\n with (EC)DHE modes) is compromised after the handshake is\n complete, this does not compromise the security of the session key\n (see [DOW92]), as long as the session key itself has been erased.\n The forward secrecy property is not satisfied when PSK is used in\n the \"psk_ke\" PskKeyExchangeMode.\n\n Key Compromise Impersonation (KCI) resistance: In a mutually\n authenticated connection with certificates, compromising the\n long-term secret of one actor should not break that actor's\n authentication of their peer in the given connection (see\n [HGFS15]). For example, if a client's signature key is\n compromised, it should not be possible to impersonate arbitrary\n servers to that client in subsequent handshakes.\n\n Protection of endpoint identities: The server's identity\n (certificate) should be protected against passive attackers. The\n client's identity should be protected against both passive and\n active attackers.\n\n Informally, the signature-based modes of TLS 1.3 provide for the\n establishment of a unique, secret, shared key established by an\n (EC)DHE key exchange and authenticated by the server's signature over\n the handshake transcript, as well as tied to the server's identity by\n a MAC. If the client is authenticated by a certificate, it also\n signs over the handshake transcript and provides a MAC tied to both\n identities. [SIGMA] describes the design and analysis of this type\n of key exchange protocol. If fresh (EC)DHE keys are used for each\n connection, then the output keys are forward secret.\n\n The external PSK and resumption PSK bootstrap from a long-term shared\n secret into a unique per-connection set of short-term session keys.\n This secret may have been established in a previous handshake. If\n PSK with (EC)DHE key establishment is used, these session keys will\n also be forward secret. The resumption PSK has been designed so that\n the resumption master secret computed by connection N and needed to\n form connection N+1 is separate from the traffic keys used by\n\n\n\nRescorla Standards Track [Page 143]\n\f\nRFC 8446 TLS August 2018\n\n\n connection N, thus providing forward secrecy between the connections.\n In addition, if multiple tickets are established on the same\n connection, they are associated with different keys, so compromise of\n the PSK associated with one ticket does not lead to the compromise of\n connections established with PSKs associated with other tickets.\n This property is most interesting if tickets are stored in a database\n (and so can be deleted) rather than if they are self-encrypted.\n\n The PSK binder value forms a binding between a PSK and the current\n handshake, as well as between the session where the PSK was\n established and the current session. This binding transitively\n includes the original handshake transcript, because that transcript\n is digested into the values which produce the resumption master\n secret. This requires that both the KDF used to produce the\n resumption master secret and the MAC used to compute the binder be\n collision resistant. See Appendix E.1.1 for more on this. Note: The\n binder does not cover the binder values from other PSKs, though they\n are included in the Finished MAC.\n\n TLS does not currently permit the server to send a\n certificate_request message in non-certificate-based handshakes\n (e.g., PSK). If this restriction were to be relaxed in future, the\n client's signature would not cover the server's certificate directly.\n However, if the PSK was established through a NewSessionTicket, the\n client's signature would transitively cover the server's certificate\n through the PSK binder. [PSK-FINISHED] describes a concrete attack\n on constructions that do not bind to the server's certificate (see\n also [Kraw16]). It is unsafe to use certificate-based client\n authentication when the client might potentially share the same\n PSK/key-id pair with two different endpoints. Implementations\n MUST NOT combine external PSKs with certificate-based authentication\n of either the client or the server unless negotiated by some\n extension.\n\n If an exporter is used, then it produces values which are unique and\n secret (because they are generated from a unique session key).\n Exporters computed with different labels and contexts are\n computationally independent, so it is not feasible to compute one\n from another or the session secret from the exported value.\n Note: Exporters can produce arbitrary-length values; if exporters are\n to be used as channel bindings, the exported value MUST be large\n enough to provide collision resistance. The exporters provided in\n TLS 1.3 are derived from the same Handshake Contexts as the early\n traffic keys and the application traffic keys, respectively, and thus\n have similar security properties. Note that they do not include the\n client's certificate; future applications which wish to bind to the\n client's certificate may need to define a new exporter that includes\n the full handshake transcript.\n\n\n\nRescorla Standards Track [Page 144]\n\f\nRFC 8446 TLS August 2018\n\n\n For all handshake modes, the Finished MAC (and, where present, the\n signature) prevents downgrade attacks. In addition, the use of\n certain bytes in the random nonces as described in Section 4.1.3\n allows the detection of downgrade to previous TLS versions. See\n [BBFGKZ16] for more details on TLS 1.3 and downgrade.\n\n As soon as the client and the server have exchanged enough\n information to establish shared keys, the remainder of the handshake\n is encrypted, thus providing protection against passive attackers,\n even if the computed shared key is not authenticated. Because the\n server authenticates before the client, the client can ensure that if\n it authenticates to the server, it only reveals its identity to an\n authenticated server. Note that implementations must use the\n provided record-padding mechanism during the handshake to avoid\n leaking information about the identities due to length. The client's\n proposed PSK identities are not encrypted, nor is the one that the\n server selects.\n\nE.1.1. Key Derivation and HKDF\n\n Key derivation in TLS 1.3 uses HKDF as defined in [RFC5869] and its\n two components, HKDF-Extract and HKDF-Expand. The full rationale for\n the HKDF construction can be found in [Kraw10] and the rationale for\n the way it is used in TLS 1.3 in [KW16]. Throughout this document,\n each application of HKDF-Extract is followed by one or more\n invocations of HKDF-Expand. This ordering should always be followed\n (including in future revisions of this document); in particular, one\n SHOULD NOT use an output of HKDF-Extract as an input to another\n application of HKDF-Extract without an HKDF-Expand in between.\n Multiple applications of HKDF-Expand to some of the same inputs are\n allowed as long as these are differentiated via the key and/or the\n labels.\n\n Note that HKDF-Expand implements a pseudorandom function (PRF) with\n both inputs and outputs of variable length. In some of the uses of\n HKDF in this document (e.g., for generating exporters and the\n resumption_master_secret), it is necessary that the application of\n HKDF-Expand be collision resistant; namely, it should be infeasible\n to find two different inputs to HKDF-Expand that output the same\n value. This requires the underlying hash function to be collision\n resistant and the output length from HKDF-Expand to be of size at\n least 256 bits (or as much as needed for the hash function to prevent\n finding collisions).\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 145]\n\f\nRFC 8446 TLS August 2018\n\n\nE.1.2. Client Authentication\n\n A client that has sent authentication data to a server, either during\n the handshake or in post-handshake authentication, cannot be sure\n whether the server afterwards considers the client to be\n authenticated or not. If the client needs to determine if the server\n considers the connection to be unilaterally or mutually\n authenticated, this has to be provisioned by the application layer.\n See [CHHSV17] for details. In addition, the analysis of\n post-handshake authentication from [Kraw16] shows that the client\n identified by the certificate sent in the post-handshake phase\n possesses the traffic key. This party is therefore the client that\n participated in the original handshake or one to whom the original\n client delegated the traffic key (assuming that the traffic key has\n not been compromised).\n\nE.1.3. 0-RTT\n\n The 0-RTT mode of operation generally provides security properties\n similar to those of 1-RTT data, with the two exceptions that the\n 0-RTT encryption keys do not provide full forward secrecy and that\n the server is not able to guarantee uniqueness of the handshake\n (non-replayability) without keeping potentially undue amounts of\n state. See Section 8 for mechanisms to limit the exposure to replay.\n\nE.1.4. Exporter Independence\n\n The exporter_master_secret and early_exporter_master_secret are\n derived to be independent of the traffic keys and therefore do not\n represent a threat to the security of traffic encrypted with those\n keys. However, because these secrets can be used to compute any\n exporter value, they SHOULD be erased as soon as possible. If the\n total set of exporter labels is known, then implementations SHOULD\n pre-compute the inner Derive-Secret stage of the exporter computation\n for all those labels, then erase the [early_]exporter_master_secret,\n followed by each inner value as soon as it is known that it will not\n be needed again.\n\nE.1.5. Post-Compromise Security\n\n TLS does not provide security for handshakes which take place after\n the peer's long-term secret (signature key or external PSK) is\n compromised. It therefore does not provide post-compromise security\n [CCG16], sometimes also referred to as backward or future secrecy.\n This is in contrast to KCI resistance, which describes the security\n guarantees that a party has after its own long-term secret has been\n compromised.\n\n\n\n\nRescorla Standards Track [Page 146]\n\f\nRFC 8446 TLS August 2018\n\n\nE.1.6. External References\n\n The reader should refer to the following references for analysis of\n the TLS handshake: [DFGS15], [CHSV16], [DFGS16], [KW16], [Kraw16],\n [FGSW16], [LXZFH16], [FG17], and [BBK17].\n\nE.2. Record Layer\n\n The record layer depends on the handshake producing strong traffic\n secrets which can be used to derive bidirectional encryption keys and\n nonces. Assuming that is true, and the keys are used for no more\n data than indicated in Section 5.5, then the record layer should\n provide the following guarantees:\n\n Confidentiality: An attacker should not be able to determine the\n plaintext contents of a given record.\n\n Integrity: An attacker should not be able to craft a new record\n which is different from an existing record which will be accepted\n by the receiver.\n\n Order protection/non-replayability: An attacker should not be able\n to cause the receiver to accept a record which it has already\n accepted or cause the receiver to accept record N+1 without having\n first processed record N.\n\n Length concealment: Given a record with a given external length, the\n attacker should not be able to determine the amount of the record\n that is content versus padding.\n\n Forward secrecy after key change: If the traffic key update\n mechanism described in Section 4.6.3 has been used and the\n previous generation key is deleted, an attacker who compromises\n the endpoint should not be able to decrypt traffic encrypted with\n the old key.\n\n Informally, TLS 1.3 provides these properties by AEAD-protecting the\n plaintext with a strong key. AEAD encryption [RFC5116] provides\n confidentiality and integrity for the data. Non-replayability is\n provided by using a separate nonce for each record, with the nonce\n being derived from the record sequence number (Section 5.3), with the\n sequence number being maintained independently at both sides; thus,\n records which are delivered out of order result in AEAD deprotection\n failures. In order to prevent mass cryptanalysis when the same\n plaintext is repeatedly encrypted by different users under the same\n key (as is commonly the case for HTTP), the nonce is formed by mixing\n\n\n\n\n\nRescorla Standards Track [Page 147]\n\f\nRFC 8446 TLS August 2018\n\n\n the sequence number with a secret per-connection initialization\n vector derived along with the traffic keys. See [BT16] for analysis\n of this construction.\n\n The rekeying technique in TLS 1.3 (see Section 7.2) follows the\n construction of the serial generator as discussed in [REKEY], which\n shows that rekeying can allow keys to be used for a larger number of\n encryptions than without rekeying. This relies on the security of\n the HKDF-Expand-Label function as a pseudorandom function (PRF). In\n addition, as long as this function is truly one way, it is not\n possible to compute traffic keys from prior to a key change (forward\n secrecy).\n\n TLS does not provide security for data which is communicated on a\n connection after a traffic secret of that connection is compromised.\n That is, TLS does not provide post-compromise security/future\n secrecy/backward secrecy with respect to the traffic secret. Indeed,\n an attacker who learns a traffic secret can compute all future\n traffic secrets on that connection. Systems which want such\n guarantees need to do a fresh handshake and establish a new\n connection with an (EC)DHE exchange.\n\nE.2.1. External References\n\n The reader should refer to the following references for analysis of\n the TLS record layer: [BMMRT15], [BT16], [BDFKPPRSZZ16], [BBK17], and\n [PS18].\n\nE.3. Traffic Analysis\n\n TLS is susceptible to a variety of traffic analysis attacks based on\n observing the length and timing of encrypted packets [CLINIC]\n [HCJC16]. This is particularly easy when there is a small set of\n possible messages to be distinguished, such as for a video server\n hosting a fixed corpus of content, but still provides usable\n information even in more complicated scenarios.\n\n TLS does not provide any specific defenses against this form of\n attack but does include a padding mechanism for use by applications:\n The plaintext protected by the AEAD function consists of content plus\n variable-length padding, which allows the application to produce\n arbitrary-length encrypted records as well as padding-only cover\n traffic to conceal the difference between periods of transmission and\n periods of silence. Because the padding is encrypted alongside the\n actual content, an attacker cannot directly determine the length of\n the padding but may be able to measure it indirectly by the use of\n timing channels exposed during record processing (i.e., seeing how\n long it takes to process a record or trickling in records to see\n\n\n\nRescorla Standards Track [Page 148]\n\f\nRFC 8446 TLS August 2018\n\n\n which ones elicit a response from the server). In general, it is not\n known how to remove all of these channels because even a\n constant-time padding removal function will likely feed the content\n into data-dependent functions. At minimum, a fully constant-time\n server or client would require close cooperation with the\n application-layer protocol implementation, including making that\n higher-level protocol constant time.\n\n Note: Robust traffic analysis defenses will likely lead to inferior\n performance due to delays in transmitting packets and increased\n traffic volume.\n\nE.4. Side-Channel Attacks\n\n In general, TLS does not have specific defenses against side-channel\n attacks (i.e., those which attack the communications via secondary\n channels such as timing), leaving those to the implementation of the\n relevant cryptographic primitives. However, certain features of TLS\n are designed to make it easier to write side-channel resistant code:\n\n - Unlike previous versions of TLS which used a composite MAC-then-\n encrypt structure, TLS 1.3 only uses AEAD algorithms, allowing\n implementations to use self-contained constant-time\n implementations of those primitives.\n\n - TLS uses a uniform \"bad_record_mac\" alert for all decryption\n errors, which is intended to prevent an attacker from gaining\n piecewise insight into portions of the message. Additional\n resistance is provided by terminating the connection on such\n errors; a new connection will have different cryptographic\n material, preventing attacks against the cryptographic primitives\n that require multiple trials.\n\n Information leakage through side channels can occur at layers above\n TLS, in application protocols and the applications that use them.\n Resistance to side-channel attacks depends on applications and\n application protocols separately ensuring that confidential\n information is not inadvertently leaked.\n\n\n\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 149]\n\f\nRFC 8446 TLS August 2018\n\n\nE.5. Replay Attacks on 0-RTT\n\n Replayable 0-RTT data presents a number of security threats to TLS-\n using applications, unless those applications are specifically\n engineered to be safe under replay (minimally, this means idempotent,\n but in many cases may also require other stronger conditions, such as\n constant-time response). Potential attacks include:\n\n - Duplication of actions which cause side effects (e.g., purchasing\n an item or transferring money) to be duplicated, thus harming the\n site or the user.\n\n - Attackers can store and replay 0-RTT messages in order to reorder\n them with respect to other messages (e.g., moving a delete to\n after a create).\n\n - Exploiting cache timing behavior to discover the content of 0-RTT\n messages by replaying a 0-RTT message to a different cache node\n and then using a separate connection to measure request latency,\n to see if the two requests address the same resource.\n\n If data can be replayed a large number of times, additional attacks\n become possible, such as making repeated measurements of the speed of\n cryptographic operations. In addition, they may be able to overload\n rate-limiting systems. For a further description of these attacks,\n see [Mac17].\n\n Ultimately, servers have the responsibility to protect themselves\n against attacks employing 0-RTT data replication. The mechanisms\n described in Section 8 are intended to prevent replay at the TLS\n layer but do not provide complete protection against receiving\n multiple copies of client data. TLS 1.3 falls back to the 1-RTT\n handshake when the server does not have any information about the\n client, e.g., because it is in a different cluster which does not\n share state or because the ticket has been deleted as described in\n Section 8.1. If the application-layer protocol retransmits data in\n this setting, then it is possible for an attacker to induce message\n duplication by sending the ClientHello to both the original cluster\n (which processes the data immediately) and another cluster which will\n fall back to 1-RTT and process the data upon application-layer\n replay. The scale of this attack is limited by the client's\n willingness to retry transactions and therefore only allows a limited\n amount of duplication, with each copy appearing as a new connection\n at the server.\n\n\n\n\n\n\n\nRescorla Standards Track [Page 150]\n\f\nRFC 8446 TLS August 2018\n\n\n If implemented correctly, the mechanisms described in Sections 8.1\n and 8.2 prevent a replayed ClientHello and its associated 0-RTT data\n from being accepted multiple times by any cluster with consistent\n state; for servers which limit the use of 0-RTT to one cluster for a\n single ticket, then a given ClientHello and its associated 0-RTT data\n will only be accepted once. However, if state is not completely\n consistent, then an attacker might be able to have multiple copies of\n the data be accepted during the replication window. Because clients\n do not know the exact details of server behavior, they MUST NOT send\n messages in early data which are not safe to have replayed and which\n they would not be willing to retry across multiple 1-RTT connections.\n\n Application protocols MUST NOT use 0-RTT data without a profile that\n defines its use. That profile needs to identify which messages or\n interactions are safe to use with 0-RTT and how to handle the\n situation when the server rejects 0-RTT and falls back to 1-RTT.\n\n In addition, to avoid accidental misuse, TLS implementations MUST NOT\n enable 0-RTT (either sending or accepting) unless specifically\n requested by the application and MUST NOT automatically resend 0-RTT\n data if it is rejected by the server unless instructed by the\n application. Server-side applications may wish to implement special\n processing for 0-RTT data for some kinds of application traffic\n (e.g., abort the connection, request that data be resent at the\n application layer, or delay processing until the handshake\n completes). In order to allow applications to implement this kind of\n processing, TLS implementations MUST provide a way for the\n application to determine if the handshake has completed.\n\nE.5.1. Replay and Exporters\n\n Replays of the ClientHello produce the same early exporter, thus\n requiring additional care by applications which use these exporters.\n In particular, if these exporters are used as an authentication\n channel binding (e.g., by signing the output of the exporter), an\n attacker who compromises the PSK can transplant authenticators\n between connections without compromising the authentication key.\n\n In addition, the early exporter SHOULD NOT be used to generate\n server-to-client encryption keys because that would entail the reuse\n of those keys. This parallels the use of the early application\n traffic keys only in the client-to-server direction.\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 151]\n\f\nRFC 8446 TLS August 2018\n\n\nE.6. PSK Identity Exposure\n\n Because implementations respond to an invalid PSK binder by aborting\n the handshake, it may be possible for an attacker to verify whether a\n given PSK identity is valid. Specifically, if a server accepts both\n external-PSK handshakes and certificate-based handshakes, a valid PSK\n identity will result in a failed handshake, whereas an invalid\n identity will just be skipped and result in a successful certificate\n handshake. Servers which solely support PSK handshakes may be able\n to resist this form of attack by treating the cases where there is no\n valid PSK identity and where there is an identity but it has an\n invalid binder identically.\n\nE.7. Sharing PSKs\n\n TLS 1.3 takes a conservative approach to PSKs by binding them to a\n specific KDF. By contrast, TLS 1.2 allows PSKs to be used with any\n hash function and the TLS 1.2 PRF. Thus, any PSK which is used with\n both TLS 1.2 and TLS 1.3 must be used with only one hash in TLS 1.3,\n which is less than optimal if users want to provision a single PSK.\n The constructions in TLS 1.2 and TLS 1.3 are different, although they\n are both based on HMAC. While there is no known way in which the\n same PSK might produce related output in both versions, only limited\n analysis has been done. Implementations can ensure safety from\n cross-protocol related output by not reusing PSKs between TLS 1.3 and\n TLS 1.2.\n\nE.8. Attacks on Static RSA\n\n Although TLS 1.3 does not use RSA key transport and so is not\n directly susceptible to Bleichenbacher-type attacks [Blei98], if TLS\n 1.3 servers also support static RSA in the context of previous\n versions of TLS, then it may be possible to impersonate the server\n for TLS 1.3 connections [JSS15]. TLS 1.3 implementations can prevent\n this attack by disabling support for static RSA across all versions\n of TLS. In principle, implementations might also be able to separate\n certificates with different keyUsage bits for static RSA decryption\n and RSA signature, but this technique relies on clients refusing to\n accept signatures using keys in certificates that do not have the\n digitalSignature bit set, and many clients do not enforce this\n restriction.\n\n\n\n\n\n\n\n\n\n\nRescorla Standards Track [Page 152]\n\f\nRFC 8446 TLS August 2018\n\n\nContributors\n\n Martin Abadi\n University of California, Santa Cruz\n [redacted-email]\n\n Christopher Allen\n (co-editor of TLS 1.0)\n Alacrity Ventures\n [redacted-email]\n\n Richard Barnes\n Cisco\n [redacted-email]\n\n Steven M. Bellovin\n Columbia University\n [redacted-email]\n\n David Benjamin\n Google\n [redacted-email]\n\n Benjamin Beurdouche\n INRIA & Microsoft Research\n [redacted-email]\n\n Karthikeyan Bhargavan\n (editor of [RFC7627])\n INRIA\n [redacted-email]\n\n Simon Blake-Wilson\n (co-author of [RFC4492])\n BCI\n [redacted-email]\n\n Nelson Bolyard\n (co-author of [RFC4492])\n Sun Microsystems, Inc.\n [redacted-email]\n\n Ran Canetti\n IBM\n [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 153]\n\f\nRFC 8446 TLS August 2018\n\n\n Matt Caswell\n OpenSSL\n [redacted-email]\n\n Stephen Checkoway\n University of Illinois at Chicago\n [redacted-email]\n\n Pete Chown\n Skygate Technology Ltd\n [redacted-email]\n\n Katriel Cohn-Gordon\n University of Oxford\n [redacted-email]\n\n Cas Cremers\n University of Oxford\n [redacted-email]\n\n Antoine Delignat-Lavaud\n (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n Tim Dierks\n (co-author of TLS 1.0, co-editor of TLS 1.1 and 1.2)\n Independent\n [redacted-email]\n\n Roelof DuToit\n Symantec Corporation\n [redacted-email]\n\n Taher Elgamal\n Securify\n [redacted-email]\n\n Pasi Eronen\n Nokia\n [redacted-email]\n\n Cedric Fournet\n Microsoft\n [redacted-email]\n\n\n\n\n\n\nRescorla Standards Track [Page 154]\n\f\nRFC 8446 TLS August 2018\n\n\n Anil Gangolli\n [redacted-email]\n\n David M. Garrett\n [redacted-email]\n\n Illya Gerasymchuk\n Independent\n [redacted-email]\n\n Alessandro Ghedini\n Cloudflare Inc.\n [redacted-email]\n\n Daniel Kahn Gillmor\n ACLU\n [redacted-email]\n\n Matthew Green\n Johns Hopkins University\n [redacted-email]\n\n Jens Guballa\n ETAS\n [redacted-email]\n\n Felix Guenther\n TU Darmstadt\n [redacted-email]\n\n Vipul Gupta\n (co-author of [RFC4492])\n Sun Microsystems Laboratories\n [redacted-email]\n\n Chris Hawk\n (co-author of [RFC4492])\n Corriente Networks LLC\n [redacted-email]\n\n Kipp Hickman\n\n Alfred Hoenes\n\n David Hopwood\n Independent Consultant\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 155]\n\f\nRFC 8446 TLS August 2018\n\n\n Marko Horvat\n MPI-SWS\n [redacted-email]\n\n Jonathan Hoyland\n Royal Holloway, University of London\n [redacted-email]\n\n Subodh Iyengar\n Facebook\n [redacted-email]\n\n Benjamin Kaduk\n Akamai Technologies\n [redacted-email]\n\n Hubert Kario\n Red Hat Inc.\n [redacted-email]\n\n Phil Karlton\n (co-author of SSL 3.0)\n\n Leon Klingele\n Independent\n [redacted-email]\n\n Paul Kocher\n (co-author of SSL 3.0)\n Cryptography Research\n [redacted-email]\n\n Hugo Krawczyk\n IBM\n [redacted-email]\n\n Adam Langley\n (co-author of [RFC7627])\n Google\n [redacted-email]\n\n Olivier Levillain\n ANSSI\n [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 156]\n\f\nRFC 8446 TLS August 2018\n\n\n Xiaoyin Liu\n University of North Carolina at Chapel Hill\n [redacted-email]\n\n Ilari Liusvaara\n Independent\n [redacted-email]\n\n Atul Luykx\n K.U. Leuven\n [redacted-email]\n\n Colm MacCarthaigh\n Amazon Web Services\n [redacted-email]\n\n Carl Mehner\n USAA\n [redacted-email]\n\n Jan Mikkelsen\n Transactionware\n [redacted-email]\n\n Bodo Moeller\n (co-author of [RFC4492])\n Google\n [redacted-email]\n\n Kyle Nekritz\n Facebook\n [redacted-email]\n\n Erik Nygren\n Akamai Technologies\n [redacted-email]\n\n Magnus Nystrom\n Microsoft\n [redacted-email]\n\n Kazuho Oku\n DeNA Co., Ltd.\n [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 157]\n\f\nRFC 8446 TLS August 2018\n\n\n Kenny Paterson\n Royal Holloway, University of London\n [redacted-email]\n\n Christopher Patton\n University of Florida\n [redacted-email]\n\n Alfredo Pironti\n (co-author of [RFC7627])\n INRIA\n [redacted-email]\n\n Andrei Popov\n Microsoft\n [redacted-email]\n\n Marsh Ray\n (co-author of [RFC7627])\n Microsoft\n [redacted-email]\n\n Robert Relyea\n Netscape Communications\n [redacted-email]\n\n Kyle Rose\n Akamai Technologies\n [redacted-email]\n\n Jim Roskind\n Amazon\n [redacted-email]\n\n Michael Sabin\n\n Joe Salowey\n Tableau Software\n [redacted-email]\n\n Rich Salz\n Akamai\n [redacted-email]\n\n David Schinazi\n Apple Inc.\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 158]\n\f\nRFC 8446 TLS August 2018\n\n\n Sam Scott\n Royal Holloway, University of London\n [redacted-email]\n\n Thomas Shrimpton\n University of Florida\n [redacted-email]\n\n Dan Simon\n Microsoft, Inc.\n [redacted-email]\n\n Brian Smith\n Independent\n [redacted-email]\n\n Brian Sniffen\n Akamai Technologies\n [redacted-email]\n\n Nick Sullivan\n Cloudflare Inc.\n [redacted-email]\n\n Bjoern Tackmann\n University of California, San Diego\n [redacted-email]\n\n Tim Taubert\n Mozilla\n [redacted-email]\n\n Martin Thomson\n Mozilla\n [redacted-email]\n\n Hannes Tschofenig\n Arm Limited\n [redacted-email]\n\n Sean Turner\n sn3rd\n [redacted-email]\n\n Steven Valdez\n Google\n [redacted-email]\n\n\n\n\nRescorla Standards Track [Page 159]\n\f\nRFC 8446 TLS August 2018\n\n\n Filippo Valsorda\n Cloudflare Inc.\n [redacted-email]\n\n Thyla van der Merwe\n Royal Holloway, University of London\n [redacted-email]\n\n Victor Vasiliev\n Google\n [redacted-email]\n\n Hoeteck Wee\n Ecole Normale Superieure, Paris\n [redacted-email]\n\n Tom Weinstein\n\n David Wong\n NCC Group\n [redacted-email]\n\n Christopher A. Wood\n Apple Inc.\n [redacted-email]\n\n Tim Wright\n Vodafone\n [redacted-email]\n\n Peter Wu\n Independent\n [redacted-email]\n\n Kazu Yamamoto\n Internet Initiative Japan Inc.\n [redacted-email]\n\nAuthor's Address\n\n Eric Rescorla\n Mozilla\n\n Email: [redacted-email]\n\n\n\n\n\n\n\nRescorla Standards Track [Page 160]", "text_sha256": "d596886622bceee8500b3f3c9d869d947b4e70a2f05ea65fdd75666b46debd81"}], "relations": [{"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5077 H. Zhou", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5077", "source_sha256": "e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef"}, {"char_end": 3675, "char_start": 3419, "evidence_quote": "{\"id\": 1269006, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269006/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}", "fact_ids": ["informative_reference:5077"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "informative_reference", "relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5077"}, {"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5246 Independent", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5246", "source_sha256": "2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb"}, {"char_end": 3417, "char_start": 3161, "evidence_quote": "{\"id\": 1269005, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269005/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}", "fact_ids": ["informative_reference:5246"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "informative_reference", "relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5246"}, {"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5705 RTFM, Inc.", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5705", "source_sha256": "29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a"}, {"char_end": 5223, "char_start": 4967, "evidence_quote": "{\"id\": 1269016, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269016/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}", "fact_ids": ["normative_reference:5705"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "normative_reference", "relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5705"}, {"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6066 Huawei", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:6066", "source_sha256": "6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0"}, {"char_end": 12717, "char_start": 12466, "evidence_quote": "{\"id\": 1591472, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591472/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}", "fact_ids": ["normative_reference:6066"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "normative_reference", "relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:6066"}, {"evidence": [{"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6961 June 2013", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:6961", "source_sha256": "c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db"}, {"char_end": 10425, "char_start": 10169, "evidence_quote": "{\"id\": 1269043, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269043/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}", "fact_ids": ["normative_reference:6961"], "record_id": "ietf:datatracker-relation:rfc8446", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}], "kind": "normative_reference", "relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:6961"}, {"evidence": [{"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_ids": ["obsoletes:5077"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5077 H. Zhou", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5077", "source_sha256": "e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef"}], "kind": "obsoletes", "relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5077"}, {"evidence": [{"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_ids": ["obsoletes:5246"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5246 Independent", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5246", "source_sha256": "2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb"}], "kind": "obsoletes", "relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5246"}, {"evidence": [{"char_end": 218, "char_start": 146, "evidence_quote": "Obsoletes: 5077, 5246, 6961 August 2018", "fact_ids": ["obsoletes:6961"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6961 June 2013", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:6961", "source_sha256": "c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db"}], "kind": "obsoletes", "relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:6961"}, {"evidence": [{"char_end": 454, "char_start": 431, "evidence_quote": "draft-ietf-tls-tls13-28", "fact_ids": ["draft_revision"], "record_id": "ietf:draft:draft-ietf-tls-tls13-28", "source_sha256": "9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 8446 Mozilla", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 13945, "char_start": 13673, "evidence_quote": "{\"id\": 1295893, \"originaltargetaliasname\": \"rfc8446\", \"relationship\": \"/api/v1/name/docrelationshipname/became_rfc/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1295893/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8446/\"}", "fact_ids": ["became_rfc"], "record_id": "ietf:datatracker-relation:draft-ietf-tls-tls13:rfc8446", "source_sha256": "3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131"}], "kind": "published_as", "relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "target_record_id": "ietf:rfc:8446"}, {"evidence": [{"char_end": 238, "char_start": 219, "evidence_quote": "Updates: 5705, 6066", "fact_ids": ["updates:5705"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 5705 RTFM, Inc.", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:5705", "source_sha256": "29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a"}], "kind": "updates", "relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:5705"}, {"evidence": [{"char_end": 238, "char_start": 219, "evidence_quote": "Updates: 5705, 6066", "fact_ids": ["updates:6066"], "record_id": "ietf:rfc:8446", "source_sha256": "47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd"}, {"char_end": 145, "char_start": 73, "evidence_quote": "Request for Comments: 6066 Huawei", "fact_ids": ["rfc_number"], "record_id": "ietf:rfc:6066", "source_sha256": "6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0"}], "kind": "updates", "relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_record_id": "ietf:rfc:8446", "target_record_id": "ietf:rfc:6066"}], "schema_version": "longworld.ietf-workflow-manifest.v1", "semantic_facts_train_ready": false, "source_status": "public_api_export", "supporting_records": [{"draft_name": "draft-ietf-tls-tls13", "facts": [{"char_end": 673, "char_start": 643, "evidence_quote": "\"name\": \"draft-ietf-tls-tls13\"", "fact_id": "draft_name", "field": "draft_name", "raw_char_end": 673, "raw_char_start": 643, "value": "draft-ietf-tls-tls13"}], "kind": "datatracker_document", "occurred_at": "2020-03-07T08:26:41Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 2, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:a632cb66584c8f5d364d909b76eed121c1a9d00e341908463e019efad0faee0f", "record_id": "ietf:datatracker:draft-ietf-tls-tls13", "retrieval_url": "https://datatracker.ietf.org/api/v1/doc/document/draft-ietf-tls-tls13/", "revision": "28", "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "a632cb66584c8f5d364d909b76eed121c1a9d00e341908463e019efad0faee0f", "source_url": "https://datatracker.ietf.org/api/v1/doc/document/draft-ietf-tls-tls13/", "temporal_semantics": "source_event_time", "text": "{\"abstract\": \"This document specifies version 1.3 of the Transport Layer Security (TLS) protocol. TLS allows client/server applications to communicate over the Internet in a way that is designed to prevent eavesdropping, tampering, and message forgery.\\n\\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077, 5246, and 6961. This document also specifies new requirements for TLS 1.2 implementations.\", \"ad\": \"/api/v1/person/person/104851/\", \"expires\": \"2018-09-21T21:26:15Z\", \"external_url\": \"\", \"group\": \"/api/v1/group/group/1326/\", \"id\": 55251, \"intended_std_level\": \"/api/v1/name/intendedstdlevelname/ps/\", \"keywords\": \"[]\", \"name\": \"draft-ietf-tls-tls13\", \"note\": \"\", \"notify\": \"Sean Turner <[redacted-email]>\", \"pages\": 160, \"resource_uri\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"rev\": \"28\", \"rfc\": null, \"rfc_number\": null, \"shepherd\": \"/api/v1/person/email/[redacted-email]/\", \"states\": [\"/api/v1/doc/state/3/\", \"/api/v1/doc/state/108/\", \"/api/v1/doc/state/112/\", \"/api/v1/doc/state/7/\", \"/api/v1/doc/state/44/\"], \"std_level\": \"/api/v1/name/stdlevelname/ps/\", \"stream\": \"/api/v1/name/streamname/ietf/\", \"submissions\": [\"/api/v1/submit/submission/59397/\", \"/api/v1/submit/submission/59398/\", \"/api/v1/submit/submission/61729/\", \"/api/v1/submit/submission/64520/\", \"/api/v1/submit/submission/65644/\", \"/api/v1/submit/submission/67695/\", \"/api/v1/submit/submission/70123/\", \"/api/v1/submit/submission/70989/\", \"/api/v1/submit/submission/71776/\", \"/api/v1/submit/submission/72481/\", \"/api/v1/submit/submission/73392/\", \"/api/v1/submit/submission/74860/\", \"/api/v1/submit/submission/77260/\", \"/api/v1/submit/submission/78266/\", \"/api/v1/submit/submission/79977/\", \"/api/v1/submit/submission/80556/\", \"/api/v1/submit/submission/81099/\", \"/api/v1/submit/submission/81687/\", \"/api/v1/submit/submission/81857/\", \"/api/v1/submit/submission/84956/\", \"/api/v1/submit/submission/86252/\", \"/api/v1/submit/submission/92891/\", \"/api/v1/submit/submission/88268/\", \"/api/v1/submit/submission/91422/\", \"/api/v1/submit/submission/92066/\", \"/api/v1/submit/submission/93465/\", \"/api/v1/submit/submission/93544/\", \"/api/v1/submit/submission/94182/\", \"/api/v1/submit/submission/94319/\"], \"tags\": [], \"time\": \"2020-03-07T08:26:41Z\", \"title\": \"The Transport Layer Security (TLS) Protocol Version 1.3\", \"type\": \"/api/v1/name/doctypename/draft/\", \"uploaded_filename\": \"\", \"words\": 35738}", "text_sha256": "e51b82d004e563f85becf3bc398f5dea916940ec55b92cf7b728f78a625a293b"}, {"facts": [{"char_end": 708, "char_start": 691, "evidence_quote": "\"name\": \"rfc8446\"", "fact_id": "rfc_name", "field": "rfc_name", "raw_char_end": 708, "raw_char_start": 691, "value": "rfc8446"}], "kind": "datatracker_document", "occurred_at": "2026-05-20T15:34:35Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 0, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:89b01010d1ff76dc260539d5a453dde0b116d857db43c9e09d4a3dc062c9ce31", "record_id": "ietf:datatracker:rfc8446", "retrieval_url": "https://datatracker.ietf.org/api/v1/doc/document/rfc8446/", "rfc_number": 8446, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "89b01010d1ff76dc260539d5a453dde0b116d857db43c9e09d4a3dc062c9ce31", "source_url": "https://datatracker.ietf.org/api/v1/doc/document/rfc8446/", "temporal_semantics": "source_event_time", "text": "{\"abstract\": \"This document specifies version 1.3 of the Transport Layer Security (TLS) protocol. TLS allows client/server applications to communicate over the Internet in a way that is designed to prevent eavesdropping, tampering, and message forgery.\\n\\n This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077, 5246, and 6961. This document also specifies new requirements for TLS 1.2 implementations.\", \"ad\": \"/api/v1/person/person/104851/\", \"expires\": null, \"external_url\": \"\", \"group\": \"/api/v1/group/group/1326/\", \"id\": 130263, \"intended_std_level\": null, \"keywords\": \"['international data algorithm', 'symmetric', 'transport protocol layer', 'authentication', 'privacy']\", \"name\": \"rfc8446\", \"note\": \"\", \"notify\": \"\", \"pages\": 160, \"resource_uri\": \"/api/v1/doc/document/rfc8446/\", \"rev\": \"\", \"rfc\": \"8446\", \"rfc_number\": 8446, \"shepherd\": null, \"states\": [\"/api/v1/doc/state/177/\"], \"std_level\": \"/api/v1/name/stdlevelname/ps/\", \"stream\": \"/api/v1/name/streamname/ietf/\", \"submissions\": [], \"tags\": [\"/api/v1/name/doctagname/errata/\", \"/api/v1/name/doctagname/verified-errata/\"], \"time\": \"2026-05-20T15:34:35Z\", \"title\": \"The Transport Layer Security (TLS) Protocol Version 1.3\", \"type\": \"/api/v1/name/doctypename/rfc/\", \"uploaded_filename\": \"\", \"words\": 35738}", "text_sha256": "89b01010d1ff76dc260539d5a453dde0b116d857db43c9e09d4a3dc062c9ce31"}, {"draft_name": "draft-ietf-tls-tls13", "facts": [{"char_end": 13945, "char_start": 13673, "evidence_quote": "{\"id\": 1295893, \"originaltargetaliasname\": \"rfc8446\", \"relationship\": \"/api/v1/name/docrelationshipname/became_rfc/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1295893/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8446/\"}", "fact_id": "became_rfc", "field": "became_rfc", "raw_char_end": 13945, "raw_char_start": 13673, "value": "draft-ietf-tls-tls13->RFC 8446"}, {"char_end": 3140, "char_start": 2873, "evidence_quote": "{\"id\": 630806, \"originaltargetaliasname\": \"rfc6066\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630806/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}", "fact_id": "informative_reference:6066", "field": "informative_reference", "raw_char_end": 3140, "raw_char_start": 2873, "value": "/api/v1/doc/document/rfc6066/"}, {"char_end": 3409, "char_start": 3142, "evidence_quote": "{\"id\": 630807, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630807/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}", "fact_id": "informative_reference:5246", "field": "informative_reference", "raw_char_end": 3409, "raw_char_start": 3142, "value": "/api/v1/doc/document/rfc5246/"}, {"char_end": 3678, "char_start": 3411, "evidence_quote": "{\"id\": 630805, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630805/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}", "fact_id": "informative_reference:5077", "field": "informative_reference", "raw_char_end": 3678, "raw_char_start": 3411, "value": "/api/v1/doc/document/rfc5077/"}, {"char_end": 6098, "char_start": 5831, "evidence_quote": "{\"id\": 630795, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630795/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}", "fact_id": "normative_reference:5705", "field": "normative_reference", "raw_char_end": 6098, "raw_char_start": 5831, "value": "/api/v1/doc/document/rfc5705/"}, {"char_end": 12864, "char_start": 12597, "evidence_quote": "{\"id\": 630768, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630768/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}", "fact_id": "normative_reference:6961", "field": "normative_reference", "raw_char_end": 12864, "raw_char_start": 12597, "value": "/api/v1/doc/document/rfc6961/"}], "kind": "datatracker_relation", "occurred_at": "2026-09-07T04:14:32.295913Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 0, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131", "record_id": "ietf:datatracker-relation:draft-ietf-tls-tls13:rfc8446", "retrieval_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=draft-ietf-tls-tls13&limit=100", "rfc_number": 8446, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131", "source_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=draft-ietf-tls-tls13&limit=100", "temporal_semantics": "retrieval_observation_only", "text": "{\"meta\": {\"limit\": 100, \"next\": null, \"offset\": 0, \"previous\": null, \"total_count\": 51}, \"objects\": [{\"id\": 630809, \"originaltargetaliasname\": \"bcp72\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630809/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/bcp72/\"}, {\"id\": 630802, \"originaltargetaliasname\": \"bcp26\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630802/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/bcp26/\"}, {\"id\": 630782, \"originaltargetaliasname\": \"bcp106\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630782/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/bcp106/\"}, {\"id\": 630774, \"originaltargetaliasname\": \"bcp14\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630774/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/bcp14/\"}, {\"id\": 282421, \"originaltargetaliasname\": \"draft-ietf-tls-rfc5246-bis\", \"relationship\": \"/api/v1/name/docrelationshipname/replaces/\", \"resource_uri\": \"/api/v1/doc/relateddocument/282421/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/draft-ietf-tls-rfc5246-bis/\"}, {\"id\": 630813, \"originaltargetaliasname\": \"rfc7568\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630813/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7568/\"}, {\"id\": 630810, \"originaltargetaliasname\": \"draft-ietf-tls-iana-registry-updates\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630810/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/draft-ietf-tls-iana-registry-updates/\"}, {\"id\": 630812, \"originaltargetaliasname\": \"rfc5116\", \"relationship\": \"/api/v1/name/docrelationshipname/refunk/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630812/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5116/\"}, {\"id\": 630811, \"originaltargetaliasname\": \"rfc5280\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630811/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5280/\"}, {\"id\": 630808, \"originaltargetaliasname\": \"rfc5756\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630808/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5756/\"}, {\"id\": 630806, \"originaltargetaliasname\": \"rfc6066\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630806/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}, {\"id\": 630807, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630807/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}, {\"id\": 630805, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630805/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}, {\"id\": 630804, \"originaltargetaliasname\": \"rfc8174\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630804/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8174/\"}, {\"id\": 630803, \"originaltargetaliasname\": \"rfc8032\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630803/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8032/\"}, {\"id\": 630801, \"originaltargetaliasname\": \"rfc2119\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630801/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc2119/\"}, {\"id\": 630800, \"originaltargetaliasname\": \"rfc6347\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630800/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6347/\"}, {\"id\": 630799, \"originaltargetaliasname\": \"rfc4492\", \"relationship\": \"/api/v1/name/docrelationshipname/refunk/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630799/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc4492/\"}, {\"id\": 630798, \"originaltargetaliasname\": \"rfc7539\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630798/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7539/\"}, {\"id\": 630797, \"originaltargetaliasname\": \"rfc7301\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630797/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7301/\"}, {\"id\": 630796, \"originaltargetaliasname\": \"rfc6176\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630796/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6176/\"}, {\"id\": 630795, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630795/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}, {\"id\": 630794, \"originaltargetaliasname\": \"rfc6979\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630794/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6979/\"}, {\"id\": 630793, \"originaltargetaliasname\": \"rfc5764\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630793/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5764/\"}, {\"id\": 630792, \"originaltargetaliasname\": \"rfc8017\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630792/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8017/\"}, {\"id\": 630791, \"originaltargetaliasname\": \"draft-ietf-tls-tls13-vectors\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630791/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/draft-ietf-tls-tls13-vectors/\"}, {\"id\": 630790, \"originaltargetaliasname\": \"rfc6520\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630790/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6520/\"}, {\"id\": 630789, \"originaltargetaliasname\": \"rfc5869\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630789/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5869/\"}, {\"id\": 630788, \"originaltargetaliasname\": \"rfc6960\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630788/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6960/\"}, {\"id\": 630787, \"originaltargetaliasname\": \"rfc7748\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630787/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7748/\"}, {\"id\": 630786, \"originaltargetaliasname\": \"rfc8126\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630786/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8126/\"}, {\"id\": 630785, \"originaltargetaliasname\": \"rfc7924\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630785/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7924/\"}, {\"id\": 630784, \"originaltargetaliasname\": \"rfc5929\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630784/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc5929/\"}, {\"id\": 630783, \"originaltargetaliasname\": \"rfc8305\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630783/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8305/\"}, {\"id\": 630781, \"originaltargetaliasname\": \"rfc6091\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630781/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6091/\"}, {\"id\": 630780, \"originaltargetaliasname\": \"rfc4086\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630780/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc4086/\"}, {\"id\": 630779, \"originaltargetaliasname\": \"rfc7627\", \"relationship\": \"/api/v1/name/docrelationshipname/refunk/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630779/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7627/\"}, {\"id\": 630778, \"originaltargetaliasname\": \"rfc6655\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630778/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6655/\"}, {\"id\": 630777, \"originaltargetaliasname\": \"rfc7250\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630777/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7250/\"}, {\"id\": 630776, \"originaltargetaliasname\": \"rfc7507\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630776/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7507/\"}, {\"id\": 630775, \"originaltargetaliasname\": \"rfc7685\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630775/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7685/\"}, {\"id\": 630773, \"originaltargetaliasname\": \"rfc2104\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630773/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc2104/\"}, {\"id\": 630771, \"originaltargetaliasname\": \"rfc4346\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630771/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc4346/\"}, {\"id\": 630772, \"originaltargetaliasname\": \"rfc4366\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630772/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc4366/\"}, {\"id\": 630770, \"originaltargetaliasname\": \"rfc6962\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630770/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6962/\"}, {\"id\": 630769, \"originaltargetaliasname\": \"rfc3552\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630769/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc3552/\"}, {\"id\": 630768, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630768/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}, {\"id\": 630767, \"originaltargetaliasname\": \"rfc7919\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630767/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7919/\"}, {\"id\": 630766, \"originaltargetaliasname\": \"rfc7465\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630766/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7465/\"}, {\"id\": 630814, \"originaltargetaliasname\": \"rfc7230\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/630814/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc7230/\"}, {\"id\": 1295893, \"originaltargetaliasname\": \"rfc8446\", \"relationship\": \"/api/v1/name/docrelationshipname/became_rfc/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1295893/\", \"source\": \"/api/v1/doc/document/draft-ietf-tls-tls13/\", \"target\": \"/api/v1/doc/document/rfc8446/\"}]}", "text_sha256": "3878cfbede9a852d9bc01fde7f1f5d982afb1fedf9511208f908601121075131"}, {"facts": [{"char_end": 3417, "char_start": 3161, "evidence_quote": "{\"id\": 1269005, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269005/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}", "fact_id": "informative_reference:5246", "field": "informative_reference", "raw_char_end": 3417, "raw_char_start": 3161, "value": "/api/v1/doc/document/rfc5246/"}, {"char_end": 3675, "char_start": 3419, "evidence_quote": "{\"id\": 1269006, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269006/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}", "fact_id": "informative_reference:5077", "field": "informative_reference", "raw_char_end": 3675, "raw_char_start": 3419, "value": "/api/v1/doc/document/rfc5077/"}, {"char_end": 5223, "char_start": 4967, "evidence_quote": "{\"id\": 1269016, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269016/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}", "fact_id": "normative_reference:5705", "field": "normative_reference", "raw_char_end": 5223, "raw_char_start": 4967, "value": "/api/v1/doc/document/rfc5705/"}, {"char_end": 10425, "char_start": 10169, "evidence_quote": "{\"id\": 1269043, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269043/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}", "fact_id": "normative_reference:6961", "field": "normative_reference", "raw_char_end": 10425, "raw_char_start": 10169, "value": "/api/v1/doc/document/rfc6961/"}, {"char_end": 12717, "char_start": 12466, "evidence_quote": "{\"id\": 1591472, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591472/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}", "fact_id": "normative_reference:6066", "field": "normative_reference", "raw_char_end": 12717, "raw_char_start": 12466, "value": "/api/v1/doc/document/rfc6066/"}], "kind": "datatracker_relation", "occurred_at": "2026-09-07T04:14:41.148823Z", "privacy_review": {"credential_test_vector_redaction_count": 0, "email_redaction_count": 0, "emails": "redacted", "private_key_redaction_count": 0, "private_key_redaction_policy": "exact_sha256_allowlist", "scanner": "longworld-public-secret-patterns", "scanner_revision": "v3", "secrets": "fail_closed"}, "provenance_id": "sha256:d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7", "record_id": "ietf:datatracker-relation:rfc8446", "retrieval_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=rfc8446&limit=100", "rfc_number": 8446, "source_family": "ietf_standards", "source_origin": "real_public", "source_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7", "source_url": "https://datatracker.ietf.org/api/v1/doc/relateddocument/?source__name=rfc8446&limit=100", "temporal_semantics": "retrieval_observation_only", "text": "{\"meta\": {\"limit\": 100, \"next\": null, \"offset\": 0, \"previous\": null, \"total_count\": 57}, \"objects\": [{\"id\": 1269002, \"originaltargetaliasname\": \"bcp72\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269002/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/bcp72/\"}, {\"id\": 1269009, \"originaltargetaliasname\": \"bcp26\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269009/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/bcp26/\"}, {\"id\": 1269029, \"originaltargetaliasname\": \"bcp106\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269029/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/bcp106/\"}, {\"id\": 1269037, \"originaltargetaliasname\": \"bcp14\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269037/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/bcp14/\"}, {\"id\": 666521, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/obs/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666521/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}, {\"id\": 666523, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/updates/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666523/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}, {\"id\": 666520, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/obs/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666520/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}, {\"id\": 666524, \"originaltargetaliasname\": \"rfc6066\", \"relationship\": \"/api/v1/name/docrelationshipname/updates/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666524/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}, {\"id\": 666522, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/obs/\", \"resource_uri\": \"/api/v1/doc/relateddocument/666522/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}, {\"id\": 1268998, \"originaltargetaliasname\": \"rfc7568\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1268998/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7568/\"}, {\"id\": 1269001, \"originaltargetaliasname\": \"rfc5280\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269001/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5280/\"}, {\"id\": 1269003, \"originaltargetaliasname\": \"rfc5756\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269003/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5756/\"}, {\"id\": 1269005, \"originaltargetaliasname\": \"rfc5246\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269005/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5246/\"}, {\"id\": 1269006, \"originaltargetaliasname\": \"rfc5077\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269006/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5077/\"}, {\"id\": 1269007, \"originaltargetaliasname\": \"rfc8174\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269007/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8174/\"}, {\"id\": 1269008, \"originaltargetaliasname\": \"rfc8032\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269008/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8032/\"}, {\"id\": 1269010, \"originaltargetaliasname\": \"rfc2119\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269010/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc2119/\"}, {\"id\": 1269011, \"originaltargetaliasname\": \"rfc6347\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269011/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6347/\"}, {\"id\": 1269015, \"originaltargetaliasname\": \"rfc6176\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269015/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6176/\"}, {\"id\": 1269016, \"originaltargetaliasname\": \"rfc5705\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269016/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5705/\"}, {\"id\": 1269018, \"originaltargetaliasname\": \"rfc5764\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269018/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5764/\"}, {\"id\": 1269019, \"originaltargetaliasname\": \"rfc8017\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269019/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8017/\"}, {\"id\": 1269020, \"originaltargetaliasname\": \"draft-ietf-tls-tls13-vectors\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269020/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/draft-ietf-tls-tls13-vectors/\"}, {\"id\": 1269021, \"originaltargetaliasname\": \"rfc6520\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269021/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6520/\"}, {\"id\": 1269023, \"originaltargetaliasname\": \"rfc6960\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269023/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6960/\"}, {\"id\": 1269024, \"originaltargetaliasname\": \"rfc7748\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269024/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7748/\"}, {\"id\": 1269025, \"originaltargetaliasname\": \"rfc8126\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269025/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8126/\"}, {\"id\": 1269026, \"originaltargetaliasname\": \"rfc7924\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269026/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7924/\"}, {\"id\": 1269027, \"originaltargetaliasname\": \"rfc5929\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269027/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5929/\"}, {\"id\": 1269028, \"originaltargetaliasname\": \"rfc8305\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269028/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8305/\"}, {\"id\": 1269030, \"originaltargetaliasname\": \"rfc6091\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269030/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6091/\"}, {\"id\": 1269031, \"originaltargetaliasname\": \"rfc4086\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269031/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc4086/\"}, {\"id\": 1269034, \"originaltargetaliasname\": \"rfc7250\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269034/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7250/\"}, {\"id\": 1269035, \"originaltargetaliasname\": \"rfc7507\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269035/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7507/\"}, {\"id\": 1269036, \"originaltargetaliasname\": \"rfc7685\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269036/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7685/\"}, {\"id\": 1269038, \"originaltargetaliasname\": \"rfc2104\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269038/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc2104/\"}, {\"id\": 1269039, \"originaltargetaliasname\": \"rfc4346\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269039/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc4346/\"}, {\"id\": 1269040, \"originaltargetaliasname\": \"rfc4366\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269040/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc4366/\"}, {\"id\": 1269042, \"originaltargetaliasname\": \"rfc3552\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269042/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc3552/\"}, {\"id\": 1269043, \"originaltargetaliasname\": \"rfc6961\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269043/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6961/\"}, {\"id\": 1269044, \"originaltargetaliasname\": \"rfc7919\", \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269044/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7919/\"}, {\"id\": 1269045, \"originaltargetaliasname\": \"rfc7465\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269045/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7465/\"}, {\"id\": 1269046, \"originaltargetaliasname\": \"rfc7230\", \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1269046/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7230/\"}, {\"id\": 1591467, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591467/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5869/\"}, {\"id\": 1591468, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591468/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6655/\"}, {\"id\": 1591469, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591469/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7301/\"}, {\"id\": 1591470, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591470/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6962/\"}, {\"id\": 1591471, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591471/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6979/\"}, {\"id\": 1591472, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591472/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6066/\"}, {\"id\": 1591473, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591473/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8449/\"}, {\"id\": 1591474, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591474/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc5116/\"}, {\"id\": 1591475, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591475/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8422/\"}, {\"id\": 1591476, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591476/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc4492/\"}, {\"id\": 1591477, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refnorm/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591477/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8439/\"}, {\"id\": 1591478, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591478/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc8447/\"}, {\"id\": 1591479, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591479/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc7627/\"}, {\"id\": 1591480, \"originaltargetaliasname\": null, \"relationship\": \"/api/v1/name/docrelationshipname/refinfo/\", \"resource_uri\": \"/api/v1/doc/relateddocument/1591480/\", \"source\": \"/api/v1/doc/document/rfc8446/\", \"target\": \"/api/v1/doc/document/rfc6101/\"}]}", "text_sha256": "d0ebf7faa70a4a9b17e88b1ea892717959f503b4610b0f7c6d416d72c2409ce7"}]}, "source_manifest_sha256": "287ca4031b223ad818d21781c575905d5ed30eb7e4e0e38fef196cf1ca1865b4"}, "length_bucket": "64k", "local_probe_trust_isolation": "non_independent_local_diagnostic", "motif": "tls13_handshake_succession+obsoletes_updates_closure", "pipeline_capabilities": {"generic_promotion": true, "generic_strict_replay": true}, "production_eligible": false, "promoted": true, "promotion": {"candidate_sha256": "b5f0e0ba4612029242d9b8451f7f0560b57a323f65336eba2281eba4efbc8c21", "dense_audit_sha256": "fd88cf67f1f03db08590a7ed02c981c764dbcaecd171bb07f7a8ec23b9e6e037", "dense_chunking": {"aggregation": "max_similarity", "max_tokens": 192, "overlap_tokens": 32, "strategy": "tokenizer_token_windows"}, "dense_model_backend": "sentence-transformers-6.0.0", "dense_model_id": "sentence-transformers/all-MiniLM-L6-v2", "dense_model_provider": "huggingface", "dense_model_revision": "1110a243fdf4706b3f48f1d95db1a4f5529b4d41", "dense_ranking_sha256": "bd1e8beb9ba383486945101b359a5f72b812eeb7a458cf6ad80ab23dbbadfacc", "dense_score_metric": "dot_product", "dense_top_k": 3, "real_source_verified": true, "release_selection_sha256": "46c412e55a5e391ac46af172f4a11623a01c552075d5ba2d80f6604f48b5f89d", "schema_version": "train-ready-promotion-v2", "strict_replay_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "strict_replay_revision": "longworld-strict-replay-v5", "task_candidate_content_commitment": {"content_sha256": "03a51274983df80c4bafd3265ee3551813545131396a63e59b4801c0941afec8", "length_bucket": "64k", "view": "ordered_artifact_view", "world_id": "ietf-tls13-handshake-succession-v4"}, "task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "sidecar_schema_version": "longworld.task-replay-sidecar.v3"}, "task_semantic_commitment_sha256": "dfad22d5f03f0a1bb543ca914cd6b7e0461f5629cb0c87f3c05816c3befde9bd", "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb"}, "promotion_eligible": true, "query_id": "ietf-tls13-handshake-succession-v4:64k:standard:ordered_artifact_view", "query_timing": "first", "question": "Resolve the effective TLS 1.3 handshake protocol succession at 2026-01-31T00:00:00Z for this scenario (canonical JSON): {\"protocol\":\"tls\",\"publication\":\"rfc8446\",\"succession\":\"obsoletes_and_updates\",\"version\":\"1.3\"}. Return exactly one JSON object with these keys in this order: [\"current_protocol\",\"obsoletes_session_tickets\",\"obsoletes_tls12\",\"obsoletes_ocsp_stapling\",\"updates_exporters\",\"updates_extensions\"]. Use this exact per-field output codebook (canonical JSON): {\"current_protocol\":{\"code\":\"TLS_1_3_RFC8446\",\"meaning\":\"RFC 8446 is TLS protocol version 1.3\"},\"obsoletes_ocsp_stapling\":{\"code\":\"RFC6961\",\"meaning\":\"RFC 8446 obsoletes multiple certificate status requests\"},\"obsoletes_session_tickets\":{\"code\":\"RFC5077\",\"meaning\":\"RFC 8446 obsoletes TLS session resumption tickets\"},\"obsoletes_tls12\":{\"code\":\"RFC5246\",\"meaning\":\"RFC 8446 obsoletes TLS protocol version 1.2\"},\"updates_exporters\":{\"code\":\"RFC5705\",\"meaning\":\"RFC 8446 updates TLS keying material exporters\"},\"updates_extensions\":{\"code\":\"RFC6066\",\"meaning\":\"RFC 8446 updates TLS extension definitions\"}}. Resolve each field independently from the supplied RFC graph; use the string \"UNKNOWN\" only for a field whose required evidence or relation is absent.", "real_source_family_ids": ["ietf_standards"], "real_source_token_ratio": 0.994230152794102, "real_source_verified": true, "real_source_workflow_ids": ["task-source:73e262c5a2939b145819"], "schema_version": "longworld.ietf-tls13-generation-candidate.v1", "semantic_base_task_id": "250a7cfbf1fb2c0a24b9", "semantic_growth_group_id": "ietf-tls13-handshake-succession-v4|ietf-tls13-handshake-succession", "semantic_tokens": {"causal_supporting": 16788, "event_bearing": 48342, "generic_background": 0, "internal": 65134, "measurement_basis": "exact_pinned_tokenizer_replay_relation_path_whole_artifact_upper_bound", "proof_bearing": 48342}, "source_binding": {"signed_manifest_sha256": "287ca4031b223ad818d21781c575905d5ed30eb7e4e0e38fef196cf1ca1865b4"}, "source_family_ids": ["ietf_standards"], "source_origins": ["real_public"], "source_record_ids": ["ietf:draft:draft-ietf-tls-tls13-28", "ietf:rfc:5077", "ietf:rfc:5246", "ietf:rfc:5705", "ietf:rfc:6066", "ietf:rfc:6961", "ietf:rfc:8446"], "source_record_ids_by_artifact": {"ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2": ["ietf:draft:draft-ietf-tls-tls13-28"], "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8": ["ietf:rfc:5077"], "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681": ["ietf:rfc:5246"], "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1": ["ietf:rfc:5246"], "ietf:rfc:5246:chars:0072013-0109924:596a422d1201": ["ietf:rfc:5246"], "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44": ["ietf:rfc:5705"], "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44": ["ietf:rfc:6066"], "ietf:rfc:6961:chars:0000000-0021461:bde82964b134": ["ietf:rfc:6961"], "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4": ["ietf:rfc:8446"], "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093": ["ietf:rfc:8446"]}, "source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2018-03-01T00:00:00Z", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_relation_kind": "published_as"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "updates"}], "source_relation_id": "6ee046ba0ec12a545607", "source_relation_ids": ["ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "ietf:updates:ietf:rfc:8446:ietf:rfc:6066"], "split": "train", "split_strategy": "world", "strict_replay_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "strict_support_event_count": 6, "task_proof_receipt": {"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "artifact_aligned_windows": {"16k": {"enumeration_sha256": "b7d1fa0abebbb673c57714fbe2454484c04d3b9b0c414a92ad071b6529d3bcde", "max_artifact_count": 4, "max_span_tokens": 16324, "status": "insufficient", "strict_window_count": 19, "window_tokens": 16384}, "4k": {"enumeration_sha256": "a06d80f8ba4acda5c7d701e00f394bb5dcca6370e3c3d9e99768251ab7b309cc", "max_artifact_count": 1, "max_span_tokens": 4053, "status": "insufficient", "strict_window_count": 3, "window_tokens": 4096}, "8k": {"enumeration_sha256": "0dd1c5d0253aba64bd0a6937dc169a3a91956a5ecc4e59e9ed83bbe6fb80f9de", "max_artifact_count": 2, "max_span_tokens": 8187, "status": "insufficient", "strict_window_count": 9, "window_tokens": 8192}}, "artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "end_token": 9837, "start_token": 0, "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "end_token": 18029, "start_token": 9842, "text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "end_token": 26166, "start_token": 18033, "text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "end_token": 34312, "start_token": 26170, "text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "end_token": 38369, "start_token": 34316, "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "end_token": 51231, "start_token": 38374, "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "end_token": 56417, "start_token": 51236, "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "end_token": 56525, "start_token": 56422, "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "end_token": 64670, "start_token": 56529, "text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "end_token": 65134, "start_token": 64673, "text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5"}], "checks": {"answer_surface_free": true, "artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_top3_prefixes_insufficient": true, "body_and_classification_valid": true, "counterfactual_changes_answer": true, "empty_selection_insufficient": true, "exact_token_count_recomputed": true, "full_counterfactual_sufficient": true, "full_replay_sufficient": true, "lexical_tfidf_top3_prefixes_insufficient": true, "minimal_counterfactual_sufficient": true, "minimal_replay_sufficient": true, "minimum_complexity_met": true, "no_shortcut": true, "raw_token_executable_windows_insufficient": true, "remove_one_fails": true, "single_essential_insufficient": true, "source_relations_replayed": true}, "essential_artifact_ids": ["ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"], "proof_input_sha256": "b5f0e0ba4612029242d9b8451f7f0560b57a323f65336eba2281eba4efbc8c21", "raw_token_offset_windows": {"16k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "5bc01bf000da7ac2ade7ec5898db05447246702b32b64dbdb88f40a618290a69", "max_artifact_count": 5, "monotonicity_proven": false, "semantic_state_count": 14, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "9a9eb35d89e2afee75c3eaadb5075ff576d4bb7216063b1a25bfbf9890abd41c", "exact_raw_executable": {"enumeration_sha256": "9a9eb35d89e2afee75c3eaadb5075ff576d4bb7216063b1a25bfbf9890abd41c", "exhaustive_start_count": 48751, "semantic_state_count": 14, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 14, "window_tokens": 16384}, "4k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "19520947cfb7d39e7ba09721be0e6bfe2f65b2c67daa8eed8f61436684c0bdce", "max_artifact_count": 3, "monotonicity_proven": false, "semantic_state_count": 18, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "022aba6f56ecbfde2808c571823ddf1efd3b9b39d9069817e25e69c73c914de2", "exact_raw_executable": {"enumeration_sha256": "022aba6f56ecbfde2808c571823ddf1efd3b9b39d9069817e25e69c73c914de2", "exhaustive_start_count": 61039, "semantic_state_count": 4, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 4, "window_tokens": 4096}, "8k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "4db7b6ec0c02879b159738a1b136d4260bf4a23424959ca13f7b1633f4e86094", "max_artifact_count": 4, "monotonicity_proven": false, "semantic_state_count": 18, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "cea077c73f3dc2a8273cea811e0c0ce3ea7b1f7f469986e8ac5190c3159e15c0", "exact_raw_executable": {"enumeration_sha256": "cea077c73f3dc2a8273cea811e0c0ce3ea7b1f7f469986e8ac5190c3159e15c0", "exhaustive_start_count": 56943, "semantic_state_count": 10, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 10, "window_tokens": 8192}}, "receipt_sha256": "0426657dedc1fdd08e4844fdc1a48cad07666ba7ce72f08cfb811f9b55964be1", "replay": {"counterfactual_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "empty_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "full_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "minimal_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "remove_one": [{"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}], "single_essential": [{"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}]}, "retrieval": {"bm25": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "score": "117.4899053298914"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "score": "85.661410221096119"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "score": "73.286337280605935"}]}, "lexical_tfidf": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "c1af44132ab8172e277a7aeb52229454aa629ca8cbfe8e7ea88b299909f9abb7"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "score": "0.42605918578853391"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "score": "0.41657976134259173"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "score": "0.32513082216618144"}]}}, "schema_version": "longworld.task-proof-receipt.v5", "sidecar_schema_version": "longworld.task-replay-sidecar.v3", "task_replay_sidecar_sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "tokenizer": {"asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "document_context_tokens": 65134, "model_id": "Qwen/Qwen3.5-4B", "revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "serialized_context_tokens": 65513}, "window_scope": "exact_raw_slice_replay_with_separate_intersection_upper_bound"}, "task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "sidecar_schema_version": "longworld.task-replay-sidecar.v3"}, "task_view_projection": {"adapter_context_header": null, "adapter_replay_answer_sha256": "a627dbccb9ab1c93890f576e49c0775aaf1e09dabcad115191e8972295391099", "artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5"}], "chronology": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "order_key": "2008-01-01T00:00:00Z|ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "order_key": "2008-08-01T00:00:00Z|ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "order_key": "2008-08-01T00:00:00Z|ietf:rfc:5246:chars:0035597-0072013:3f180e49add1"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "order_key": "2008-08-01T00:00:00Z|ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "order_key": "2010-03-01T00:00:00Z|ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "order_key": "2011-01-01T00:00:00Z|ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "order_key": "2013-06-01T00:00:00Z|ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "order_key": "2018-03-01T00:00:00Z|ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "order_key": "2018-08-01T00:00:00Z|ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "order_key": "2018-08-01T00:00:00Z|ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093"}], "derivation_revision": "longworld.task-view-derivation.v4", "document_context_sha256": "17f118290fad008405f1d83e43e9740fe9b8b7775606faf288249a728f7dcbfc", "dossier_id": "a00388a585cd2f4d89b3", "materialized_replay_answer_sha256": "a627dbccb9ab1c93890f576e49c0775aaf1e09dabcad115191e8972295391099", "parent_artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "provenance_id": "source-span-sha256:e9f9cb247c4930a773dadf87b4cedf06cad9bbeec5e364d08fd5a83f678817ef:0:41965:80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49", "source_origin": "real_public", "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:0:35597:ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6", "source_origin": "real_public", "text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:35597:72013:3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69", "source_origin": "real_public", "text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "provenance_id": "source-span-sha256:2ebe1b8e651696aeacc8ab3848bedb465505737901ea3b691acf40c392958feb:72013:109924:596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef", "source_origin": "real_public", "text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "provenance_id": "source-span-sha256:29c829cbdf29f4f6a1a0f59eacddf0503bfa67c22af28f127e25b502b3eb2c7a:0:16341:4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473", "source_origin": "real_public", "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "provenance_id": "source-span-sha256:6a33cceaa5e1b6518923d9d04ffaa738bad0417806b3c94a995484e140f5cbc0:0:55067:a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813", "source_origin": "real_public", "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "provenance_id": "source-span-sha256:c33e7e30760c8852509fbdae89f2daf3efcc424d6a90651de048efd6685745db:0:21461:bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a", "source_origin": "real_public", "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "provenance_id": "source-span-sha256:9d3143dcb56e38c71f088e1e534858a5956888bae4dae536f4c16646f84a55a6:0:342:d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b", "source_origin": "real_public", "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "provenance_id": "source-span-sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd:0:38698:61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e", "source_origin": "real_public", "text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "provenance_id": "source-span-sha256:47871bc8820a2c3b6ea89f061055577058862cf543686b82d10131239702b3bd:335784:337410:0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5", "source_origin": "real_public", "text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5"}], "parent_candidate_sha256": "3a56e24f453b328fb4635d85eaa9c95fe023fe8be199a806064111960bea07df", "parent_source_binding_sha256": "fc34a690fcb8ffffb64df92f1df724ea75d43ce88c10daf432fb085fee09ddbe", "parent_task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "e77fe08f470f227b433656f0fac14e677653bdb1f53faa939175c5c883608e76", "sidecar_schema_version": "longworld.task-replay-sidecar.v1"}, "schema_version": "longworld.task-view-projection.v1", "source_token_measurement_receipt": {"final_prompt_tokens": 65513, "measurement_basis": "final_prompt_real_source_marginal", "parent_artifact_token_contributions": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "parent_text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49", "retained_as_real": true, "token_contribution": 9837}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "parent_text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6", "retained_as_real": true, "token_contribution": 8192}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "parent_text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69", "retained_as_real": true, "token_contribution": 8137}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "parent_text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef", "retained_as_real": true, "token_contribution": 8146}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "parent_text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473", "retained_as_real": true, "token_contribution": 4057}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "parent_text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813", "retained_as_real": true, "token_contribution": 12862}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "parent_text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a", "retained_as_real": true, "token_contribution": 5186}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "parent_text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b", "retained_as_real": true, "token_contribution": 108}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "parent_text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e", "retained_as_real": true, "token_contribution": 8145}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "parent_text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5", "retained_as_real": true, "token_contribution": 464}], "parent_document_context_tokens": 65134, "parent_real_source_token_ratio": 1.0, "projected_document_context_tokens": 65134, "real_source_marginal_tokens": 65135, "real_source_token_ratio": 0.994230152794102, "retained_parent_document_tokens": 65134, "schema_version": "longworld.source-token-measurement-receipt.v2", "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "tokenizer_model_id": "Qwen/Qwen3.5-4B", "tokenizer_revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "without_real_prompt_tokens": 378}, "view": "ordered_artifact_view"}, "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "tokenizer_context_tokens": 65513, "tokenizer_model_id": "Qwen/Qwen3.5-4B", "tokenizer_revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "train_ready": true, "training_objective": "sft", "trust_scope": "local_probe", "trust_valid_for_production": false, "verification": {"artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_topk_insufficient": true, "candidate_mode": false, "closed_book_unsolved": true, "contiguous_windows_insufficient": true, "counterfactual_changes_answer": true, "counterfactual_replay_sufficient": true, "distractor_invariance_gold": true, "embedding_topk_insufficient": true, "essential_single_doc_insufficient": true, "essential_surface_gold_free": true, "essential_text_grounded": true, "full_sufficient": true, "lexical_tfidf_topk_insufficient": true, "local_window_insufficient": true, "min_complexity": true, "minimal_sufficient": true, "no_shortcut": true, "production_mode": true, "question_only_unsolved": true, "remove_one_fails": true, "schema_ok": true, "semantic_sufficient": true, "strict_executable_sufficient": true, "surface_match": true}, "verified_derived_order_relation_edges": [{"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2018-03-01T00:00:00Z", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_relation_kind": "published_as"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "updates"}], "view": "ordered_artifact_view", "view_verification": {"classification_ok": true, "content_gate_eligible": true, "essential_present": true, "expected_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "global_proof_green": true, "production_eligible": false, "semantic_text_grounded": true, "strict_replay_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}"}, "workflow_ids": ["task-source:73e262c5a2939b145819"], "workflow_kinds": ["real_source_derived"], "world_id": "ietf-tls13-handshake-succession-v4"} diff --git a/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/quality_report.json b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/quality_report.json new file mode 100644 index 0000000000000000000000000000000000000000..8badb251aa8633fb4c4d1a568528f8cd175427bd --- /dev/null +++ b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/quality_report.json @@ -0,0 +1,81 @@ +{ + "attestation": { + "digest": "df31553322f2f079d7a59282ae871f213c1212f4580a381a044a1859501d3f97", + "environment": "probe", + "key_id": "probe-report-882ad33b463bdee162c2f5c2738cfe18", + "purpose": "quality_report", + "role": "report", + "scheme": "hmac-sha256-v2" + }, + "by_domain": { + "standards": 6 + }, + "by_length": { + "128k": 3, + "64k": 3 + }, + "by_motif": { + "tls13_handshake_succession+obsoletes_updates_closure": 6 + }, + "by_split": { + "train": 6 + }, + "candidate_generation_retention": 1.0, + "candidate_quality_report_sha256": "810fa0a21659c7e60ea037520c9640977fb1b07a406e6a7b373edabc1887d843", + "candidate_row_set_sha256": "ac7a533fa940f69d59e313b6ddf0a34aa54c92fe072f4484de2c7e9afc327460", + "content_gate_eligible": true, + "data_product": "worldlong_release_union_v1", + "data_stage": "train_ready", + "diagnostic_only": true, + "local_probe_trust_isolation": "non_independent_local_diagnostic", + "n_candidate_rows": 6, + "n_candidate_worlds": 1, + "n_candidates": 6, + "n_clones": 0, + "n_promoted": 6, + "n_rows": 6, + "n_world_ids": 1, + "n_worlds": 1, + "predecessor_gate_receipt_sha256": null, + "predecessor_profile_id": null, + "predecessor_report_sha256": null, + "production_eligible": false, + "promoted_domain_world_quotas": { + "standards": 1 + }, + "promoted_row_set_sha256": "cb4990de2b7b129cc7c0455e4c94576ffaa5b59759150be07bad2f75c5f6aeea", + "promoted_split_row_set_sha256": { + "eval": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", + "train": "cb4990de2b7b129cc7c0455e4c94576ffaa5b59759150be07bad2f75c5f6aeea" + }, + "promotion_retention": 1.0, + "real_worlds_by_split": { + "eval": 0, + "train": 1 + }, + "release_profile_id": "p57-ietf-tls13-64k-128k-extension-probe-1-v1", + "release_profile_sha256": "e6498c7886a440afd68272229deb5d5e02643115abc831bb58605196a637a7bc", + "release_selection_sha256": "46c412e55a5e391ac46af172f4a11623a01c552075d5ba2d80f6604f48b5f89d", + "report_binding_revision": "longworld-quality-binding-v4", + "retention": 1.0, + "schema_version": "longworld-release-train-ready-report-v1", + "split_worlds_by_domain": { + "eval": { + "standards": [] + }, + "train": { + "standards": [ + "ietf-tls13-handshake-succession-v4" + ] + } + }, + "target_promoted_worlds": 1, + "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", + "tokenizer_model_id": "Qwen/Qwen3.5-4B", + "tokenizer_revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", + "trust_scope": "local_probe", + "trust_valid_for_production": false, + "worlds_by_domain": { + "standards": 1 + } +} diff --git a/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/quality_report.jsonl b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/quality_report.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..ead672f01a61f4032721e75d7d44ecc204acba3e --- /dev/null +++ b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/quality_report.jsonl @@ -0,0 +1 @@ +{"attestation": {"digest": "df31553322f2f079d7a59282ae871f213c1212f4580a381a044a1859501d3f97", "environment": "probe", "key_id": "probe-report-882ad33b463bdee162c2f5c2738cfe18", "purpose": "quality_report", "role": "report", "scheme": "hmac-sha256-v2"}, "by_domain": {"standards": 6}, "by_length": {"128k": 3, "64k": 3}, "by_motif": {"tls13_handshake_succession+obsoletes_updates_closure": 6}, "by_split": {"train": 6}, "candidate_generation_retention": 1.0, "candidate_quality_report_sha256": "810fa0a21659c7e60ea037520c9640977fb1b07a406e6a7b373edabc1887d843", "candidate_row_set_sha256": "ac7a533fa940f69d59e313b6ddf0a34aa54c92fe072f4484de2c7e9afc327460", "content_gate_eligible": true, "data_product": "worldlong_release_union_v1", "data_stage": "train_ready", "diagnostic_only": true, "local_probe_trust_isolation": "non_independent_local_diagnostic", "n_candidate_rows": 6, "n_candidate_worlds": 1, "n_candidates": 6, "n_clones": 0, "n_promoted": 6, "n_rows": 6, "n_world_ids": 1, "n_worlds": 1, "predecessor_gate_receipt_sha256": null, "predecessor_profile_id": null, "predecessor_report_sha256": null, "production_eligible": false, "promoted_domain_world_quotas": {"standards": 1}, "promoted_row_set_sha256": "cb4990de2b7b129cc7c0455e4c94576ffaa5b59759150be07bad2f75c5f6aeea", "promoted_split_row_set_sha256": {"eval": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", "train": "cb4990de2b7b129cc7c0455e4c94576ffaa5b59759150be07bad2f75c5f6aeea"}, "promotion_retention": 1.0, "real_worlds_by_split": {"eval": 0, "train": 1}, "release_profile_id": "p57-ietf-tls13-64k-128k-extension-probe-1-v1", "release_profile_sha256": "e6498c7886a440afd68272229deb5d5e02643115abc831bb58605196a637a7bc", "release_selection_sha256": "46c412e55a5e391ac46af172f4a11623a01c552075d5ba2d80f6604f48b5f89d", "report_binding_revision": "longworld-quality-binding-v4", "retention": 1.0, "schema_version": "longworld-release-train-ready-report-v1", "split_worlds_by_domain": {"eval": {"standards": []}, "train": {"standards": ["ietf-tls13-handshake-succession-v4"]}}, "target_promoted_worlds": 1, "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "tokenizer_model_id": "Qwen/Qwen3.5-4B", "tokenizer_revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "trust_scope": "local_probe", "trust_valid_for_production": false, "worlds_by_domain": {"standards": 1}} diff --git a/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/release_gate_receipt.json b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/release_gate_receipt.json new file mode 100644 index 0000000000000000000000000000000000000000..c9ca39bea21caae420c3f1d9c18b20b923396255 --- /dev/null +++ b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/release_gate_receipt.json @@ -0,0 +1 @@ +{"attestation": {"digest": "ac9e4e5cc92dd944e98b6ea311377ed339b6e3eb3e8d456d1fc12620134eb415", "environment": "probe", "key_id": "probe-auditor-039d83adbb5136b28aac3f3590b47b41", "purpose": "release_gate_pass", "role": "auditor", "scheme": "hmac-sha256-v2"}, "content_gate_eligible": true, "diagnostic_only": true, "errors": [], "gate_revision": "longworld-quality-gate-v6", "local_probe_trust_isolation": "non_independent_local_diagnostic", "metrics_sha256": "b454b0052c9d3ff35f94826a854b5eb720b8e087935c504a3acd5430073948a6", "n_rows": 6, "n_worlds": 1, "ok": true, "predecessor_profile_id": null, "production_approval": null, "production_eligible": false, "promoted_row_set_sha256": "cb4990de2b7b129cc7c0455e4c94576ffaa5b59759150be07bad2f75c5f6aeea", "promoted_split_row_set_sha256": {"eval": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", "train": "cb4990de2b7b129cc7c0455e4c94576ffaa5b59759150be07bad2f75c5f6aeea"}, "quality_report_sha256": "08b37d29cce4bd2c09c7864cbe56e4f14dbec8beb599be1dd365b5f4641cb98a", "release_profile_id": "p57-ietf-tls13-64k-128k-extension-probe-1-v1", "release_profile_sha256": "e6498c7886a440afd68272229deb5d5e02643115abc831bb58605196a637a7bc", "schema_version": "longworld-release-gate-pass-v1", "source_file_sha256": {"eval.jsonl": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", "quality_report.json": "08b37d29cce4bd2c09c7864cbe56e4f14dbec8beb599be1dd365b5f4641cb98a", "train.jsonl": "01ef1c0f38aa021a47017902146a908f5cd62d729a8f0a7640581ed4538cccf4"}, "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "tokenizer_model_id": "Qwen/Qwen3.5-4B", "tokenizer_revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "trust_scope": "local_probe", "trust_valid_for_production": false} diff --git a/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/release_selection.json b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/release_selection.json new file mode 100644 index 0000000000000000000000000000000000000000..71ac53186dd32b97bbc0c06b4fc12cf21df699d8 --- /dev/null +++ b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/release_selection.json @@ -0,0 +1,83 @@ +{ + "attestation": { + "digest": "5f0aa117c74a77d9cc02c2be52636b0d2fa65f09edd669b331d4e0bb79ae5c1c", + "environment": "probe", + "key_id": "probe-auditor-039d83adbb5136b28aac3f3590b47b41", + "purpose": "release_world_selection", + "role": "auditor", + "scheme": "hmac-sha256-v2" + }, + "audit_row_set_sha256": "b59187879ba32ed40ef7743ccfd46bc227297e84b55cbec9c3276a0b0e82351c", + "audit_sha256_by_candidate": { + "6f2957d4a31198559985b79b011d9558f53a691eb1b4d276ff292762c80fb06f": "4a828a66915b2c1c09806c53c22e3083ef9b4af5134fc9c2dff1fe4e79b96afd", + "893fff21771fb6e8114c8cdf1c3232eaa30553e0e56ecf8e999b6732d17e53ea": "7ca44fd6571878614de11ab969d206ee348e5c25ea50b89e4309341b2aceaaa7", + "b5f0e0ba4612029242d9b8451f7f0560b57a323f65336eba2281eba4efbc8c21": "fd88cf67f1f03db08590a7ed02c981c764dbcaecd171bb07f7a8ec23b9e6e037", + "b9ffa84d0b0c4f27ec539da80a163bd8f89346a6400a51d7652440c323dd61d6": "9b72d2270d99637f687703da54d985c0b10cf9755010352b634a7eb0bfb342c9", + "c338b94c9a4a7398e8a3393f0a232154d2ca62be348c4dd6055ac4c81cdc35d4": "0cfbc91eb32dcfcf8b1bbb0bbc494135b9f491fa4533f783c1948fe3476b6615", + "da2ef78789b9102dd483f37fb474731f5273c926c417426c8c60b0232a0bcca2": "139ff6ba954658642a32a439476993587d0b670031ba644a6458111e17145f80" + }, + "candidate_row_set_sha256": "ac7a533fa940f69d59e313b6ddf0a34aa54c92fe072f4484de2c7e9afc327460", + "local_probe_trust_isolation": "non_independent_local_diagnostic", + "n_candidate_worlds": 1, + "n_eligible_worlds": 1, + "n_real_eligible_worlds": 1, + "n_selected_worlds": 1, + "predecessor_gate_receipt_sha256": null, + "predecessor_profile_id": null, + "predecessor_report_sha256": null, + "promoted_domain_world_quotas": { + "standards": 1 + }, + "real_worlds_by_split": { + "eval": [], + "train": [ + "ietf-tls13-handshake-succession-v4" + ] + }, + "release_profile_id": "p57-ietf-tls13-64k-128k-extension-probe-1-v1", + "release_profile_sha256": "e6498c7886a440afd68272229deb5d5e02643115abc831bb58605196a637a7bc", + "schema_version": "longworld-release-world-selection-v2", + "selected_candidate_sha256": [ + "6f2957d4a31198559985b79b011d9558f53a691eb1b4d276ff292762c80fb06f", + "893fff21771fb6e8114c8cdf1c3232eaa30553e0e56ecf8e999b6732d17e53ea", + "b5f0e0ba4612029242d9b8451f7f0560b57a323f65336eba2281eba4efbc8c21", + "b9ffa84d0b0c4f27ec539da80a163bd8f89346a6400a51d7652440c323dd61d6", + "c338b94c9a4a7398e8a3393f0a232154d2ca62be348c4dd6055ac4c81cdc35d4", + "da2ef78789b9102dd483f37fb474731f5273c926c417426c8c60b0232a0bcca2" + ], + "selected_worlds_by_domain": { + "standards": [ + "ietf-tls13-handshake-succession-v4" + ] + }, + "split_by_world": { + "ietf-tls13-handshake-succession-v4": "train" + }, + "split_strategy": "world_atomic_hash_v1", + "split_worlds_by_domain": { + "eval": { + "standards": [] + }, + "train": { + "standards": [ + "ietf-tls13-handshake-succession-v4" + ] + } + }, + "task_semantic_commitment_sha256_by_candidate": { + "6f2957d4a31198559985b79b011d9558f53a691eb1b4d276ff292762c80fb06f": "d0ba9fce5d7a14d72aa724fb05a7f9c6c81ffcf19090647fc0fca5aa6cfe21a5", + "893fff21771fb6e8114c8cdf1c3232eaa30553e0e56ecf8e999b6732d17e53ea": "6aa677d5563f05cc7d13c073a5366b4565cc6393e73aa135963a92717777b09d", + "b5f0e0ba4612029242d9b8451f7f0560b57a323f65336eba2281eba4efbc8c21": "dfad22d5f03f0a1bb543ca914cd6b7e0461f5629cb0c87f3c05816c3befde9bd", + "b9ffa84d0b0c4f27ec539da80a163bd8f89346a6400a51d7652440c323dd61d6": "2a663ac9d67af388be26c067991445dbb2eacca593c56829c9ca45ed70649209", + "c338b94c9a4a7398e8a3393f0a232154d2ca62be348c4dd6055ac4c81cdc35d4": "86095c2bf0d6795f5d1c7616e3508f5ba1cb48f76084cfa84f3e369a5fe03b7f", + "da2ef78789b9102dd483f37fb474731f5273c926c417426c8c60b0232a0bcca2": "1b81a2b66c899825befe3f3db3755d142bedf21f47f3f47d9d64caba4f50b400" + }, + "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", + "tokenizer_model_id": "Qwen/Qwen3.5-4B", + "tokenizer_revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", + "worlds_by_domain": { + "standards": [ + "ietf-tls13-handshake-succession-v4" + ] + } +} diff --git a/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/release_selection.jsonl b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/release_selection.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..f9d277cd17302bfedcf6038dee71cd14bbf3c714 --- /dev/null +++ b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/release_selection.jsonl @@ -0,0 +1 @@ +{"attestation": {"digest": "5f0aa117c74a77d9cc02c2be52636b0d2fa65f09edd669b331d4e0bb79ae5c1c", "environment": "probe", "key_id": "probe-auditor-039d83adbb5136b28aac3f3590b47b41", "purpose": "release_world_selection", "role": "auditor", "scheme": "hmac-sha256-v2"}, "audit_row_set_sha256": "b59187879ba32ed40ef7743ccfd46bc227297e84b55cbec9c3276a0b0e82351c", "audit_sha256_by_candidate": {"6f2957d4a31198559985b79b011d9558f53a691eb1b4d276ff292762c80fb06f": "4a828a66915b2c1c09806c53c22e3083ef9b4af5134fc9c2dff1fe4e79b96afd", "893fff21771fb6e8114c8cdf1c3232eaa30553e0e56ecf8e999b6732d17e53ea": "7ca44fd6571878614de11ab969d206ee348e5c25ea50b89e4309341b2aceaaa7", "b5f0e0ba4612029242d9b8451f7f0560b57a323f65336eba2281eba4efbc8c21": "fd88cf67f1f03db08590a7ed02c981c764dbcaecd171bb07f7a8ec23b9e6e037", "b9ffa84d0b0c4f27ec539da80a163bd8f89346a6400a51d7652440c323dd61d6": "9b72d2270d99637f687703da54d985c0b10cf9755010352b634a7eb0bfb342c9", "c338b94c9a4a7398e8a3393f0a232154d2ca62be348c4dd6055ac4c81cdc35d4": "0cfbc91eb32dcfcf8b1bbb0bbc494135b9f491fa4533f783c1948fe3476b6615", "da2ef78789b9102dd483f37fb474731f5273c926c417426c8c60b0232a0bcca2": "139ff6ba954658642a32a439476993587d0b670031ba644a6458111e17145f80"}, "candidate_row_set_sha256": "ac7a533fa940f69d59e313b6ddf0a34aa54c92fe072f4484de2c7e9afc327460", "local_probe_trust_isolation": "non_independent_local_diagnostic", "n_candidate_worlds": 1, "n_eligible_worlds": 1, "n_real_eligible_worlds": 1, "n_selected_worlds": 1, "predecessor_gate_receipt_sha256": null, "predecessor_profile_id": null, "predecessor_report_sha256": null, "promoted_domain_world_quotas": {"standards": 1}, "real_worlds_by_split": {"eval": [], "train": ["ietf-tls13-handshake-succession-v4"]}, "release_profile_id": "p57-ietf-tls13-64k-128k-extension-probe-1-v1", "release_profile_sha256": "e6498c7886a440afd68272229deb5d5e02643115abc831bb58605196a637a7bc", "schema_version": "longworld-release-world-selection-v2", "selected_candidate_sha256": ["6f2957d4a31198559985b79b011d9558f53a691eb1b4d276ff292762c80fb06f", "893fff21771fb6e8114c8cdf1c3232eaa30553e0e56ecf8e999b6732d17e53ea", "b5f0e0ba4612029242d9b8451f7f0560b57a323f65336eba2281eba4efbc8c21", "b9ffa84d0b0c4f27ec539da80a163bd8f89346a6400a51d7652440c323dd61d6", "c338b94c9a4a7398e8a3393f0a232154d2ca62be348c4dd6055ac4c81cdc35d4", "da2ef78789b9102dd483f37fb474731f5273c926c417426c8c60b0232a0bcca2"], "selected_worlds_by_domain": {"standards": ["ietf-tls13-handshake-succession-v4"]}, "split_by_world": {"ietf-tls13-handshake-succession-v4": "train"}, "split_strategy": "world_atomic_hash_v1", "split_worlds_by_domain": {"eval": {"standards": []}, "train": {"standards": ["ietf-tls13-handshake-succession-v4"]}}, "task_semantic_commitment_sha256_by_candidate": {"6f2957d4a31198559985b79b011d9558f53a691eb1b4d276ff292762c80fb06f": "d0ba9fce5d7a14d72aa724fb05a7f9c6c81ffcf19090647fc0fca5aa6cfe21a5", "893fff21771fb6e8114c8cdf1c3232eaa30553e0e56ecf8e999b6732d17e53ea": "6aa677d5563f05cc7d13c073a5366b4565cc6393e73aa135963a92717777b09d", "b5f0e0ba4612029242d9b8451f7f0560b57a323f65336eba2281eba4efbc8c21": "dfad22d5f03f0a1bb543ca914cd6b7e0461f5629cb0c87f3c05816c3befde9bd", "b9ffa84d0b0c4f27ec539da80a163bd8f89346a6400a51d7652440c323dd61d6": "2a663ac9d67af388be26c067991445dbb2eacca593c56829c9ca45ed70649209", "c338b94c9a4a7398e8a3393f0a232154d2ca62be348c4dd6055ac4c81cdc35d4": "86095c2bf0d6795f5d1c7616e3508f5ba1cb48f76084cfa84f3e369a5fe03b7f", "da2ef78789b9102dd483f37fb474731f5273c926c417426c8c60b0232a0bcca2": "1b81a2b66c899825befe3f3db3755d142bedf21f47f3f47d9d64caba4f50b400"}, "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "tokenizer_model_id": "Qwen/Qwen3.5-4B", "tokenizer_revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "worlds_by_domain": {"standards": ["ietf-tls13-handshake-succession-v4"]}} diff --git a/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/train.jsonl b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/train.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..59b8b060e947f8c6c02805db3073abfd95bfa258 --- /dev/null +++ b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/train.jsonl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:01ef1c0f38aa021a47017902146a908f5cd62d729a8f0a7640581ed4538cccf4 +size 12404181 diff --git a/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/train_audits.jsonl b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/train_audits.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..b489aa4337e1fdf59977d7498b71ff488671246e --- /dev/null +++ b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/train_audits.jsonl @@ -0,0 +1,6 @@ +{"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_audit_sha256": "568522427dfdd3adcc062ce5e6c951910a4666e4e0dae230b0dd9626ff00d1e1", "attestation": {"digest": "6337f81da0edb89a09861e3ca6e0a773fbac0fc8b6a3f948c8104e94183cea9b", "environment": "probe", "key_id": "probe-auditor-039d83adbb5136b28aac3f3590b47b41", "purpose": "dense_retrieval_audit", "role": "auditor", "scheme": "hmac-sha256-v2"}, "candidate_sha256": "c338b94c9a4a7398e8a3393f0a232154d2ca62be348c4dd6055ac4c81cdc35d4", "counterfactual_replay_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "embedding_topk_insufficient": true, "expected_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "full_pool_strict_replay_sufficient": true, "k": 3, "local_probe_trust_isolation": "non_independent_local_diagnostic", "model": {"backend": "sentence-transformers-6.0.0", "chunking": {"aggregation": "max_similarity", "max_tokens": 192, "overlap_tokens": 32, "strategy": "tokenizer_token_windows"}, "model_id": "sentence-transformers/all-MiniLM-L6-v2", "provider": "huggingface", "revision": "1110a243fdf4706b3f48f1d95db1a4f5529b4d41", "score_metric": "dot_product"}, "near_dup_sentence_ratio": 0.047, "query_id": "ietf-tls13-handshake-succession-v4:128k:standard:cf", "ranker_type": "dense_embedding", "ranking_sha256": "affa42313ef99ec1dd55d8f2c3343476429fcaf274de1ab3778118e11edaa670", "schema_version": "train-ready-promotion-v2", "source_binding_sha256": "fc34a690fcb8ffffb64df92f1df724ea75d43ce88c10daf432fb085fee09ddbe", "strict_growth_metrics": {"authentic_source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}], "graph": {"hop_count": 2, "n_essential_artifacts": 7, "n_essential_events": 6, "proof_depth": 2}, "semantic_growth_group_id": "ietf-tls13-handshake-succession-v4|ietf-tls13-handshake-succession", "semantic_tokens": {"causal_supporting": 76236, "event_bearing": 51745, "generic_background": 0, "internal": 127986, "measurement_basis": "exact_pinned_tokenizer_replay_relation_path_whole_artifact_upper_bound", "proof_bearing": 51745}, "strict_support_event_count": 5}, "strict_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "strict_replay_prefix_answers": ["{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}"], "strict_replay_revision": "longworld-strict-replay-v5", "task_proof": {"task_proof_receipt": {"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "artifact_aligned_windows": {"16k": {"enumeration_sha256": "78be99a9cdfb0a90424c602706306c9481c3431a23dda1076a0cb029b5b8ae86", "max_artifact_count": 18, "max_span_tokens": 16358, "status": "insufficient", "strict_window_count": 197, "window_tokens": 16384}, "4k": {"enumeration_sha256": "892b843b9fa2b58c8b78b9c70ae97f16003fb18fe35f47269ef60599a90d8cee", "max_artifact_count": 7, "max_span_tokens": 4093, "status": "insufficient", "strict_window_count": 83, "window_tokens": 4096}, "8k": {"enumeration_sha256": "31ef4c15de8c532a7e19685e7383e5a6dc34cf1dce4aa0e80ac9494964d2e47f", "max_artifact_count": 13, "max_span_tokens": 8190, "status": "insufficient", "strict_window_count": 153, "window_tokens": 8192}}, "artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "end_token": 9837, "start_token": 0, "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:8446:chars:0271846-0337410:525df48ab2b8", "end_token": 25957, "start_token": 9842, "text_sha256": "525df48ab2b8a9141892cf0ea81d94e92c6c503602649613667dea066e4c73b7"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0072013:18cf3066a9a8", "end_token": 42282, "start_token": 25962, "text_sha256": "18cf3066a9a822587a9996a13ac9baa0218b710ad2b6b2d72542dd0c55cc7b0e"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76", "end_token": 58644, "start_token": 42286, "text_sha256": "3b5af3e300e9e80b4ba711c6dcc2ee07e3b0c0a6a634e080cff043fbbec595ee"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0146106:45e3925fff8a", "end_token": 74931, "start_token": 58648, "text_sha256": "45e3925fff8ae399cb3f2e0b96b6e6dcff94e81aa983435cad0e85f809644f1f"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0037551-0039075:d5814b6d659f", "end_token": 75309, "start_token": 74935, "text_sha256": "d5814b6d659f9a5240bf000853f24aecce2275048c196f4dae64f9c4b3f52946"}, {"artifact_id": "ietf:rfc:5246:chars:0146106-0211487:5bc5f9c58046", "end_token": 91670, "start_token": 75313, "text_sha256": "5bc5f9c580462aee827e715243a198ef8b8bb8e6791304831fd9b9fffca38836"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0035288-0037549:e6b945af4872", "end_token": 92077, "start_token": 91673, "text_sha256": "e6b945af4872e2e58c4d4500cf144262bc5b756325bbd1e8bec2715018c389ad"}, {"artifact_id": "ietf:rfc:5246:chars:0211487-0222287:1fec83bd80c4", "end_token": 95222, "start_token": 92081, "text_sha256": "1fec83bd80c406dea553fd8aa891d498ec5171c06d4c4e47e478f6d4df864f93"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0034186-0035264:ea4bc5bf90fb", "end_token": 95473, "start_token": 95227, "text_sha256": "ea4bc5bf90fbc9a98bd30de1b380a2bcddb55efc7991827525e8918247c32d44"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "end_token": 99530, "start_token": 95477, "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0032066-0034182:a7ee330ff65b", "end_token": 100039, "start_token": 99535, "text_sha256": "a7ee330ff65b9a4736a75b752c27dc7baf1b38d2075c41bc1ebc67c6c3cb4539"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "end_token": 112900, "start_token": 100043, "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0030088-0032062:0daf34faf915", "end_token": 113353, "start_token": 112905, "text_sha256": "0daf34faf9151b141002856e0089f037d35cbb1bd9a02e974f48f5d6daa68903"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "end_token": 118538, "start_token": 113357, "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0028840-0030086:6dbd252139f9", "end_token": 118864, "start_token": 118543, "text_sha256": "6dbd252139f95ab7ae310ead187349e86670800a92937e56c35503f671d4ed88"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "end_token": 118971, "start_token": 118868, "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0027405-0028838:bc5a97345326", "end_token": 119333, "start_token": 118975, "text_sha256": "bc5a973453261596484cbb35ac076d116307b38107867b20a4fe7cb0fa890a80"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000342-0001925:6f2334143226", "end_token": 119762, "start_token": 119336, "text_sha256": "6f2334143226841033b156c3219652b33427cba2d05d41529eb244fbc8849722"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0025909-0027403:23af0821b330", "end_token": 120137, "start_token": 119766, "text_sha256": "23af0821b3308b31b17edc94216c7df796330fecc4cd29a7998cd2237e8f3447"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0001927-0005175:6149256b2cfa", "end_token": 121379, "start_token": 120141, "text_sha256": "6149256b2cfa7496a4ec85e7f1258f2616c027927e56f292eff509a7db2de23b"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0024557-0025907:a7c1fb548c43", "end_token": 121751, "start_token": 121383, "text_sha256": "a7c1fb548c43c43dcf22c0d1b4cd542818793145466953b34d3d54e813786c0d"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0005177-0008832:99b199c9a269", "end_token": 123429, "start_token": 121755, "text_sha256": "99b199c9a269cb43809af4fe953e95bee61051e0a97386dd7f30f1ac346d3c73"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0023096-0024555:30941310b1c9", "end_token": 123799, "start_token": 123432, "text_sha256": "30941310b1c94bf116d1afdde08636f3f47a304ade1d9842866a1b444bb86652"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0008834-0012489:cc329629d8b2", "end_token": 125381, "start_token": 123803, "text_sha256": "cc329629d8b261ba8b3de2c325c68053d96351a1a87606b0ca24781662ded12b"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0021517-0023092:77d95940d32f", "end_token": 125777, "start_token": 125384, "text_sha256": "77d95940d32fa6a8f09c56e40f3cf2a6d1dc027d954bc9e5a522f47072ef39e5"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0012491-0014984:478ed25b7e97", "end_token": 126321, "start_token": 125781, "text_sha256": "478ed25b7e97a17b2766f4058786bf7659ffcfd24c9dab96ec5464541a452c98"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0019884-0021515:0506cd161eae", "end_token": 126733, "start_token": 126324, "text_sha256": "0506cd161eae4f16da9c1af675eb969006bb7c68ed6bbfbd8844408c1356b037"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0014986-0017326:cb3c7c610ea6", "end_token": 127335, "start_token": 126737, "text_sha256": "cb3c7c610ea654500d77b09fffe18f0de9133aa75f2b97ab1948d9ddbb3b4c14"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0018419-0019882:45aa4da2fa30", "end_token": 127693, "start_token": 127338, "text_sha256": "45aa4da2fa307189092173f231056a3b41f31b52c1ebf25ca4d7b29ef86506a6"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0017328-0018417:1e264a52137b", "end_token": 127986, "start_token": 127696, "text_sha256": "1e264a52137bb90f2430fc1d583fa7c20e22682f7a6afc746ec67aa945a4a1c9"}], "checks": {"answer_surface_free": true, "artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_top3_prefixes_insufficient": true, "body_and_classification_valid": true, "counterfactual_changes_answer": true, "empty_selection_insufficient": true, "exact_token_count_recomputed": true, "full_counterfactual_sufficient": true, "full_replay_sufficient": true, "lexical_tfidf_top3_prefixes_insufficient": true, "minimal_counterfactual_sufficient": true, "minimal_replay_sufficient": true, "minimum_complexity_met": true, "no_shortcut": true, "raw_token_executable_windows_insufficient": true, "remove_one_fails": true, "single_essential_insufficient": true, "source_relations_replayed": true}, "essential_artifact_ids": ["ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76", "ietf:rfc:5246:chars:0211487-0222287:1fec83bd80c4", "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "ietf:draft:draft-ietf-tls-tls13-28:chars:0017328-0018417:1e264a52137b"], "proof_input_sha256": "c338b94c9a4a7398e8a3393f0a232154d2ca62be348c4dd6055ac4c81cdc35d4", "raw_token_offset_windows": {"16k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "3017763731fcff5d9c560f256c92562e0a49cff62d3335bea836dca40ddd73c0", "max_artifact_count": 19, "monotonicity_proven": false, "semantic_state_count": 42, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "3663ef3e7a606886963aaa2eb5870e47db23a19a6c28afc1b5aa929f67ac1753", "exact_raw_executable": {"enumeration_sha256": "3663ef3e7a606886963aaa2eb5870e47db23a19a6c28afc1b5aa929f67ac1753", "exhaustive_start_count": 111603, "semantic_state_count": 38, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 38, "window_tokens": 16384}, "4k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "9dc357ecf41b6ddda2109a160501f3b2bfba013d088e716e8a782d077fd75777", "max_artifact_count": 9, "monotonicity_proven": false, "semantic_state_count": 55, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "d292f0b9a969dffd854e621cbc36a65c308331f9ffd4aa07eeffa6ffdd0776f9", "exact_raw_executable": {"enumeration_sha256": "d292f0b9a969dffd854e621cbc36a65c308331f9ffd4aa07eeffa6ffdd0776f9", "exhaustive_start_count": 123891, "semantic_state_count": 36, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 36, "window_tokens": 4096}, "8k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "c98f2effa30cc2e74a4d95b604d71b945196613c3a746eec6dd867baf5ed30cc", "max_artifact_count": 14, "monotonicity_proven": false, "semantic_state_count": 50, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "f0b6669d06e78066f82a04851e53d0e086a6040a5ebbaafe9066f2d7e9a637a0", "exact_raw_executable": {"enumeration_sha256": "f0b6669d06e78066f82a04851e53d0e086a6040a5ebbaafe9066f2d7e9a637a0", "exhaustive_start_count": 119795, "semantic_state_count": 36, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 36, "window_tokens": 8192}}, "receipt_sha256": "a79e77d24dda2d42fc5cf9c7913014ccefef0988001e5fdf8556b13a3cf4ff82", "replay": {"counterfactual_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "empty_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "full_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "minimal_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "remove_one": [{"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5246:chars:0211487-0222287:1fec83bd80c4"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0017328-0018417:1e264a52137b"}], "single_essential": [{"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5246:chars:0211487-0222287:1fec83bd80c4"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0017328-0018417:1e264a52137b"}]}, "retrieval": {"bm25": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76", "score": "168.90947504619558"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "score": "128.12174529307543"}, {"artifact_id": "ietf:rfc:8446:chars:0271846-0337410:525df48ab2b8", "score": "125.97385615168373"}]}, "lexical_tfidf": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "score": "0.43146207004410869"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0014986-0017326:cb3c7c610ea6", "score": "0.36921192709526263"}, {"artifact_id": "ietf:rfc:8446:chars:0271846-0337410:525df48ab2b8", "score": "0.32585473443537483"}]}}, "schema_version": "longworld.task-proof-receipt.v5", "sidecar_schema_version": "longworld.task-replay-sidecar.v3", "task_replay_sidecar_sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "tokenizer": {"asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "document_context_tokens": 127986, "model_id": "Qwen/Qwen3.5-4B", "revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "serialized_context_tokens": 128364}, "window_scope": "exact_raw_slice_replay_with_separate_intersection_upper_bound"}, "verification": {"artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_topk_insufficient": true, "candidate_mode": true, "closed_book_unsolved": true, "contiguous_windows_insufficient": true, "counterfactual_changes_answer": true, "counterfactual_replay_sufficient": true, "distractor_invariance_gold": true, "embedding_topk_insufficient": false, "essential_single_doc_insufficient": true, "essential_surface_gold_free": true, "essential_text_grounded": true, "full_sufficient": true, "lexical_tfidf_topk_insufficient": true, "local_window_insufficient": true, "min_complexity": true, "minimal_sufficient": true, "no_shortcut": true, "production_mode": false, "question_only_unsolved": true, "remove_one_fails": true, "schema_ok": true, "semantic_sufficient": true, "strict_executable_sufficient": true, "surface_match": true}, "view_verification": {"classification_ok": true, "essential_present": true, "expected_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "global_proof_green": true, "production_eligible": false, "semantic_text_grounded": true, "strict_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}"}}, "task_proof_sha256": "72c22aaeed8e822a95d9e3c57d56c821c891644227a0bb492647d01b38a430c8", "task_quality_metadata": {"answer_program_id": "ietf.tls13_handshake_succession.v1", "authentic_source_relation_id": "e0dcb1ad1463850449f0", "base_task_id": "776021e7d8e7f94c336a", "context_source_relation_count": 22, "domain": "standards", "executable_proof_id": "0c64de0f18afd43b8e63", "hybrid_causal_edges": [], "length_bucket": "128k", "motif": "tls13_handshake_succession+obsoletes_updates_closure", "real_source_family_ids": ["ietf_standards"], "real_source_token_ratio": 0.8696207659468387, "real_source_verified": true, "real_source_workflow_ids": ["task-source:73e262c5a2939b145819"], "semantic_base_task_id": "250a7cfbf1fb2c0a24b9", "source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2018-03-01T00:00:00Z", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_relation_kind": "published_as"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "updates"}], "source_relation_id": "6ee046ba0ec12a545607", "world_id": "ietf-tls13-handshake-succession-v4"}, "task_replay_adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "task_replay_payload_sha256": "58034cb5ca73384b58d248a0221969f5c7a78a51df1c3237d0106f3f7b05ce96", "task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "sidecar_schema_version": "longworld.task-replay-sidecar.v3"}, "task_semantic_commitment_sha256": "86095c2bf0d6795f5d1c7616e3508f5ba1cb48f76084cfa84f3e369a5fe03b7f", "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "top_k": [{"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0014986-0017326:cb3c7c610ea6", "chunk_count": 4, "rank": 1, "score": 0.7148211002349854, "text_sha256": "cb3c7c610ea654500d77b09fffe18f0de9133aa75f2b97ab1948d9ddbb3b4c14"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0072013:18cf3066a9a8", "chunk_count": 108, "rank": 2, "score": 0.7057527303695679, "text_sha256": "18cf3066a9a822587a9996a13ac9baa0218b710ad2b6b2d72542dd0c55cc7b0e"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76", "chunk_count": 114, "rank": 3, "score": 0.6841272711753845, "text_sha256": "3b5af3e300e9e80b4ba711c6dcc2ee07e3b0c0a6a634e080cff043fbbec595ee"}], "verification_replay_sha256": "ecad4bd03faa55a09dc28e70c1aa51e24fa1bb5bc5b958fb8bbac62cd873bdbf"} +{"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_audit_sha256": "568522427dfdd3adcc062ce5e6c951910a4666e4e0dae230b0dd9626ff00d1e1", "attestation": {"digest": "2b1fa678a78706a04e49748121ea50dfc60c41bb67f1a58a5f572da9596ea1d4", "environment": "probe", "key_id": "probe-auditor-039d83adbb5136b28aac3f3590b47b41", "purpose": "dense_retrieval_audit", "role": "auditor", "scheme": "hmac-sha256-v2"}, "candidate_sha256": "da2ef78789b9102dd483f37fb474731f5273c926c417426c8c60b0232a0bcca2", "counterfactual_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "embedding_topk_insufficient": true, "expected_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "full_pool_strict_replay_sufficient": true, "k": 3, "local_probe_trust_isolation": "non_independent_local_diagnostic", "model": {"backend": "sentence-transformers-6.0.0", "chunking": {"aggregation": "max_similarity", "max_tokens": 192, "overlap_tokens": 32, "strategy": "tokenizer_token_windows"}, "model_id": "sentence-transformers/all-MiniLM-L6-v2", "provider": "huggingface", "revision": "1110a243fdf4706b3f48f1d95db1a4f5529b4d41", "score_metric": "dot_product"}, "near_dup_sentence_ratio": 0.0473, "query_id": "ietf-tls13-handshake-succession-v4:128k:standard:full", "ranker_type": "dense_embedding", "ranking_sha256": "f1217a87b7652195dd5c296882bc595bae14e2425e24dba07eb35520df4f8524", "schema_version": "train-ready-promotion-v2", "source_binding_sha256": "fc34a690fcb8ffffb64df92f1df724ea75d43ce88c10daf432fb085fee09ddbe", "strict_growth_metrics": {"authentic_source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}], "graph": {"hop_count": 2, "n_essential_artifacts": 7, "n_essential_events": 6, "proof_depth": 2}, "semantic_growth_group_id": "ietf-tls13-handshake-succession-v4|ietf-tls13-handshake-succession", "semantic_tokens": {"causal_supporting": 76236, "event_bearing": 51765, "generic_background": 0, "internal": 128006, "measurement_basis": "exact_pinned_tokenizer_replay_relation_path_whole_artifact_upper_bound", "proof_bearing": 51765}, "strict_support_event_count": 6}, "strict_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "strict_replay_prefix_answers": ["{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}"], "strict_replay_revision": "longworld-strict-replay-v5", "task_proof": {"task_proof_receipt": {"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "artifact_aligned_windows": {"16k": {"enumeration_sha256": "4fde8ae65a2dc16c8af9abf7131ec2ac616a2737162b3b94ed470fd41962068c", "max_artifact_count": 18, "max_span_tokens": 16378, "status": "insufficient", "strict_window_count": 197, "window_tokens": 16384}, "4k": {"enumeration_sha256": "892b843b9fa2b58c8b78b9c70ae97f16003fb18fe35f47269ef60599a90d8cee", "max_artifact_count": 7, "max_span_tokens": 4093, "status": "insufficient", "strict_window_count": 83, "window_tokens": 4096}, "8k": {"enumeration_sha256": "31ef4c15de8c532a7e19685e7383e5a6dc34cf1dce4aa0e80ac9494964d2e47f", "max_artifact_count": 13, "max_span_tokens": 8190, "status": "insufficient", "strict_window_count": 153, "window_tokens": 8192}}, "artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "end_token": 9837, "start_token": 0, "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:8446:chars:0271846-0337410:525df48ab2b8", "end_token": 25957, "start_token": 9842, "text_sha256": "525df48ab2b8a9141892cf0ea81d94e92c6c503602649613667dea066e4c73b7"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0072013:18cf3066a9a8", "end_token": 42282, "start_token": 25962, "text_sha256": "18cf3066a9a822587a9996a13ac9baa0218b710ad2b6b2d72542dd0c55cc7b0e"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76", "end_token": 58664, "start_token": 42286, "text_sha256": "ebd38261da7659280f75a3d6a07e45c4c554ffd74c2b622be1b2d10a49db7fe4"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0146106:45e3925fff8a", "end_token": 74951, "start_token": 58668, "text_sha256": "45e3925fff8ae399cb3f2e0b96b6e6dcff94e81aa983435cad0e85f809644f1f"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0037551-0039075:d5814b6d659f", "end_token": 75329, "start_token": 74955, "text_sha256": "d5814b6d659f9a5240bf000853f24aecce2275048c196f4dae64f9c4b3f52946"}, {"artifact_id": "ietf:rfc:5246:chars:0146106-0211487:5bc5f9c58046", "end_token": 91690, "start_token": 75333, "text_sha256": "5bc5f9c580462aee827e715243a198ef8b8bb8e6791304831fd9b9fffca38836"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0035288-0037549:e6b945af4872", "end_token": 92097, "start_token": 91693, "text_sha256": "e6b945af4872e2e58c4d4500cf144262bc5b756325bbd1e8bec2715018c389ad"}, {"artifact_id": "ietf:rfc:5246:chars:0211487-0222287:1fec83bd80c4", "end_token": 95242, "start_token": 92101, "text_sha256": "1fec83bd80c406dea553fd8aa891d498ec5171c06d4c4e47e478f6d4df864f93"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0034186-0035264:ea4bc5bf90fb", "end_token": 95493, "start_token": 95247, "text_sha256": "ea4bc5bf90fbc9a98bd30de1b380a2bcddb55efc7991827525e8918247c32d44"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "end_token": 99550, "start_token": 95497, "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0032066-0034182:a7ee330ff65b", "end_token": 100059, "start_token": 99555, "text_sha256": "a7ee330ff65b9a4736a75b752c27dc7baf1b38d2075c41bc1ebc67c6c3cb4539"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "end_token": 112920, "start_token": 100063, "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0030088-0032062:0daf34faf915", "end_token": 113373, "start_token": 112925, "text_sha256": "0daf34faf9151b141002856e0089f037d35cbb1bd9a02e974f48f5d6daa68903"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "end_token": 118558, "start_token": 113377, "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0028840-0030086:6dbd252139f9", "end_token": 118884, "start_token": 118563, "text_sha256": "6dbd252139f95ab7ae310ead187349e86670800a92937e56c35503f671d4ed88"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "end_token": 118991, "start_token": 118888, "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0027405-0028838:bc5a97345326", "end_token": 119353, "start_token": 118995, "text_sha256": "bc5a973453261596484cbb35ac076d116307b38107867b20a4fe7cb0fa890a80"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000342-0001925:6f2334143226", "end_token": 119782, "start_token": 119356, "text_sha256": "6f2334143226841033b156c3219652b33427cba2d05d41529eb244fbc8849722"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0025909-0027403:23af0821b330", "end_token": 120157, "start_token": 119786, "text_sha256": "23af0821b3308b31b17edc94216c7df796330fecc4cd29a7998cd2237e8f3447"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0001927-0005175:6149256b2cfa", "end_token": 121399, "start_token": 120161, "text_sha256": "6149256b2cfa7496a4ec85e7f1258f2616c027927e56f292eff509a7db2de23b"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0024557-0025907:a7c1fb548c43", "end_token": 121771, "start_token": 121403, "text_sha256": "a7c1fb548c43c43dcf22c0d1b4cd542818793145466953b34d3d54e813786c0d"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0005177-0008832:99b199c9a269", "end_token": 123449, "start_token": 121775, "text_sha256": "99b199c9a269cb43809af4fe953e95bee61051e0a97386dd7f30f1ac346d3c73"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0023096-0024555:30941310b1c9", "end_token": 123819, "start_token": 123452, "text_sha256": "30941310b1c94bf116d1afdde08636f3f47a304ade1d9842866a1b444bb86652"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0008834-0012489:cc329629d8b2", "end_token": 125401, "start_token": 123823, "text_sha256": "cc329629d8b261ba8b3de2c325c68053d96351a1a87606b0ca24781662ded12b"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0021517-0023092:77d95940d32f", "end_token": 125797, "start_token": 125404, "text_sha256": "77d95940d32fa6a8f09c56e40f3cf2a6d1dc027d954bc9e5a522f47072ef39e5"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0012491-0014984:478ed25b7e97", "end_token": 126341, "start_token": 125801, "text_sha256": "478ed25b7e97a17b2766f4058786bf7659ffcfd24c9dab96ec5464541a452c98"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0019884-0021515:0506cd161eae", "end_token": 126753, "start_token": 126344, "text_sha256": "0506cd161eae4f16da9c1af675eb969006bb7c68ed6bbfbd8844408c1356b037"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0014986-0017326:cb3c7c610ea6", "end_token": 127355, "start_token": 126757, "text_sha256": "cb3c7c610ea654500d77b09fffe18f0de9133aa75f2b97ab1948d9ddbb3b4c14"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0018419-0019882:45aa4da2fa30", "end_token": 127713, "start_token": 127358, "text_sha256": "45aa4da2fa307189092173f231056a3b41f31b52c1ebf25ca4d7b29ef86506a6"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0017328-0018417:1e264a52137b", "end_token": 128006, "start_token": 127716, "text_sha256": "1e264a52137bb90f2430fc1d583fa7c20e22682f7a6afc746ec67aa945a4a1c9"}], "checks": {"answer_surface_free": true, "artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_top3_prefixes_insufficient": true, "body_and_classification_valid": true, "counterfactual_changes_answer": true, "empty_selection_insufficient": true, "exact_token_count_recomputed": true, "full_counterfactual_sufficient": true, "full_replay_sufficient": true, "lexical_tfidf_top3_prefixes_insufficient": true, "minimal_counterfactual_sufficient": true, "minimal_replay_sufficient": true, "minimum_complexity_met": true, "no_shortcut": true, "raw_token_executable_windows_insufficient": true, "remove_one_fails": true, "single_essential_insufficient": true, "source_relations_replayed": true}, "essential_artifact_ids": ["ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76", "ietf:rfc:5246:chars:0211487-0222287:1fec83bd80c4", "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "ietf:draft:draft-ietf-tls-tls13-28:chars:0017328-0018417:1e264a52137b"], "proof_input_sha256": "da2ef78789b9102dd483f37fb474731f5273c926c417426c8c60b0232a0bcca2", "raw_token_offset_windows": {"16k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "441278967dadf3db3b16a192c0a22239bcc8faa8d0ec804ee503a987d0a1c635", "max_artifact_count": 19, "monotonicity_proven": false, "semantic_state_count": 42, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "58e7eeae8deb8205b6f5a58a2c12016d95dfaeba6d410e4e22e9fe351209487c", "exact_raw_executable": {"enumeration_sha256": "58e7eeae8deb8205b6f5a58a2c12016d95dfaeba6d410e4e22e9fe351209487c", "exhaustive_start_count": 111623, "semantic_state_count": 38, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 38, "window_tokens": 16384}, "4k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "e7c0a0d776a7df3081a93079ad09db1f82f334fb321e861b2139aef590eb306b", "max_artifact_count": 9, "monotonicity_proven": false, "semantic_state_count": 55, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "f378c023b9d17d479413fd329574fe9889580be30f92f12e1f416cc91d48f3b2", "exact_raw_executable": {"enumeration_sha256": "f378c023b9d17d479413fd329574fe9889580be30f92f12e1f416cc91d48f3b2", "exhaustive_start_count": 123911, "semantic_state_count": 36, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 36, "window_tokens": 4096}, "8k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "f47b9a176eda57f5885e342cebcc52cbca495eb8f6f0dc6ca20afe78225730aa", "max_artifact_count": 14, "monotonicity_proven": false, "semantic_state_count": 50, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "a330bc16e01cb24a866bb26240b2504f95e3cc32811913dab97c863bf6b1e8ad", "exact_raw_executable": {"enumeration_sha256": "a330bc16e01cb24a866bb26240b2504f95e3cc32811913dab97c863bf6b1e8ad", "exhaustive_start_count": 119815, "semantic_state_count": 36, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 36, "window_tokens": 8192}}, "receipt_sha256": "489395d4891b8af427edf29daea425157f53535e8dc9e924eaa21c49493e4fbc", "replay": {"counterfactual_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "empty_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "full_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "minimal_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "remove_one": [{"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5246:chars:0211487-0222287:1fec83bd80c4"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0017328-0018417:1e264a52137b"}], "single_essential": [{"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5246:chars:0211487-0222287:1fec83bd80c4"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0017328-0018417:1e264a52137b"}]}, "retrieval": {"bm25": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76", "score": "168.86507971976266"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "score": "128.12877489877692"}, {"artifact_id": "ietf:rfc:8446:chars:0271846-0337410:525df48ab2b8", "score": "125.97876729711818"}]}, "lexical_tfidf": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "score": "0.43146207004410869"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0014986-0017326:cb3c7c610ea6", "score": "0.36921192709526263"}, {"artifact_id": "ietf:rfc:8446:chars:0271846-0337410:525df48ab2b8", "score": "0.32585473443537483"}]}}, "schema_version": "longworld.task-proof-receipt.v5", "sidecar_schema_version": "longworld.task-replay-sidecar.v3", "task_replay_sidecar_sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "tokenizer": {"asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "document_context_tokens": 128006, "model_id": "Qwen/Qwen3.5-4B", "revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "serialized_context_tokens": 128384}, "window_scope": "exact_raw_slice_replay_with_separate_intersection_upper_bound"}, "verification": {"artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_topk_insufficient": true, "candidate_mode": true, "closed_book_unsolved": true, "contiguous_windows_insufficient": true, "counterfactual_changes_answer": true, "counterfactual_replay_sufficient": true, "distractor_invariance_gold": true, "embedding_topk_insufficient": false, "essential_single_doc_insufficient": true, "essential_surface_gold_free": true, "essential_text_grounded": true, "full_sufficient": true, "lexical_tfidf_topk_insufficient": true, "local_window_insufficient": true, "min_complexity": true, "minimal_sufficient": true, "no_shortcut": true, "production_mode": false, "question_only_unsolved": true, "remove_one_fails": true, "schema_ok": true, "semantic_sufficient": true, "strict_executable_sufficient": true, "surface_match": true}, "view_verification": {"classification_ok": true, "essential_present": true, "expected_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "global_proof_green": true, "production_eligible": false, "semantic_text_grounded": true, "strict_replay_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}"}}, "task_proof_sha256": "086ffce8293530577bcdddf9ce46f8ad036a95300b153208af252f9ee0931379", "task_quality_metadata": {"answer_program_id": "ietf.tls13_handshake_succession.v1", "authentic_source_relation_id": "e0dcb1ad1463850449f0", "base_task_id": "776021e7d8e7f94c336a", "context_source_relation_count": 22, "domain": "standards", "executable_proof_id": "0c64de0f18afd43b8e63", "hybrid_causal_edges": [], "length_bucket": "128k", "motif": "tls13_handshake_succession+obsoletes_updates_closure", "real_source_family_ids": ["ietf_standards"], "real_source_token_ratio": 0.9970557078763709, "real_source_verified": true, "real_source_workflow_ids": ["task-source:73e262c5a2939b145819"], "semantic_base_task_id": "250a7cfbf1fb2c0a24b9", "source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2018-03-01T00:00:00Z", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_relation_kind": "published_as"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "updates"}], "source_relation_id": "6ee046ba0ec12a545607", "world_id": "ietf-tls13-handshake-succession-v4"}, "task_replay_adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "task_replay_payload_sha256": "58034cb5ca73384b58d248a0221969f5c7a78a51df1c3237d0106f3f7b05ce96", "task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "sidecar_schema_version": "longworld.task-replay-sidecar.v3"}, "task_semantic_commitment_sha256": "1b81a2b66c899825befe3f3db3755d142bedf21f47f3f47d9d64caba4f50b400", "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "top_k": [{"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0014986-0017326:cb3c7c610ea6", "chunk_count": 4, "rank": 1, "score": 0.7148211002349854, "text_sha256": "cb3c7c610ea654500d77b09fffe18f0de9133aa75f2b97ab1948d9ddbb3b4c14"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0072013:18cf3066a9a8", "chunk_count": 108, "rank": 2, "score": 0.7057527303695679, "text_sha256": "18cf3066a9a822587a9996a13ac9baa0218b710ad2b6b2d72542dd0c55cc7b0e"}, {"artifact_id": "ietf:rfc:5246:chars:0146106-0211487:5bc5f9c58046", "chunk_count": 92, "rank": 3, "score": 0.6802471280097961, "text_sha256": "5bc5f9c580462aee827e715243a198ef8b8bb8e6791304831fd9b9fffca38836"}], "verification_replay_sha256": "47ad1c09b074ca6e46a856dd22b1f67a6627e1170b3b768b621ddd2139837aa1"} +{"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_audit_sha256": "568522427dfdd3adcc062ce5e6c951910a4666e4e0dae230b0dd9626ff00d1e1", "attestation": {"digest": "386c648d9d550825e2ce0f4a70e458217f05f79c1fd014e5beae085d18b5062a", "environment": "probe", "key_id": "probe-auditor-039d83adbb5136b28aac3f3590b47b41", "purpose": "dense_retrieval_audit", "role": "auditor", "scheme": "hmac-sha256-v2"}, "candidate_sha256": "893fff21771fb6e8114c8cdf1c3232eaa30553e0e56ecf8e999b6732d17e53ea", "counterfactual_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "embedding_topk_insufficient": true, "expected_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "full_pool_strict_replay_sufficient": true, "k": 3, "local_probe_trust_isolation": "non_independent_local_diagnostic", "model": {"backend": "sentence-transformers-6.0.0", "chunking": {"aggregation": "max_similarity", "max_tokens": 192, "overlap_tokens": 32, "strategy": "tokenizer_token_windows"}, "model_id": "sentence-transformers/all-MiniLM-L6-v2", "provider": "huggingface", "revision": "1110a243fdf4706b3f48f1d95db1a4f5529b4d41", "score_metric": "dot_product"}, "near_dup_sentence_ratio": 0.0473, "query_id": "ietf-tls13-handshake-succession-v4:128k:standard:ordered_artifact_view", "ranker_type": "dense_embedding", "ranking_sha256": "1c2e2a73cc6f56ea401a988b52ec4b562c1bb9c8103c31aaa52acabb1f9d800e", "schema_version": "train-ready-promotion-v2", "source_binding_sha256": "fc34a690fcb8ffffb64df92f1df724ea75d43ce88c10daf432fb085fee09ddbe", "strict_growth_metrics": {"authentic_source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}], "graph": {"hop_count": 2, "n_essential_artifacts": 7, "n_essential_events": 6, "proof_depth": 2}, "semantic_growth_group_id": "ietf-tls13-handshake-succession-v4|ietf-tls13-handshake-succession", "semantic_tokens": {"causal_supporting": 76151, "event_bearing": 51850, "generic_background": 0, "internal": 128005, "measurement_basis": "exact_pinned_tokenizer_replay_relation_path_whole_artifact_upper_bound", "proof_bearing": 51850}, "strict_support_event_count": 6}, "strict_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "strict_replay_prefix_answers": ["{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}"], "strict_replay_revision": "longworld-strict-replay-v5", "task_proof": {"task_proof_receipt": {"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "artifact_aligned_windows": {"16k": {"enumeration_sha256": "f8e9b1dfd39cec8b3321133302c07d9528f013912ec37e58ccf4784b6e6fb773", "max_artifact_count": 21, "max_span_tokens": 16378, "status": "insufficient", "strict_window_count": 262, "window_tokens": 16384}, "4k": {"enumeration_sha256": "36534e3237c390e28a906379b93a87112dcbb00e633f95fe5bbd9dbd330f35c7", "max_artifact_count": 10, "max_span_tokens": 4080, "status": "insufficient", "strict_window_count": 135, "window_tokens": 4096}, "8k": {"enumeration_sha256": "639fc30080db27b34c662ba82b6881c3f538a84942f7313609cff46e6aeb0f2c", "max_artifact_count": 17, "max_span_tokens": 8117, "status": "insufficient", "strict_window_count": 208, "window_tokens": 8192}}, "artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "end_token": 9837, "start_token": 0, "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0072013:18cf3066a9a8", "end_token": 26162, "start_token": 9842, "text_sha256": "18cf3066a9a822587a9996a13ac9baa0218b710ad2b6b2d72542dd0c55cc7b0e"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0146106:45e3925fff8a", "end_token": 42449, "start_token": 26166, "text_sha256": "45e3925fff8ae399cb3f2e0b96b6e6dcff94e81aa983435cad0e85f809644f1f"}, {"artifact_id": "ietf:rfc:5246:chars:0146106-0211487:5bc5f9c58046", "end_token": 58810, "start_token": 42453, "text_sha256": "5bc5f9c580462aee827e715243a198ef8b8bb8e6791304831fd9b9fffca38836"}, {"artifact_id": "ietf:rfc:5246:chars:0211487-0222287:1fec83bd80c4", "end_token": 61955, "start_token": 58814, "text_sha256": "1fec83bd80c406dea553fd8aa891d498ec5171c06d4c4e47e478f6d4df864f93"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "end_token": 66013, "start_token": 61960, "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "end_token": 78875, "start_token": 66018, "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "end_token": 84061, "start_token": 78880, "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "end_token": 84169, "start_token": 84066, "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000342-0001925:6f2334143226", "end_token": 84598, "start_token": 84172, "text_sha256": "6f2334143226841033b156c3219652b33427cba2d05d41529eb244fbc8849722"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0001927-0005175:6149256b2cfa", "end_token": 85840, "start_token": 84602, "text_sha256": "6149256b2cfa7496a4ec85e7f1258f2616c027927e56f292eff509a7db2de23b"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0005177-0008832:99b199c9a269", "end_token": 87518, "start_token": 85844, "text_sha256": "99b199c9a269cb43809af4fe953e95bee61051e0a97386dd7f30f1ac346d3c73"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0008834-0012489:cc329629d8b2", "end_token": 89100, "start_token": 87522, "text_sha256": "cc329629d8b261ba8b3de2c325c68053d96351a1a87606b0ca24781662ded12b"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0012491-0014984:478ed25b7e97", "end_token": 89644, "start_token": 89104, "text_sha256": "478ed25b7e97a17b2766f4058786bf7659ffcfd24c9dab96ec5464541a452c98"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0014986-0017326:cb3c7c610ea6", "end_token": 90246, "start_token": 89648, "text_sha256": "cb3c7c610ea654500d77b09fffe18f0de9133aa75f2b97ab1948d9ddbb3b4c14"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0017328-0018417:1e264a52137b", "end_token": 90539, "start_token": 90249, "text_sha256": "1e264a52137bb90f2430fc1d583fa7c20e22682f7a6afc746ec67aa945a4a1c9"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0018419-0019882:45aa4da2fa30", "end_token": 90897, "start_token": 90542, "text_sha256": "45aa4da2fa307189092173f231056a3b41f31b52c1ebf25ca4d7b29ef86506a6"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0019884-0021515:0506cd161eae", "end_token": 91309, "start_token": 90900, "text_sha256": "0506cd161eae4f16da9c1af675eb969006bb7c68ed6bbfbd8844408c1356b037"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0021517-0023092:77d95940d32f", "end_token": 91705, "start_token": 91312, "text_sha256": "77d95940d32fa6a8f09c56e40f3cf2a6d1dc027d954bc9e5a522f47072ef39e5"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0023096-0024555:30941310b1c9", "end_token": 92075, "start_token": 91708, "text_sha256": "30941310b1c94bf116d1afdde08636f3f47a304ade1d9842866a1b444bb86652"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0024557-0025907:a7c1fb548c43", "end_token": 92447, "start_token": 92079, "text_sha256": "a7c1fb548c43c43dcf22c0d1b4cd542818793145466953b34d3d54e813786c0d"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0025909-0027403:23af0821b330", "end_token": 92822, "start_token": 92451, "text_sha256": "23af0821b3308b31b17edc94216c7df796330fecc4cd29a7998cd2237e8f3447"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0027405-0028838:bc5a97345326", "end_token": 93184, "start_token": 92826, "text_sha256": "bc5a973453261596484cbb35ac076d116307b38107867b20a4fe7cb0fa890a80"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0028840-0030086:6dbd252139f9", "end_token": 93509, "start_token": 93188, "text_sha256": "6dbd252139f95ab7ae310ead187349e86670800a92937e56c35503f671d4ed88"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0030088-0032062:0daf34faf915", "end_token": 93961, "start_token": 93513, "text_sha256": "0daf34faf9151b141002856e0089f037d35cbb1bd9a02e974f48f5d6daa68903"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0032066-0034182:a7ee330ff65b", "end_token": 94469, "start_token": 93965, "text_sha256": "a7ee330ff65b9a4736a75b752c27dc7baf1b38d2075c41bc1ebc67c6c3cb4539"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0034186-0035264:ea4bc5bf90fb", "end_token": 94719, "start_token": 94473, "text_sha256": "ea4bc5bf90fbc9a98bd30de1b380a2bcddb55efc7991827525e8918247c32d44"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0035288-0037549:e6b945af4872", "end_token": 95126, "start_token": 94722, "text_sha256": "e6b945af4872e2e58c4d4500cf144262bc5b756325bbd1e8bec2715018c389ad"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0037551-0039075:d5814b6d659f", "end_token": 95504, "start_token": 95130, "text_sha256": "d5814b6d659f9a5240bf000853f24aecce2275048c196f4dae64f9c4b3f52946"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76", "end_token": 111886, "start_token": 95508, "text_sha256": "ebd38261da7659280f75a3d6a07e45c4c554ffd74c2b622be1b2d10a49db7fe4"}, {"artifact_id": "ietf:rfc:8446:chars:0271846-0337410:525df48ab2b8", "end_token": 128005, "start_token": 111890, "text_sha256": "525df48ab2b8a9141892cf0ea81d94e92c6c503602649613667dea066e4c73b7"}], "checks": {"answer_surface_free": true, "artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_top3_prefixes_insufficient": true, "body_and_classification_valid": true, "counterfactual_changes_answer": true, "empty_selection_insufficient": true, "exact_token_count_recomputed": true, "full_counterfactual_sufficient": true, "full_replay_sufficient": true, "lexical_tfidf_top3_prefixes_insufficient": true, "minimal_counterfactual_sufficient": true, "minimal_replay_sufficient": true, "minimum_complexity_met": true, "no_shortcut": true, "raw_token_executable_windows_insufficient": true, "remove_one_fails": true, "single_essential_insufficient": true, "source_relations_replayed": true}, "essential_artifact_ids": ["ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "ietf:rfc:5246:chars:0211487-0222287:1fec83bd80c4", "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "ietf:draft:draft-ietf-tls-tls13-28:chars:0037551-0039075:d5814b6d659f", "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76"], "proof_input_sha256": "893fff21771fb6e8114c8cdf1c3232eaa30553e0e56ecf8e999b6732d17e53ea", "raw_token_offset_windows": {"16k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "f2078463482657f1dceb678f86429bde12017f22e4288d7f0d2b2828f33efe80", "max_artifact_count": 23, "monotonicity_proven": false, "semantic_state_count": 59, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "87f26a47129ed8028f30d6a04d359ec40817f10749adcca50bd2e45a35d4ee2e", "exact_raw_executable": {"enumeration_sha256": "87f26a47129ed8028f30d6a04d359ec40817f10749adcca50bd2e45a35d4ee2e", "exhaustive_start_count": 111622, "semantic_state_count": 54, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 54, "window_tokens": 16384}, "4k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "23f5d0541cd4e5cb8f1dff6fcbf4b3398cb6d547d7c4ab7f29580d90470998fc", "max_artifact_count": 12, "monotonicity_proven": false, "semantic_state_count": 61, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "667bafafe7c00d9e5fc467224431565527506cea5865b02a6926854811036f2d", "exact_raw_executable": {"enumeration_sha256": "667bafafe7c00d9e5fc467224431565527506cea5865b02a6926854811036f2d", "exhaustive_start_count": 123910, "semantic_state_count": 44, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 44, "window_tokens": 4096}, "8k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "f8b50d688d1688fcca525dc0b704a0c456279d895bc9fc210d9eb02c665d987c", "max_artifact_count": 19, "monotonicity_proven": false, "semantic_state_count": 61, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "be94884b71b7f0c54536263d028a58cef3566b65567cdc5e2f802762f9d754d9", "exact_raw_executable": {"enumeration_sha256": "be94884b71b7f0c54536263d028a58cef3566b65567cdc5e2f802762f9d754d9", "exhaustive_start_count": 119814, "semantic_state_count": 47, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 47, "window_tokens": 8192}}, "receipt_sha256": "f1923aea05e64d27696b9404a0a294718fb762a2a5b687d7ce093768cde5968b", "replay": {"counterfactual_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "empty_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "full_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "minimal_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "remove_one": [{"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5246:chars:0211487-0222287:1fec83bd80c4"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0037551-0039075:d5814b6d659f"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76"}], "single_essential": [{"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5246:chars:0211487-0222287:1fec83bd80c4"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0037551-0039075:d5814b6d659f"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76"}]}, "retrieval": {"bm25": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:rfc:8446:chars:0000000-0072617:ebd38261da76", "score": "168.86507971976266"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "score": "128.12877489877692"}, {"artifact_id": "ietf:rfc:8446:chars:0271846-0337410:525df48ab2b8", "score": "125.97876729711818"}]}, "lexical_tfidf": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "score": "0.43146207004410869"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0014986-0017326:cb3c7c610ea6", "score": "0.36921192709526263"}, {"artifact_id": "ietf:rfc:8446:chars:0271846-0337410:525df48ab2b8", "score": "0.32585473443537483"}]}}, "schema_version": "longworld.task-proof-receipt.v5", "sidecar_schema_version": "longworld.task-replay-sidecar.v3", "task_replay_sidecar_sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "tokenizer": {"asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "document_context_tokens": 128005, "model_id": "Qwen/Qwen3.5-4B", "revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "serialized_context_tokens": 128384}, "window_scope": "exact_raw_slice_replay_with_separate_intersection_upper_bound"}, "verification": {"artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_topk_insufficient": true, "candidate_mode": true, "closed_book_unsolved": true, "contiguous_windows_insufficient": true, "counterfactual_changes_answer": true, "counterfactual_replay_sufficient": true, "distractor_invariance_gold": true, "embedding_topk_insufficient": false, "essential_single_doc_insufficient": true, "essential_surface_gold_free": true, "essential_text_grounded": true, "full_sufficient": true, "lexical_tfidf_topk_insufficient": true, "local_window_insufficient": true, "min_complexity": true, "minimal_sufficient": true, "no_shortcut": true, "production_mode": false, "question_only_unsolved": true, "remove_one_fails": true, "schema_ok": true, "semantic_sufficient": true, "strict_executable_sufficient": true, "surface_match": true}, "view_verification": {"classification_ok": true, "essential_present": true, "expected_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "global_proof_green": true, "production_eligible": false, "semantic_text_grounded": true, "strict_replay_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}"}}, "task_proof_sha256": "eeea00d9c950381828d6c02c4b3ca96629e3c74162f2cedb06723206b71bcdda", "task_quality_metadata": {"answer_program_id": "ietf.tls13_handshake_succession.v1", "authentic_source_relation_id": "e0dcb1ad1463850449f0", "base_task_id": "776021e7d8e7f94c336a", "context_source_relation_count": 22, "domain": "standards", "executable_proof_id": "0c64de0f18afd43b8e63", "hybrid_causal_edges": [], "length_bucket": "128k", "motif": "tls13_handshake_succession+obsoletes_updates_closure", "real_source_family_ids": ["ietf_standards"], "real_source_token_ratio": 0.9970557078763709, "real_source_verified": true, "real_source_workflow_ids": ["task-source:73e262c5a2939b145819"], "semantic_base_task_id": "250a7cfbf1fb2c0a24b9", "source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2018-03-01T00:00:00Z", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_relation_kind": "published_as"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "updates"}], "source_relation_id": "6ee046ba0ec12a545607", "world_id": "ietf-tls13-handshake-succession-v4"}, "task_replay_adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "task_replay_payload_sha256": "58034cb5ca73384b58d248a0221969f5c7a78a51df1c3237d0106f3f7b05ce96", "task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "sidecar_schema_version": "longworld.task-replay-sidecar.v3"}, "task_semantic_commitment_sha256": "6aa677d5563f05cc7d13c073a5366b4565cc6393e73aa135963a92717777b09d", "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "top_k": [{"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0014986-0017326:cb3c7c610ea6", "chunk_count": 4, "rank": 1, "score": 0.7148211002349854, "text_sha256": "cb3c7c610ea654500d77b09fffe18f0de9133aa75f2b97ab1948d9ddbb3b4c14"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0072013:18cf3066a9a8", "chunk_count": 108, "rank": 2, "score": 0.7057527303695679, "text_sha256": "18cf3066a9a822587a9996a13ac9baa0218b710ad2b6b2d72542dd0c55cc7b0e"}, {"artifact_id": "ietf:rfc:5246:chars:0146106-0211487:5bc5f9c58046", "chunk_count": 92, "rank": 3, "score": 0.6802471280097961, "text_sha256": "5bc5f9c580462aee827e715243a198ef8b8bb8e6791304831fd9b9fffca38836"}], "verification_replay_sha256": "2c39b2e16de5cb7eebaa5f89cb6b304eb6801945981494cc946887a78f874b93"} +{"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_audit_sha256": "568522427dfdd3adcc062ce5e6c951910a4666e4e0dae230b0dd9626ff00d1e1", "attestation": {"digest": "5a147d130b558b06fb12fa0963611e8f7daf9dbde8200b14e491ba8da1dec794", "environment": "probe", "key_id": "probe-auditor-039d83adbb5136b28aac3f3590b47b41", "purpose": "dense_retrieval_audit", "role": "auditor", "scheme": "hmac-sha256-v2"}, "candidate_sha256": "6f2957d4a31198559985b79b011d9558f53a691eb1b4d276ff292762c80fb06f", "counterfactual_replay_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "embedding_topk_insufficient": true, "expected_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "full_pool_strict_replay_sufficient": true, "k": 3, "local_probe_trust_isolation": "non_independent_local_diagnostic", "model": {"backend": "sentence-transformers-6.0.0", "chunking": {"aggregation": "max_similarity", "max_tokens": 192, "overlap_tokens": 32, "strategy": "tokenizer_token_windows"}, "model_id": "sentence-transformers/all-MiniLM-L6-v2", "provider": "huggingface", "revision": "1110a243fdf4706b3f48f1d95db1a4f5529b4d41", "score_metric": "dot_product"}, "near_dup_sentence_ratio": 0.0296, "query_id": "ietf-tls13-handshake-succession-v4:64k:standard:cf", "ranker_type": "dense_embedding", "ranking_sha256": "f6f5ee8a3d729bacde9a1680bd7689f3606e3835489081031ff1c2a345631171", "schema_version": "train-ready-promotion-v2", "source_binding_sha256": "fc34a690fcb8ffffb64df92f1df724ea75d43ce88c10daf432fb085fee09ddbe", "strict_growth_metrics": {"authentic_source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}], "graph": {"hop_count": 2, "n_essential_artifacts": 7, "n_essential_events": 6, "proof_depth": 2}, "semantic_growth_group_id": "ietf-tls13-handshake-succession-v4|ietf-tls13-handshake-succession", "semantic_tokens": {"causal_supporting": 16788, "event_bearing": 48321, "generic_background": 0, "internal": 65114, "measurement_basis": "exact_pinned_tokenizer_replay_relation_path_whole_artifact_upper_bound", "proof_bearing": 48321}, "strict_support_event_count": 5}, "strict_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "strict_replay_prefix_answers": ["{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}"], "strict_replay_revision": "longworld-strict-replay-v5", "task_proof": {"task_proof_receipt": {"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "artifact_aligned_windows": {"16k": {"enumeration_sha256": "176dd7ff0b642099e09047b3cdbb45efc68edb43dd0eb6792ae30e102a6d56b8", "max_artifact_count": 3, "max_span_tokens": 16365, "status": "insufficient", "strict_window_count": 20, "window_tokens": 16384}, "4k": {"enumeration_sha256": "c17059cf0a6db37f54932b489e8b206cba9297c08835f29fb38c17d5547143a3", "max_artifact_count": 1, "max_span_tokens": 4053, "status": "insufficient", "strict_window_count": 3, "window_tokens": 4096}, "8k": {"enumeration_sha256": "17c2a9b44867d08be887c7aa164a1f5a325af0400cd974e3d29c316e3e490030", "max_artifact_count": 1, "max_span_tokens": 8187, "status": "insufficient", "strict_window_count": 8, "window_tokens": 8192}}, "artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "end_token": 9837, "start_token": 0, "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "end_token": 10302, "start_token": 9841, "text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "end_token": 18494, "start_token": 10307, "text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "end_token": 26619, "start_token": 18498, "text_sha256": "f9892786fca5daa1aca587f899ae9a4d74912e13b89740492729f5c3fe986064"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "end_token": 34756, "start_token": 26623, "text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "end_token": 34863, "start_token": 34760, "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "end_token": 43009, "start_token": 34867, "text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "end_token": 48194, "start_token": 43013, "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "end_token": 52252, "start_token": 48199, "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "end_token": 65114, "start_token": 52257, "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}], "checks": {"answer_surface_free": true, "artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_top3_prefixes_insufficient": true, "body_and_classification_valid": true, "counterfactual_changes_answer": true, "empty_selection_insufficient": true, "exact_token_count_recomputed": true, "full_counterfactual_sufficient": true, "full_replay_sufficient": true, "lexical_tfidf_top3_prefixes_insufficient": true, "minimal_counterfactual_sufficient": true, "minimal_replay_sufficient": true, "minimum_complexity_met": true, "no_shortcut": true, "raw_token_executable_windows_insufficient": true, "remove_one_fails": true, "single_essential_insufficient": true, "source_relations_replayed": true}, "essential_artifact_ids": ["ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"], "proof_input_sha256": "6f2957d4a31198559985b79b011d9558f53a691eb1b4d276ff292762c80fb06f", "raw_token_offset_windows": {"16k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "159605de9d850aaa5e65289a749ab780b61cb505a50fe9f8a06abe85ea650063", "max_artifact_count": 5, "monotonicity_proven": false, "semantic_state_count": 16, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "44a4330b371decef1948dfc2a46a5eac9baa66e8ce2c93d0105748d586b4e537", "exact_raw_executable": {"enumeration_sha256": "44a4330b371decef1948dfc2a46a5eac9baa66e8ce2c93d0105748d586b4e537", "exhaustive_start_count": 48731, "semantic_state_count": 18, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 18, "window_tokens": 16384}, "4k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "57865553c94928c8b274e05a7294df1411c0d970a68bbdccd0f02f1916766cf6", "max_artifact_count": 3, "monotonicity_proven": false, "semantic_state_count": 19, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "6c49fd92d918b166dc6519a3ec1bbb751453b4e2b032ed36b69e633b48a42914", "exact_raw_executable": {"enumeration_sha256": "6c49fd92d918b166dc6519a3ec1bbb751453b4e2b032ed36b69e633b48a42914", "exhaustive_start_count": 61019, "semantic_state_count": 4, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 4, "window_tokens": 4096}, "8k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "1c8147135bb36caef0215ba3a71e1280b3adc7557e4d8e6578451b1f411afb4a", "max_artifact_count": 3, "monotonicity_proven": false, "semantic_state_count": 19, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "520eda04c4f0e40b887017e30f18ba13c85a092576996e26b44a5a8b767c9c6b", "exact_raw_executable": {"enumeration_sha256": "520eda04c4f0e40b887017e30f18ba13c85a092576996e26b44a5a8b767c9c6b", "exhaustive_start_count": 56923, "semantic_state_count": 9, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 9, "window_tokens": 8192}}, "receipt_sha256": "da66ada2a0f5addaa6922c37d1b27e4c41d5db85ebfb4d98bbd79da4a26b0b67", "replay": {"counterfactual_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "empty_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "full_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "minimal_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "remove_one": [{"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}], "single_essential": [{"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}]}, "retrieval": {"bm25": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "score": "117.51975065902222"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "score": "85.654555114976148"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "score": "73.28241578146303"}]}, "lexical_tfidf": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "score": "0.42605918578853391"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "score": "0.41657976134259173"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "score": "0.32400943442123614"}]}}, "schema_version": "longworld.task-proof-receipt.v5", "sidecar_schema_version": "longworld.task-replay-sidecar.v3", "task_replay_sidecar_sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "tokenizer": {"asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "document_context_tokens": 65114, "model_id": "Qwen/Qwen3.5-4B", "revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "serialized_context_tokens": 65493}, "window_scope": "exact_raw_slice_replay_with_separate_intersection_upper_bound"}, "verification": {"artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_topk_insufficient": true, "candidate_mode": true, "closed_book_unsolved": true, "contiguous_windows_insufficient": true, "counterfactual_changes_answer": true, "counterfactual_replay_sufficient": true, "distractor_invariance_gold": true, "embedding_topk_insufficient": false, "essential_single_doc_insufficient": true, "essential_surface_gold_free": true, "essential_text_grounded": true, "full_sufficient": true, "lexical_tfidf_topk_insufficient": true, "local_window_insufficient": true, "min_complexity": true, "minimal_sufficient": true, "no_shortcut": true, "production_mode": false, "question_only_unsolved": true, "remove_one_fails": true, "schema_ok": true, "semantic_sufficient": true, "strict_executable_sufficient": true, "surface_match": true}, "view_verification": {"classification_ok": true, "essential_present": true, "expected_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "global_proof_green": true, "production_eligible": false, "semantic_text_grounded": true, "strict_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}"}}, "task_proof_sha256": "c97caef789864c42b54fe2d92d8710eff7a6d53307bea127742fa20baa73887d", "task_quality_metadata": {"answer_program_id": "ietf.tls13_handshake_succession.v1", "authentic_source_relation_id": "e0dcb1ad1463850449f0", "base_task_id": "776021e7d8e7f94c336a", "context_source_relation_count": 22, "domain": "standards", "executable_proof_id": "0c64de0f18afd43b8e63", "hybrid_causal_edges": [], "length_bucket": "64k", "motif": "tls13_handshake_succession+obsoletes_updates_closure", "real_source_family_ids": ["ietf_standards"], "real_source_token_ratio": 0.8702304063029637, "real_source_verified": true, "real_source_workflow_ids": ["task-source:73e262c5a2939b145819"], "semantic_base_task_id": "250a7cfbf1fb2c0a24b9", "source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2018-03-01T00:00:00Z", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_relation_kind": "published_as"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "updates"}], "source_relation_id": "6ee046ba0ec12a545607", "world_id": "ietf-tls13-handshake-succession-v4"}, "task_replay_adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "task_replay_payload_sha256": "58034cb5ca73384b58d248a0221969f5c7a78a51df1c3237d0106f3f7b05ce96", "task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "sidecar_schema_version": "longworld.task-replay-sidecar.v3"}, "task_semantic_commitment_sha256": "d0ba9fce5d7a14d72aa724fb05a7f9c6c81ffcf19090647fc0fca5aa6cfe21a5", "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "top_k": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "chunk_count": 52, "rank": 1, "score": 0.6783159971237183, "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "chunk_count": 69, "rank": 2, "score": 0.6729809641838074, "text_sha256": "f9892786fca5daa1aca587f899ae9a4d74912e13b89740492729f5c3fe986064"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "chunk_count": 29, "rank": 3, "score": 0.6694298386573792, "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}], "verification_replay_sha256": "33efcf62f09e25af2fd2519f2a8811ee7793f298b9122079bd5a21c1a23a1b99"} +{"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_audit_sha256": "568522427dfdd3adcc062ce5e6c951910a4666e4e0dae230b0dd9626ff00d1e1", "attestation": {"digest": "2b5b694f465b238bcecbc101db341ccc714d4daa3e2fe163be0d0082bfcfd6e6", "environment": "probe", "key_id": "probe-auditor-039d83adbb5136b28aac3f3590b47b41", "purpose": "dense_retrieval_audit", "role": "auditor", "scheme": "hmac-sha256-v2"}, "candidate_sha256": "b9ffa84d0b0c4f27ec539da80a163bd8f89346a6400a51d7652440c323dd61d6", "counterfactual_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "embedding_topk_insufficient": true, "expected_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "full_pool_strict_replay_sufficient": true, "k": 3, "local_probe_trust_isolation": "non_independent_local_diagnostic", "model": {"backend": "sentence-transformers-6.0.0", "chunking": {"aggregation": "max_similarity", "max_tokens": 192, "overlap_tokens": 32, "strategy": "tokenizer_token_windows"}, "model_id": "sentence-transformers/all-MiniLM-L6-v2", "provider": "huggingface", "revision": "1110a243fdf4706b3f48f1d95db1a4f5529b4d41", "score_metric": "dot_product"}, "near_dup_sentence_ratio": 0.0296, "query_id": "ietf-tls13-handshake-succession-v4:64k:standard:full", "ranker_type": "dense_embedding", "ranking_sha256": "3d2a8fb375fb7f15e5d4f03a10a16f4113bc075c539066e56b1cb1d8863a8548", "schema_version": "train-ready-promotion-v2", "source_binding_sha256": "fc34a690fcb8ffffb64df92f1df724ea75d43ce88c10daf432fb085fee09ddbe", "strict_growth_metrics": {"authentic_source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}], "graph": {"hop_count": 2, "n_essential_artifacts": 7, "n_essential_events": 6, "proof_depth": 2}, "semantic_growth_group_id": "ietf-tls13-handshake-succession-v4|ietf-tls13-handshake-succession", "semantic_tokens": {"causal_supporting": 16788, "event_bearing": 48341, "generic_background": 0, "internal": 65134, "measurement_basis": "exact_pinned_tokenizer_replay_relation_path_whole_artifact_upper_bound", "proof_bearing": 48341}, "strict_support_event_count": 6}, "strict_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "strict_replay_prefix_answers": ["{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}"], "strict_replay_revision": "longworld-strict-replay-v5", "task_proof": {"task_proof_receipt": {"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "artifact_aligned_windows": {"16k": {"enumeration_sha256": "0dfdf47f3975b6557021b0ac9fa141d5b0a482dbe8b837666f40decdfcb359d1", "max_artifact_count": 3, "max_span_tokens": 16332, "status": "insufficient", "strict_window_count": 19, "window_tokens": 16384}, "4k": {"enumeration_sha256": "c17059cf0a6db37f54932b489e8b206cba9297c08835f29fb38c17d5547143a3", "max_artifact_count": 1, "max_span_tokens": 4053, "status": "insufficient", "strict_window_count": 3, "window_tokens": 4096}, "8k": {"enumeration_sha256": "9827342ea228a6bdd0a545fb68b476c6abfdb1407ae92f5a47381679372bb2e6", "max_artifact_count": 1, "max_span_tokens": 8187, "status": "insufficient", "strict_window_count": 8, "window_tokens": 8192}}, "artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "end_token": 9837, "start_token": 0, "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "end_token": 10302, "start_token": 9841, "text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "end_token": 18494, "start_token": 10307, "text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "end_token": 26639, "start_token": 18498, "text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "end_token": 34776, "start_token": 26643, "text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "end_token": 34883, "start_token": 34780, "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "end_token": 43029, "start_token": 34887, "text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "end_token": 48214, "start_token": 43033, "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "end_token": 52272, "start_token": 48219, "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "end_token": 65134, "start_token": 52277, "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}], "checks": {"answer_surface_free": true, "artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_top3_prefixes_insufficient": true, "body_and_classification_valid": true, "counterfactual_changes_answer": true, "empty_selection_insufficient": true, "exact_token_count_recomputed": true, "full_counterfactual_sufficient": true, "full_replay_sufficient": true, "lexical_tfidf_top3_prefixes_insufficient": true, "minimal_counterfactual_sufficient": true, "minimal_replay_sufficient": true, "minimum_complexity_met": true, "no_shortcut": true, "raw_token_executable_windows_insufficient": true, "remove_one_fails": true, "single_essential_insufficient": true, "source_relations_replayed": true}, "essential_artifact_ids": ["ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"], "proof_input_sha256": "b9ffa84d0b0c4f27ec539da80a163bd8f89346a6400a51d7652440c323dd61d6", "raw_token_offset_windows": {"16k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "4f0b1843910fd703db5d39e76976b5f85e5a61ed1666bf8ab3295611fd0b2b45", "max_artifact_count": 5, "monotonicity_proven": false, "semantic_state_count": 16, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "215935ba84770df4367dad9a46f1ec1181cdc7e4b093098c18f7b2990a0cd5a9", "exact_raw_executable": {"enumeration_sha256": "215935ba84770df4367dad9a46f1ec1181cdc7e4b093098c18f7b2990a0cd5a9", "exhaustive_start_count": 48751, "semantic_state_count": 17, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 17, "window_tokens": 16384}, "4k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "02fa86d64b71a9758c4ff96b5d522ad44bc885838009f1e86e9be94861702d27", "max_artifact_count": 3, "monotonicity_proven": false, "semantic_state_count": 19, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "4c39eda337b321f26a594cbd77edb4ae2fe1bcdcf8a543e1d44e9130cc544740", "exact_raw_executable": {"enumeration_sha256": "4c39eda337b321f26a594cbd77edb4ae2fe1bcdcf8a543e1d44e9130cc544740", "exhaustive_start_count": 61039, "semantic_state_count": 4, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 4, "window_tokens": 4096}, "8k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "74e2a505eb70d9fcdb90b3f195cd69a2713f6b9c3a76c121b450dbc516debccf", "max_artifact_count": 3, "monotonicity_proven": false, "semantic_state_count": 19, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "649f2e876f8201780e2eb55d677038a44195e5a513717b76679ee51fe67baecb", "exact_raw_executable": {"enumeration_sha256": "649f2e876f8201780e2eb55d677038a44195e5a513717b76679ee51fe67baecb", "exhaustive_start_count": 56943, "semantic_state_count": 9, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 9, "window_tokens": 8192}}, "receipt_sha256": "40397fb32e82c7dcf296f3d3f53620ad1d3be0908173b1b96fd226c9c0e4922d", "replay": {"counterfactual_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "empty_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "full_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "minimal_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "remove_one": [{"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}], "single_essential": [{"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}]}, "retrieval": {"bm25": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "score": "117.4899053298914"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "score": "85.661410221096119"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "score": "73.286337280605935"}]}, "lexical_tfidf": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "c1af44132ab8172e277a7aeb52229454aa629ca8cbfe8e7ea88b299909f9abb7"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "score": "0.42605918578853391"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "score": "0.41657976134259173"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "score": "0.32513082216618144"}]}}, "schema_version": "longworld.task-proof-receipt.v5", "sidecar_schema_version": "longworld.task-replay-sidecar.v3", "task_replay_sidecar_sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "tokenizer": {"asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "document_context_tokens": 65134, "model_id": "Qwen/Qwen3.5-4B", "revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "serialized_context_tokens": 65513}, "window_scope": "exact_raw_slice_replay_with_separate_intersection_upper_bound"}, "verification": {"artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_topk_insufficient": true, "candidate_mode": true, "closed_book_unsolved": true, "contiguous_windows_insufficient": true, "counterfactual_changes_answer": true, "counterfactual_replay_sufficient": true, "distractor_invariance_gold": true, "embedding_topk_insufficient": false, "essential_single_doc_insufficient": true, "essential_surface_gold_free": true, "essential_text_grounded": true, "full_sufficient": true, "lexical_tfidf_topk_insufficient": true, "local_window_insufficient": true, "min_complexity": true, "minimal_sufficient": true, "no_shortcut": true, "production_mode": false, "question_only_unsolved": true, "remove_one_fails": true, "schema_ok": true, "semantic_sufficient": true, "strict_executable_sufficient": true, "surface_match": true}, "view_verification": {"classification_ok": true, "essential_present": true, "expected_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "global_proof_green": true, "production_eligible": false, "semantic_text_grounded": true, "strict_replay_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}"}}, "task_proof_sha256": "553c60df6ac357c3cf9640f56cc81c8c7067df49962e620bd5604bbc4a11a596", "task_quality_metadata": {"answer_program_id": "ietf.tls13_handshake_succession.v1", "authentic_source_relation_id": "e0dcb1ad1463850449f0", "base_task_id": "776021e7d8e7f94c336a", "context_source_relation_count": 22, "domain": "standards", "executable_proof_id": "0c64de0f18afd43b8e63", "hybrid_causal_edges": [], "length_bucket": "64k", "motif": "tls13_handshake_succession+obsoletes_updates_closure", "real_source_family_ids": ["ietf_standards"], "real_source_token_ratio": 0.994230152794102, "real_source_verified": true, "real_source_workflow_ids": ["task-source:73e262c5a2939b145819"], "semantic_base_task_id": "250a7cfbf1fb2c0a24b9", "source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2018-03-01T00:00:00Z", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_relation_kind": "published_as"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "updates"}], "source_relation_id": "6ee046ba0ec12a545607", "world_id": "ietf-tls13-handshake-succession-v4"}, "task_replay_adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "task_replay_payload_sha256": "58034cb5ca73384b58d248a0221969f5c7a78a51df1c3237d0106f3f7b05ce96", "task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "sidecar_schema_version": "longworld.task-replay-sidecar.v3"}, "task_semantic_commitment_sha256": "2a663ac9d67af388be26c067991445dbb2eacca593c56829c9ca45ed70649209", "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "top_k": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "chunk_count": 52, "rank": 1, "score": 0.6783159971237183, "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "chunk_count": 29, "rank": 2, "score": 0.6694298386573792, "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "chunk_count": 69, "rank": 3, "score": 0.6620718240737915, "text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e"}], "verification_replay_sha256": "7a939d2bc8621aca75d4c2f062298f9952d4723d641f751638ada01c40d7ce1c"} +{"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_audit_sha256": "568522427dfdd3adcc062ce5e6c951910a4666e4e0dae230b0dd9626ff00d1e1", "attestation": {"digest": "90c03e396ccdfcc43cbde0bc65e00067f156b6c197046ae388be13be9a45d2a8", "environment": "probe", "key_id": "probe-auditor-039d83adbb5136b28aac3f3590b47b41", "purpose": "dense_retrieval_audit", "role": "auditor", "scheme": "hmac-sha256-v2"}, "candidate_sha256": "b5f0e0ba4612029242d9b8451f7f0560b57a323f65336eba2281eba4efbc8c21", "counterfactual_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "embedding_topk_insufficient": true, "expected_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "full_pool_strict_replay_sufficient": true, "k": 3, "local_probe_trust_isolation": "non_independent_local_diagnostic", "model": {"backend": "sentence-transformers-6.0.0", "chunking": {"aggregation": "max_similarity", "max_tokens": 192, "overlap_tokens": 32, "strategy": "tokenizer_token_windows"}, "model_id": "sentence-transformers/all-MiniLM-L6-v2", "provider": "huggingface", "revision": "1110a243fdf4706b3f48f1d95db1a4f5529b4d41", "score_metric": "dot_product"}, "near_dup_sentence_ratio": 0.0296, "query_id": "ietf-tls13-handshake-succession-v4:64k:standard:ordered_artifact_view", "ranker_type": "dense_embedding", "ranking_sha256": "bd1e8beb9ba383486945101b359a5f72b812eeb7a458cf6ad80ab23dbbadfacc", "schema_version": "train-ready-promotion-v2", "source_binding_sha256": "fc34a690fcb8ffffb64df92f1df724ea75d43ce88c10daf432fb085fee09ddbe", "strict_growth_metrics": {"authentic_source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}], "graph": {"hop_count": 2, "n_essential_artifacts": 7, "n_essential_events": 6, "proof_depth": 2}, "semantic_growth_group_id": "ietf-tls13-handshake-succession-v4|ietf-tls13-handshake-succession", "semantic_tokens": {"causal_supporting": 16788, "event_bearing": 48342, "generic_background": 0, "internal": 65134, "measurement_basis": "exact_pinned_tokenizer_replay_relation_path_whole_artifact_upper_bound", "proof_bearing": 48342}, "strict_support_event_count": 6}, "strict_replay_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "strict_replay_prefix_answers": ["{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}"], "strict_replay_revision": "longworld-strict-replay-v5", "task_proof": {"task_proof_receipt": {"adapter_audit": {"counterfactual_parent_binding_valid": true, "projection_artifact_bindings_valid": true, "projection_candidate_only": true, "projection_chronology_valid": true, "projection_document_digest_valid": true, "serialized_context_valid": true, "source_token_measurement_valid": true}, "adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "artifact_aligned_windows": {"16k": {"enumeration_sha256": "b7d1fa0abebbb673c57714fbe2454484c04d3b9b0c414a92ad071b6529d3bcde", "max_artifact_count": 4, "max_span_tokens": 16324, "status": "insufficient", "strict_window_count": 19, "window_tokens": 16384}, "4k": {"enumeration_sha256": "a06d80f8ba4acda5c7d701e00f394bb5dcca6370e3c3d9e99768251ab7b309cc", "max_artifact_count": 1, "max_span_tokens": 4053, "status": "insufficient", "strict_window_count": 3, "window_tokens": 4096}, "8k": {"enumeration_sha256": "0dd1c5d0253aba64bd0a6937dc169a3a91956a5ecc4e59e9ed83bbe6fb80f9de", "max_artifact_count": 2, "max_span_tokens": 8187, "status": "insufficient", "strict_window_count": 9, "window_tokens": 8192}}, "artifact_bindings": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "end_token": 9837, "start_token": 0, "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:5246:chars:0000000-0035597:ec6aa21eb681", "end_token": 18029, "start_token": 9842, "text_sha256": "ec6aa21eb681a7f7a455e34c5b28ae70e044ed84f5d803214ba1287d032d2df6"}, {"artifact_id": "ietf:rfc:5246:chars:0035597-0072013:3f180e49add1", "end_token": 26166, "start_token": 18033, "text_sha256": "3f180e49add131e1f4ebafc634bb5a75a6ba75e0a641aced256cdc9f38ae5c69"}, {"artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "end_token": 34312, "start_token": 26170, "text_sha256": "596a422d120147050a3cb8cad3bd8c74eb421566bd856d37d7036570eba63cef"}, {"artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "end_token": 38369, "start_token": 34316, "text_sha256": "4d5c8d8dca44bdb215ec6e020ba634197fb5c57d466d21d513421b814242d473"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "end_token": 51231, "start_token": 38374, "text_sha256": "a5ee89216c44759cd4f2b4fc9beab1a712e7a2b81aeefc67da0764260e69a813"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "end_token": 56417, "start_token": 51236, "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "end_token": 56525, "start_token": 56422, "text_sha256": "d68f107e55d255476f3f7762fd990f02be7cdbcfa5eabd434f94f0489e75e98b"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "end_token": 64670, "start_token": 56529, "text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "end_token": 65134, "start_token": 64673, "text_sha256": "0c8ef99440930f6638bfeeaa5af230fbefbd614206443ea33bc8fdd4a56598c5"}], "checks": {"answer_surface_free": true, "artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_top3_prefixes_insufficient": true, "body_and_classification_valid": true, "counterfactual_changes_answer": true, "empty_selection_insufficient": true, "exact_token_count_recomputed": true, "full_counterfactual_sufficient": true, "full_replay_sufficient": true, "lexical_tfidf_top3_prefixes_insufficient": true, "minimal_counterfactual_sufficient": true, "minimal_replay_sufficient": true, "minimum_complexity_met": true, "no_shortcut": true, "raw_token_executable_windows_insufficient": true, "remove_one_fails": true, "single_essential_insufficient": true, "source_relations_replayed": true}, "essential_artifact_ids": ["ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "ietf:rfc:5246:chars:0072013-0109924:596a422d1201", "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44", "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"], "proof_input_sha256": "b5f0e0ba4612029242d9b8451f7f0560b57a323f65336eba2281eba4efbc8c21", "raw_token_offset_windows": {"16k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "5bc01bf000da7ac2ade7ec5898db05447246702b32b64dbdb88f40a618290a69", "max_artifact_count": 5, "monotonicity_proven": false, "semantic_state_count": 14, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "9a9eb35d89e2afee75c3eaadb5075ff576d4bb7216063b1a25bfbf9890abd41c", "exact_raw_executable": {"enumeration_sha256": "9a9eb35d89e2afee75c3eaadb5075ff576d4bb7216063b1a25bfbf9890abd41c", "exhaustive_start_count": 48751, "semantic_state_count": 14, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 14, "window_tokens": 16384}, "4k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "19520947cfb7d39e7ba09721be0e6bfe2f65b2c67daa8eed8f61436684c0bdce", "max_artifact_count": 3, "monotonicity_proven": false, "semantic_state_count": 18, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "022aba6f56ecbfde2808c571823ddf1efd3b9b39d9069817e25e69c73c914de2", "exact_raw_executable": {"enumeration_sha256": "022aba6f56ecbfde2808c571823ddf1efd3b9b39d9069817e25e69c73c914de2", "exhaustive_start_count": 61039, "semantic_state_count": 4, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 4, "window_tokens": 4096}, "8k": {"conservative_intersection_upper_bound": {"can_establish_pass": false, "enumeration_sha256": "4db7b6ec0c02879b159738a1b136d4260bf4a23424959ca13f7b1633f4e86094", "max_artifact_count": 4, "monotonicity_proven": false, "semantic_state_count": 18, "status": "gold_absent_non_certifying"}, "enumeration_sha256": "cea077c73f3dc2a8273cea811e0c0ce3ea7b1f7f469986e8ac5190c3159e15c0", "exact_raw_executable": {"enumeration_sha256": "cea077c73f3dc2a8273cea811e0c0ce3ea7b1f7f469986e8ac5190c3159e15c0", "exhaustive_start_count": 56943, "semantic_state_count": 10, "state_partition": "complete_canonical_record_visibility_v1", "status": "insufficient"}, "status": "executable_insufficient", "strict_window_count": 10, "window_tokens": 8192}}, "receipt_sha256": "0426657dedc1fdd08e4844fdc1a48cad07666ba7ce72f08cfb811f9b55964be1", "replay": {"counterfactual_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "empty_answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "full_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "minimal_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "remove_one": [{"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}, {"answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "removed_artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "removed_artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}], "single_essential": [{"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5246:chars:0072013-0109924:596a422d1201"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:5705:chars:0000000-0016341:4d5c8d8dca44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2"}, {"answer": "{\"current_protocol\":\"UNKNOWN\",\"obsoletes_ocsp_stapling\":\"UNKNOWN\",\"obsoletes_session_tickets\":\"UNKNOWN\",\"obsoletes_tls12\":\"UNKNOWN\",\"updates_exporters\":\"UNKNOWN\",\"updates_extensions\":\"UNKNOWN\"}", "artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4"}]}, "retrieval": {"bm25": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "score": "117.4899053298914"}, {"artifact_id": "ietf:rfc:6066:chars:0000000-0055067:a5ee89216c44", "score": "85.661410221096119"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "score": "73.286337280605935"}]}, "lexical_tfidf": {"all_prefixes_insufficient": true, "prefix_answer_sha256": ["1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "1e96f319a0c5c9ddae535d7c7ae2c312fa5f83819ca2da62e872ecccfd099ba3", "c1af44132ab8172e277a7aeb52229454aa629ca8cbfe8e7ea88b299909f9abb7"], "top1_insufficient": true, "top3": [{"artifact_id": "ietf:draft:draft-ietf-tls-tls13-28:chars:0000000-0000342:d68f107e55d2", "score": "0.42605918578853391"}, {"artifact_id": "ietf:rfc:8446:chars:0335784-0337410:0c8ef9944093", "score": "0.41657976134259173"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "score": "0.32513082216618144"}]}}, "schema_version": "longworld.task-proof-receipt.v5", "sidecar_schema_version": "longworld.task-replay-sidecar.v3", "task_replay_sidecar_sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "tokenizer": {"asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "document_context_tokens": 65134, "model_id": "Qwen/Qwen3.5-4B", "revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", "serialized_context_tokens": 65513}, "window_scope": "exact_raw_slice_replay_with_separate_intersection_upper_bound"}, "verification": {"artifact_aligned_windows_insufficient": true, "bm25_top1_insufficient": true, "bm25_topk_insufficient": true, "candidate_mode": true, "closed_book_unsolved": true, "contiguous_windows_insufficient": true, "counterfactual_changes_answer": true, "counterfactual_replay_sufficient": true, "distractor_invariance_gold": true, "embedding_topk_insufficient": false, "essential_single_doc_insufficient": true, "essential_surface_gold_free": true, "essential_text_grounded": true, "full_sufficient": true, "lexical_tfidf_topk_insufficient": true, "local_window_insufficient": true, "min_complexity": true, "minimal_sufficient": true, "no_shortcut": true, "production_mode": false, "question_only_unsolved": true, "remove_one_fails": true, "schema_ok": true, "semantic_sufficient": true, "strict_executable_sufficient": true, "surface_match": true}, "view_verification": {"classification_ok": true, "essential_present": true, "expected_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}", "global_proof_green": true, "production_eligible": false, "semantic_text_grounded": true, "strict_replay_answer": "{\"current_protocol\":\"TLS_1_3_RFC8446\",\"obsoletes_ocsp_stapling\":\"RFC6961\",\"obsoletes_session_tickets\":\"RFC5077\",\"obsoletes_tls12\":\"RFC5246\",\"updates_exporters\":\"RFC5705\",\"updates_extensions\":\"RFC6066\"}"}}, "task_proof_sha256": "911e4bf3f5340c91c8cf2229414013f3706ac7751a1659e690070d754fd0002f", "task_quality_metadata": {"answer_program_id": "ietf.tls13_handshake_succession.v1", "authentic_source_relation_id": "e0dcb1ad1463850449f0", "base_task_id": "776021e7d8e7f94c336a", "context_source_relation_count": 22, "domain": "standards", "executable_proof_id": "0c64de0f18afd43b8e63", "hybrid_causal_edges": [], "length_bucket": "64k", "motif": "tls13_handshake_succession+obsoletes_updates_closure", "real_source_family_ids": ["ietf_standards"], "real_source_token_ratio": 0.994230152794102, "real_source_verified": true, "real_source_workflow_ids": ["task-source:73e262c5a2939b145819"], "semantic_base_task_id": "250a7cfbf1fb2c0a24b9", "source_relation_edges": [{"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "informative_reference"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "normative_reference"}, {"child_record_id": "ietf:rfc:5077", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:5246", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:6961", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "obsoletes"}, {"child_record_id": "ietf:rfc:8446", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "published_as"}, {"child_record_id": "ietf:rfc:5705", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_record_id": "ietf:rfc:6066", "parent_record_id": "ietf:rfc:8446", "relation_provenance": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:informative_reference:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "informative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:normative_reference:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "normative_reference"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:5077", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5077", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2008-08-01T00:00:00Z", "parent_record_id": "ietf:rfc:5246", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:5246", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2013-06-01T00:00:00Z", "parent_record_id": "ietf:rfc:6961", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:obsoletes:ietf:rfc:8446:ietf:rfc:6961", "source_relation_kind": "obsoletes"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2018-03-01T00:00:00Z", "parent_record_id": "ietf:draft:draft-ietf-tls-tls13-28", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:published_as:ietf:draft:draft-ietf-tls-tls13-28:ietf:rfc:8446", "source_relation_kind": "published_as"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2010-03-01T00:00:00Z", "parent_record_id": "ietf:rfc:5705", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:5705", "source_relation_kind": "updates"}, {"child_occurred_at": "2018-08-01T00:00:00Z", "child_record_id": "ietf:rfc:8446", "parent_occurred_at": "2011-01-01T00:00:00Z", "parent_record_id": "ietf:rfc:6066", "relation_provenance": "authenticated_relation_timestamp_order", "source_relation_id": "ietf:updates:ietf:rfc:8446:ietf:rfc:6066", "source_relation_kind": "updates"}], "source_relation_id": "6ee046ba0ec12a545607", "world_id": "ietf-tls13-handshake-succession-v4"}, "task_replay_adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "task_replay_payload_sha256": "58034cb5ca73384b58d248a0221969f5c7a78a51df1c3237d0106f3f7b05ce96", "task_replay_sidecar": {"adapter_id": "standards.ietf_tls13_handshake_succession.v1", "adapter_revision": "longworld.ietf-tls13-handshake-succession-replay.v1", "sha256": "a96ec5a46b1ab0d298738ca6372e14aa2c2bcd232ab9c922f97dc1fdedd2200b", "sidecar_schema_version": "longworld.task-replay-sidecar.v3"}, "task_semantic_commitment_sha256": "dfad22d5f03f0a1bb543ca914cd6b7e0461f5629cb0c87f3c05816c3befde9bd", "tokenizer_asset_manifest_sha256": "bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb", "top_k": [{"artifact_id": "ietf:rfc:5077:chars:0000000-0041965:80ff79980ee8", "chunk_count": 52, "rank": 1, "score": 0.6783159971237183, "text_sha256": "80ff79980ee80597e0553814e1a3d8639a39df6ee67b5238cea301e3383e5d49"}, {"artifact_id": "ietf:rfc:6961:chars:0000000-0021461:bde82964b134", "chunk_count": 29, "rank": 2, "score": 0.6694298386573792, "text_sha256": "bde82964b13437810c6cae298919e39322ad3a788481abf5cfef813cf713867a"}, {"artifact_id": "ietf:rfc:8446:chars:0000000-0038698:61f9e133f4c4", "chunk_count": 69, "rank": 3, "score": 0.6620718240737915, "text_sha256": "61f9e133f4c4a4636778976b454b1df6fc10392a3c60969990fec1776156105e"}], "verification_replay_sha256": "b0fb9f93f1a7c4335494d26b2a82fa747c25094adee693d1f91e567d9bdc0b94"} diff --git a/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/train_candidates.jsonl b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/train_candidates.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..d19d2f4e7dd727b6b0224940cbd59840faa8ddaa --- /dev/null +++ b/local-probe-train-ready/p57-ietf-tls13-handshake-succession-probe-1-v1-promoted-v1/train_candidates.jsonl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6c202213590e1bfae6d21c55fc76f28bd6a36118742d59516f1dd535eb3b4866 +size 12250510 diff --git a/local-probe-train-ready/p64-primary-training-v2/P64_TRAINING_MANIFEST.json b/local-probe-train-ready/p64-primary-training-v2/P64_TRAINING_MANIFEST.json new file mode 100644 index 0000000000000000000000000000000000000000..4a08db933b7f59468804ee45f9906b55db8471bb --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/P64_TRAINING_MANIFEST.json @@ -0,0 +1 @@ +{"advisory_dependency_audit_bindings":[],"arguments":{"code_catalog":"/WORKSPACE/longworld-worlds/configs/p64_codeforge_taskbank_catalog_v1.json","code_root":"/WORKSPACE/longworld/data/candidates/p64_codeforge_sft_v1","finance_catalog":"/WORKSPACE/longworld-worlds/configs/p64_finance_taskbank_catalog_v1.json","finance_root":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2","recipe_path":"/WORKSPACE/longworld-worlds/configs/llamafactory/TASKBANK_P64.yaml"},"attestation":{"digest":"988333f6de8b242a7973355622ee8f41ca64f9ab340c6ed48ded8895ef20743a","environment":"probe","key_id":"probe-report-ae02be9c17f2c0dc2110745c9eb86acb","purpose":"training_export_manifest","role":"report","scheme":"hmac-sha256-v2"},"canonical_tasks":2486,"combined_counts":{"eval":533,"train":1953},"counts":{"codeforge":{"eval":147,"train":657},"finance":{"eval":386,"train":1296}},"dataset_modes":{"codeforge":"p64_codeforge_train","combined":"p64_finance_train,p64_codeforge_train","finance":"p64_finance_train"},"default_mode":"separate_domains","excluded_accounting":"CodeForge short.jsonl and overlength rejects are excluded from this primary handoff","framework_preprocessing_verified":false,"input_bindings":[{"bytes":1625,"path":"/WORKSPACE/longworld-worlds/configs/p13_codeforge_dprint_three_release_recovery_workspace_v2_bundle.json","sha256":"7520312ccc079e25a7b40e19957ea7e9b318991a9acec5f56bd00d0a4921aa78"},{"bytes":4432,"path":"/WORKSPACE/longworld-worlds/configs/p13_codeforge_oxc_three_release_recovery_workspace_v10_bundle.json","sha256":"7e54e16d65cbb9a0880b670204f362102d70ced6f514a5600269c35181764b44"},{"bytes":2638,"path":"/WORKSPACE/longworld-worlds/configs/p13_codeforge_ruff_first_three_view_workspace_v4_bundle.json","sha256":"48b7237dbcedcd01858791a1fddb30e3b0903c002d170d37e861e35c0c9853e8"},{"bytes":1980,"path":"/WORKSPACE/longworld-worlds/configs/p15_codeforge_uv_release_validation_v1_bundle.json","sha256":"17bd6cccd8abcd929c4b25dde48944ea175c13be689ac67c87a37b243f11bfa6"},{"bytes":2044,"path":"/WORKSPACE/longworld-worlds/configs/p16_codeforge_pulumi_failure_recovery_v1_bundle.json","sha256":"d23382ce6922a1415ee6dec685c8796632a71c93c55eb3137c69f68a97803418"},{"bytes":2131,"path":"/WORKSPACE/longworld-worlds/configs/p17_codeforge_transformers_failure_recovery_v1_bundle.json","sha256":"1bd7c439f2bc3f3483b6334a3bfae024ffc0575749bc3ba37689eb166445a166"},{"bytes":1646,"path":"/WORKSPACE/longworld-worlds/configs/p64_codeforge_opensearch_source_bundle_v1.json","sha256":"79e12caa25b4985911be4b091a58d20c0940e8dea178f038f1555515112c3254"},{"bytes":1705,"path":"/WORKSPACE/longworld-worlds/configs/p64_codeforge_taskbank_catalog_v1.json","sha256":"9664e83ad47437377aaaa6a7b99b5413697397e3a4517e823409474f8e592aef"},{"bytes":481,"path":"/WORKSPACE/longworld-worlds/configs/p64_codeforge_taskbank_dprint_v1.json","sha256":"fb3be6dcf914d60e4827a4523e1720706ed869b796ab83cff6543477d44001b3"},{"bytes":478,"path":"/WORKSPACE/longworld-worlds/configs/p64_codeforge_taskbank_opensearch_v1.json","sha256":"a6150666acbe63ed26c2a9435fcf2f0e70eab376b2b409a93c8dc11df6b6aaab"},{"bytes":481,"path":"/WORKSPACE/longworld-worlds/configs/p64_codeforge_taskbank_oxc_v1.json","sha256":"655fce2f461c2cf36e0464aecada47b98918ba33a8425e2202431408319c4492"},{"bytes":465,"path":"/WORKSPACE/longworld-worlds/configs/p64_codeforge_taskbank_pulumi_v1.json","sha256":"df412916b23d90b7a83b04565cdfaf88bda3cd4b5d0d4561af1ef3207464c817"},{"bytes":474,"path":"/WORKSPACE/longworld-worlds/configs/p64_codeforge_taskbank_ruff_v1.json","sha256":"0560e4eea6ea30b4b4e72e73a2061604863ff69d7a587c0e326ca33035221d78"},{"bytes":482,"path":"/WORKSPACE/longworld-worlds/configs/p64_codeforge_taskbank_transformers_v1.json","sha256":"1f8637c1c101f57a94487a96d14cf024e50b86cc58603d5d053c6f9651b83d49"},{"bytes":461,"path":"/WORKSPACE/longworld-worlds/configs/p64_codeforge_taskbank_uv_v1.json","sha256":"add502548abd6842beffede9eac43a521a3129f498064d41ae6fa07e1611d861"},{"bytes":542,"path":"/WORKSPACE/longworld-worlds/configs/p64_finance_taskbank_alphabet_v1.json","sha256":"37b41f8d73bb8e9f9b7330befc9e5899a0c2d7b501a2a59f84584998c0e1418d"},{"bytes":540,"path":"/WORKSPACE/longworld-worlds/configs/p64_finance_taskbank_amazon_v1.json","sha256":"d12a2846f7ad26f4eabff1a9688513a3ac6d591699cde188559f4432deec76d4"},{"bytes":571,"path":"/WORKSPACE/longworld-worlds/configs/p64_finance_taskbank_amd_v1.json","sha256":"664ac3f5c92af21fcc47a18b4e9884c38745a37d4a477894c591132b4c97f67c"},{"bytes":4332,"path":"/WORKSPACE/longworld-worlds/configs/p64_finance_taskbank_catalog_v1.json","sha256":"eccac00ade1d286ffedee91e53e5c41ea93c6e44f69f2d013871df3f4ecaec9b"},{"bytes":518,"path":"/WORKSPACE/longworld-worlds/configs/p64_finance_taskbank_intel_v1.json","sha256":"382a20c11211a0640d2cf6c341311735e08286c8d7119d7c0fc49d4c136ae983"},{"bytes":538,"path":"/WORKSPACE/longworld-worlds/configs/p64_finance_taskbank_meta_v1.json","sha256":"5bb660c2bde8eb61f67ab8b1749c3715bc2e70dd54ea825c13978c66a9d1aa46"},{"bytes":540,"path":"/WORKSPACE/longworld-worlds/configs/p64_finance_taskbank_micron_v1.json","sha256":"78dc57f5056f3d3b8c008a529844607b374ef805cc27cb6c7ccc99a0f640b7b2"},{"bytes":540,"path":"/WORKSPACE/longworld-worlds/configs/p64_finance_taskbank_nvidia_v1.json","sha256":"5bcca736d11447f2131f56d9974c0f469235b9c971eb647172f9b6a7811ca245"},{"bytes":547,"path":"/WORKSPACE/longworld-worlds/configs/p64_finance_taskbank_siriusxm_v1.json","sha256":"84e448f337573008fcfb2fa7cea069ee17bdc0b5bfce6d083b40421b811ce14a"},{"bytes":11430,"path":"/WORKSPACE/longworld-worlds/longworld/core/attestation.py","sha256":"12d0b762e86a6f14f21665f4f3152f3c93aa909b2efc41def209fe1233b80fa8"},{"bytes":16229,"path":"/WORKSPACE/longworld-worlds/longworld/core/codeforge_taskbank.py","sha256":"bbcd4c8fd7f6ce3201d16bca116b3b380ee6d279a79fa9dda56851510b61df42"},{"bytes":40769,"path":"/WORKSPACE/longworld-worlds/longworld/core/finance_taskbank.py","sha256":"943b8e4d80baf5937599fe53d0432c560785586771e1437a3b7fbd1b327228c3"},{"bytes":16914,"path":"/WORKSPACE/longworld-worlds/longworld/core/finance_taskbank_v2.py","sha256":"d1dcfa2bc5ce05be179980bf3595a03b782ac497f58199e396a5c42e387e877d"},{"bytes":17739,"path":"/WORKSPACE/longworld-worlds/longworld/core/issuer_generic_inline.py","sha256":"bf506eefb942fec2baa975eee95489117f651d6587359b6209286ad8b9950480"},{"bytes":107875,"path":"/WORKSPACE/longworld-worlds/longworld/core/issuerfilingworkflow.py","sha256":"5aadaf3ff4f55328a51377b0bf5ce153d12d3c4b7fbdd124930a914f0e35cee3"},{"bytes":32202,"path":"/WORKSPACE/longworld-worlds/longworld/core/issuerinlineworkflow.py","sha256":"0c45a936304d8f78ea7cb4451810bd1e5f9f4688efcfaaa3718485c42468b4db"},{"bytes":10513,"path":"/WORKSPACE/longworld-worlds/longworld/core/provenance.py","sha256":"361248fc1e1737bfac88e0bed9b01c15aa883a241dfe53b32bd155c79572290b"},{"bytes":22210,"path":"/WORKSPACE/longworld-worlds/longworld/core/realworkflow.py","sha256":"cfaba8b9172a54f881b47ec94e59e819ebfcd39337408b0c2c15a1d2f542bea9"},{"bytes":12852,"path":"/WORKSPACE/longworld-worlds/longworld/core/taskbank_context.py","sha256":"2720171270f1bbe9983bf423156334a85cfea3e6e31c40d2533be1ba739fe2b1"},{"bytes":4076,"path":"/WORKSPACE/longworld-worlds/longworld/core/taskbank_long_context.py","sha256":"596650b7726c4da8b90e54d7887b1917fb29f5d4b5d125de1c444797a2981dba"},{"bytes":4288,"path":"/WORKSPACE/longworld-worlds/longworld/core/tokenizer_assets.py","sha256":"792ab0ffe4b77439dda2f5ccaf90b63aa327c760464a2e272a9b9bb4b74179f6"},{"bytes":963,"path":"/WORKSPACE/longworld-worlds/pyproject.toml","sha256":"222a003027a49cdb02965c30940d05d40e0890b75ec9c706331124662966c730"},{"bytes":215698,"path":"/WORKSPACE/longworld-worlds/reports/p15_codeforge_uv_release_validation_sources_v1/uv_pr20804_0.12.4.json","sha256":"4ac2e9abe813bc62c218ac783f1e3e5ea13d6d51c17f5c7a5c7e8b4fcbd1d5fe"},{"bytes":222761,"path":"/WORKSPACE/longworld-worlds/reports/p15_codeforge_uv_release_validation_sources_v1/uv_pr20921_0.12.2.json","sha256":"7d2df9879d0d62fb4db13ccf05f6446dba1842dbb27b678e5ae3090d1b63c5c4"},{"bytes":103317,"path":"/WORKSPACE/longworld-worlds/reports/p15_codeforge_uv_release_validation_sources_v1/uv_pr20988_0.12.3.json","sha256":"f7654548868521848b4a45b3ef6e78de48bdcc3f78a51a985c6c06903597c45e"},{"bytes":127894,"path":"/WORKSPACE/longworld-worlds/reports/p15_codeforge_uv_release_validation_sources_v1/uv_pr21138_0.12.5.json","sha256":"f15dcfc3340a3d9accd479ed9828031511e2668172a0dc61ee64c9ba329ac048"},{"bytes":138458,"path":"/WORKSPACE/longworld-worlds/reports/p15_codeforge_uv_release_validation_sources_v1/uv_pr21186_0.12.7.json","sha256":"ad7b5f350158cbe0813d92695bd4ea82fe9f91ac88ad8850051af85ec2425539"},{"bytes":78459,"path":"/WORKSPACE/longworld-worlds/reports/p15_codeforge_uv_release_validation_sources_v1/uv_pr21295_0.12.6.json","sha256":"b0a345bf532e8541f6f34d2f6d1fd9e7109aa65635c1324da45ebddcdf9c67ff"},{"bytes":225366,"path":"/WORKSPACE/longworld-worlds/reports/p15_codeforge_uv_release_validation_sources_v1/uv_pr21373_0.12.8.json","sha256":"6a4c1963b96bf3a750639e4c55f829818fc2528a41abc729f49ee6ea91e7c278"},{"bytes":107641,"path":"/WORKSPACE/longworld-worlds/reports/p15_codeforge_uv_release_validation_sources_v1/uv_pr21408_0.12.9.json","sha256":"e44cfcbb3bfcd2f61f6d2dca97d3b01d49d176861d05742300f21ca44bf43e4a"},{"bytes":344558,"path":"/WORKSPACE/longworld-worlds/reports/p16_codeforge_pulumi_failure_recovery_sources_v1/pulumi_pr22812_v3.235.0.json","sha256":"29ba9ac88dda40b94070685ed6bbbe4aaad13220702dc9008571a1b141f6cc90"},{"bytes":102255,"path":"/WORKSPACE/longworld-worlds/reports/p16_codeforge_pulumi_failure_recovery_sources_v1/pulumi_pr22862_v3.237.0.json","sha256":"7b6a7e4600d046f6b835904be52ec4f583f52930aca6853f8488d899bf1a2e00"},{"bytes":89885,"path":"/WORKSPACE/longworld-worlds/reports/p16_codeforge_pulumi_failure_recovery_sources_v1/pulumi_pr22864_v3.238.0.json","sha256":"45092015a98c3cf85b29e7b97a14179fb794278010aa4c64addabb4a9a619b00"},{"bytes":379958,"path":"/WORKSPACE/longworld-worlds/reports/p16_codeforge_pulumi_failure_recovery_sources_v1/pulumi_pr23092_v3.241.0.json","sha256":"f3378bf2954893fdcb4d3e444ee937f187cda082cd512509321b0337c5649ef3"},{"bytes":95279,"path":"/WORKSPACE/longworld-worlds/reports/p16_codeforge_pulumi_failure_recovery_sources_v1/pulumi_pr23245_v3.242.0.json","sha256":"815d457816a62340a9d16f53c3657311aa3f2b17d2aec9ccf1c229846d8c6e02"},{"bytes":59777,"path":"/WORKSPACE/longworld-worlds/reports/p16_codeforge_pulumi_failure_recovery_sources_v1/pulumi_pr23425_v3.245.0.json","sha256":"af5d8b9665267b3e227f4e0e9cf0c9120d1b3ddbd020717c415e8c560b49a90b"},{"bytes":106516,"path":"/WORKSPACE/longworld-worlds/reports/p16_codeforge_pulumi_failure_recovery_sources_v1/pulumi_pr23450_v3.246.0.json","sha256":"1382bb5ac2b0c2bba9b5d7f7b481143c453767954e2c15b85301b9bf91779018"},{"bytes":63832,"path":"/WORKSPACE/longworld-worlds/reports/p16_codeforge_pulumi_failure_recovery_sources_v1/pulumi_pr23831_v3.252.0.json","sha256":"9977fcb88d75b751c3f31f85f4c22f143c55416b1d1b1064bdb9260913a0f014"},{"bytes":116144,"path":"/WORKSPACE/longworld-worlds/reports/p17_codeforge_transformers_failure_recovery_sources_v1/transformers_pr40744_v4.57.0.json","sha256":"ef089162f4e70694367679a93282643b69307335240aa62135ebc245559f8dc1"},{"bytes":132400,"path":"/WORKSPACE/longworld-worlds/reports/p17_codeforge_transformers_failure_recovery_sources_v1/transformers_pr44907_v5.12.0.json","sha256":"d383dd741ddd1aa5596d07bf7639c3635f50385099e9e28bd6e98cdbad7a1d36"},{"bytes":44302,"path":"/WORKSPACE/longworld-worlds/reports/p17_codeforge_transformers_failure_recovery_sources_v1/transformers_pr45921_v5.9.0.json","sha256":"fde48892f9f9146576f51f94dd1e1d07f1fb28cb1239c8ac9e3e3abb4b75fe6b"},{"bytes":88592,"path":"/WORKSPACE/longworld-worlds/reports/p17_codeforge_transformers_failure_recovery_sources_v1/transformers_pr46142_v5.10.1.json","sha256":"2faade63aff90132e9eef7ac91864ed7d7b59f3f6e054d7682601fcfc1cdcdd1"},{"bytes":115344,"path":"/WORKSPACE/longworld-worlds/reports/p17_codeforge_transformers_failure_recovery_sources_v1/transformers_pr46830_v5.13.0.json","sha256":"7547382c55d150a2193a492ade468aaeef9838b6a5291056f75ccc5a9124bfd2"},{"bytes":141818,"path":"/WORKSPACE/longworld-worlds/reports/p17_codeforge_transformers_failure_recovery_sources_v1/transformers_pr46965_v5.14.0.json","sha256":"891b9d99c921f8c8f55a61fbecd770e2cf6d415d39ae2d872d48bdd268d2944e"},{"bytes":133379,"path":"/WORKSPACE/longworld-worlds/reports/p17_codeforge_transformers_failure_recovery_sources_v1/transformers_pr47727_v5.15.0.json","sha256":"ac39bc1fe22a7f40c5bef757e3e543eaa0889d7fa074520293331a0aa8174ea5"},{"bytes":153458,"path":"/WORKSPACE/longworld-worlds/reports/p17_codeforge_transformers_failure_recovery_sources_v1/transformers_pr48023_v5.16.0.json","sha256":"7a1053ffaccd2e4006a53407274f87210877954f637b5de27b6c65fd0f88ec0c"},{"bytes":16967,"path":"/WORKSPACE/longworld-worlds/scripts/export_codeforge_taskbank_sft.py","sha256":"754ac43a437680aac166dc3a5e915eda2d7095d871a4590d53e913e61d104d9e"},{"bytes":6866,"path":"/WORKSPACE/longworld-worlds/scripts/materialize_codeforge_taskbank.py","sha256":"669d5295d93f84ebe64d04e994e0b72fbebec0a149fa420f9af36e336f170a9d"},{"bytes":12241,"path":"/WORKSPACE/longworld-worlds/scripts/materialize_finance_histories.py","sha256":"e441bcb922d29f10c4c2339125a082e7b1d460c8f72de7d1ed8d9893f05dd797"},{"bytes":22858,"path":"/WORKSPACE/longworld-worlds/scripts/materialize_finance_taskbank.py","sha256":"582982f8dfd34e086a6a46cd316778d27650ccf18a6b6541ec682075f95c5867"},{"bytes":14472,"path":"/WORKSPACE/longworld-worlds/scripts/materialize_finance_taskbank_v2.py","sha256":"8f5e64e445309006dde8e5a43f4ba5e0df19f6d019af607afad194d1d29a07a8"},{"bytes":13978,"path":"/WORKSPACE/longworld-worlds/scripts/train_sft.py","sha256":"749f72b641021cb79940fcf09c11f5c7861459617c8c6a724f093888cb184d41"},{"bytes":890515,"path":"/WORKSPACE/longworld-worlds/uv.lock","sha256":"b8518611546a53105169cb3d8d508cfab352b4ac0f29088f7be01b4a6f3bfc7a"},{"bytes":30083,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_sft_v1/BUILD_RECEIPT.json","sha256":"6a90bddcf6a96a2654329c35e425efb76c0c5e3af26c4bf0593faf3ac0ae463b"},{"bytes":454,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_sft_v1/dataset_info.json","sha256":"fded4de52642a51c898f216abbb5ed39fe038fe2699f5532965653cbc79e59e6"},{"bytes":56646863,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_sft_v1/eval.jsonl","sha256":"7d25be28abf066d10d51d2c37f4ed6abe97211270f8288a154dd069f21302333"},{"bytes":947784,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_sft_v1/metadata.jsonl","sha256":"cc9b17cba2fbba045b9ee2863ff73a2529faa06b0c094e1d745878c5f6c6496d"},{"bytes":1327,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_sft_v1/rejects.jsonl","sha256":"451777457585d705cd8c50a17732462bac01cc270431b81ea1573e548fd887db"},{"bytes":1151560,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_sft_v1/short.jsonl","sha256":"52ce5df60e1bf7820aaf7b9e208acdc7019e5a33052bb16308c5e9680b4eb89c"},{"bytes":242682477,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_sft_v1/train.jsonl","sha256":"4547ea4692dd32148636b04f01d0cdf400bf3c00e330eed3bd9cc5cf03f828ff"},{"bytes":3288,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/dprint/BUILD_RECEIPT.json","sha256":"dd8331fa8d11f30d617e7cb5aef1b016800650c68e2b94bc67dda8532a6ae1d2"},{"bytes":6318,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/dprint/contexts.jsonl","sha256":"7039e5d8c0952eb625483095df7afc0d7c6904fffa79942f6e69af46283f31fe"},{"bytes":709497,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/dprint/tasks.jsonl","sha256":"30564dea9d936e722e2a7391d32e954497aa7fb8ff69ac22bd3a8a18281ee4dc"},{"bytes":1254812,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/dprint/world.json","sha256":"37553dcad5e1c7901eed02fdc0dddc03ee1c758f32a8599d25ef31a8809b1547"},{"bytes":3356,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/opensearch/BUILD_RECEIPT.json","sha256":"a03cb642b4a9f1fc581eb58ba2316589d94dbe45d25f33b039db14b5f92acfba"},{"bytes":6380,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/opensearch/contexts.jsonl","sha256":"7ebe740c38c342ec14bde3baa699482267b04b0748cb8800e405d4eb6e1854f3"},{"bytes":763980,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/opensearch/tasks.jsonl","sha256":"5f3c5091573cd1f7097e7e5eda7f3d1257a39c72ce8b88d80f0ae33fd74021b1"},{"bytes":1287693,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/opensearch/world.json","sha256":"2f42f8ac5152da249c8a7f1da166fa03a5fe1522858d6d5e83170b5c57593d06"},{"bytes":6327,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/oxc/BUILD_RECEIPT.json","sha256":"d458cb35029bb8f43ed2a7fd19940936ab31aceb073930030ef9683112362a46"},{"bytes":43425,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/oxc/contexts.jsonl","sha256":"0f0e2ba184f78a1e4c071832b7fe3eb4206f390080dd6c9f81412e5da8aa3b67"},{"bytes":5120728,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/oxc/tasks.jsonl","sha256":"d6cc7c276115de12b88830ebe3673f706c6f50c4827188fa1205f9a9df2ea8dc"},{"bytes":2736778,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/oxc/world.json","sha256":"0853ef80ad822a0b5813a912bfe3f3f720e3f9bf6ed721d7aad6932a1ef93165"},{"bytes":3766,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/pulumi/BUILD_RECEIPT.json","sha256":"44c68a88e69910827163ad0a683ba4dd642c2fce1cc23a7ebf5f390b1dde2490"},{"bytes":10998,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/pulumi/contexts.jsonl","sha256":"464e8e1e2d9442da0466c8a6dae3ef61591db9502ee6f0f87451dc7c55b3446e"},{"bytes":2798250,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/pulumi/tasks.jsonl","sha256":"a219e5ee9f8403480c542439af44f13afacf9322e96d2b41c5dbaba1a5764a66"},{"bytes":1487743,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/pulumi/world.json","sha256":"87c3a1ba957161d15011d796d2ed47989e063c607b739ddbe0f76d4fe5224b8c"},{"bytes":4385,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/ruff/BUILD_RECEIPT.json","sha256":"5a7f0f31eb4115028eefe36439183dab5d4835b426dbd981210066f9638bb66d"},{"bytes":19422,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/ruff/contexts.jsonl","sha256":"f9cd95d6ec96a81edb5c240cb86d6692ee66ff81f2391d05d747602219453f26"},{"bytes":2687546,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/ruff/tasks.jsonl","sha256":"7b68704d2ee69f0588c7e66056fda048fb178b15578bebc0758415c72ac1db82"},{"bytes":2009767,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/ruff/world.json","sha256":"ad89b8c60338d8ad57419c5a7376ce7ffb8ad6353bbbcb9911722c4d7effec21"},{"bytes":3852,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/transformers/BUILD_RECEIPT.json","sha256":"51307df0624fde249ec3f677e05b33b66926b6a60ec99c9097706a9777ccd1de"},{"bytes":11057,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/transformers/contexts.jsonl","sha256":"50abbdf8476b6cbf25e45682de175c25a38b9f9109d3e860a39c789ad4784e0f"},{"bytes":1889219,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/transformers/tasks.jsonl","sha256":"f316d78c9645838c885c74e45b749d8c1089dd9eb2b6914d278768782cf53488"},{"bytes":1058172,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/transformers/world.json","sha256":"c9dbb3149825721d30c1702198474a107cf329b2880d18516903be7ff9392932"},{"bytes":3669,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/uv/BUILD_RECEIPT.json","sha256":"e120122624321643b3d2ccb4bd20a05f7d41c8939611bbfe7e3c2e44378f213f"},{"bytes":11010,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/uv/contexts.jsonl","sha256":"9237b27355ac1dc0ec9223850cc5fb046510e527a3471bc4a6ee3cc5c7908edb"},{"bytes":2173424,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/uv/tasks.jsonl","sha256":"1b8e0868b132c6dded710d8485381a29cc8318119f6a3c3afd641635fd34e086"},{"bytes":1396736,"path":"/WORKSPACE/longworld/data/candidates/p64_codeforge_taskbank_v2/uv/world.json","sha256":"52c48186dd2658709ce64a2f78db12b1959b27a0690e15d4a6dac0cd5e55de04"},{"bytes":6929,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/BUILD_RECEIPT.json","sha256":"300d01e98793c4e999a92a17b435ecad4c3b13b658e9247c745b13064069ae68"},{"bytes":271072,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/154e09d6e4c237af30cff5dadc2a891340432572b1d315096b57824d0d5b66d1.txt","sha256":"154e09d6e4c237af30cff5dadc2a891340432572b1d315096b57824d0d5b66d1"},{"bytes":279740,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/2dcedd0caf9cd18d9013ba85880580b22a3d8969b51e27a305f6eaf34191ca6e.txt","sha256":"2dcedd0caf9cd18d9013ba85880580b22a3d8969b51e27a305f6eaf34191ca6e"},{"bytes":257521,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/3a0996291839ebe9371cd550499923e30d9bca2086c7ea9acc6d9fd43fa6fb9f.txt","sha256":"3a0996291839ebe9371cd550499923e30d9bca2086c7ea9acc6d9fd43fa6fb9f"},{"bytes":270865,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/4eb5304fff1054c1f7aaf5bb03eef7e98d2bc493a8ae9e85c6ac9e5d0bf21a51.txt","sha256":"4eb5304fff1054c1f7aaf5bb03eef7e98d2bc493a8ae9e85c6ac9e5d0bf21a51"},{"bytes":771173,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/5cee07fd60d85988a7b983a72d4c569377370035c69c5d1d95ad270657cebfe9.txt","sha256":"5cee07fd60d85988a7b983a72d4c569377370035c69c5d1d95ad270657cebfe9"},{"bytes":289840,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/738d4240d33665554d603213ff613f18fba704e434675e62a2e6da772aa6708a.txt","sha256":"738d4240d33665554d603213ff613f18fba704e434675e62a2e6da772aa6708a"},{"bytes":289858,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/797ff6d8b623e8ce33edc43bcbb6743cf674ec7ef7f418065640e4dc9038abc1.txt","sha256":"797ff6d8b623e8ce33edc43bcbb6743cf674ec7ef7f418065640e4dc9038abc1"},{"bytes":290047,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/84db9f14e644fa1b78d035512438699341a2df3c32b6a2ff6c134d4b20a17e68.txt","sha256":"84db9f14e644fa1b78d035512438699341a2df3c32b6a2ff6c134d4b20a17e68"},{"bytes":267828,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/d5a456f882b7a581770ca204ce751402a7981eff5c9efaf3eec1848bc9ce2ca2.txt","sha256":"d5a456f882b7a581770ca204ce751402a7981eff5c9efaf3eec1848bc9ce2ca2"},{"bytes":252887,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/d6dec4f613fc2cea14de35b3e8b741d429edfcafb80e4fda78c588f29903f3da.txt","sha256":"d6dec4f613fc2cea14de35b3e8b741d429edfcafb80e4fda78c588f29903f3da"},{"bytes":260765,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/edd2b66ab221e24cb3e4e95fa92ca32d69fd4e8f9aed8ef006a55902bd898e90.txt","sha256":"edd2b66ab221e24cb3e4e95fa92ca32d69fd4e8f9aed8ef006a55902bd898e90"},{"bytes":790148,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/f37780863cbb857250dab5f98304bf4772588ae84962486aae13baf0cf3215db.txt","sha256":"f37780863cbb857250dab5f98304bf4772588ae84962486aae13baf0cf3215db"},{"bytes":808333,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/contexts/ffc2ed4cbddd229ca16aa1a49b787f14bd9309b70c0b4b7c11f13521cca40440.txt","sha256":"ffc2ed4cbddd229ca16aa1a49b787f14bd9309b70c0b4b7c11f13521cca40440"},{"bytes":96222,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/rejections.json","sha256":"984763fe3dfc7ae0911fa528fd497ba70d62e3d2bc08ac6e5d4e6f1cb15e7b78"},{"bytes":92662826,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/sft_candidates.jsonl","sha256":"c7ece52b99f3be549b6e4eecdae3b39cd4b5fd0ba24e2c7d27ad49139053603f"},{"bytes":1540383,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/alphabet/tasks.jsonl","sha256":"f4531a5325846bb25e2f81023d452c9e9c6bc462b7f6aace8f1020deda0a07be"},{"bytes":7037,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/BUILD_RECEIPT.json","sha256":"9a53e7e304af1a0a5edf9acfc60d57ef70d8f39ae49327257e2a2ed21494d71e"},{"bytes":861001,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/0632def39a5833143266a2ca455e0b00f5630113e958b555f2f586f092aa0dbf.txt","sha256":"0632def39a5833143266a2ca455e0b00f5630113e958b555f2f586f092aa0dbf"},{"bytes":352058,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/1e562c17a48d348b6d5ad1b0aba9e7e34d226a278351d97839650c7084851ab7.txt","sha256":"1e562c17a48d348b6d5ad1b0aba9e7e34d226a278351d97839650c7084851ab7"},{"bytes":353054,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/29f96fe64c8390eff076eea810b08661a2bbf23feebd44d6b30b64fd940fd93b.txt","sha256":"29f96fe64c8390eff076eea810b08661a2bbf23feebd44d6b30b64fd940fd93b"},{"bytes":261968,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/307b3490ffe95ad74d9b7619f2b18fab0c7d1919fd1d8c44317faf91fe9ded85.txt","sha256":"307b3490ffe95ad74d9b7619f2b18fab0c7d1919fd1d8c44317faf91fe9ded85"},{"bytes":345971,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/37919f816651d50bf7222787ce63c7212b7d98eaa896c70039a2873d34882fbe.txt","sha256":"37919f816651d50bf7222787ce63c7212b7d98eaa896c70039a2873d34882fbe"},{"bytes":746395,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/3cc05a1fc51020bbb9174f4ce2c9742777ace87064284ea49cc1e767a4ddcf94.txt","sha256":"3cc05a1fc51020bbb9174f4ce2c9742777ace87064284ea49cc1e767a4ddcf94"},{"bytes":343962,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/4dfdd7cafeed8282cc4c928b070b09e56c101115edbd47e41e3d78dc1c05f2f7.txt","sha256":"4dfdd7cafeed8282cc4c928b070b09e56c101115edbd47e41e3d78dc1c05f2f7"},{"bytes":328969,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/6653a3a7d73e650d07a12252afb82b485844e2fd45f89348ac2623fdfbe37661.txt","sha256":"6653a3a7d73e650d07a12252afb82b485844e2fd45f89348ac2623fdfbe37661"},{"bytes":785094,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/83ada2118cffbaeeafb5ab369dd42dce9b47a3a74b096ac4fd8d0acfb99bfdb9.txt","sha256":"83ada2118cffbaeeafb5ab369dd42dce9b47a3a74b096ac4fd8d0acfb99bfdb9"},{"bytes":337875,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/bf4d5242c8e976c042f0af4acc44166c6db621c1b7cd94b4b135177d3a38a5e1.txt","sha256":"bf4d5242c8e976c042f0af4acc44166c6db621c1b7cd94b4b135177d3a38a5e1"},{"bytes":231365,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/c3a2fd4c51be6fac062adad576bbddf2af272f961e8b6c95b319c36fef637988.txt","sha256":"c3a2fd4c51be6fac062adad576bbddf2af272f961e8b6c95b319c36fef637988"},{"bytes":270064,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/e34de302d303e045ca84eab9af36fe728b7469dbbe2ffa2f8e890dbcc4f3ebe3.txt","sha256":"e34de302d303e045ca84eab9af36fe728b7469dbbe2ffa2f8e890dbcc4f3ebe3"},{"bytes":253062,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/contexts/f6faf8116be02c41fb4dc317d45e422f369cc0ff535fff45b5160830bc5340d8.txt","sha256":"f6faf8116be02c41fb4dc317d45e422f369cc0ff535fff45b5160830bc5340d8"},{"bytes":111047,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/rejections.json","sha256":"cc3708c686967f1e8ef8ff6e4adbd447e7b6812fd23256bd6f3b18574111ce9c"},{"bytes":126386509,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/sft_candidates.jsonl","sha256":"22c9728ff464dafa7a51b2cb81af4649885c8cb29d5d5e43ef3573e7418a8a7d"},{"bytes":1959078,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amazon/tasks.jsonl","sha256":"2cd06e8daddfee36dcefbbbe512c447949fa767c3ae33ae575428fc748ce8991"},{"bytes":7129,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/BUILD_RECEIPT.json","sha256":"6192b9c4402a2a17180fe4fcc5c3914d42890a244d27defc45feb129b0017545"},{"bytes":384693,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/1cdb9111ecd6c4d21b949556278d36bdcde0a2087c6e1906e4bdd3d1d74a588b.txt","sha256":"1cdb9111ecd6c4d21b949556278d36bdcde0a2087c6e1906e4bdd3d1d74a588b"},{"bytes":1157698,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/43fec308edae3c6a728e567560ddca3946fd435d25b432d24409ba07c4d6ad05.txt","sha256":"43fec308edae3c6a728e567560ddca3946fd435d25b432d24409ba07c4d6ad05"},{"bytes":1171076,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/48222c28b5835cd14ef2be9e6167c4430b1998f2f91f0a20b260e0f6e90bc9b3.txt","sha256":"48222c28b5835cd14ef2be9e6167c4430b1998f2f91f0a20b260e0f6e90bc9b3"},{"bytes":362046,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/8f4b08522043e0e27a94b7636f7ac8c053e84ddc509a331619af973644d2d403.txt","sha256":"8f4b08522043e0e27a94b7636f7ac8c053e84ddc509a331619af973644d2d403"},{"bytes":815705,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/9e16fa165616150878dfac4805ebb2afb73162f2656d11460f0ebdcfdc16d254.txt","sha256":"9e16fa165616150878dfac4805ebb2afb73162f2656d11460f0ebdcfdc16d254"},{"bytes":410631,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/a5497b5cd96887fb3f2d3fb486e044bdbd5381a1856f465c2c868b067bbef80e.txt","sha256":"a5497b5cd96887fb3f2d3fb486e044bdbd5381a1856f465c2c868b067bbef80e"},{"bytes":411134,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/d8b27164be72bdc8aba4ed271004543c58bae6e66e9fbc16228d1b93c2ec7abb.txt","sha256":"d8b27164be72bdc8aba4ed271004543c58bae6e66e9fbc16228d1b93c2ec7abb"},{"bytes":410519,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/dbfbe018c9ab9e8655fe882d853fcd91bf5143322e100cae6da0de2db5dcc705.txt","sha256":"dbfbe018c9ab9e8655fe882d853fcd91bf5143322e100cae6da0de2db5dcc705"},{"bytes":404459,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/dc4c37f3ed30da834bc73e4a0cc6b88b42e0cec01e9754b91748ed073a735fe9.txt","sha256":"dc4c37f3ed30da834bc73e4a0cc6b88b42e0cec01e9754b91748ed073a735fe9"},{"bytes":411246,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/f1ba4c1acbf0c8f8d35705c6d6f65ce05dfbc9b696e87076ad6d4fe9e7464b5a.txt","sha256":"f1ba4c1acbf0c8f8d35705c6d6f65ce05dfbc9b696e87076ad6d4fe9e7464b5a"},{"bytes":410959,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/f4d0291a1c134ad466300db27612e42d7d30d5de08842516c5312b20663d0685.txt","sha256":"f4d0291a1c134ad466300db27612e42d7d30d5de08842516c5312b20663d0685"},{"bytes":404571,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/fc5e876f21ee068dbdcf9f6496574fe5f36e9dbee7cef0d69bfd3e59add6be54.txt","sha256":"fc5e876f21ee068dbdcf9f6496574fe5f36e9dbee7cef0d69bfd3e59add6be54"},{"bytes":390753,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/contexts/fcb4719410a5e317c51dcfb9a2d9c580de812e90af0c719d702833320b0d932a.txt","sha256":"fcb4719410a5e317c51dcfb9a2d9c580de812e90af0c719d702833320b0d932a"},{"bytes":81035,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/rejections.json","sha256":"dc1cde31ff2d5486b01acc2e652a5c6da563cd06fc12982c6a8e3c34a5a34f00"},{"bytes":126403215,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/sft_candidates.jsonl","sha256":"5d66d8f0b325677cdfff2a3abeaf5f482c5ee0a710167515788edc67676588cb"},{"bytes":1584823,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/amd/tasks.jsonl","sha256":"846c5021c75d14c50b09abe98b1e2a4d5ecca88633861e0b8c5a9a5c035b4c3f"},{"bytes":7062,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/BUILD_RECEIPT.json","sha256":"bd7b40ff31a68d310d5c30fb2912fcd26ede7278bc486bf4357a3e79a631506e"},{"bytes":490417,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/0fe5fcf882ee39eb26140f7ace258ef73721bf6ba8912aba4a99fef982539206.txt","sha256":"0fe5fcf882ee39eb26140f7ace258ef73721bf6ba8912aba4a99fef982539206"},{"bytes":489283,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/18e7415f0fee766c56acf1401f2b5f49ebbb2abb3f74cf4a370cf555a27b8012.txt","sha256":"18e7415f0fee766c56acf1401f2b5f49ebbb2abb3f74cf4a370cf555a27b8012"},{"bytes":987942,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/2d778c955a4e36f37279a36231fc48df29b385dbc92492e0ca3b44b7095b949b.txt","sha256":"2d778c955a4e36f37279a36231fc48df29b385dbc92492e0ca3b44b7095b949b"},{"bytes":483780,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/47f2c3b156f36aa907f6cc64ff90c465fe8f9c656d036f77edc8efd089efd062.txt","sha256":"47f2c3b156f36aa907f6cc64ff90c465fe8f9c656d036f77edc8efd089efd062"},{"bytes":492022,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/509090fd119d65b0f2d53cd76cb6869790db986d13bf15e31ab4deb51c759515.txt","sha256":"509090fd119d65b0f2d53cd76cb6869790db986d13bf15e31ab4deb51c759515"},{"bytes":495920,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/712034bb82f54d01c07a8e9a5c8f78ce22dd934fe6060c9a7c62222e66fa5cc2.txt","sha256":"712034bb82f54d01c07a8e9a5c8f78ce22dd934fe6060c9a7c62222e66fa5cc2"},{"bytes":498980,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/7160a8761039df6462802461b899ae7674cf56e4e5e08d0c1d4a304fb5e1a7c4.txt","sha256":"7160a8761039df6462802461b899ae7674cf56e4e5e08d0c1d4a304fb5e1a7c4"},{"bytes":466483,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/7f28ebf52c2d18efba81b01c8e30e7c283b152a5b068803892004ecf20f62bf7.txt","sha256":"7f28ebf52c2d18efba81b01c8e30e7c283b152a5b068803892004ecf20f62bf7"},{"bytes":989397,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/98c243f24e02183e752ef9642aadebc20a85a027a8bbb8f3f751038e94f25610.txt","sha256":"98c243f24e02183e752ef9642aadebc20a85a027a8bbb8f3f751038e94f25610"},{"bytes":490738,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/a4b989117d67e9b31a95c5b05e8138bca39f7411c147ced417ecc546e4bcf786.txt","sha256":"a4b989117d67e9b31a95c5b05e8138bca39f7411c147ced417ecc546e4bcf786"},{"bytes":498659,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/b9a9eb7ab86d14791695238b3dd57e753e3f7751dc6dc1be311d35d01872599a.txt","sha256":"b9a9eb7ab86d14791695238b3dd57e753e3f7751dc6dc1be311d35d01872599a"},{"bytes":982760,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/baac012fdaf754957b30260b81b1193ee13c04cdffffd3e63e53299f16a42019.txt","sha256":"baac012fdaf754957b30260b81b1193ee13c04cdffffd3e63e53299f16a42019"},{"bytes":499050,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/contexts/cb4f17d19453df18acf28ceb619378d8ddfd5bd8ffe3abca1254a96a0a7f270f.txt","sha256":"cb4f17d19453df18acf28ceb619378d8ddfd5bd8ffe3abca1254a96a0a7f270f"},{"bytes":78246,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/rejections.json","sha256":"32d50aab04db9523d7e83ce10a1078bdf0224ecee924f1ecf23755c7b0895c1a"},{"bytes":138386995,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/sft_candidates.jsonl","sha256":"b9fb2b4b55a87a76cde6897821f1ed241a326bd222755ca7f5141268a0c1d40b"},{"bytes":1614376,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/intel/tasks.jsonl","sha256":"624b196f793cac0c72812582431d65ea487a3d838deed2727f0c691fcc976536"},{"bytes":6924,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/BUILD_RECEIPT.json","sha256":"446c344f98c522d33ef846234fca9f6f440014631ede8f17f98a73f66f3d0044"},{"bytes":281863,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/1409620bb96634a441fa0d2ff5b852ffdf2c59a1e75593062d5ef8f30ccad641.txt","sha256":"1409620bb96634a441fa0d2ff5b852ffdf2c59a1e75593062d5ef8f30ccad641"},{"bytes":249039,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/1f0b6d7e72196d8ee02d590ecd5d7f6383f7495597d6a0abff2f703fa5e09c28.txt","sha256":"1f0b6d7e72196d8ee02d590ecd5d7f6383f7495597d6a0abff2f703fa5e09c28"},{"bytes":281077,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/24239ce1eb422cee8cb5a0adf2538e8df300308df664faaeabd53d183be384d1.txt","sha256":"24239ce1eb422cee8cb5a0adf2538e8df300308df664faaeabd53d183be384d1"},{"bytes":238191,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/2d2965cbc54c07b1eb0df3f4ce7ae42aa83724a1bfed415eca0e0a66c9ab7b1e.txt","sha256":"2d2965cbc54c07b1eb0df3f4ce7ae42aa83724a1bfed415eca0e0a66c9ab7b1e"},{"bytes":237035,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/71c8a3f488196eedbd962fa000a6e563304a0b460b3cb786a826127c475f7c16.txt","sha256":"71c8a3f488196eedbd962fa000a6e563304a0b460b3cb786a826127c475f7c16"},{"bytes":724096,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/8cb6cedeae2ae5552c03623d3a40a05ecaff6bcb00cf7eb56a1bce78c7d93bd2.txt","sha256":"8cb6cedeae2ae5552c03623d3a40a05ecaff6bcb00cf7eb56a1bce78c7d93bd2"},{"bytes":755764,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/8cbf1d6546470e661a71850c540bda6a053681dc6e7dd810c43be8dfc867c60b.txt","sha256":"8cbf1d6546470e661a71850c540bda6a053681dc6e7dd810c43be8dfc867c60b"},{"bytes":269859,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/aba6f83d394ff463cec02caec6bae25c8bde103180b65a6a3b9c55b2fd3aff3a.txt","sha256":"aba6f83d394ff463cec02caec6bae25c8bde103180b65a6a3b9c55b2fd3aff3a"},{"bytes":260088,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/bd5cadf836f8ed39e0f3200cce7a5a78dfc8c36838abd22732d020e22d9016a7.txt","sha256":"bd5cadf836f8ed39e0f3200cce7a5a78dfc8c36838abd22732d020e22d9016a7"},{"bytes":767768,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/c23de6700996101b559a6ef80a3c31e04a7ff77252cd8d9db0dd1861a2548ddd.txt","sha256":"c23de6700996101b559a6ef80a3c31e04a7ff77252cd8d9db0dd1861a2548ddd"},{"bytes":260874,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/e53be3c61f6a80023b896d1701777c8259071eb015394442ce2857a512b81321.txt","sha256":"e53be3c61f6a80023b896d1701777c8259071eb015394442ce2857a512b81321"},{"bytes":280526,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/ec48cdae156bed15aaa756198b498862dca8090673cc548850fc4c1b0fe6630b.txt","sha256":"ec48cdae156bed15aaa756198b498862dca8090673cc548850fc4c1b0fe6630b"},{"bytes":248870,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/contexts/f10e67788aa3ccd4963f03508eed256453e645da412222f685cd5b2ca13314ba.txt","sha256":"f10e67788aa3ccd4963f03508eed256453e645da412222f685cd5b2ca13314ba"},{"bytes":89648,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/rejections.json","sha256":"1e6e1179096b06ae4ce1d2e5f995cc7452885dce5342d5afa501b2b7a389f2cc"},{"bytes":96712765,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/sft_candidates.jsonl","sha256":"09d85cc147784cbd2f294f49e7c6e60be76b0f23d1f3aff847475b724abc5f5f"},{"bytes":1687642,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/meta/tasks.jsonl","sha256":"3add7cb0d9f78fd223bcf5c61fe34e490e654983f36b6711d8b0a4334e725df3"},{"bytes":7031,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/BUILD_RECEIPT.json","sha256":"e6e67e6c31936812f5f05eb221b2693795f329ee1dfd65293ed0f872b6cbe15e"},{"bytes":229553,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/02b831ba0c9ef57a27e79a8dc553009e9c03c3f3fb43de386d2d3c11492a0761.txt","sha256":"02b831ba0c9ef57a27e79a8dc553009e9c03c3f3fb43de386d2d3c11492a0761"},{"bytes":217796,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/0531bffdeefcda1e6fe8e61633c0ada4a318f4023844d2da634d6c7471f984fb.txt","sha256":"0531bffdeefcda1e6fe8e61633c0ada4a318f4023844d2da634d6c7471f984fb"},{"bytes":238072,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/12112d1f10abd53a2a33c661077e248edc610b1b46b1bf104aafccebe507b2b1.txt","sha256":"12112d1f10abd53a2a33c661077e248edc610b1b46b1bf104aafccebe507b2b1"},{"bytes":226488,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/15368484810fa291fc8b27d8645bd1dc477db50839ca1f58d5a5386530bd4118.txt","sha256":"15368484810fa291fc8b27d8645bd1dc477db50839ca1f58d5a5386530bd4118"},{"bytes":657225,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/49de1d040b5b40c2cc560237d6802556cbe73fb842f891c7f8af25d6c0ec4b2d.txt","sha256":"49de1d040b5b40c2cc560237d6802556cbe73fb842f891c7f8af25d6c0ec4b2d"},{"bytes":238245,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/515d1a6bef4877d8ce4f1d800e8856336aca2d54f61a8a7107f416d4b9b9d421.txt","sha256":"515d1a6bef4877d8ce4f1d800e8856336aca2d54f61a8a7107f416d4b9b9d421"},{"bytes":249784,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/54873b601347cf5c1955dc1633fe646e10e20e5fd9f05f96c3749c92517e06ac.txt","sha256":"54873b601347cf5c1955dc1633fe646e10e20e5fd9f05f96c3749c92517e06ac"},{"bytes":212941,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/59215993c1c553bd70fd6eaae6b259cc0cbff0276ed8fc9adf7d1fff5484d039.txt","sha256":"59215993c1c553bd70fd6eaae6b259cc0cbff0276ed8fc9adf7d1fff5484d039"},{"bytes":249624,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/79e6077675db888e504aea5a0896f97c59393812cbaeb2bc61e9f596f584ca96.txt","sha256":"79e6077675db888e504aea5a0896f97c59393812cbaeb2bc61e9f596f584ca96"},{"bytes":694241,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/93358faeff4273f33b19d59cea9f7ccb8568b90f09f78dd96f16f3fca1359ca8.txt","sha256":"93358faeff4273f33b19d59cea9f7ccb8568b90f09f78dd96f16f3fca1359ca8"},{"bytes":249957,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/a80ede423100486eeb7e153db6da086d6f7da95e8f61b5fc73e83459165113fa.txt","sha256":"a80ede423100486eeb7e153db6da086d6f7da95e8f61b5fc73e83459165113fa"},{"bytes":677629,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/b468dfca7ad2b2f3590418c3a82a4105b2b7be50d9ac95408f9468a0d32e9c05.txt","sha256":"b468dfca7ad2b2f3590418c3a82a4105b2b7be50d9ac95408f9468a0d32e9c05"},{"bytes":238200,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/contexts/bffecd8eb4e5fd80c472b76f82433a37cbe9212faef375241b1ae9979a45d462.txt","sha256":"bffecd8eb4e5fd80c472b76f82433a37cbe9212faef375241b1ae9979a45d462"},{"bytes":78402,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/rejections.json","sha256":"70bbdcda4a29c9ff42b114edcf2fec422d83c94566b594b3c1cd67ef88b8f7ff"},{"bytes":107464338,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/sft_candidates.jsonl","sha256":"a3c806ce050b0f1451cc3bad5d56467f1d10445ec5b95ba925e63d471390bc4f"},{"bytes":1967950,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/micron/tasks.jsonl","sha256":"73a9b3fb0b4acc3f30358729cb964d4c5b7329c1df829101c685df2dfcc3bce8"},{"bytes":7094,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/BUILD_RECEIPT.json","sha256":"ecfff3114427f2cd7ebafa26e3d181a0f24fd12424969185a088d0241edefd9f"},{"bytes":605780,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/23e46e6920784fbfa766fee79a8506644ec3f61ac570acb5abbab871e7e96a1a.txt","sha256":"23e46e6920784fbfa766fee79a8506644ec3f61ac570acb5abbab871e7e96a1a"},{"bytes":417350,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/36e5a18b2c5c860221f41e2572cd51891b5cf9b33e0e52d77ba213d98fa37cdc.txt","sha256":"36e5a18b2c5c860221f41e2572cd51891b5cf9b33e0e52d77ba213d98fa37cdc"},{"bytes":213731,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/4bfc587827a33b7154a97e36cabb668e6efb1f1ac947ecf793067e33ca1e8bf9.txt","sha256":"4bfc587827a33b7154a97e36cabb668e6efb1f1ac947ecf793067e33ca1e8bf9"},{"bytes":620310,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/4e030a8a451fc6c43286f9909a15ca845c451f3983df1eb4aadca70cc1c7b4c0.txt","sha256":"4e030a8a451fc6c43286f9909a15ca845c451f3983df1eb4aadca70cc1c7b4c0"},{"bytes":412932,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/6d8e60958839292df2c6cb5bce79848a9ebb6a277236a7392bd761d6a2e8be23.txt","sha256":"6d8e60958839292df2c6cb5bce79848a9ebb6a277236a7392bd761d6a2e8be23"},{"bytes":202960,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/72b2e4eac6de9360c73ab106eae565072e2484e8ce7cba64a18de61ed0b3dbd8.txt","sha256":"72b2e4eac6de9360c73ab106eae565072e2484e8ce7cba64a18de61ed0b3dbd8"},{"bytes":199201,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/7bfee258f50525e164d22f046596adfb2736b9f8b5f44d23410df729755bfdd6.txt","sha256":"7bfee258f50525e164d22f046596adfb2736b9f8b5f44d23410df729755bfdd6"},{"bytes":402820,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/7ffd1e559dd278614b516ab27fb12a4855f75f8950f97e559559ea90f4e93e42.txt","sha256":"7ffd1e559dd278614b516ab27fb12a4855f75f8950f97e559559ea90f4e93e42"},{"bytes":819511,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/85cbb0fa0b3c5ffcb37025f6e68ece45fda913f615d6bb8f098e3088312da0aa.txt","sha256":"85cbb0fa0b3c5ffcb37025f6e68ece45fda913f615d6bb8f098e3088312da0aa"},{"bytes":406579,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/b033f71f145509f44854677befd31abacca41e15aef8788c1952c46263506f14.txt","sha256":"b033f71f145509f44854677befd31abacca41e15aef8788c1952c46263506f14"},{"bytes":203619,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/b6d8e0735daa89d6c015f8c8b78b182d123307a22cf8e07ab925385512fd7d32.txt","sha256":"b6d8e0735daa89d6c015f8c8b78b182d123307a22cf8e07ab925385512fd7d32"},{"bytes":416691,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/bfb3958beb465efd44eabd711783f8d3a718e1d1c366addf13fa6a160ac45a78.txt","sha256":"bfb3958beb465efd44eabd711783f8d3a718e1d1c366addf13fa6a160ac45a78"},{"bytes":402161,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/contexts/d65b3e28c82c6c98647ea7ad6364387d87d2f739f4a22dc952a36d80c53048e1.txt","sha256":"d65b3e28c82c6c98647ea7ad6364387d87d2f739f4a22dc952a36d80c53048e1"},{"bytes":82256,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/rejections.json","sha256":"52d42c3f01a9f20f8d93f4c2c57aeb89ec6dae57817d7409217cfc2d1399f5b2"},{"bytes":91476117,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/sft_candidates.jsonl","sha256":"2815791f1d4071ced06ca2994c43b2d8163a95aa4f2516c23fbf0370a4efed5c"},{"bytes":1490209,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/nvidia/tasks.jsonl","sha256":"ee97904cb08c893e64e00f1e787ba648de154d25bcacdba30b8c39775caa950c"},{"bytes":7103,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/BUILD_RECEIPT.json","sha256":"ee7f9fa1669da97ab829c264d5f0a717ae4c2ec026cb5b4e0c5760214320f86c"},{"bytes":878662,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/1b2db7db658cebc9bca047e9439fed92af6862fc21582df0ff3b1dd3fb96fd26.txt","sha256":"1b2db7db658cebc9bca047e9439fed92af6862fc21582df0ff3b1dd3fb96fd26"},{"bytes":468887,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/408f77f0a8ccb940cf6b014244cae03ce0f56e07427ab0f903e124c6f5cc14bc.txt","sha256":"408f77f0a8ccb940cf6b014244cae03ce0f56e07427ab0f903e124c6f5cc14bc"},{"bytes":881408,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/469b7500510a65c3656953bb3c8becfca18f9c0f92440d139c07460e4b447369.txt","sha256":"469b7500510a65c3656953bb3c8becfca18f9c0f92440d139c07460e4b447369"},{"bytes":404934,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/49711ee0105f6f24e2c7c8532adbddfa4d454fef910b9ea88c18d5c451843373.txt","sha256":"49711ee0105f6f24e2c7c8532adbddfa4d454fef910b9ea88c18d5c451843373"},{"bytes":402188,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/5171071832cdd87ce91d13ab685305b81f4b3081233582ab53f475315940c517.txt","sha256":"5171071832cdd87ce91d13ab685305b81f4b3081233582ab53f475315940c517"},{"bytes":476477,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/59aadaa31d2d07ebed807d1bac9daae57737c2dcc3c22651b72f005ebc7099a6.txt","sha256":"59aadaa31d2d07ebed807d1bac9daae57737c2dcc3c22651b72f005ebc7099a6"},{"bytes":405039,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/752a9701280c2e1e62684a1619dc749d84ac39de065fdf2e378d12e97480fd0a.txt","sha256":"752a9701280c2e1e62684a1619dc749d84ac39de065fdf2e378d12e97480fd0a"},{"bytes":397452,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/77927889247d6ceac48c8753a7b5e22130c688b7ad411d2811abb28833f42f39.txt","sha256":"77927889247d6ceac48c8753a7b5e22130c688b7ad411d2811abb28833f42f39"},{"bytes":409775,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/8f7dd7c369e937c93b4e4cb765cb2b76b3f98311f286adf595d3079d0016f32c.txt","sha256":"8f7dd7c369e937c93b4e4cb765cb2b76b3f98311f286adf595d3079d0016f32c"},{"bytes":476369,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/9d57886c8a8f16bbad1e6e16375a50ec013b50237ecc3f3c2d255732bc9dd144.txt","sha256":"9d57886c8a8f16bbad1e6e16375a50ec013b50237ecc3f3c2d255732bc9dd144"},{"bytes":807227,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/c8f905a15498d964d029665db896185101dce35e1deaa703b89b64cd0cb0dc3d.txt","sha256":"c8f905a15498d964d029665db896185101dce35e1deaa703b89b64cd0cb0dc3d"},{"bytes":391887,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/e1049abdffa19f2ca2e40b5a4d3c2d551431986edb21332fd4ceea970bc7deb8.txt","sha256":"e1049abdffa19f2ca2e40b5a4d3c2d551431986edb21332fd4ceea970bc7deb8"},{"bytes":476474,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/contexts/fc46eb798e8980db5151ca03695ca388dee18ec24e07d8986d8341a649c38f47.txt","sha256":"fc46eb798e8980db5151ca03695ca388dee18ec24e07d8986d8341a649c38f47"},{"bytes":82202,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/rejections.json","sha256":"32148c14592f85d9655be54719abf2759662eaafaf902593b22dfb3a0b49e702"},{"bytes":114173889,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/sft_candidates.jsonl","sha256":"1d238241417fefc49691316a296fa29a8fd27cef80a3e03baeb66a33538e4bc2"},{"bytes":1566613,"path":"/WORKSPACE/longworld/data/candidates/p64_finance_taskbank_v2/siriusxm/tasks.jsonl","sha256":"7da706f484ecb216d3ba6a3328e320dc77274647f50561ba63537db137049ad3"},{"bytes":30210275,"path":"/WORKSPACE/longworld/data/finance_sources/p64_intel_generic_v1/source_manifest.json","sha256":"10a717b252973dc50e5578c6964401f23c8b355a5a2dae7af83b2dff15cea2f9"},{"bytes":23551308,"path":"/WORKSPACE/longworld/data/finance_sources/p64_siriusxm_generic_v1/source_manifest.json","sha256":"8685060409f294a173fa176b6031fbab21a39ad4cf01cfe6640731792fe2e8c2"},{"bytes":178764,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_dprint_workspace_v2/dprint_pr1174_support.json","sha256":"7dd352d25c0d5fdd60b0026779a166dcd63882eac99287b5899206dd338cc1ca"},{"bytes":82699,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_dprint_workspace_v2/dprint_pr1177_support.json","sha256":"91b9e60ccad5c1426b5cf5b89023686a6ceccd7a9979d0f271d217fca87bad30"},{"bytes":210845,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_dprint_workspace_v2/dprint_pr1178_support.json","sha256":"78b9afab7739bd1eed55e288a98117001bc81b5149ac46df21920a3cd2d1c3f5"},{"bytes":161258,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_dprint_workspace_v2/dprint_pr1185_0.55.0.json","sha256":"5c875f540162289e75bd7450cc2b42878b4d4fa8e275c103ce651acc480e1d92"},{"bytes":175775,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_dprint_workspace_v2/dprint_pr1203_0.55.2.json","sha256":"9b84d84832a50d10c32941d2a438cb32a4bfce48d82c4f9e58527e0e83f338bd"},{"bytes":331726,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_dprint_workspace_v2/dprint_pr1207_0.56.0.json","sha256":"513eaa48fc34956f2d042db2a68968fbe5632ac5346a92cbc1c214298c5ebeae"},{"bytes":139619,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr24360_support.json","sha256":"750dac05c87818eb8329b58941f3ebcaa28ae89d25a02709c6eae3d9ae7284bf"},{"bytes":52615,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr25403_crates_v0.147.0.json","sha256":"1867e896dbbc2f5596ac0cedc0535010a4618214d355e1b58db9c2f26fcb613e"},{"bytes":217251,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr25585_support.json","sha256":"93edd5fc9dec0c28750965ac4fedb98c8bf3ebed4595765318f3a9b1e0d7ec14"},{"bytes":94271,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr25909_crates_v0.146.0.json","sha256":"f35a0d4f4f9cbf37f02ba5f0ea9a7e36a9ca2db507366868fb9388cc7298bd37"},{"bytes":128493,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr25930_support.json","sha256":"300a3a21c4facc21caa025f8cbb30081d6ba3f8d68bc4c3efbae6dca7adf9ee9"},{"bytes":89841,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr25944_support.json","sha256":"3f37657167ce5f4b6e89864ec8b7d4b58c038bc7f9f8038bb2a3be46f4c60348"},{"bytes":91550,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr25958_crates_v0.147.0.json","sha256":"4b3aaf966efcfb30de7408a8b964f912b61be312f187ddab9c665f0b7a4be334"},{"bytes":122856,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr25975_crates_v0.147.0.json","sha256":"5cd9bff4cf3ba1c2ad405ec726ceca23c32558fb18c0f09e77983ed99a0d1133"},{"bytes":147665,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr25986_support.json","sha256":"28ebdf55706759c6c237e0e44aec20f77f9ac553e6645e3b0087115d28046fa5"},{"bytes":88792,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr26021_support.json","sha256":"f8fa9c100cc16c792297d977477b11e337e879d5e65def91643e3ecbeba59104"},{"bytes":63183,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr26047_support.json","sha256":"c77fbd33dcb0894fec18b9b16e7dcc249c126657bcb2089c7b4aa4fa518f3c7f"},{"bytes":46515,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr26050_support.json","sha256":"94c26c0ef1fd868dcc408b8e9a5b19fb0ad94f8d214cb4767a338156764d38eb"},{"bytes":71839,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr26081_support.json","sha256":"48baa0543cf0f9447af1b69f71d6c823219cc13bfed241ef4ffc1e4f5e3cc627"},{"bytes":121927,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr26084_support.json","sha256":"308a1891884c020c6a0e69ecb5fa891e21d0e30e769205bc573216d5bad2a98c"},{"bytes":113354,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr26099_support.json","sha256":"589031e3a9349af1a44d6d8d6684082b392219e7486799c84d73b678584ad021"},{"bytes":151769,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr26128_support.json","sha256":"1a66e96094f0c238d46e0e64d3717343b7935dcb996ac496a81ad8e24e2b4b32"},{"bytes":194610,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr26142_support.json","sha256":"b20a925048daddf27b5767ca983fb19535d153901220f81695dbccddc0e3770e"},{"bytes":93474,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr26144_support.json","sha256":"4fbacb2266abe4e2eeb6a6b5c6c672cce9b815b69e3274a5ae78694ad3d42cf0"},{"bytes":78446,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr26169_support.json","sha256":"31595b27ba1a2ed9efa2b489bd4c117cfd6d729aa0bd2ae4f85d97ff1124fc84"},{"bytes":181078,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_oxc_first_workspace_v2/oxc_pr26172_crates_v0.148.0.json","sha256":"3266f218512c83c46e53d09e7b6b80dee8da605fe1dfeaac19cbcdd29ad43223"},{"bytes":284119,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_ruff_first_workspace_v1/ruff_pr27190_support.json","sha256":"b23ff3a99707b8de1b3870d33ba7e3f02a98059be0276ab6a043e5c51b085342"},{"bytes":75280,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_ruff_first_workspace_v1/ruff_pr27626_0.16.3.json","sha256":"de1e8c619d08ab1064262a6c6762cd9e847ed3130a46611dd4da5dcc6ca9707a"},{"bytes":405139,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_ruff_first_workspace_v1/ruff_pr27666_support.json","sha256":"e52354bf60fc27ee2d409be7463fed8b207f97544f6ddb10427d7f854e765c21"},{"bytes":116422,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_ruff_first_workspace_v1/ruff_pr27687_support.json","sha256":"f9f7470a1559b8bcc6e7cb2248fb29bec5836fa969691671762062156f1bed4e"},{"bytes":103273,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_ruff_first_workspace_v1/ruff_pr27835_0.16.4.json","sha256":"9369eb774f40955e7036a1564e53b90460c674a513c3127eeb1b254b54804042"},{"bytes":229514,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_ruff_first_workspace_v1/ruff_pr27877_0.16.5.json","sha256":"ef0925074421dffa95095b3a1c9ed636745a4eb3cdc7f1713d4f2d98852028fb"},{"bytes":113640,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_ruff_first_workspace_v1/ruff_pr27906_support.json","sha256":"46d0d401ef44660a3b91971c165edfae114355b9e800655badecf2b6ea281492"},{"bytes":159163,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_ruff_first_workspace_v1/ruff_pr27910_support.json","sha256":"71e51a03604049f3ce7e0de6a33b62cb1ad10544c708bd1ac09d2791d5f66441"},{"bytes":67243,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_ruff_first_workspace_v1/ruff_pr27924_support.json","sha256":"b84f6dd98c3483b9668c7cadeb78aeea1af50cdbf5fdff8d2c588549db2bdbda"},{"bytes":82491,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_ruff_first_workspace_v1/ruff_pr28000_support.json","sha256":"c64e38c198e64f1dede167cb79eef3787b4f3978e7942573973fefeaff9e7b23"},{"bytes":66283,"path":"/WORKSPACE/longworld/data/source_inventory/p13_codeforge_ruff_first_workspace_v1/ruff_pr28023_support.json","sha256":"0df3df5eba3e8cd02fc6ccfb5b50461afff95d7a5a54ab75016341a1e59f0705"},{"bytes":14762056,"path":"/WORKSPACE/longworld/data/source_inventory/p57_finance_amazon_ir_fy2021_2024_v1/issuer_ir_manifest.signed.json","sha256":"8e48f6c0688ae3bcbd9b9e9eade195e8ba70e297f7bf806573e2d95c90645bb6"},{"bytes":13896575,"path":"/WORKSPACE/longworld/data/source_inventory/p58_finance_meta_reconstruction_v1/issuer_ir_manifest.signed.json","sha256":"7871f9310cc39bb94a465c7f960b538ebbf53ae752ecc1b439e641cebb4f2b5d"},{"bytes":16381734,"path":"/WORKSPACE/longworld/data/source_inventory/p59_finance_alphabet_reconstruction_v1/issuer_ir_manifest.signed.json","sha256":"4028a54d58dca726657487ee1b4238e3213c3c0b59c5044cb2e1cc50faf31e3c"},{"bytes":16760485,"path":"/WORKSPACE/longworld/data/source_inventory/p59_finance_micron_reconstruction_v1/issuer_ir_manifest.signed.json","sha256":"d85ebdb650cb37a9b7e7d0fddc783cf167e4d1b7c5f38109b80236d843cd15cd"},{"bytes":13897529,"path":"/WORKSPACE/longworld/data/source_inventory/p59_finance_nvidia_reconstruction_v1/issuer_ir_manifest.signed.json","sha256":"5bb796cc5d1d69fee06bf7ecdde7b86d165738c40f9f688a72aaf7cd0070b793"},{"bytes":19060553,"path":"/WORKSPACE/longworld/data/source_inventory/p60_finance_amd_inline_fy2021_2024_v1/issuer_inline_manifest.signed.json","sha256":"4b8756fa1819883bbb706fb520516a64c17516cf1c44dbcfe018a5ebdfd67258"},{"bytes":652679,"path":"/WORKSPACE/longworld/data/source_inventory/p64_codeforge_opensearch_workflows_v1/opensearch_php_pr398.json","sha256":"bf0379a144e795497ad82040146be9bf6a26b65f7f458f25c821b922cb89a91d"},{"bytes":265306,"path":"/WORKSPACE/longworld/data/source_inventory/p64_codeforge_opensearch_workflows_v1/opensearch_php_pr399.json","sha256":"75eb2dc3817b0aa95045b555d152b85e589366e2f708ca0606209d6bb4ece97d"},{"bytes":42117,"path":"/WORKSPACE/longworld/data/source_inventory/p64_codeforge_opensearch_workflows_v1/opensearch_php_pr403.json","sha256":"cfe4a4572a519c445a975dda8a187501b637d25c1793a24c7c7ae3a9e6fecfbb"},{"bytes":42941,"path":"/WORKSPACE/longworld/data/source_inventory/p64_codeforge_opensearch_workflows_v1/opensearch_php_pr405.json","sha256":"42c342cb6308e2da5b50c4236b7bc8f78d28b1d6c12c8b68fe4a029b4c73006b"},{"bytes":136730,"path":"/WORKSPACE/longworld/data/source_inventory/p64_codeforge_opensearch_workflows_v1/opensearch_php_pr421.json","sha256":"4095941246cbf4b56f2abd2f8699d74a132ebc6b143baa33373e596324e216d2"},{"bytes":49606,"path":"/WORKSPACE/longworld/data/source_inventory/p64_codeforge_opensearch_workflows_v1/opensearch_php_pr423.json","sha256":"2ba4e443307f70ab26bd32d571bca661c7bb94c1f9d5f44bb5333c8e494f0278"}],"local_training_eligible":true,"manifest_path":"P64_TRAINING_MANIFEST.json","max_full_message_tokens":261954,"outputs":[{"bytes":653089550,"path":"finance_train.jsonl","sha256":"18cce1a697eeacd468b1a1ac278c42734a91711894ec65c31cadd79bb170da57"},{"bytes":240577104,"path":"finance_eval.jsonl","sha256":"3ee5d59d48d6f87c032b1afb8ded4f9b255aa6be12cf51d79eb54d32b6875b89"},{"bytes":242734380,"path":"codeforge_train.jsonl","sha256":"e78852013aaa49ea54e65e0d475d5b8e0c61dc1e096ddcfcd4a28ccb3cd7cbae"},{"bytes":56658476,"path":"codeforge_eval.jsonl","sha256":"b6705df17f1c099084dcb2d0a56a3e29376ac80fb4a83eefa4a7c9a81f9d7bee"},{"bytes":796305,"path":"sample_index.jsonl","sha256":"1e767e5f1fe489887db917615aa071155e33aa3859909ce9a5563a551f9debec"},{"bytes":850,"path":"dataset_info.json","sha256":"c41d28e54498091aaf1a8d3f8931a27c298923e84800ed53afb0e5b8315ed301"},{"bytes":387,"path":"TASKBANK_P64.yaml","sha256":"fee14c63bc6396991ac6b145ca7837c563c592070b832443a2a9490d05380460"}],"production_eligible":false,"profile_id":"p64-finance-codeforge-primary-local-v1","samples":2486,"schema_version":"longworld.p64-local-training-manifest.v1","source_validation":[{"capacity_bins":{"131072":139,"262144":61},"domain":"finance","exact_token_ranges":{},"name":"alphabet","production_eligible":false,"semantic_tasks":200,"status":"PASS"},{"capacity_bins":{"131072":163,"262144":86},"domain":"finance","exact_token_ranges":{},"name":"amazon","production_eligible":false,"semantic_tasks":249,"status":"PASS"},{"capacity_bins":{"131072":124,"262144":70},"domain":"finance","exact_token_ranges":{"262144":47},"name":"amd","production_eligible":false,"semantic_tasks":194,"status":"PASS"},{"capacity_bins":{"131072":128,"262144":71},"domain":"finance","exact_token_ranges":{},"name":"intel","production_eligible":false,"semantic_tasks":199,"status":"PASS"},{"capacity_bins":{"131072":145,"262144":71},"domain":"finance","exact_token_ranges":{},"name":"meta","production_eligible":false,"semantic_tasks":216,"status":"PASS"},{"capacity_bins":{"131072":149,"262144":94},"domain":"finance","exact_token_ranges":{},"name":"micron","production_eligible":false,"semantic_tasks":243,"status":"PASS"},{"capacity_bins":{"131072":72,"262144":65,"65536":52},"domain":"finance","exact_token_ranges":{"65536":13},"name":"nvidia","production_eligible":false,"semantic_tasks":189,"status":"PASS"},{"capacity_bins":{"131072":126,"262144":66},"domain":"finance","exact_token_ranges":{},"name":"siriusxm","production_eligible":false,"semantic_tasks":192,"status":"PASS"},{"canonical_source_tasks":818,"counts":{"eval":147,"rejected_overflow":1,"short":13,"train":657},"domain":"codeforge","exact_context_token_range_sample_counts":{"128k":9,"256k":16,"64k":12,"other_long":767},"full_message_capacity_counts":{"cap128k":305,"cap256k":358,"cap64k":141},"local_training_eligible":false,"max_full_message_tokens":261954,"max_observed_full_message_tokens":263244,"name":"codeforge","primary_long_samples":804,"production_eligible":false,"status":"PASS"}],"strict_long_dependency_verified":false,"token_count_basis":"pinned chat template, thinking disabled, complete assistant target, truncation=false","tokenizer_asset_manifest_sha256":"bbcbdfe073f579453f3c891f989a43fbb15cc88952e9f8ae294f04f6ca2036cb","training_code_bindings":[{"bytes":387,"path":"/WORKSPACE/longworld-worlds/configs/llamafactory/TASKBANK_P64.yaml","sha256":"fee14c63bc6396991ac6b145ca7837c563c592070b832443a2a9490d05380460"},{"bytes":23250,"path":"/WORKSPACE/longworld-worlds/longworld/core/taskbank_training.py","sha256":"c4dd87b5cb73466702ab432a7cf8054e07a61e84822cae93e928529b2d142c54"},{"bytes":4288,"path":"/WORKSPACE/longworld-worlds/longworld/core/tokenizer_assets.py","sha256":"792ab0ffe4b77439dda2f5ccaf90b63aa327c760464a2e272a9b9bb4b74179f6"},{"bytes":12241,"path":"/WORKSPACE/longworld-worlds/scripts/materialize_finance_histories.py","sha256":"e441bcb922d29f10c4c2339125a082e7b1d460c8f72de7d1ed8d9893f05dd797"},{"bytes":30511,"path":"/WORKSPACE/longworld-worlds/scripts/prepare_p64_training.py","sha256":"e34255ef827dd30fa0eb595dfffb5b2628681debe2df313619d4de01d3559aad"},{"bytes":13978,"path":"/WORKSPACE/longworld-worlds/scripts/train_sft.py","sha256":"749f72b641021cb79940fcf09c11f5c7861459617c8c6a724f093888cb184d41"},{"bytes":16527,"path":"/WORKSPACE/longworld-worlds/scripts/validate_training_export.py","sha256":"ecfb109e1798ed27f8193d1be49c82a2243c907c77b9b8254a3074543b621201"}],"trust_scope":"local_probe"} diff --git a/local-probe-train-ready/p64-primary-training-v2/TASKBANK_P64.yaml b/local-probe-train-ready/p64-primary-training-v2/TASKBANK_P64.yaml new file mode 100644 index 0000000000000000000000000000000000000000..db4d07fb025025acf98c987e502fa4653c6a09b3 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/TASKBANK_P64.yaml @@ -0,0 +1,12 @@ +# Local P64 primary handoff. Default separate Finance; CodeForge and combined dataset aliases are manifest-bound. +model_name_or_path: Qwen/Qwen3.5-4B +model_revision: a7b0d22b993d71000cf2eadfb37222a67cee521e +trust_remote_code: false +stage: sft +dataset: p64_finance_train +template: qwen3_5_nothink +cutoff_len: 262144 +packing: false +neat_packing: false +train_on_prompt: false +val_size: 0.0 diff --git a/local-probe-train-ready/p64-primary-training-v2/codeforge_eval.jsonl b/local-probe-train-ready/p64-primary-training-v2/codeforge_eval.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..1b240e5e7a461b5a081baf963d2d96faa0293c23 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/codeforge_eval.jsonl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6705df17f1c099084dcb2d0a56a3e29376ac80fb4a83eefa4a7c9a81f9d7bee +size 56658476 diff --git a/local-probe-train-ready/p64-primary-training-v2/codeforge_train.jsonl b/local-probe-train-ready/p64-primary-training-v2/codeforge_train.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..f09b57c89b02d5b5fa2c728d62fde62cc79a3da9 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/codeforge_train.jsonl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e78852013aaa49ea54e65e0d475d5b8e0c61dc1e096ddcfcd4a28ccb3cd7cbae +size 242734380 diff --git a/local-probe-train-ready/p64-primary-training-v2/dataset_info.json b/local-probe-train-ready/p64-primary-training-v2/dataset_info.json new file mode 100644 index 0000000000000000000000000000000000000000..bd4bd4c68610d116b4f0b195d969947a932d6500 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/dataset_info.json @@ -0,0 +1 @@ +{"p64_codeforge_eval":{"columns":{"messages":"messages"},"file_name":"codeforge_eval.jsonl","formatting":"sharegpt","tags":{"assistant_tag":"assistant","content_tag":"content","role_tag":"role","user_tag":"user"}},"p64_codeforge_train":{"columns":{"messages":"messages"},"file_name":"codeforge_train.jsonl","formatting":"sharegpt","tags":{"assistant_tag":"assistant","content_tag":"content","role_tag":"role","user_tag":"user"}},"p64_finance_eval":{"columns":{"messages":"messages"},"file_name":"finance_eval.jsonl","formatting":"sharegpt","tags":{"assistant_tag":"assistant","content_tag":"content","role_tag":"role","user_tag":"user"}},"p64_finance_train":{"columns":{"messages":"messages"},"file_name":"finance_train.jsonl","formatting":"sharegpt","tags":{"assistant_tag":"assistant","content_tag":"content","role_tag":"role","user_tag":"user"}}} diff --git a/local-probe-train-ready/p64-primary-training-v2/finance_eval.jsonl b/local-probe-train-ready/p64-primary-training-v2/finance_eval.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..cb7eb11d2f5102b40539b44707eda2b114ee55f0 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/finance_eval.jsonl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3ee5d59d48d6f87c032b1afb8ded4f9b255aa6be12cf51d79eb54d32b6875b89 +size 240577104 diff --git a/local-probe-train-ready/p64-primary-training-v2/finance_train.jsonl b/local-probe-train-ready/p64-primary-training-v2/finance_train.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..02151bdfc7ada218eba897c684176a83b21c587b --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/finance_train.jsonl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:18cce1a697eeacd468b1a1ac278c42734a91711894ec65c31cadd79bb170da57 +size 653089550 diff --git a/local-probe-train-ready/p64-primary-training-v2/sample_index.jsonl b/local-probe-train-ready/p64-primary-training-v2/sample_index.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..9c994746808d0384120024c66ce531f116fa66e4 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/sample_index.jsonl @@ -0,0 +1,2486 @@ +{"domain":"finance","full_message_tokens":81992,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":0,"sample_id":"sample:5b559821ed599de7a82bdf67e6606fe5ad4ce51d1698cbd315b78a61579ee198","semantic_task_id":"00925024f8789ea6716fa7c7d9e9464fd4bb2b58968b013e0aaa97beb0c8a3a4","split":"train"} +{"domain":"finance","full_message_tokens":88155,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":1,"sample_id":"sample:9bbfc77926cf154f487c8ab040ac6a9e3b2f297ad96549ff0631e6cd445ecd1f","semantic_task_id":"05b1439f07f32bc4ff6d86ed8bb7ba7bb05d2928ed2260fb98516f61572d5a37","split":"train"} +{"domain":"finance","full_message_tokens":77330,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":2,"sample_id":"sample:8ed0a22494e81e54cd2de8650340d6fe7fb64630cb12f46b43bb7a954e3dc393","semantic_task_id":"060d10ac8e8d8c15e32ba15be64955683f5d8d1159ce7f88cde0ca48afbadeb8","split":"train"} +{"domain":"finance","full_message_tokens":83335,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":3,"sample_id":"sample:3ec7329bac9f1e58caa89db7d31099c2c8a0efb5d6b596319477af1133fa4f9c","semantic_task_id":"06d9ea0fdffaa318d5c1059a274d29503344b7a01dd7259b0c0bf0c76a283b40","split":"train"} +{"domain":"finance","full_message_tokens":82094,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":4,"sample_id":"sample:5b3f5637a5be6eebbe744acb5c316ce6d186c0dbeb75cd5bef8e0004ce9de799","semantic_task_id":"08f8aa9cac3ca88138818f7704064a91a106906b94f68dc4fc09af3797a9b8c1","split":"train"} +{"domain":"finance","full_message_tokens":242431,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":5,"sample_id":"sample:ff42f60a8966524daa16814344a5c71b49f5e9e66d853c9c5b75ffbc25eeab2c","semantic_task_id":"0b68305b108125132906cd281c1238f667e280fbe5bb117f8f3529fb2b09b400","split":"train"} +{"domain":"finance","full_message_tokens":76602,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":6,"sample_id":"sample:de8069cef67d1b5c7f0174c8663b3aecb7878477fd8740f3cdf8f4e2b22476f2","semantic_task_id":"0e41f2418d4989e607a4fb716bc95ea5bbf73a8fd0ebf4fa5a91bd583a035f47","split":"train"} +{"domain":"finance","full_message_tokens":242431,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":7,"sample_id":"sample:5ced33290b5d0e097610b330e69ef899d9df93c2f3bfa2bb44725a36cf22bbb2","semantic_task_id":"0f3b88380c86f93f4e3cb938743fd52d0fb5b2073139a19c94131a7730ae60ed","split":"train"} +{"domain":"finance","full_message_tokens":88050,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":8,"sample_id":"sample:33c44952fe101a8672f26b68e776f1ffab23de5a58a7ea20c26862c3cbf932d7","semantic_task_id":"109db02db3ee3abdcc24531b9f2518aa2dbd36551022cc8fdfaec113f0ee6f66","split":"train"} +{"domain":"finance","full_message_tokens":83377,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":9,"sample_id":"sample:ae04ba5a12543a5092827fc0d13d77cc3eb9509b736c460c9b2f4c2ac1299057","semantic_task_id":"12080e82b1253c3b7b7ae9c5245a50f66edabfe80f3dcdfb64df98c723e1d554","split":"train"} +{"domain":"finance","full_message_tokens":82093,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":10,"sample_id":"sample:87076701f8e2102f9e294bc27b1edd91c50dbd3b9ffe45bc7277364c998bb71f","semantic_task_id":"15b757cb04ccba7f8bf0aecaead95e608b823f807e4ad933fffc81a81913a6ef","split":"train"} +{"domain":"finance","full_message_tokens":77221,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":11,"sample_id":"sample:a96291d7c1bd98a9bd5905536782b0d7604bc90fd2e70eb64a9aa3720d2b5e67","semantic_task_id":"16dabb973e98669b90275ccf1cab0ebb65c43ef28be67b097908c8e96f4dfd46","split":"train"} +{"domain":"finance","full_message_tokens":88054,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":12,"sample_id":"sample:6dd66ac9a86e6661a837cd06f335a4d08353d75416248a1aa7db2b6b32faa6b9","semantic_task_id":"17e64a8935a01882dee540c88e37396d841eed38f7597b6a89c8de1c40c51121","split":"train"} +{"domain":"finance","full_message_tokens":230871,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":13,"sample_id":"sample:5a68d7d69e3402af62e4c22403bf5b8928dd572a902d32665eb2b40fe31df507","semantic_task_id":"1929755c2d69c5f8f6bf0a735ddde413630f3c6a7de1d93584978616ffa5426c","split":"train"} +{"domain":"finance","full_message_tokens":82093,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":14,"sample_id":"sample:4cdfb1a9be686a5a0fbeeda87677f114fb7cadff7d14e25d719720c7ef4f7a20","semantic_task_id":"198c2f6abd4e0709867587d1eb036d185a178e3061e69f48909ce38ab381a250","split":"train"} +{"domain":"finance","full_message_tokens":77275,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":15,"sample_id":"sample:88d12df2fac0a745b25b76b2b385e89aad17f27b39fd5c0930d8e62611371d68","semantic_task_id":"1a4f6049096d09854af3e25dfc0e393276179539b9ca6bd384b4f4cce2c28c9a","split":"train"} +{"domain":"finance","full_message_tokens":81992,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":16,"sample_id":"sample:e2bcfeca4809d472a3f44094fbbae3174c23f9f4dd29e1b3b6128aabd7c191db","semantic_task_id":"1c4570812659ef650c89b666b2cdd82a11b775f967fb2851473aac2e987d5642","split":"train"} +{"domain":"finance","full_message_tokens":88052,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":17,"sample_id":"sample:640986dc52564cf827d80fa7e3cbfa411fcac2d8e11c1e4aedb50e8f7b09362b","semantic_task_id":"1f3a4eeb4007d0a9fee71a4058f9f753d50c64ca10dc3f3529bcedc583fe4921","split":"train"} +{"domain":"finance","full_message_tokens":76551,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":18,"sample_id":"sample:6e2037b3b5a7b9bba059179508086579303d682c024079d838fe87cb2cbc4c3b","semantic_task_id":"1fb10c71df72b8778fc2292f2153e6d11c0b2a4e6136fd966281659d45ea1d4e","split":"train"} +{"domain":"finance","full_message_tokens":230872,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":19,"sample_id":"sample:a5d650424b056e3a7cd39039966e4f342ad6951bb84a643f76f2b083e573258a","semantic_task_id":"201ccb188eba5918e30072a418957e41120844abd13a1188cbc9537a7e25adc7","split":"train"} +{"domain":"finance","full_message_tokens":82039,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":20,"sample_id":"sample:33fe87adfc4e5a853eb80468e66e4cc0c4c2034777015f4a0dd775b204b2cd93","semantic_task_id":"22aa6842e0209c7afae59aacc4f4fe60555cd8a76176daef35631f3ecbfa33e7","split":"train"} +{"domain":"finance","full_message_tokens":242450,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":21,"sample_id":"sample:ec55f852fa7efa7b89095b58207d34c7756d9e028933234e7bb79b0fa0bf2e51","semantic_task_id":"22ae101db8dc86f6f0df1f8f4f4590dba2936ae0ff7f9479bf1a8067e38c02b1","split":"train"} +{"domain":"finance","full_message_tokens":236932,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":22,"sample_id":"sample:711314abc1f974e951aa8ae1f6ea500a19e8fa9d031af1e561f9e59b38d7bfef","semantic_task_id":"2345952ee8eea222d89a064887ef3471f1c42116bfe521563e48a0784f6c0eb3","split":"train"} +{"domain":"finance","full_message_tokens":236931,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":23,"sample_id":"sample:81fd2ce7acd6cdcb1ae415ce6c2f14423f31e36110e8a20f8fb3487972de0007","semantic_task_id":"236c206024d76867a394c236cdaabc044e5fed74e844dcb7455677fc87f874aa","split":"train"} +{"domain":"finance","full_message_tokens":88039,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":24,"sample_id":"sample:9682155b1ab3556bf2a114c34d2e5ac496572e857e6b53310ffb21b1a050bcc5","semantic_task_id":"23d288e3a0966f80eb797a95e7fca1d584a98b64fa425b6a19bb3680296006f3","split":"train"} +{"domain":"finance","full_message_tokens":77274,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":25,"sample_id":"sample:b14b77580501527f604c1c90d661375f31b797f79cf6a02c4bed7bc96c78ca40","semantic_task_id":"24d3064331b4a8f25e3751946825ff6897346f8d7d2f2e1cf325131ce9a830a0","split":"train"} +{"domain":"finance","full_message_tokens":77273,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":26,"sample_id":"sample:f46f225b223f73199795ac0b2b6063e7e83e826d11255f1bfc153ed36108ffb9","semantic_task_id":"25c8a4fe668ea872a036d671538fc72f540937221d305437c7bd3dfbfc0faaad","split":"train"} +{"domain":"finance","full_message_tokens":82040,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":27,"sample_id":"sample:8827f6c18b1be6596574f6862eb09ebec6e9b1488eb5b32dbdd4c1c7f2870d8c","semantic_task_id":"290818144bbdf5b49ce7dbef7684657ccd3dd8aead321af883565ec0ec9cc14c","split":"train"} +{"domain":"finance","full_message_tokens":77274,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":28,"sample_id":"sample:d055406a8276c831ea42a6d53ef780beba92f7983879d4bc66adad0cdeb0cfdd","semantic_task_id":"29fb7e982a37d18a1cb4d6d13c0d4d3ba83718f19512e123138ede12bc037d7c","split":"train"} +{"domain":"finance","full_message_tokens":236922,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":29,"sample_id":"sample:a0adeccf7c94ce10aa563f7b49700cac0953a97168b31edc8984d0a8e82a40b1","semantic_task_id":"2a263c871858771df6c00057962376f7507bb252519f5997b1aafed50523c5ec","split":"train"} +{"domain":"finance","full_message_tokens":82095,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":30,"sample_id":"sample:5b8670ae86d8faf2375e766b193ce0dc8c81b1c6212a83485e9a76f1263ef118","semantic_task_id":"2a46fd5fb7cfa4c9a57f29b18007a07c9ad6fb77d48b78a1092c12cb80a64b16","split":"train"} +{"domain":"finance","full_message_tokens":242432,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":31,"sample_id":"sample:22e913cdb710c8310cdb3f2f2ceb4810946a938c696919fe91bbcf2d700a9b12","semantic_task_id":"2aef6087e2bc2a6a842bfe91d64bdcae4cda899b6abb75b6fc41a6458e644576","split":"train"} +{"domain":"finance","full_message_tokens":230850,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":32,"sample_id":"sample:cdc135c7e74757b0f9be2e4a33d207d638d2b96e0a5ec44077867ead4eb5c801","semantic_task_id":"2dcbce1cb05b9ffcd52471c0b53f9eb623ded93c66a0bc963cc9801d4883316c","split":"train"} +{"domain":"finance","full_message_tokens":236911,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":33,"sample_id":"sample:1be1cc3ec19ed99fe3c854166f993490da9472fa91d6e19d860692889a3a7a8a","semantic_task_id":"2f9b37e69d2e15f33b0bf3887ed8f38fe8f18ec9b088ad16ff469edbb6059541","split":"train"} +{"domain":"finance","full_message_tokens":82035,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":34,"sample_id":"sample:a7d0b17f698d88ead72b94461f5fe2dd6780eae621141448deec1a32e1206548","semantic_task_id":"30082d83b21ca1aebfd68a64d286c1f7d373f64a76c0529d9f4927adafd1812c","split":"train"} +{"domain":"finance","full_message_tokens":76554,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":35,"sample_id":"sample:346946646313225b23b4c6178c390d804f467d1a94fb904ee1035f92009a404a","semantic_task_id":"302a66858537bf276f0420580decc715fcf8fbbefc95420ccc4e72fd9117023f","split":"train"} +{"domain":"finance","full_message_tokens":82091,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":36,"sample_id":"sample:d5588ef872136401231420b9d98f7d19069c328789fe9c392cc25ed47aa23a1d","semantic_task_id":"30e230c6c53057b679cc8e5bd321443e943c9bb99068ab39ca1ef1070b5e5384","split":"train"} +{"domain":"finance","full_message_tokens":230885,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":37,"sample_id":"sample:7cb5d3ef73b86ff3bc252a87f0ec98ec265b3c4bb81c63300c2134096fb842fe","semantic_task_id":"31801e5b05abc06d73befac567e3f8e90193ab036cd28b6cafdef69c301f1460","split":"train"} +{"domain":"finance","full_message_tokens":230869,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":38,"sample_id":"sample:ab4ba2a9cf7d1a721fc9df6de496b9d5a845ebeeeec745831dc875734850e3b9","semantic_task_id":"319193ed7e600cc3ecfc1210b4fbc00b0821407ef1cc59345252f580914b055e","split":"train"} +{"domain":"finance","full_message_tokens":82038,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":39,"sample_id":"sample:9b6c1d16ecea5a25b58a17ae288e41e7daa19045fda444ec6ad188a1837ee0c2","semantic_task_id":"33000ae118501be4cbc2b0e2db6afbeabaa1a61c249b2f43c80cfc295a590a5a","split":"train"} +{"domain":"finance","full_message_tokens":81994,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":40,"sample_id":"sample:82e9d5fa5bf8ce5fe5c4e0fa00526daa46b6483553708e73834b4f94d454c5cc","semantic_task_id":"33227d5b4144c325e7505c1ed53eeb15c0d176023a347d323714506d503fadf0","split":"train"} +{"domain":"finance","full_message_tokens":88157,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":41,"sample_id":"sample:5586b12f576b79bb6373dcdc58f00823196867aa6ba52cda22cf32e2b2f0789a","semantic_task_id":"332b1ae3e0a0b41aaeac4c13b93e395f60f44a90bc58caccd3387bd9a1feeb0c","split":"train"} +{"domain":"finance","full_message_tokens":76553,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":42,"sample_id":"sample:dfd3114d97d105a1f239ed99114d7bde11ffccd092e903727feb9c7b2d49a5cf","semantic_task_id":"3439fa0baf1993958e26e065aef2b631cce78cc96014d09682b3bf1648384d95","split":"train"} +{"domain":"finance","full_message_tokens":77314,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":43,"sample_id":"sample:92100817e4bac7808c755f658251956902843612fa0188710fd3088a1366b2fb","semantic_task_id":"344ffab65a71d34a82597a70b8007ad3f00af9981af6eddf2bc693bad260eb56","split":"train"} +{"domain":"finance","full_message_tokens":242467,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":44,"sample_id":"sample:07a1add3784be968e3abef9fe3b9073833d4f61d644d3d495a5ded8a510699dd","semantic_task_id":"3587fb46e41ca7b09420a845c9079faf38ed668fb5327cb36b467ccca53235fa","split":"train"} +{"domain":"finance","full_message_tokens":77219,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":45,"sample_id":"sample:a7172c724ba5e97be62b277636d31568cee7ec3893a30e03f02964d3f2f4fe35","semantic_task_id":"3711b2497ccd176eef8e24eac56d4df4e4c291296b0bca57ae294a0ce96ef6e3","split":"train"} +{"domain":"finance","full_message_tokens":88108,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":46,"sample_id":"sample:e2ca98a1c676e6c7167d5a83e2b78ca82bb4a1a845a398e63f2a6bee33024f5b","semantic_task_id":"3715b72e06d030093adf22c6d7b8e8bc2c41a89d3fbe40dcfbb5f0c981ac976e","split":"train"} +{"domain":"finance","full_message_tokens":82091,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":47,"sample_id":"sample:541a1361f72508e944289c7ac1f4d925aafff33c965ce2828bee960575d7744c","semantic_task_id":"3864700aa57a9a3230fc400537df7f01ec707fb62646f6266a7ff20fa1a7f40e","split":"train"} +{"domain":"finance","full_message_tokens":88155,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":48,"sample_id":"sample:a7be5e5490b4b672db1ed0ad13d06626b564bf165e600911575661e842bd5487","semantic_task_id":"3bbf15c875e522f87413900a0fa2d5ddf12da831025bb4eb3240d00f770dd4e6","split":"train"} +{"domain":"finance","full_message_tokens":230848,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":49,"sample_id":"sample:83121eb925c73b629d514487bf1c2b06e43bacf346057062cf7dc97e27a32006","semantic_task_id":"3bff35b82776f113db3bc270aa4266b8e086b7b3f67687e3d5be5cdee8193c45","split":"train"} +{"domain":"finance","full_message_tokens":77373,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":50,"sample_id":"sample:66788dcb9847e04d28c53e8e50d4e1cba1861440c7ef4316cdb4cba21d277d75","semantic_task_id":"3daabef51fcd63df9f70d1da3d4da84db58285e806837a2f96a4560df9544b4d","split":"train"} +{"domain":"finance","full_message_tokens":236911,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":51,"sample_id":"sample:3cc5633243fb9d4a57416c4423e34336b100dd1b545bdbb834f99596423d8509","semantic_task_id":"3e139fba7c3082c380e3f1461b8204fdf183e5a5c5ba2c38bd2e618a4de97821","split":"train"} +{"domain":"finance","full_message_tokens":77319,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":52,"sample_id":"sample:26951ca7a427286731f957baac7c3deb21821a85ccf402da65de65971df5eb45","semantic_task_id":"3ee35b0795c0498f2ab0115e033f78e22c92461448ab9bddc9a07b9d423de323","split":"train"} +{"domain":"finance","full_message_tokens":83337,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":53,"sample_id":"sample:feb54c9502223669affd2e1dd73c179ef5355e83b3c104aee2693dcb73780249","semantic_task_id":"40c9ef89339fd01ee90256cde6ec5db370056740c7cbb4894f1eafea442d1968","split":"train"} +{"domain":"finance","full_message_tokens":242462,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":54,"sample_id":"sample:f85445b7a25fa3d71d05e9fc12aa1f74132585d4ea109ff0e0316ce1612eb9c0","semantic_task_id":"418741188078ef432f6f562ea322c581a37dd0a3b33d938f0a1dd10292f58315","split":"train"} +{"domain":"finance","full_message_tokens":88056,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":55,"sample_id":"sample:f2edffc0405ca1c90298b0cfaee8a23c9086d504e04206d20c969995ec7781a4","semantic_task_id":"43b89eb5cedcee5a8f3d1da1d733acde29bcb6763ba4313de61fecaf0da8a6c6","split":"train"} +{"domain":"finance","full_message_tokens":88107,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":56,"sample_id":"sample:0f697426a5584c49b0093c34e1fca920b2b100097338e2db6dcfe8d37adeb96c","semantic_task_id":"467cc379bab060bec66eba1ec99772ba3b7a41628c913f43aacf4ccec6b6a92d","split":"train"} +{"domain":"finance","full_message_tokens":88156,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":57,"sample_id":"sample:503c4823e87f5e321b32bf4a2715c8272e3fa0d75a4df68dac2a739b39891856","semantic_task_id":"48951474722342e80473ba9e24601838c2771c83b2843e0a4b24e36046a8234e","split":"train"} +{"domain":"finance","full_message_tokens":77274,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":58,"sample_id":"sample:bf5342fca39f4ad37c8aa8a31780e42405d515dad052fa2311c34973bd707582","semantic_task_id":"498a328e6897f67e9fbc23e73dd577f0c491873363e9c2ccaed2a36fbbb9b8d6","split":"train"} +{"domain":"finance","full_message_tokens":76608,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":59,"sample_id":"sample:4266ac94e0d568a48ef3cf0a11f180387b55d253b2dd196348afc6b8ecf1525d","semantic_task_id":"4aaad2f3060cdc90395f90a2ad17b463e50a093a2046f2ecccb0c3fb6e0ac9ab","split":"train"} +{"domain":"finance","full_message_tokens":236932,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":60,"sample_id":"sample:9f616c1d3c4688d8633f1b17ace84dd329328f3389502c99c7e9624ce5138c2b","semantic_task_id":"4becdbf98ed212401d460dac39125fe775007757b23af55be96e537ed9f3b283","split":"train"} +{"domain":"finance","full_message_tokens":76621,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":61,"sample_id":"sample:4d7281177a13bbd3de02a87b032c5824337dbdcfedb6888b86be56f9535a3db5","semantic_task_id":"50d83f73355e24a369dc1a04c27b76362c4ce42f553f2314d9a75290fb2ee6da","split":"train"} +{"domain":"finance","full_message_tokens":81993,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":62,"sample_id":"sample:3de89c8640bf65d70012d98ea257dfae4e49b262c72de8b72f87fff9f35945ae","semantic_task_id":"51be81af591674778c5f1256dc7fa9fa0081982dfb34707012f738225f9decb1","split":"train"} +{"domain":"finance","full_message_tokens":230851,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":63,"sample_id":"sample:ef02a19260beecc92ecb47069bd1fbe9013bedde62dc462a309601f274a8feac","semantic_task_id":"522b557064446d516b85a756f5e90fe09f140e47bc852b7cdc1b15b62c724b5d","split":"train"} +{"domain":"finance","full_message_tokens":88108,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":64,"sample_id":"sample:541b9226d9fa8c5d71854fb001513dd062c31044261e7ae952f18a4b8b50f97d","semantic_task_id":"546f54abcf1158722cd1792bb6171f2cbce119da7c86577e18ec89d3529f04d6","split":"train"} +{"domain":"finance","full_message_tokens":77344,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":65,"sample_id":"sample:2b1aa033e0c14ceab4c546b4b25dc93b660576eec9184d7d0c1014918684d024","semantic_task_id":"557917c0e11fed05eb58fddf8a39053c9a6b35c5e514cd9ed32847060f48ae87","split":"train"} +{"domain":"finance","full_message_tokens":82089,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":66,"sample_id":"sample:a66f0574850f202da7d9d9ab827271b55b93600cb781e1132492414ba41cd70d","semantic_task_id":"5a572e36260501df544c632485924333510e03ec28c5e9e7fdb4498bacfaed19","split":"train"} +{"domain":"finance","full_message_tokens":236948,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":67,"sample_id":"sample:721b6caca6143abc7e8d79e4cebbaa26c3ac8ddf04798fe57cc432713671105f","semantic_task_id":"5c1ba42c1b04137e7225581aeac18cb35069a13ad7f03b219a5df892326468d8","split":"train"} +{"domain":"finance","full_message_tokens":82079,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":68,"sample_id":"sample:cedf8b13eaf58c69120bcf8eca11c4498cb5fcf8d63c8cd40e47640a981f660a","semantic_task_id":"600481dce9396647e6d3f7f934cde85f86b4c945d36c074b8f23ef6c20db870b","split":"train"} +{"domain":"finance","full_message_tokens":81992,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":69,"sample_id":"sample:31585e5660ad257fd9c2ca852b7f55db1171c55e226538b5355a3813e0fe7ce1","semantic_task_id":"612c5d4ad31be861b027ffab59db24b34864c611ba8f60daa00af71528ece7e6","split":"train"} +{"domain":"finance","full_message_tokens":82034,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":70,"sample_id":"sample:9ef1398a442d735907a410f8935984f252c136cdaa4f2f782de14642fe5c2310","semantic_task_id":"62df82d0b0b2e3b2f83dfc0c59910f310c20c8db9acf3b347d51eaa16cd57959","split":"train"} +{"domain":"finance","full_message_tokens":82035,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":71,"sample_id":"sample:55157176d33f57971952c1ead4d61a04f0fcfb8db02228d13764afb55b10517f","semantic_task_id":"64271beb40d38899565b31e0e7ccce0020856c30606a650269c853510ad5a06c","split":"train"} +{"domain":"finance","full_message_tokens":82088,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":72,"sample_id":"sample:38e3bffa606b67ae2f522d320eac21c9a4a67dfa5f28341d9532f1371228ee16","semantic_task_id":"65e8708e58483404b430f8cd991e41d030713bcc7452ecc277e7136dfaefdfc3","split":"train"} +{"domain":"finance","full_message_tokens":82094,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":73,"sample_id":"sample:8442ae30aa1948a23a052d7b044de147e4b03ef183915cd3c7db58c9199caa30","semantic_task_id":"6b1e8e1f61eec4c1b40dd216d09beb75240ea31adc464083e212a44a8305d53a","split":"train"} +{"domain":"finance","full_message_tokens":82094,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":74,"sample_id":"sample:5ea91a9f85a4178b08b840f75d9b9dff9bdd80adbabc369b48af46ed21d0ac9d","semantic_task_id":"6caed6e776a850760ee5369cdc1e6c72b62950a42e2ce9d072d83141ed736530","split":"train"} +{"domain":"finance","full_message_tokens":77290,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":75,"sample_id":"sample:c660dd00c8972f6a3ae40d0a5238fcc2c11ae08ac5c9c07a40bedb402faf05d9","semantic_task_id":"6d7c496160667809bbbe184ff51ae8784505760975044416004de90cfd88800c","split":"train"} +{"domain":"finance","full_message_tokens":82090,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":76,"sample_id":"sample:3b159786867964b5ae72acb68b73db0a5b2c093bfe9fd5112d0d4ab645ab6968","semantic_task_id":"6def3608998f4501f11e5eeae96aae3cad86299e458933b364518e9bcdf9f869","split":"train"} +{"domain":"finance","full_message_tokens":230849,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":77,"sample_id":"sample:e5070f5efdc2d57dc5b0f7d21388bf711298763fd74e142841771c527ce734e2","semantic_task_id":"6f8a4fd02458fab360458bdc1ee9c759c3bda676a17da80e52e7a196590e8246","split":"train"} +{"domain":"finance","full_message_tokens":88050,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":78,"sample_id":"sample:e1044fc56c5aefb9844c42cfcd3871f0592164863f85a53c43a90981569536d8","semantic_task_id":"6fc5fc878e9cd2d465ebd74c9a27f68e453646bd0c8298d47f41be494431a3e4","split":"train"} +{"domain":"finance","full_message_tokens":88108,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":79,"sample_id":"sample:6f77954e7ccf06291502301e8756eb6324221132c3563375ba49da98295c313a","semantic_task_id":"6ffdb882834de09ad0851326a7a0e704fdcf949ada546a62239b22dbc5ecbc11","split":"train"} +{"domain":"finance","full_message_tokens":81992,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":80,"sample_id":"sample:e1f99cf5a734245d8c21eea3e4820802faaf0f0222d55cfc7e67675c68651a6c","semantic_task_id":"73a7411a04561dfd03a1d757676a65ffef505ff083871d116af550c9bdfb755e","split":"train"} +{"domain":"finance","full_message_tokens":83337,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":81,"sample_id":"sample:268c1cffee42b4cd52a555901740d8ab6bf6aaa9e62e1f286915e161c11616e5","semantic_task_id":"74893cc8db352d248f38a94c46bd533d66851cb09ac8d6e6c1bcf8ab1bdea74d","split":"train"} +{"domain":"finance","full_message_tokens":236934,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":82,"sample_id":"sample:2a3af3f558c98c7b3ea256486732474062b4e930e91850719813c266bee2e9b3","semantic_task_id":"7641b1adf67a5f5b307faf9963f263dc140ae9db5edcdcfeaeaca6132d40c2a6","split":"train"} +{"domain":"finance","full_message_tokens":88106,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":83,"sample_id":"sample:53047f83a9c2f6497ddc906fa45fb0db8873410015a94c7182ba0f81640bc328","semantic_task_id":"78354da55c118bf5247f42f85a9f3d12ae582a3709889133bc5697148f60243f","split":"train"} +{"domain":"finance","full_message_tokens":77220,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":84,"sample_id":"sample:240f6a311a81b81922dca5094bfd933f8364fbb4818cabaaca4ca78e79acc2fa","semantic_task_id":"79dae03047f7cee3e95a20fc1ce625403a9d4a56e1abcaabf9cf6e21b7fb4420","split":"train"} +{"domain":"finance","full_message_tokens":230849,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":85,"sample_id":"sample:a3d0c8524c3c19e6a38897c51051c8fb47711a984d46c72ca7ba853a55d2443b","semantic_task_id":"7a117abb5a59f68a8d9e1ff6932d6582a328c861db1e8ae456f0ff583d4b187a","split":"train"} +{"domain":"finance","full_message_tokens":82034,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":86,"sample_id":"sample:9d337246e590d2ec75f024b67f270883203786c7fb07c611aa8af2c3a63a1ecb","semantic_task_id":"7d68712f2d5156f8db39bf685794781ff9ae5c60590b9d5cc4da9346529fdb85","split":"train"} +{"domain":"finance","full_message_tokens":236912,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":87,"sample_id":"sample:5b4188035a13693b2100780a26a8f6b231d63c14efe75ea47059040822af69c3","semantic_task_id":"7e63634c325a097b5703eb7e472ddf2e11d6baf336ef7b9f67554d14ef9fa7a7","split":"train"} +{"domain":"finance","full_message_tokens":83338,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":88,"sample_id":"sample:2f5f7a8dc66d3bc0fe314bcd8e08ab70b87989b5f29f8ab227aedb41215fcd2a","semantic_task_id":"7ee6b81ebb7c1d9a3275dc8b37d4c8d168500505593f0e10b0a1f9463f57d4d0","split":"train"} +{"domain":"finance","full_message_tokens":88156,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":89,"sample_id":"sample:491bf97538119d2618d42a7aec592fde533a2496e2e34da8e087538dd33c0d06","semantic_task_id":"7ee82300f78538108dc1e339b1ffdd1789a4f7d6041772e35fa3bcafd888488f","split":"train"} +{"domain":"finance","full_message_tokens":82092,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":90,"sample_id":"sample:97c5a4f70035cc60f63802f8574671425dd6862b6293d1a19a82f474afde8bc7","semantic_task_id":"805f372c2a317459d0c2f99c93a608243a6be329ffe79eb26f52cec0a036c078","split":"train"} +{"domain":"finance","full_message_tokens":242430,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":91,"sample_id":"sample:07f6743bf53b13d395929e0c9e95f5badc57a2bfdafd7c608d48dca81a6cb3da","semantic_task_id":"80a963ef0a0f812d4075fe45895dcefdfda3eaedf3c2c01046786020d3eb1ceb","split":"train"} +{"domain":"finance","full_message_tokens":82037,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":92,"sample_id":"sample:32d797ee8dfd6123177a6dcc8ebf0d0052a14f75df72c613c8022c7e3e87a48d","semantic_task_id":"82d4b4c433bd4f928d4c28edb045ace9f2b20309a5a70be99eef697aa777582b","split":"train"} +{"domain":"finance","full_message_tokens":83335,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":93,"sample_id":"sample:442387fcfb8741bb89b7ffd032e633116b2e1c7d6d8ecff4c708804cbd5a87c5","semantic_task_id":"843346204b6d0ad4f7b6a2c0368711379a49bc779e10f619e38935e9bae98a57","split":"train"} +{"domain":"finance","full_message_tokens":77270,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":94,"sample_id":"sample:1f112b90a64e53071a57cc980f008b3cd0bd94acb2ae6e8890a5e20cdc60cf14","semantic_task_id":"877a3b79cb0e54f4689f6c7cf9916c81dbfb8437cab986a272302891ba8e15bf","split":"train"} +{"domain":"finance","full_message_tokens":88157,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":95,"sample_id":"sample:c648b8888c92762bbe7a76d0af3d64730aba0e488013859d981d1beec66dc2db","semantic_task_id":"89e5dbb2472d20913ad629f93d4817983e68d5ecc1ea1f4f33b7d3f68c6ebd1d","split":"train"} +{"domain":"finance","full_message_tokens":77219,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":96,"sample_id":"sample:497263ae65164af60422d9b07b0a41c29e7798b64dc5225a0fc99026272e9ea7","semantic_task_id":"8aadeec7139b15f624306454d4f5a7268a6e25ae3943e04e10b9093f5c33a8b9","split":"train"} +{"domain":"finance","full_message_tokens":242463,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":97,"sample_id":"sample:2c9b4946b085ff1e931e92433e98ae52b1fd21e4e1f38f9ad0829854659e133e","semantic_task_id":"8ab558195478bc6a2efb42bec57eb0947e8995dd9a44083378f3d4312a1a5843","split":"train"} +{"domain":"finance","full_message_tokens":242462,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":98,"sample_id":"sample:d5954c650b1fb561982cc3607f24df81aaa019633527d08241be5038911a8113","semantic_task_id":"8b1b9db1bb576d7dbd277d6b0a4376c0f7e5d2841697c5eea0ae467a327f4fae","split":"train"} +{"domain":"finance","full_message_tokens":77219,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":99,"sample_id":"sample:38e557bc2fcc29009b87ed5718eae54533e6743736048da753e016b7af282beb","semantic_task_id":"8c648cff3811c283944b253af9b1cc1a35f3bcf5036e7371f5f6b7f1a54d62a9","split":"train"} +{"domain":"finance","full_message_tokens":77275,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":100,"sample_id":"sample:a84e3b0925e5b17139b3b1c76c082060c6d945b0824f59a5b32bc00516f0d495","semantic_task_id":"8ebe488a6ff23e70e04dc8a76a73d643177dfe81801812e0b29ad9404faa7abd","split":"train"} +{"domain":"finance","full_message_tokens":76553,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":101,"sample_id":"sample:1aff9ea66695d362fe6912f130ec1b07d849898b580728e5fb75f596acdac66c","semantic_task_id":"8fe5ae389392acd98e8004aa9638bdcfaad70096e8810d032bd2f5f3e371d694","split":"train"} +{"domain":"finance","full_message_tokens":76551,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":102,"sample_id":"sample:dfde58e7042b65a341e8ef8eae710baaf92b29d591784d8714794f06ce8aefae","semantic_task_id":"9002a1b36bc7dc0f13d0114a728cce6384491734447fbef6191c20d61eca1215","split":"train"} +{"domain":"finance","full_message_tokens":236913,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":103,"sample_id":"sample:ec266c43043884bdc664f50664b1df02acea68b6f90d591c606f2e19bdd05015","semantic_task_id":"90147adbeb74ae66edd56b8b3b421e107c2773fbb05ffe5586edd4b1acdacb97","split":"train"} +{"domain":"finance","full_message_tokens":82033,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":104,"sample_id":"sample:236937593ddbcea9f13ccebd62402b8df509e848f925402f321d2a43fd9bda57","semantic_task_id":"903cbf89cf1c2bd98f83542ab01c6aee7678388f05f3cfa8d468d00a539ebc98","split":"train"} +{"domain":"finance","full_message_tokens":82040,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":105,"sample_id":"sample:2256ab3cb42e058701c6c71411898f5abd943e4b3da2aed6cd17da62ca84a65c","semantic_task_id":"909f2a642017ef4da151c9eccc9e8760aef4cb1357afeb19677a48ea99d559fa","split":"train"} +{"domain":"finance","full_message_tokens":88106,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":106,"sample_id":"sample:3b2cf6cbdbe4ded1f3de84763a2a84be110ed6a9118a04dc21b5a87022b28213","semantic_task_id":"90fcd8f7af8f177745a7e9e3a4dc3c3689d90e667ca53f66c7f7b38b37f4251e","split":"train"} +{"domain":"finance","full_message_tokens":242463,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":107,"sample_id":"sample:f119d8b1234505d41a1a9223feb60f9b3aa088196841171dcea4f1cd3bb7f8c7","semantic_task_id":"952e3a57611b2a5853a6089aca73842c5cf47b64949ed362349aa114de6c4fe9","split":"train"} +{"domain":"finance","full_message_tokens":82039,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":108,"sample_id":"sample:2b38130b7d431e0312b23b55b954e2cb73196ba7eb6524d77baf7fc5ec606a4c","semantic_task_id":"9738aa402ebe7ee1664007a593c9be3549376a791c5e3914e815a1b43d64fe23","split":"train"} +{"domain":"finance","full_message_tokens":88051,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":109,"sample_id":"sample:866f18a8421e5b5cfab40fb6d91cfe734f6ec3eccf8c348b66508af50c0e3edc","semantic_task_id":"992ccb995adc2115f6149bcedc2b653ef073c789da44d484a1cdd0b7bda4e23e","split":"train"} +{"domain":"finance","full_message_tokens":88055,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":110,"sample_id":"sample:36735f81decc3c183f999619a839493671d2dd2be40938175cfe83c22264da3a","semantic_task_id":"9b52dce4577db2a5e8de62fd5c03448dc91b48f6c2cf2f53ca507ca8e11e0600","split":"train"} +{"domain":"finance","full_message_tokens":76552,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":111,"sample_id":"sample:68f5510a53f639dc4a5e4f64711a8215278f1419857e0fce1102428e68e859c5","semantic_task_id":"9b7cebf47fb0a03f273f8995d0e01b236ad98fc448339e0d9b67d575cbfa616b","split":"train"} +{"domain":"finance","full_message_tokens":76553,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":112,"sample_id":"sample:aa1635c808e771155f36c07e84122fedb713c05d132875beaf060babff32551b","semantic_task_id":"9bbe174550cd77ff4977d8ea0189a63c64f1df0ce50ca4e9a97eef68d902d182","split":"train"} +{"domain":"finance","full_message_tokens":236933,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":113,"sample_id":"sample:74e7abbb06440a9a0a2fd066642ae9500faf0f1e0ec16bf52c406c10c10a1f57","semantic_task_id":"9bf3717c4e302c0cea1b72ada4b43abb276ba2910f791d87fffd332dc8f9ca35","split":"train"} +{"domain":"finance","full_message_tokens":83434,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":114,"sample_id":"sample:6e17adfe2ddba8f20aec31e471383dcf97301059262747e727adbc4ab3416e88","semantic_task_id":"9d143151986a3f3369929df6ad77dc50bc4fd02fded99b7ae41a0c3c5bfd7575","split":"train"} +{"domain":"finance","full_message_tokens":81994,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":115,"sample_id":"sample:e70262b1745059e4322fbb02c3a0897f7e1720afec820a5117a56d44bdb7f861","semantic_task_id":"9d5624f8c6b353d2f308149c55ebf8eb192d29a93c56f376e5be286061ccc3c8","split":"train"} +{"domain":"finance","full_message_tokens":83391,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":116,"sample_id":"sample:bd74243a13fe89818cf559e2270a41e30195eb7bab7f55d4ea06e0c5e6d2ebb4","semantic_task_id":"9e0da522caa06d9ae8512fe0ca7d48f56755e4ff55a62fc466f230ff38ddca19","split":"train"} +{"domain":"finance","full_message_tokens":230849,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":117,"sample_id":"sample:1b54190937d0761742ef33e9d481c222f73834aaf639c4055c33bf24fc3ef17c","semantic_task_id":"9e5de180648c52788207cdc99e7e4fe1107f0d070ca947274837fb0cef3b2269","split":"train"} +{"domain":"finance","full_message_tokens":76600,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":118,"sample_id":"sample:1da10fc3e163224476c576141ca2275ca5215d8d25b5428b264fbee32c12037a","semantic_task_id":"9fa8ee207d5643f9029a639efb9bc5b0314b6c616755e9c529d0d68a3413b010","split":"train"} +{"domain":"finance","full_message_tokens":77276,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":119,"sample_id":"sample:82409545864bcf1ed8b406c4093bb4751e912d89b81fb78075753ef2232f18e4","semantic_task_id":"9fd292328ddcf7e653afde612fefc0e70a80c81af0238f4e2ec8f4242408e545","split":"train"} +{"domain":"finance","full_message_tokens":77276,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":120,"sample_id":"sample:452e39fcdade2f45918f692fed5799eb8058a0752b8a6ba9205661a400881806","semantic_task_id":"a0050f2711f3d5af01d95b534b69b3e72f4bda820a86923be1acc597ba3a572b","split":"train"} +{"domain":"finance","full_message_tokens":236910,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":121,"sample_id":"sample:19615b6d3b274c811241e56fad6c730783345b4aa78de9bf8d8d1687ed542cbb","semantic_task_id":"a179dd0ddbd7e55b848f50ec107545aea16eab99607cac76c445d82c1e05d809","split":"train"} +{"domain":"finance","full_message_tokens":88156,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":122,"sample_id":"sample:6ecdbd1a449dafa40fbe4e2057bf1b19fa740c6ecf9328eb970ba92e4cd39e18","semantic_task_id":"a3ec93cb5374af933277e30a99be12279ef283507389a1ec741ffec4e634156d","split":"train"} +{"domain":"finance","full_message_tokens":242464,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":123,"sample_id":"sample:89e2064a18f29327f269b5bc07c707a85e42e448838591a4657fc7f5f96e646c","semantic_task_id":"a4d4ce7db9085858b8abb416085c649c3d6e6c6f5869e5c7b81df15d47f4eeeb","split":"train"} +{"domain":"finance","full_message_tokens":82093,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":124,"sample_id":"sample:749f75c86fdf37251cc90f51a346b867a1e4d3f30d0f6a03a0585abdf2a300a6","semantic_task_id":"a4e330ce22a198a6ca5327e38b6d7ac7104ba19ad4a17dbdda73e54bb7e522cc","split":"train"} +{"domain":"finance","full_message_tokens":88055,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":125,"sample_id":"sample:76b3b1df455737ff44cf21a1a3385e80d6258bf62fefbb8ffec7b478ce41ec35","semantic_task_id":"a5a993ad162cab181e245917143d5737d41969e9cb2403dd01f602395fa96162","split":"train"} +{"domain":"finance","full_message_tokens":230872,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":126,"sample_id":"sample:c8734d024cb3cf55f8efadf7766b3e0a62371df741b57f21505e3f4c65a0fccc","semantic_task_id":"a5c7ff50f06b357892a1aec3b0a7c8304d41d884c4115e309c399b995bc8db5c","split":"train"} +{"domain":"finance","full_message_tokens":236934,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":127,"sample_id":"sample:a842a77a10b267d4232fdb7a7f0f47dd7a9b98f683a5d503ce7ce70fe5d2e3ec","semantic_task_id":"a65d67147fcf6751ebe78efcfb220e1b25e3335ca8b53066ed1eaaf650e86bed","split":"train"} +{"domain":"finance","full_message_tokens":83336,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":128,"sample_id":"sample:3750402d324026500f15504500b40263f16eb45bee386209b268bd6762f88d60","semantic_task_id":"a65dd037384fe07ca3150b1f31e94da8ab042f0414953164982629d6ea1e694c","split":"train"} +{"domain":"finance","full_message_tokens":77260,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":129,"sample_id":"sample:ea4a62b15d1db85dc098b0d925d534dae2fe7e0bac79f91fc4e28e6b0aac8b4c","semantic_task_id":"a75278f57b0eccee1839fdd73d562d99110241051830d31c6bba83659cf686af","split":"train"} +{"domain":"finance","full_message_tokens":242432,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":130,"sample_id":"sample:4ba7f34fd4c20f7a739a54361684698f23641ea1fa7e745823d329f471a0f640","semantic_task_id":"abfbde65da19f74af9a9d4010b9fb7c0d527e0dc26e4eff78bdedd4c131ada9d","split":"train"} +{"domain":"finance","full_message_tokens":77274,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":131,"sample_id":"sample:295d6c702aa38f038cb5d64f77f19874b6b5eab9e97b881458f41f47781b11ee","semantic_task_id":"af933d51221a686c250039012c5a43150986d81f85f12383cf1cf0b319caee6a","split":"train"} +{"domain":"finance","full_message_tokens":83406,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":132,"sample_id":"sample:c40b0a7e1dc355a705f6da1c4a1e3df9ac2819a0493ed26be5398cc4f5d0087b","semantic_task_id":"b14b09b70465292eba4f85c06f67bc029c59195dd4e82e1b4c6808fe30b72d6a","split":"train"} +{"domain":"finance","full_message_tokens":82078,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":133,"sample_id":"sample:876464853eca41aafedb6f2c38239bd0b237c83af7dfc537338e527750c4baed","semantic_task_id":"b2b9ceeb637956e1a18234e135d72ac4c6ef97c6a30ff45daadc11178acbd4b0","split":"train"} +{"domain":"finance","full_message_tokens":242450,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":134,"sample_id":"sample:4eb02f7969ccf3fa02fd7870229614808107f86fe160dbcc6ff8a091a736f44e","semantic_task_id":"b3271af4a1b579f37733bb8f4c5a798e91bd94c5fa517db5ca8be8176b881ad7","split":"train"} +{"domain":"finance","full_message_tokens":242441,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":135,"sample_id":"sample:5db9f11a170cc7168c2be6f2b0f25792cd394d48990efac944bd1f651bdbdb45","semantic_task_id":"b3796976cee64a04cd98836a428291c85d17f6a10743f6f6f94600cc5c003f3f","split":"train"} +{"domain":"finance","full_message_tokens":82036,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":136,"sample_id":"sample:8f74d77d91dfc5a62b30fcbbf94546a673169470cbcaad71ac5c658608a6ae6d","semantic_task_id":"b3e1ea034f0cfcc088e0e2c83133913d7ffb15f53ad1fa7de067afdabaa9a0c9","split":"train"} +{"domain":"finance","full_message_tokens":82089,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":137,"sample_id":"sample:be5c9b72282e87a403f3b1a5143bf20227abc7c7dc8ab8da66dacc9b017d6f4a","semantic_task_id":"b5e336a40223bf3cfadf9812865eeb614605608b4d13a90139e2d8918219df11","split":"train"} +{"domain":"finance","full_message_tokens":88052,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":138,"sample_id":"sample:c398e2a7a29e143a73ed3c12e84e973f91a5870ad83ad10ca1c386023b796417","semantic_task_id":"b6ef21a878ceab9b2720403e783e102f7bbfdbcd9cece5e8b3549e6610b0ebea","split":"train"} +{"domain":"finance","full_message_tokens":236912,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":139,"sample_id":"sample:fcc4f37e0349fcabde610fff79f9bbc9768a87e6ffbaf148dcda77ccc56d406c","semantic_task_id":"b76840e97f88c788e9f169d805f359f2ad7f59b2c12718e751d36ebd7112ee95","split":"train"} +{"domain":"finance","full_message_tokens":230870,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":140,"sample_id":"sample:46ae1a67ef4f6ecaa8980d25e1923902ce6c30848b096e1f60a2589686e3fa51","semantic_task_id":"b9f5e57bf0e165539a3c833f9aaf31c591d2bf370830c9764be550c8c1e3b72c","split":"train"} +{"domain":"finance","full_message_tokens":77324,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":141,"sample_id":"sample:f3456e41b756af1966cd5033f4cfd7ae6cfdc79257098229751fc04cf903e4d7","semantic_task_id":"ba20cd2dfcbc201dd17e5722f18a6c38e16d846781e2ab5e786f9b3f7095b55e","split":"train"} +{"domain":"finance","full_message_tokens":83336,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":142,"sample_id":"sample:f9c37f3d6ec2d1b667f7a5e03ac5455d0bd9158c68a3d088a15131a07b4a0d0f","semantic_task_id":"ba279427cf32f0000dd2b9dc7eb7725b74a8308986c1124b36bcc3bfb8ee7d1d","split":"train"} +{"domain":"finance","full_message_tokens":242429,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":143,"sample_id":"sample:1c4562e3f5b3fac81e6b3e6e2f71d433dee95a543bd114bd01628952f77c6c38","semantic_task_id":"ba468f19a154c06e1843e066c46bc2036e1da7c450046715218848b28fe974d1","split":"train"} +{"domain":"finance","full_message_tokens":88056,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":144,"sample_id":"sample:2c654a5e1cb00dcf3da16ef5be4bceb4daef2fb938b46b4937bed875ebb2146c","semantic_task_id":"bada9f5fc2fbf30619affec26ad2954e7409dcc5fbdb0448340abe0ccea58e9e","split":"train"} +{"domain":"finance","full_message_tokens":76649,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":145,"sample_id":"sample:f760fbf6c97fcbfe00a8f09f31db88d08f0a24226d613c9308616bb2e8a28724","semantic_task_id":"bb93bb7c734c7f8e02f3d880f26c58099addc3d05a8c8e1e3b03ca72d23c1ef2","split":"train"} +{"domain":"finance","full_message_tokens":230871,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":146,"sample_id":"sample:822843d86dfd96eafa50240f2a2394708652ebcb3eb3c90c4650b8e0f2574842","semantic_task_id":"bd1ee4bcf20e00bbb1f605b09ac8c64ecd2d15f2086513cdadb375e1833e514d","split":"train"} +{"domain":"finance","full_message_tokens":77267,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":147,"sample_id":"sample:0b8d141893469147ad2e922002677b0d5785442013e32db6b0fd91a40eee4a93","semantic_task_id":"bd9e3b9701e37253340814b622772c7ae1dfc93018085897f644126b2074b574","split":"train"} +{"domain":"finance","full_message_tokens":242430,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":148,"sample_id":"sample:1bc16a4bbecaf14151203093a4a5b1942db38ad5bad3b860c46332b677c64c68","semantic_task_id":"bdb8a3b10bf1e0dd3c7188c1f305d828b40fa6434a5a74299a83ac02368ebd0a","split":"train"} +{"domain":"finance","full_message_tokens":236911,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":149,"sample_id":"sample:a1532db173d0a41c2c70b88776f6e0ca01d961307d903acbe2494ec000e1ecb5","semantic_task_id":"bde7c9bf1fac782ffa68b82922679ac88df4a2a2baf98493f53647fbd5d38465","split":"train"} +{"domain":"finance","full_message_tokens":230849,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":150,"sample_id":"sample:681f69352a8b280e29ee22babbee1ac14a14638f5bc9cc5160a53cdf966b24ef","semantic_task_id":"be8e120329c67ef3e040cab0a77b82dc975e6855575d9d9f0654392f7893caa9","split":"train"} +{"domain":"finance","full_message_tokens":77219,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":151,"sample_id":"sample:78e9ac2c34d046388bad20ac7ca36aea9987f9d90d379d4215b6fa0fd0a92fda","semantic_task_id":"bff96c4418234962242163e95f18379fd56c9086a188e8ca92a6221a53a56c47","split":"train"} +{"domain":"finance","full_message_tokens":242441,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":152,"sample_id":"sample:b947ed69e6f0dea2473c318755810f58c2120c13c3011abcc27b681d42f0e654","semantic_task_id":"c072686a5db7a7af9ff2815a93ae5ec49776675d5415ed113c9cb9aa25d4d9ed","split":"train"} +{"domain":"finance","full_message_tokens":88054,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":153,"sample_id":"sample:bc33ccef0658dfdff7604f95e15d0bca14bfdfba475da156132587c7dff529e5","semantic_task_id":"c0d6662f77b4df22b5c6fe93c25784d6dd09ef84d9e95b7ad2e78a6502283021","split":"train"} +{"domain":"finance","full_message_tokens":77275,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":154,"sample_id":"sample:3d443bf586da885f25aaba9daabf3e25b5aefd22cdd922875fba848f5d7f1ed9","semantic_task_id":"c0de1099eb6c36d59606a7b35f826e84d2b0bff26f2d8feaa790307e94e43d52","split":"train"} +{"domain":"finance","full_message_tokens":88050,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":155,"sample_id":"sample:ca1506d186aa5c329e004024262976f1c1e86cba0fd7f7afc9a6a46402f58dcb","semantic_task_id":"c21955d7eaedf47c1eef2bb2fd4225a940e4f1fe012b0ca6561425a1b2587807","split":"train"} +{"domain":"finance","full_message_tokens":242464,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":156,"sample_id":"sample:0ec854bf01fbc9893418c2cf9df9dc10285c967a470f33ef712c5c2212a612e5","semantic_task_id":"c4dcafca40fb8e960bddf7108c9f70c7ae4b6f422ca11f0fc4ed5b73ec479bbd","split":"train"} +{"domain":"finance","full_message_tokens":83338,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":157,"sample_id":"sample:6a06aaf29898c5364aa21d75eaa71599a559e72bc8a649bfc553b71fb76914ea","semantic_task_id":"c5a0c0d42cbd4d4a34585487773914f29c0f9587b5b75f6771a89a06791c3d52","split":"train"} +{"domain":"finance","full_message_tokens":242464,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":158,"sample_id":"sample:778ff930ee01e8750c2a966e7e9c43292a39bb64dc596468ab6b5350ec3a9493","semantic_task_id":"c640d54fbcaa1f475a2b31e97a30d6db467110643725a318533110c4810d874d","split":"train"} +{"domain":"finance","full_message_tokens":242430,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":159,"sample_id":"sample:e4d5373fee2c51932a8aa5578bbb4cefb3e48482c31b92d90a22d4c10d586dc9","semantic_task_id":"cc20fcd4b0b4d098ec665d40b79d50b9386b3e03d2d374ce579bd1a378731748","split":"train"} +{"domain":"finance","full_message_tokens":230850,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":160,"sample_id":"sample:f9f6ce6719f8ebf1002d023034d249baca47ee460e39e91582997301b35310ec","semantic_task_id":"cc91dcdc32ebed8c2876148c1be1deaddefcb76207042486f2e2f910cccd6c2e","split":"train"} +{"domain":"finance","full_message_tokens":88054,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":161,"sample_id":"sample:faa5082f76b0b15929f04688d75d469eb3bfdfd30d01c9a42dadd071d3dc667c","semantic_task_id":"cd6612c1c40354dda51186e0fc597e9571ea0687003ef3a04aace467760d10c8","split":"train"} +{"domain":"finance","full_message_tokens":230849,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":162,"sample_id":"sample:cb4ef24e9d2324d1d26a6700535c476eff6a1f45a9df0016d58f1c03598b61f5","semantic_task_id":"cf377bead5def5fafe7e78ec208876876a1906644354bd8506ac36de21826edd","split":"train"} +{"domain":"finance","full_message_tokens":88055,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":163,"sample_id":"sample:210fa5e04b5bc23920f8a1284833d965977d2f50ba19c05d2d24904a820ecf97","semantic_task_id":"cf596b1f4fe222751eabc17f40fa4b49c288604e7b25ab6e63735a2e9b971392","split":"train"} +{"domain":"finance","full_message_tokens":230870,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":164,"sample_id":"sample:6cae071f746fa7795acbb8a1fbea2e526b11da917ed6f94701504a7be21e2689","semantic_task_id":"d0a15f9501935303596a71634ed70f2d182d527e91023294a684cc82063a30f8","split":"train"} +{"domain":"finance","full_message_tokens":88056,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":165,"sample_id":"sample:313604c2420ea1178db20ce84def24eebbe697ad662ffdd2358817602c7f9dd3","semantic_task_id":"d1afe2a1c04f291af3eacaa5817c6dca5cd0b1190363d4f2b1ca9eadac738a48","split":"train"} +{"domain":"finance","full_message_tokens":77221,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":166,"sample_id":"sample:ce1715c51bc28bb6c8bd598d1e07ebdff64f3fd6d42245b2e4370de8c6d0928e","semantic_task_id":"d1ed7509e0d264ebe700313118d5d73446d4be3ddc1d9f260ef30f63915f38ca","split":"train"} +{"domain":"finance","full_message_tokens":82090,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":167,"sample_id":"sample:14c52bfd08ccacc6f888455b53ffee7546f1e49d2312c419ed6ad19aef811503","semantic_task_id":"d299aeeebee2b97da3872dc75fedcbf067e6086727f50adbaae7e9010fdb2e39","split":"train"} +{"domain":"finance","full_message_tokens":88106,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":168,"sample_id":"sample:14ebbc6b202a4b0311180fb0bccb758db23d47cdc2a287056214d095e0749173","semantic_task_id":"d3d11a8a2f92cefb42e8e64122a3cd192d22e8bb5affebc2e11b986b24cb2e53","split":"train"} +{"domain":"finance","full_message_tokens":77218,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":169,"sample_id":"sample:c53d87e460918635475169c607b30a61b51df15589dd8f4a4da840d73e6bcbd1","semantic_task_id":"d4b688756712afeef6dada27b7207294b714d831d44a74486063cd0c96c8d44f","split":"train"} +{"domain":"finance","full_message_tokens":88157,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":170,"sample_id":"sample:fb72c2184b43173d57282512be99496681233313a1fbc161ee28a86d349f1bc0","semantic_task_id":"d4b812fcbb9bd962531386eb98c7a5108925af08a7cf6744ab11c370a767c0f5","split":"train"} +{"domain":"finance","full_message_tokens":82038,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":171,"sample_id":"sample:d9e6a9f6042f0b065aec6299dcaa207e6130b024627e439cf5f48b36277f0e4d","semantic_task_id":"d4f5cbe14b86df06f2bbde7018e3adde36580741b4c35c701cdd48f35be67019","split":"train"} +{"domain":"finance","full_message_tokens":88040,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":172,"sample_id":"sample:bcedeb7be7aedcb1e2bbe1bfb3077dd42e32088d9aa16a58d44cdd3d02b3a83e","semantic_task_id":"d5c07a517b06778069983a88816820e18101fb87cb68d4774232698c9a2e8e74","split":"train"} +{"domain":"finance","full_message_tokens":83387,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":173,"sample_id":"sample:1fdf3dff2abaef246801dd8638eb0eb7c2da67a7d60257b50a51dd806a51008b","semantic_task_id":"d75da2e1d81838c9f4a839eae2fdac9a27c26fb8567c5906e57930cc790800b0","split":"train"} +{"domain":"finance","full_message_tokens":242431,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":174,"sample_id":"sample:829626c5b02607d3cef2ad22a196b43c8be0cc9a993f85b3320704aa97d0624c","semantic_task_id":"d883db0e4c112714344aef93ac975dfe594cda1574cf277e52a0d426e117255c","split":"train"} +{"domain":"finance","full_message_tokens":88108,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":175,"sample_id":"sample:17ee8da73a0b5af2f2c2cc5211c17a945ca2012d1e266df6ec710afa73744460","semantic_task_id":"dbbe68c87f57dad971e07e622303f723894d2467afce2a47f6d06535c907a960","split":"train"} +{"domain":"finance","full_message_tokens":83385,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":176,"sample_id":"sample:24f00a1a9dfd9070f75191c8b1a65060473500c15a66ca96fa86685ef17ae933","semantic_task_id":"dd6cbbddc490c88958196ae00b8167386636a427734e31dcd463e883306cd6e7","split":"train"} +{"domain":"finance","full_message_tokens":82093,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":177,"sample_id":"sample:ab4c4bdc9a5e15ee40df60f400d6ddd071c9f391e94477e646738cb7facc6661","semantic_task_id":"e179a8e5379e372a3d9d8d5c549b5e8b0b8efe371c6926c5f66d046b0eefdfb6","split":"train"} +{"domain":"finance","full_message_tokens":236913,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":178,"sample_id":"sample:f48b1f0ec80492be30f98c12f482e52991fdcff086078c6cbd8d8c0dccd0b9ff","semantic_task_id":"e620b80a60b2200f3a9d4c647b756477097b2b31200789cff253d6cc6bdd2da2","split":"train"} +{"domain":"finance","full_message_tokens":236933,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":179,"sample_id":"sample:dcb1a0fbedd1f077f601c574eb31dce6c5725b9220e57917f18ad6d034787720","semantic_task_id":"ea0bb324def85bb93c7d535778712a84bb43ecc567536e59cc98991e0d31f18d","split":"train"} +{"domain":"finance","full_message_tokens":230860,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":180,"sample_id":"sample:edf76d0d5f1067280310515c69027f9ad89db30874ccdb20d5a84eac01cc1cf3","semantic_task_id":"ea182398338639bce0b5214c2650e7ff8c4023a42ab0d709e27100874cdec72e","split":"train"} +{"domain":"finance","full_message_tokens":77220,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":181,"sample_id":"sample:1ba8e9a63ce7c1d7db7d26b0fb3828780ac54a6ef1ab770f4f27e071d3318518","semantic_task_id":"eb664ed4ada451c3b7ad1b8b3d493750d0d7f3943d8b66f2d4704bdbb0674d5d","split":"train"} +{"domain":"finance","full_message_tokens":76552,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":182,"sample_id":"sample:ad36081d60bb88987553deeb8930defbb26fd09b07a73372a1a32f023ef446e7","semantic_task_id":"ebb737fa7804c59a5aa38c6e8e103bdaf23ca7f5a33ce0594cd91534e34fd702","split":"train"} +{"domain":"finance","full_message_tokens":76592,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":183,"sample_id":"sample:6d25436844d70dfc5eb646dcbbab779cfbb490f24cb0cb7c9d74e03cd0e015df","semantic_task_id":"ef49bd73c806c1567f7f7f350874e9a57ee2d5ffc2a4f10cad2a424fda130791","split":"train"} +{"domain":"finance","full_message_tokens":88051,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":184,"sample_id":"sample:305ce3ef2ae5a68045d5526047bc8624b2b50609a0575f32eb3a8422ff28c5e7","semantic_task_id":"efc636acdf3a66032878285d1f9a4326d7bf464fd05905375d5aeb7fbf1bb767","split":"train"} +{"domain":"finance","full_message_tokens":82036,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":185,"sample_id":"sample:5a36b6c6e4da279de489346a8b418b573da9aa2b374aeab611846f6188d65494","semantic_task_id":"f0ca2f2c99a41a4f68a7f84440e714af26fd59a3bd42af15d1175db8079e0a4a","split":"train"} +{"domain":"finance","full_message_tokens":81994,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":186,"sample_id":"sample:487d38e97288ca01a46263037a5a3e197a12c971bca5da681d453e9985163215","semantic_task_id":"f0ca6db27dcba30f6fa2ab60d52476dcdc7d194b94f8e2383205c36a4fc1071f","split":"train"} +{"domain":"finance","full_message_tokens":230870,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":187,"sample_id":"sample:41ec4be1f779414e682a36c62896cbab1407cbb9097df53c092e81bd700b5370","semantic_task_id":"f1f3e6b1054ab42fd1273452f2cce0a3e728a6c1829b123265ff3e87572b1616","split":"train"} +{"domain":"finance","full_message_tokens":82038,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":188,"sample_id":"sample:1fdaefb4342d421dcdbf067d59610fff84e6c216c76ef6a55cc2b9414ab0becc","semantic_task_id":"f223d1c422893c8eab6705d813bf02bee227f36e2776dfecd18c15e067e65857","split":"train"} +{"domain":"finance","full_message_tokens":230851,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":189,"sample_id":"sample:0c39dcc54a86c0e84c68dfe9e190f0a74ef1e522ae63a564afdb882124ea97c6","semantic_task_id":"f38b69d049ac0c41f64ee87f3130522471ebf5a043c517309ed312bbce38ce4f","split":"train"} +{"domain":"finance","full_message_tokens":76552,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":190,"sample_id":"sample:3eaa8d44534d7664ecbb48fa948fcc3c522291c117bb09c3ffa2f5b1f46f5f14","semantic_task_id":"f3e2284278a44892707132ec7ee0ab24c6903ce785278afe212034c017606c4d","split":"train"} +{"domain":"finance","full_message_tokens":82039,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":191,"sample_id":"sample:09b331ccdd52f65a1b8a23205b50991f9e93265be97f8bcdca022f6e36f38a96","semantic_task_id":"f4cc7f0748299061d5cfc26842437b02aa621b2f2da5deac9cc5b7f59405b01e","split":"train"} +{"domain":"finance","full_message_tokens":81993,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":192,"sample_id":"sample:8800c7f47e2956cc16340b5473e6dee60405643af28819aa56cc5c76f423b704","semantic_task_id":"f64a1543303f88d63ecc2c18b957d0c9bdd4c77b68d5df9cfe355853de72fba2","split":"train"} +{"domain":"finance","full_message_tokens":236932,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":193,"sample_id":"sample:7ae81f30349406776de1967c736f00b00b12589a730e24ae2b0b268500678f6e","semantic_task_id":"f720a1cddd2207b960cda908965c2f7f8177ae96719688e558a34390cfb153b2","split":"train"} +{"domain":"finance","full_message_tokens":236912,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":194,"sample_id":"sample:fec9c41561e27c6094bcb2d405a5415de35dd3ba1f7e13658b9d8e987426cfce","semantic_task_id":"f73785b71bbdda1a337089d6d450930415160188939d8123f25dab9ef1ed89ac","split":"train"} +{"domain":"finance","full_message_tokens":83337,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":195,"sample_id":"sample:6f776f94839a78c64b5fb2dfc9fa1a1e0ff765c59b12005796a018f06160fdbe","semantic_task_id":"fb8e6f6d9316ffdaf864927a07e2dc181cca02f8302c36776c5c0d551fea89e6","split":"train"} +{"domain":"finance","full_message_tokens":88155,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":196,"sample_id":"sample:ab04e010f8ba645a5d30eb6f836495454507153afda007394f79e93e3c61ea26","semantic_task_id":"fcb71ca1c26c2c0d83c27e82a736c36138443ba637ae8d039197bbd204cd957d","split":"train"} +{"domain":"finance","full_message_tokens":88107,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":197,"sample_id":"sample:5660b8b8d64482e9e3cd37c5d73dac2ff786cdad76c57656f760679ac7f7ffad","semantic_task_id":"fd3e49fa026e63c1e91cec9ed0d91e3b27e1b2af018042a7c68d339f72a0a1bd","split":"train"} +{"domain":"finance","full_message_tokens":77322,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":198,"sample_id":"sample:bc3ab386c224c5e56dc01573b983ef4c6443422902e1b4785519e87e7e0b5d2d","semantic_task_id":"fd9455d8e97a0ca777479977468cba7c7612793861d95c999988b63f4cf50bd6","split":"train"} +{"domain":"finance","full_message_tokens":236910,"group_id":"0001652044","output_file":"finance_train.jsonl","row_index":199,"sample_id":"sample:e4651a9809ce90136d8ffb7f79413985385d038d6440cc2e6bff0dd93e8ca521","semantic_task_id":"ffc4467f0e74c16eccfec2ba0aaf204070e957465e5c6b60326617d864f0fe58","split":"train"} +{"domain":"finance","full_message_tokens":254821,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":200,"sample_id":"sample:04f515bc1c70e92d9dc290d0fb6dba2d7d542aaaa2ac8ecb363fc23f09bba683","semantic_task_id":"009d1502904e8f290bd7b0ca6f9479651c8db540ea1dd1a3137d9669be6335e1","split":"train"} +{"domain":"finance","full_message_tokens":254850,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":201,"sample_id":"sample:4468094df806f1da1a5318a261205acff7423d4707e456579930f442fa03bba7","semantic_task_id":"01c5e6c56121472560a79ba5ed3665780ec1c3f9b662e50d22bc8e1405fb220b","split":"train"} +{"domain":"finance","full_message_tokens":222187,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":202,"sample_id":"sample:ac7b873d11c25af0397c432e7c7024121bed5fb25e67d31de2af0a1021749264","semantic_task_id":"024c4e26bced7941b4450a811f6736f1b2d1ded04c53240acccc4cf3a31b894a","split":"train"} +{"domain":"finance","full_message_tokens":99669,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":203,"sample_id":"sample:6db8960bb3d69a21d370ff5944db1f08780b0def09d22143663fca248045cd82","semantic_task_id":"029107c8e6e0c12fe46ec420777e762412a23445a590798a397ff66d7925ae67","split":"train"} +{"domain":"finance","full_message_tokens":103802,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":204,"sample_id":"sample:1d74bd6855642139671670aed2f166c782a151b48ad84b73a4d94fdd3c300d82","semantic_task_id":"031917280b70c908fbc6b4ea507c9cd80034820bc434a2b90661a979df243324","split":"train"} +{"domain":"finance","full_message_tokens":222174,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":205,"sample_id":"sample:e0e0eec192b54e628dcc72857c8ce5ad0d1e2858d909df64f7e21e76f03b25af","semantic_task_id":"034cf73dd1efa0ad20e274f014f65dde9a302b2a99e1125add872318157c26aa","split":"train"} +{"domain":"finance","full_message_tokens":254809,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":206,"sample_id":"sample:76c1e150adb8e6e3923d43a7615ad6c7e4200a7d627d27abe8307cbd3d8cca5c","semantic_task_id":"039794e8de74d21ca7c6f8f76f6081982b10787eb6dc3f87cb55813959336161","split":"train"} +{"domain":"finance","full_message_tokens":103799,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":207,"sample_id":"sample:072d2efb9e6a7c9fb6895e87e7e8f36068cd23afaced1aff53d68b7edc3c783f","semantic_task_id":"07a1e3f4fa8cc78264865a81a688584509411703454986e7cfc72d99bd77b5b4","split":"train"} +{"domain":"finance","full_message_tokens":222198,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":208,"sample_id":"sample:5e133631c7b95f977dd7ae1c5180e51369d79bb185e2b36ad1aa4ab7b8cf7acc","semantic_task_id":"07c4860b96f1ea09e8da9c21f8a49e82e6e9f13e672426d5407ad5a6ed89f1fb","split":"train"} +{"domain":"finance","full_message_tokens":99671,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":209,"sample_id":"sample:904d8db2c09f1cc98bc8d2cb1755bcc681f76eca48fd7f232c03359ef51fcc54","semantic_task_id":"082a2ba674ec938bd6c24225acc2b58923005db79f6252383b07949b392e5ab3","split":"train"} +{"domain":"finance","full_message_tokens":254786,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":210,"sample_id":"sample:0ac864f947a36fe3a8fe02243715b6d3853a05a5c555f911692c9164e60d79ac","semantic_task_id":"08566bb65ac8dc66b07909605332e97d0b60592f2da7f7adc2029d4ce9ba332a","split":"train"} +{"domain":"finance","full_message_tokens":103800,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":211,"sample_id":"sample:5c8c171568e2f0d04e0ccccb109247c5adad6b317565ab56cfcb6038be976c11","semantic_task_id":"095bff12307b3f13a5b9027c485c339753274390347634df1b0958633f59edf2","split":"train"} +{"domain":"finance","full_message_tokens":78127,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":212,"sample_id":"sample:0664287921380a2623cbfdd33a00311e6e5e591b2e0779cfae43b6aab3ec06d9","semantic_task_id":"0c94a6a0e3129d6d2ef35016b752a56434fc3ed3c1a79fbeeee4b9320dec977d","split":"train"} +{"domain":"finance","full_message_tokens":102275,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":213,"sample_id":"sample:a6b53749872172b9cbecd88377621cc2951c681b7f80b0f1bab715d1b5dca30e","semantic_task_id":"0cb4b1dbb7516ca05ffd1b6b735b84c861bd2e0d52102472618e41bf782f033e","split":"train"} +{"domain":"finance","full_message_tokens":103992,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":214,"sample_id":"sample:03d159bf13a8d70dfc7a4ab69e48283406aed379d6fea8b1ee04568e58fa4fe6","semantic_task_id":"0ccab38b0821d1fa2b9210ae90993353b41026ad7e41eb0152096f6732e3548e","split":"train"} +{"domain":"finance","full_message_tokens":230662,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":215,"sample_id":"sample:b32d9bbecb93c0152ddc51a0b3ece20bea51f3bcf0ca0000485e410d323ce01d","semantic_task_id":"0d04ae31377347e766d04b79219b2a0f4fbc4270d6ca1900cd214dc2798cb5e5","split":"train"} +{"domain":"finance","full_message_tokens":102869,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":216,"sample_id":"sample:8ef443205de27ab1331239ae9bdc7b9ee3f69c88050add3d7224e9b36fb6ce97","semantic_task_id":"0da709f0fd3e0f4432da507110d22564eb7841696d0c16b276320da5f4e7811d","split":"train"} +{"domain":"finance","full_message_tokens":102880,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":217,"sample_id":"sample:a5a4263cbcf98f8740ea1819ab5fe0603859a7b192b8a2668f9524820c498cf5","semantic_task_id":"0dfbedb5bbcb0e528bbfe6ab3f85cb5cd852ca0a6762f551cf1f80930f2ecae0","split":"train"} +{"domain":"finance","full_message_tokens":78127,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":218,"sample_id":"sample:587b829c18f9e5e8241afd6f60fe7eb57f2e6a2401442b2907278388422de125","semantic_task_id":"1029007dfbd088f24330e284ed4f10ba20b75183ab3fd579bffcff05404b2c5f","split":"train"} +{"domain":"finance","full_message_tokens":78127,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":219,"sample_id":"sample:f336ef33ba56ad99bdf781eac70bb9f2b92cf9a81b117f504d34d0d17f40f76f","semantic_task_id":"1166019a55e42b13478a68807f69d30d91075fd2c9160aa8f60633725203a307","split":"train"} +{"domain":"finance","full_message_tokens":75593,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":220,"sample_id":"sample:c2405bef9fd1bc34b2744adf5883f58c86d93774a654cf50aa3567351c495be3","semantic_task_id":"11f69bb93f492664eddd387eaf5a88c78a5280df6dfa224a6d8f0901a55919df","split":"train"} +{"domain":"finance","full_message_tokens":77211,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":221,"sample_id":"sample:758a1df8c9f2f533d35826f0b4bff0d98ad1f44827a5967258bedd2ffa3597e0","semantic_task_id":"12c6a121b7be1365fa33126e558b999fbff2be3335377ce7c0d33141ba626456","split":"train"} +{"domain":"finance","full_message_tokens":75569,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":222,"sample_id":"sample:e5473fe72b0eeb85f3b3de0d5645d110a22e7c775160f7b5496970e0cd186fb0","semantic_task_id":"1337e23ba8b06847ea97bb79f7cd3c7bbb7ff4bb2c850bb324a77239ea7a8a3a","split":"train"} +{"domain":"finance","full_message_tokens":69738,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":223,"sample_id":"sample:76f923dc26fc2f8dc843eda4074c4b437ba4fa6a23cea0ee09d804b4c08b2dff","semantic_task_id":"143cc0b0bd3e9beafdb7d89657cba0b5497cf11a511cd7e085dfa39157768fe3","split":"train"} +{"domain":"finance","full_message_tokens":102274,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":224,"sample_id":"sample:de79662c08206af01d64642e5fcb291946ec09723898316100eb6ea1eda614ec","semantic_task_id":"14c9a632248476bc2db400ae1fede8d7e749951366b06927a55369b558d3e6ae","split":"train"} +{"domain":"finance","full_message_tokens":254799,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":225,"sample_id":"sample:59ac009f49be34ec772492fcd378d238d5e97469d879ba5e374cdc1b9b5f46b3","semantic_task_id":"15c439e6fa06dd987fd8e98f44a75646defc9918528a219828574a5136fdef76","split":"train"} +{"domain":"finance","full_message_tokens":103990,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":226,"sample_id":"sample:c64202b53307cb410d0dce5abbf142f2abea2c60415e695393c654c54b47700d","semantic_task_id":"16e27879ada982b85b5cae86927dd0d232f918693a3b46ded2807b3db0f6d20d","split":"train"} +{"domain":"finance","full_message_tokens":75578,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":227,"sample_id":"sample:51680bbaa9c6f6275f3b06dfdcf5495114ec8fd97481ebff8cfea369a62672c5","semantic_task_id":"17c93c90a482ae2f260cd818946f5046d9710dc0b9a98f51fe5e3884902de73b","split":"train"} +{"domain":"finance","full_message_tokens":230684,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":228,"sample_id":"sample:129a3fcca8c85b16b255beddc5f1b95ba6ccffa463e34744762d562333087e43","semantic_task_id":"17e1c0a6a5c716a395ee6cfa9d42230f373896761d1554a0871a1b07f5368f0e","split":"train"} +{"domain":"finance","full_message_tokens":222236,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":229,"sample_id":"sample:78641db950cfbd1cfbe4a1b2d883103c69c17084aa065e9be87338ae9f419f4b","semantic_task_id":"17fb21e6f780987dc53a97711dc4d81b902906cc79841c0a29ee0484bbdff207","split":"train"} +{"domain":"finance","full_message_tokens":230673,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":230,"sample_id":"sample:8b4f22cc7f15b0af792af98d6afa4771d9b2f0b1afa650125dc20911ffec1cd3","semantic_task_id":"18936ec326bb345d3159d267284c465ae7ba36bde3a94be105569bd4dafd7bbc","split":"train"} +{"domain":"finance","full_message_tokens":77211,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":231,"sample_id":"sample:883c5be5dfd82d8f48cebda6111425142ede53fd881d72716405b7a3088c96b0","semantic_task_id":"19a575933c88494941cdd8b7a040ed393e58dbe6abda4d2b60ba5558e99ea279","split":"train"} +{"domain":"finance","full_message_tokens":103801,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":232,"sample_id":"sample:6244e6e493d88d049d1bdf53a1063361b0ec31678ab6efbdda5bd2975b8a450b","semantic_task_id":"19db14b33db19fd111a528618e9c08fb633d702449f7049aa1157d5b79897b8f","split":"train"} +{"domain":"finance","full_message_tokens":102275,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":233,"sample_id":"sample:94e895bad41c82b05e0a473199d09bfc7b498918f8c77f5fc6f54188aa0a139d","semantic_task_id":"1a3d1ce9d4661c3b77e2f81b00dc26404a159810dd825297c3a629809cd17ca5","split":"train"} +{"domain":"finance","full_message_tokens":222176,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":234,"sample_id":"sample:6a079a1cb23464eabc70457d65ebfcc8d91047ebcbeeaec92cff4a48dc77ab78","semantic_task_id":"1aecc58b8b6c3e299ad5787309ebe4ba75529c632955c51ee8f4df4e725071b8","split":"train"} +{"domain":"finance","full_message_tokens":77212,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":235,"sample_id":"sample:8de315c24b1e60dbd81b779fde1bda9bdf9475aa27814566d63ff2bc449675bb","semantic_task_id":"1f705dbe874779dee041392b86ed39100b9c5443f8bc04c6754d3fe26990a9af","split":"train"} +{"domain":"finance","full_message_tokens":222197,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":236,"sample_id":"sample:58273077659a363684ce88fa8af2df15c22322ea3d87f645d6528d54f1f210d0","semantic_task_id":"21951a9990d6b1370d97589fc7f53dc32de88f33e30f026d7727cc1ffec7827c","split":"train"} +{"domain":"finance","full_message_tokens":69695,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":237,"sample_id":"sample:e0c3d5fbe2abbf9a26944b8995c868e35a26941d6b72bc37c64eadc7f8e01eaf","semantic_task_id":"244d198045ab97407f2db4193b1e03be82b5a2150440b8f24f0a97ad783b1212","split":"train"} +{"domain":"finance","full_message_tokens":102884,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":238,"sample_id":"sample:3adaeb920180de871f0885c5eec722c30314808d915c10df33a05b0353b610d7","semantic_task_id":"24c9d8ee5fe5eeb7d446b0f3d1450d3b0229768e2e3ca5f634dd3f87a3115168","split":"train"} +{"domain":"finance","full_message_tokens":75522,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":239,"sample_id":"sample:7a26b320578453034e97d2abba99b4dc9ad82cebdf2c2f0cb355c43ade28e3bd","semantic_task_id":"274b7d10708c6983a2215c904a7e9efaf4df25db4dd86ce3b8d2ea04bca29765","split":"train"} +{"domain":"finance","full_message_tokens":78127,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":240,"sample_id":"sample:b47b9595a3a87f342e5c005f10b8624290fe891e084ec31936908ae2516cf117","semantic_task_id":"286832b8e22f4e7aa91b1a77c9eee3c57639f07d835ce12c37f639269b276144","split":"train"} +{"domain":"finance","full_message_tokens":99665,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":241,"sample_id":"sample:26bbd352c5c3fcaf9bb9dfeb5803f343d9aa1f4a1e68b43f006858ecaaf40f6b","semantic_task_id":"298d9927472cd286beff59a7e7d6ee91c191e15b8ce9505b1d9b44bcc818210c","split":"train"} +{"domain":"finance","full_message_tokens":254828,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":242,"sample_id":"sample:90c3cd4f93e1a980f17f3e30385b9c4b6281456bf7a251ee77b22f91c14dc5df","semantic_task_id":"29d0957a137a5d1fe47f26bb2fa5cc62299f15a84ed35a1481dc303d35b273fc","split":"train"} +{"domain":"finance","full_message_tokens":77251,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":243,"sample_id":"sample:58d282d2e27f97bfa69737aed84ff12cfed82a1678856286cb7ac4a389d8cd9b","semantic_task_id":"2b9554cfae3e4ef9dd3fe322e3e5f857fe747b3eff112982c2bdbcbd712ed648","split":"train"} +{"domain":"finance","full_message_tokens":102884,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":244,"sample_id":"sample:7544671dd3cc6be62d01428b16dfc078467de51c5587077b3d31fdb5f52e05fa","semantic_task_id":"2e78cced714e9b7a0982b946159341f9ad43fdafb9f254eadd5178eb89de5eda","split":"train"} +{"domain":"finance","full_message_tokens":254819,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":245,"sample_id":"sample:ddc1150f0ff93f54bbbf3bbd544099f8eb30d192c85256ea5210916648ddb4d7","semantic_task_id":"2ea52d3728067a2abcc6b9e0b31b3e1de13680c012082b60e02938e4e63d0f37","split":"train"} +{"domain":"finance","full_message_tokens":69709,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":246,"sample_id":"sample:29f1d331953bff5ffc2932045e780aa7912c0c9c7bfca589a40b0b2d13e86000","semantic_task_id":"2ebeb87508bc1fc60ac638f3bfb83616f8ef3aabad0dabd18dbfa7f03436881e","split":"train"} +{"domain":"finance","full_message_tokens":75522,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":247,"sample_id":"sample:0a38451f4c5e9e32682c4dc3d037f6e2ffc248661849bd4b8896a1c79b695531","semantic_task_id":"2ed36e7c09cea179182ac12b5205e08560ee9f3ec41b031ca57ecde3ed7770fb","split":"train"} +{"domain":"finance","full_message_tokens":75523,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":248,"sample_id":"sample:3406456e501f89594425136e089bf3e21d5e7cc9dfefcef742b331f877ad2ddc","semantic_task_id":"2effbe7693e17090065a53f2e4d64c663b1f2e86a8806c26af1ed72b0a66a498","split":"train"} +{"domain":"finance","full_message_tokens":222175,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":249,"sample_id":"sample:6f8206043089a537921334d5e35cab53c92d19f86374f8a15793b22f251bb49d","semantic_task_id":"2fa1a5c220a49b780f34893635f40272e1e1c725d5c56c0fde4381d15643a30a","split":"train"} +{"domain":"finance","full_message_tokens":230663,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":250,"sample_id":"sample:b2900c8125322c6d571e12c3cb149dc7e971ed88e4da611346f4eb74fb661b3a","semantic_task_id":"31287e413e6e04f9dfc4594a2ab5d228925d4dc4461ce6ca4418c16d56b5eeab","split":"train"} +{"domain":"finance","full_message_tokens":69642,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":251,"sample_id":"sample:2ae0648c4a2b9f362ad72c3aebf11de2bb296924ed7603e2c6c053e2a03bd807","semantic_task_id":"324c9e18cc00e3d43f89d50354b90bdeb777e0380cac7f987d97c50f051fcf2e","split":"train"} +{"domain":"finance","full_message_tokens":254820,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":252,"sample_id":"sample:c082a551b03195f65c16569b1a95934ed0155165171c526abc5ba6f863d5b87a","semantic_task_id":"343883a31fd2ff009c461a493a251204a2eebc93a402529a145957b4a51d5799","split":"train"} +{"domain":"finance","full_message_tokens":222196,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":253,"sample_id":"sample:fc1cf9130975634a3a285e8dd40270e2d9a544b557e1ff288f605b398d042c2c","semantic_task_id":"3af4d8bd4c538fe2d23cdeccfd8a515a1cad723c3ffa9ce9d2b43e160087ae92","split":"train"} +{"domain":"finance","full_message_tokens":254810,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":254,"sample_id":"sample:a3646ee943f15197e98aaa9902bc85b1b8f38298ce596606025251adaa15be54","semantic_task_id":"3b96066091c6677fa17757121ad147c22e6d6244ba74b9a79edd7effcc7fbfe6","split":"train"} +{"domain":"finance","full_message_tokens":77282,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":255,"sample_id":"sample:7e62cfc917718fba51f1b1d16d1e826052f07c314ba9d0883dc209b33dcb070f","semantic_task_id":"3cc2691ec63388865bdaae65b11ccac9bbd3c673c9b3ebb45e314409059f0c0f","split":"train"} +{"domain":"finance","full_message_tokens":230662,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":256,"sample_id":"sample:c3749f0d760537497eda2608bce0e3c70ee779c47c44f146a1a72523cb9752cf","semantic_task_id":"3f14968b81749bf6074061039271add0c5086f002e589b08878fd0c33321acd0","split":"train"} +{"domain":"finance","full_message_tokens":230663,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":257,"sample_id":"sample:d9110a0f19a46a72518fa1cbc28ef6ad2b6f20f652b4d4e8f29cef195d9aec58","semantic_task_id":"3fef36cea9bd74112b41b253b85414d646c121e5670de94e083a5837ba06902e","split":"train"} +{"domain":"finance","full_message_tokens":254787,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":258,"sample_id":"sample:37400a4d4a8b65fd9bfdcc1991cf6e98d845a61829f51360dfdacb0c3a66821f","semantic_task_id":"40252aff57f3d85d3beeeb5d794930c386091076278280278cbf5cb7d47c0b89","split":"train"} +{"domain":"finance","full_message_tokens":230725,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":259,"sample_id":"sample:5758b7042bf5523541e497d8ea5219c27c1263d86ca03b609ad65980b992a2c1","semantic_task_id":"41f049cf13e275adb75eaba0cf5764e5ea113cdef79a56164f4bffb8d6d12cd3","split":"train"} +{"domain":"finance","full_message_tokens":230663,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":260,"sample_id":"sample:fb3729dfdcc97771e89b858b2e41c3c5b94a2f2c9984b06675f72df1665a5d4c","semantic_task_id":"424d35d64ad0b0a7fd5ece744a1c1121c1e96119c56314194ba1aa52685feae7","split":"train"} +{"domain":"finance","full_message_tokens":78129,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":261,"sample_id":"sample:348ec7aee2455cb599cf9091da7694a881a07515f096c4d537b890f1e569977c","semantic_task_id":"43120de4be7d0c11e7ece86393fe60bddf1f9499d44498072252ed7ec084344d","split":"train"} +{"domain":"finance","full_message_tokens":102879,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":262,"sample_id":"sample:e6fe1bb7cceb6b96808ecb315bf78d9460438c0e0f8701f85c5d275b5c942b1f","semantic_task_id":"434af71803b1c82700c8f0991ace7caaa4bd031904a5b32a7cb821987fdd1d7c","split":"train"} +{"domain":"finance","full_message_tokens":99669,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":263,"sample_id":"sample:a898f43cf4aebd9f70ff88b7e15b5bcbe650e2b745da6b999d829f3724eb2ba5","semantic_task_id":"44c554f40ac3d32c1a45d8290a376c25cbd1c5958b7dc39f0b7f4c5d46d79733","split":"train"} +{"domain":"finance","full_message_tokens":101359,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":264,"sample_id":"sample:9d1166c07055ab2bdaeda0c941dd2935210e7cb31517afe6829f0f836a0a77c0","semantic_task_id":"450af3d3bb9b720dd00570db4470e6e4598eaa8f789b8927cbb33a6c72029d6e","split":"train"} +{"domain":"finance","full_message_tokens":222175,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":265,"sample_id":"sample:757b5132fbe36c1fcad69df1994e6d72910c90e861719f0fe6cffaf970c563b5","semantic_task_id":"45733f146baaa5e2829decf76719e24d4ffce4e598fd0f057dd75dd293d23b78","split":"train"} +{"domain":"finance","full_message_tokens":77212,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":266,"sample_id":"sample:dbaf23c9d50b24927fdcb80249d15debd12c248f692fbb2b4211b2f28611e12b","semantic_task_id":"45a2f521d5321956fffb5710244990ac75a4ed2b27ecb82612a234225b45683a","split":"train"} +{"domain":"finance","full_message_tokens":102276,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":267,"sample_id":"sample:8e37f527a9b1d246d373041f21c1b0230736093825bff72bc27623cf74fd6e1b","semantic_task_id":"46069949ac647c41c61e8bc3b02b376bee1ffde6d007edfaa7043e5408da9a7d","split":"train"} +{"domain":"finance","full_message_tokens":99669,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":268,"sample_id":"sample:c5c20c0ec929cfeabc15c9fd4945fadd0608455d5b8b196d23e9919110030ee6","semantic_task_id":"48b477db27749800ae94814f26fc82d8b05166cea8fefdec1fac6722c9256c87","split":"train"} +{"domain":"finance","full_message_tokens":254824,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":269,"sample_id":"sample:2b2d6d0ef25b88739b0c48d7a5e68b9afd09c91d832069a105e3a6309a803c08","semantic_task_id":"490964b5aff93642862fdd18b7c6d21222fd0ac76ded2514db76411c4e87d9d8","split":"train"} +{"domain":"finance","full_message_tokens":103995,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":270,"sample_id":"sample:be00eda28c368a170280dc0819d103b8235e4b98d0de1d299dd84d3cd4118b79","semantic_task_id":"4bb0e914796667e1c1f5a522c573827a5ad0d266a9bb50a0a16a69eeaf3b5360","split":"train"} +{"domain":"finance","full_message_tokens":101358,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":271,"sample_id":"sample:2afbfb641f72c8d467a2f00dd2da3e50655971082ab796ff77a24a66e3f69a06","semantic_task_id":"4c126fe8084c05c6d8e924cf75eea0d65cd5a60e2b96d58edffd897823bdcd69","split":"train"} +{"domain":"finance","full_message_tokens":69680,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":272,"sample_id":"sample:965668e13dfb94818049fd5561404be07717aca9847151aefc3e6cd568e7f781","semantic_task_id":"4cf08d5f107aa7781e1ca49e19b4f71d6c0545d85fa3309521d23a33877ac184","split":"train"} +{"domain":"finance","full_message_tokens":99655,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":273,"sample_id":"sample:49706eec3ad6a723d7e5503882f470530d1f1babcee5a6580851c0ca85eec836","semantic_task_id":"4ec496ca2c85945b38e9b3922147e0fae3b4f1e0982beb69a984be66dc56959c","split":"train"} +{"domain":"finance","full_message_tokens":77268,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":274,"sample_id":"sample:472e9dce552a7014226ef483ed94d444304ce62e27b6528eb35fd3667158f7ba","semantic_task_id":"4ee23e542b2858204baf1c39b9b958d9364f7c1589b37641adca06147b2b0ab5","split":"train"} +{"domain":"finance","full_message_tokens":230662,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":275,"sample_id":"sample:7e31b0074bfd5ea23c044a833cc028f5c591d9586053190eb472553010efd69c","semantic_task_id":"4f8ef8e0a20bcf7af8a9ef4e5f5cc90830a863ee1dc4edf2e19c8484b9381eea","split":"train"} +{"domain":"finance","full_message_tokens":103801,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":276,"sample_id":"sample:1692f0e870c682e03c97d24ce71f21b0e7c45f7fd7cb4c2ed03108f3152a1120","semantic_task_id":"4fb16040e969653d6c2f5875f68b55648d6a2f0725a845de6aff22d5ff9bf6fc","split":"train"} +{"domain":"finance","full_message_tokens":254810,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":277,"sample_id":"sample:7f71fd52ff03e79796ddc8cf82030d6a39d2e716a31a24251ea96b2df1a550ff","semantic_task_id":"507449d6c6338cf59af96925fbf40cf61404e7e395c732dad938467e184eaf74","split":"train"} +{"domain":"finance","full_message_tokens":254786,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":278,"sample_id":"sample:f15a112d11c8c9ae50b451c59c8d275c847387d8722661465ac9738918bc6730","semantic_task_id":"52d0501d3340bd435af4cfcb4e3ffbb93b0990e20fac25e0e35158ace491e175","split":"train"} +{"domain":"finance","full_message_tokens":102870,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":279,"sample_id":"sample:69c06b12141466fa9e44c278f8a867ddf64417f88f5a2a8783755bef87797bd7","semantic_task_id":"52faa47c5e2381a28fa4fd462cfb395004cdc18846bbffdf5d9f0962a72f6391","split":"train"} +{"domain":"finance","full_message_tokens":101360,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":280,"sample_id":"sample:b3bf392cdc6904116debe8749d899755c8cb1dc7d1bfe5887e1876daf649ce7a","semantic_task_id":"53900a66ea491167faeb73069e0031c9d263a4c8a39511880ba84d5d4b6d6c17","split":"train"} +{"domain":"finance","full_message_tokens":254787,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":281,"sample_id":"sample:1236bfb4a9c80f364b79f430436563ca3651126cb4e302d8f566187a21ee5a44","semantic_task_id":"555c327d2f28053af2b83f57c69b664ec6a3d155fa0d27c78be7294175f1cd36","split":"train"} +{"domain":"finance","full_message_tokens":103993,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":282,"sample_id":"sample:f6a07a50ca947ffb2d7802dd541bccc5f803d99ff812db6789fe017e5bad74bb","semantic_task_id":"556cf1024b5e48a0094f43c4ef5b5c30ea394f40801b0b633d1ee5ccd2bc2679","split":"train"} +{"domain":"finance","full_message_tokens":101358,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":283,"sample_id":"sample:2e6fd1c9c5768f5fda5774289d94fa47b7ae17be48e6a190655fb73b03012ac9","semantic_task_id":"574647c81de9f5ffdbfa73e6a86f3f0877bfd0f79d5a706accffdc2cd03d0b14","split":"train"} +{"domain":"finance","full_message_tokens":230661,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":284,"sample_id":"sample:931f3ade6c74c1abd431b1ca41f33e99d168cdd0bd0fafb3e4c9a1448b1aa00e","semantic_task_id":"592589da5904337e90f57639b94ca565388e32ec627636d6e316b24b4adaf01c","split":"train"} +{"domain":"finance","full_message_tokens":103997,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":285,"sample_id":"sample:8e5507d473eaf2ece464ba0a981119e9b7031e3ad3651612e3590991580d3728","semantic_task_id":"5b4fef9a0d77c0c3e4f29b6a753e9b87d82a89c49ebb02a130b2d8c0e007b037","split":"train"} +{"domain":"finance","full_message_tokens":103800,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":286,"sample_id":"sample:b694ad20152dcb3672c6b3cfc93d216f0e981de2ff31dc968122606e3f8214ae","semantic_task_id":"5c843da79040dde4bf3fe1378392f943eede90779cfb4d3797799c9ab571c205","split":"train"} +{"domain":"finance","full_message_tokens":103996,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":287,"sample_id":"sample:b5d536967c4e255840986aeafb202d2574668c644b231e8b20349f63f9746cc7","semantic_task_id":"5c920765bcb355d7afe89f2ec51966021c5b88663de222ba333e945710f8883e","split":"train"} +{"domain":"finance","full_message_tokens":102878,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":288,"sample_id":"sample:4f44fdb27559e626d7abf95462d9d332c1706e4dbedddc0da6fa1ef273700f73","semantic_task_id":"5ee954f031a22ff352fed912154bc97e849841b38933ba177ec4f83beea1d6be","split":"train"} +{"domain":"finance","full_message_tokens":230662,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":289,"sample_id":"sample:c95a55617c76f772640914e655923e515219141667f39150430b947ff3cca657","semantic_task_id":"5f8b0aaa64b82ec3a18c70cd0e42dc3ebb2319248fa4228fa12086acd652b541","split":"train"} +{"domain":"finance","full_message_tokens":78175,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":290,"sample_id":"sample:300d508b7ff0b0cdc30fde58c05404b38d8718d6b73b7779f6d56334b35251ae","semantic_task_id":"613bcb8055b8282e4ece7b8a34ed233cbacf9f5518c443dbb439601a7a90856c","split":"train"} +{"domain":"finance","full_message_tokens":103996,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":291,"sample_id":"sample:103cf47b11ab761879dee21fa7243671c3cd0c40c4b885e0ff186316e3940799","semantic_task_id":"624eeee892db58a287c3c7471f111740cc10a4f3f23690d1db1e470ae3324925","split":"train"} +{"domain":"finance","full_message_tokens":230683,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":292,"sample_id":"sample:6e4b82b9b4e4811b8e4803bcd983e046a4dfcf5ddb7fdff10729ea27c12cfc23","semantic_task_id":"632988140f45dd15e21d273a1dabaa8c668295507a7af53c2829110eaff9937c","split":"train"} +{"domain":"finance","full_message_tokens":69639,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":293,"sample_id":"sample:8f28e98c37ce04f4990ad65382b7b33690d3faca580b53b742764e2399a80936","semantic_task_id":"6349debaacdb0ff231e9cdb39b3057b9b1eea1d5b6db40c359910f5ad1b25084","split":"train"} +{"domain":"finance","full_message_tokens":77211,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":294,"sample_id":"sample:4e51a7fb5ae9f2c0d6b84724160c5eb92d58779bf2ba579bda899fa384ba424c","semantic_task_id":"63e5610ae789a06a1f47260c0e37feb47b1da47298eedb8b453cd2930ab99506","split":"train"} +{"domain":"finance","full_message_tokens":222196,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":295,"sample_id":"sample:23eb4bb60b13ea372738a82f1393761f69007a30768e5664a82a6957e5202be0","semantic_task_id":"6478b271cdfac761be7cb6ec6dc312698d9bd714732dcf61518deea511ec3267","split":"train"} +{"domain":"finance","full_message_tokens":254787,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":296,"sample_id":"sample:180e886189bdbff04d041e0cfaa2b11d51f6ed91a37c494860416d03b4046933","semantic_task_id":"6768a3e59a41ea18567b2908ff8647d1f84e9404d4315d7e686b8f6c4d9ed92f","split":"train"} +{"domain":"finance","full_message_tokens":102885,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":297,"sample_id":"sample:51e3cae170817a386646b6c894e3d1e1cc105516af16521867bc4d0b703c6230","semantic_task_id":"69632d25db77d1e2931668aef5848ea66912651bee1538623dbb7c1b32802056","split":"train"} +{"domain":"finance","full_message_tokens":99669,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":298,"sample_id":"sample:90c2430649959662ebeab45ac530d08f61c5cb2dc2e066679b652e75a1604aba","semantic_task_id":"6ac71d57e1d9d6cef65722c704a157ffe147d4f2afc78db2994974f9d9d5c62d","split":"train"} +{"domain":"finance","full_message_tokens":77212,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":299,"sample_id":"sample:d46248e4eaaf58d1a91399cff8ba6f9dcaeec4c1ceccd0581461195a7f286b18","semantic_task_id":"6ba43058bfe63b31201647b41fa6e083e34428c3bff97fb6541bc0490a264d4f","split":"train"} +{"domain":"finance","full_message_tokens":75522,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":300,"sample_id":"sample:f1a0a678e8d6ca16588c2e266cba291f5e0fb25262ca08e986bffb74f829b0e6","semantic_task_id":"6c66f8db132071da3d8e59e1bd39428322cb48fe92804dc80433b3de3e46f26f","split":"train"} +{"domain":"finance","full_message_tokens":103997,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":301,"sample_id":"sample:68f4f405ed20e4ba5e73c8883e1efa42504aefb81f47a52182a578a680ca3cdb","semantic_task_id":"6d42296fa7f308bc3f05d8143373e5a33333ceb8f9a74265f156ecddaa9f888f","split":"train"} +{"domain":"finance","full_message_tokens":102885,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":302,"sample_id":"sample:38ae970b2dceed714abe89368b41c5ebd5c35a8694fffdac947431f7aac69270","semantic_task_id":"6d9df5252fad45fa660e06fd18ba718cc86db5e38e63937b9bd53be0e15be2dc","split":"train"} +{"domain":"finance","full_message_tokens":103801,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":303,"sample_id":"sample:01c6aea49d11e2fc1ee53d05a6505653a2878bc83c28f2f13dec226f3c090670","semantic_task_id":"6e6ac21bdb78b0ad110a2181cf4c7948b6cd57d60b5f702eeaccd1f64c3f9ec5","split":"train"} +{"domain":"finance","full_message_tokens":78128,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":304,"sample_id":"sample:200bc38bee82eabc18f1dccfc94f42ff179f68d36edd045f080d23a4446e634a","semantic_task_id":"6f45c98d9d8914aad5af3a8db267360fae3838116eecfd2f970fca2ef41dca2c","split":"train"} +{"domain":"finance","full_message_tokens":102277,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":305,"sample_id":"sample:b649b70814a8039d6989dd15ff62815ef58278410561d8f879a1544bb0854841","semantic_task_id":"70dad9a735aab72d434fa4bacd3b078ae816aa74006b30224ef8f62dcc6ee9ff","split":"train"} +{"domain":"finance","full_message_tokens":99656,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":306,"sample_id":"sample:c957c61c70b73ada2b12dec6fa281e3a18a4ae1bd470b1a47d7c99429a9310b6","semantic_task_id":"70e081abe8c8e0baa6623e3366bccb2b2ac7f09e7d95a4fa9f98c8de2e6d2014","split":"train"} +{"domain":"finance","full_message_tokens":75561,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":307,"sample_id":"sample:9427ffe489faa32dcdbe15c05a4b0fbdb11b3567419170661221709c2a557cf4","semantic_task_id":"71d95981ef6111bce56bce5fb86e9a7601b3ae1e3dc8a08fa01418b0332b37a3","split":"train"} +{"domain":"finance","full_message_tokens":230662,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":308,"sample_id":"sample:0fc4dbbe4a3519a1441de9e31686831933a08f638e855706e08e2eab9eb85f7e","semantic_task_id":"720d4c88b48c5edc42abb6c9e39a22bccde17d5390f057d2da0dd66985f205a6","split":"train"} +{"domain":"finance","full_message_tokens":75523,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":309,"sample_id":"sample:06b37bbae5daec854d2e6b80ba9d37c79fc74baa6d673be0df5710398df41147","semantic_task_id":"723a174b4553240b892e547e1135f3c8b1f6c523e297f357b6e53eed7b50448d","split":"train"} +{"domain":"finance","full_message_tokens":77212,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":310,"sample_id":"sample:ae116ffb5fa28bd1ddc2158a7e08982863f7644be1f6b78dca8d8225ec23bd8d","semantic_task_id":"72afc34cd2000ee7abbcfeb8c4310fdad3b43f2f9d2ec4f1d20028248a3e2851","split":"train"} +{"domain":"finance","full_message_tokens":101361,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":311,"sample_id":"sample:567f6028b85e428bec8ae5cd34089cbcf7bf5143d824d1968213da0151ce7808","semantic_task_id":"72f89c685a732db9fee0df8a319578aa8629560c6c1b1e6f16a7fce1b1eb5c49","split":"train"} +{"domain":"finance","full_message_tokens":69640,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":312,"sample_id":"sample:c47e483f08125a4a18cd2ab51628fb17d325759d62c1f2c7d6d8563c3fcd5b10","semantic_task_id":"738686f780eaae05b97fd3ea999699ee4c7d0abb32bcd6e3e41b95d87d803291","split":"train"} +{"domain":"finance","full_message_tokens":102276,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":313,"sample_id":"sample:16b115df079dd4208f0be16dc8997718f407a0bc5b35c7a5ce321a72275d6b10","semantic_task_id":"73ac4af7d9bf34fdd849fb9bde7f773ea8345bdb1c7fcc0a6cb9e6b065d81029","split":"train"} +{"domain":"finance","full_message_tokens":78129,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":314,"sample_id":"sample:d4c78e4666070e4a2c77eaf33cc0cd7d2ba46952a005c15b53d7d3f4cbdbac19","semantic_task_id":"744077e8a1e2b825c851aaf791b3795850bc76b064d3cc3098a71e047182fe06","split":"train"} +{"domain":"finance","full_message_tokens":222175,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":315,"sample_id":"sample:dc3a814c595043c57786adf2b133f905493b8d94a6c4c5e6d8185fa4ef4b5d1c","semantic_task_id":"7508dfa82750a1b35653d9573329f23aea96177187e388443c92429b0a482cc7","split":"train"} +{"domain":"finance","full_message_tokens":102882,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":316,"sample_id":"sample:8b54aa5ace2df71ea999019eb598b92fccdb86d2d9e7e5a91e7025cb7f13c5e3","semantic_task_id":"753163d7297b72dd2c9950e8414c7be4fe8a065f5ef4bdeb223ce833b5d17716","split":"train"} +{"domain":"finance","full_message_tokens":101359,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":317,"sample_id":"sample:d194e0ef55bdf2da4a112255cf7ce0a4d2e3bf4bcb888e9f7e693e443b8a9a20","semantic_task_id":"7736b3fc19317357ed911658363d6d7ec74f4233ddec3d37a9e08b135c24d51f","split":"train"} +{"domain":"finance","full_message_tokens":102881,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":318,"sample_id":"sample:c10b4d6a53eb81f997bc94189c2ba92847b39be56360def82334465164dcf8b3","semantic_task_id":"779e110fc2ba05a01353c320f43d6db91b07d331297bd5e7cadecf0cb10fb14f","split":"train"} +{"domain":"finance","full_message_tokens":78168,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":319,"sample_id":"sample:ec5eaf6350c2fd876148e99d478c8beb06c3ebe30a39b300977868e455794f46","semantic_task_id":"784f175b4119ebc2405300d1ee0e899eb044d9885cb537fcb18daffb585323f5","split":"train"} +{"domain":"finance","full_message_tokens":222177,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":320,"sample_id":"sample:2cfbb7485c90cc67158a9bf4095db01359fc5515c507ee7b4cb2a52ac1c200c8","semantic_task_id":"794d548235d7589de0b21df1be07921f6c9b2dc4e8c5308210b4b11fb9c38ece","split":"train"} +{"domain":"finance","full_message_tokens":103981,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":321,"sample_id":"sample:29d8ffc47aefecbc9c86956058b3dfb2d5dd63ad35f0b6b8a1fe783cddb84246","semantic_task_id":"79663255f9abef328d354358120e938b70d8b3ee92a066c2966af98bef7607da","split":"train"} +{"domain":"finance","full_message_tokens":230683,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":322,"sample_id":"sample:2cfa0712c92979e3a55d26ab0d2b50843493eadd365cf1be5ac97c2a3a86b877","semantic_task_id":"7a105ab7cf9048a26ae67065d603fddeb2d8b7dc2c6ec5db4690df6062301351","split":"train"} +{"domain":"finance","full_message_tokens":102276,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":323,"sample_id":"sample:cc66dbac1d3d83ba10cb83e0545a546bceac5c645bb8bd811913322b06adf306","semantic_task_id":"7aac847028c6d6ab0011d0277e62d9ef453744396ce7b30e64ba96d6c0c26cc4","split":"train"} +{"domain":"finance","full_message_tokens":69640,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":324,"sample_id":"sample:1aa65472f1623515ea8a9e62ef9594d7e14e265413cb6a9115913f62a4fcd9a6","semantic_task_id":"7b5f94168a3a5f1c674a34034eaa33259c4b524a50af605e201de3d3b5d6270f","split":"train"} +{"domain":"finance","full_message_tokens":254831,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":325,"sample_id":"sample:bcc5dc211a933cf3026b9c37c01cb58c8f78027e431db570afff2464db66bffc","semantic_task_id":"7c6a36dbd811cce1fa22ffdde8e11145088ea3a165fcc1ee15b5c313e761a4f6","split":"train"} +{"domain":"finance","full_message_tokens":222195,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":326,"sample_id":"sample:6b10ad93d11eb2e84041832b1ea5c2fa6831cb3e72c2a7df2107afdf07ceefce","semantic_task_id":"7d3de1890d7913e5717e51f624025b66b5bfb4bbd606bcdb9b6014dd369b531f","split":"train"} +{"domain":"finance","full_message_tokens":103992,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":327,"sample_id":"sample:1db29a558a8e3405440d2428074af3132e606f886fbbaf0cb17586d8aa9f03ea","semantic_task_id":"7d6ef900f2b01d93f2c3e6f5707775c8e08cf52b42fae70b3c7db17f919d3f56","split":"train"} +{"domain":"finance","full_message_tokens":78127,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":328,"sample_id":"sample:ddb5f297205b410a9e5a9d3fe10de3942331a4cda5c8251901b9195a77065749","semantic_task_id":"7d83ffb3c4497b98adacd7ddf4ead9f881a7a6dfe0f716b8f4f5f727fdee0055","split":"train"} +{"domain":"finance","full_message_tokens":103997,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":329,"sample_id":"sample:229a3a55715c0f7ab4de6923af23bd7203b41fd9ca7dacebe155b32c5af2519f","semantic_task_id":"817ad881bcf90f3ffbff74003caeba010438d48551528842380baa6edce1aa16","split":"train"} +{"domain":"finance","full_message_tokens":230684,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":330,"sample_id":"sample:7345eb0186ff5c41f0a6f45560b379b415ca6b596bbe395a83aa85f9f546e464","semantic_task_id":"824d55e53b9e456666f4dc4b6ef2139c647205ed8eabb60e6d6d3da4fc9ff748","split":"train"} +{"domain":"finance","full_message_tokens":99666,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":331,"sample_id":"sample:558a19421b61d675b2902c67417913b8c69bfb4793b39fc9befc85d6015d81bc","semantic_task_id":"8352bc122b5e7e84eb013a43a301c9e65ed83c5399708f0b3fddf4356a458b1c","split":"train"} +{"domain":"finance","full_message_tokens":69640,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":332,"sample_id":"sample:b971414e0036a91bc895b035aac44a2a94b2270db71d44ee2cbfa20c0fd60d9b","semantic_task_id":"8381af023c88dc38a121aa8c879261b3271e97f41e293e257c28bc3dd5020f06","split":"train"} +{"domain":"finance","full_message_tokens":103800,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":333,"sample_id":"sample:b83af89de4d859343e99b479e5b417b0dc04f365039c7d1f9534968442b79609","semantic_task_id":"876d91d48d1ae973d043499ff3bb68ccc0d284a3725661113d326b3f3c47560a","split":"train"} +{"domain":"finance","full_message_tokens":254787,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":334,"sample_id":"sample:b000bcf92b1fee358a6248c15df078d4079e576553e1d9311e2bdb5b69e521c4","semantic_task_id":"87d45f170fc71071576783b1b73aa1905b42ae53708242850839b5c1b55e7813","split":"train"} +{"domain":"finance","full_message_tokens":254788,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":335,"sample_id":"sample:7e19443077bcb964257e82b9ab23d24819084875fc53e25cec2c39d8b7dd9424","semantic_task_id":"88f7171631211edaf45d549aebaf3679e81b7c30cff4dfcde515eb2771dacdba","split":"train"} +{"domain":"finance","full_message_tokens":78184,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":336,"sample_id":"sample:334e1b2b22da8e916baa2c7eae6ef8dd84bc89f8cb2596373698d188fd90e30e","semantic_task_id":"892ec11a1c63d21cd27ae49bda7d4957f6aae965d8bc9d87f8ad1d28c0cdadcf","split":"train"} +{"domain":"finance","full_message_tokens":75523,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":337,"sample_id":"sample:e5fd36e9c5112fd1b659f52dbe723e62f6739d17c7662c3142c5c48dacdcb52e","semantic_task_id":"8978dd64584f3d46a97749a7a2674857cbf21e9ff4ee2409c2d5a128ef1673d2","split":"train"} +{"domain":"finance","full_message_tokens":101359,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":338,"sample_id":"sample:7babfb3591a8669fa026afc5bce5b1c78ac4885492c6fab173874ec4f37f134c","semantic_task_id":"89a9713891ac6c7bca72d874a2100c82d038cee3f300e582bc7891c82ce6d9e8","split":"train"} +{"domain":"finance","full_message_tokens":69640,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":339,"sample_id":"sample:d71a9633597468210cfc86cabe499ea18ebd6c70c1a520b0d3f66d6ee1f6449d","semantic_task_id":"8c155317efee72b63a1d01242d21b28acea752573b4be30822cd3b96bf732c2c","split":"train"} +{"domain":"finance","full_message_tokens":102883,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":340,"sample_id":"sample:8a0d635c26fd98d7d83df63f90115e061604d75aa8e373b63b4602502ba5a925","semantic_task_id":"8c25e8f83195906af4da07ef596d36bbed285be08da86daef7d74d3ed9925078","split":"train"} +{"domain":"finance","full_message_tokens":102885,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":341,"sample_id":"sample:7e3468945032e5bcadd7a2336192533d1bf435b88d162e55b90f4368c192175b","semantic_task_id":"8c3fa6717871cd76d029464bda4912456865e8e3646f03c8bd19925e6d126831","split":"train"} +{"domain":"finance","full_message_tokens":77212,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":342,"sample_id":"sample:065f04b8e5518abced613472a69d30560b88bd22c97020b7ff2cf0080884a53e","semantic_task_id":"8d1dc73e89ebe8afabe1a4c6cc847653881fff7aba5cf6db72140db52c82b839","split":"train"} +{"domain":"finance","full_message_tokens":222186,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":343,"sample_id":"sample:c512ee3f1fc2f725aecd2cd4105ef27a09573895c704cac28b54559dcf3c3d42","semantic_task_id":"8e2d7b515953542815e06d6ab017819eb5733a3219afabd12e267088c6a248b1","split":"train"} +{"domain":"finance","full_message_tokens":75571,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":344,"sample_id":"sample:2cf438e33813e2fffc0433e73266f03a3197999e00b34c84c10f1c2bdd342afe","semantic_task_id":"8f5144d140f580f3d9bbc218287eb304e5a43c631163c72e59ee6f18822ac444","split":"train"} +{"domain":"finance","full_message_tokens":103980,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":345,"sample_id":"sample:40c0fd53db52ca95b559a8dce1ec6dcc9545597067cdd91b2752de2ac71a4157","semantic_task_id":"917f460543357b9aff0f4d89d6707783807dffbb480e7223ceab26c57e25c409","split":"train"} +{"domain":"finance","full_message_tokens":222211,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":346,"sample_id":"sample:62991e6b0bf18390bf8d67f41168bec8c5b4757e576e1ddad946ff780ba38e5d","semantic_task_id":"9371574cefd531293d8bc64f9f9cec27dcbc38f82ff96b9d1129d5d4628fb7e0","split":"train"} +{"domain":"finance","full_message_tokens":254789,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":347,"sample_id":"sample:fefee6448ee6618a3124f95befe95172c7f2f22f91d7d57bc11c39198c078646","semantic_task_id":"9449b780be49c8545d800fa915fd27bc4c91db4bddef72251ce4fc55a0f8fb81","split":"train"} +{"domain":"finance","full_message_tokens":103799,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":348,"sample_id":"sample:5b34f0197e179454480031ff2fc98052c86fafbc7e68c32ca4cd6d00d074e88d","semantic_task_id":"94f24af7b39d5b14e472a0fa10d5ecd5f8f980d3b34f92229ab9664da67dbbf6","split":"train"} +{"domain":"finance","full_message_tokens":103800,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":349,"sample_id":"sample:7d72a4a7be5504d7d9a3ef4c42cf01be3bbf140d8a923889e2392b39a32f9daf","semantic_task_id":"94f57643f81cc4c3de45edcd39f779b4794745d6aa58d24230484eda82ccd5d4","split":"train"} +{"domain":"finance","full_message_tokens":230685,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":350,"sample_id":"sample:2eb07cd14543b3dbde04de82663f61235c5fc46c616ef0d397cf95ba064c0383","semantic_task_id":"95b0943aea0bdae598a5a4cc854243517a39cdb6638c45cf681e3d109afb8525","split":"train"} +{"domain":"finance","full_message_tokens":222175,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":351,"sample_id":"sample:e005e69a6be2771a6a96b269b423d6af522e13d2f5e1180704081571b0e9dd71","semantic_task_id":"965ce863abca661d66838bad6a28bc37237498efd1751d6337348f3caa214028","split":"train"} +{"domain":"finance","full_message_tokens":99670,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":352,"sample_id":"sample:ad3af56361b7c5d5d6c024c6804d2283b9226f2ab9a881d2524f3a12e04d5ed8","semantic_task_id":"96ba599e153e41d1961ea3c9b467e36548ec502dd22b68725aefc81ac8cef382","split":"train"} +{"domain":"finance","full_message_tokens":78199,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":353,"sample_id":"sample:450bd8b5d6548993fd4d818c4e4777d1493acee2066e967516dead87b4b56080","semantic_task_id":"9b71cf29f70f77ec9956fc3fe5c068ef0030daa946c2a936026e40307888cf71","split":"train"} +{"domain":"finance","full_message_tokens":102880,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":354,"sample_id":"sample:128a9d503b07c4e316219f40ee2bf25970b6d62741ba037448db94529c99c2d6","semantic_task_id":"9d6b12dc664027f1785f36fa7d3e2ec4862431066936b77348407da65031a329","split":"train"} +{"domain":"finance","full_message_tokens":99666,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":355,"sample_id":"sample:7030b993016865a9f83836b7531f533d8c4752108b4aac2eee8a932081869635","semantic_task_id":"9df86b6547da0bfb330bac74424b99856253e5f01ab0bb80f081939b00ff5b23","split":"train"} +{"domain":"finance","full_message_tokens":222198,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":356,"sample_id":"sample:7852c416b46e7b315df4b9ca7041301eaf6b34d4dca9862f2641668975752876","semantic_task_id":"9e444bd81cbff4257d676a27e6dc960274be1e876496e49cd8c71bf5d53fd440","split":"train"} +{"domain":"finance","full_message_tokens":230699,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":357,"sample_id":"sample:dac6491819cc65c09ee8b59f46b22cd0adb7b3aa4f16f7bd8aaa744902d7d3ec","semantic_task_id":"9f59ac132fb405bac628a66a3c86efce07fe4e9ea675fef4e3a99372bba74833","split":"train"} +{"domain":"finance","full_message_tokens":77312,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":358,"sample_id":"sample:f380ce755a1afc0b0c496c6d3e292367e950987703e39c8aff4ba3089b77a83f","semantic_task_id":"a05733326e1f387091e9e639cb565168078afe598d1e5a1d0157afcbec396055","split":"train"} +{"domain":"finance","full_message_tokens":102868,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":359,"sample_id":"sample:f419b3ee9f0a8479861360861ac2690b43a6c4c060851d201708309eaaa05dad","semantic_task_id":"a0868ef7214ee40f75835b8ec5c919925d18e0c983400d93c0f9ed0bb45a53ba","split":"train"} +{"domain":"finance","full_message_tokens":222187,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":360,"sample_id":"sample:2d38922ba844978adeadccdb48782731f179a1b236441235f296e3785f3d28cb","semantic_task_id":"a15048bc69d3a83c49e686e954c925ac70c3727d406c12cae75c030fcc3f62bb","split":"train"} +{"domain":"finance","full_message_tokens":78128,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":361,"sample_id":"sample:106172fe89322ed2ab57d9acad618b7b8d3382fb1164ca4ed8818ee6fceb842a","semantic_task_id":"a1f262bd770b1a10ac2dde4b4fe1df6b09b3871f1f706a8b619158f23f7f4007","split":"train"} +{"domain":"finance","full_message_tokens":69641,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":362,"sample_id":"sample:b865ca900e3b4cacef5fdc84ccc46e817f3018eacae2f7d7981347fee3056677","semantic_task_id":"a2decbc2fd38027acc8bc538ce379ddf85ea2c11037e7efa6fae251d0ca9fea2","split":"train"} +{"domain":"finance","full_message_tokens":75522,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":363,"sample_id":"sample:d50377c8cd152172f4cc9e1f4760c5224859b0cb532c8a0a7779892998793e61","semantic_task_id":"a590e839ca8f4af37f9c14e2505de0a51de6f2ec43020329da2c62e8d5427860","split":"train"} +{"domain":"finance","full_message_tokens":102276,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":364,"sample_id":"sample:cebb7f738626c0476ef79acbf8856f7457bf9eb228aaf9e25352d826d21a489e","semantic_task_id":"a7f6f2ab619aaf05384d287b56e60cb2f723709d930d679b801f893dd5fea909","split":"train"} +{"domain":"finance","full_message_tokens":103801,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":365,"sample_id":"sample:125e221ca761d1ff08dd5d63509d79910027901a12e3b5e453633e5ea62c52a7","semantic_task_id":"a8a073bf9591b0652ec79d93ba3ed369e2cc443d133188ee8de6e37d4a62fa2e","split":"train"} +{"domain":"finance","full_message_tokens":102275,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":366,"sample_id":"sample:738e84b4c54c75e97e6adb139b1fa1c0ad906467f00163f5cccdbb0bba5d8f32","semantic_task_id":"aa0de4d52b5130f68c65c8cb39844a99909a33647c880082c9b0573de7828320","split":"train"} +{"domain":"finance","full_message_tokens":69640,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":367,"sample_id":"sample:1824bd51139244a5fc656bb04b56d416fa0c10ade370b3675bcadd6ec37f8001","semantic_task_id":"ad45ec3b938ffec44201b995312ce00d27f5d41249f4b7a2e101be8499335281","split":"train"} +{"domain":"finance","full_message_tokens":99669,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":368,"sample_id":"sample:ba9609f9547faec0f7783ab063415c27964d0b2e5a2a06cc97c79246c5defa55","semantic_task_id":"ae3d28bc4a1cc5b43dd411ea6e4475e6d7877027492289292c3fac0b9536bca4","split":"train"} +{"domain":"finance","full_message_tokens":102879,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":369,"sample_id":"sample:b2fd62b7df8605a0bd57bcfd88566624c0e8c45a344d488b7a622c3d1e6ee667","semantic_task_id":"b10beac1ce144b13398dfd7679f42fd0b791e4eaf322b0e61aacd70ac78d72bc","split":"train"} +{"domain":"finance","full_message_tokens":222176,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":370,"sample_id":"sample:9074816b8216ac3e8a50bc070497dd78d60283d9f17d4cde9998234a3ff0bc5e","semantic_task_id":"b389443c75ad15db4c774866b10aeb0457e973dd83830e7442e1fccaf8309827","split":"train"} +{"domain":"finance","full_message_tokens":103997,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":371,"sample_id":"sample:72ae4e12e4131d8d2669c55ee4f58e903b9edabf1676a2abf9f581b2fccd5e4b","semantic_task_id":"b4a420767fa2304dbd920f17289e90aa58ec321938774a40e7efd5c38c62ca22","split":"train"} +{"domain":"finance","full_message_tokens":230664,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":372,"sample_id":"sample:7ba592a1d38fc3a6c73001b19a3b7e12f81b358cc52c0669b1ea72d65bacf469","semantic_task_id":"b6895e70480bc1742d1a8c7e4238c441dcc555a1858df0a33e5eac2e46f7a874","split":"train"} +{"domain":"finance","full_message_tokens":230662,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":373,"sample_id":"sample:2307a9497616a11f203c8f5c08430e42f0e7806ea049a03f0f3bc20ebe874133","semantic_task_id":"b68a2de3b8ca8571504cfe890ac586c28ff4844600df88414c0c20daa05c730c","split":"train"} +{"domain":"finance","full_message_tokens":222176,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":374,"sample_id":"sample:ae3a0e1f7ca38dec6a54aa5d23cf7d16b654101e6628001763706890c1547bf3","semantic_task_id":"b6e5b1b9d5c9b21e483dce95378db5688f5ca7c66daa87f72e702a68622d6dfa","split":"train"} +{"domain":"finance","full_message_tokens":254848,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":375,"sample_id":"sample:89e5480d1e29ee78ba5dd7d310a50fd9848eeddf6a7b20c35a9ac25fe6991f89","semantic_task_id":"b8b2fa4bc85911d857f567f19744a37e727eca72740f9fa7e5dbfb6b02d989b5","split":"train"} +{"domain":"finance","full_message_tokens":75522,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":376,"sample_id":"sample:cebeba88538273a194305507c2a1f0b5f1653a58bcb6306b28eb55d4c492ac39","semantic_task_id":"ba40f5452f6f7fa3a7bdf8f38176f2f6bfd4f69cba0c0cd83d729ca9d1e8d3ed","split":"train"} +{"domain":"finance","full_message_tokens":254789,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":377,"sample_id":"sample:e2dc565c876efbd32023dd4fd28a5ef0c32ac1cf431f1f05aa2613a41f2ab53b","semantic_task_id":"ba465feeb26621fedc67a4687de1abb188d5a246ed4990815b68ea6ebac7c6ff","split":"train"} +{"domain":"finance","full_message_tokens":103995,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":378,"sample_id":"sample:bb6285f9071b5f563b31388b50956d7df5a8a917c2cb136fa22440439a02ae10","semantic_task_id":"ba6f0edc2d30e02508b98796825cc74eb14a63214c0e54c0eec9c40ef395920b","split":"train"} +{"domain":"finance","full_message_tokens":75522,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":379,"sample_id":"sample:66b796a6d1c653708d44e650c0af9ca40126a3824fdf843739165b7e42425f94","semantic_task_id":"ba9a5ebab9099992b8d9dc7444f288a4a1e735b53bee669e32fe367860033a9b","split":"train"} +{"domain":"finance","full_message_tokens":102885,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":380,"sample_id":"sample:0e39a073e6fc30a0b4f3672e211efa755d9bc91322a209ba435ca68fb3d7b1c9","semantic_task_id":"bbead9720d47e7bc4418879324d69c5094d28d447fab9c20dd5327ec3ba53bf1","split":"train"} +{"domain":"finance","full_message_tokens":99665,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":381,"sample_id":"sample:11d914d7fcb911e6c2a04b1f5c52b83c89b7a1c26b9dec02437a27cfa8f0c752","semantic_task_id":"bc248d3c73a7ff45ee100dc3add1aca80181eb52002785a3ca118d6dbfa3951f","split":"train"} +{"domain":"finance","full_message_tokens":69641,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":382,"sample_id":"sample:97bcba7db8a93c9f0d1c6760adbe6e87c4b04d1f1ad2baa1cedfa25692dbe8b0","semantic_task_id":"bdfe0b51bbd17ccdd8a3f4d2cfe17654de2740f67265d4750ecafd39afef3353","split":"train"} +{"domain":"finance","full_message_tokens":254821,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":383,"sample_id":"sample:aea58ad7eb043cbc4ebbcde69da8cab78010e32f98bb0ce0800e8d7d46f3894f","semantic_task_id":"bf3b9978164f612851636db028f4d305302503b9ca4fb4fb4e6929e9e4bf964d","split":"train"} +{"domain":"finance","full_message_tokens":69640,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":384,"sample_id":"sample:0fbdcca23e8725fbd8cea6e664492fc3c6104340507b8fcc9152fdcbccbc2ea3","semantic_task_id":"bf47dafed884262d94a04bdb325892c67bb1cb0d9d0731e88914def0f19080f8","split":"train"} +{"domain":"finance","full_message_tokens":222197,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":385,"sample_id":"sample:735a0aecaaf917a237c03b1cd67bb805b992fec7c2314128a5a0e02fc4d41f77","semantic_task_id":"c03265ed7020aeef7e86fbf8f8da27553fc143019fcf289d4ac29d3b814fe81c","split":"train"} +{"domain":"finance","full_message_tokens":99665,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":386,"sample_id":"sample:9852055daf2f0c37112667886431d4c2df40ca6d1d7b9cfe0ecff7e41f1eeddb","semantic_task_id":"c0b52b47ce162dc72c708bcaa07d88d6e878a0a184beade9e894454ca9c7dace","split":"train"} +{"domain":"finance","full_message_tokens":78228,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":387,"sample_id":"sample:b0787b40e780e069153e3baaf4a33eaa2af816b8f3b507c3504e339b7f0b5c0d","semantic_task_id":"c0dd802120407ed8a87e8111f9b55c32b7cd2be882295b718b2de4ff07ec4f07","split":"train"} +{"domain":"finance","full_message_tokens":102883,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":388,"sample_id":"sample:fa3a4e56ba249ab8412cce3f2970e4e699da6079c679ce6068fe67d8d722e95e","semantic_task_id":"c26249b2a49fc6ebbd22b809846994b5b1c8d3bdcb58e0d7e8181c534c34a126","split":"train"} +{"domain":"finance","full_message_tokens":254787,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":389,"sample_id":"sample:0bb6cd674c33eb8508f29cec2e55a2ca2e738cb9af8c1e0f9f3b8463e188dbbc","semantic_task_id":"c38a61eb6f8f73c3485b5886985e6def5a08a37e701b41212aeb06db0f957d49","split":"train"} +{"domain":"finance","full_message_tokens":230685,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":390,"sample_id":"sample:656a5c3dcf915efa9cabf079346649d0edb5f42dd8b4b953dbc012a3a6522bb7","semantic_task_id":"c432f0274d521ace18cdae9042563897882cee2c750d19f02831124a1180cc63","split":"train"} +{"domain":"finance","full_message_tokens":103996,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":391,"sample_id":"sample:04f802ee4ff4d964cf8ef6abd6c20fcf063d3d8255970ee40c266d1f87c2ad22","semantic_task_id":"c4d38c3576db7da8ca27b975fec5331a3e55e351819fb3493bc41b51a7e6cb70","split":"train"} +{"domain":"finance","full_message_tokens":99671,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":392,"sample_id":"sample:bdaca4238332825200416c873e4f0bb0e8d78aafae006f988d3f3399097db302","semantic_task_id":"c526cbbdddfd30337e0bbfbd65c83c16fc74a3bcff87e4e95830529e35ae487a","split":"train"} +{"domain":"finance","full_message_tokens":103992,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":393,"sample_id":"sample:98b21114d66382ffda9c3032da880ef493653941201964c89378969dc172b632","semantic_task_id":"c82838f34b15d93647c14bde2986189eee589bec341771cc27368b8fdfb1506d","split":"train"} +{"domain":"finance","full_message_tokens":102274,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":394,"sample_id":"sample:ea4f24dd459318f41610b8d6ee1e69fa52fc15ea28dccbcdf63da6dcc0e137ee","semantic_task_id":"c90233f2d38b1691580bc9c872a56ceeed274234b2b9ed61d71fe2b9053e1f44","split":"train"} +{"domain":"finance","full_message_tokens":222175,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":395,"sample_id":"sample:e22ac969c4998a2f3c4ecc45fc8a7e96ff85a20b6a19febf479590278c712a7c","semantic_task_id":"cabf97ce988f91ba7dac3498e75243b618911416f95bfec03dd9f735d04c4741","split":"train"} +{"domain":"finance","full_message_tokens":222174,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":396,"sample_id":"sample:220461b04f584b15e1cfb40ff55ca107253bac3a0047e09f3f9bc6a02cfce4df","semantic_task_id":"cb06b6909929f4514d9cb701b091677078c6b7e51b9cc4b8b7d49cdefc421f99","split":"train"} +{"domain":"finance","full_message_tokens":103993,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":397,"sample_id":"sample:ceba3ec0d7f92272729363563ccdaa01f8ef565419adf2ecc27969528efbd158","semantic_task_id":"cb2bf11821c6d9fb2a5f2260f00d2e5822245a8a7f008e3bd5316eeec718e827","split":"train"} +{"domain":"finance","full_message_tokens":102884,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":398,"sample_id":"sample:321b23e457317a8d74ef9b79cf8488ca65f3c3f8e78bc7604f94fb22b9632e0d","semantic_task_id":"cba459e3e4af7a28fd8933cdae723d5f508ba2a5cc5dc35f1307f29ff17153f8","split":"train"} +{"domain":"finance","full_message_tokens":102869,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":399,"sample_id":"sample:3b1e3d2a8dfb046378801cad0c480b687a7461743f037c01b2451b2e8af14792","semantic_task_id":"cc7f22e70896a9755650c8346a288b300d8e391102da62e2b46717a1e72084e4","split":"train"} +{"domain":"finance","full_message_tokens":222196,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":400,"sample_id":"sample:8705f1479dde11ef6f9b4428de85444b0844c72540ac437dd242634f561eb075","semantic_task_id":"cc9c7df39623fcb5465ca8837e4e3b28ab193df9c061fe577798aad1d73fe3a5","split":"train"} +{"domain":"finance","full_message_tokens":254820,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":401,"sample_id":"sample:a221edd6dd026ffc2cd500a9d38c4df3368341fdce764f68606b186b75960dab","semantic_task_id":"cca0c6892b2b6107e402f83b3d1cbd77dcb6ce05d842f8a246b6b6a110063bc2","split":"train"} +{"domain":"finance","full_message_tokens":254787,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":402,"sample_id":"sample:bbc532256b7f81e5159846fbd8daebc33ece3aea2865617ef8f75fd84382a91a","semantic_task_id":"ccf33295512fdd00fd1c2294572fa065adef9880d047b5199c6c47a5a12f6d8a","split":"train"} +{"domain":"finance","full_message_tokens":222195,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":403,"sample_id":"sample:a51502fd1fdb61b3ed31ac1b80e9e58d80e9d79a88b86c39d5eb32b033c9d244","semantic_task_id":"cd4cd57f44960b3c123a9837f384fea6bb67945204f49f20549a149215bbee51","split":"train"} +{"domain":"finance","full_message_tokens":102884,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":404,"sample_id":"sample:4ead7471084d31b1ce4fc794b603484dd3a51f5d4a684f50eb32087b051f23d7","semantic_task_id":"cd7af3d2f57831c0b0fc889b8a3640f3b3ace0f697788eaa14d156ada9254bb4","split":"train"} +{"domain":"finance","full_message_tokens":230664,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":405,"sample_id":"sample:b284eb38446a19d6758117e36ab19642db870fa8d16dc8b88044c7c02fba0ab0","semantic_task_id":"ce6ca1976dc22cee325fab0dd0743a78e5a740398ea3a197efddb25794200aec","split":"train"} +{"domain":"finance","full_message_tokens":102276,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":406,"sample_id":"sample:0261058f94833f37f0ea02e83ec6296f6f05c4fa8177a90b273e09d793efd547","semantic_task_id":"ce75cc2cfa19f9b91272f220ece0583e5f42923e551edc46ceabb461c54ca5cd","split":"train"} +{"domain":"finance","full_message_tokens":230661,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":407,"sample_id":"sample:9cff3fc6e0b28c163841119b466b75d37f5a549740b92e9bc10ecb55249e10cb","semantic_task_id":"cfe8f2b27f8dcc01d1b5af136e9ee9ba6a82d3734e53679efc826961b7a8e609","split":"train"} +{"domain":"finance","full_message_tokens":103995,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":408,"sample_id":"sample:4e294081f026b32799fd02619b25f8216201b7a91569c035252596eff9d6ec57","semantic_task_id":"d2e7c77b7429cf59c91c9a06f9997dae2fea28b7daf4ea5415401187fbd44a6c","split":"train"} +{"domain":"finance","full_message_tokens":222175,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":409,"sample_id":"sample:2b9ad49315c79dfcc20d9ee59507bd051ff8cb59552f4598df1c8a3fb04bedb2","semantic_task_id":"d4ab2269047335f4c98c7d6a55c96834f07fe2b32b8150d67e5a0464ac446181","split":"train"} +{"domain":"finance","full_message_tokens":99666,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":410,"sample_id":"sample:b30c444c178a0fb9c97b89455c9960c19f8bf41b960aa0e3ca628db6ec3d3c0d","semantic_task_id":"d59f290d28712da52291604cbeec129f08a98946701a1e4d84aafc3f29124803","split":"train"} +{"domain":"finance","full_message_tokens":254808,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":411,"sample_id":"sample:c495cd3fd6885d4688272952ab9674f56f1ff7da9ce481168cdf2443aa2dd392","semantic_task_id":"d6c9682e33e9862df8b89d030ae94d78a6adf2a87450cb7dee2eb16d90b53c23","split":"train"} +{"domain":"finance","full_message_tokens":77261,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":412,"sample_id":"sample:33021149dc8841916995766e4a1567cc6521d2afe561485a7a39803199d79716","semantic_task_id":"d6d77eb06922122b30d9213c982bc5256936b0d1c95dbe2bc7b8a5a338dfcc82","split":"train"} +{"domain":"finance","full_message_tokens":77211,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":413,"sample_id":"sample:bfc2a791deb620e31edf45e19bf308a13cd202c390ce982708d20be7d1b5e31d","semantic_task_id":"d722d4a20130dcf5fd5695ac545d22f66d5ba8df575f1c8ae38429537c88a651","split":"train"} +{"domain":"finance","full_message_tokens":222198,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":414,"sample_id":"sample:a25eccdf3c1fdacb110b33b1341668cce432a2ba6d80e414c2b3719c6539b996","semantic_task_id":"d80be438625803061f926b3e42e0370565c8ebcde94c0804dbf2417d9c832f0f","split":"train"} +{"domain":"finance","full_message_tokens":101359,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":415,"sample_id":"sample:70a8da49fa5e9bdc5ea64bf819706cec0aeab2e7c37db073311f9f10f4faf5a1","semantic_task_id":"d87b2fd2056135d2c17b8c89f1f6acc3c74623172f4fed5c320279575a717bff","split":"train"} +{"domain":"finance","full_message_tokens":254788,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":416,"sample_id":"sample:c8a35bb2e40a90cbd33577492dbbecf2086219d76ac9a8bf3869d5d3652a947b","semantic_task_id":"d94ef2099c9e39241d2785b32d3dc30d278ff65fb1c2195452322232af652b52","split":"train"} +{"domain":"finance","full_message_tokens":77213,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":417,"sample_id":"sample:4fe59da4efa63717b5add7efaa0e12bbf3c9d334f7e36d0805f2425ee300b425","semantic_task_id":"d9681145660e3875131c22b7b11e6c1a1314b093a94b22b6a92dbada101c77de","split":"train"} +{"domain":"finance","full_message_tokens":99670,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":418,"sample_id":"sample:7b8276995169798867054980f0db90de9a942bd18705c48592558eab256aacf0","semantic_task_id":"da51836ea024b597d82c05d194f231eba224b74367839dc2bed3c34aa12e837e","split":"train"} +{"domain":"finance","full_message_tokens":69689,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":419,"sample_id":"sample:c41adf060ae73a03d0ef42998f6403c21ba18f0525effb7ec1302c68f29f6bdd","semantic_task_id":"dcd86eef5454e28213eeae029c519e9394a3ebd7d61da52b1d90407dce1ebec6","split":"train"} +{"domain":"finance","full_message_tokens":78178,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":420,"sample_id":"sample:ab7b1bf936140fd7dd4937fe4401568589d27c30e7428d8bece2b3aefebcde8a","semantic_task_id":"dcdee460c3d221b72588e56851d5c9dc55c73bc91d4fde4d5f147a9dc84d7d54","split":"train"} +{"domain":"finance","full_message_tokens":99671,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":421,"sample_id":"sample:753fafe5d1821a681e9a0f931a4ff1138dee08bbec32b45217c41af23394053e","semantic_task_id":"df84f033b2f78a08e08f167dc2326dce7a9d7467aee7651db1081bbb6cdd2b61","split":"train"} +{"domain":"finance","full_message_tokens":102880,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":422,"sample_id":"sample:98bceab43a4da63b1bec026e2572486a0d545ebff38a37d15a6f0af37f407517","semantic_task_id":"e0b3ea92555e6a8b4780d868ba677f93350ad17be268529b3c75c6f7681b05ec","split":"train"} +{"domain":"finance","full_message_tokens":103996,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":423,"sample_id":"sample:8fe54eaa8bd526419f015f4a3fab1c5adae77052e953eba78bc77c6e49158ba8","semantic_task_id":"e14186442c073fbb467965697a645938965ae3e468b76e81c68a01ffd2be8985","split":"train"} +{"domain":"finance","full_message_tokens":99666,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":424,"sample_id":"sample:af639ed598a14cd4fb9180d8f52bfed415c33c14ef27fb32e44d014e14d49190","semantic_task_id":"e1753eff46dbaa2a90160f296f38f0161d8bf7f0abce59ce075f2ab9dc82aa51","split":"train"} +{"domain":"finance","full_message_tokens":101359,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":425,"sample_id":"sample:d7a9ccfcd7c534b0aa1f90f49a6a5062e341f0b3a5d1e806d09e77822747bcb7","semantic_task_id":"e29537d68cb711d97ba59eec23d5f81999d0267a45c2aa15ed93200e826d540d","split":"train"} +{"domain":"finance","full_message_tokens":78128,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":426,"sample_id":"sample:353a5a9d245561e4844b519d336b97e261ff3201b7dc33c1565c356c2a170498","semantic_task_id":"e3cd5bc1e30092ba81d532aebded399dd48c4aa18487b55900186e00dad152a0","split":"train"} +{"domain":"finance","full_message_tokens":254788,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":427,"sample_id":"sample:ac1b68d83c30c46011a8e54564d326b79c7271a149a917b0d137b979d356c7c9","semantic_task_id":"e3d1799589cf8d2a46120d8dd81a0de91c842836bd71ec112cbae4188c034b1e","split":"train"} +{"domain":"finance","full_message_tokens":230684,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":428,"sample_id":"sample:4d20e7be359cc3b68954556eb7b3b028a2e4f9cda728c00008f7493cf63117b9","semantic_task_id":"e4149d9b1fcc68ce3774435990e9a7c03354b96bd791fd88a1b3d0e627bf37e8","split":"train"} +{"domain":"finance","full_message_tokens":99669,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":429,"sample_id":"sample:615f0370fdfaf9a5c48967a8a957e4d9b36c4fe99812efac9cb3d5bad3b6d304","semantic_task_id":"e46d8674a77192e94da61fe456487d71edf75897223ba751e499b851a337ae4f","split":"train"} +{"domain":"finance","full_message_tokens":77258,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":430,"sample_id":"sample:4c4f38a813e99d858b3a943780ee956ae4c8abc392102c21893527200769d0ce","semantic_task_id":"e770f3f152a32bc5f01a431bb6b50db6c97d4ab2e37bfa66b66c29fcc0bb4601","split":"train"} +{"domain":"finance","full_message_tokens":75524,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":431,"sample_id":"sample:664731b51de6fc724bc779e3081069a7f5249ac43db0a53253fc384638a98dfd","semantic_task_id":"e79716c10253bcc7f1f97a68c5686f24c2006a27d782d9528f28d401aa4b533f","split":"train"} +{"domain":"finance","full_message_tokens":69641,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":432,"sample_id":"sample:d43ffb333ecc7fe98e1cd3332d19a8dd2b632a46c5a3d5eb7454efa9e2a9e0ac","semantic_task_id":"e7bc15b4e471efd32094b7c2cba43b83494a0d13001e0c855408356319e6db25","split":"train"} +{"domain":"finance","full_message_tokens":69687,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":433,"sample_id":"sample:73f7f92cc68f294eb2585350316a61a67b41ff9820c868466a0854c81e8aeca1","semantic_task_id":"e814f13efc46a10bed29a378b93929a28f2cfce84cd5f1436f9e35afad81623c","split":"train"} +{"domain":"finance","full_message_tokens":103992,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":434,"sample_id":"sample:e4265fd1813504ec54d7d49d5ff9362219fa0493615f81701206b4c3c2954200","semantic_task_id":"eb2fcb905c6183f423572c0b6deac0d49903114272e2dfa7e25cac8290e3da8b","split":"train"} +{"domain":"finance","full_message_tokens":101360,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":435,"sample_id":"sample:059e0532f83822c05a54c0aa9d3ac4a019afb03551b28dc55d004a19c30c6874","semantic_task_id":"f1d56dbc7ef7e475f1f2b045993013038290df130ae7168c2f90cbfe2c5c7d3b","split":"train"} +{"domain":"finance","full_message_tokens":230684,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":436,"sample_id":"sample:36debd3abec539de20a92ee4cfb23d4bb9031b38436950774cdbe5b8f3d0dde0","semantic_task_id":"f1f451f284899306be4a5c0f901e4cfcf4df622e9de657c43c1501e63e67c226","split":"train"} +{"domain":"finance","full_message_tokens":75523,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":437,"sample_id":"sample:9d7d23f51958e526553b0aff07dfd510a1498b6895440ad2535376ed37b70640","semantic_task_id":"f3f2f3b2a99b4b458a2b72d6c5a5a72752e0a4e372f6c665c746b22b3d5010ef","split":"train"} +{"domain":"finance","full_message_tokens":101360,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":438,"sample_id":"sample:3961c05c3309dfdc7373ebf4b4a3c838503d66229cfe14dfb7bd92a7c9ee3ecd","semantic_task_id":"f499f72806cc66df24f4c19f47b16fa2576c26c54ef094a9d714255167f34d36","split":"train"} +{"domain":"finance","full_message_tokens":230685,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":439,"sample_id":"sample:de4725bbd428307b1ca2d9673615091af3f06a569ff1fe1a8756d2c67a2b570f","semantic_task_id":"f49dbbe4d67a1720e563da03ec73832fad9989055e1612965dedc42ddef568c6","split":"train"} +{"domain":"finance","full_message_tokens":230674,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":440,"sample_id":"sample:e4fcae5b72148a62787aac125822d859ef24104a5f9b78de4a17d1e01bbeb813","semantic_task_id":"f535d14e19f32bf2a4249c1b1d7c70a7819f40ef7d66caa1eafd0c171413f0b2","split":"train"} +{"domain":"finance","full_message_tokens":254820,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":441,"sample_id":"sample:eb590f47d525524aae8ebc156059e415aa5f284c6a831684916b9f72d383ac55","semantic_task_id":"f86c7119a391855fa4a0ffbf46c7e806a11766999cb80082f3ca58d76a9c3f33","split":"train"} +{"domain":"finance","full_message_tokens":254799,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":442,"sample_id":"sample:c72655000230468368aa5bcb757f584aee261a05c81f84ae5aa32bc032ccac2b","semantic_task_id":"f8cd6498eeb05027a6f7c50a982c9b143a981eb50adb63f7ccc68bf4bc4fbaff","split":"train"} +{"domain":"finance","full_message_tokens":254820,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":443,"sample_id":"sample:b649c344098dd54e11c2cb65f146b83fac42f2ae808bdf3fbab09ff3c1c0160c","semantic_task_id":"f8eee5a1db89fa3a9c8b818c30ebacb83fb28728e58be9a0d4e3bdf1541d1685","split":"train"} +{"domain":"finance","full_message_tokens":222176,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":444,"sample_id":"sample:3c0820cfeb1f9db5f6cc1b3e83b20b45297731442404966239866c40df0748d1","semantic_task_id":"f98fa5b57f8df2e02933ff104b61dfd8a73efe8285b6cab110d10c8c35aa8552","split":"train"} +{"domain":"finance","full_message_tokens":75623,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":445,"sample_id":"sample:a53b83db4025934274eda35c218cae4774ef0e0a5909a506ee8a40d02300d506","semantic_task_id":"fa40872280d610e0db3bcc4c17a6eebc26f2ef1e83be5dabaf261bfbe0773d28","split":"train"} +{"domain":"finance","full_message_tokens":77210,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":446,"sample_id":"sample:dc52102df381aa8f0dc3b26569e7f8eeb5d2940e68fcb25220d76ea3d7ae58fc","semantic_task_id":"faf1e45fa186ef2ab2c8332d88610c09c6c62fbfa19f4e8aed80c2a3684985c2","split":"train"} +{"domain":"finance","full_message_tokens":222176,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":447,"sample_id":"sample:f9a4c91446826f38a97367a99eee31176d9a53aeaf04b6d11f91c916e59f4f6c","semantic_task_id":"fe65e930ceec01bb9ee13e1680940baa080e616c1ab261327daa0f450b025348","split":"train"} +{"domain":"finance","full_message_tokens":78127,"group_id":"0001018724","output_file":"finance_train.jsonl","row_index":448,"sample_id":"sample:6da8ae4eae66f6b7e53ca4f4ccf478cd303a82bdb4f219223e333a12777f5956","semantic_task_id":"ff86f69fdac7f0be0a3327ed121eeaaebecc69fbb82039082c91f9a3f8749be3","split":"train"} +{"domain":"finance","full_message_tokens":92930,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":0,"sample_id":"sample:dced69e76340c32392e0610b54149376ac4cc38f7ef8fb44bc6c022e88b6ca32","semantic_task_id":"039502d2360b18fc62ad85f52f0f9409ff73f201d0114d2106108134fdcc90fc","split":"eval"} +{"domain":"finance","full_message_tokens":93043,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":1,"sample_id":"sample:3a843d1fabea1716c50f21af6245f951150c7d708003172c3cffc669e88ad030","semantic_task_id":"04fca28fdd83638c6f554e33bac56bdb3a0ad9974e0032fdfc457a7bb650c351","split":"eval"} +{"domain":"finance","full_message_tokens":92950,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":2,"sample_id":"sample:3617c7a1ef6d885f246c2deb8a3f2e156c4efb43c453319e363d19e6b9eec7d2","semantic_task_id":"0698c43b31406a317b213d13161098263b189c8538947d258aa5bcec6008ecbb","split":"eval"} +{"domain":"finance","full_message_tokens":260073,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":3,"sample_id":"sample:061802260d5c6e1756a2230f711113bf2df6af6c968cac5f713710450c522524","semantic_task_id":"0aa3c9561edf7ad2d8fe4d21e5e837b9eda6a3e6d81f2ef86cfb1635e09d783f","split":"eval"} +{"domain":"finance","full_message_tokens":80416,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":4,"sample_id":"sample:ce7620271545de3ff1b2638355d2d54047f28dcdc4f8686cdfcf76cd22c60359","semantic_task_id":"0aac19f591888141130ff119c5244df2a2beb902fa2549973a289681dd1a869b","split":"eval"} +{"domain":"finance","full_message_tokens":89912,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":5,"sample_id":"sample:59a281737eac4cb336aa796abcdf813779dfca866bbfb1d27a0cbcc6c7031f5f","semantic_task_id":"0b798cc7f5b3d3f801cff44aeff8ecfd2dbcd01457c91fbe0e20e444b998d840","split":"eval"} +{"domain":"finance","full_message_tokens":92688,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":6,"sample_id":"sample:9599901fdd02f7f303b99c32945562d1cf96b5eab738ef622cab690c009e2b9f","semantic_task_id":"0d3f3483763d6fb281061b769ebe694c81b2e23505c23b92b8636633adb6f48e","split":"eval"} +{"domain":"finance","full_message_tokens":92685,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":7,"sample_id":"sample:3e89cce0a2521c9e9407029f785ead4f9f74dd1d4ae7e817e09fb27c30057a00","semantic_task_id":"0f43bd169717608f197124012a132b822c51920b89a90d73559392f132947313","split":"eval"} +{"domain":"finance","full_message_tokens":92930,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":8,"sample_id":"sample:520dc2e7ca9d053b7aec3744801ff07e0ea8909b508ab669f027d7456a07e50c","semantic_task_id":"1018cfb6d0ca9467701df767d5ce76f156f222540db3e07b1912f9a9fcdf9641","split":"eval"} +{"domain":"finance","full_message_tokens":92960,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":9,"sample_id":"sample:2491d116f6e3a14cba6e8bfa32d1911789a0ed823db350abaa372d5e926a079b","semantic_task_id":"14c8cf33b37b66972b712bc1385bbab90e79c993db6222cec8c11b2a78da8a29","split":"eval"} +{"domain":"finance","full_message_tokens":92926,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":10,"sample_id":"sample:e96fdad4ccffdb01a72b2f8e89136bf7e82a1513b3d7f57f9254311f1682dae2","semantic_task_id":"151d8ca3980e6cecbd946272b0fea6aa4d66a688a8f2c1145e75104ba511b1d8","split":"eval"} +{"domain":"finance","full_message_tokens":90079,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":11,"sample_id":"sample:03a946aab66a1dcfb0e26b5eba0bc4061097a6561ea6819af13a38073680f83d","semantic_task_id":"1699c5d09530261727785d6c639e3ad3e3490dceb593313613ee43b205773ecf","split":"eval"} +{"domain":"finance","full_message_tokens":80336,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":12,"sample_id":"sample:6cc125b4d5aa36367603d2d7df194cae45fed7899d2eaf9d6fc78827da2ef342","semantic_task_id":"1752db250d95a19fcb0d8d2d359fb2ce8a05216860663db8018a50c584ad8eef","split":"eval"} +{"domain":"finance","full_message_tokens":85740,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":13,"sample_id":"sample:b7a07437cbd7667f7a506beafd890f0b944e4579ffc405828e6c5b87906599b1","semantic_task_id":"1790980938d3193e90162fe53cbad482f3b7ee93598e7ab5751863ab517514b8","split":"eval"} +{"domain":"finance","full_message_tokens":182894,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":14,"sample_id":"sample:7e1fbb1ee5957ee49aea0e4e6d5cb8a18348389257a24e4d28d950153f746f74","semantic_task_id":"17b1cab01992b23be003a465bd89a1fe5d618ec01f6aca31d9aca9ccfa70d083","split":"eval"} +{"domain":"finance","full_message_tokens":260094,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":15,"sample_id":"sample:80373f5734f7049d94a4d1a768c36fb920624e2619f3a376b9d214bbc255dd08","semantic_task_id":"18018704846dac957171e5c33d6e27832982444285627c8d00051934724b590d","split":"eval"} +{"domain":"finance","full_message_tokens":93041,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":16,"sample_id":"sample:971c6ffbc54e924113200789976eede0d53bd30c01eadb919d11bb3f91512103","semantic_task_id":"1985e44d5cf5450a2a7e785fb14685e64fdee546bde8c33a2d24b7325f91d0c7","split":"eval"} +{"domain":"finance","full_message_tokens":92960,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":17,"sample_id":"sample:72d2b7130b5b5a39b835e3c49f5f72bd55c9bb70b030e994ade27d11da29e3a5","semantic_task_id":"1c951c984c2f3dd47e8f01986887f6335f6a19a2ee90d8e98009b59830a36011","split":"eval"} +{"domain":"finance","full_message_tokens":92674,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":18,"sample_id":"sample:53f7545344f79000d1362bf3479ff9dacd85d13d1ffcb5b0d61485fa3fa28926","semantic_task_id":"1e255527c13c8219757947740944bc22278136e5563d019234c537d0ee8e7e91","split":"eval"} +{"domain":"finance","full_message_tokens":80336,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":19,"sample_id":"sample:97f8b96260d6a4ec24d58ecf297291027a0bfe768cb70b71f00c200310b8dcb5","semantic_task_id":"1e82e290a87d56a952b91dd679f62ee1ecb02e359d198744fb185680be2fc27d","split":"eval"} +{"domain":"finance","full_message_tokens":85789,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":20,"sample_id":"sample:7e7d1ca1cdb21e1f5186a77f36fa8dfd34c906327f5df90685e49400f92b5f47","semantic_task_id":"1ecc74bef7ce4ee49a3b35cfb35cc9657335041b8f6b5a2bcdaa233e78fd026c","split":"eval"} +{"domain":"finance","full_message_tokens":85740,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":21,"sample_id":"sample:68d0a0d00e2607fdc5ce6e0e7dfdbf1e4c87a59bcf0d807060ce323cd98f7c7c","semantic_task_id":"1ee5ad2386144c7a4d88456f0b32c5eb03e891107a98fd29ead90fb8c3cb2221","split":"eval"} +{"domain":"finance","full_message_tokens":80338,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":22,"sample_id":"sample:c2263e61298da67c4e07265c3a99e16563ddb15851f60cddb7c077d6db4612db","semantic_task_id":"214cbc5fc749a65749861eafc192bec94c988846dff5754767af506ea6a4b3a5","split":"eval"} +{"domain":"finance","full_message_tokens":89913,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":23,"sample_id":"sample:e15e6200aaf32d984f473da39b6ac4094039db1e60ec1d85a1a9c397a73b096c","semantic_task_id":"2214b72961e2f2020e17d5b4e995fff9e4541c74afb18098028bd44b717fbfa6","split":"eval"} +{"domain":"finance","full_message_tokens":92960,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":24,"sample_id":"sample:d94094442db63934ec248aebdd375a1edee7e8b1be05f49a85209ff823051967","semantic_task_id":"24ca65ee41ea3a6743861cb5a5446b06e5400f3a1846217169dcb549105959a4","split":"eval"} +{"domain":"finance","full_message_tokens":92926,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":25,"sample_id":"sample:4b373eecf9f954ae788d0133ddd37d0f6038cf8142531471ef7d767af0d9b210","semantic_task_id":"26baea9bb87413b280068dadd19a84e4a47f1facaf894572a3ea97dec9f9019e","split":"eval"} +{"domain":"finance","full_message_tokens":90063,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":26,"sample_id":"sample:ce19d77107fda014ecfcbe6b7cbbe0de0f73b179abc429cbf739bc52d78d1f6a","semantic_task_id":"28a0f46f1db5dea100d31a18abb1f1ccb9c085fd79c949864dd653ea9ac2991c","split":"eval"} +{"domain":"finance","full_message_tokens":258827,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":27,"sample_id":"sample:e21a42bb782b7fcab14fec88e2713a82154f5199544be96678ac1a42f50a28a2","semantic_task_id":"2cfed66a120084cfdea5b14b4582c51b9e15ff4d14e45d740d6279b9590230d8","split":"eval"} +{"domain":"finance","full_message_tokens":92929,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":28,"sample_id":"sample:9f1d530b563a3372db3ef033e731531fabe940f229411a4bb588f61498d1075a","semantic_task_id":"2d4e20f8755af8af971840ededd7dd32199923af284d3b6b4d786ecfb97e28d3","split":"eval"} +{"domain":"finance","full_message_tokens":88404,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":29,"sample_id":"sample:18c2e538762a9b7caa19b03ae03ea7f8264da98ebeb190d254c74fa9a770cfb7","semantic_task_id":"2ffb5eb6e7de49de802e0fc357f2431ae448f72cb958fdf3c8d202ed126387ed","split":"eval"} +{"domain":"finance","full_message_tokens":85740,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":30,"sample_id":"sample:9c7c7b1a222009d47301b4c5b59d55f358f694d99a21f2702a8e8cedbbda978f","semantic_task_id":"312ee32dcb0440bf62b556bbd5dc549016112b53273e30ed677bd13ae9b14838","split":"eval"} +{"domain":"finance","full_message_tokens":182857,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":31,"sample_id":"sample:12e0a49dc33252cd0e186ad811d19de0c83a91724cc4cab1d4b6df8bd6b3ae77","semantic_task_id":"330870d6f915b842188ad914a7b0e29729a19e411f5bb43b6b8b3cadf4315007","split":"eval"} +{"domain":"finance","full_message_tokens":92965,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":32,"sample_id":"sample:ef1e9bebb78c6bd20feaa27998984c2615346037b02cbf263214a2096526e2ea","semantic_task_id":"333715204af5d2a10c2c0a64d4d578f4e938dc84ef3c2373adaa1448fbc95a23","split":"eval"} +{"domain":"finance","full_message_tokens":92684,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":33,"sample_id":"sample:6d0bc9186562b74ad3e653966c4d81bee3d0c90874049419cde2eb737faa595d","semantic_task_id":"336dcaa06bb24e6af4ac97766bd24bdfaaa2d70bddf30d5414d088627a5483bb","split":"eval"} +{"domain":"finance","full_message_tokens":258816,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":34,"sample_id":"sample:713b485de68847fee7298264604fd25200f9387a94e347dd8b4b37d174dffa72","semantic_task_id":"34c8643c7e05ce189cda2ee8e76eced72e71f9294f753587723c3d0274e3ebef","split":"eval"} +{"domain":"finance","full_message_tokens":92931,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":35,"sample_id":"sample:56d77ba18c7af1c1dfeafe985608eba4ce2aaf99d983661721c63d33bdf3e5af","semantic_task_id":"34ec104b415bb5153b8ab484049736bd3bea75d1ebd114f63cfcf4cffe40565b","split":"eval"} +{"domain":"finance","full_message_tokens":92576,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":36,"sample_id":"sample:6ba41104598be771c10d2b2cf77d5aabc52508801b68e758b2ec231b67762b7e","semantic_task_id":"3ad3c83fba6bcd5c853029aab939451a8630e9f8a20d9bcdce36ed45e9e796ac","split":"eval"} +{"domain":"finance","full_message_tokens":258815,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":37,"sample_id":"sample:dc2d28a6bc65bd869029423fff74976a9607950f042b83f9fd7565cb48049bb5","semantic_task_id":"3d055cdeac4c1685f6744e18007b386853a65deb018d550ed38027301be30a5c","split":"eval"} +{"domain":"finance","full_message_tokens":182836,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":38,"sample_id":"sample:f94fec9d81d04bce25d68e827172db638ce58c098aa01a11f50c791f98a1dc9f","semantic_task_id":"3db9163f6e0dea5012ccaa3a0d9bf4ff7ebbc90095d660c48e336b9422c08037","split":"eval"} +{"domain":"finance","full_message_tokens":92949,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":39,"sample_id":"sample:b8bdc6ac45468cb7e17526c1b33b8b078e9726ad71917450549ca0ddaa7c61b1","semantic_task_id":"3f17f4f7cf5bd2544d1e6153b09bd266f9f0fb01422505b39bf73e40b26d3ec8","split":"eval"} +{"domain":"finance","full_message_tokens":92684,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":40,"sample_id":"sample:e537b2dc86cdd81c2186fda162d3a6dd01cb00e4527fa5fdeb82c558cf061dfe","semantic_task_id":"4032aae401ed2c409b1a64f3125b4a3d9d155c42606a78f66dcade9da8757384","split":"eval"} +{"domain":"finance","full_message_tokens":258815,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":41,"sample_id":"sample:0e37f7eedea0e74e1ad32943d471059e883e713b1513c18a018fd40dc2bb32b7","semantic_task_id":"41a0255e388a07d696fe5837d677029c9a3ed7c2ce6c3f822c2ff611e7729c63","split":"eval"} +{"domain":"finance","full_message_tokens":92685,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":42,"sample_id":"sample:17ac6f626a930c03b309e56328c3509d2cccb00580637e6779118468740b0bb8","semantic_task_id":"42434028b4e2af24bb33f1de7c84dc327be258bfde8a99b4f8f24b48a58ac818","split":"eval"} +{"domain":"finance","full_message_tokens":182836,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":43,"sample_id":"sample:ad33ede2862c50c9cdab58fe4bbbe69c347fe8a96d5c5b3f4164679e1bdd00f2","semantic_task_id":"43801fbc2d74d381d8c1bcc29e70f635160fde1e847a35e6e9784e08946eb64c","split":"eval"} +{"domain":"finance","full_message_tokens":258816,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":44,"sample_id":"sample:d76f41c676cd6dccd4b0eed133d5b60a50099406b4849ff8a1db21467d985301","semantic_task_id":"43ed5dda966afd911086c83290d2cda3bde647eeab77ef91978b2ef5aa7ac342","split":"eval"} +{"domain":"finance","full_message_tokens":182836,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":45,"sample_id":"sample:02f04870679a3f1bb42d83c77dbeaa86463a05ef9f6a9d48616b9b8e07e89fd2","semantic_task_id":"4a01ed1699a0059eed561ff3544b777e2f78a6daf1ed7b1505d83a899f100326","split":"eval"} +{"domain":"finance","full_message_tokens":258835,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":46,"sample_id":"sample:4564ddade2b93a98ad9ff5ed5e78fbc149f8b2d45a99307bd039f960ad5b4b79","semantic_task_id":"4ae650176b1520923c7a54161fbb0431c77236feb22c67322f8dad18c19c11f4","split":"eval"} +{"domain":"finance","full_message_tokens":92962,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":47,"sample_id":"sample:3b9bec47967c5108888e38e03b947f23942ab1afa462fdff81a7e0d4c4b62574","semantic_task_id":"4b601f0ee63468f5c622f486016b41d0dd11496ccac95b228c4ab267cb838a74","split":"eval"} +{"domain":"finance","full_message_tokens":92690,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":48,"sample_id":"sample:119c72eae4c66c8c4383ddfe811c8609beb2970fe921c13f0d2a239d647d4512","semantic_task_id":"4b99a7a2ececa420a395b9e4015cb222efed8d18911a43d4c63e694c574f5497","split":"eval"} +{"domain":"finance","full_message_tokens":92959,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":49,"sample_id":"sample:091115f5488562a4e5d472a8204d32d3512d9bc8419e280320ebe73b2fb000f6","semantic_task_id":"4df1c3292d757c9bb608f1ac67403874e0f31b5ca783be907a67fc0beedca732","split":"eval"} +{"domain":"finance","full_message_tokens":260093,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":50,"sample_id":"sample:a9aae746e43115a25052341f00a38f33984e5d94176646cb9e1376f7502c38db","semantic_task_id":"4df4a47d31a64c5ee9a695fd32bf51224b7fc22415267a1c4a5fd368498176a1","split":"eval"} +{"domain":"finance","full_message_tokens":89914,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":51,"sample_id":"sample:045e1550b4eebcbe93a3edd789ed2fd4adfa39ed7838b810e883f4b54a5c2234","semantic_task_id":"4ef2972a410bc71864338452ea906d585790f29a166f9c577d1e53fe54ef5c54","split":"eval"} +{"domain":"finance","full_message_tokens":88403,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":52,"sample_id":"sample:0eafddfd5f5ce80401d2aa10f2adec126c6d99726dba733fb7f66626c29fb315","semantic_task_id":"532f277d7ee121fad06e4398e073a4f411b3fde414361618b632e61e4959a428","split":"eval"} +{"domain":"finance","full_message_tokens":92959,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":53,"sample_id":"sample:bc263448c999f5588755c664670535c4c74e01c91e12c9fe0fe2e27e89ae7051","semantic_task_id":"5413673b978cd93f001e73d7da87bd7cd3dfdf88ebf1b9157d6a4feb075358f2","split":"eval"} +{"domain":"finance","full_message_tokens":90027,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":54,"sample_id":"sample:53c1e297e42c6355da38a6e78268d53b879a49f232bc074b0683fc53cff3af94","semantic_task_id":"5418ac198e134c8e5c18c3c66b675557dcdc11dcd0ca0d7566465f9dcb5f0830","split":"eval"} +{"domain":"finance","full_message_tokens":260073,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":55,"sample_id":"sample:eecbf84ba7d1b180d2fc33fb48efb5de9ac338079bd6060c30634531ba3d85df","semantic_task_id":"54a90551679e51bef25086a9ff79ac74ff897871b2074d1a2e7940f65d0aa37d","split":"eval"} +{"domain":"finance","full_message_tokens":182835,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":56,"sample_id":"sample:e40313031f25c37295f9c2be2c054e2d77c29dec35c92334e22fecdeef66cd18","semantic_task_id":"5675f5f6995206490559aaced73f4acef5bd43abb581a513e1e46a7f19fc76f4","split":"eval"} +{"domain":"finance","full_message_tokens":92962,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":57,"sample_id":"sample:2c157ad9889980901c28d090000879fd2df527f14b506a4fdf47c80e13c8852a","semantic_task_id":"57ad1fffb436e39b17389c9d4aab55e5bb89aa4f0795dc20e0dbd211c0a2bdb8","split":"eval"} +{"domain":"finance","full_message_tokens":88404,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":58,"sample_id":"sample:4c38b2e2819de4c1e04697488fc8fde9ab7f1e9f826b0c135027a25018433c0b","semantic_task_id":"57f25746d426dee23686684d94d536d095fcd2dcdbb5888518aea98ee6a7fac9","split":"eval"} +{"domain":"finance","full_message_tokens":260093,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":59,"sample_id":"sample:fe4642ab43a4dad9f8dfbddba546d34aa1b30c64bef3334e55201df5c73a7eae","semantic_task_id":"58c9ddfab9aad27ee9487423043a44d6e50505068296cbe5877f39ac9e5773a2","split":"eval"} +{"domain":"finance","full_message_tokens":93043,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":60,"sample_id":"sample:e97e42646f80385f2df0844b276e856025bbedaa6a5fdd3234304cf9acf94279","semantic_task_id":"59f046913472c518c915b283146b1e90e89d13523f74f05d0ebb1d0c8dd85250","split":"eval"} +{"domain":"finance","full_message_tokens":85821,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":61,"sample_id":"sample:c457dde9b7cc8e2078ec057a1c1739064a0d885f78d3619a1b81e8fa3ad4f68f","semantic_task_id":"5a6ff6ce11eae5441d5ea582b8dc9471580e19e9197c963faa87e7addbde98d3","split":"eval"} +{"domain":"finance","full_message_tokens":258895,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":62,"sample_id":"sample:1a142ff9f4d0fa89c45b71fb1b5a9e4597ffc99874971f3406241c7d46db4e1c","semantic_task_id":"5ac19614c9b941c4479c5a9aa7174c5fb8e3a2c4f8c8be36a3e80fb11fb2bfad","split":"eval"} +{"domain":"finance","full_message_tokens":258848,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":63,"sample_id":"sample:98f0afc91a06303ca3cbcc1da006c9dee0fa1d898577aa819d9ce7cd2ea17215","semantic_task_id":"5f67a1252fadcffd854612cf57ce08fd069a300392fcfb5b7688c01ebec503b8","split":"eval"} +{"domain":"finance","full_message_tokens":182847,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":64,"sample_id":"sample:0b90a041758e2c1ba07c044347b52f3ff46cb170c3a82f6c1944929c7824edc1","semantic_task_id":"602595e950114acc0989055d0689a4323bd43b309f46b37f172ad685276b42fd","split":"eval"} +{"domain":"finance","full_message_tokens":260074,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":65,"sample_id":"sample:9728e836576641de409727c624dea34b6578b03202f9074194fe6e87e38121c7","semantic_task_id":"605505b6e825b030b19f0d73df7de6e467b8ff1a35228b90c58f579974455f95","split":"eval"} +{"domain":"finance","full_message_tokens":258827,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":66,"sample_id":"sample:3d46356652dd363573953a85b61baafcd3802fa5e0feaa863eeec8b300d21432","semantic_task_id":"60eda5cde9730a87ffe721f54ba0d9822adbd329cbcfc5b8e34f38fd2d7197e9","split":"eval"} +{"domain":"finance","full_message_tokens":258851,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":67,"sample_id":"sample:9d3e382fd2a39f2dacce3e1841522af35a00c09c953e5891cb29b58c6607a9ba","semantic_task_id":"637c3eba1da1e8af6c791ebf17fbab934d2bc4f67a54bcd14a76e8b256aa8163","split":"eval"} +{"domain":"finance","full_message_tokens":182912,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":68,"sample_id":"sample:0528b0e09a1d5c1c862f549b97aa1105a48517c2b233f79e0bb8abcfb5309c2d","semantic_task_id":"666941f1f94ae5cc715332e4b98541180ff53529161039830749eb3bd0d56875","split":"eval"} +{"domain":"finance","full_message_tokens":182835,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":69,"sample_id":"sample:75e28dcab719e7180a7df8d88e769d6c8cadb911c2008116dfdc91aacd8e07ec","semantic_task_id":"668869377bd62cac108058c883f0b1da0315e01652f296c496e47a0a6808e9b4","split":"eval"} +{"domain":"finance","full_message_tokens":92689,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":70,"sample_id":"sample:04619454133cd4b03a4271a7199711ea44f7eea31f50352ce802e0000e711df2","semantic_task_id":"674b33c9d2957f69abb68e49e666b536427c34faf53ee0e3ca23de7c86f6e85a","split":"eval"} +{"domain":"finance","full_message_tokens":92691,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":71,"sample_id":"sample:8e3f500477348edd8c17e669ee653ed87953a019ddf5248c6438ab2a48dee59e","semantic_task_id":"67be02d97da9713d565886a21fe49a998bcd07633b89ce4aad5597ea4c8246b1","split":"eval"} +{"domain":"finance","full_message_tokens":90071,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":72,"sample_id":"sample:4142f693e65cf3fd5fa788281a7e009d1d397ffb752b41207331ce8d9bfc2e3c","semantic_task_id":"69205803e46b198c22f24fbc38ea79fbc988500317616d53a9e12ba8552afaa7","split":"eval"} +{"domain":"finance","full_message_tokens":80336,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":73,"sample_id":"sample:a0ce6b5df053b99adf6ee057fc2215d5017d346a3cd5a30e15bf8b890640f52c","semantic_task_id":"69bdd75d10bda0072b88f5728fbd52febd5c5e007a44078bebc3a3666665666f","split":"eval"} +{"domain":"finance","full_message_tokens":92929,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":74,"sample_id":"sample:a161dd84711c6cdb369939d9dbe1b3fef0d22954b36e98bf1d6ca359327dcb9c","semantic_task_id":"6a8e26a6f2778a2786627e6e32d6341e4068ca0564fb0c068e8f6b6dc8adcc76","split":"eval"} +{"domain":"finance","full_message_tokens":90025,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":75,"sample_id":"sample:86cdb71fc76b78be9cf354eb0dd49a3c02e3c1d2cd0984e7b5cad3e86cff4fb8","semantic_task_id":"6f5f4efcb9e404c788f349a0f1c9be5d5f833215d028a1584c257996c0bb33ec","split":"eval"} +{"domain":"finance","full_message_tokens":260084,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":76,"sample_id":"sample:615b24b50b7a8959d3d365daabb13ce55ee3a9d7008c41b209849363293ec7c3","semantic_task_id":"715f6993a32d70998aeac4c38b6d339632eda7b3541fbc93dce88b3cc41383cc","split":"eval"} +{"domain":"finance","full_message_tokens":182855,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":77,"sample_id":"sample:398a5e1b126ed1e6aeeb026b9a458d66260ad6b8113b372717e749623b0e6af9","semantic_task_id":"72f89bf1db5ff06b085004a733fd40ffb93e1d7d7719e70fef67d5a1f546fabd","split":"eval"} +{"domain":"finance","full_message_tokens":88404,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":78,"sample_id":"sample:7372107376b70eafc908329bf963c84b7f78ddac5ddd75649c407a51a0a95227","semantic_task_id":"73879e87ee10d6e885c5f2d17b3a27b769dce21fbded0d2194188b6dac88563b","split":"eval"} +{"domain":"finance","full_message_tokens":92964,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":79,"sample_id":"sample:a0f566b696eab54aa55178cb71032d8ea60b96b039634abc82409f228a91747b","semantic_task_id":"739ab589c6be5515b4ab4d6531e4fb7dab74649a360717f0a3e3f072507c699b","split":"eval"} +{"domain":"finance","full_message_tokens":260072,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":80,"sample_id":"sample:1be52322f4992067c54b294b5428af536bfe0ca3572b37a52c9ccadcccf15c86","semantic_task_id":"74d22f329b966db7c4bfca7f18ae2598e7c7cac0cf0e3febd8716d290c7f5003","split":"eval"} +{"domain":"finance","full_message_tokens":258815,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":81,"sample_id":"sample:b7951d817d7d0953ed735d8f8d19c7347e7af08e62dc4ffdaa555f3dfb4cde4f","semantic_task_id":"75a87e667b6963c485139839fb5f40a13c89c63185432282ed110ad089f2b6ea","split":"eval"} +{"domain":"finance","full_message_tokens":182856,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":82,"sample_id":"sample:4ad13790ee9342cfe8e5b4b57e04a80375c6741a98144176db65d663e92f24dd","semantic_task_id":"76bad27f1c8cd5aaf345226f08506cccac208af15c0dde2ac76f231fff8eab80","split":"eval"} +{"domain":"finance","full_message_tokens":92926,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":83,"sample_id":"sample:8e276cd1482e428d3fc136dd149a01155cbbba2de138c6f77515fa0a69e53614","semantic_task_id":"76db2a45d03292b108bb81ee82014769906b70b24d7d24b2ed404d09ad8c7721","split":"eval"} +{"domain":"finance","full_message_tokens":80388,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":84,"sample_id":"sample:63f6f2c158c706ba2810daec23d20bc6e56de9715d1aa5e58d2d837ac9154d0b","semantic_task_id":"7a1340a6669666997f34a3d959f9809d3c5fb032ecefb63b740aa58ff94e9f95","split":"eval"} +{"domain":"finance","full_message_tokens":260108,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":85,"sample_id":"sample:6bd5f2462a081c0ee3fcc6c7349edad5aa516638d2234e336445e2b38b0d2db5","semantic_task_id":"7b8af4bcfd76ea49789fa50e726183ef49365bfc0ce18900525987c26ebbdf4c","split":"eval"} +{"domain":"finance","full_message_tokens":92578,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":86,"sample_id":"sample:86cdf3756e9f2626f6865e10d0b6e4e54058b3de0291218b74444111d2990e4c","semantic_task_id":"7cca20cb7e026bdaaa83075554e4e7dd6e695eb5563bde44ec220242f13bcdaa","split":"eval"} +{"domain":"finance","full_message_tokens":258836,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":87,"sample_id":"sample:06b8ab86be2cf2b5f4e5cd64a58faae09ffbbc80eec32c92b04806a35da34e4a","semantic_task_id":"7cde732aeb1454a69148e33a9d80dbd905bf9b05378449aa9d17ea85be8eab70","split":"eval"} +{"domain":"finance","full_message_tokens":88404,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":88,"sample_id":"sample:4b5589ebb5dff14ea4fa50c8d651c09b9140a9e3b231d5d40f873f121a0083bb","semantic_task_id":"805f5d91fa3bcc803f134827e833558216069d13984d83691b162c2d4f207eaf","split":"eval"} +{"domain":"finance","full_message_tokens":182871,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":89,"sample_id":"sample:80f6db0b7106c706d68c20954cbf4e8ea877c8d64bf442f593424f63edf8e8db","semantic_task_id":"848622793510d49088838b3b909ad49cc737d38a5edc2f5df67f17b5c0e6e633","split":"eval"} +{"domain":"finance","full_message_tokens":92965,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":90,"sample_id":"sample:ef90939f7a1a1e06965459d809a7ad00fc0131c36a10d06105d89c47afa4e017","semantic_task_id":"85cba71f78f27374f8cb4b161877a6bbdb89e8427e8ddeef366d34a90e5f8a5b","split":"eval"} +{"domain":"finance","full_message_tokens":80336,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":91,"sample_id":"sample:6a67b9e0c95b91dde95e75eb4f40c4d4d77793635105a07e6f13c8639601585f","semantic_task_id":"862dd72710f86bc94318ef3874be5e91ea6a13ac05b0365d3f9bdfc5cda25c74","split":"eval"} +{"domain":"finance","full_message_tokens":92925,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":92,"sample_id":"sample:5c92df955d5105ab5215e6ea79f7cef4300166e36655eb036af86ddbf70c4b0e","semantic_task_id":"86d3eefa9c69bda7f4007745876a6e006c279a87279cbc59a6d106c92bc51012","split":"eval"} +{"domain":"finance","full_message_tokens":90025,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":93,"sample_id":"sample:af3443245268eb6f34aaea171b91157cc8f436d71db8abdd5ee01030e1d27bc6","semantic_task_id":"872249562a64f664dc748d0b9419facabe5ae43d610a1c6ede4dae972aa4ab2e","split":"eval"} +{"domain":"finance","full_message_tokens":182856,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":94,"sample_id":"sample:75d4c0386cfe021bfacf5484dea72f39d40c58ba3fd24fe07cab1016a5030496","semantic_task_id":"8736773477884120b1dc1fe9a830da8a94fd97decf67a49c57459aa1073e8f06","split":"eval"} +{"domain":"finance","full_message_tokens":92685,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":95,"sample_id":"sample:795e7a30c70f42a94fae1e53d2eef68ca1bc508941cd100da686121973a81098","semantic_task_id":"897ad9a47388273d8299ddcb4b1404c8cd301c17ac37a0a12f32642b37ac88a4","split":"eval"} +{"domain":"finance","full_message_tokens":93043,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":96,"sample_id":"sample:81fb1dd1721bb36cbb379a75e659bfb4749b6bae4d1ce1b27cfcf2af7015d3c9","semantic_task_id":"89989f7d3dafe20e8f6f0e96e6e87ee42247609a1562470e35fbc3667bad5adf","split":"eval"} +{"domain":"finance","full_message_tokens":92690,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":97,"sample_id":"sample:c4f0478209e91a490344ecb4d64193a5a5a90c839827a32b120120061db426c8","semantic_task_id":"8a7c04c1400b4e5aef24cc688c1edb80848979b3acba9b4487d3fa727b544792","split":"eval"} +{"domain":"finance","full_message_tokens":260073,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":98,"sample_id":"sample:781341504e0f2dafd8bbbaf2cab9cc4b9023a1de990de683e635e3eb76ce58e4","semantic_task_id":"8aa4fea28d3911ecd44f7733409ab85bbe077a419c78dc61a2114d589e11a094","split":"eval"} +{"domain":"finance","full_message_tokens":92577,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":99,"sample_id":"sample:ebe7e0a9a06444b561d280cbc522f92937da70be21899f7a8caa2ea9a96423c4","semantic_task_id":"8b36cc2018b2fcf4daccf47cf08bacbfcf433967ef1f208974daa83530c956b3","split":"eval"} +{"domain":"finance","full_message_tokens":92579,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":100,"sample_id":"sample:c86ce5ad6e4f7367bf0483e253a24774629c273476e548f38219763e5ba97991","semantic_task_id":"8b92421d8571064e4471b734374048da22f917880db0d81afb53502e08aa9401","split":"eval"} +{"domain":"finance","full_message_tokens":80385,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":101,"sample_id":"sample:637c2ac0f4c55c037942ebc65bccd54d76647aa8e28b9ad2c89d284772659210","semantic_task_id":"8ff23c3f51abed8e5d45856a20e9e51e474d8127b7a3d8157f1da7b8797020dc","split":"eval"} +{"domain":"finance","full_message_tokens":90108,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":102,"sample_id":"sample:a68611b5b4523325db1ab34793e46418d4354b841bb4e09497d823f631c3e655","semantic_task_id":"90d2cf0dc79d2a6e5b7a5313826525349e70d9d89b43eecd362f5391cb95ea26","split":"eval"} +{"domain":"finance","full_message_tokens":92685,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":103,"sample_id":"sample:98595027c42b2e22992034662f9d8d0c9545917d7d332deff74d845a9b00b4d3","semantic_task_id":"9205003156d95d4dd48b92a694b73a3b391081725f73fd2670a3d5bb4b685ac0","split":"eval"} +{"domain":"finance","full_message_tokens":258817,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":104,"sample_id":"sample:cfaec3e0c9066b426950e990e193e01a11247c8786a613c230c5580293c91027","semantic_task_id":"929787671912607287c31add14d0f74a211c554e8f62d4141e30f98fdaaf87aa","split":"eval"} +{"domain":"finance","full_message_tokens":260072,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":105,"sample_id":"sample:906de748468619205ad03e641c585738d544f357dd01103980cef667d3444b7c","semantic_task_id":"9405939f0208c51857c0851723d298078701ca4058777a1648ae9a23aac69231","split":"eval"} +{"domain":"finance","full_message_tokens":92577,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":106,"sample_id":"sample:43fca55bfbfadcaef0a0b935cc1dc5227a8f6732871fb2556237cb9b7c8778bd","semantic_task_id":"95e757312235842eb8cb05fe7abdfd0c62380f021ec38ca2ad512bc462523ceb","split":"eval"} +{"domain":"finance","full_message_tokens":89913,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":107,"sample_id":"sample:07c0c9a3fd649e5c262d19ef8851e4fd8b2c035bd9ec9b9a06313fb751af776b","semantic_task_id":"95fa3d7905abff5d098a6e3982076440ff168a7349ff364fd2809e1c67815f66","split":"eval"} +{"domain":"finance","full_message_tokens":90025,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":108,"sample_id":"sample:ea5554108572ac58994b02a532f7e3162b3d9c6f66ba2740e596b0de2afc4388","semantic_task_id":"96580bc23bb9183a5bed25e54c3871695cc3756adb35ebc6726ba1aa86fe02e5","split":"eval"} +{"domain":"finance","full_message_tokens":182857,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":109,"sample_id":"sample:b0b7b25503d446cdff9757959d7b1d4da066e59047690b39b1e4f93b55ec667b","semantic_task_id":"96a0c4e5589aa40b131dfd889009101e9d5e015c420ebb653a9f20e8a20f8a33","split":"eval"} +{"domain":"finance","full_message_tokens":80373,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":110,"sample_id":"sample:25dfd86362713e2ce5795d6e20d83c9918075941fe9d65dfb71acdc613a6a91e","semantic_task_id":"96b6b1ba443252c8a77999ca41c397dfab237749d5b28490f45e3225536635f5","split":"eval"} +{"domain":"finance","full_message_tokens":92925,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":111,"sample_id":"sample:2cf9f4fa6c31fcd8ea75a1c97a8aa7b79d243b50ec1991c523a8b29fbb91cd87","semantic_task_id":"9768309eb355cf7781c80072dfdb4d728e77860fd4babf251556c976cfd6b3f9","split":"eval"} +{"domain":"finance","full_message_tokens":260073,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":112,"sample_id":"sample:d1155fb8da98a5d9f2bf88607530da43a815fbc5c7524740982130472a2efd98","semantic_task_id":"993986875a1d29504fa48024e231c60de9db2dbbb475251bf37e2782e67518f3","split":"eval"} +{"domain":"finance","full_message_tokens":93041,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":113,"sample_id":"sample:3b99f8425d09f4a0fea2e2aa1bafec96b237567197c0d5baa79a8009cfce6a7a","semantic_task_id":"9ad2ae30982d14caacdd780e731c9a8c20f70635622c74be4a8beca4e3537f92","split":"eval"} +{"domain":"finance","full_message_tokens":260094,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":114,"sample_id":"sample:36ab86eec55301e864e23c4e7746a2e8d674c4d81f7cd1e79fd4e413230f7ae1","semantic_task_id":"9b87851bb0e2a828fd7e97b049259fe6262a7e2eff7d0a0b528da780a5f8f77f","split":"eval"} +{"domain":"finance","full_message_tokens":182857,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":115,"sample_id":"sample:078b2b2f75df7429f4b1542a926670d7e4bbac9654b2ef6423fab64834b5a6f0","semantic_task_id":"9b8b82fa789008225f68bae71837d219e62bdd8a974256dc5212d9c2e9b3df82","split":"eval"} +{"domain":"finance","full_message_tokens":90026,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":116,"sample_id":"sample:d134c5928281d5007ed1b9eb7a93d9e3082280a32d952918b6e281fd4ddb0d85","semantic_task_id":"9ea02d9a30d08955ab206cbf894bb0409cdaf4ebcb66451b2b32af1c1648356e","split":"eval"} +{"domain":"finance","full_message_tokens":92926,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":117,"sample_id":"sample:c262e5eb98aea9844df5143059f9cd1a096571a3b400a58d66e290a7b905cbbe","semantic_task_id":"a0501f23912e1eb72f0b755e4ffae990cee4d633f67719db10f5b90a52f1ce74","split":"eval"} +{"domain":"finance","full_message_tokens":85786,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":118,"sample_id":"sample:de46063ae567bab5206274dba5351e59f0c817714e1a1bf12cf6775d1c1a3170","semantic_task_id":"a10a6829ca89eca5b1620034f73f0f4f510bd2030cb532434fd1b41c385db734","split":"eval"} +{"domain":"finance","full_message_tokens":260093,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":119,"sample_id":"sample:aa559e71002a8593934f764ea3d87f19664883b525ba1492c16f8d9a45a7867a","semantic_task_id":"a2db94353160c82a21845245fb97902ddf25f8889d560e2deede5689d7ec28eb","split":"eval"} +{"domain":"finance","full_message_tokens":80336,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":120,"sample_id":"sample:b945e9ad09527d68976ee3373fb6dff9d9d82b24208214e91d6ad29b393a2e8a","semantic_task_id":"a2fb9dc63756ca0fc55f8c92490e9d3da465095705ca94081b2f57b8b4884934","split":"eval"} +{"domain":"finance","full_message_tokens":258849,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":121,"sample_id":"sample:7961d4975b4cb3b6f6d03a670119672372a499e3758ec35c7eb0d9481cb208d8","semantic_task_id":"a44f92f0d46841aaad69577e8d976db7e00ceb3d5280f13d954c7c5ded4cb70c","split":"eval"} +{"domain":"finance","full_message_tokens":182847,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":122,"sample_id":"sample:736c405c8e08fbe6debeac4c9c40b3b8b833806bed0451482d0511ba53d98081","semantic_task_id":"a48a3ae45e8deb168976068346587c483dd8c1499da87d620328c9bab647c24d","split":"eval"} +{"domain":"finance","full_message_tokens":182857,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":123,"sample_id":"sample:a1f315ad7698b37a2547e7f3d3572979f557a346ab71696cc658f75bccbad246","semantic_task_id":"a58a7593e80e0c73fef6e943dffd9cdf6a452cfe7d41ee11d06c3cb0411dd14a","split":"eval"} +{"domain":"finance","full_message_tokens":258857,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":124,"sample_id":"sample:18649d39c4345757c6cca7584b775e37d0805b247bafabc02841d9c08ae47ca5","semantic_task_id":"a9b4c86a89e3e8050313043cc70c9e1b9a4aa20c30c339a55751ff768e5afa45","split":"eval"} +{"domain":"finance","full_message_tokens":80337,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":125,"sample_id":"sample:e33d6ab239f96fa588c1ad229335dc60d0a96999542311f40392dd05ed6715b7","semantic_task_id":"a9b5a8a45b446562d86a2450a5cf279f31e29e9f57342f845ac6d039118ca269","split":"eval"} +{"domain":"finance","full_message_tokens":260072,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":126,"sample_id":"sample:620dc40eb94165f7375d96b606bac1a7ad12360f7b0c8368db7490372c04bbd5","semantic_task_id":"ac391848be43ca1997f8a4aa090761661126632f4733f727b341ea7fcbd59438","split":"eval"} +{"domain":"finance","full_message_tokens":92959,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":127,"sample_id":"sample:4b28eff6e532be215ac9484124a6f7f8cf1d83d1f34cb395ed341dc6fe8acac6","semantic_task_id":"ae78cab11e9f6b5eaef67b9488fddac87b29d4f52e90a2042712d4a35e4c65e5","split":"eval"} +{"domain":"finance","full_message_tokens":258816,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":128,"sample_id":"sample:48d68d1d39a56628df0e303d82f70c7a8e5b3de1c173e96758e149c9e093df72","semantic_task_id":"b05bc24a770cf08a8b68e4b70009f4d3fc84378fee32a2802077062e6d6d3ee9","split":"eval"} +{"domain":"finance","full_message_tokens":93042,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":129,"sample_id":"sample:11eda0423ea4f34599e742de87dff29867c92fde13346707ca432733b86e95ec","semantic_task_id":"b1dcfe0f2010124ca24186922b549e8bd34be6ce3df502190510b42576376fa8","split":"eval"} +{"domain":"finance","full_message_tokens":92688,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":130,"sample_id":"sample:b21548472ad07d87f63524b7df5e74cea3c15eaf1021af12554413781c61b4c9","semantic_task_id":"b2788f7e77c4ac309a0ce66977b2f23c81dbbd61e88068927620b8931d056b75","split":"eval"} +{"domain":"finance","full_message_tokens":85778,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":131,"sample_id":"sample:5bc51fe0a4d61622c0b1a9b0f9b97e412d41348f7ffbf9dbd048fff8af0d27ad","semantic_task_id":"b31dce02bcc5fbd9845afcc3b716d171a337f29dcf3ac1a5fc76990972861f88","split":"eval"} +{"domain":"finance","full_message_tokens":92950,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":132,"sample_id":"sample:f99ca3f6cfd9551272a750bbd86517cbf7a751336c80913c0edb61dbde4bd103","semantic_task_id":"b37efcd94da0935cf19c9a61dc98a707ee5fb9247d64d48cde6512740cc36d26","split":"eval"} +{"domain":"finance","full_message_tokens":258816,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":133,"sample_id":"sample:4d285ae2014df12f5368feaafbde1ee4912752b97282f539a7857cac94b7425b","semantic_task_id":"b5f844be2dc1e7297a989689887a8de7acfbc063bed62725cade0fc657de73e2","split":"eval"} +{"domain":"finance","full_message_tokens":89994,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":134,"sample_id":"sample:efaf9d976ef4fef5d46ac983191aa16df4cc27c674ae5f4c84dff8e40f2bba45","semantic_task_id":"b78b0ffcfd06fd02c056a8fd8aa7b56370b4015be8887173aa37b58ffb653ae4","split":"eval"} +{"domain":"finance","full_message_tokens":92931,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":135,"sample_id":"sample:a8b768f21defaf1936b4a4e5f1c5494a1b7a5523d6773f8be6a1ae512689da43","semantic_task_id":"b81765eedbf9e16d7c1c227bb44f58b449abc002c6a280a4705a2148655f7fd6","split":"eval"} +{"domain":"finance","full_message_tokens":260084,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":136,"sample_id":"sample:eddf964cc15cbda2b948a95066c7ab765193923563cf6f7a6136ec5f5e30bb94","semantic_task_id":"bb7201a37ce7df8e0d41f14710ef0ca35b71b4afcc22cb8a091a5abc37207793","split":"eval"} +{"domain":"finance","full_message_tokens":182837,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":137,"sample_id":"sample:df98e9b73f88f73cb048fde483e580f66f98b8dfcb3989d7a4a5f58dbe5424c3","semantic_task_id":"bd2a0f27fc62855cb9dcba424c55d7d222eba8d5975998554147b06d53e3fefa","split":"eval"} +{"domain":"finance","full_message_tokens":182835,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":138,"sample_id":"sample:ac3d4143cff11a4cd0216ef837ac5c5c4018b05a1533d3e05f73d31d7612f34f","semantic_task_id":"bdb1061f90bd1fad8029366bc455c38336060d68bf9ec67c63a1c226642850c6","split":"eval"} +{"domain":"finance","full_message_tokens":182836,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":139,"sample_id":"sample:8c89e26a37bf0f205606ad70639a2b59d3c83900e8c6f4c225da1d26737b18aa","semantic_task_id":"be9fccb64f9c0bab1bd8e42635c582a0668a2c08ddb124a0df3c0b4be09c9c0a","split":"eval"} +{"domain":"finance","full_message_tokens":89915,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":140,"sample_id":"sample:5365c5158d62cbce5176c558d4a5f3f16c41ccc06cdca3750d946920e1608e93","semantic_task_id":"c239ab66f4fafb4f56f8b76c70afc7d06945bc48eb23f2585cd4f29b166ab829","split":"eval"} +{"domain":"finance","full_message_tokens":258847,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":141,"sample_id":"sample:102ecebfa923cd864fb03d2e78f1ce150939e29d4da084777772862fd7be6a7a","semantic_task_id":"c32483243a55f16105e5fc56d16ca2d1efba14820e821fe54575532296ee31b9","split":"eval"} +{"domain":"finance","full_message_tokens":85741,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":142,"sample_id":"sample:a40bfc423895880138739e2986f9833e92032c7b050f6545b5f437cc1b16a9f7","semantic_task_id":"c526ba2b0e6ee901422e1f6db6fd7683f2e4e757dbba60fd97bb16cf2bd597b9","split":"eval"} +{"domain":"finance","full_message_tokens":92950,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":143,"sample_id":"sample:98d690c2883ce6f092faf52f387f2e2d3477fdc031c04324b67c1c3abab04446","semantic_task_id":"c5f8f87196d89cf3497478225813d4aaa2286db4e811ca2a9bf30cf765ca10a7","split":"eval"} +{"domain":"finance","full_message_tokens":92963,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":144,"sample_id":"sample:dcd14a7599c064d0ae877a9dd69ac5edb6b7c056021555a2552b2dec9071a419","semantic_task_id":"c653e7b9d9d816788b9bd674fc1bea843e9fa381fead7fa54746e38948aac6b1","split":"eval"} +{"domain":"finance","full_message_tokens":260093,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":145,"sample_id":"sample:8108e23109ba9dc512d87f965dd58c83e9276d63c8d3bbec61a6f38cfa2e7352","semantic_task_id":"c6bde291b57f59c9ff698c5a0dc0d8d02d62a4a936828b08c33a5e008951fada","split":"eval"} +{"domain":"finance","full_message_tokens":258847,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":146,"sample_id":"sample:818dcd2879768f9d872a15a0cfce09a7ec729a7dbc8e8d780625e5c265ec4764","semantic_task_id":"c6c4293d7200b6f9360cf0bdbec82320cae4f1cb39281b2ab616f0bc1b1d8c6c","split":"eval"} +{"domain":"finance","full_message_tokens":85742,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":147,"sample_id":"sample:eb9bf01892b246b3f583fa79369982d196b9a6fd47223afc83f031ba8b9df0d5","semantic_task_id":"c6d6bf958f10fc554c261f167b2f2ef8d6e87a8583b144cca8fd8ed3bba23b5f","split":"eval"} +{"domain":"finance","full_message_tokens":90026,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":148,"sample_id":"sample:db53d1f25cedf292d12cb7c62bb8f12e636af8a3eb02b011c1532808a8c29f3d","semantic_task_id":"c737af06a0818ab3f0e4ce91ede3aab34319354d3b3d8934efe9737d689f5728","split":"eval"} +{"domain":"finance","full_message_tokens":92689,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":149,"sample_id":"sample:08e4b7e35f477010ed47f4d135bb16c1dd30a69c5c4f198f08ec4292e3d16e99","semantic_task_id":"c74ced88f5b30cfbdf5b9a315e95d43e6f292d0ae9b35b37ac9b77fc16c928d7","split":"eval"} +{"domain":"finance","full_message_tokens":260072,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":150,"sample_id":"sample:0005870ff0b3fa042c2c666bb49378e6bdcc202699ad812284e0c360d5dbf8f6","semantic_task_id":"c77fcae02ae0b79b4da5508af820cacd4f9eadf11bbd2dd418ff162e7c9f1311","split":"eval"} +{"domain":"finance","full_message_tokens":260073,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":151,"sample_id":"sample:05ff569b54c068b29cde50a12310ecee21cc33acb68280e98241940611ec6dfa","semantic_task_id":"c7ab2e595f5e24151735bbae05dee7e1144b8ef2c6b8988c994d201dc0662a66","split":"eval"} +{"domain":"finance","full_message_tokens":90074,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":152,"sample_id":"sample:bcaf66d2608a8482d8c2d4dc0e86bd00f8c387f09dfa6471df09336888aa9a5d","semantic_task_id":"c933928e7ad09c212f51b3a99b857934ea57fa2106e83b6d43ece9d0ac1cd10b","split":"eval"} +{"domain":"finance","full_message_tokens":258848,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":153,"sample_id":"sample:649cf4b0ac648f954926c7240abd3acb940a61aecbdae52dcb772094315d179b","semantic_task_id":"ca3fdcd6e6f5ffa35f391c97cf26f5c1ea4e8d75df03f038be66a755724ba34a","split":"eval"} +{"domain":"finance","full_message_tokens":80383,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":154,"sample_id":"sample:6633a2308e85ba15e385b6874bf9c0055f2d7d97df645a2489d58dabcb75ba78","semantic_task_id":"ca941c74ecf141da91f910cd1dfc8fdafee5f61ff71a5d3c6ea6d6808e3153b0","split":"eval"} +{"domain":"finance","full_message_tokens":89966,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":155,"sample_id":"sample:ea7441d5036698171a122bbb72d085f201ce316d00db652beb6eab42eecc60c1","semantic_task_id":"cc73cf095a0ab6aa7bdd49d73b8ef918273daeac9d8176f7fe7b29dbbfbbee18","split":"eval"} +{"domain":"finance","full_message_tokens":258838,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":156,"sample_id":"sample:6be26d5d06504ea4bddb4fd247f7ae31d3c829a8076c2edfc3a03edcd6685075","semantic_task_id":"ccb93f07b0a5d6439497b4b9b5cfc84c93fc2749e750823eb60e020c49b7f978","split":"eval"} +{"domain":"finance","full_message_tokens":89950,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":157,"sample_id":"sample:c31114474cd47f4216f2c918335fa5843837e1b72a5219a8f6b9099a697c09eb","semantic_task_id":"cd165b67f2792c33f11c466fb531b49b42b967d7612e4b04477a8d2521d332cc","split":"eval"} +{"domain":"finance","full_message_tokens":92930,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":158,"sample_id":"sample:273d6fbf5cc2b596349d5ae3010b933e0034c4e6d5aac60121e86b4a03d8d1ed","semantic_task_id":"cd7713d1414682ef8868572245b2a64b5194d2f98c591b784d40456d732b01ac","split":"eval"} +{"domain":"finance","full_message_tokens":88404,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":159,"sample_id":"sample:2808182360c4c8c8f9748753e0073441aee7431d6188f2ccb4ee3b8edadfc057","semantic_task_id":"cd7e03b6d574eefa6c818d6d099a48acee5d386627686e59bb243d5913b0a1de","split":"eval"} +{"domain":"finance","full_message_tokens":260094,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":160,"sample_id":"sample:d77c9a4b5d9e8f46c327218cbeb067f248008f829f70d3b5a1255bed8d7bfafc","semantic_task_id":"cebfbe04ccab0733423bf9e40322bc545859c683a63ea871e4ea473a93020589","split":"eval"} +{"domain":"finance","full_message_tokens":88406,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":161,"sample_id":"sample:35b9b7811d2c0704b8b729e62652648485f9945edbdbee978ba5478a8303032e","semantic_task_id":"cfd534f7050da2153913fc564b0ca70d87982c4f628e849f5c9a58083315ec7b","split":"eval"} +{"domain":"finance","full_message_tokens":92965,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":162,"sample_id":"sample:e32c9ef574455166e55e6543a485bdc2dc0d578cc12f60f7cf020044473f8c2a","semantic_task_id":"d04ffed16f737b66159699ba632ce97daaffba759174a3087a76ce5740fc78b3","split":"eval"} +{"domain":"finance","full_message_tokens":182856,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":163,"sample_id":"sample:eac4433e35b4da613d513a01e9735987734928194fe2a05216b7f256b97cb636","semantic_task_id":"d067b4df47f5fbeeb60fe6d6d50971719f14335aa62fb619cb8ac82ecdb99b88","split":"eval"} +{"domain":"finance","full_message_tokens":92965,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":164,"sample_id":"sample:a10e4ec3b7b1ba7af8ce55a3a1fa267ed02041012735b34a4dd519fb1a5cdcd8","semantic_task_id":"d3fe7df90ffaf33de48fbfbba085214e1f28e45afad3a8713a3401e4873b0ee6","split":"eval"} +{"domain":"finance","full_message_tokens":89961,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":165,"sample_id":"sample:6b7f2755874226e77b17206392263a61dc07be10d0766807f96bef1185269163","semantic_task_id":"d4f15c8a6a02dd80ad977df7a66bbf3f3f7119bb7fe3f08404f18de5b177518d","split":"eval"} +{"domain":"finance","full_message_tokens":90026,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":166,"sample_id":"sample:110b1a1d9a2e7dce3d216f34899fb7dda8c4f45b0ad85b9f12e972d4556f840b","semantic_task_id":"d712a1f774225d5b4cbfd64a52c7eee1af58c048b38eba2c661ce1a333d30afc","split":"eval"} +{"domain":"finance","full_message_tokens":92578,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":167,"sample_id":"sample:cddaf97832e66203a5f4f199e94f093877054d6f2a4196ca11a100d472c9fb4c","semantic_task_id":"d9d482de7c877b07d274675c35bcbba442de740be6f575e4e3dfb3bd4cc96b30","split":"eval"} +{"domain":"finance","full_message_tokens":260093,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":168,"sample_id":"sample:77b0149fabfe3d0d0390e596e817e8a644f71c0f1b071fd74f84baa0f835632c","semantic_task_id":"dd2dc3b323c5997c88f96fd087a819ef95f634673ae370cf0fa4c4f302e7cac6","split":"eval"} +{"domain":"finance","full_message_tokens":89913,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":169,"sample_id":"sample:2aca760c65b9d964d5014c3b9e16f37483fde93910defb63f3de99a310cba154","semantic_task_id":"dd801d08a60975908c92c4b8904e1de6f77c0f85a963eddd038307be4a17072b","split":"eval"} +{"domain":"finance","full_message_tokens":90026,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":170,"sample_id":"sample:a7ab5bdd908246a86711aa85eb3f3408c9f1974e6df9c54f09696b7f8e3e143e","semantic_task_id":"dd81593f9abdb729d54f1fd69a66f6dc15aa947d3a59e24f1035254547acf2c3","split":"eval"} +{"domain":"finance","full_message_tokens":89914,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":171,"sample_id":"sample:6edd6154cb34e20cc73309c2e1bbf7fc6fdcd763bb3f16cd1a63e9908fe6f388","semantic_task_id":"e334292b1f23d3898701e8da67ec848a0d5bd90b9375742a7a7c91b3efd47248","split":"eval"} +{"domain":"finance","full_message_tokens":92675,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":172,"sample_id":"sample:9801c8188c739edb9e200fd4718bdef48c118f5422933dfe4be0d9066e3f5974","semantic_task_id":"e52466cfe0bde1dd16762b0c12e34082d9092b45ce9f57a890738d2be046c464","split":"eval"} +{"domain":"finance","full_message_tokens":182836,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":173,"sample_id":"sample:344fb579b0ee6f2766eafaf3df8be013040729de830af9760be7e0b7b635b8db","semantic_task_id":"e544fef899cb4394d954813ee575cfac6ddec430f602d609d66c1103e96c9b22","split":"eval"} +{"domain":"finance","full_message_tokens":93041,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":174,"sample_id":"sample:a442657cca1f8f083f549adb93c515fe9c5feac5f8e0d86456133ec12d971b92","semantic_task_id":"e5a8bdc504d9c8f4daf7978085636e01036f77aed8d28ad38c98b21c88ef6fd3","split":"eval"} +{"domain":"finance","full_message_tokens":89958,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":175,"sample_id":"sample:f78bf0ae4aff7b4db1c9e3924d1c3ca73d01b53f9a540b879aa963fd3ca79519","semantic_task_id":"e781e4ef5fee8a95a4b49e7f1040fd96d5386b9c474349415cc09bd7e9c7c25f","split":"eval"} +{"domain":"finance","full_message_tokens":88405,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":176,"sample_id":"sample:f6b31a4369e64b444bc767a8a3e01c793dc69117fe47c4fa4391cd50179ef877","semantic_task_id":"e870ef32f4337fcca40aea616a7e06a113a29cabadca9cde2f577a4eebf889b3","split":"eval"} +{"domain":"finance","full_message_tokens":258816,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":177,"sample_id":"sample:3866564885a0561bc2d83e2a5936393142a247f7c54cefb2fa49c6c1a1df6c6d","semantic_task_id":"e8af231536db78791a772051bfc266378a10aaed93e2bdbe1917f8e93dc17601","split":"eval"} +{"domain":"finance","full_message_tokens":85741,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":178,"sample_id":"sample:b69dedc5f042402ea5aa3325ce845f3b74e5675eb40e26212f78d16efc141394","semantic_task_id":"e9540333302ff08664a08644dc91e385b0cdfdca14371ecb583f55909bc1f379","split":"eval"} +{"domain":"finance","full_message_tokens":258892,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":179,"sample_id":"sample:1a6f8b9587dd2e4442805319ee90469059e53cfec6d95dab54fe5a331925b248","semantic_task_id":"e9cda7f836f293d733b9184b8a473e6f57af79cf95df4042dfee3fb1db8619fd","split":"eval"} +{"domain":"finance","full_message_tokens":92577,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":180,"sample_id":"sample:74d9b3d7950a030e6a12c1f1dfb26813c06c1553481aad13041f7d90990ab3bb","semantic_task_id":"ed9af43406b18a190465d3fe065cbe13ce5baf3542df25b769fc34c15367cfdc","split":"eval"} +{"domain":"finance","full_message_tokens":260131,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":181,"sample_id":"sample:18ab5a828b1f64db658c2d69702f8b9bc636ab218c8a9786674faef108d4aac1","semantic_task_id":"ef8f164949f6af46b8a3ef5e1d6b6e6358fafe48e370d623cdbab3889eba49ad","split":"eval"} +{"domain":"finance","full_message_tokens":260149,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":182,"sample_id":"sample:c6c9253bb98d342332988a9c1c3c389b281bafe1709dcb38cb9ef091124af5ed","semantic_task_id":"f00c4be5de94a4f7561402fa7c61ca478217580d74efa8d88e36481a96b84261","split":"eval"} +{"domain":"finance","full_message_tokens":80337,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":183,"sample_id":"sample:dcf28cca30aa188cd3edc0fdafde3b7fb967022f6a1dbaa844286a62b5a2711f","semantic_task_id":"f0c0e36358ab409a4aedd4d5458dcb3c19940115e1568fb9ecce671a7cef1aab","split":"eval"} +{"domain":"finance","full_message_tokens":258848,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":184,"sample_id":"sample:1167e6ff3842dd7e0b9891ea035aef4655540ad4f523bbdcab192a2289a1be3c","semantic_task_id":"f2b87e227e9765aaedc5c8ddce3048bd54cdbd9f8d2e68ebbfde6e5d8239ece2","split":"eval"} +{"domain":"finance","full_message_tokens":93043,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":185,"sample_id":"sample:4895c6744b8e64dae260efb28d717e68f1f9b5e1889c367e8ae65658cf001f84","semantic_task_id":"f4db1d18f70f9fd4a3130f90027af7c47b4e164a28f86e8aa5f06408cb9ffe0b","split":"eval"} +{"domain":"finance","full_message_tokens":92690,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":186,"sample_id":"sample:3b21d07781fec13368ef6604d6bbcad65e3be7183cbcb3a96f2ec14cb4c65a62","semantic_task_id":"f5d288a753f4c8238e3c6e0d236d2d34766a770a4cc8c8f9b1c6bc23a4bb67fc","split":"eval"} +{"domain":"finance","full_message_tokens":92577,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":187,"sample_id":"sample:38ba024b11f75356b8fb47db477cd8cffdb2a50ac5d6bb84dd1ea63174c1bf30","semantic_task_id":"f9e3a46a2a10eb2988fcaeb06b71f9a7e73341c7042df4264221bf73d2c72cef","split":"eval"} +{"domain":"finance","full_message_tokens":85793,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":188,"sample_id":"sample:d6fbebe681f661ac946aafbbf7a5bf678b7a072f3e6332404da1f577ed240825","semantic_task_id":"fa5b307bd95b934b9a27c470c78fb8ba6053293225092e8cfabc89104a4b6e46","split":"eval"} +{"domain":"finance","full_message_tokens":182836,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":189,"sample_id":"sample:30078681e729c15204feac06069ca1b6d2cb85d2e36b3aac28c121874e15a264","semantic_task_id":"fad2a258390164aa13aa8447ff8b23b15a14a080e00bd0faedb20b019bd47180","split":"eval"} +{"domain":"finance","full_message_tokens":92930,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":190,"sample_id":"sample:dac5b89779b211974a8ea496454271b805938e1c31a0002f864f521fc2744af4","semantic_task_id":"fb6f1dc8ef07653bd2b3f52dd2b199c7bdf966d2ac180b060b67eb8d5f1b3772","split":"eval"} +{"domain":"finance","full_message_tokens":89914,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":191,"sample_id":"sample:a8d4b3a0d3f2fcac7f817d7b3f697b98d883cf99e6463fcf6813ae1ce3b08199","semantic_task_id":"fd89a4dfcd5f1e243f57c858acf5ad9802a9a601cbc6f97f499e7fc17df6c270","split":"eval"} +{"domain":"finance","full_message_tokens":85740,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":192,"sample_id":"sample:e76f676698402ac59d82d4e74a5c6cc4c10ece64fcb2def471190e443a1253ae","semantic_task_id":"fed4d72da91c8026b0efc29fbf26ed8322ad9a08fefb03a8d46f52db96f7e5f6","split":"eval"} +{"domain":"finance","full_message_tokens":85741,"group_id":"0000002488","output_file":"finance_eval.jsonl","row_index":193,"sample_id":"sample:fc46367def8716e647585db2d937cec04782f9b90e6abf055f3147e936a0884d","semantic_task_id":"ff56dc01fbcbb4e3434871bc215b27719a01d67882eab52be862fe000c8ad2eb","split":"eval"} +{"domain":"finance","full_message_tokens":227393,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":449,"sample_id":"sample:b10341e1907e35887cd45f2aa34fbb174cd8109b2da3ab2293f7062dd41f9527","semantic_task_id":"04392190acdf228b290c60f55439cbd96defc0cec30114fc832ae2c5d7564f06","split":"train"} +{"domain":"finance","full_message_tokens":115338,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":450,"sample_id":"sample:40a7b174077bfc18299d564d397c9a4f533cf4952ddd2c8d58b845c6053fd088","semantic_task_id":"077cb227155d622d4d03d2176342d6a89f7b93be924403c72efbc6e0a18ae037","split":"train"} +{"domain":"finance","full_message_tokens":230786,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":451,"sample_id":"sample:f4adaec485f078df323bc6bcdf8231d524389d5a7ec83eeffe8fee5e8be79e16","semantic_task_id":"079795b24cabccb281d759feefb4d4923865cf25ddada65c6a80c53e77fd1e3b","split":"train"} +{"domain":"finance","full_message_tokens":115171,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":452,"sample_id":"sample:8243ffcfb221bd27d983b37fc0c9d007a2e767c1ca10a454c3fb3e0a299ce477","semantic_task_id":"08db0d352febfcede763c7528e7c28a9b1cad3b35d100a687b1c51fc5bf39100","split":"train"} +{"domain":"finance","full_message_tokens":115573,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":453,"sample_id":"sample:86309459e73185aeb27648fa9fbbca2e49730e3c0480be728cd40477ca5656cf","semantic_task_id":"092a6626b680e1a912f94ff22d7d72966b2639547573d9986bd68691bdf4bbf7","split":"train"} +{"domain":"finance","full_message_tokens":108302,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":454,"sample_id":"sample:357c8814469ec0f4b2be1d7a1d5325838f789053cdcb7481036bbaa3bb673d3a","semantic_task_id":"094c5d34ac7d197a1124d2ebe565af9c5dd9fecfcff2bbeb98ff81e39cfa9d8a","split":"train"} +{"domain":"finance","full_message_tokens":230820,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":455,"sample_id":"sample:fd97df9d56cc187f2d751da0180f7f2dcf5a01973a0a5865a2eb36b0175442cb","semantic_task_id":"0aeca7b97bd5e1da54abe24e1212b6746c8f1f47788ede4f45094a9488367737","split":"train"} +{"domain":"finance","full_message_tokens":230788,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":456,"sample_id":"sample:78edafcb4ffd95db2dd02ef8462da633cda4d84c9a88c9c7175684ff332adea2","semantic_task_id":"0b934ff4f48ced8705befd77d5dce1b61ad282bc9fc2cefc2ed847f418134fd0","split":"train"} +{"domain":"finance","full_message_tokens":227392,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":457,"sample_id":"sample:4b1dc726f0312f0a0bcdbe7bd163fc6431a56ff23907659b2a83257fad5f6df9","semantic_task_id":"0d29b1b6da7530dabbbac6b17d4c82e53952e08b4cfc97234ac6673a4b7b26d7","split":"train"} +{"domain":"finance","full_message_tokens":230798,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":458,"sample_id":"sample:689403eadd0bd9fb3f24b07d502baba5da5ca07eda19293284e0d855797e4101","semantic_task_id":"0f69d514a3b2137a8391fcab638383a7ec56425fdb7f9295260ba8661ced93a9","split":"train"} +{"domain":"finance","full_message_tokens":230788,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":459,"sample_id":"sample:d1701070aba5b9f56810fe669dc9241981bf274da3555f29f9299aee18243c33","semantic_task_id":"1181d9f5cd0a506cd06b92743b6e9657581065c54e0394b71d52db95600c202f","split":"train"} +{"domain":"finance","full_message_tokens":227681,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":460,"sample_id":"sample:ee7c4512453c992f2248b316dd695b1397b593a8291427e0c00581efa12832ef","semantic_task_id":"11c40f135d9a911e7a7a3e85ee27b2f2cbb3d53fc9dcf8f8ae3ef455c1dd81f1","split":"train"} +{"domain":"finance","full_message_tokens":230809,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":461,"sample_id":"sample:b0d119ec64e03042fbdf4d09564bd20a06c245e62325754760d7ff3d9e3ec362","semantic_task_id":"12b27b4d72dbd9703e16cf98ef81952f644d25ec33038ded03ad658e94a782f9","split":"train"} +{"domain":"finance","full_message_tokens":115636,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":462,"sample_id":"sample:159fbb87a81ed837c2b8ffe86fd0e3b3d92dd2a6b848067e83acd45f548305d9","semantic_task_id":"12cea479c585a9cbf688bb691588f82a792140bfb1126240822974508711cfe3","split":"train"} +{"domain":"finance","full_message_tokens":115641,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":463,"sample_id":"sample:b8da2790a369934a58616dc4d9c9b5b6223d5ca6352c750a0b1a9e9581f5aae2","semantic_task_id":"13aee5c1529634c0df0dee4632c1d792e7882456c34c9169c09964995bb69748","split":"train"} +{"domain":"finance","full_message_tokens":112366,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":464,"sample_id":"sample:efcdfb2e2b98ea37f039cacd4c5cd1e3b1dfb0d5a2ea18d118374f643b63ccc0","semantic_task_id":"152b1f201add965d1a47e1be269a167ab4b62e0dc340ecd940c8ccd191b9f29e","split":"train"} +{"domain":"finance","full_message_tokens":227454,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":465,"sample_id":"sample:c8325e8c6085b42517e8f7455d21d31b33173f2d5789ed00879d9886e6e6c841","semantic_task_id":"180fa4d4bc9d1b1d189f41553abdcfa8620c3271e598e9869fcf8aab9a26ec86","split":"train"} +{"domain":"finance","full_message_tokens":115584,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":466,"sample_id":"sample:38e898e2dcadfde7864e7c820160369116b0272eff5a013eae240a8681b6fc23","semantic_task_id":"181055324b8a231153b5a465929f85b44722dd6b44ed29b5e9c41f0b599f867e","split":"train"} +{"domain":"finance","full_message_tokens":115187,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":467,"sample_id":"sample:6a52b12638c45301f2ad37302d3a1f27b152c0e03f4020ff5f31ee3b0d5d9830","semantic_task_id":"1841e12bb78cbe9af3175410b1bae3b09d7665e4c2bc5a8902ab09bc560187bc","split":"train"} +{"domain":"finance","full_message_tokens":227429,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":468,"sample_id":"sample:e5fcb80fa755a784cd5162759383ab502945c421cdb3372c2044dfd324993c13","semantic_task_id":"1866b004e710e4e0325b9c3c1fbd3e77633d3da52bb8e8d7cb35a444ab79ca52","split":"train"} +{"domain":"finance","full_message_tokens":230808,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":469,"sample_id":"sample:561162013893df5f4add08214ea29e6d13ce0a10fe2a969bbe36ef06a00eac32","semantic_task_id":"1bebf3150db261e9a7c4f15747e30889ba6eac61f31c3a18382123f79018aa89","split":"train"} +{"domain":"finance","full_message_tokens":227662,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":470,"sample_id":"sample:622c19ff8b2ff34a5bae8058b311ad5cede8c78a0e547bd36f1a6cc7b38af143","semantic_task_id":"1c96f10c04a740bd31c7514f90ba0efc0baa4dc01627850e8d96dbdcb1016bc9","split":"train"} +{"domain":"finance","full_message_tokens":227394,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":471,"sample_id":"sample:427649499d614991f6dc5eb31b5dd865c6a3f798029ef7eef365e77e12415867","semantic_task_id":"1d0926b074cff976f7884e4a06029c9e8f81444ad1b03c03085118047d290c7b","split":"train"} +{"domain":"finance","full_message_tokens":227661,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":472,"sample_id":"sample:cfca304dc43ac05c2a30c7d3450981b744527b7f6372c20d1cd4843e23004982","semantic_task_id":"1e66cbdf8ef3e295577b7451331dc59e771ace83081946b3ac53ce951ee9d03e","split":"train"} +{"domain":"finance","full_message_tokens":112189,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":473,"sample_id":"sample:578b7530feb627297c5203352fde318b1596729f6419c6407bf98340c015b287","semantic_task_id":"1f7f4e30d995f68066a4489eef3fb791d6e83302dc827e2d89b4d35cb3931906","split":"train"} +{"domain":"finance","full_message_tokens":108303,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":474,"sample_id":"sample:d15d79be0ddcff07b21f3f541863219cc4351d39f638c34af2bb1900ad2f77f8","semantic_task_id":"20eb3734835c897ed8b4ee3d137c84cf9acd6d2872ae920e81b7437b72338260","split":"train"} +{"domain":"finance","full_message_tokens":115181,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":475,"sample_id":"sample:93e02f7f4a7516afe87b34a8802505d235301a7f929ef3822087c557aecce9f6","semantic_task_id":"243640e493b2172700bee92b97062f3614e5af7c91e9b66d8a0f15e92841aed5","split":"train"} +{"domain":"finance","full_message_tokens":108304,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":476,"sample_id":"sample:fca34d55c0d218633b7c3cbf125b14ddb31e10feb72adb411c7e5d75d7db799f","semantic_task_id":"24efadfa0be810ed14b6de5a2cacdccbd37e08451ef000f7baccdd5a2594ef7f","split":"train"} +{"domain":"finance","full_message_tokens":115637,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":477,"sample_id":"sample:e01e1c86f908f69c71fb017b8fe5fe7197b1a57166fb96f7b8254b83b6c686ea","semantic_task_id":"259460b9e248ac8afbff9888f14f701016b2e5ba5c79676608a3c05ba1dbc633","split":"train"} +{"domain":"finance","full_message_tokens":115586,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":478,"sample_id":"sample:e81f4078d28f288e1e7db885f8fc69ed41b1b4aaa1cf4795b6828a97f42288bb","semantic_task_id":"26bca14db7f049f31723c7f5a4ef870b855c849c79e753e7c2b510a681428570","split":"train"} +{"domain":"finance","full_message_tokens":115187,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":479,"sample_id":"sample:d7a61e3f0c96c2a0cdc2c991b0a69335ebec794e4c861d27d922213bedea7246","semantic_task_id":"271933f622222de2c07d3c4e5e4b4135097337260c2c53aba843f78b3e3256b9","split":"train"} +{"domain":"finance","full_message_tokens":115636,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":480,"sample_id":"sample:2f4d27514821c09e598f4c70613e0d15e9b07bb3c8db22b929fce98669a4553c","semantic_task_id":"2812c9f6d7b5dbb5b88f366e50da9fd67be24f7937191a8edd4efee94922ce4f","split":"train"} +{"domain":"finance","full_message_tokens":112227,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":481,"sample_id":"sample:449f8f9492ea4842ebf90cd10cefd25908e1e7a73f98a6d017af531d709d9066","semantic_task_id":"2a1408c939f118824b223329b67c2ffbd4a81d90043a64b087da3127f97ef9d5","split":"train"} +{"domain":"finance","full_message_tokens":112244,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":482,"sample_id":"sample:94180d3fc16391a24af6d8259e9eb23e84506a7f03c7df93f290eda42ce1542c","semantic_task_id":"2be22b36931a05d88285e03adb850f8f527241761696bf10dcabf0bb11165af5","split":"train"} +{"domain":"finance","full_message_tokens":112189,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":483,"sample_id":"sample:c8d724133b95f9f0fc27e51341673fe95aebdb039f5d8ced201a661b04857df0","semantic_task_id":"2bff674bbc29294384274f9d0c7c19ef35b82dafe6f011e7c060dfd99b068766","split":"train"} +{"domain":"finance","full_message_tokens":115339,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":484,"sample_id":"sample:31e26475dd27b587e1dbbc1c70435a96eef82053f68e4eaff3b199eee8de450f","semantic_task_id":"2c133e0a7304f4adc29ce6061118c6a159ad0c6d44eeab44dd4a5061cd6ea5a6","split":"train"} +{"domain":"finance","full_message_tokens":227660,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":485,"sample_id":"sample:01d2d186b310761f9c348d3cb59b451527e1ddd79f3bbf07ee9cbf9ea183b924","semantic_task_id":"2eda607dff02e801a8552a0727987158bebe7afd8c070553c6f759d205787e8b","split":"train"} +{"domain":"finance","full_message_tokens":115186,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":486,"sample_id":"sample:562f09b6245625e9bef10fadc81534383ec5b0f67af1e8f87c7aee516cb8e326","semantic_task_id":"2fedb5dafc1a4a48d6fab27b0b36755660e9ca555c57446d312009e22fa07cd5","split":"train"} +{"domain":"finance","full_message_tokens":115588,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":487,"sample_id":"sample:12f23e685075d3f1659379d0aca48ef69f24ce66fca8778439aa21900bff30f7","semantic_task_id":"3024245f5fecd4f8b246c7ebcb616131158c4912cb0dee00085d6b0608115e91","split":"train"} +{"domain":"finance","full_message_tokens":230787,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":488,"sample_id":"sample:920d6397ce102c49ab0945d62c332b0836439239d11ed4d6d3ae8cddc154771b","semantic_task_id":"32a2d7e27eed7ff3c89acf1ac30de03c4c04a7dfe67db258ba6171e27d8b8e8a","split":"train"} +{"domain":"finance","full_message_tokens":108387,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":489,"sample_id":"sample:5660a67e50a3b455d100c7bb2e1e1ba36c525069b6512838b9fd5fa5517eb817","semantic_task_id":"32c4230368786030d08c70bc2df84927be0b1fbcd0b548b176093d5e9b1bd92c","split":"train"} +{"domain":"finance","full_message_tokens":115453,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":490,"sample_id":"sample:38dac2b8fe1bad3bf209a51866c98016b6a599ee16af8316f4558c68bbf47ee5","semantic_task_id":"336feb1ec5d439fe0dbfdd4d6315b3ee06303c7eca8c2b29f9897a647ebfeb29","split":"train"} +{"domain":"finance","full_message_tokens":115186,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":491,"sample_id":"sample:b50bd80427dbb5314d050e5f0a4fbe35debdaed4ffbe9b69f7a9f4496884817f","semantic_task_id":"339fd4ee296221dc36f4161842b778d887c8fc6f1c6cb35ead4c963f0072a839","split":"train"} +{"domain":"finance","full_message_tokens":230787,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":492,"sample_id":"sample:3f5b0c3611e7fcb1330311ed6a9dd985e3ee596d6976ff503ad6687adb1f4553","semantic_task_id":"33d33a285860376cec23d6c622da61cc278fe21f5c789ced7d7dcce0a6036a01","split":"train"} +{"domain":"finance","full_message_tokens":227671,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":493,"sample_id":"sample:45a3b307f10de3c7cc952588b98c2b75e3c7d875635ea43b88158499e74d5baf","semantic_task_id":"3790359b65215364b63197e76405da7c56c1df10733374756d5c445bc07e16c4","split":"train"} +{"domain":"finance","full_message_tokens":115626,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":494,"sample_id":"sample:656d8805c2014f1fd705e6e598a70c43ce77cb5362a106334079207b696e5ac7","semantic_task_id":"38633bd22fa9c80a4c0cb58994b50ac08982246f098f3cc57c5f33e24a1ff04c","split":"train"} +{"domain":"finance","full_message_tokens":115340,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":495,"sample_id":"sample:ccf6b22b810aedf145ee1106765318181925bc6f11c57b623d6067eb12140e00","semantic_task_id":"389690c7d65cc4c48ec91a79cbbdaaac04389951d50bcb4b82a032e89fd5a467","split":"train"} +{"domain":"finance","full_message_tokens":230819,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":496,"sample_id":"sample:d0a377877f3b1a6dfd1fa776782bf1d4d94c698de599cf66c2355326dd7692ea","semantic_task_id":"3ac9860dfbdaccba46d3bcc44549a4233a70b86af0c35df9c9ce7d847019225f","split":"train"} +{"domain":"finance","full_message_tokens":115475,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":497,"sample_id":"sample:7a268d153b3da3db661995c1cf72cfb9fdbabf6ecd55944d0b8de07e25594050","semantic_task_id":"3aead7f6a961d45b4497301c10cd9767501d76bd0b844e6a156e7ee1fb211981","split":"train"} +{"domain":"finance","full_message_tokens":112273,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":498,"sample_id":"sample:7d0c7684564d8f961ad0684ce80ec9199577bef78679994b7ed86f9f59897473","semantic_task_id":"3c41f393d317caf3d93279c7fac926d527b5ea05d2a7048a00cdb7287a3dd71f","split":"train"} +{"domain":"finance","full_message_tokens":112091,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":499,"sample_id":"sample:85f59178b7a0d5f8a2aa043d4262ff90415932e9198ee96b56baad2ce31c40e7","semantic_task_id":"3ca0741c5e070f2e5d17e0eefcbae2d97b5b2375779c33aefc0bfb282c0534c2","split":"train"} +{"domain":"finance","full_message_tokens":112036,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":500,"sample_id":"sample:185e13b6f08bcccbce30ef5faf241bd45b9d9ee8f361dfe5d0fc3f4c94b6eded","semantic_task_id":"3d423911fb30c3a75cbed34767d1a322ee414e5f23a2fafa7f1cbe878fcffa66","split":"train"} +{"domain":"finance","full_message_tokens":227682,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":501,"sample_id":"sample:5151aeb009fd280884c317dd861a32f06233ffb1c6dfaa2da1ab7b6dda55c204","semantic_task_id":"3d7d1da1b2a393bc3fa7a83e0018ff528febb6c7fa99e0474e8ab286910e7217","split":"train"} +{"domain":"finance","full_message_tokens":112190,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":502,"sample_id":"sample:d8b6204581cbf79cfeb55cee7da9205647855a227c2fb2539d09cc7556b44966","semantic_task_id":"3df10a60ada5d8239a3df9ed4628ebfc45e23174c31e6cc9fe9de066c4ac4f23","split":"train"} +{"domain":"finance","full_message_tokens":112190,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":503,"sample_id":"sample:a694e2695925bcbfd00e775afa4ba4a9d90cacd863b62d0e83a1b0cef479af3f","semantic_task_id":"3e6aef3f3698b71589696627872a74308a95189809820659d415dc5085e0b5b0","split":"train"} +{"domain":"finance","full_message_tokens":112075,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":504,"sample_id":"sample:c1e607c88220cf583437f95dd315382c1a5add32066776b2549eeb9ffc9b757b","semantic_task_id":"3f125ff788ca731c20a07c41f2218f23933e181d77d1162d6cba02aef7f0b6ed","split":"train"} +{"domain":"finance","full_message_tokens":112120,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":505,"sample_id":"sample:70c8e746925e690492ffadb23a7abb1a4779e676a691cccf35da43a11693322d","semantic_task_id":"3f621edfd78dbc4627b2fe3041b15b469da482dc2074a7e704e40ebc3cca86e0","split":"train"} +{"domain":"finance","full_message_tokens":227661,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":506,"sample_id":"sample:b57e345653dbfb63fec45430b285c02e8aa7e6256ecfc84d391f1bb1299d96c1","semantic_task_id":"450a0b57aa687a213778dcb7d7f1b6ac4258fe8b0c619864beb414789b396580","split":"train"} +{"domain":"finance","full_message_tokens":112036,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":507,"sample_id":"sample:0e39cd545e2f4831b42ee5b88694ce6de460a9fcbe8e6d95f76b445290db3cda","semantic_task_id":"478ddc2f7242a65e146f7ec0f136a59fdda3ca8a689552a695e87b60a3941285","split":"train"} +{"domain":"finance","full_message_tokens":112037,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":508,"sample_id":"sample:e1fc0892dd647e6ed3a00ee7dc7037f995cb21b25923d7fda3e4ffbef9e10430","semantic_task_id":"47a80fdfb859cbc39a702fa70aa8448b5f0f3ae7ef8160863df721470890ca17","split":"train"} +{"domain":"finance","full_message_tokens":115454,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":509,"sample_id":"sample:d069be250d494f3f82e30b3e307d721671404c011ac99622bb21f50c73578cd4","semantic_task_id":"47fa11328c17331f42bbc595b6858cb17c0b98b21cb35273909f6e9fd9b13ba8","split":"train"} +{"domain":"finance","full_message_tokens":112188,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":510,"sample_id":"sample:56f4a8857d65680d8cba492f13180096483eb07186f7cace9f9f170dc8966718","semantic_task_id":"49094b63f4ef555dec898a9c8eab87f45374d47e09557f8ffa4165802e1d11f7","split":"train"} +{"domain":"finance","full_message_tokens":115584,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":511,"sample_id":"sample:e98a690bc92bad5d559b40998e7b87df404751c754177f5da53d22ac15bff3c0","semantic_task_id":"4ad533768e892c9c256d99a1d1a6a242f8775ad14e5f743185692a34b7dd9697","split":"train"} +{"domain":"finance","full_message_tokens":115587,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":512,"sample_id":"sample:b4e1e6265640e30bf15ed7f6733448b8d474188e9dab153455adae7889d9083c","semantic_task_id":"4b76019bf6311a200a3bae5ee7860b5b55493d951d4dd8372447fc7264f8a0ff","split":"train"} +{"domain":"finance","full_message_tokens":230788,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":513,"sample_id":"sample:6085942c160d29602461f17d090e2736d9dc40b10f24344518463bc57149a4ea","semantic_task_id":"4dd02cb92a6144c2a5fe447ec23e09cc2d49c678215274ffb7f44751ea073fb3","split":"train"} +{"domain":"finance","full_message_tokens":112324,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":514,"sample_id":"sample:cb79d321a75c2987f2bba778a5897dd696851cc4d748a262de636ca338ef0954","semantic_task_id":"519162bc7286e3f2f4b996ad93ad2470f0a777be0c5d6f2de894196d90ee8e5d","split":"train"} +{"domain":"finance","full_message_tokens":230787,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":515,"sample_id":"sample:bcd9c58f8fd7002617ea15c5ae2bd6d6b1711318997c284a97d30015b1d82023","semantic_task_id":"52b776f87a0b509ac0053ee00502d94f8b8ff82c79b1e240723352206e4b0c0a","split":"train"} +{"domain":"finance","full_message_tokens":112325,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":516,"sample_id":"sample:b8ca77be96e0fb758f3f075b21ce4837b331474387f79a4517441d9375f351ce","semantic_task_id":"55124e04048c21ea2e8ba508f1ebdb478d73def3fbbff8991c79f66e65d5ce2b","split":"train"} +{"domain":"finance","full_message_tokens":115626,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":517,"sample_id":"sample:0e55fd6538fd1cdb1e4390e479a01430fc0ba8e0b30578f9df459d948d086891","semantic_task_id":"55965e7d1f16faffe35b7661c6bc8b28c10683e5749699648c9fc0204331d26e","split":"train"} +{"domain":"finance","full_message_tokens":227680,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":518,"sample_id":"sample:e21f2c4fde509e6599226eff1c378e6844372a77d9546625480edf859f4e6ed3","semantic_task_id":"57e07a1be2decbaa0b3e5299f2330af67999a70d477109ba485f5ad3b06e4541","split":"train"} +{"domain":"finance","full_message_tokens":112037,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":519,"sample_id":"sample:b60f4c007bad65e444bcae879f79a002a2b7de53fc748515a2e9c46975361d42","semantic_task_id":"5808ea5d02a5a0c9854fc89d5092f5cf1e542872c5f946ca60496a78e6c13d31","split":"train"} +{"domain":"finance","full_message_tokens":227670,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":520,"sample_id":"sample:da8ccaf8ec07a08c389a73718b054ed4dc6e43c933416fb37c1e7293c17dde73","semantic_task_id":"583d1dc9001b6f7855e0bb846c717c6393474312fda1fc801ea10ab5de92b7a7","split":"train"} +{"domain":"finance","full_message_tokens":227716,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":521,"sample_id":"sample:d955ee352d20530106527c99fe020734c21170b7ac6d3ebaf72118d72ad1e14f","semantic_task_id":"59a52f90675f1b3dca3e00281a00fd2a77e43dd57d81cc169148ec5df21ce9bc","split":"train"} +{"domain":"finance","full_message_tokens":115171,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":522,"sample_id":"sample:90622903fd1f30e18682d8714095e4b42feb23297b17ccdf0012ab3d3ac8a7f3","semantic_task_id":"5b106a73025babdcfc59cb7cabb379c79b64eef879204f8656a36af32b45fe17","split":"train"} +{"domain":"finance","full_message_tokens":115588,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":523,"sample_id":"sample:80b54eef8da83565b3db5ab346ab1b2f9b73e57d4b06fa46772d5a775ddac9e5","semantic_task_id":"5bedc14269908ace177577848d036c0e5e588f7eb54b2803b5c36182fc3f034d","split":"train"} +{"domain":"finance","full_message_tokens":115171,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":524,"sample_id":"sample:080d3879677568703c7aea3dcd6d2345b9d2e0d8c636e37dd3c4564ac511d6be","semantic_task_id":"5c7bc5b0003eda969a8398743ed3abf27fb008e8c063376a897d0b04af72434e","split":"train"} +{"domain":"finance","full_message_tokens":115181,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":525,"sample_id":"sample:518146c4e4ebd13980164a326b70fa18002d2c3484fead9711a180922012d4dc","semantic_task_id":"5f2f77d6c67e7f07d9c4b51ee45c235455587461232080e6094923243d12e690","split":"train"} +{"domain":"finance","full_message_tokens":230869,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":526,"sample_id":"sample:fbf9402d1c8d8bc1184250d143a85ce171777532ce8aed12ef0e77bc25b234b8","semantic_task_id":"5fa83c665c2721c204bdb221da223b7e53bc259d61faed8d70536dc888a46ea7","split":"train"} +{"domain":"finance","full_message_tokens":230787,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":527,"sample_id":"sample:9217aa5ec0c1fb2aa407311ba9ff1c7f481bcea2613fb745e55e892cbfafc52a","semantic_task_id":"5fb336036dcd008aa20a62c3bcac48f354aa02d790568eb9237170b9739c8ca7","split":"train"} +{"domain":"finance","full_message_tokens":115639,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":528,"sample_id":"sample:38b5cc1adbe7643a38a6aa3aef90d81e677bd755a48701e2beaba6f718896760","semantic_task_id":"606aceaf62102524a813d668e03ca9d1abf477870de77ec5f5e937cb7a6ffd81","split":"train"} +{"domain":"finance","full_message_tokens":115475,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":529,"sample_id":"sample:679e068a1f87650b5742700bb920c64e33829f4af773f6544e7367c7b7479653","semantic_task_id":"6254e59e8b4fed5405cb99419653fce8714a213fbafdf56a0990a818d23ccd06","split":"train"} +{"domain":"finance","full_message_tokens":227414,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":530,"sample_id":"sample:fd27de9c46aa38db870fca633e8add903e39a5731d6ca946dbe48311a1b8db0b","semantic_task_id":"62fcd6c751d7c0bc116fd2813d654f016ef17764691b8becfbf09a027af4ac8d","split":"train"} +{"domain":"finance","full_message_tokens":230827,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":531,"sample_id":"sample:ed3c6a10df6f12387409329c93886161f9fd26b186fa15c3301fee82b470ad6a","semantic_task_id":"6464f7e695b64150d281422a3f33ef16792c97a7b6c179529d75e6c0cd177884","split":"train"} +{"domain":"finance","full_message_tokens":115453,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":532,"sample_id":"sample:0f910488a335866fb432652142ef58ccbb6fbdcf2505dcd831f8d698187c2587","semantic_task_id":"647d9a2ce6a4382504dd4791af81089b94e745d4b9f1e6fa4db5519d417a80c0","split":"train"} +{"domain":"finance","full_message_tokens":115583,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":533,"sample_id":"sample:c9bfea16cf44005b6d9f875545a26fbbaedc240905712e57911fd53d3a966dd7","semantic_task_id":"6576b937f3d3603ec95b06e9c945a80ebc8bf06e54314fab73210aeb0f8080a5","split":"train"} +{"domain":"finance","full_message_tokens":227404,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":534,"sample_id":"sample:e4cde922f9ba06e837d959ddf2d0d499c685012852c10fb8f5ad59df818af4d5","semantic_task_id":"66ca4ab1cb6987862ac24a57768177ee82e6756dde5138cbbfa2a314b33fc126","split":"train"} +{"domain":"finance","full_message_tokens":227678,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":535,"sample_id":"sample:2578456181aa346ad012dcfd6365b6acf1289f984744dbc9f2d467575ed95741","semantic_task_id":"68855ecce6eef0bf8d4b96440d3157aa832bef87a3083c85ae86c0235124909d","split":"train"} +{"domain":"finance","full_message_tokens":230810,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":536,"sample_id":"sample:ce12ecfabc4d8403d2222f128130a92b09b6fcb5d76c875599f3df1f902cbba1","semantic_task_id":"69c96ea10b445807f521691540b96bb43c3068694d3d122ee709274088dfb0d8","split":"train"} +{"domain":"finance","full_message_tokens":112379,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":537,"sample_id":"sample:fbf8e2eac97034d10b474b9990f2e0141a775e59aa886b72a3dad1a1edaca976","semantic_task_id":"6c8d4ae94f3831f79622dd29eddf6fee020bdeb071b42f8028ee5fc50df9d281","split":"train"} +{"domain":"finance","full_message_tokens":108350,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":538,"sample_id":"sample:559ad4652e4cce6d523539c815f7288ffb10e66f8db25d75ca9211d60d7f77d9","semantic_task_id":"6d6fbe9dc242d861c6376a0ca5bb84c7af9a67736ca25a9dde174b9e9e069325","split":"train"} +{"domain":"finance","full_message_tokens":227393,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":539,"sample_id":"sample:eb7e97f0c5d8973525e9371419439f9203f4f5081e154de5a75b539eb61d982a","semantic_task_id":"6df015fd57eb291e00a199bc70f2d75be453e34aba649b8437a281f5614da71b","split":"train"} +{"domain":"finance","full_message_tokens":112325,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":540,"sample_id":"sample:70cb3c8fc132d6fb8fe60bcf198a0600518f6ee30ac78699ec703c58cd1b8f0c","semantic_task_id":"6ecb9a5052683a2a547e7d60acaf35bd806ab43ceb7abe75c13892c43578fedd","split":"train"} +{"domain":"finance","full_message_tokens":115339,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":541,"sample_id":"sample:d54f85d5f2a93349f3c5d9932f014ba36c84e0ec67512e9aa85a3497753e2d81","semantic_task_id":"700026f82a53a51ed35bcbd8ed19919d0af33f926e0c711bc92b94c2db300d90","split":"train"} +{"domain":"finance","full_message_tokens":108303,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":542,"sample_id":"sample:f3c52b636d614d7fdae6de7a5ad61e05bb7388329366a6841503cdf773f55f4b","semantic_task_id":"714d0b8958b12233b84a27aa973cac71626efee942b22dfbfbb7df34e4db5315","split":"train"} +{"domain":"finance","full_message_tokens":230820,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":543,"sample_id":"sample:f30a091b3a9bba2ed527f612aa680c81b067e0634a6fc348031e0f02a3051eae","semantic_task_id":"71564ab031bbb73318eaf1d112fb948c3b45d0a79a186affda17434d3fd42528","split":"train"} +{"domain":"finance","full_message_tokens":115589,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":544,"sample_id":"sample:ea35f7f885fe1fedd088aa2286b9f1641e90c4b4d383fb9118ef5779286059de","semantic_task_id":"71f9c9e8fabd721104d1829c8180279365546175c4e0cb1d6fde7dc4c4094e53","split":"train"} +{"domain":"finance","full_message_tokens":115181,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":545,"sample_id":"sample:0336ac3e11fa5779349d5982293555acd6bdb4406f43a4fab6f6722bf2e5bc8d","semantic_task_id":"721bb73f34c7988bd47801ae6d6b2326fdc98344611cad24a7f19419d6f0b850","split":"train"} +{"domain":"finance","full_message_tokens":112324,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":546,"sample_id":"sample:1ccf49d123a5be9396c50038f922174b46a44c77ab6acd7303210557cc76dfb0","semantic_task_id":"739a455c85c0ad96351dae2ae3f432a02be04f0deaf7b77da60aa1b48d588305","split":"train"} +{"domain":"finance","full_message_tokens":112236,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":547,"sample_id":"sample:bb9e0c90a564c894a4c68456bd18b07b46edcbf865349463a648f79ae9eb0d4d","semantic_task_id":"75ab211ac985db78139f6486c4e3e74fe789a87ce5430a7316a566ea5676622b","split":"train"} +{"domain":"finance","full_message_tokens":112085,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":548,"sample_id":"sample:b904c27badeadd08c7023ed81ba8b840470170c7c0ee27c0722decc2ab8a45bc","semantic_task_id":"76886375d1787ab648e1bfbfc10f5af69aab94b6f3f3f7a6da63a5bedca8517f","split":"train"} +{"domain":"finance","full_message_tokens":115588,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":549,"sample_id":"sample:6472ad1acf1b884b849cbd2183bfb5e9d8641536461b99ceb5398a4e3ad41301","semantic_task_id":"76f7c5addac16f1df832f7ec2d02ef634559b3e4ff879981bc6d05f27e7013bf","split":"train"} +{"domain":"finance","full_message_tokens":108303,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":550,"sample_id":"sample:a3ad0713f647eec9bf8c422731dac90e4c1a787046e3cf8800756146c9bf1538","semantic_task_id":"773fe613e01a577fec145acfbf2fe2a819d33fb37cc83ef5b5dd04157b584224","split":"train"} +{"domain":"finance","full_message_tokens":115586,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":551,"sample_id":"sample:36cbe62f8405540a5ef0c9cf0738373f851dd80234872227aa273d1773e07ea4","semantic_task_id":"77de73e0f2134abd78783d532979f0884598eb09f5760bfec24d3b79240071b5","split":"train"} +{"domain":"finance","full_message_tokens":227660,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":552,"sample_id":"sample:781f4b1fb2509b7e5f31c22c8c870be8fc9e0a65cdd9937e4fb260613a1b2d86","semantic_task_id":"799ebc9b6435789d4ef94a2fb1278ca256714a7f133ed5d3a9f9fe986dda6c0f","split":"train"} +{"domain":"finance","full_message_tokens":227394,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":553,"sample_id":"sample:62581e57684ca41f8c94c14816eb37bda83379058d94a31b6df554ab8f9d6404","semantic_task_id":"79a211750638ffb5e5986e8ce77e5f903f6afcecfb0d9082883c3add2b3d53fe","split":"train"} +{"domain":"finance","full_message_tokens":115182,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":554,"sample_id":"sample:d2c173aeefc1b3a8ad716fb62509739e9290832907f340e8c5f6b2d8e9f4dc1d","semantic_task_id":"7c6253643da0d108e01b95b1233b51234b36e680ae3a9f689297c41499e96892","split":"train"} +{"domain":"finance","full_message_tokens":115641,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":555,"sample_id":"sample:ff6f758f1ab9ff289ff64a18d84e062009ada64c1247e7d38a0da4aa0df6f9c2","semantic_task_id":"8228ecf10769c92d07891246a43994d4ea3992f4f14c1ea4eeea0f275c271524","split":"train"} +{"domain":"finance","full_message_tokens":115453,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":556,"sample_id":"sample:88afcdcadb9f9cc4fbd9a30482024c0c903188fea91bc89549ed1f302fec082a","semantic_task_id":"82460efeb58e1bbf3e053c723ae19ffc1d5daad7accb2e382d25b9880e48f1f8","split":"train"} +{"domain":"finance","full_message_tokens":115181,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":557,"sample_id":"sample:a236fc148b1172b90b007b1e4bba5cd628163a5dcbf9533a70bcda2866f387be","semantic_task_id":"84a84ce8b31f11abc50ec2847bae6611cacf86e755529b475f69b90724d6a5d4","split":"train"} +{"domain":"finance","full_message_tokens":230798,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":558,"sample_id":"sample:5963777fb191d19a28a8ca4184c9473ff980a6a313ea4e059baeecfc5dc4ee4c","semantic_task_id":"85123d534c96cc8419687dc441fd98ced7465a90941d4480be84e3f7d94584ef","split":"train"} +{"domain":"finance","full_message_tokens":115639,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":559,"sample_id":"sample:b786a7114169d8690f9fb586e14113f9dc8d22f3fa3fe283d562d3051e5e1295","semantic_task_id":"88dbc5d7006a0555d4703bbf2888d0b9065be0937bd61f38969904ee79fa9d8c","split":"train"} +{"domain":"finance","full_message_tokens":115186,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":560,"sample_id":"sample:e115bd7488a74fc0d523f30d6a31eb9c5c8cdfc3f1b31b9ca42656002fd4b5f2","semantic_task_id":"893a6da9a2531af1dcc67db91eb0800fbef043fa58e2b799368086568b7aec85","split":"train"} +{"domain":"finance","full_message_tokens":230819,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":561,"sample_id":"sample:ee7ee7bce61fa4ab643bc323e4bb6aa3617ed89e6484e673a1e97f13eacca4e6","semantic_task_id":"8a317f35a51d54314092d9e06753a8f2a7e07dc180a47cad3d4b385f646a81e1","split":"train"} +{"domain":"finance","full_message_tokens":115641,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":562,"sample_id":"sample:56e459b10ead90ac4af023ed4b76dac491ceddec690d6e645e974e223bab3362","semantic_task_id":"8b2fcac46d54aaab1de14fac3323772dd3e62caec984519d02b499db323d60e2","split":"train"} +{"domain":"finance","full_message_tokens":115583,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":563,"sample_id":"sample:9c470f7c990040c34bb09136d39624aee7d703c4e976f19de4f290d4285fc3cb","semantic_task_id":"8b78e09fecc0bc7b420abfb6361c9784966140c180d2278dee0e1897cda80e3b","split":"train"} +{"domain":"finance","full_message_tokens":227393,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":564,"sample_id":"sample:4dc96675869673102de3718bbd0972346cc1bc6d44ebd8956e48129f80a829da","semantic_task_id":"8f0865e7a09448810b569b0ccc43524e3bee447edb8ed8ed30163e5ff6f6ad0f","split":"train"} +{"domain":"finance","full_message_tokens":230787,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":565,"sample_id":"sample:06c9aff89bfafc103dbfaaa093c6120d14e799a9cdcb5de2fcdbb1b4373dc326","semantic_task_id":"8fc7270fd138a17e258333dfd479368c0b3374fbcedcbb15288edbc2d49c2cd1","split":"train"} +{"domain":"finance","full_message_tokens":112081,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":566,"sample_id":"sample:c5aa3ac14554bc5531f6308475f11dfc8c6d078d9e92d4e4cda0b75c1f785927","semantic_task_id":"92d412ce40bcf38ce2281f4fe153462c194eec10c8d0e1ac6dbedc09bcab29f4","split":"train"} +{"domain":"finance","full_message_tokens":227414,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":567,"sample_id":"sample:871a3b9190b7279c6b8c2a2e6d2852bc6e18c172453dda5bba5ea8ceb7eb34ca","semantic_task_id":"931daf48c715b58356331615c62911f7b659f341731787063a211de2af6fe6f8","split":"train"} +{"domain":"finance","full_message_tokens":115338,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":568,"sample_id":"sample:a24a787563d94785ad6239919f28c27314bc33aa33521c7327a517ca940726dd","semantic_task_id":"93de5d44a82919de0f585442174c26a1eed167b4241a1c177a60e04502232146","split":"train"} +{"domain":"finance","full_message_tokens":115473,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":569,"sample_id":"sample:bf788b465bcb162c7df121900eb0decd9e04894010c510bad67b7c819186266b","semantic_task_id":"94020b2ff7c6175bcafa9fa7818ba0229c34be3dddc1fae7eb6825136804c17e","split":"train"} +{"domain":"finance","full_message_tokens":230819,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":570,"sample_id":"sample:80b7b43be3c919e7be234a970bacffc4d10a1b3889c79adbfab6fbf7c18e6136","semantic_task_id":"94190550896beaf562bf7e7048473799ccaaf1a69e9bd027361b4c7f2310114f","split":"train"} +{"domain":"finance","full_message_tokens":227414,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":571,"sample_id":"sample:fb0fc6bc8361083cd86941708b2f9ac37b9c3b9fe350c30770acc449a667470d","semantic_task_id":"9544b690010785721e9df222668c0129936a69906c9455b037da91c064815a6e","split":"train"} +{"domain":"finance","full_message_tokens":227659,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":572,"sample_id":"sample:dfe077641660b02f1bca508da177000134daf96ce216709246664c7c964af3cc","semantic_task_id":"95a3aa9ead06d115352b4577bd5730ac3c63f03c19d24ec7245393a1ecb0e452","split":"train"} +{"domain":"finance","full_message_tokens":112191,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":573,"sample_id":"sample:3b0b8971386dd9c08c59b891690fb1f11088dfe2da850c35e1779a61160f4191","semantic_task_id":"9686f62f84a7125083ef72a02fe3f4cb76aff38f3353ee55889d77eefb54e5c0","split":"train"} +{"domain":"finance","full_message_tokens":227681,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":574,"sample_id":"sample:492bb0a7bfd31562269db9a5fccb1e298c837628f174de9e179276bfec91fee6","semantic_task_id":"999d1a8a1eed502f7a21474b0a4155106571aa404a4368ff391bc28e4fbef656","split":"train"} +{"domain":"finance","full_message_tokens":227393,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":575,"sample_id":"sample:ffbedfaeb3ec96a7976a87c1a7e2f620befddfb833d9a1ac076ae08b067f7f10","semantic_task_id":"9b50e69a2ccb58a5b754660981ad5db4dd6699cf25d859cba38f89591d82cbc0","split":"train"} +{"domain":"finance","full_message_tokens":115573,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":576,"sample_id":"sample:8b57711723269e7d1e1a0d480bbb2e57e4a4a12b0b658766641c2cfbda07d169","semantic_task_id":"9c1c5ad622c125b33289e9ded05b0f9322cc6c5587fea475bc1f540770cc1f14","split":"train"} +{"domain":"finance","full_message_tokens":227660,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":577,"sample_id":"sample:abd2051092aaae8693a2b18cd0b88a853ec31b2f5d5a13f4c3eedee6d784eff3","semantic_task_id":"9e7c8af3871d0826452fb2e5936a29c87966c8053f1c478055621a609e36a05f","split":"train"} +{"domain":"finance","full_message_tokens":108342,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":578,"sample_id":"sample:94c68d419c2ef0e43e91199e39732708ff11fc2f3a028c16f17e7901d87946f6","semantic_task_id":"9ec2738d70a1d526729fbb4217b405ed6ed8360daa72e736de87bcea54a87cb8","split":"train"} +{"domain":"finance","full_message_tokens":108304,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":579,"sample_id":"sample:b222324435b553a64f17a38bfa6d4a5a92196f33e07807ab50503402a6e78519","semantic_task_id":"9f7d35431c8e463acdc5cc8dd8785c74e3835922a311936e58ce6d963cfa7634","split":"train"} +{"domain":"finance","full_message_tokens":112321,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":580,"sample_id":"sample:5635364e05f4b8bb3ff8b594a5351a10866576103873da07faa4a30e5eb033b2","semantic_task_id":"a0efd3764b678c6b3d4831db989d9336d24b79ca0ae2bb4463a03ad6b9cccc59","split":"train"} +{"domain":"finance","full_message_tokens":230865,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":581,"sample_id":"sample:e42fcb78fd097c567b7d4a0b775897233959c1aa694e10be38110a0a9508895c","semantic_task_id":"a153d2c838f5209bae954bc9cac84dd5fd930c9d5af04c4a8e3e52eb8f8b4304","split":"train"} +{"domain":"finance","full_message_tokens":112035,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":582,"sample_id":"sample:43f327c6d762a9d44f45eacf425465213d5bc94885686366a4ca29ccf0defc54","semantic_task_id":"a391c2974bc205b47bba548eb2dc7d96d8f717cf911d3e2bc804941f3b783805","split":"train"} +{"domain":"finance","full_message_tokens":230820,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":583,"sample_id":"sample:63176193639397ff35fea33d247455b16cf42211172ce9212b1835c6bb37b4c6","semantic_task_id":"a42334f1c1ad28e754cd5eeddbc26f39bd3abc9e78237d9e7e5e6c8b34290ea6","split":"train"} +{"domain":"finance","full_message_tokens":115626,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":584,"sample_id":"sample:63f8493977759d076a24610a25ff5e11149087dce27bbf6cfac43c5345cc43f1","semantic_task_id":"a65edfa0ad32a068847188517ca2863caa663792f65b57a08da84615f01413dc","split":"train"} +{"domain":"finance","full_message_tokens":115582,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":585,"sample_id":"sample:7364f318e5d53f2052cb1327898be01510c9a0356eb57b930634fb89bb41f8ed","semantic_task_id":"a759e6d19e3821761037f20387055835a68dc5338ea0cd9d4a3f6a03b0991c64","split":"train"} +{"domain":"finance","full_message_tokens":115451,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":586,"sample_id":"sample:287aea5bd88fe7a401ed99841b6633a7578bff453c302e66fde1d94f6115cdaa","semantic_task_id":"a9830eadeca05acffffafd85255c34d1d7ea9244551a896420ad4a52ed51b18e","split":"train"} +{"domain":"finance","full_message_tokens":227696,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":587,"sample_id":"sample:06a8dffd2cb97e2156df82c1fe6df4161b448882d10a27c11f57bf7ff30c377e","semantic_task_id":"a9a929dfb3fc5830b657c196f75c1c738902c983a031710923b75c22038a7e31","split":"train"} +{"domain":"finance","full_message_tokens":112364,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":588,"sample_id":"sample:0c4c8426a017301e8702ee6a84f14a9840522e343f5bd90f8459fba3f7783957","semantic_task_id":"ab1dcd6ef62d9ed808a84bccd0d2553fe553f3cb53cd909f91963de816e37b1e","split":"train"} +{"domain":"finance","full_message_tokens":115641,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":589,"sample_id":"sample:44068dcd31a87d5547ace79f1f40cc472949a4bd250843032b8912f65b5544c4","semantic_task_id":"ac874d9cfc999f52967b69832b0b40dd9c0235eaf04f9cf8c4296906cfc8cfc3","split":"train"} +{"domain":"finance","full_message_tokens":115453,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":590,"sample_id":"sample:4e14cd27f2e02c78494474f9becf83e5eacf9d842fab8faa5c688aec20c0e26e","semantic_task_id":"adf9d4fe032c30ac0e694105a355d96e09e4858ad549050b7cbd1aeafdbc0049","split":"train"} +{"domain":"finance","full_message_tokens":112189,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":591,"sample_id":"sample:864635a93ccbf5562bdf4b7fc2fb192bf86d5a2d05877e623d2b5f0cb3431bc6","semantic_task_id":"afa761207379a07e9d5bb49ecd9d93717cb73dcf4a96f46f0ec8b5e7a92eb4d5","split":"train"} +{"domain":"finance","full_message_tokens":115637,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":592,"sample_id":"sample:fbffdbc7f96bf4b7ecf45b661e25991cd868dd02c4726bec66e1ab769bbef2b3","semantic_task_id":"b0815b0698c58bb7c3a6fa78c5aa35d82283d03accb3785ba6a93f44299fe6b7","split":"train"} +{"domain":"finance","full_message_tokens":115589,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":593,"sample_id":"sample:2c82957ad72da31b45fd1bff2009f8bd4a845b09347d53ff4629237d7587bce7","semantic_task_id":"b4811ca8aaef7cee4bb4fc229ec87ad4155c8bb89df129c153c21cc97b8f78be","split":"train"} +{"domain":"finance","full_message_tokens":227682,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":594,"sample_id":"sample:00ab31f9926ebe54c89b97bffaf602d25ba47b32a5b6bb98804c70c44d2f7eed","semantic_task_id":"b73b33241defa15d13299ed5a12e4feebbdf1858ccab66fe3d005443b46fa733","split":"train"} +{"domain":"finance","full_message_tokens":115474,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":595,"sample_id":"sample:9794fce76d494cb8592f9df248063af5a23b02248d68071eb7a92c8ff461cfba","semantic_task_id":"b8d2b41d93b616a2b19ff88d4bf13166b5e56132625cf2f503884c821eabaee4","split":"train"} +{"domain":"finance","full_message_tokens":115339,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":596,"sample_id":"sample:905ecaa344c2d5e825366bc079d45d4b3b5fc3b8a6d4854c0e4a6f694efd5557","semantic_task_id":"ba1172a0c19144a7d633f1ed22a5dee68a840ffe73bb5536d154df4c21bcbbad","split":"train"} +{"domain":"finance","full_message_tokens":115583,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":597,"sample_id":"sample:7bb98eed0842ac4a01e2096513e58937baaba9e5b3da50bcedcf03f186bb1b34","semantic_task_id":"bc37b9a5f2e9d968e8f3a9dd14c5c29286ef13da33e966c4d643fee49f526091","split":"train"} +{"domain":"finance","full_message_tokens":115187,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":598,"sample_id":"sample:91a5c96dbd15f6f27d4a5d05fec57a028f71448c81268cdf60dcf97913fdbb52","semantic_task_id":"bf3338e7fce618099e08d8b05a3976e090cd0557ac40a2008645fff4d80ef0ce","split":"train"} +{"domain":"finance","full_message_tokens":115623,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":599,"sample_id":"sample:f030d3c59cd88fc24abbe69c97553c0afe40ae76a94206947a44dac9ed35009f","semantic_task_id":"c0afbb6142282719c9ce5a992fa794e4f381b0f899e567fe6a12cd9e7f9b356e","split":"train"} +{"domain":"finance","full_message_tokens":112324,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":600,"sample_id":"sample:f385f9581fe8700c5e0518f820215354cb42f00de4060b67732fce754a9c7a8f","semantic_task_id":"c1fb230fefcb1595d19338cfdb69f0fea113bb2ba70303c789c2e4370dd64adb","split":"train"} +{"domain":"finance","full_message_tokens":227393,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":601,"sample_id":"sample:4ab2120dca6b57fb5dc6a771efdf40bdac77380bad6974d69b060cc12f99dddc","semantic_task_id":"c200819f031615b96fcad591b74098b825eedce0257c2ca19fa4ae3799e4b2a2","split":"train"} +{"domain":"finance","full_message_tokens":112326,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":602,"sample_id":"sample:236ee0ac3f77a41d1c88f9fe9d1af6b04b320248c61a554e02776856c805b224","semantic_task_id":"c3961cfa27d4bf2be27ff65bc6a5b16e390c691dca1d45000052de5df0ab3179","split":"train"} +{"domain":"finance","full_message_tokens":108305,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":603,"sample_id":"sample:60264e60b8f2dcea505cd33453d1eb7e2aa0c027eb877cc4510d9d7d50141d20","semantic_task_id":"c442af78d9447f9685d9bd0bc1f22b892726a57664a24e94c89a05f42ec7fc94","split":"train"} +{"domain":"finance","full_message_tokens":115453,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":604,"sample_id":"sample:32bda15ec977511870721668b9439cd3d127d4a33e649b43b25a0bfc3383128d","semantic_task_id":"c470f3547c328914f5684d18befda9ae7333a081cb4f6e2f12fdf80838eb32b1","split":"train"} +{"domain":"finance","full_message_tokens":230830,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":605,"sample_id":"sample:ad4f5ea3ddef8160fcfaa9a3a5902e6180bdb97b4281fd491da1b87fb541a032","semantic_task_id":"c53d297a1f18b0c8804cf1d63208c1bc06bb0d9505be066ce07f34dbc965d156","split":"train"} +{"domain":"finance","full_message_tokens":227415,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":606,"sample_id":"sample:0e51d5fd4de4fd54a98171fce327114ac1cf693232b10bae88ed48d364d005b8","semantic_task_id":"c95479d3aa5670d907494fc2d6a3a7f4060a6d24dbe75c16adf749ab86b1d5d3","split":"train"} +{"domain":"finance","full_message_tokens":230823,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":607,"sample_id":"sample:d6d3185b66fb4cc375fcbffa5961cb9dea6ec5a941d6ec91347640ec474be0c2","semantic_task_id":"cb567274b19ce6765fa2ca99a799ce1dc4b6f4a79f9af7f25316f611e23fb049","split":"train"} +{"domain":"finance","full_message_tokens":227414,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":608,"sample_id":"sample:978a18d132a6a42a14cef979bc625705b030020c055a8edeec0c68ac6d1da1fe","semantic_task_id":"cb6e2bc3019e856d91b46eee94cf434829a86e1459c035bbceb08c1ac0bcb462","split":"train"} +{"domain":"finance","full_message_tokens":115637,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":609,"sample_id":"sample:d8859d609432d2ab76d5be039166331371ac63f5f651033528126281da605316","semantic_task_id":"ce774a60a74cb2e1498c98152c6e710b81f92a73c7474109afb1c7ce80f09b6b","split":"train"} +{"domain":"finance","full_message_tokens":112408,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":610,"sample_id":"sample:3d58fc24766993108dadd1b6df8a76f0e4b817490cd2157e3ab5d09b104216eb","semantic_task_id":"d0280710b1c612c361c588d631561f41c0f915c0334bd45a3eb024f376f49b4c","split":"train"} +{"domain":"finance","full_message_tokens":230789,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":611,"sample_id":"sample:cd99d7e614c673f5ab9c2b6bf2c6d19962e57b1f5fdada0b94b4a24773ba8002","semantic_task_id":"d118ecab3aef7574c613fa4a571192cd4db108cf17b6f7cbeffad4b241097466","split":"train"} +{"domain":"finance","full_message_tokens":227404,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":612,"sample_id":"sample:ff4bf0a870d1ff950f80b63c557bb8c0bbde0028e0cb649291d9016595649b00","semantic_task_id":"d18d7975348f0afc11249924b69d2406864bbd7af1ac0b96ab017085b90b7c09","split":"train"} +{"domain":"finance","full_message_tokens":115636,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":613,"sample_id":"sample:e785cba5e5885edeeeeb6002062db410ce4fba6d1dcfdc2da115e7693a893da7","semantic_task_id":"d3433ced20d0f3daec4ba752d66ac5001f851adc8e8164f582782e1263924653","split":"train"} +{"domain":"finance","full_message_tokens":112323,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":614,"sample_id":"sample:fc193afbe090cf5fdd65458e7f0b4d0837bfeb331a79a6b0a893d4868b82b186","semantic_task_id":"d55c4f32b33bb727dcfa594689a1c2b4d1e087c9408701efd293686a5595e321","split":"train"} +{"domain":"finance","full_message_tokens":115474,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":615,"sample_id":"sample:d4860c51f93558b3f092414b8bb89eb84d808aeb20c8a26aec8c113a0ba1765d","semantic_task_id":"d695ef74c686d2aec0cb54c58a25da6d3fe8d7e01d93927a088eb438a3d907bf","split":"train"} +{"domain":"finance","full_message_tokens":112373,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":616,"sample_id":"sample:569106e9f4727ebdb7ab78b39e7cf28a643050477b054f85223e6f70bd3f63cb","semantic_task_id":"d6b034deee937435f8c2537b4249e637cf96a88c4c24bc02173c92014503f797","split":"train"} +{"domain":"finance","full_message_tokens":108352,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":617,"sample_id":"sample:0412bbf39ce68d345ac5667dae275a5d236aa97711ae8efd0760af72af6a5f48","semantic_task_id":"dd1f3159e2dc9c0341cbc6d092ab71799175cb8ab036f4c0bab112e027cbad7e","split":"train"} +{"domain":"finance","full_message_tokens":227393,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":618,"sample_id":"sample:77c3d1c4ac66692d807f00e50a6e2fb68da3e9ff61af703a8ce6efebaa82db4c","semantic_task_id":"deaebf86143fb550e710c4539a9c8231960c9689795030594ae6465764d53210","split":"train"} +{"domain":"finance","full_message_tokens":115641,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":619,"sample_id":"sample:bbfbaef7e4c1ff6a76afbbc59069cdf55140950882c6210c30c3fd23adf3327b","semantic_task_id":"df0caf5546795b59946189469bf10f7e9e8ee69c7e0433f9e49c133ec32c6a35","split":"train"} +{"domain":"finance","full_message_tokens":115171,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":620,"sample_id":"sample:b05b36d96011cf1d2adbc32c1a8ed0652dab67fbe8811b311fa04ba91ccd8b7c","semantic_task_id":"e0fe903ccf4625e8d9ce219ef5055b155b768cfebfaaf879f5f5261533e4c051","split":"train"} +{"domain":"finance","full_message_tokens":227659,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":621,"sample_id":"sample:88d7f63cf674c724038522cc5dcea4f46471958dcc621632a26f1bddc522f4e3","semantic_task_id":"e35d44e3f43593db576059a4a4ec12eb5e32c83b73da320bb9e5d03ba2364116","split":"train"} +{"domain":"finance","full_message_tokens":115340,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":622,"sample_id":"sample:4223c802f4475f5fe45d5bd9b31cbc11a4589dfcd8129a974ffb8db3f074b229","semantic_task_id":"e35fe4049ec5125bdc3294cad5945c1301e0c61e32912c90d0166cad85adad51","split":"train"} +{"domain":"finance","full_message_tokens":227661,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":623,"sample_id":"sample:54f609a0272f47b16ea1ce780f2a89e028eb39ac1337e3dd7dd7240444203f1d","semantic_task_id":"e368df259ceac4bdab4117d73d6e5ffd70f0c9583eca60cc00624d74ecc89cae","split":"train"} +{"domain":"finance","full_message_tokens":112035,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":624,"sample_id":"sample:dca213e88386542bd59256714935eaf285485407af4bc627d4006a01884f2097","semantic_task_id":"e4e546f908d62d2720400536ecc46eff1ca4b88f31892308c726c8f1c62869dd","split":"train"} +{"domain":"finance","full_message_tokens":112189,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":625,"sample_id":"sample:2a26004d2d7084d59fcc3ea9436bc32d94064deadda8af80e40212ca28f3788d","semantic_task_id":"e5d42aa4b158c2904a7d91de7d1f7aa8866d3b5d7d6a3a06b5cb06fcfc34ad81","split":"train"} +{"domain":"finance","full_message_tokens":227395,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":626,"sample_id":"sample:f16ea75b8b24d9fdc7f6e5d5663615731bdddde5ff549334a305eb6091590072","semantic_task_id":"e75faf8abc80763c6582908ed91d3da646500132bda1541f36c6d1fb4ceb9262","split":"train"} +{"domain":"finance","full_message_tokens":112036,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":627,"sample_id":"sample:6f6d7e96defd6a0c5343ef522f2ebc1221839cb5f23330282c65bf2da6efa982","semantic_task_id":"e84e509d2f48886eabc03c467e246a8e2fb7d5b5e4aa1cb6bd0c02040ea1c65a","split":"train"} +{"domain":"finance","full_message_tokens":115475,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":628,"sample_id":"sample:f0b707c50ded90a6959dd6ec7603dee173210075efb7350c0bad6866c8b86a63","semantic_task_id":"eba3f70a0a7e02a2eb8743744981af25336e1bcff92c4795e6b6e0e177cfac1a","split":"train"} +{"domain":"finance","full_message_tokens":112237,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":629,"sample_id":"sample:01e990d10db3709ce6d1af8ed950decc34ed12e59bdba9a081d31c0bb7c52842","semantic_task_id":"ec3432639da7a03c22757e86c149ef51a8f25a226bf395afd0dd35cfc42605f6","split":"train"} +{"domain":"finance","full_message_tokens":115641,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":630,"sample_id":"sample:c53d74560005c734d734bbe6c0f992dc2e8a55bbf7139a87ae48c0a19d1e3de1","semantic_task_id":"ec5704ff4b0bb34c009405f69ec366cedc701fc4c6edf09167c4138ae0ed49c3","split":"train"} +{"domain":"finance","full_message_tokens":227415,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":631,"sample_id":"sample:884e2fd99eb3bb1f6c408095173bad552abefe0e89b29abe805ff8d4b3437510","semantic_task_id":"ec8e393fdfdec684c58b20c306cb432e1b9283f8e4b277531cc29a571c249536","split":"train"} +{"domain":"finance","full_message_tokens":115185,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":632,"sample_id":"sample:c0264e0dc0fa403828e31f91fa8795295e35de7a4f268802c404a38f51f4e7bf","semantic_task_id":"ede44d7bbdec8f70cbad7add40e595b95dd709dc6bba16b1213bc48ee15aabf4","split":"train"} +{"domain":"finance","full_message_tokens":115452,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":633,"sample_id":"sample:c45ad67455b5fc17313e9cd632794d032a6682f606488957f690bb86ca1b2c57","semantic_task_id":"ee11d818908a8d32139ebcfaf9eb2950b5ed1483326a21003231a36b834bb98f","split":"train"} +{"domain":"finance","full_message_tokens":115340,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":634,"sample_id":"sample:d0b6c40a8911cfc84c54c7d51722556cec39196a5c59a2ae7bd98d1a77ba873d","semantic_task_id":"ee500d92376442246030088e1f372a311bc3917615d93c4224df33073cef27cc","split":"train"} +{"domain":"finance","full_message_tokens":230808,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":635,"sample_id":"sample:53000f8d9cd469c08ac4a446b1ae08fc035ca270b1cb468cdae4e2e25c9587c2","semantic_task_id":"efb7a710b13ca6d682616d700bafc04fcdd94e81358ecdd45fa3eda9815522f5","split":"train"} +{"domain":"finance","full_message_tokens":115185,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":636,"sample_id":"sample:bdccad69e0ee2c7bd9f49e71e3b718cfeed3ce96c1d75674de087d42223a06e0","semantic_task_id":"efebc9aa08d3a2645efc96b342122aaa0a7d22ac79d2dfc9fd2115f94425ae06","split":"train"} +{"domain":"finance","full_message_tokens":227660,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":637,"sample_id":"sample:9d29e6903d5c01630cab87a3fdb8c28c3c72bb15fe3c864279136961322266ac","semantic_task_id":"f2c608197673375e80765b5a1804437d6ad3e6f78a5802569c32f10105e75a19","split":"train"} +{"domain":"finance","full_message_tokens":227682,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":638,"sample_id":"sample:e98f9b2857c586894580ea0a47f55459610b1d1108375590c80bc74570b4727e","semantic_task_id":"f2ca7801bc546984bb333bf3ae91f5477ebf243b80eb12c314d960147a8ea6c2","split":"train"} +{"domain":"finance","full_message_tokens":227414,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":639,"sample_id":"sample:1b910a14d26050c85efa44470babccd5ff555e27e72ae95f0d3d1b048e592355","semantic_task_id":"f2d7c2da9fe813235e1d15dc1256a956acd00bd272e4adf246d11dd695a42a1a","split":"train"} +{"domain":"finance","full_message_tokens":115182,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":640,"sample_id":"sample:229d827d0bb924bf725f9a770684aed7521e9b427d97c0c26d30ac531d5e0236","semantic_task_id":"f3ad9c13c97f9811e505b78325dca5232509f89b5e370473bffdf916744d7a8f","split":"train"} +{"domain":"finance","full_message_tokens":227414,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":641,"sample_id":"sample:096a779d41d2dba0c9c2b5b3f636a26ee1a76f337f44ee7fecddfd73371843e1","semantic_task_id":"f4c5b376c4e1e6aacd765f126c81a37d08e5cf7ec9caaf2d00b1e0c62aeb2793","split":"train"} +{"domain":"finance","full_message_tokens":115474,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":642,"sample_id":"sample:8148d9f4e5dc164e2cf3289a2bbf01c0c5ad4723f37be40685b19125a038be05","semantic_task_id":"f57cc621e5fad66e0d90020fefd39d5a99d1ce4fd106641cfbd5b07bc3896b51","split":"train"} +{"domain":"finance","full_message_tokens":108358,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":643,"sample_id":"sample:adc868d02df37862c39af27280332f7165c19fac62931ba584d5dd39686130b2","semantic_task_id":"f653c6e129054c6d712be664d2a70a03ed978faedc7a907bd0dfc81a095df735","split":"train"} +{"domain":"finance","full_message_tokens":108303,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":644,"sample_id":"sample:add7ac349ecf6e7223b5ea0134c4df3f05bc57f32beeae604b46026951f966f8","semantic_task_id":"f93709a3331324915f2c9f79797b6fcfbd94575a1129b7b6501c3326146ef69a","split":"train"} +{"domain":"finance","full_message_tokens":112038,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":645,"sample_id":"sample:1904b185f9ea38a93ab408adeed89660c13940b4d6ff26fe42f71721d1a8efcf","semantic_task_id":"fcdd25b199f5bea284b396396b3c2d824917d935b7374a95f6d77fe4c26a73da","split":"train"} +{"domain":"finance","full_message_tokens":115473,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":646,"sample_id":"sample:f3b66251668b7b9770f7af10339f055e38773152b969aed50cb1fffd06b5f9e4","semantic_task_id":"fe052e6550d6737d53f69f4502a6525a47f8c734916bf02025957c2bae0ae16a","split":"train"} +{"domain":"finance","full_message_tokens":227681,"group_id":"0000050863","output_file":"finance_train.jsonl","row_index":647,"sample_id":"sample:57885ad87f3ea47a5f6bdd134c22df4ea1b8e4148021b83bd64da8fb711ff3d7","semantic_task_id":"fea537fe2367aeb6804e4c480bf1add726ec5e4a745241bc3be9eabff14da06c","split":"train"} +{"domain":"finance","full_message_tokens":80673,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":648,"sample_id":"sample:9efd8be50a1323fb6a2b6fb0e6c4d1eaf926d3366e9af91290f66e37de642210","semantic_task_id":"0098304f8f1102acd2987d43c6f8a8ccd095f0c01477f09ab85ecdc6d4f1f57a","split":"train"} +{"domain":"finance","full_message_tokens":67061,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":649,"sample_id":"sample:940758d6c2f291e0b9d0cdb8d0410338f4d61c4938738049d18f1610b63f951e","semantic_task_id":"0361b34371868c956689c47f270e76492f3032bffc49120df83ce78f42279324","split":"train"} +{"domain":"finance","full_message_tokens":214460,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":650,"sample_id":"sample:0d45e2e8050aa336ec44a8faf5b55bf401b6c32228fe30c431320fe090f74e0f","semantic_task_id":"03a439f073a82cf47051aea4fceebb18c345b89d82dbaa22d043278fc0539faf","split":"train"} +{"domain":"finance","full_message_tokens":73655,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":651,"sample_id":"sample:62c30e192733d3bb40f70ec8cc3d1f47632a31be950d5547e5bed0e7a7b47fa3","semantic_task_id":"03f0c1f06b2df437addc8be9277825103cc091b498fc9578707e83728a995b01","split":"train"} +{"domain":"finance","full_message_tokens":205428,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":652,"sample_id":"sample:7eb846bd9ac62cd235763ad786a74f1fca842442a8ee097d43db40ff304ed140","semantic_task_id":"043a23d8b0b2f423453f18c160ed80abdb967328cb6cffe68d5d205bc8ac0b17","split":"train"} +{"domain":"finance","full_message_tokens":205451,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":653,"sample_id":"sample:85bd1de3a6ca9cb561ff77da9afd511cf8faba0269aa02089b50f741b5d78405","semantic_task_id":"0463c05c875ab107f3b8fda1276c45880fb1c07a5de5f00cbf54d38be6c8ad32","split":"train"} +{"domain":"finance","full_message_tokens":73654,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":654,"sample_id":"sample:1508a3cbeb10b0562abc8c6aae96c03493c65f39a5071e9980de92d2934b64c1","semantic_task_id":"05ba7bbb9636ea4d810be52fd9d3da4dd44dcac1eaccb20498ed2786c86ef593","split":"train"} +{"domain":"finance","full_message_tokens":74963,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":655,"sample_id":"sample:c6b3056ef3cfb10a2f103a655c43e387a737635a450bea55607a5fb5c4a6c327","semantic_task_id":"064c38602bc2873f1a4f1d17e8fb5d4a288578bf7d951995b88b941c4e0ae86d","split":"train"} +{"domain":"finance","full_message_tokens":74962,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":656,"sample_id":"sample:00bcca6013efcb650dd0dd72c70c237a4989a39c71e1a2482ce7e78e11af4d50","semantic_task_id":"07deaecf6173916f53afb797a5797497916281fcc1947a0e8b179ffd53538801","split":"train"} +{"domain":"finance","full_message_tokens":219454,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":657,"sample_id":"sample:7d3747945405260c9fd833fd314a84940d77e1cfa8f87a8380b39bfc2108e860","semantic_task_id":"08c2fbdf0de1f19c523211eefefaa2d4eddda64579ee6ef52d6772ee66a74164","split":"train"} +{"domain":"finance","full_message_tokens":73658,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":658,"sample_id":"sample:c5f916de79d06b62ed1ddcc740f82c8f5e0f29af624625f8e8eead6b8fd597a5","semantic_task_id":"08c494a9afdb6859f408524de2ce0b73b88b013ac55a28d9962f6c8a8fe58847","split":"train"} +{"domain":"finance","full_message_tokens":73644,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":659,"sample_id":"sample:583d472f19b3270e5a84c6c5eba8b0c8093ea4ce23769a7f830315c5f61af58d","semantic_task_id":"08da9c5e9069341446828a8f1bff8ec655bb99cf863520800b1ad0da537c43a2","split":"train"} +{"domain":"finance","full_message_tokens":69995,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":660,"sample_id":"sample:ffbff551838b455937804f29f5540dc7a55ab430cdd9590a25f10cac40274f4b","semantic_task_id":"09af1f03a5888b694ccfda46c4dfcd9b79b1999ed64b3173eb5169ed8dd6ae16","split":"train"} +{"domain":"finance","full_message_tokens":80808,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":661,"sample_id":"sample:ae787a9b0c2ee30be8b73e6e1ae9224f0b9816452afa3dfb00cd49fde8cc7d9b","semantic_task_id":"09ecb5d2be3057b2ff5baa1fa5c8c2a465f4f31ff3e7cf358485d2b83f9cb220","split":"train"} +{"domain":"finance","full_message_tokens":81036,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":662,"sample_id":"sample:ff79d330da6ca119b66c52906ab3225334b52657939b0880eb671e06880baeb8","semantic_task_id":"09f2e5b426e2b4114a6a0c3d5e6a3b1fe41270f7f093618152a29c8ebc4afac4","split":"train"} +{"domain":"finance","full_message_tokens":205430,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":663,"sample_id":"sample:fae513dec6bdc69e3084b05046f6dfd21d980cbf59047c77ce17d1d261cb89a4","semantic_task_id":"0a0d3e9b6a3fa25717574573198363464946811dc347689f399f37d4632ed2fa","split":"train"} +{"domain":"finance","full_message_tokens":70025,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":664,"sample_id":"sample:bc052706b27d0adf14e492b88c011f86f51bc0fb868a9950029afd9053f90475","semantic_task_id":"0a80fd55eab81d61af1b88301dd814af9bab6eb009711cb20dda053428dc3d49","split":"train"} +{"domain":"finance","full_message_tokens":80671,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":665,"sample_id":"sample:ea10f766dd6ac3f48909afa626bb40a9f790b3dd85d1daebfee0d2593409bd6d","semantic_task_id":"0f1a8af4a9b3984c3719631961bc36ec5d744e2d1f1d3809de72644e42a501d2","split":"train"} +{"domain":"finance","full_message_tokens":67120,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":666,"sample_id":"sample:d10f930ad084fa037dbb09ce46ef07147fa592067dda2fd87561f771501454dc","semantic_task_id":"10a869b754f052398cfc3eb15d0186cea8650b809f54e3aa4c88db1899bf2274","split":"train"} +{"domain":"finance","full_message_tokens":205452,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":667,"sample_id":"sample:fc8828c9165944067e5cb184ae9e65e5ce735049a150a733708645497b0ab1ed","semantic_task_id":"1290b401693a50a1765692b542ba493814ba0b683fb9d23420e7bb8017733d4f","split":"train"} +{"domain":"finance","full_message_tokens":74964,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":668,"sample_id":"sample:995d8013eab90ac697d89f98eda896f7795b7227cd01c8d1bb3fdf1c5e4319db","semantic_task_id":"13db283671278f8e53e0469e8e98b8b2c8e227adae9722db34ca4eb235ba0b1f","split":"train"} +{"domain":"finance","full_message_tokens":73655,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":669,"sample_id":"sample:d67189f9cd9a06c399966dccb0f085464e510db47261ab4f91f16a3dc1be5985","semantic_task_id":"142b870eb3b142d14f322a973dac8e6eaa9a1e729fc2995e1b3097f082720af8","split":"train"} +{"domain":"finance","full_message_tokens":73654,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":670,"sample_id":"sample:56322a654c7673140ae06ebed4785b20d91ce4df268146b09140cf996cdec9a3","semantic_task_id":"17cb1249025853e58a2aa8467f15979d0d2281e8e1a7ba37f6826dcd59c35551","split":"train"} +{"domain":"finance","full_message_tokens":214456,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":671,"sample_id":"sample:8d22b87c2bcf63a212561e0638fa64642c1674cbdcb7f3f9947fba08b720846c","semantic_task_id":"18dd91e7e929616f4dc71fe02cac32b014877c3d9dfe463040f4599749947078","split":"train"} +{"domain":"finance","full_message_tokens":68652,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":672,"sample_id":"sample:f628c60926102d482397a69212cb2d8bf4b1ee0d065ab3adffa7788779c5d957","semantic_task_id":"1aa0f48b959b976c2f6eb9cd4b009cb2552bec5765ab6139a9710852f496e1bd","split":"train"} +{"domain":"finance","full_message_tokens":74962,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":673,"sample_id":"sample:d77273b4a045ed5f4c843f61cfd33133c5cf1f6590926de7c69aef892b2a2f25","semantic_task_id":"1c3d52cb10aaa617ff6749a6d9f386cdd55364daf950a1de29e739863fe0a3c7","split":"train"} +{"domain":"finance","full_message_tokens":69957,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":674,"sample_id":"sample:8594695d84195d752abdadc4e205a1ac31eaa7c9495f43de66bad1346fba6441","semantic_task_id":"1e8adeaad8930006a627f571817ceb4f2725cb81232c1fcfb1423c6d583beece","split":"train"} +{"domain":"finance","full_message_tokens":214458,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":675,"sample_id":"sample:a431797a1e026f9c4dfbe9ea8bb382974cc3bc74207f2ce3a42692ec5712029e","semantic_task_id":"2061dbca682b3716315ee7bb74c9b0071f8d15dc7096ee22232867774714ea47","split":"train"} +{"domain":"finance","full_message_tokens":80672,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":676,"sample_id":"sample:1fc3479a436921ae18d1a5ffc6da27879259929c30e03f95e6ee02789c94827f","semantic_task_id":"206b96bda5ef83367cf2016e2a0a28530c97c8b99f5cb82cd30b3eef76651015","split":"train"} +{"domain":"finance","full_message_tokens":214457,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":677,"sample_id":"sample:dd36b4e99459f30d1283c9992971ce90aebd745d521a0ab836af8a2406ae6543","semantic_task_id":"235895b74c44ceccf87815046c48b1f7dc3f37217b2b4c4b9fe435f0b8c8893a","split":"train"} +{"domain":"finance","full_message_tokens":68699,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":678,"sample_id":"sample:1d10ddc72ff1dbd4cd7fba72356fa7a86b68d9b101c7522200ead272df9d54be","semantic_task_id":"23b175eaf90ff6785d059ffb0025eff41bae1241fc507ebf1db416430d64d13c","split":"train"} +{"domain":"finance","full_message_tokens":74734,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":679,"sample_id":"sample:a13166aab89ec9660a5ebb9e66a2cddb1a7bd0ec2f274f322c8a0d6d6d69aa27","semantic_task_id":"2400456bb03213305337d41f70aa860ca31e7384b55768a3aee4869168a54906","split":"train"} +{"domain":"finance","full_message_tokens":80673,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":680,"sample_id":"sample:9d03a5fc93cf7ab4534095f325332983f3c438edb6f2dd11d50b5b3fee3520f5","semantic_task_id":"2439d3e15b990bd1554aa0441da531b2fe106cdae42d4f5f6ad147844fb5bbcb","split":"train"} +{"domain":"finance","full_message_tokens":74964,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":681,"sample_id":"sample:da42eda355c2f607f77f648f877653efb04b826eea90cc30f062b9b0bf1d4706","semantic_task_id":"2972d1beea311c155bb3a19184fc4ba7b46762ea1d039a2e174521863c7128dd","split":"train"} +{"domain":"finance","full_message_tokens":219443,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":682,"sample_id":"sample:dbb89aebc25c16f4596e7c3ee9a6844ab308b8bed7f6bb9a9e5b95a623c2956e","semantic_task_id":"2cd92d43e57f9ae53e0f763e5f0a40a2c031e05f0ba3c30a4695f9524a8d1f63","split":"train"} +{"domain":"finance","full_message_tokens":81037,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":683,"sample_id":"sample:43654db0c9833d8b5fdc3fdc56ec6d4aa95dcf42e13bcc9edb86ce715d968511","semantic_task_id":"2ed1afc3acee97131c348f4f62a2f817dbf79fa2d65199e7dc8d2f0fc8ad41a1","split":"train"} +{"domain":"finance","full_message_tokens":219455,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":684,"sample_id":"sample:d9572dbe583e9908afe37ca8a50b1cd7011ed9ba00d9ff87bcac74d1422db402","semantic_task_id":"2fa53ebaa6ce87cb2d1ebc733fe7b8c572293f848022be178c7d19d089f1681a","split":"train"} +{"domain":"finance","full_message_tokens":76030,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":685,"sample_id":"sample:f5faf176970913580cac00831ef4d555b3cdc0649247e713719641d9d40799fc","semantic_task_id":"2fa6a0a55ae8e13035c11e59539b109a1c67ad86f9f487ce6830bbf7e116a03e","split":"train"} +{"domain":"finance","full_message_tokens":73658,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":686,"sample_id":"sample:c4551832bb6a4152f9842e7d889eeb3e5fa807315cb002ee3083c11db81e7b59","semantic_task_id":"335f035d3bf3c59d5ad08b5e2b7f3873292fcd89df39b1c08ab5ab8a5713667a","split":"train"} +{"domain":"finance","full_message_tokens":67091,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":687,"sample_id":"sample:a9bae047a9d6493f7d58aa5461f8c3bbe4e786da235c38a8243e7f6def4b192f","semantic_task_id":"337652dbf5a4caa53ca71f84032e03f9d64430f13b7a3a8cba82ccaed06671d2","split":"train"} +{"domain":"finance","full_message_tokens":67021,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":688,"sample_id":"sample:310d781ea26fe382f2341478efca32a6faf0ce161ce517b3ef556fa3ed535c0f","semantic_task_id":"35eb6202016ba3e00dbf141f1f073a3f9aa065c831d2ad18c7ac716cfc827bba","split":"train"} +{"domain":"finance","full_message_tokens":80667,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":689,"sample_id":"sample:b468dbf91819c97eb648e8480fcf0a20c28f7d909a7102c2930bc34055bd6eb8","semantic_task_id":"36201c39c936eba1cc7ed06a6a73573ba92d5942ca43845aab390d736de8ebcd","split":"train"} +{"domain":"finance","full_message_tokens":76029,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":690,"sample_id":"sample:be16622d08567c3b32c0ff63f2c26b64b25543735ede314568a728ebfd86bac1","semantic_task_id":"375260fd3dcfcf09dd99f97ac718982858ea7027d1f723e40c3aa1f42d1af89d","split":"train"} +{"domain":"finance","full_message_tokens":219441,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":691,"sample_id":"sample:4dc1a013b61c6a03b7c8be4be307ab089b085c42b0a3de89ec496305aaf9639b","semantic_task_id":"3b3f2195ee63f6cdf25c3e473b05f1f19ccb461c66aa257b25679a5ee929ea4b","split":"train"} +{"domain":"finance","full_message_tokens":214437,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":692,"sample_id":"sample:446980a523034613d1d1b0205d7e3965298b7f31efcee7ee9211b71ccc488b1c","semantic_task_id":"3baef49192f355592fdfcfb8b3ae7a4cb39663e28f4a91e41c0f9cd4b0b03862","split":"train"} +{"domain":"finance","full_message_tokens":80656,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":693,"sample_id":"sample:af04ff96978fa09bdea610d717b260fb6473b66566f4b0dd5f8aa6841feb4843","semantic_task_id":"3bc74f12801d9b28d9d874483ce81362e6edd6864d31fa5ce05652881283b3b1","split":"train"} +{"domain":"finance","full_message_tokens":214438,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":694,"sample_id":"sample:061794eb52996ee9887c58e069cff2dcc330b05017b10dd70616a6b86f197524","semantic_task_id":"3bcb0341c056dc76a9405469b7c7b5b1b2bd4ab897725daeedb658b41478396b","split":"train"} +{"domain":"finance","full_message_tokens":214439,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":695,"sample_id":"sample:3e1262b88f0746394c6f6aeab0426df7b8df92227908ca1071268a8b57bd129d","semantic_task_id":"3caeeea3e09789255482b2f7f1440d99fd44120556d5121c7c849eded2fb7237","split":"train"} +{"domain":"finance","full_message_tokens":73660,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":696,"sample_id":"sample:1bfd221293331b56b5a76938c8b08162ab36838412c7a2f654a0f26c39313cd0","semantic_task_id":"3caf432d80496b626fa68362f42a214b09ef30740454b4c44d85ce01ba770700","split":"train"} +{"domain":"finance","full_message_tokens":76031,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":697,"sample_id":"sample:d3b0318d1bf37f343631b61d6fdf8ea45c17077f3583a044894b1121bd3628b4","semantic_task_id":"3cdbe1830aadc829a5f04e639d95f94a6eeef96b8d9931e3da4f4b4558d98c99","split":"train"} +{"domain":"finance","full_message_tokens":214460,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":698,"sample_id":"sample:b7827177f46c5db96c1aecdc30276e8eded1862f8cbb9baf06ac49bb3505640d","semantic_task_id":"3d93b1c4ae789337eb6941cf508f8ef94464bb6017749c4a8aebafce45a7da8c","split":"train"} +{"domain":"finance","full_message_tokens":214436,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":699,"sample_id":"sample:a86727540f98e743dfee8f209005846fd5ccb4b828d45e529a3470ce7a308897","semantic_task_id":"3e780f999cda26ff28484223aa214a24d91cc7d16f06e141395710eb073d5637","split":"train"} +{"domain":"finance","full_message_tokens":81037,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":700,"sample_id":"sample:61582e0d0640d0640d7141c00b26207ee5e20df246295d4caf06afee0ad591a5","semantic_task_id":"4001ec591a1ecfb4a7e68907b90b87eaf6c2d7e665c3ef5757ea171b05ed192b","split":"train"} +{"domain":"finance","full_message_tokens":80806,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":701,"sample_id":"sample:175a3c5ca5f6151c481049e20a25a07ac1ec267e64665618effc2e025d185225","semantic_task_id":"4038f28cd90a8e86f91fec323cc197ac4e28d3d8dae26e5dbb86dca6ea3593c2","split":"train"} +{"domain":"finance","full_message_tokens":80673,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":702,"sample_id":"sample:09994683182b2349f1bcc0d1275b9896c138f8f6dc1574e7b250e9810758efb4","semantic_task_id":"4192651bbb4dae0c1bec2ca72915cefabf728d07466378030b91ab194d7d1e41","split":"train"} +{"domain":"finance","full_message_tokens":73660,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":703,"sample_id":"sample:832b7b747f1365efff9b10a67c33b25431fffbe73fbc3e090e87a938def5688b","semantic_task_id":"42d4bcf37d674610500ca298001e2f87626a8bb5b994868e6e1479eef2fccc32","split":"train"} +{"domain":"finance","full_message_tokens":205450,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":704,"sample_id":"sample:05f24ef116898ee490d65de7b3c8bdf81fbe9e8b9a7436d0d6b1d402e6263e85","semantic_task_id":"4646f516fc2660b87c7a2c7499183737b12ef04a41318bb1c34705518ec55ca7","split":"train"} +{"domain":"finance","full_message_tokens":68722,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":705,"sample_id":"sample:d1e6fed113d6f79a5e0d3da0a4dc07942b4cfc2e581c513f513ecd651e4047ae","semantic_task_id":"47575fe019c97380911cd925a083ee90850566ebef9147345c7ab0a30c060591","split":"train"} +{"domain":"finance","full_message_tokens":73643,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":706,"sample_id":"sample:bc281aaf8d0b153e6a46ac070d14bd9796bb3264ef12b5658a022a554347f899","semantic_task_id":"4926aaa4efa4aba411012e718deffd8c9fe273cf534067fab364205552ff4a36","split":"train"} +{"domain":"finance","full_message_tokens":219423,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":707,"sample_id":"sample:97e6c0d96fa4835e0af195af8f72624dd6f33aace850130b55acf3f6b169e953","semantic_task_id":"4a53ebc6533238f12ae453f66a476ef83f30793e48b111f083211dcfa096b72b","split":"train"} +{"domain":"finance","full_message_tokens":70003,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":708,"sample_id":"sample:6ff82b92e50830797e5ad98a10c309abf5078e1ec8c0e415cfa1f99ed24291ab","semantic_task_id":"4b335aff79cc5a4e139ecde1ebb16907f12a2fd31ee1b4034b1fb78a5ce00e13","split":"train"} +{"domain":"finance","full_message_tokens":205431,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":709,"sample_id":"sample:bce160af93214dd2d1fb8c57c6e98869818cea38e9bc3756aea35b41ccc1d15e","semantic_task_id":"4de325e6acad19060cf81eff1619a4e7567b03ce42635a172a9e670ad36a4be5","split":"train"} +{"domain":"finance","full_message_tokens":73659,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":710,"sample_id":"sample:3bd698035f70be68d5755652a45308bc489576f3fa7a7deeda25a72a95aab16c","semantic_task_id":"4e3f95cf8e22ebc4935d5f5fedff50adef935a165c5462ec1404d060ba387e5c","split":"train"} +{"domain":"finance","full_message_tokens":80673,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":711,"sample_id":"sample:ab776101ef8cf62f1e04772cbb101280c832c675c77ca2a1dad61507990fd154","semantic_task_id":"4f0f96c959b6e46b3fddb1c4bcf4b6bca2beaed2fbdaa154930893b13f424654","split":"train"} +{"domain":"finance","full_message_tokens":205441,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":712,"sample_id":"sample:24baa2f8d2c5e62866e104f1aa9a538eac0eac308cff1d8f705ef84cd74fbdc2","semantic_task_id":"501954a0c846397e2e9a487c403f582b5eafddb873c0f3c9ed7dfa80bbdc8289","split":"train"} +{"domain":"finance","full_message_tokens":73659,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":713,"sample_id":"sample:0cc8916a79bbaa60635f2af45fc9d62b22818f01ad548fd69c96f4ba7e959536","semantic_task_id":"53846c9de0fab4e7df8cc8677d4c49b0ff47a1b6fdda13928ffe477c0fcd8ada","split":"train"} +{"domain":"finance","full_message_tokens":219422,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":714,"sample_id":"sample:869d9d2661480bd91eacaaa3d89c4946658dc6625819fc9c98936ffdda2af4a9","semantic_task_id":"53b8ad810213fef4c32d40e3e6547c20261dbaa8526f91d91e8fcbaa9b512956","split":"train"} +{"domain":"finance","full_message_tokens":74732,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":715,"sample_id":"sample:f5c7625083479c488c16df5261a3f58db79b4e546009d7319fd1e576ea7def68","semantic_task_id":"56a475ab54ebca88a5084cd192c1cbcb0bdf6778f8c6b0b82a3e0d69c05927e1","split":"train"} +{"domain":"finance","full_message_tokens":214437,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":716,"sample_id":"sample:1cb1a21e4a8194261910cde926752a7409a4cfe773f6cff0cb51f63d18f43692","semantic_task_id":"5994f95ec5acf7cb9b589fc572d57800753ecf35f32576c3332ad3804a6eb791","split":"train"} +{"domain":"finance","full_message_tokens":67021,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":717,"sample_id":"sample:a31b4e3df5cbafc49e9d532503734c0c7cf201b240c6ab1a2def814a8804166b","semantic_task_id":"5a44cab41ae135f3162024700f764f66c165d13ba433b2c5af6d91a49e362a16","split":"train"} +{"domain":"finance","full_message_tokens":80672,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":718,"sample_id":"sample:87411a7dd4866d247d76cd35b46ede9a63c3322af9d7f0301f75fbea79efc5a8","semantic_task_id":"5a8979f5b79cf741260bcbedc206bb0f5c6f69dcb8f8e9172ea4b55facd8fa18","split":"train"} +{"domain":"finance","full_message_tokens":69955,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":719,"sample_id":"sample:5ce461fe2a9c41c78d055db5e219a7f26af736bb873130a593852ce9be565f3f","semantic_task_id":"5cd0ad148e572fa9efcec1460fd6dfe97f20f9242b26fcf0a1abeb6182b22a86","split":"train"} +{"domain":"finance","full_message_tokens":219456,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":720,"sample_id":"sample:b81d6ec28e6e34a1e1ac7d0ee66b6bd4c6f2600b32ba02c99a43eef97bf65bd0","semantic_task_id":"5cf93669fe2522d26b010600fe9e6a9e9c346c2f3fc5aa43a3ff2f9f71267c65","split":"train"} +{"domain":"finance","full_message_tokens":76101,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":721,"sample_id":"sample:a2b6c30816adee49b979599162e215c64d480c6db01916f10c757c955601b7ea","semantic_task_id":"5d4aa4771b1eb9dc7b99aea5ae491046d80d716d3275b0bd51ba15c6744304d8","split":"train"} +{"domain":"finance","full_message_tokens":81037,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":722,"sample_id":"sample:668d69374768f38f3010b0e7ece135b0af0212ab0316ca9e5069d5a607bde484","semantic_task_id":"5d611b8f3b0fb4cf382876728e9858077dc4e1821c2e53b5a3ec0db2cc9e41d8","split":"train"} +{"domain":"finance","full_message_tokens":219433,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":723,"sample_id":"sample:36c83e9aec05a45547b54d55316c25262fcd7273db55b769af8b5c46fa39afcf","semantic_task_id":"5fb0440dc5e16ab6ca7a0e95e1ec32e2dbfc3e3c3a7f0a17f73ffdb6db463fd7","split":"train"} +{"domain":"finance","full_message_tokens":214439,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":724,"sample_id":"sample:5a50a158249aa29f91176cf28ed0d5b1050235f1225470c0769a15bc7d520ac2","semantic_task_id":"6099369157692b9f398af246bd74c50259cdfda77abd7cc6aa6249e4c1983047","split":"train"} +{"domain":"finance","full_message_tokens":74717,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":725,"sample_id":"sample:34ef111eca3035cfcd3e34a770fbb2dbf70331d6c1564be0d2abfc59bd103758","semantic_task_id":"6294a8a275b2f012f8494e3f118e88bf671fa5f01e8cffdc8321c9ce88e2592d","split":"train"} +{"domain":"finance","full_message_tokens":68651,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":726,"sample_id":"sample:b245f5f82d35020e20be86ee6298c5d0d15df03e2e84eb17971a93b18b160ce7","semantic_task_id":"62a8ce4c4fc71936374c83be4b5602799247081c70439dfe39974112b92b4943","split":"train"} +{"domain":"finance","full_message_tokens":70005,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":727,"sample_id":"sample:fd43fa9c01027cf1a5a92887d50cba1dcbe487abf0d64bf3fa9a9d0541f88292","semantic_task_id":"63ffec08dc2c2831ff21a4c22cab97075c52bc183f56dd9597ed5a266cd5d266","split":"train"} +{"domain":"finance","full_message_tokens":205448,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":728,"sample_id":"sample:1f8831acd4f0b948036a15690d7c9dd3b1080a89941da39aef44698d34907778","semantic_task_id":"65befd8e4a5ca23eb2744c718ff5fb88d6d5b7f72279ce436f64809a293f4bff","split":"train"} +{"domain":"finance","full_message_tokens":76080,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":729,"sample_id":"sample:1c659f920f7fdd7d4d6c49edd6e69cbf58ac2ffdb9ed9d8f517e7d35178ecd5e","semantic_task_id":"66a3d80cc1a12e10b39feef23586dba6223d2626fc55d75ee8e1b4b698751a49","split":"train"} +{"domain":"finance","full_message_tokens":76031,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":730,"sample_id":"sample:e15f7dc38512e0aa591dac3e2dd466f169bbe1a77bde026727ee3dc53c22a433","semantic_task_id":"677b62e4f9ac9bbe5d0c1a89a4197260c54e76c7f013eece6298d7ddf1b7cee4","split":"train"} +{"domain":"finance","full_message_tokens":219456,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":731,"sample_id":"sample:f454c5f59e72e3d6efc95be78ec2aa0b7dd8f13a8a468e3c53ef9f882ff66ea9","semantic_task_id":"67e13dda4f2c375b6d031d88f4f01c041a3f6dd1cf83111dee34946213ab5989","split":"train"} +{"domain":"finance","full_message_tokens":214473,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":732,"sample_id":"sample:54d7b178fc4df3d2f975cf74136638955d8ea787445de2730912c9666a411736","semantic_task_id":"6a39d5958437c9a01260aadc2d67028c54aa1d8bdcc7a108989fba0ac4f6b8be","split":"train"} +{"domain":"finance","full_message_tokens":67022,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":733,"sample_id":"sample:11a9783d03f4248d59c5f907f834330638c48d587455f26e15c8a649d9229769","semantic_task_id":"6a5f15fbd58cc3f19b00ac8cc12e09f0a28a367ca22d7cc556193db7e0842cae","split":"train"} +{"domain":"finance","full_message_tokens":76028,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":734,"sample_id":"sample:ccb667e287de20bd2189d0af5574fca94ca3800a37173c71a5eb8d531068b88e","semantic_task_id":"6a8597e496a2e317b5c355381ed0227c1ea2166c5a5e863b7816600060bfea7e","split":"train"} +{"domain":"finance","full_message_tokens":219424,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":735,"sample_id":"sample:9f04261dc28421c2c4d84f1639f31b17ff721305896c1068ac212b8915adbe44","semantic_task_id":"6b29ed316a9d06f79b77cc27d82ed60b82a662d383a34f890bae300acfcb029f","split":"train"} +{"domain":"finance","full_message_tokens":219455,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":736,"sample_id":"sample:50cce290ca30a56af1c296da2381892c46e2911e6f744e48a2dbb26fba716a55","semantic_task_id":"6e35a5f74261dd8af4b441ad6255b9923942fc96799ce1dd0170ee54a14f640a","split":"train"} +{"domain":"finance","full_message_tokens":80667,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":737,"sample_id":"sample:8bedc7abf4207fd721d23c9c8a553f3d0bdbc18dbdc06244af952229c851f5d0","semantic_task_id":"6e6675fa53a872022a09b938e663fb39b8473a4820555ee753d2d55539879cb3","split":"train"} +{"domain":"finance","full_message_tokens":69955,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":738,"sample_id":"sample:32f507390c5ffcd511d0e87484564652ec38ff7358af3b14f35506bd32c7c585","semantic_task_id":"6e774132d35ba3a49fd3667bc2b4098aae7f9f9495e57eeb7518fa77dbb0a299","split":"train"} +{"domain":"finance","full_message_tokens":80807,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":739,"sample_id":"sample:141099fde0b519ff95e8b42ba33df39b0d48bcc3a1ebb1d2319aefe6df7bfac2","semantic_task_id":"7136d4be569a05b24379b02ea55946a5abb02f90048f65780ddae2fa1aa19e71","split":"train"} +{"domain":"finance","full_message_tokens":74732,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":740,"sample_id":"sample:8631729c3bb69842f5659bb7e2a8ede928a4c4047bb01bf49f7c6df34e45c60a","semantic_task_id":"716ab381bce6e5aaa93f433a0e9b45ab011b37e1150b1b80dd7ee92088409652","split":"train"} +{"domain":"finance","full_message_tokens":73658,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":741,"sample_id":"sample:2fd2de401e32cdb50093515a523ffe325044091ce0728158e02d0a52a1981a90","semantic_task_id":"72c4bf996c21e3bbff46d311c375dd280e10c8b2312c9072634e22689f5111f9","split":"train"} +{"domain":"finance","full_message_tokens":205452,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":742,"sample_id":"sample:fbbf162f34c531bd85530d2df3ffc1e120070f739c463407ff9722a17767f150","semantic_task_id":"73663760c56a37d17e6c31e400c594c6c9a879da5b9b5a239936b051c3261adb","split":"train"} +{"domain":"finance","full_message_tokens":219434,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":743,"sample_id":"sample:3549e4b47153222f6d077f662233eeea7f42db75a462cac89c9f77704fcb7f7a","semantic_task_id":"75770168cd9cf3bd2990727516a04aa554c518162b8b8bcc2dd3c5715cc0f21e","split":"train"} +{"domain":"finance","full_message_tokens":67069,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":744,"sample_id":"sample:bcf7d20ebcee98b2667100bb2945053005e0db677bea68774522874232e3814e","semantic_task_id":"75ee4c19052cba1eb80be37c6c93e96f75e02cba9752ba38c6825488b41ffdc8","split":"train"} +{"domain":"finance","full_message_tokens":205440,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":745,"sample_id":"sample:494db4ff6ec36f2e4fa4b2d5883a41258f8be6557efb66dc5bfd74618465ebbb","semantic_task_id":"76eeebe6ebf3ca320547920ca1a0d76b1257bded165a989b79762ede04ca0a5d","split":"train"} +{"domain":"finance","full_message_tokens":80657,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":746,"sample_id":"sample:c5ca08c3759683dd6b6c8f577af23a242e4038b0bf3afbac6116981c24c3bf0c","semantic_task_id":"7a1d37d765a16a643c3454002194638fc0ef15df25bb87e5b9cde168d10ab5a7","split":"train"} +{"domain":"finance","full_message_tokens":68652,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":747,"sample_id":"sample:e8197a93e2c3cd19953884c17494ace1e9555872ca03b9ee90c7b73a1ed91300","semantic_task_id":"7a8234c027ce5e5d63f2fcec63496cdd91ffc65c5275c21f7f02985fc4ca550f","split":"train"} +{"domain":"finance","full_message_tokens":68652,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":748,"sample_id":"sample:8982ed3181089be6d24a7fcec3a2387d1832190808b811ba255855d36b1bef6c","semantic_task_id":"7b438d4ca575d56b8d94a6d0bb4f53cb086d045d9ca0b97b67b363911b9308f5","split":"train"} +{"domain":"finance","full_message_tokens":214459,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":749,"sample_id":"sample:cd01cf153d3640de931679c1615fc5663d93d96c461b969276d0f0eef552f707","semantic_task_id":"7b998cf7835936f0b5ab962827b1bc17349991a9361175be40543c4bfce6fef2","split":"train"} +{"domain":"finance","full_message_tokens":69955,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":750,"sample_id":"sample:336d02db595cf64f7421cde8a2acbf10c05b14e0dd96c2d8c582b403b877c9bb","semantic_task_id":"7ba1c107ff90d20f1965f0d5c890c5ab72879259c56e6d8f2838fd62926b7a3a","split":"train"} +{"domain":"finance","full_message_tokens":74734,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":751,"sample_id":"sample:a88d6ac9aae6551476f4b983b8be2dfdc537dd7c6b563092126b0ece499655d4","semantic_task_id":"7bc50462a69517ca587dbc83c0e6a302da0f3828264c60250fe57a05d138f42f","split":"train"} +{"domain":"finance","full_message_tokens":76031,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":752,"sample_id":"sample:04275bc9f17a5f0280d0e8f862127a7c28be19b28a93df686b81695a0c88a876","semantic_task_id":"7f57f4ed0c1a5f00ada1436ea19807c66274448277e2b4090bf2088372d5ba26","split":"train"} +{"domain":"finance","full_message_tokens":67071,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":753,"sample_id":"sample:c97dd47a7610d0506736194a64092f4804f2fab9de24c393b142be847a0c3d40","semantic_task_id":"7fcb31243a3a25dc3e284a047209e2d84f400a59255163ff350459f5ddd184b9","split":"train"} +{"domain":"finance","full_message_tokens":214459,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":754,"sample_id":"sample:0b530e2ee940f77eaa3b5a955e7dec6372152519db1fe8f41f16d06edf13e25e","semantic_task_id":"7feb2f1c9735d757d2084513dac528a19093b779f01e8f3ad62bab2ec059a671","split":"train"} +{"domain":"finance","full_message_tokens":214459,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":755,"sample_id":"sample:c453e23f07d39a2a53e2cdde09c79f74611e90a4751b578a7ee18efc6f1f460a","semantic_task_id":"817adaff5a25363709d78f4b7a0431004fb8da3e2f20dbd32987fe750a140a10","split":"train"} +{"domain":"finance","full_message_tokens":73655,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":756,"sample_id":"sample:01387de826740a7037c2f7127081f22e076d4c3199d54e3858c4fec1c764d70f","semantic_task_id":"818893444df44023148863b7a154ab3b09d9e93c13ed2c7b79bf9b4d5b41e759","split":"train"} +{"domain":"finance","full_message_tokens":205430,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":757,"sample_id":"sample:5457f0f9bfaae6112f89edf0e90f4c5f582697346b3a628971e348305ca737bc","semantic_task_id":"82322c8b7008b13125d4bc8cb7eb2869e552dfb528bf0557fc341bc6ab299cc1","split":"train"} +{"domain":"finance","full_message_tokens":74728,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":758,"sample_id":"sample:e213db8e67137ecab02693ed4320bcf93c6dc77e970c134de332377b381ef53b","semantic_task_id":"834d461a021d00346a6c60366cfb3ec6f7d3af3417095d093b3331a02a854ea7","split":"train"} +{"domain":"finance","full_message_tokens":205430,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":759,"sample_id":"sample:b18edc8a3f9b75ed223979f6a77e07a46ac8a43bcc490dd9d4ae5c5d1e3ac6c1","semantic_task_id":"83892788f2904fffc67686fdc00f7622807df6aca08007844dd6f9c0e8be4bfd","split":"train"} +{"domain":"finance","full_message_tokens":74963,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":760,"sample_id":"sample:625a8c2a68c8f35dd5d8438d18accd56880f0969a779ac59a4eb216cc38dbc55","semantic_task_id":"844c2fe33b665da0e92206caf69be709ec0840c4875ee9821cf5bbe16f42d98b","split":"train"} +{"domain":"finance","full_message_tokens":74964,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":761,"sample_id":"sample:d885988f6e82ec59b064320f34a29fdf1df15ce6206cfa2a2ddbce901e92052b","semantic_task_id":"859ab4951b2c3e150669b92ffdd057d98425441400e14c713785530b1eaae038","split":"train"} +{"domain":"finance","full_message_tokens":80668,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":762,"sample_id":"sample:b8806571e357ae908b79130a0d4a36d2807fff5b60892ebdd8f0f64d133e6a60","semantic_task_id":"85a1a98b877eeee00c9c9d53fcda37a5497d4a689178ae07f0a138cd0e0b8359","split":"train"} +{"domain":"finance","full_message_tokens":214438,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":763,"sample_id":"sample:314f6c12a873f3bece9f58354dec8459a96d5d5c76066388389be5dec3d98a05","semantic_task_id":"861532b7a7ea99a5613363e7775bb35c8898309745e4c08f0b2696f0f1b9ea35","split":"train"} +{"domain":"finance","full_message_tokens":219421,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":764,"sample_id":"sample:f6f9da3a8416e0e4661980e0aeb576e02de362d87cdba5bf3abc9a9b71ba45c7","semantic_task_id":"8a171e5a49dc5f8c7303e93dedd06828839d6b02953425bd2afc9b3d3c3e8889","split":"train"} +{"domain":"finance","full_message_tokens":73659,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":765,"sample_id":"sample:3039adda4704be9dab04c63e9bee5cf077dafe27500fc7fa4a757a988a7b56b1","semantic_task_id":"8aa00c3c75ff6305f7343672f7088fb231543c1d9c55e4a1af555930f7370712","split":"train"} +{"domain":"finance","full_message_tokens":74733,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":766,"sample_id":"sample:151b7687c99c79dd1c5a53c0ceef9b9d215322944244680afd1b0d5caf22c725","semantic_task_id":"8d9ed9f9f2a0a17ec303606c50c610ac1e9651db377313b9ae260f07f715fdcc","split":"train"} +{"domain":"finance","full_message_tokens":73643,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":767,"sample_id":"sample:985bdd81af63f199a1e19a4f5e1304858a4c8bb64e249d106304284644b211ce","semantic_task_id":"8e4e9449c6b486d96788e2998e32c7108e365bf438e09059fff81470ba13a14f","split":"train"} +{"domain":"finance","full_message_tokens":73645,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":768,"sample_id":"sample:e2ea35820a6156d277fa837994b7b2ec1fa0e5ea6bffb3c1bc54b023baba9651","semantic_task_id":"8ee923d632df4d68d0b3d1bbd2b385d69256bdd59001b89cb35fb28bb6a8000e","split":"train"} +{"domain":"finance","full_message_tokens":67076,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":769,"sample_id":"sample:ac91d4ae34f72c3b65406d69ebcfe335655f8f3e66cbaee7d30adb5fb89aae5b","semantic_task_id":"8f7a91a91ae6cdfc7b2756383f14b37fdb98dc3c9d0c8663380beaeb2c811a57","split":"train"} +{"domain":"finance","full_message_tokens":74962,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":770,"sample_id":"sample:92e7c51f3c19a448fb0add6f787b15e44bda35952ce1c2c333aa85432592cd5c","semantic_task_id":"9078388833e8eefec0d0383ee4a101ac71af8dcb2070d081b24f1b44d09805ba","split":"train"} +{"domain":"finance","full_message_tokens":67022,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":771,"sample_id":"sample:31bb939a5dabef4c3de68427963cb81f43114b625cf15a2e4d42c06243e041ae","semantic_task_id":"92c82663c7cb277ab8acf634e3e197e30109cd381b3942dd637ef2389719a54e","split":"train"} +{"domain":"finance","full_message_tokens":219433,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":772,"sample_id":"sample:643dab062f857fd155451c10ebf96f9e367daeee51fddfc4a7103b13ae6528be","semantic_task_id":"93980a34e6cd1a041a8fc4b722d9cd5d540ddd6c0ca1162f800abf95cd101fbf","split":"train"} +{"domain":"finance","full_message_tokens":81039,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":773,"sample_id":"sample:272a8dc958650cc2e80d8a16cc567fcd93f7e54bc5a53e1665ede8da9db23255","semantic_task_id":"9446fa54098183532405a879b0009270d309ded91910a39d02f7a6eb7e2a05c9","split":"train"} +{"domain":"finance","full_message_tokens":68651,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":774,"sample_id":"sample:1cb86b3b31a23888decf0312fc78c23753656a9875a5204c9bb5cce4a366faf4","semantic_task_id":"95c59d23c52a200dea203ab4949572c1333900a27921e832b614289f8c95fc7b","split":"train"} +{"domain":"finance","full_message_tokens":80669,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":775,"sample_id":"sample:cb96e1465299b341d6e6e81f41206b61e023546cfaf5da0bff558e3a645ff099","semantic_task_id":"9ab07e7be8343580deae5278cc58e140e8cda21ac76ad8f7172cd69d98b9d09f","split":"train"} +{"domain":"finance","full_message_tokens":214447,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":776,"sample_id":"sample:575bc4c097901167f931c1b79c917413f5419d4b93ac32393f4628fa5b59038f","semantic_task_id":"9bf15b109e0e236b8d19a36f4f689257f105648b1d7a62412c707f4588133f5a","split":"train"} +{"domain":"finance","full_message_tokens":80809,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":777,"sample_id":"sample:5c0e982cea3a1056feaa2b78ecfb84ac94f2eea4a58a4e4c7988c31a70a12c1a","semantic_task_id":"9c18b80027647d70e099b989ccab1e4d889d722a1ff09dc80dd3081ec6952819","split":"train"} +{"domain":"finance","full_message_tokens":69955,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":778,"sample_id":"sample:70d471fbac44ecde06c57327166c03375f9fbcf750844c920498434cd33d1826","semantic_task_id":"9df62413d5b5b4c719397be4511d34074f96a0e4542e88eaf356b47e5ffbe3bb","split":"train"} +{"domain":"finance","full_message_tokens":80669,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":779,"sample_id":"sample:c5c747c327a8de704865b11e0c14d75b75a8351d25e1779864fb932b85c0de6a","semantic_task_id":"9f20807ffe04c6253bf4dfb9c1ea8448a1ab7bb77c9560dc11a27651bd3e477e","split":"train"} +{"domain":"finance","full_message_tokens":76029,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":780,"sample_id":"sample:cf27fb670c97a8f3774967d5253182807e16e38b8061b5459bee02ac7b2caed7","semantic_task_id":"a444e9a24f0dc3d13bc8c60b959fa9061644bc28360a739302bcfa94c871cff6","split":"train"} +{"domain":"finance","full_message_tokens":68651,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":781,"sample_id":"sample:25623b5866e8ebb28db459c96fc15067c430a558415da77e33d8b38ea5ce3c0e","semantic_task_id":"a44715974d69aaa15bee06ab4442b9212f90644bd741e2f734f61b5dbee7a6a6","split":"train"} +{"domain":"finance","full_message_tokens":69956,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":782,"sample_id":"sample:f23d7b28b37befb8169b9b1c22eb930989b5b2f0bc495e2c1ba5144d16453775","semantic_task_id":"a4812b92d49831344823a0cb56c7eb3b6e501de8205fafeda14975639cdaf6cb","split":"train"} +{"domain":"finance","full_message_tokens":73653,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":783,"sample_id":"sample:40ccb27f562467532a9a38ee69721098042a0ecdea1d55bfd1deb22385a58a26","semantic_task_id":"a72e88adc73aa96257773c3b21f70de24c1a22378de8e2d662216c9669a5c660","split":"train"} +{"domain":"finance","full_message_tokens":74734,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":784,"sample_id":"sample:6a45d776f5750994263b54aadbd872b26339037edbdfbad68ea6a2458fbc2c83","semantic_task_id":"a7b93420d0f7a68de84020178bd6e8ac4f9553fb8ee51910dc1018fa8294b7be","split":"train"} +{"domain":"finance","full_message_tokens":70011,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":785,"sample_id":"sample:93f0a729c399cbdd0ecc950134c988dbd3a8156ee214226cc97724762b05d118","semantic_task_id":"a98461d1aa2e8510b7a7a7eb2eda2a809aeb7b2f4d7ae0dad6bcb82be3126226","split":"train"} +{"domain":"finance","full_message_tokens":205429,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":786,"sample_id":"sample:888c1e90232968cc73412fdfbc613deca8c4ccfa7bb01bb0f81bc7b30db5265c","semantic_task_id":"aa9acfd800f58402a0a400c8ef41593c2c0394aec0d422ec86199f6112a2b3e1","split":"train"} +{"domain":"finance","full_message_tokens":74734,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":787,"sample_id":"sample:2f39c443f8aad73acf088ac0f89064adcdc49666fa8f6d4f5ac7cdd3ab89f658","semantic_task_id":"aae73b1f62b4766319fed14fcd52cc53031108fb5cbcb5bc64d28a96638ba070","split":"train"} +{"domain":"finance","full_message_tokens":80806,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":788,"sample_id":"sample:3a051511d482b4d71fcabeaf80cf9614cb6dcf796defd67b58c9b7ba2b0cb634","semantic_task_id":"ade43e2d05a58d2f9fc804ed0972ebfaeba11791ebe4ce8582796f46b5aa0a63","split":"train"} +{"domain":"finance","full_message_tokens":219458,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":789,"sample_id":"sample:48685c7e8421ff4b92a4452b8f9a635d37820c22f1668508f1ba044e4fc5446c","semantic_task_id":"af5f30e7469f0dc4fae0efbbccee80db6905afd2d9915bcb55807ecdc86f649c","split":"train"} +{"domain":"finance","full_message_tokens":80672,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":790,"sample_id":"sample:cb05508f1a73506aff5d38cfc92dc876da33b098efca70110399a92c57e2321e","semantic_task_id":"af868f8cc7333650c2c2ff818389ab2439850f68660e9c1127aea3623853e2e0","split":"train"} +{"domain":"finance","full_message_tokens":219444,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":791,"sample_id":"sample:dece5038fd673ba5ce501a44d3d9c60cbb345208b5d0d01beaba650c9efde6e9","semantic_task_id":"b2d0deaf2dff2e5df70fc61092affbb5d47a67b09808220e86e52d8b1c345b8f","split":"train"} +{"domain":"finance","full_message_tokens":74730,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":792,"sample_id":"sample:9f4e703c224c1da22a19e4897e3ae6873f9fb0f287e17e5e6e6e70774bd70a10","semantic_task_id":"b33ec1e2808820f8666b262b952c0c20dfce7f813a53b15f07b07cb2c3c75a99","split":"train"} +{"domain":"finance","full_message_tokens":219441,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":793,"sample_id":"sample:a92b20147c1b0a644e8c1335cd3e619c50c2b44eb9768396a86b3b21c3c1ca6c","semantic_task_id":"b916b1ae8ef134d5bca182a60f61f3c1e9b102637fb0fee10a4f755221005fae","split":"train"} +{"domain":"finance","full_message_tokens":205450,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":794,"sample_id":"sample:23cb26ff8dd2ce105254896350237af117586f5646bd94fa47ee517b27469ff5","semantic_task_id":"b9bb426532b82b3531baa7077c773a0a3d61035e2b5dec7939c0e3279277a2d8","split":"train"} +{"domain":"finance","full_message_tokens":69956,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":795,"sample_id":"sample:236d9f2a7b32b306163b1dc2b7ca086240ead4af73147e2bfcd75bf6995d587c","semantic_task_id":"bb4b1dc5a687c357252ed98103ca6d35d8a62aa6ae1b079623ce4f9026fbeb70","split":"train"} +{"domain":"finance","full_message_tokens":214438,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":796,"sample_id":"sample:c763a2cd09a5b0d3bde7c87c0027cbd8a0abeee190cabc88b34fa95a87b41856","semantic_task_id":"bcb75e33990a78f1aac103a290cfc4d7436df21ec91d277a34a512ba5e2b28d1","split":"train"} +{"domain":"finance","full_message_tokens":214437,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":797,"sample_id":"sample:ba5ba5bc1118a0e35956b2142debf692ffbb9faefd9b2640d4c8f6b14085d3db","semantic_task_id":"bcc2c6129b4a5d0eb8646e722a2b2726a089d8691ecd2f229636644fe4c8630e","split":"train"} +{"domain":"finance","full_message_tokens":80656,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":798,"sample_id":"sample:22a0b65f2b38060b3b797f099ecb622fde49706ed4608457299c49fb92f72c58","semantic_task_id":"bcd22886cfe6595ea230cd1c582964b5c5e8d3c5fd28425eb45aa63a6ae1dca4","split":"train"} +{"domain":"finance","full_message_tokens":67020,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":799,"sample_id":"sample:93353361588f807f119ab5e6df5a94227fd51e260890756a213d07682cd2a42e","semantic_task_id":"bd4302d85e5e30035294ee3bef17de18feecdb9ad1ac88841fc06c7f29731039","split":"train"} +{"domain":"finance","full_message_tokens":219424,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":800,"sample_id":"sample:2121a6c2cba8484ec73576a1e1254cff4cdbe6e08c6f3b00aee28d2887f3b811","semantic_task_id":"be7263ce0b774daf709f058569fd629f68aa6c08bff3d2b3fb81528a506647ef","split":"train"} +{"domain":"finance","full_message_tokens":80808,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":801,"sample_id":"sample:2b8d230747b2b59d150f6caaebf321bbb7ac2aadec6c4c570d5be4d111a6220e","semantic_task_id":"be9b175fd768d27098617d90e5fae6fe4f88de8a91da7fe6d858323e1cb6202d","split":"train"} +{"domain":"finance","full_message_tokens":68701,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":802,"sample_id":"sample:437fd6158cd559cbcd6e6129ad959b421496315fbad2e138febfc4601de1f128","semantic_task_id":"bff777bed8f8eb158b767cf684dee8f1c81ff63a467440a573359d72efb46a5d","split":"train"} +{"domain":"finance","full_message_tokens":80671,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":803,"sample_id":"sample:22c6fb0df72021891675c171f806d993bbbc83f34019c9fa388bbcdaf84369c0","semantic_task_id":"c07d6f3e9815b603b90b692e4230e7f8b86b6ecefdd83454f902dbeed05d58dd","split":"train"} +{"domain":"finance","full_message_tokens":214459,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":804,"sample_id":"sample:a9ac2e7d94f1d30ce1e887115eb5d1f8ae2c393c38af54bfb18cdae900cf7271","semantic_task_id":"c0d615f7e6ca831ed98f974721977fcbefe04ab79de8eb819da819234fbf0633","split":"train"} +{"domain":"finance","full_message_tokens":68691,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":805,"sample_id":"sample:414ca06e3baa67e2d3e9e6b823584463cc9568231714b81177362c4b88e78ce0","semantic_task_id":"c176d739b956b77e16bfd8d553a5cc68afb13566bf49e39f7abe119843b89e62","split":"train"} +{"domain":"finance","full_message_tokens":74729,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":806,"sample_id":"sample:66b1d7a476d8f53243599fc33bdf70c805688ff0780ecb5ae19830663f95e14a","semantic_task_id":"c1bfbc611b687246fa659b89f938db884559333f83236b9933c533059ccc3218","split":"train"} +{"domain":"finance","full_message_tokens":76077,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":807,"sample_id":"sample:2168f04fe363809b236546ef154dc553659b7475c461a41460cce99a873f61de","semantic_task_id":"c291500f7b5d3a9f7ab7ae9c69521aca810c38fee88906ce6402b2f9064d3c19","split":"train"} +{"domain":"finance","full_message_tokens":76029,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":808,"sample_id":"sample:48d3d35f751b29ca6c11f2b512df788ffb2e73ea388142ae91b2a589ee129095","semantic_task_id":"c41d3a2161dadd4ff6fdeb09028d102c6a43f081e2dafa757969e1baa3746303","split":"train"} +{"domain":"finance","full_message_tokens":74727,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":809,"sample_id":"sample:ae1b50724110ad3d958a7c5704e8e77d19d74b2c313211ba985cfec9d9a24538","semantic_task_id":"c4a7b054a339e6cc8130ee93ed9680a11d93c96e2a73d0f142365bda5a6be5df","split":"train"} +{"domain":"finance","full_message_tokens":74717,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":810,"sample_id":"sample:f3e1585ec588e116c3eb55da4054d5b6befd865ae41c2ccc375af218b92047d3","semantic_task_id":"c4f5f79ad94891de3336dc48d805526f2f7cc30276030936fe20d4fdc9a21c8c","split":"train"} +{"domain":"finance","full_message_tokens":205451,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":811,"sample_id":"sample:102882cf780d106e2bd4d4cb36ce7dde009d16578187de128d987c0fff989255","semantic_task_id":"c57b2b68f8ce7d0e238dae55e9599b85736ea4527ebaf60cd228974cbd6db9bb","split":"train"} +{"domain":"finance","full_message_tokens":68752,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":812,"sample_id":"sample:67449ed123d238433e20995ea9b6c393bd13b8aa91fc3cf0dfbeb222dc8798a8","semantic_task_id":"c57ea42629b56244a5716f0f77ffbeedd90440c48c52163ac3130c39300674cc","split":"train"} +{"domain":"finance","full_message_tokens":70054,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":813,"sample_id":"sample:634bda7c93773588cb4266a91472d8d19174df0fb74fa1fab12aff69726208fc","semantic_task_id":"c66dda63af832fd720f14f54a0a8af5ac5856a28c5e116c9b91ddb922549cb84","split":"train"} +{"domain":"finance","full_message_tokens":67023,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":814,"sample_id":"sample:53a431186eec29e9193aed2ccf7976fd53eeae5f23a3f090f7181c2bc023f35a","semantic_task_id":"c7a1892f6594b626006e114626335b37fe5752f9192b88936bb1b752cfb015aa","split":"train"} +{"domain":"finance","full_message_tokens":214436,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":815,"sample_id":"sample:a7670a4522ded5b8e40dbb9ca0e2b678b05ada046053796b85b6e762e868d4d3","semantic_task_id":"c7f69381b7e99bd801f96d142b446b7372076cfbc457256698c4e800bcdcc135","split":"train"} +{"domain":"finance","full_message_tokens":69954,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":816,"sample_id":"sample:e75da994bf2814d064cdd5924fb0f6f91d1bf57a2733b74381736c462eae7c88","semantic_task_id":"c9f3568304f9754125ff27433e3945d7e938920b3b7cdc4e399389090441d37a","split":"train"} +{"domain":"finance","full_message_tokens":68652,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":817,"sample_id":"sample:712a41fe1b3872e10d8cae4a809d06238cd543073ee8476f55e9fcf89a9ae92e","semantic_task_id":"cb9491afb6ada4f64e1f3d023d9c21ecdba438d86b6e6f87eedbbe4c58dedc59","split":"train"} +{"domain":"finance","full_message_tokens":219454,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":818,"sample_id":"sample:da109b8a5e206a348643c9d35899e853bb24c27eed4ae52e607cd6dd9fcfd9a7","semantic_task_id":"ccf33db5e2d17613b7406913114fc1729e548a30ec6da9d0e88f5d8aacb856e6","split":"train"} +{"domain":"finance","full_message_tokens":68708,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":819,"sample_id":"sample:bfe90dd586f7c8e691f891f6bce93bf7f0eb501f0d09e7eb7b5e09657f2d5bf2","semantic_task_id":"cf0f55f22eb63798fd26eb90f09044a2a59a844fd5d1174e2a5f4f247e8188f9","split":"train"} +{"domain":"finance","full_message_tokens":205430,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":820,"sample_id":"sample:7ce31765b9ba2cb48df9f4b321fb722a8dcd33d70808208711c5ef16a06f218a","semantic_task_id":"d1c01b9c5fe180c61982345f89b7cb50e47f7fdc407925b397315ed3d9d71d1d","split":"train"} +{"domain":"finance","full_message_tokens":74729,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":821,"sample_id":"sample:5df0bc06d7dde38b5bcd3c4a351a578de3975b6c0931d6052cee0d24a81e8a45","semantic_task_id":"d24a8410580d0ca5e913835376173c7e4d1b98218d59706b575b97b07e3431d7","split":"train"} +{"domain":"finance","full_message_tokens":219422,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":822,"sample_id":"sample:1682436f459c8abaefd911dda118b37e8754a02daed3fdc37214f0146e6e33b6","semantic_task_id":"d3d5c2b5ff7e16332f4da8e7f1928c17bf7eba7ac1fd8ce391bace55df706c51","split":"train"} +{"domain":"finance","full_message_tokens":80807,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":823,"sample_id":"sample:54745f41618da1f417ee1d4054ff45e90fbf1b30a019b28cbd9638078ac9999b","semantic_task_id":"d45e4bc042cd76f01ca85956dd5ea9bd4eb3d02f169ae602a8171bb960e15f1c","split":"train"} +{"domain":"finance","full_message_tokens":81036,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":824,"sample_id":"sample:6b7f3af54795cc4d7d010d9e29684da02ee214066d209a2bdb5dc57dee210f5a","semantic_task_id":"d58165735792dd8ea264f08388d5d0f4d9445fc9914c67608c63204677e3bb2e","split":"train"} +{"domain":"finance","full_message_tokens":205451,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":825,"sample_id":"sample:cbd075000f6cd67298176ecfeec0167506bbdbc5099ef788d8b1a3cf1721d9e3","semantic_task_id":"d899296480d55acb4939dd2038a662212724be7f67d803f446586d5d9ecd25a8","split":"train"} +{"domain":"finance","full_message_tokens":214449,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":826,"sample_id":"sample:67f5bc073903ec6d845611703f59ba319121769e14ab2250188895dc9bf7f4fd","semantic_task_id":"d98e0948182a00ecf34768256e258c87326cda88e85cc48d66be739d00749f74","split":"train"} +{"domain":"finance","full_message_tokens":205429,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":827,"sample_id":"sample:3ef56c5a9c78f9c7dc6d844df293c682fbd29f6bfaf32305a22453474f754374","semantic_task_id":"da141a1deaf7f277947a37fabb17df6cfd7fe31d54e419691e0cc10bb10dd273","split":"train"} +{"domain":"finance","full_message_tokens":80808,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":828,"sample_id":"sample:c9bc05afddeddff5e044848b806f1c9372e25b1cc1f148348c3240e8fe840829","semantic_task_id":"dc5071d44e1c00a5c1c832faacd1322695b792fc4111584b2c8be6a42090f5f4","split":"train"} +{"domain":"finance","full_message_tokens":67021,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":829,"sample_id":"sample:4f77a05bad90e8d5bd41134f263a80c29604794eb96f0a0b92bb1712b817b3f4","semantic_task_id":"dcc5deb4ab60c1dff2a8f10ba010aea9ff0efac07b66b22b000a7eb7b86bda94","split":"train"} +{"domain":"finance","full_message_tokens":205465,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":830,"sample_id":"sample:4fb0296a98d86d8d610b3df089832eec10a2cc2d6d35a89539863c2b1a8b199b","semantic_task_id":"deb81f0c3cbb1c2198f2acb448204cbe7db33719bbdba38423d08afe8d8b009c","split":"train"} +{"domain":"finance","full_message_tokens":74729,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":831,"sample_id":"sample:b3100f5aaac6140cbb6f7776324321c5db8965d857e759a19280e2533e1f8460","semantic_task_id":"debab1cbeb2f410eda80a488378add0cfc4717897f3375a2fb5c2ee24a35301b","split":"train"} +{"domain":"finance","full_message_tokens":74734,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":832,"sample_id":"sample:0324521b962a294a73a471b609fad3a7c72408048f6ac657685143ec410af3d9","semantic_task_id":"e1792d267a3f44543b839e20def5f501b3b9708ceabe392dfa295880d8ebe881","split":"train"} +{"domain":"finance","full_message_tokens":73660,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":833,"sample_id":"sample:fd7b340bbd695d7728f9a7e705dd100feb1511842c4ee2df47a9011a3481a7b6","semantic_task_id":"e1e071f2995cd53c258908ebbf751768ccccc3a2d34e9e0377c2f37aadc91c5f","split":"train"} +{"domain":"finance","full_message_tokens":219423,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":834,"sample_id":"sample:e41f39e150e60818773079aae865a013cc4a98069fbf39fae308f2bb3fe37a41","semantic_task_id":"e2a97b9f550752c98abcd5fb9b426908e563d5b234e77f048f45170e8eaa1f1a","split":"train"} +{"domain":"finance","full_message_tokens":80806,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":835,"sample_id":"sample:0197c67be5aadb27b6b976980dc5b1f469ffb5eb7949bb70788a8ea3866fb80d","semantic_task_id":"e3bd5d00eb5d05597c0f65cae6e5802035812a572dfe2dada9ecc743187f9f1f","split":"train"} +{"domain":"finance","full_message_tokens":205428,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":836,"sample_id":"sample:4b856052e2a63796a7ad4b321bb62018ff05c8d1e8a095e3d79dc437bdec7d4d","semantic_task_id":"e4211f20b1618a9658ebbc1f8de5bebf9e0882039c41bd4f0af27fa3bf5721a7","split":"train"} +{"domain":"finance","full_message_tokens":80668,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":837,"sample_id":"sample:2af6e33a587f18b0061d515b036432604b074aa3da8b32265257c308b8f306e0","semantic_task_id":"e43cb836561d8ff7445e8eb457711bdb025ef515b6b4ee33c095f57e3972aa7b","split":"train"} +{"domain":"finance","full_message_tokens":74729,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":838,"sample_id":"sample:3476a31f64c2914ffa61bfaffb78037cbc2b31ac99b9fedf952277ab825238f5","semantic_task_id":"e456984e2c8fc0d6133982bac5c59264eb3259d952259008f515afab2606ad70","split":"train"} +{"domain":"finance","full_message_tokens":214437,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":839,"sample_id":"sample:3b0abb7d5b0e71ad0500a9e72dcd4b43acb43d1912cd87ca729a3c87e2a29599","semantic_task_id":"e48504e158314c80472d4e03ca4d56366f1b1ab2c943a97e41dfe4138a24a3b4","split":"train"} +{"domain":"finance","full_message_tokens":76087,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":840,"sample_id":"sample:b886fdfc6ba04535df892286ef4fbbb42d8d6a045ee22f42e7751e1b510dec71","semantic_task_id":"e7a7da7ed08d6524704666bc0440e64e082252b6a2f5c2c7b9efeb3b59a9baf4","split":"train"} +{"domain":"finance","full_message_tokens":76071,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":841,"sample_id":"sample:d2be1fa4ee1f9f662992c6381eece453cbd029f1ad682cfe2cb4566855f05f2a","semantic_task_id":"eb9cc7b8dd1bf8c9eb3ee501d657461d1b436795a02555620aed07ebb8cb229e","split":"train"} +{"domain":"finance","full_message_tokens":219422,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":842,"sample_id":"sample:80291c21ed2ff48f5a7255786fc99da274348ad87fa59a598792341e09ade08c","semantic_task_id":"edc755ab992dd4f3f0c3242c1ba79915dece685ad217e8e2e55cab2c009e8a6b","split":"train"} +{"domain":"finance","full_message_tokens":81038,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":843,"sample_id":"sample:0aa2f3c37a7aaaa2bdf29cc1e93ecca05983d98b9be718816b5691ed094578df","semantic_task_id":"f0d43c1924732e565aea075554c930ac0fe3a0b9089071a7fbea87d5a4f8be30","split":"train"} +{"domain":"finance","full_message_tokens":67022,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":844,"sample_id":"sample:901b322ca726b37a17b7f68a69bd5c9df084ce2786d81566da957fe982cd892b","semantic_task_id":"f2f4e6c64d96244df0e4a176c658f2aeb42ee3cd58846c3da940e41528a889f1","split":"train"} +{"domain":"finance","full_message_tokens":67021,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":845,"sample_id":"sample:ac921c96f973e2d3cf5960962662f179097e7f1a86f66d756ac7d869d986c097","semantic_task_id":"f33f253655f23ab00bcb47bde78384284417620631c6ce0970d717c24a8581ae","split":"train"} +{"domain":"finance","full_message_tokens":74964,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":846,"sample_id":"sample:ef8e72b2ab80ba86075a78bd54d499fea62a18ead1fb1092bf7245464e18b3d0","semantic_task_id":"f351dc18a68eef05bee72bc92aa9c0ee96ae82e562e49c82c8e9fdbc5d487bd7","split":"train"} +{"domain":"finance","full_message_tokens":205429,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":847,"sample_id":"sample:8bd95478160556eec6ad0b29acf98b325070277edbe633116da94320e452d838","semantic_task_id":"f3c67de1f77c520f6c7c2865695e89c35f3cc594204edee2e8c4c5f4d1fbecb0","split":"train"} +{"domain":"finance","full_message_tokens":81037,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":848,"sample_id":"sample:b181b7815a0e84aa4a669206e73a88ff48866a2ec992db27e54725e8e7593b9b","semantic_task_id":"f41df14fc7bd1d9a27b72b352c2b6dc94609e3c276f99567c3d1feb6f9aa8719","split":"train"} +{"domain":"finance","full_message_tokens":205451,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":849,"sample_id":"sample:018ee2581e1a0db76681ea122c1996796c55ba4d752e296c95284f96315517c6","semantic_task_id":"f43a41e5ec5839ad1e95e815b58e0310a35e84a45d7240e5e7125fa2f1bfc43d","split":"train"} +{"domain":"finance","full_message_tokens":80666,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":850,"sample_id":"sample:5da5cea936dec3bac0e5e0cfa45b1ea3fc67060795414866349c9c39e78b7c67","semantic_task_id":"f47833d8e4f08a2d82c2240f7fa5850870973ba2eaf5d90ed041a44b8bcbec4e","split":"train"} +{"domain":"finance","full_message_tokens":73656,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":851,"sample_id":"sample:542a1450fbbe3c043f3a8b8a073c6c10588a1642191adf668a5b6645fb8f3d89","semantic_task_id":"f4a92a275f0fc707131e2127e63487de6edc5ae22c44767006b2ae2723ef3264","split":"train"} +{"domain":"finance","full_message_tokens":205429,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":852,"sample_id":"sample:6728f0d48d618c5dea36997c46f1f69fbab0f3d0d75d4d1476fbbad1bb804fd3","semantic_task_id":"f515ce98e992708ff4bd9be62d7a25f009c42bb7b57d6bf0d324ca55eab64671","split":"train"} +{"domain":"finance","full_message_tokens":74732,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":853,"sample_id":"sample:1f4c36747d69956099c7e2943b811a75a67b3ac75f4d9e4f6ce70cfa0cf66bd0","semantic_task_id":"f6951f1b1973003cc12237b3c84d548ea6253b06238ba5da17690084e15f8258","split":"train"} +{"domain":"finance","full_message_tokens":69956,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":854,"sample_id":"sample:bb3cbc35c73007aff1dcc4c4759b088c32b97925d0e6aa5bcd99ac8c5e814855","semantic_task_id":"f7c7f9385a9e7b5840ef5e5ad1c210e8dee634b0ee7151cea994ebf37a8b6e64","split":"train"} +{"domain":"finance","full_message_tokens":80657,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":855,"sample_id":"sample:0e0ad32d9b01adc833c7348dd60a07517647e4680f0ad468c8fc4404c925ae2d","semantic_task_id":"fb4821b29a00f63f9e9b8dcd9112c043664a1d5f2a8c8a5e907102b23c0db2cd","split":"train"} +{"domain":"finance","full_message_tokens":76029,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":856,"sample_id":"sample:952bac099843be696f8343b2a2c998202c66961c4ccd25406ed2e50a6372166f","semantic_task_id":"fb604fed8b05d0eff2c7510960637520cc484017b5de027ed192d3b03d312f0c","split":"train"} +{"domain":"finance","full_message_tokens":219456,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":857,"sample_id":"sample:82005181edfaf0940d4d4a5ba341ed1a6966502daafeda8f0460c660fcd9fa9e","semantic_task_id":"fba5055901e7957d602ce1a6e6e2192f10fc416f523d79298af056b85cb20399","split":"train"} +{"domain":"finance","full_message_tokens":68653,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":858,"sample_id":"sample:dd6b225c82b2d3210dfb992224b043f2358109fe9c90227ff349ba5d8ba1a0a6","semantic_task_id":"fc7e761e2cb4044ecad49cd5037977a417e0170946bca3b607f1b01eca2e49f1","split":"train"} +{"domain":"finance","full_message_tokens":76130,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":859,"sample_id":"sample:f3dee2c236c59498f89fb4e5a530a7c699dc6e173b38dbb2a9dc44c3898290fb","semantic_task_id":"fc92247e9f5e41c7aa263b599f01f0fa1004361fbd47b41f4f2ee4afddb19469","split":"train"} +{"domain":"finance","full_message_tokens":219423,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":860,"sample_id":"sample:67f22cd5c8cc243dab4d3f9895aad20a7322f8e489c7caebfd7a34af5401c0eb","semantic_task_id":"fd2f1c10c243aba592c2e5f6d249dc021d5b1cfb071b53ba1ef5c3580fb278b0","split":"train"} +{"domain":"finance","full_message_tokens":68650,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":861,"sample_id":"sample:f100058e2a6983bc063ff3c61c8f0a0b72b8dc081f98918578992a8dbcfb5502","semantic_task_id":"fdd128051034ef70d07b3d5dedf484fb89acad627d8d6364d43a96bc478c6667","split":"train"} +{"domain":"finance","full_message_tokens":219432,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":862,"sample_id":"sample:6ef022433ca1b83af3ab9e0ba2034228451d611e6c2bb4ac7fe6df927b0c21d7","semantic_task_id":"ff41551df2e3c20b48e8a02a30bcbbf63716a3cbed50afc86a3ca8dd84a6a9ab","split":"train"} +{"domain":"finance","full_message_tokens":81038,"group_id":"0001326801","output_file":"finance_train.jsonl","row_index":863,"sample_id":"sample:a666129287b7febe78d096444208775635cb401758b4b3fe95de1c5c0c207aba","semantic_task_id":"ffcb08acbe3f0147b6b7f8ae0a5d638828ba23df488cea68768393508c787857","split":"train"} +{"domain":"finance","full_message_tokens":221048,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":864,"sample_id":"sample:017e2ed1161c7f79fc24035a51b860e13e41f069b9cb7d31175639e94ab3d90c","semantic_task_id":"003533bcc43582664fa8a891ca47a0f7a6ee52717a8e57c506dcc9f2d475d480","split":"train"} +{"domain":"finance","full_message_tokens":68089,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":865,"sample_id":"sample:27d903570382f0f1cd05f41f9397d784b08f9beeeece7c8297c28798ea0ba657","semantic_task_id":"034e661ae38981cbe353a720be770265fbef3668ed6be87499d7c25e191ea30f","split":"train"} +{"domain":"finance","full_message_tokens":208918,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":866,"sample_id":"sample:f45d8fac18f4e52282c74422572040db3f9f9708881ad401828c6cc1b600b730","semantic_task_id":"070868ab97bf76086c2a1be21c8a6873a4061b56bc828fc35af5d393d4f10f74","split":"train"} +{"domain":"finance","full_message_tokens":80112,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":867,"sample_id":"sample:3d1a3730c17cae99ddc197ace90e5f8054688d242cf898645c93be5e5d8ab848","semantic_task_id":"0756f1e9128e3cfec0d882e56cb801aa5d705406b6cc01a0df8e451e624e53db","split":"train"} +{"domain":"finance","full_message_tokens":75225,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":868,"sample_id":"sample:489da480ce4f9e6933bb2a79df615d80ef5f754fd5e3d134bd62e41d7a33c510","semantic_task_id":"07c19e261e73bb6f7c10cccd865a18a065aa64365c65358b73616fd7efa5fbb3","split":"train"} +{"domain":"finance","full_message_tokens":80182,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":869,"sample_id":"sample:d2ecd2f27d4c7f7e591b352025f76642c1b677dd5ef04692b95b1835276296fe","semantic_task_id":"0b01441b204adc2ce2aec2aa42370b03635fd717dc10983593f4a2e6e59ad8c3","split":"train"} +{"domain":"finance","full_message_tokens":215607,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":870,"sample_id":"sample:be0a528b97a4811299279e2813c42476455db816a70a14b84eceda17d45a30e7","semantic_task_id":"0c260a73528be9a5e42e1c48608c69f8ac33c4ff5b2a14c459b80f12ef9d65c5","split":"train"} +{"domain":"finance","full_message_tokens":80215,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":871,"sample_id":"sample:14992e803a0d89966dd23058aa69a828fb634fec7039f4815f193b2aebd82e27","semantic_task_id":"0d29e3b693d48048ae9ef73035e4474e00f882d927fd4f757fe9129faed60c9e","split":"train"} +{"domain":"finance","full_message_tokens":72540,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":872,"sample_id":"sample:f85a0d74c7dfec874de31beb8eed9db866768053c3983138a9533d64d8189fec","semantic_task_id":"0d7ae0783088fdaefc615ac211ab80a37921136d6b15efef84a600f2d5aa235e","split":"train"} +{"domain":"finance","full_message_tokens":68090,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":873,"sample_id":"sample:b82e8d93510e93385254cd7a0c19c05ce68aade9101abd1b3146a043dfb9ddea","semantic_task_id":"11887f895977faa4632f60be831d7cb5f1113fc1017cc6f2e7933369bf4fba51","split":"train"} +{"domain":"finance","full_message_tokens":215662,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":874,"sample_id":"sample:0537ecd19454186ddcdb3f66ba59e740bbd82c73c529487223a8b08a4e353fad","semantic_task_id":"12a0399888f47972737003e1b4c27a199b8e61c705fd18765bad0cb8bdc7b1c4","split":"train"} +{"domain":"finance","full_message_tokens":221029,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":875,"sample_id":"sample:48a9bb7768a9a3cff95c9a2dd31d86ae763bc6752b80d8857c04433cae1802e9","semantic_task_id":"1388879535830b57d14230b07daf3c937487d8c168d255a20f142deb0097474d","split":"train"} +{"domain":"finance","full_message_tokens":68516,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":876,"sample_id":"sample:76fe67a7bbf40ff633a58be6b4b3086f5259ac0ee08b02793bd2a839e11811b8","semantic_task_id":"13f69c44ee55c671ed498605b6dfeb285afcd648e6d0ee934f15659287b9288a","split":"train"} +{"domain":"finance","full_message_tokens":77472,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":877,"sample_id":"sample:a88954c68622db8577be2cf2bd0ca5e468a78898399feb7484c63580d119367c","semantic_task_id":"146de6142dc87214fe21223156cd33f25b8bd54cd5218da4dc91f5ed323ecda3","split":"train"} +{"domain":"finance","full_message_tokens":208918,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":878,"sample_id":"sample:fbd72abe69111124e5b3b14cb7229f6c66d98588750d8d2648f97d915e2c11b5","semantic_task_id":"148e4bf2547312ed8fab403b011873e93575979edeb7a70b4b3193cd882653c8","split":"train"} +{"domain":"finance","full_message_tokens":80112,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":879,"sample_id":"sample:ed989d27b2ffe9827b6d7a3ee7b88496102f2bb2bdbdf23e2a356b29c31aa984","semantic_task_id":"165e782cf81d2620232777f000085394a29801f9d18c4013c72c2b2ae448ca8a","split":"train"} +{"domain":"finance","full_message_tokens":68089,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":880,"sample_id":"sample:5f2bfc80c33a90c779e551123788779b22c05e754262eb3fcc9a02ac110f7448","semantic_task_id":"1855c779f633f1aeea25759f6aebf5c206376c8ea33f1b724a418d59b4f41de1","split":"train"} +{"domain":"finance","full_message_tokens":221016,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":881,"sample_id":"sample:a2171653395af8e08df6c6800880caa9e36fcbdb4dddfce45cc58c37fcf16c4b","semantic_task_id":"193a91f023c6b65de4dfd193d2cf7c211471908ed9432b58f24af2f43c59aec8","split":"train"} +{"domain":"finance","full_message_tokens":80221,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":882,"sample_id":"sample:fb4b52f351a080e027117e5f6d494e6a765d97c462fcf7aee6d783362744dc0b","semantic_task_id":"19d7061e8ce6e11325e48e12fb466665b49754627d4b6c767651d3d266b5f3f3","split":"train"} +{"domain":"finance","full_message_tokens":215628,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":883,"sample_id":"sample:5dd2d42980571c26949cf43493d92e4f013a52a079dd10f32b3f9fea10a1cd70","semantic_task_id":"1aa4fb420e030264cb73f7e331f8ca21bc7c8cba466ac40efa41b58e8247bb98","split":"train"} +{"domain":"finance","full_message_tokens":77486,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":884,"sample_id":"sample:c55c784bf1c2612643c4b9d8665a2858a3a92f955600807540d2f65dff225415","semantic_task_id":"1b74895c6d6aa2d81e51458567e5a756239bc1078494b9e04916d45474536c77","split":"train"} +{"domain":"finance","full_message_tokens":215607,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":885,"sample_id":"sample:4719d6bfcc106487f0c8ff7927ba9881da0646f5eacf91adb2d36027371cb4d6","semantic_task_id":"1b9623c7f70d4d08dad93fbd6c6a0e0248ae27ea577ba3555c83285055a2a7c5","split":"train"} +{"domain":"finance","full_message_tokens":68088,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":886,"sample_id":"sample:4312b65709dc598bfe2cfc4124ad8ab273266aae3289ba2a160cd097b139be37","semantic_task_id":"1baf6891a32c588df9603510574fc88bd28ea66946bba8d58868c51901bb0767","split":"train"} +{"domain":"finance","full_message_tokens":68137,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":887,"sample_id":"sample:b6dd8c262d65147ddb07bfb542b08c28a13c6521e27801bea4abde8b120c9f8d","semantic_task_id":"1deed7b2865f415f62b4c360499c12da8b915adbc785691fb8e93120f39652e2","split":"train"} +{"domain":"finance","full_message_tokens":215616,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":888,"sample_id":"sample:ef4b2fb1974ce2fe1571a32f875ab14cae9759ea005e0343f1c1177c40b11c61","semantic_task_id":"1f9623287ee98dd2777b39a9f8bf66b5c3f9d6d601cd9bf0464f8c8ff99fabbe","split":"train"} +{"domain":"finance","full_message_tokens":77442,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":889,"sample_id":"sample:c5909c2a1e720e20b95dd8c2efc4cc6b58f79667abd28bdf14d2f77116d32020","semantic_task_id":"20b898098b6059ac8f015231b90a44a76509d11e49682a316f70e506111b64ec","split":"train"} +{"domain":"finance","full_message_tokens":80181,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":890,"sample_id":"sample:5689883b27f7b03821614dc2a04cf925cc5e0d87e798d6af03afeb362a8c1ba8","semantic_task_id":"21e025a7b476a10ce701009cc9e673f81c2832bd01ae780e90fd0e2b74b9eb27","split":"train"} +{"domain":"finance","full_message_tokens":80216,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":891,"sample_id":"sample:99353200de15b55f60a3f3f9cfd2b42fd898b207851afedfa6d32be6f76a6c4a","semantic_task_id":"235969fca7a339a6fd55e01eaab8a232a5b5d134b461d557ae023c95a6bfb5f5","split":"train"} +{"domain":"finance","full_message_tokens":68089,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":892,"sample_id":"sample:e5051050a394b1c40d3f40eb23cc84ffe6058139182f6d5cfae7f78a47e58c61","semantic_task_id":"23a81f033fcda09d830daaac29079bbd04b7d6b1d2ad2c8fef0be02ae88f0159","split":"train"} +{"domain":"finance","full_message_tokens":215626,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":893,"sample_id":"sample:3bce8d37c3ebe10a29c129a795621c85f5029de46c5453d3dfd9ed8bac9e5ea3","semantic_task_id":"247b42c38076bc191085f5115037e6e64e92fff375c7acc1e6034543e0c72217","split":"train"} +{"domain":"finance","full_message_tokens":221008,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":894,"sample_id":"sample:d62d76caf539eda0d735a0a3a3922703961590ffe2cce386b2c1ee07e086bed2","semantic_task_id":"254eacdfc11f9cb77739ab12ae291118b114e175eba9bd110a8eddbae88af0d7","split":"train"} +{"domain":"finance","full_message_tokens":68088,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":895,"sample_id":"sample:4c17568e0ac8c57d332f6c3253f942854bfc43d83970e121e38fca9ff5e972fb","semantic_task_id":"271dd30683c0a36201f43c73c24cbed4b40eb37cd4e18305f5f4f2aec4dbf833","split":"train"} +{"domain":"finance","full_message_tokens":77448,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":896,"sample_id":"sample:605a2961ec03b923975093cc4028c622508e75209d9e9da3a7019a87355e2512","semantic_task_id":"2774d94fc76e672ec6537d6e1527c51e49214b00d8855ec3ffa3f405809ff1c4","split":"train"} +{"domain":"finance","full_message_tokens":77448,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":897,"sample_id":"sample:855fbd27c308babe5f547e0144ea13f16568905446fbcb244ef8b12e440e81e0","semantic_task_id":"299eb750cb701c27cb4f3af20bd5a5a05ef11a0c1844b8d8642145141f625131","split":"train"} +{"domain":"finance","full_message_tokens":73512,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":898,"sample_id":"sample:62d0efd6fb50c5d17b51345f95f0fb01ea45fff4f8ee2d733fd0138367da22d7","semantic_task_id":"2a0b17e17e1e9b01aaeb14c9fbd0329c8635430a2491d23c114f664acc87049e","split":"train"} +{"domain":"finance","full_message_tokens":68516,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":899,"sample_id":"sample:a7ff06304dc8f4ba3e7c34f55de9d69845a9ca658659e7e2d34a73231b0513c8","semantic_task_id":"2a34e854c43143f87a724d15296b038d55a51cd669128adad6906165404e2e8e","split":"train"} +{"domain":"finance","full_message_tokens":221006,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":900,"sample_id":"sample:9fabf4b5763ef3fe67d2797f0d1a4eb2eaf1a3491a30018643486273e5ab8e2d","semantic_task_id":"2a492e44cef98448c1d37f22d586a1d35b720b80dfd4fcef2bf492c99af4ed3d","split":"train"} +{"domain":"finance","full_message_tokens":80182,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":901,"sample_id":"sample:3ecee80f4bccbc3bce16e17779873e563a5bcd344f4b747a97c701893e36197f","semantic_task_id":"2a7979623b0983034769e33825e1f54379e2bee25a9e8864301318833a4e3829","split":"train"} +{"domain":"finance","full_message_tokens":75227,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":902,"sample_id":"sample:e01c6d44b228db5cd1e7bf1af85ef20e44edbaf79b094e937fa94c1af2b23483","semantic_task_id":"2b0a4738ae36df27ebd13857f8f7c06bf69f2858a970b1b1684318a7a36c646c","split":"train"} +{"domain":"finance","full_message_tokens":80110,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":903,"sample_id":"sample:0148c1c7666fcf6eebc47598ab53ec46de8f03c89047ba19c3e27ecffa54730b","semantic_task_id":"2b49003f742307854a34576cfb83750f2efbfc6051f988bde3428cc273aa3733","split":"train"} +{"domain":"finance","full_message_tokens":77485,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":904,"sample_id":"sample:12e965391a003d351f0104d5f680c8a6f350245c26bf75d329dcb90ff6891213","semantic_task_id":"2bee2c526b758a1488a59b539ad122a8630fb6375da822cde1111aad7a864aeb","split":"train"} +{"domain":"finance","full_message_tokens":72493,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":905,"sample_id":"sample:94240f8c3fc6d9d8fd724c12904490e4125376edb4010d3ae3e277534b675aa3","semantic_task_id":"2c4970dc36bc3e246a4f27cef380f41dfd5a010b32f1994d1235067829c3e9c8","split":"train"} +{"domain":"finance","full_message_tokens":73510,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":906,"sample_id":"sample:1badd68abfeb411d899a7a0b325e3fa53b5153da0b14ed1434ba0420a029f97b","semantic_task_id":"3043ddc59eca5be6ccb60cc3f7f1bf4dfc8b008e95bf4da851cf32742b1e6d09","split":"train"} +{"domain":"finance","full_message_tokens":77449,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":907,"sample_id":"sample:4aac16cf5490b920999e3c6899f6cb0cf89cecd866ca6aee123c4b09c2da6ea2","semantic_task_id":"32df5b611b3861a0a47b72621881d584487fcd5e36514b493fd178fbf1b88a99","split":"train"} +{"domain":"finance","full_message_tokens":221006,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":908,"sample_id":"sample:b0975b98a53c1563892ac6acb924f293b78d5feb9bf99120b06ac571a173b4da","semantic_task_id":"332b99127fefc9845db27dc877d8176bf4eced5f5bb9fe788a8df6fd2e5e0f8e","split":"train"} +{"domain":"finance","full_message_tokens":221008,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":909,"sample_id":"sample:cb7f25cd66eb7ac306e215f79e566578b09bc3d10a3ed5c3b7d05269165f9f72","semantic_task_id":"3370bf28a340a55185f4b511564c6befea405993e5ef5f62c052a4a1f3d1c483","split":"train"} +{"domain":"finance","full_message_tokens":208897,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":910,"sample_id":"sample:cf4b212335334832788e98c036bfeebcc71adaebe5bc5ab3db3915f885d2b005","semantic_task_id":"355dcd530bc6247f601f0800e2f52ab81b52611558704f75b9e87f16aba2bd57","split":"train"} +{"domain":"finance","full_message_tokens":77483,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":911,"sample_id":"sample:f6cd547a60d93614f1b0de02e7ff4bbe660808bc41a680ae53a4b5174f56ecb2","semantic_task_id":"36f01cb282fe37fde2cf5076f584c2f57030306037ca8619b23edbbc2e529e6e","split":"train"} +{"domain":"finance","full_message_tokens":208898,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":912,"sample_id":"sample:df9aa3d5dff41073ced74ed555202e73bf8e1cbc2e1e151e068773448736936c","semantic_task_id":"3725e9934a4a8d6cf3c7d11fdf5d1ed284ec4d6455453d18a968d45ad1154532","split":"train"} +{"domain":"finance","full_message_tokens":77443,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":913,"sample_id":"sample:51989ca701e02ddf7e3dbe125a1b9f651d313655a6cb57033be6c4d58bbd08d3","semantic_task_id":"372e88e82068d5c369c3b5beaf81e753df4881db486e1fdc644eae0c50016793","split":"train"} +{"domain":"finance","full_message_tokens":77481,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":914,"sample_id":"sample:b6c0510d67aeee851f590aa2daab6aadcf5c3296eaa9305d49d62d1e8f6664f5","semantic_task_id":"37392bbda14860302afd994cb3080c4816e05b896faea2001d75dd29ed6af461","split":"train"} +{"domain":"finance","full_message_tokens":68516,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":915,"sample_id":"sample:eff6c41af523478826bb6a897848ec8f3a595d70dd699ac473e5bb98f7d156f2","semantic_task_id":"384b71c775cf6ef6f00cdfb158620809171efd7a12954dbb271c06a66bfb3300","split":"train"} +{"domain":"finance","full_message_tokens":80220,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":916,"sample_id":"sample:8613294a59c514ae36e7eb949066f19044cf5a64c6055aa4c8a4868e4544da7f","semantic_task_id":"38a53234300f53709a634a399fe43e30708bc7ff383019826ba8e2701d21cd87","split":"train"} +{"domain":"finance","full_message_tokens":77487,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":917,"sample_id":"sample:4bc6bdbf18be6b140784d0054cc3117af4fa60b953bb2890c625e753480b8942","semantic_task_id":"394a8b77e2ee16ffe86f1a88bd5e5102ba0d1dc9f76d12bf2ff1987df6dd6aa5","split":"train"} +{"domain":"finance","full_message_tokens":221017,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":918,"sample_id":"sample:e0b7dfcba384c9ce6041d2612502e60b9b41f41feaf5630f6a5c638e24894c6f","semantic_task_id":"3c3bb3bbd5d8fb0f4275fc5d176fc1cc8903b48f7c52c21949c7d75945cc3a3e","split":"train"} +{"domain":"finance","full_message_tokens":221023,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":919,"sample_id":"sample:abc5dbc0fc175e80e3287418e87f9ebd02e6495f2c1133f6edf137d854d8cd3e","semantic_task_id":"3df183f454d12e7a1ecc5b400092be61b58065af96d53a111ce5920f246ce754","split":"train"} +{"domain":"finance","full_message_tokens":77442,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":920,"sample_id":"sample:4a9358e267653eef5a7a7bb50dd45ab23815a218dce28165fc4a90b36d89c6a7","semantic_task_id":"401db19e3e5822f8530a4ac9a66bc1e0543f6006cde163caa25c3a0035077568","split":"train"} +{"domain":"finance","full_message_tokens":77433,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":921,"sample_id":"sample:8a9e710f4d854742d69ebb754418af32ea1450a08d5602f71692275fdca1c0c8","semantic_task_id":"411ab737d8d799bd175fb0fe2f1b3455f24153788efbf89490b79a90903937b5","split":"train"} +{"domain":"finance","full_message_tokens":221040,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":922,"sample_id":"sample:61ce82927c74741d27ab32230e01149e7360cde253370467b74f861b84f325ad","semantic_task_id":"4404cdcb278c376a1fd3236c5d194f435cd775382f4f7356055d94e564eaac6d","split":"train"} +{"domain":"finance","full_message_tokens":221040,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":923,"sample_id":"sample:95d672ec6e4f9094bf31821ddf1531fb9e382940f52501535dbc2c114a5775e2","semantic_task_id":"466a6fec724b9719347e0156ffe9269e18f4d5340ebb98a6c92b768bd98183ab","split":"train"} +{"domain":"finance","full_message_tokens":208899,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":924,"sample_id":"sample:5f4f2f4f12098f5fb10b525c97e1871e550830634be8e7fabbda91081f5e01da","semantic_task_id":"47af80bd046da23d38daacdad42bdf769bfa4b5deef63e2adb6920855704333d","split":"train"} +{"domain":"finance","full_message_tokens":72529,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":925,"sample_id":"sample:bcaea54d891a080cb5a16ecc4baa4816fa396ea7339540a394bdf0d920052c91","semantic_task_id":"48189c2419b00b756abef6f36a1de336d396b1076479dcfebf7ce1ecd842b4c0","split":"train"} +{"domain":"finance","full_message_tokens":221005,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":926,"sample_id":"sample:ee937e26b844b708cd5d0987aa4a839ccae5487f88165e48aeab754837591bf2","semantic_task_id":"4888fe67083f3e3c0b6290f2277105a5054cdfda67e0884793e395aa56ae6d47","split":"train"} +{"domain":"finance","full_message_tokens":221039,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":927,"sample_id":"sample:1ea4970d5ddc499a8f28e78e66628889ecd9a1a5f53e41edbe47e774463537fc","semantic_task_id":"4ad86cf691f41e26d1f88816c05ed5723f509a53318672fc62610210dcb2261a","split":"train"} +{"domain":"finance","full_message_tokens":77432,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":928,"sample_id":"sample:83224888e03615ef203c5e70c8d6a03bf8130ae6f32531fc9fd3cb7babd374a4","semantic_task_id":"4c4507555db6caa95ffa054e33f3369c019902acf507e3cad41ebad0e3fe39f5","split":"train"} +{"domain":"finance","full_message_tokens":68516,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":929,"sample_id":"sample:76b46e9b5ff60502ee6fc3bf966d27cb228de3888138ad6f056b097d3a93bde1","semantic_task_id":"4c68707917cd36e3dcd4b08781385fc4f01715463bffaf6667b989f1d8bf91b1","split":"train"} +{"domain":"finance","full_message_tokens":77432,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":930,"sample_id":"sample:98291b7e85b7f236569630e6db778399de3f50bcafb3559e44bf6465a1d49f5e","semantic_task_id":"4ca143437b6a174c02c492197104748031a08e45c22424c663b7e28c83ef58e3","split":"train"} +{"domain":"finance","full_message_tokens":68518,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":931,"sample_id":"sample:3f5d784dbdef88b56abbab4da4bc5adf56488940d5e8086fb5f83c343e69ab62","semantic_task_id":"4dbbe60687aa8745eba6be0bfb775caabf33caa6e6f1c65ebde69facca9ba689","split":"train"} +{"domain":"finance","full_message_tokens":221043,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":932,"sample_id":"sample:7c4a9f0d13a629aeb4caf925e9ebdf65444799532ca7703fd78ebe368fc0e78b","semantic_task_id":"4e442e223657825995a513becd6b989c2100e9defae7c467de6b834ae0e5cfca","split":"train"} +{"domain":"finance","full_message_tokens":215618,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":933,"sample_id":"sample:5ce5f7d90a03a3a932e0a96359274b4bdc26e3cc0fe2c46a8ff3b4f302516b84","semantic_task_id":"4f372e3449e84506530f92ab915ad932810950628d1a167e74459fecfe064b3d","split":"train"} +{"domain":"finance","full_message_tokens":221039,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":934,"sample_id":"sample:5762b1f5657eb105ccb74575a865eb65eb077f1f5c22e25f9599847f72cbedfd","semantic_task_id":"536d5e07da18a17211020a7b43948b15e13bbf036d29733da6e18e27dffe3355","split":"train"} +{"domain":"finance","full_message_tokens":80182,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":935,"sample_id":"sample:315c092faff89d742bc54c3d5cd36ed911eb9bf304488f212be2745c4efb8a8f","semantic_task_id":"5770c57737156e345ff3f2bfafe3881aa2c62aeacecabf6988e7c18da11941eb","split":"train"} +{"domain":"finance","full_message_tokens":68155,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":936,"sample_id":"sample:f0dae5bf3a1dc5309ed150dd6e9aba66f3962c7d6a023fae59f7e6bb58309c3c","semantic_task_id":"577d9bc5aa402f4a388ebe36ecfc675f72d9173294e491793dc2035115b83c99","split":"train"} +{"domain":"finance","full_message_tokens":221007,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":937,"sample_id":"sample:8d0a12d04fbd21bce081b1122bf3dd5c744cd2250365e9674b4bcae1708fa7c9","semantic_task_id":"57ce0b41a05b8eb956d33ddcc0eb1d776924da5c69a885f70da33af34df5a32b","split":"train"} +{"domain":"finance","full_message_tokens":215606,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":938,"sample_id":"sample:9ef00b0b01aae67f9a71900de7b054aea1073a4fd861cdfd06aa57044fb32154","semantic_task_id":"5a38ce1703fdf9828bdc9e02f66fc3105b863d397eba2acd38e9fe4cace271b6","split":"train"} +{"domain":"finance","full_message_tokens":80183,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":939,"sample_id":"sample:274442c83d70039628c4305395587ed0f3115abcdf76c2df4cf6383184e70b5a","semantic_task_id":"5b288881409b438b872ee70d47d6d4231af9a809f692b38b25f2c28c4a583da1","split":"train"} +{"domain":"finance","full_message_tokens":80113,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":940,"sample_id":"sample:d1c93c5959ca0ab00c3e8c017ee1d77925982db00aa310d2879a67476f7bed6d","semantic_task_id":"5b759325a3426e43173f17694484cbf7798054151fd2f707928b08dd32c1e97a","split":"train"} +{"domain":"finance","full_message_tokens":208908,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":941,"sample_id":"sample:bef757afef6f75a6f3f55878a85c2abfa02b425c275e5c9db7e34912b18b3665","semantic_task_id":"5dda5d9c57320b0ca19e6dad6021d874d2ae0e56d8f476a1d1d98de3b7923ed8","split":"train"} +{"domain":"finance","full_message_tokens":77470,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":942,"sample_id":"sample:4b6859e38a165d174b801b5997aa0add597c91d700b53e042c792fa2aac6c1f5","semantic_task_id":"5dfdfc51b3b8d556f182b1396c2f56bb515ea5a8ad4410023861fc2cf050ae32","split":"train"} +{"domain":"finance","full_message_tokens":75226,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":943,"sample_id":"sample:66c1b0a308c480ddac794c5cd275c0097dabf466c2d5295a2a738c3685d453be","semantic_task_id":"5eff5f0b36488b600764d3187a31541e8ddd37fe020ede312b240a96092f6535","split":"train"} +{"domain":"finance","full_message_tokens":80180,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":944,"sample_id":"sample:7931649123ff5f4eceac08a562a13c957eb1188bf4bddf6bf20ddfb33931a309","semantic_task_id":"630a256291bf4b843cefb4f2ce455a5aad5291dabff89703a555befb29138918","split":"train"} +{"domain":"finance","full_message_tokens":77446,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":945,"sample_id":"sample:2f822c5f156d0ea567027f175ebc5c1b2026d374ee58c32f8308b3423cef14d0","semantic_task_id":"6310f7b6c7bfde0594f01f2c80504d8519a19a358053f557aff8dc05b48c1d9d","split":"train"} +{"domain":"finance","full_message_tokens":77448,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":946,"sample_id":"sample:7de06885d3910fae95d3629239700e6534544660ab4a7a709b3dbb072f5f9f0c","semantic_task_id":"63b67e1ca34b039d036a2aabdabd0319ce5073f67be87362b6efd323007c2504","split":"train"} +{"domain":"finance","full_message_tokens":75280,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":947,"sample_id":"sample:6f5e7fc5cf4ef9aa2bba77dd2776efbd6cbbfe224d8979147e27bf3f0c8d58ba","semantic_task_id":"64baf479340eaa0c9e55174527374bf2f6799dfaa56e3c4e960bc5ac80c76f07","split":"train"} +{"domain":"finance","full_message_tokens":80112,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":948,"sample_id":"sample:a268e93711904b09aef93593c42088615abef2f0c6492c447d0cee4a523adc89","semantic_task_id":"64c7ea8ff9665d93ea1097b5a2fab09b9218750ba4045177a94e0f7e8357ef68","split":"train"} +{"domain":"finance","full_message_tokens":75272,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":949,"sample_id":"sample:11e32cfdbcbbe35f9027551fbc83c0677e54146859678cf7baaf5c7b67296b27","semantic_task_id":"6580505b5d857cca28eefba65db7f5b638d89287f27fb2bb2f91a18cb6235849","split":"train"} +{"domain":"finance","full_message_tokens":208896,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":950,"sample_id":"sample:3673d474c774d20c96e29c1097460f7a69a687f7d7daba85063463f5be6460f4","semantic_task_id":"6a63c722ccf3331440dab7e12673f0f548496289273b4ccdfce83bdb4e857975","split":"train"} +{"domain":"finance","full_message_tokens":215606,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":951,"sample_id":"sample:54f1fdbd8cb913ccebfe301e57277cb10824c2a99288f66968b40651d5b19693","semantic_task_id":"6ae790a076194927b5835bcf634e5c2be534a9eb9bfdd5c589d0cef85bb1dc5f","split":"train"} +{"domain":"finance","full_message_tokens":215618,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":952,"sample_id":"sample:be2186e42635a00455c9307a7bb1ee972ddb8209e41da7efa961ce6fd65b9e4a","semantic_task_id":"71e57aafbab6b0de0579513020af56c542e5e1abe518ea23200571e11e32c83b","split":"train"} +{"domain":"finance","full_message_tokens":80113,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":953,"sample_id":"sample:a86e2014a513456e75101b886a8f57fc6560bef92c4cea9c9400f2520baaf6d5","semantic_task_id":"7283448355785195a96c945fc9536d1869a9c0534434ad6d9fdc06e0821da18e","split":"train"} +{"domain":"finance","full_message_tokens":68563,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":954,"sample_id":"sample:f4901d1c0a44840d076851ec394a529f1377b3bd3ce94ec30d2a6e420e287356","semantic_task_id":"73de5595ceb73a349985400c0b8f31942b167e0ddfdd375d118a0aae7007138f","split":"train"} +{"domain":"finance","full_message_tokens":80221,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":955,"sample_id":"sample:ac88808b3c8987bb021b2f59e67e1b6a7f84e62f20eb4c2b46f0de7df35abfe3","semantic_task_id":"75a72d7d243015c4d6c237c62d7c17cad230f0bcab3384016cd27de137e4c1c5","split":"train"} +{"domain":"finance","full_message_tokens":72492,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":956,"sample_id":"sample:b0fbea214130f66e328ad4a35e9f1897af61eaa9a3201f2e9b15dc75984a18b2","semantic_task_id":"75c2ba8d02a3261e43eccdb77d3337ae983b73642f522182f2e4895aa74e681b","split":"train"} +{"domain":"finance","full_message_tokens":77483,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":957,"sample_id":"sample:7306973ca74fbee0c617c2849556b889760631a472ca54af79c173330a6e67aa","semantic_task_id":"767cca189d45a24ace0550d9249bbe5f0a98fed3cb1f256bc3d3e63a47737d06","split":"train"} +{"domain":"finance","full_message_tokens":221007,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":958,"sample_id":"sample:421b9d803551edf0a88e1d926a60d80b6f4d50a6ec111e94fa92d7662b54e397","semantic_task_id":"76a7548bf6ba37590a72201b7db30153dc66d73fdf5a93ed7a9d48162dae7635","split":"train"} +{"domain":"finance","full_message_tokens":77485,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":959,"sample_id":"sample:9f8ce3600c8fd45da706a6a81b1552c5998911e62ddba07cfb8479e4360312a9","semantic_task_id":"773f2deefb32624507d17c6330ab5ae33f22716c06bc926c001b8d44e0812b52","split":"train"} +{"domain":"finance","full_message_tokens":221037,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":960,"sample_id":"sample:fd93ffd90a2d9b8cdd8cc5a556088569c5dd5f425279c14f04a7c4e24798fe35","semantic_task_id":"775641adc879c1168e4846cdb91eb01944a0e711925ad8d5befbe33a7d13fcc5","split":"train"} +{"domain":"finance","full_message_tokens":68518,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":961,"sample_id":"sample:dcfd6e4b01a5cf02ee7eb4f2d10a5d920f6770144643ddd88dc8543307fe4e23","semantic_task_id":"79bc05ff17267f2d34706bda79e74ec60ad11c3efe89deaa6ba8a78e47f42f4e","split":"train"} +{"domain":"finance","full_message_tokens":68087,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":962,"sample_id":"sample:887a633719b33b9a769d09e07816f20fd4030e334be402f02aad242826fd255e","semantic_task_id":"79c5fd30a6b7a78f172f815601a658de7337ab62259e44c2ebd6438c2434c2a9","split":"train"} +{"domain":"finance","full_message_tokens":77444,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":963,"sample_id":"sample:033738500b28c2a4e07189c63f28fca9aed128d834197b4cd120e88f56210104","semantic_task_id":"7abbaa0b5a031d2595da6feeb92531fdabaf93f0f9cc1fa0c40eadf254b0c564","split":"train"} +{"domain":"finance","full_message_tokens":221039,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":964,"sample_id":"sample:350b06dec2768efb253f447807d2781ae6065f514a5687efb87a55982ed6fb78","semantic_task_id":"7abc73a57ce05de0df98f4193789f5d06c33ba5cbe8d6a49cc6df11b2547e545","split":"train"} +{"domain":"finance","full_message_tokens":221007,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":965,"sample_id":"sample:42b2b3d22f5ba93137893b69d5e928788991f809b2c79d4b5a000bc4283e4e6c","semantic_task_id":"7d93a34890a72b33c327b47b3472759ceb445fb39efc09eb909b9be94379fb0f","split":"train"} +{"domain":"finance","full_message_tokens":80217,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":966,"sample_id":"sample:3be3ce65303ab112c276c36cb6da1245dc6103a9ae05558379e9a949a6b22ce5","semantic_task_id":"7dfac58faef899cb677cfeeb49aceb96c1be9d70e0eaa440b91d5b3debd729a7","split":"train"} +{"domain":"finance","full_message_tokens":215605,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":967,"sample_id":"sample:17cb64011a1f28747f5183e5102a6c1ef19384f0f4a50e020b698930d1416f5e","semantic_task_id":"7eb4d74693b2cba5bbaa3c18350d5e7ef2266b6e68e79d6e52c9ad58e0ea96b4","split":"train"} +{"domain":"finance","full_message_tokens":73512,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":968,"sample_id":"sample:e0aceee023b95976b62cdebac78c41f716e2369e10920bbd30b04c998eecb5d6","semantic_task_id":"7ef86017b419a4a31556104f1a1a45c63f286b25f2e4286273b7275916b8646d","split":"train"} +{"domain":"finance","full_message_tokens":215627,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":969,"sample_id":"sample:8e301c05f7ecd7e0af886fca7a6e8d84a697a864e9a04bdd0ec7470611f4decb","semantic_task_id":"7f21652fa35890c16b0ea47cbf6b727bfb57486bb67a26fccd495643c72e6e2f","split":"train"} +{"domain":"finance","full_message_tokens":77487,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":970,"sample_id":"sample:05e40a8f4afd6cec09a1072578af6dec83284c5d0a8fd9e6f56ceb0fff710651","semantic_task_id":"80d88380c6e6befe9f42bcd89f0f9b27d7333ea7c9ddbfc91c8e10b73f91aedd","split":"train"} +{"domain":"finance","full_message_tokens":208898,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":971,"sample_id":"sample:582ea05e69fcd1cecd8a13bdd7ab6f84fd7b24339f8821bae8d7a221742ca687","semantic_task_id":"81d5f043726f6875bef942c76915e2282d83bde993a2c74222619409fd70947d","split":"train"} +{"domain":"finance","full_message_tokens":77487,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":972,"sample_id":"sample:e3b6ea787ddc8752f55ac34da4883e9d2b135d761e43ab6b277929a2e9c1206e","semantic_task_id":"8298fd1437e3cb399e11ccf583c8a6571aeb0012acd0830baa30a093beaa3fb7","split":"train"} +{"domain":"finance","full_message_tokens":80179,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":973,"sample_id":"sample:ca76e6b2c89615497af390b64a0f1d1d58c3fa10b7b7f6f696d89c677eee5720","semantic_task_id":"8299307f92603abedd43ef8613c8c775a6d3eb951accb0cb49493e12593140d4","split":"train"} +{"domain":"finance","full_message_tokens":68556,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":974,"sample_id":"sample:9673303c8357b44d802cf5b74bb6a10d2bd8701ba9e1f8f5e5c7ffb25c47e01b","semantic_task_id":"83837bc81ac170cc672fdae9d1d09e10826200eab0bc29d74139fbf07ddb9635","split":"train"} +{"domain":"finance","full_message_tokens":208909,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":975,"sample_id":"sample:c6ee50e500cc267b01245a71b314528b3def711e6796a30e57d178d02648d45b","semantic_task_id":"844a381511f1b1cbd5ff8f410157b2fb0972375d747ef784c0a244b0db011edb","split":"train"} +{"domain":"finance","full_message_tokens":80220,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":976,"sample_id":"sample:fa5e747f34c22241ff436ea7bbe7af7de3c706e4107108978b408af701801426","semantic_task_id":"84d7dd154215a3b556fc17ec1b09bb737daa0c4e3296def8fbea17f036600a69","split":"train"} +{"domain":"finance","full_message_tokens":75223,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":977,"sample_id":"sample:e43720ff2b9605cb529d830af41db53b7d212f14899b7703d79590796f111c35","semantic_task_id":"85365d50a63faa33ac877ace98cf2fef23a71ce99095b9e1ab965e348a9d651d","split":"train"} +{"domain":"finance","full_message_tokens":208897,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":978,"sample_id":"sample:033ceeb3d7586459207935d83914ee6af3ef139dd1abc5c4a114ec5360654719","semantic_task_id":"85eab971f3d3990e9368a7f5fd996752df6d0221dd72165feccb66185b4647dd","split":"train"} +{"domain":"finance","full_message_tokens":215628,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":979,"sample_id":"sample:9334b7e5ae5ce97e84a60b8098a369f2dbf7448def9db7d5ff514035cec57e2c","semantic_task_id":"8711792dc23a6b497292cea50572059d13db40e74bc4a551195cdbf24cad8d60","split":"train"} +{"domain":"finance","full_message_tokens":215623,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":980,"sample_id":"sample:34005aa7d13c341d7dcd03a309ca698f870d51234553c310af2437b9a0150903","semantic_task_id":"87297dbc5396749cebab21176dda0a5420b65ac5c0c37edc20fe91676594cb9e","split":"train"} +{"domain":"finance","full_message_tokens":75265,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":981,"sample_id":"sample:0ea50ec343a20f27f3ed1328eacddc50b20d3ad1bd7713c62825fd05226db639","semantic_task_id":"89c4499f17609ed2aecc21af4775a5aa7636601021f16aafdb841e1bf31d87f2","split":"train"} +{"domain":"finance","full_message_tokens":221007,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":982,"sample_id":"sample:accbb1eee418eb9fbb90f9490d585c65946729dea2de156a8390d479085d4d30","semantic_task_id":"8a4151042c8c7ee52bb8fcdb2efd5d0c27024b6c21dd1dc0fecaf46976a43b7f","split":"train"} +{"domain":"finance","full_message_tokens":215608,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":983,"sample_id":"sample:7c5ddea1720943902922cd44b014de770305f0934cb952d34aa0ed263bb29e41","semantic_task_id":"8b1718c67855c0df5dccf20e1b0342fd954db1a5c12d3a39208a59f63df181c6","split":"train"} +{"domain":"finance","full_message_tokens":208898,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":984,"sample_id":"sample:54fd30cef3a4eae644d0e0d075b0898050618fedbd234141e2f46af37d77f208","semantic_task_id":"8d09a6bb228698a755568d918b68c3fa9dbaad41469320b0d969f0f939af49d5","split":"train"} +{"domain":"finance","full_message_tokens":72492,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":985,"sample_id":"sample:9708e6388fbbab1298f69f081e7af443cfffe0aaf67a46abd9e440349b78e9c1","semantic_task_id":"9009e26a6090e4959a85aaf2c1007641accb68b01855874d74bb002c7843553f","split":"train"} +{"domain":"finance","full_message_tokens":221084,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":986,"sample_id":"sample:f464cbbe546bb7c6c4625dd3180c9875610b6fb6213044be20f3c7e5dd1b11a7","semantic_task_id":"903da9e9df12cf0630868daf74513f31ef72864e31ebc8d6d802f4dc0f854e4d","split":"train"} +{"domain":"finance","full_message_tokens":208909,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":987,"sample_id":"sample:868fa65b59f7791d86d69314fa28498c3b628e45531a79661102aa8aaef9ac82","semantic_task_id":"904b4e48082d49f3b792b3edafe7cbf3254b0b332236e52e27c469e7f5227692","split":"train"} +{"domain":"finance","full_message_tokens":68088,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":988,"sample_id":"sample:7daab48623de236829a35a6eea215dc4437883f4f260a3dc0f7cff501891b991","semantic_task_id":"919144c12db7bdf019652712a43f42846a8c5f9e450382e50a8fcb5b6c085bec","split":"train"} +{"domain":"finance","full_message_tokens":77442,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":989,"sample_id":"sample:ca5340c2962abaec4d7b3d551e784a9e413fbe0e6e7d49f97244b99dbe73e8b8","semantic_task_id":"9241adb3973fd62ce9f17f0fa2889ffd0c5604779cebcf60c5f614697159f113","split":"train"} +{"domain":"finance","full_message_tokens":77472,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":990,"sample_id":"sample:4d62ef7aba76e0eb6f2d815d8dace9c8d66cd843c541cca5f14bfb7120bbff16","semantic_task_id":"9308f864e9a003c24f37a7644895e7f692b5fc23f7fe6ac760ea294964916179","split":"train"} +{"domain":"finance","full_message_tokens":221007,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":991,"sample_id":"sample:f8d29adaf4d720e217aabfe6c50dbc554aa51551210736180966e69634718c80","semantic_task_id":"94465a1a3a1477e0b349a95b32323639a850e25b6622a7430b90c709097325da","split":"train"} +{"domain":"finance","full_message_tokens":77487,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":992,"sample_id":"sample:2fdb56499f1de5e44eafd4607b1327af4a9d5165dffbee07d872db732e36d7ea","semantic_task_id":"94c875050eeaecfa5283e433ccf9a6ba3f7f157485940351d75e6828d2fb806c","split":"train"} +{"domain":"finance","full_message_tokens":73510,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":993,"sample_id":"sample:59b217bff72e1b22e314c61e1cd2b6ca9814b19b2d050224032e8eb7f5d89ff7","semantic_task_id":"96311d88b9e8d7a5f32c6cffbc5824e5792c478c4e32e988c2176d4d6206bd5d","split":"train"} +{"domain":"finance","full_message_tokens":77444,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":994,"sample_id":"sample:5ad681a456e5ae56c585f8c9d9eea35ec23f8aefb89a99364a6e21df74c28f70","semantic_task_id":"969d3f6cf29bcf50eb49bbea531dfaf78f96bcd49fe4dc86ac1f9872a4ff3af2","split":"train"} +{"domain":"finance","full_message_tokens":80206,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":995,"sample_id":"sample:51cff8797426724c1c946c35def1284930ffceabdc25dee3b1ffd3dc9a76b852","semantic_task_id":"96b12d2127e50afa91c170097f7e7fef2a68ce070ea02e5ec6ec2ff6cd6c0920","split":"train"} +{"domain":"finance","full_message_tokens":215627,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":996,"sample_id":"sample:fded505e4e43aac639645a3cff9bf5e4e3401f68f5e2f06bcc9c3081d3a41183","semantic_task_id":"97736766fa6fab32123fe88e55843c657f5888ef1cbdf126baf8490385c5706e","split":"train"} +{"domain":"finance","full_message_tokens":221089,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":997,"sample_id":"sample:f2a26bebba4b0a4742bbe5436d92e688a5484fd6de2b161ba0f3cdc985ce0f1a","semantic_task_id":"9854c7debcde6db177edc0c6faa1f517750e295315000c908701a791f4f1d136","split":"train"} +{"domain":"finance","full_message_tokens":75227,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":998,"sample_id":"sample:7b8100abae3b32519b665a6aa10d13ae9b22a2f6127343bb88c5c6d5a54c3e16","semantic_task_id":"98bf40c4ad6e8e17af341ff1151a3eaf62078068e938903e8ef2361d99b02bcd","split":"train"} +{"domain":"finance","full_message_tokens":72491,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":999,"sample_id":"sample:145d82bc47f920deaed9c8c618c0cd8f7b09f8efb485e9af8c35d0f951b94877","semantic_task_id":"98fd37058447a923c9ceea1b36e77c05b26fa19d9c4b34c5a9d96895ece08553","split":"train"} +{"domain":"finance","full_message_tokens":77487,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1000,"sample_id":"sample:74e0de3e6d3f3d9a34563ea02d8129fa2da70f7c1d4b55ea330d75ff984e99fd","semantic_task_id":"9b22a6a12b8aa63530ef95eb423fef82642fabe1bcc887bc1cf5fd84d8316114","split":"train"} +{"domain":"finance","full_message_tokens":221025,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1001,"sample_id":"sample:1c714e8150e5a8fad84a740fce9e34e836cad4eb135aefa07f482ded590521bb","semantic_task_id":"9b58b7ffcaa54e5d2fd492090bef2cd53b46b06b837b83f661a06c5e2edd7dbe","split":"train"} +{"domain":"finance","full_message_tokens":215605,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1002,"sample_id":"sample:4c5f714966174692d4ab3bd4bfa705b45d8b24111aa5db9d83cf296b9e8de2ee","semantic_task_id":"9d7db94a7d931b3227174eb413b78db66baa665d4846543af59dd9cc206f7bef","split":"train"} +{"domain":"finance","full_message_tokens":215643,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1003,"sample_id":"sample:d8c100c2f0d9e144299b639551e1e6e6ec16a19545cf3a89b570f872968b6d76","semantic_task_id":"9d8c8282d643f4e3228f607c129a3d8b63896197decd04d3cb678b596718b9bc","split":"train"} +{"domain":"finance","full_message_tokens":75225,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1004,"sample_id":"sample:dd6269f051e9f78dcf4409660c307197ad2fbb9427e1799c0b3c3e1473d2d6e3","semantic_task_id":"a01ed059cdf10ab21ff1e68a11bc07a47cb3beb74badd8ba77aa2f28a2e5f721","split":"train"} +{"domain":"finance","full_message_tokens":68518,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1005,"sample_id":"sample:3732516352f6f265baa04d63339560028ee2ca6ad26f9ff05425513499df9291","semantic_task_id":"a103f5581eda2f0201577935bf2870b9205ff4eede66a8ef2a331dc3d6f1a69e","split":"train"} +{"domain":"finance","full_message_tokens":77447,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1006,"sample_id":"sample:4a77220541d045fb27fde42d8724155ff625fda55b883d34f5c574c366e382e6","semantic_task_id":"a23822998a3e465ce1e2237542ad367cf1885ae54149a02f5895877def721790","split":"train"} +{"domain":"finance","full_message_tokens":73512,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1007,"sample_id":"sample:112078f47235a0997442b6851203fcf1bb195a73f3b53d139c218ec4d71cc4a4","semantic_task_id":"a2e649e60bd40ab30b1c1ef758369ff74e179da33667ca0ccf676d6cee26f0af","split":"train"} +{"domain":"finance","full_message_tokens":208953,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1008,"sample_id":"sample:f871795741305e812a8783af9987e6b9ad642e7d798155251bf0dd6421b3451f","semantic_task_id":"a2e6d48956b75d8a73d8dc04b938f6e7466cd1eeb59ed6def7fbaa9b626135ab","split":"train"} +{"domain":"finance","full_message_tokens":77485,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1009,"sample_id":"sample:46128d7dccd21c857ee3050b441b75bb6a32d36685ea04514ee78a5de37be4d2","semantic_task_id":"a38d4532b3612ced4b0cbab7708b7f52c2c3bc3d19c26f4f864cfc109a3d87dd","split":"train"} +{"domain":"finance","full_message_tokens":68142,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1010,"sample_id":"sample:1bc3ee5003036de1687c83551389489b52daee7690ab0f54a6cc182f0ac84076","semantic_task_id":"a3a106873ca755330828aa1bc2426ed6db9a1b1fd1866282ea19aa1e1cb85502","split":"train"} +{"domain":"finance","full_message_tokens":208896,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1011,"sample_id":"sample:0c44179cf84e143aa883b1f937d292cd838c8bc490570cf19291644658322c5f","semantic_task_id":"a84792f103a2629f1d8936e883f6e8695c8972fff8fe31868237980e23b5bcd8","split":"train"} +{"domain":"finance","full_message_tokens":208897,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1012,"sample_id":"sample:a36f5ced4d3404b35ce6547e6bee975cf5cf83e2a402ddfb7e7c0d9595ac8e4b","semantic_task_id":"a95d2332232d91811af2d601ccf54cc33a9eaccc681b56ff4bb006384cf40c5e","split":"train"} +{"domain":"finance","full_message_tokens":215627,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1013,"sample_id":"sample:3033b614068a7bc2847e793092c589daad76b37f7546b37ca8898f5813e0f48b","semantic_task_id":"a97d8a26aeb2e6e66a1d4dba3a7351019a972f2b6a81bf3c716b5680c1d9fbf1","split":"train"} +{"domain":"finance","full_message_tokens":208917,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1014,"sample_id":"sample:18e8bab53e8d577054ad0a1526e6a1a0813a1053e45bca940b46257ecff52af1","semantic_task_id":"aba5b09646266afd2e86db90770f88ffc6c430adee36fb54c801f1a35e58570d","split":"train"} +{"domain":"finance","full_message_tokens":73511,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1015,"sample_id":"sample:914a2a856526d0df6d22154ce255dc617c9e500e090caa1f3e274bdc1595be0f","semantic_task_id":"abeb19a778ef4ab3dae295f5cd230c4b4f1eb5b711056787f237a56a477fe97f","split":"train"} +{"domain":"finance","full_message_tokens":80182,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1016,"sample_id":"sample:103daa894c88bb5e36424a015ca0b8b1e3fc58adae7ad908ce8a7aa858335dfe","semantic_task_id":"ac023316e021aa84ee61347825981d1d1c2719e6c0fae8b90c3bd48cbcc38609","split":"train"} +{"domain":"finance","full_message_tokens":80181,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1017,"sample_id":"sample:4f37afdb2d7c23ef1aac35cb1982ddd59c9a09a1440e9c75e659e4dbe3aaa346","semantic_task_id":"ad53ea286d9806a9e61be40d5a73118b3e7f7990b419df0a0a65c85a209d52cc","split":"train"} +{"domain":"finance","full_message_tokens":77470,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1018,"sample_id":"sample:debc065a3f955e68d81353d2fb4c3937dd28032f086c503f9eb37b4bef96c586","semantic_task_id":"aecf095069eff119d22636e5555719c578d65ac61a43d09312fc2779916730fc","split":"train"} +{"domain":"finance","full_message_tokens":215608,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1019,"sample_id":"sample:961e1d89467afb5226cd5d0a4b14afc065fe742fb03e3b58ef96a0e36d7abe06","semantic_task_id":"b11a99b24333d8a759bbd788912a17c3eadaa6cc6d8a78bbf899195a54d04010","split":"train"} +{"domain":"finance","full_message_tokens":68585,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1020,"sample_id":"sample:74394b123c34ea4f34246955c0562004ae245b7c832fcafe19d684ddd9cf598c","semantic_task_id":"b1c7a1f62f3504118298e121b2d1f0df60d73a411eb21e65f7b000ab45487ab9","split":"train"} +{"domain":"finance","full_message_tokens":77447,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1021,"sample_id":"sample:dc8e2871aba398c109eef303ed189390600874147fd6f6e40ccda46727617861","semantic_task_id":"b4b6dd4750a31825805dd0c86ee4334e6d251824e3fa7f6cbe1a605ef4bf21ef","split":"train"} +{"domain":"finance","full_message_tokens":215607,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1022,"sample_id":"sample:6af3eaef2731bb7cb75476b150a8562e6ca3b5bdcad441bd17bb1f3aa85bf2b9","semantic_task_id":"b5c79b5c5227b2245133d2f0a61dd165ffd69f92b676557075d3747e506dfeeb","split":"train"} +{"domain":"finance","full_message_tokens":80219,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1023,"sample_id":"sample:c6cc4523895f23223b2fa65d120e2392acadf6eef2aa66c2918a08f642d460d0","semantic_task_id":"b6a2d0a739e820cd0c06f0bbc5613cace936f557de2e8205091180467f6c9cc0","split":"train"} +{"domain":"finance","full_message_tokens":80111,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1024,"sample_id":"sample:8e8053f984f315c4ada3c43c8f5b9108d38f405a6b5a6c6c35d729157b199afd","semantic_task_id":"b6d8e3dc4d51703a0ed4b50bb5d65c8718c44248f824bcff8722a7bb6a2cb8df","split":"train"} +{"domain":"finance","full_message_tokens":208958,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1025,"sample_id":"sample:c63e895675257c50463b3e7b65f045451e03b7e19f4afe1d10286afe69742b4d","semantic_task_id":"b6f9e6a38b068406b922e5af199095abf080d82e57c155992cdd67f24a28522f","split":"train"} +{"domain":"finance","full_message_tokens":80111,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1026,"sample_id":"sample:2244f900bde41d6375d889d347602630a201665fe2702b24f3d0a5a1f8ad69ec","semantic_task_id":"b7843c5a19609446747c1b8d303d9a86576c87171a4e7781a056db2aafc147d2","split":"train"} +{"domain":"finance","full_message_tokens":77442,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1027,"sample_id":"sample:ba978fd13f6d569fc23b6ff44f8d00ad975ce5353c76722a64754f6400a88639","semantic_task_id":"b9588d521352039ac40b8d6f42f0b401368992d52231bb40b62490c7c5327260","split":"train"} +{"domain":"finance","full_message_tokens":80206,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1028,"sample_id":"sample:f84003a6b739c90f96c427d8dca47bd0d9cff63fe0c5254d52af9fb5918e9dd6","semantic_task_id":"bc8ee005a4799e95ec142aba2b28714d0592112b8495160a734127af300e823d","split":"train"} +{"domain":"finance","full_message_tokens":80222,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1029,"sample_id":"sample:e8037d4d2fbaaf666c56e97434da6bfe689ac5d478118df7c8eb97fa640ec5a0","semantic_task_id":"be24fa3f52ceb0a3aa1fb96e49bf2a4e083cf9d552fe5ae0dbc35da81dbffc50","split":"train"} +{"domain":"finance","full_message_tokens":208918,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1030,"sample_id":"sample:6a87657339414da75261e8f095550a5daa36c2c944302d6db2cbd22d78499cd1","semantic_task_id":"beffaef3961fbfbaaa9afb38779098157ac8d760e7b130023c6411d390d5942d","split":"train"} +{"domain":"finance","full_message_tokens":221029,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1031,"sample_id":"sample:ad91b4174d4fed7e7d910ac789caf908ae2b134a60172c6fcb57f4667450c589","semantic_task_id":"bf7761e8094b515082cde2154db514cea565709fa9f1aa8c5c75f71d6ad128f8","split":"train"} +{"domain":"finance","full_message_tokens":80217,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1032,"sample_id":"sample:6c226f3e770de09d311aff6ece6e76db0150904d8dae0c7e6706e971e4e1ac16","semantic_task_id":"c03d48458becd954756e5cb043ebf93b8ddf74c0582d9402eb18fd82a64efeac","split":"train"} +{"domain":"finance","full_message_tokens":80220,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1033,"sample_id":"sample:19a76e8a052f0a01024c4eb2cf37cb057dacfe9b275ee793371b3de325e08d94","semantic_task_id":"c19fa4edc18b86bc42125516adbb39cd1589bf64d93d064cb6a7e80ec348718b","split":"train"} +{"domain":"finance","full_message_tokens":215607,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1034,"sample_id":"sample:25ccecee591603d59a5a3d9dc31bf2d60b20bc8cd137d1969334745082d8c658","semantic_task_id":"c253349cc20a0b99d2114b038b50c17d2546e82e8c559ddca4bca09ab0adda5c","split":"train"} +{"domain":"finance","full_message_tokens":215627,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1035,"sample_id":"sample:47d2b248fad698dadfec06788aaa5b6817a43d78db08649b6b059c6cff092492","semantic_task_id":"c2fb3540477c1e97d9c663ca7dbcc83b92edbb5cbf7be1a8716bf1bfb3fe8080","split":"train"} +{"domain":"finance","full_message_tokens":77479,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1036,"sample_id":"sample:7b4eab125cc193226bce977e25f6d473e0252d59b7e4eda784b434ab8e786138","semantic_task_id":"c371cf5387a4476269d0e139da1ddc8c75b85b654611485402a6041b2c1a1d84","split":"train"} +{"domain":"finance","full_message_tokens":73512,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1037,"sample_id":"sample:9af55a5b683fab879ea46d7b5e51fe9ffecd023fc1edde6ce43c3f3b6b838b10","semantic_task_id":"c42e2489395d4e01d2c201be05b69e338f8fc962447a7c1d3619012bcb7d9169","split":"train"} +{"domain":"finance","full_message_tokens":208918,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1038,"sample_id":"sample:70ecbed7a3daa41ce77f8b85bd5af714bd289bf1364005834fd3723884e333d1","semantic_task_id":"c48b966590a73509cecbf9eaa527d4ac159ba60abf4c77cc68eb94d39273e735","split":"train"} +{"domain":"finance","full_message_tokens":77471,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1039,"sample_id":"sample:5bc863230169757872a76a2987ebeff9f12b5dd1a130733069b92cee77b855fd","semantic_task_id":"c5a4f16ad36aa249adb20d3714533b65afae16d7f82c11e26aa4abf6bb40af10","split":"train"} +{"domain":"finance","full_message_tokens":208919,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1040,"sample_id":"sample:20406609854c5c221e05c5bca3c66b36e7ad00d72fd70d56650a84b7978be6d1","semantic_task_id":"c5c2a0d7d7f68705e66f077c6cbda9675a3698999aa4e3edfc0bb3d30e7b0f2a","split":"train"} +{"domain":"finance","full_message_tokens":72538,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1041,"sample_id":"sample:a5a61980a8a39777756806ccfecc801bdb5392dbcf281b6b6ab66e30fa85e953","semantic_task_id":"c61962cbb3c4c03c5d5a711dc4766a8771393e3c9c0a6ccb11f3869978b9dd8e","split":"train"} +{"domain":"finance","full_message_tokens":215607,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1042,"sample_id":"sample:b6502c28c77f49505f495651faf6f2a848930443877efd51749dd2aead03d477","semantic_task_id":"c63ac55c1d8b1df0c881062e52a1da0be6a83897f0a965ea785b4c7c323be6f7","split":"train"} +{"domain":"finance","full_message_tokens":221048,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1043,"sample_id":"sample:6c711327a83fae3cfbc71d97b6a7f030246fb6ec500f022d1b87f966c5e246aa","semantic_task_id":"c670074689d527f9294d29dc6a83ca155e8f0d1dc41a0ffc3f8145ed443e3a83","split":"train"} +{"domain":"finance","full_message_tokens":208918,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1044,"sample_id":"sample:9323bf2494998a2bb22f0a5168aeb6de2a84c9219d1cdf1d2bfea71d32aab248","semantic_task_id":"c700b7ef96e7a3d558d05a0c1c2fc7ff6ed4f7c4106db54cbf9a03926093ea15","split":"train"} +{"domain":"finance","full_message_tokens":68089,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1045,"sample_id":"sample:2c98213d5edbe945601a14eee1b9aae1fa5cd906c992c3abbef3c13ed7df5f6c","semantic_task_id":"c80de9fe5fb71a56c342d595b71ef0b5814bc84e3386669d722ae88019079708","split":"train"} +{"domain":"finance","full_message_tokens":68517,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1046,"sample_id":"sample:3f00d1f61ce7ef5cca8648a65655967126d14661b9c13307deba2929bda951f5","semantic_task_id":"c8a289538d3f1b7fb436faddede94e8a38918cac5a809abc0ee5a511d92b65e5","split":"train"} +{"domain":"finance","full_message_tokens":68125,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1047,"sample_id":"sample:c441d6a758c87664377986c6362a042ed8293f1dcabd849d3079e396dc333be9","semantic_task_id":"c8af0c886f36489013b3f75a196c9566a6dddff54a08739fef957562ac01f8f2","split":"train"} +{"domain":"finance","full_message_tokens":208920,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1048,"sample_id":"sample:cb69b85a958de986773a7711427792786f3ee29219bcaa763b21b8208a50f6b1","semantic_task_id":"ca5323e2c2ec042b095938b338b36ed9dbda3cf29d0bcf62dfc6eaf32b5cb7e8","split":"train"} +{"domain":"finance","full_message_tokens":72491,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1049,"sample_id":"sample:57b57a2050d80a595387d023403bc9dc8f5ce7893fbf6e5c17e9501f59cdc2fc","semantic_task_id":"ca66159b5e8f61638d959e75c927004a0a87b63682e7dea0e65b1d53d5fc9de5","split":"train"} +{"domain":"finance","full_message_tokens":77448,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1050,"sample_id":"sample:6652136042f74e1431ae23960532f02db91d7f4e8414d54f58005430a32e380f","semantic_task_id":"caf3b8c930a67807dc46105f5c988d8ccd38e6b82721bcd43524628483df77f0","split":"train"} +{"domain":"finance","full_message_tokens":208897,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1051,"sample_id":"sample:2aa370a63a24bf5e95387b081e91222d9e74553493bed0822739866c76831e3d","semantic_task_id":"cb62d404ff9b8c0b4212fa104b25196ee740ba6e58e6d2ccc4d29e67b8dccef3","split":"train"} +{"domain":"finance","full_message_tokens":75275,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1052,"sample_id":"sample:5f11864e9ae39524ca5ba4089848529ff5565ee46d0bb3f2714a9888329c8497","semantic_task_id":"cca9f938b0e0c3c8dcacc43e17f9b0a97250e75955c181311427b53832440034","split":"train"} +{"domain":"finance","full_message_tokens":68566,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1053,"sample_id":"sample:5ed215a399f0f48a42e569bfbe98a892480d56d1fa38bd775d43fa4ac222167a","semantic_task_id":"cd413474b262f1bc45d99187210b17ae36ead15115f9354534fc4320803f0288","split":"train"} +{"domain":"finance","full_message_tokens":73512,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1054,"sample_id":"sample:5b6e2786b6953e1ac16d44217984c74bfbd3ded05fceb345aff27318cfb49173","semantic_task_id":"cedafeaad21cf4cae1e7b8869b68ae1cee14ce2b1b33e753e1e7c9526d06c364","split":"train"} +{"domain":"finance","full_message_tokens":77471,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1055,"sample_id":"sample:27b4a57f664f5c2db8a38087358523da684ab4a6a061a06e541f14f24300d1bc","semantic_task_id":"d0bb532ca68754f89bc69009c2b3a9ac5d023e0dab4037f95dbbe9d62c5b8796","split":"train"} +{"domain":"finance","full_message_tokens":215618,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1056,"sample_id":"sample:3a8de2936bbf9e4da973e689d8f456f75bb40ebc452aa83cd5f4d1d81c4992ca","semantic_task_id":"d15da063586a129c047e297034e460d07ced41dc8cde45377a507d541ec63448","split":"train"} +{"domain":"finance","full_message_tokens":77446,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1057,"sample_id":"sample:d469cb41886f43c9c2b86260ef3e639861ae93777101e1c0b6b0a146268259b3","semantic_task_id":"d250f3d794ce9505dc1ab6bcc701a08560e7b59e681c385ad23c0ad8ab877933","split":"train"} +{"domain":"finance","full_message_tokens":221006,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1058,"sample_id":"sample:b2439be015c32eb933fa7c897849e0b847eb8d8317aaa3815d56b36e00de9b0f","semantic_task_id":"d33f3f572e22649eed733ab59b4099366dadbb8725f9a54d321af716d395a49f","split":"train"} +{"domain":"finance","full_message_tokens":68134,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1059,"sample_id":"sample:8ada505643eb513eea4af5dceb865cbbbccdc4f65a614c9edb70c2f2d384fe69","semantic_task_id":"d4d5b1b1e6d9105fdd095e1aed7fcebe6945475d364794d0907ea9e9c81531fb","split":"train"} +{"domain":"finance","full_message_tokens":73511,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1060,"sample_id":"sample:c7d0eadddab31c85720bb3ba9deba69fecd71e8ab6cc81453ee759da80cdf0c3","semantic_task_id":"d50c5bbcd873ff6fc2fff6dc6a6c0b2e95b69fb31fb901672f3f2839cfc6e0e2","split":"train"} +{"domain":"finance","full_message_tokens":80215,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1061,"sample_id":"sample:7eee68247c5b97c72d62cf8ea814fbbfef6ab42f05331b1c0bc25e27cd5aca20","semantic_task_id":"d5e5d5fbbfaf3ca96958eb8a71bb5d9f0161e7a7768456bc4805a8531b8f0f0b","split":"train"} +{"domain":"finance","full_message_tokens":75227,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1062,"sample_id":"sample:37d31e40354d727508c06668cdf0b8927499744ea568c6607371575220245aa2","semantic_task_id":"d61ced8aea61f6ec7c7c4f661b365a804369710d8efd8ff4010866bb77519d1c","split":"train"} +{"domain":"finance","full_message_tokens":72491,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1063,"sample_id":"sample:0cfe156864648a91a1b5907c6dd35ab3f3138265a423e5965c11c45a78e9cc88","semantic_task_id":"d7fb3b2ce6436b029e75818bbf1c1cf4aad11ea36c55c610ac349f21f31787da","split":"train"} +{"domain":"finance","full_message_tokens":77472,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1064,"sample_id":"sample:4bc5a69091f6616cd3c7558802031924e92d2e249ce71b0b83417b7833650851","semantic_task_id":"d97d04b20fab14908b3a4199a089ef51338cb34c988738816e5f5198d168eab2","split":"train"} +{"domain":"finance","full_message_tokens":80112,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1065,"sample_id":"sample:3401f31a5b26612b07d0b09e454ad4cb660e23ec0d14c2e937ac2d449c309794","semantic_task_id":"da1c74bd8154f1660362cee0eedfb6afe50e124e21dffff3dc8b92da18fddd6a","split":"train"} +{"domain":"finance","full_message_tokens":80216,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1066,"sample_id":"sample:136171fcf2bcbd14525d060ba57b96fc8827afeb3a9f1c5cc27f8accadb0ff1b","semantic_task_id":"da9e7566c34b79fbb239a1df4d7b48a3afd497f702a5b7bdbabc75aa5ddedf88","split":"train"} +{"domain":"finance","full_message_tokens":215627,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1067,"sample_id":"sample:fb9fe0d963277ceed0fd0956ce4e95c0d5a1543657e81b13ed70c346968e4ab1","semantic_task_id":"db0afadc65ecb6a87dea78a187942ab588f4ba06223b5f0196ad635657d503cd","split":"train"} +{"domain":"finance","full_message_tokens":208898,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1068,"sample_id":"sample:5d6af46b909c967faea9883575b8c64ec40cb3f6821fa1b5129d925aa01a88f1","semantic_task_id":"dc56e4c4f0109db165c7bcdba047219053f48efaef720c7a788a843eb7a85f0a","split":"train"} +{"domain":"finance","full_message_tokens":75292,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1069,"sample_id":"sample:1f5ba7a4cb0f496fcd3940c659de3537c225a12951cbda423eb9bc48f7b8a2c6","semantic_task_id":"dd02887e3a0b009fa4ce08138f81b309a4caf806562637615db6bbed33fed6c2","split":"train"} +{"domain":"finance","full_message_tokens":72544,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1070,"sample_id":"sample:cf8f782134885b5c538648dd3227b0fb91c5be0ee087118a0537e1d132325fec","semantic_task_id":"ddae557e79e422287177f6ebfd0a661dafdbb58624595b4109a9cc5b56776756","split":"train"} +{"domain":"finance","full_message_tokens":215606,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1071,"sample_id":"sample:debc1cb4e8f809b75ad5fd8ca936d0465743e76e448e5e9a19ee20b5b6710ca2","semantic_task_id":"dde1799ec2d962b20c4790b594d599b9bb74cf85b54014dd15eccbe6bc815c07","split":"train"} +{"domain":"finance","full_message_tokens":221038,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1072,"sample_id":"sample:732884d34b2e6ca4490e6fd81a3042d4dd74af0494cd825a3de80bec12765c4b","semantic_task_id":"dde5448c6a144644eb89c1db324a410b45bfccd6f3e6adcec187f7522a874327","split":"train"} +{"domain":"finance","full_message_tokens":68571,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1073,"sample_id":"sample:be18f4e37d14a46c1fc056962c7731ac3a027ac4b1c1108f8c126fadbe83f489","semantic_task_id":"de9a29efc58fe1fb1373a83d12a53dac95b79afe6965627688260660e9af558d","split":"train"} +{"domain":"finance","full_message_tokens":77433,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1074,"sample_id":"sample:9c101d283940c635e5ef4adc0e390bd9b9dcca2f5ae701ba05bc625803dcf554","semantic_task_id":"ded9fd61bdf436f67a56b2471865ee9117f6c26c95f98c935fe8048ce913a84f","split":"train"} +{"domain":"finance","full_message_tokens":221047,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1075,"sample_id":"sample:f1e07439a2b39b66c30f4ab97e6e82edc7ea504ed17b24d5c0a3ae4e517ee2a0","semantic_task_id":"defe2809099bb905c98b23ab64a7028c221c9443031e03712edebf338e43473e","split":"train"} +{"domain":"finance","full_message_tokens":208916,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1076,"sample_id":"sample:dbc11c3c951e9d2b1a7ca43aa39ac29e55ef2a1ab6c6ab7f9c9a00f010659861","semantic_task_id":"df50892d702ad369052ec30c2269ef07dfa6afb776adb181a4532d3122775f2f","split":"train"} +{"domain":"finance","full_message_tokens":80214,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1077,"sample_id":"sample:5dc1125053a5201a6e5ae1f3820b33c9391d6277f2055efcc30fae0f366a3fad","semantic_task_id":"e080e6f18c7578611bad67b26973e2d3cc8f9a0864d44842adb5796048e105e4","split":"train"} +{"domain":"finance","full_message_tokens":72557,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1078,"sample_id":"sample:41d44c1caad10ba8994fd48f354f11bc13989ee600feaf98cc7413911fe20261","semantic_task_id":"e0f8c1d40b134f7cd0e7b8ad8d16a159029f3755332bf8af579edf6478b90da8","split":"train"} +{"domain":"finance","full_message_tokens":72492,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1079,"sample_id":"sample:e3d2c973ac3274c6ebbaa7121fa92f2b3801afc7f5c53f55248c2e4987359ec0","semantic_task_id":"e2304f4a8cc153d8c05c4ff64ab39154b419e195cd7877dd7155739a1e535013","split":"train"} +{"domain":"finance","full_message_tokens":72491,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1080,"sample_id":"sample:b021644e7b022f96c337c4ba93518d46a1a28264ca40d68a5c4faca0061e2b7f","semantic_task_id":"e3005b9448e66a39d44a04d4ef8614e385ed93db40c9e4336717e4ba6788427d","split":"train"} +{"domain":"finance","full_message_tokens":221029,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1081,"sample_id":"sample:a855e6f447f4d903f7b8e902b5d3767f047aca3f16db9be00cabb7f176b3c090","semantic_task_id":"e3027aabf302fda4365b41f2eb9acf87786ffb0cd2685e4e0a105813cccc75c8","split":"train"} +{"domain":"finance","full_message_tokens":208918,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1082,"sample_id":"sample:7d0e3838bfb07ea1f70d7879e434d82540b9c02a0ef07902afb51c633a146592","semantic_task_id":"e31b4701f07ef9ddbdc99c39a17d8d10fa8e73cc1e031e41163ad85268c1234e","split":"train"} +{"domain":"finance","full_message_tokens":215629,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1083,"sample_id":"sample:3f3e4ca3798ba9faf4de24f50d86ba98ade0fe5cd5b65533bcdc68203957b499","semantic_task_id":"e4ffe9ea7f10fa461628e42e880ee1ed6348cf0eec6d1161b6b28baa0071db91","split":"train"} +{"domain":"finance","full_message_tokens":75225,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1084,"sample_id":"sample:68cd4e4198834468e4e6853f2256537d2d5d98171a92da0c3fcde06ea3ed49b1","semantic_task_id":"e532bb6f0f0869420d46ffe95da2aebcb3882eacec8f6787cd3ddcc002af806c","split":"train"} +{"domain":"finance","full_message_tokens":208934,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1085,"sample_id":"sample:28365222c7af515c90d30e82d4aa7daeab1986c9a9d7a92adc6c990cc805fd03","semantic_task_id":"e81dabc7792dbe4be19d955fd13226982d551bee32a45214cc81f19bc6b1fef7","split":"train"} +{"domain":"finance","full_message_tokens":215629,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1086,"sample_id":"sample:229de1f6f2b41cbf83cb27b1d4915b60b6751e94e1a13e0c80cdedf8d62aece5","semantic_task_id":"e892c4f4fcfe114968e9d21b613490ce37150dac9d1415560d3bcd2b7200e83a","split":"train"} +{"domain":"finance","full_message_tokens":215606,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1087,"sample_id":"sample:fbbfdfab0c0297e7410f0e3a02494dc2bf388d61e7797bc98e56d489e27ee977","semantic_task_id":"e8e5a145b18b5e58ed939154991e881be5ec980f2bbaed0aeb8b777ad5cbadab","split":"train"} +{"domain":"finance","full_message_tokens":68184,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1088,"sample_id":"sample:9e69d8c32f2b9eced1e4c587c8ebb31813b4958168fcc841d062ad3c7efbf7a6","semantic_task_id":"e912b3617bcd46d911d7467d8407081cd919b127423db60a77f07079988b09d7","split":"train"} +{"domain":"finance","full_message_tokens":77471,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1089,"sample_id":"sample:ea36ebf48b59f722e6a4f7c3058b59fdaa8b70657331c41683908e07e3397bdc","semantic_task_id":"eb990eb6d15ef3675af33962ba406ab2b652756a62ffdd8aff102bd7cf9719a3","split":"train"} +{"domain":"finance","full_message_tokens":77482,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1090,"sample_id":"sample:211317d01aefd056d17aecd92fd6184287f75fad4abdc099c235d863ddc8c7a9","semantic_task_id":"ed6ff8ca16568860a1cc6d3779b2b5a19b796ebaf6bbf78756c522f7beb9f7d0","split":"train"} +{"domain":"finance","full_message_tokens":68516,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1091,"sample_id":"sample:95856d752e8f1f183f8898e60c0f9e7b29226e8b7e77b4dd75772ea695044a66","semantic_task_id":"ed73412e95d16c64511c2ff231c276275c64881639bfa47d42bba6ddcd9e5a99","split":"train"} +{"domain":"finance","full_message_tokens":215667,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1092,"sample_id":"sample:190c9b8b34e9fb508789c2b441f2e6f6d22c0d65fd7ce266766bf9a51c25871f","semantic_task_id":"edeb279685f89bcfe157b831e2d983a59fb30a4712bf190e854c552682c30b4e","split":"train"} +{"domain":"finance","full_message_tokens":75321,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1093,"sample_id":"sample:daa3f258a6e33f272118bab71272135c4995f463fbb2598dda517f4c9e0f8828","semantic_task_id":"ee99054ff40f114dc9829274b11acbe317b33c5c1e6829d9dc10300d9ae0d7f6","split":"train"} +{"domain":"finance","full_message_tokens":221006,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1094,"sample_id":"sample:db6f92e60ccad82167230047d76ae155e9a771e7ba1106a81145533d4647f67c","semantic_task_id":"eec495fde33465ba38c6b148c64ee7692193531ba5ef462c2fc102c836d70ba4","split":"train"} +{"domain":"finance","full_message_tokens":208908,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1095,"sample_id":"sample:9150b30bec0888e325d53ac28185335f7a283cbf12fa7c27b27dae2b30b74264","semantic_task_id":"f2d4d5d25cce608bf7f009b540cc153dba5a0dcb9f7793b9e91eb126696fa825","split":"train"} +{"domain":"finance","full_message_tokens":208899,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1096,"sample_id":"sample:53c3fc47dfa0b280f7c600e2544908ae72ab7cf22c3e90446d7a05bf942c2c4e","semantic_task_id":"f384a239ba605f1a0ada226d5f26f19f7d7edc0f2fff3060a6da1be37089a5c1","split":"train"} +{"domain":"finance","full_message_tokens":77481,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1097,"sample_id":"sample:b12f8471ac7a1d79d871bf58e8f4a7379cd6c92bafa2e1d9a98c6587f34d571d","semantic_task_id":"f5fd87dd95a59a747c3da3ab22d8290dcac8c4ee8a9df54e0a363af3ff459c3c","split":"train"} +{"domain":"finance","full_message_tokens":72490,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1098,"sample_id":"sample:c52bd109aa52516b4c7fc0e2f21914c76d2b603b5070e1f6d82d7529e64110d4","semantic_task_id":"f79f2e51dbcc2f19025526a9589c78aed1d13ca0f8420d0f3b0e05ff9f93fa17","split":"train"} +{"domain":"finance","full_message_tokens":77482,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1099,"sample_id":"sample:839f3e8b2299c22f2ed20a5df1cb63e28d1a64d2f2363f9d00d06f16701c81de","semantic_task_id":"f821e5a03bc51ba1e3361e1cc057e89ad0235ed7de1716dc5d4b3f40130775d5","split":"train"} +{"domain":"finance","full_message_tokens":208897,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1100,"sample_id":"sample:4dd0a5dacaffbab1b10e245b188ce0e0a5cd17577872f5601acc7d99ee769a13","semantic_task_id":"fb9e96e64f690641fd92e7227511023d48836033ed91bd1a8cc38ef7ca850be7","split":"train"} +{"domain":"finance","full_message_tokens":75225,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1101,"sample_id":"sample:1cc5feb96fbb8a135d93e8fb5f9224125967b9f6fe079006a864c9770a5968d2","semantic_task_id":"fbcb68f737e5ef6d6b049b19725999d864cef591d0702f57c847bc49eee54643","split":"train"} +{"domain":"finance","full_message_tokens":80221,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1102,"sample_id":"sample:34d3644fd3800770d2976861144c2f1a534102a42dd0e14884488aced3e2ba4e","semantic_task_id":"fbd96a031467bf08bb7f52c542fc331352cfe759f743e63db72ccb3c579eca54","split":"train"} +{"domain":"finance","full_message_tokens":208920,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1103,"sample_id":"sample:27a8167661b3e215013eee5f31fe2c80a4458064eb4fd67d022c9364c04c8d4a","semantic_task_id":"fcdd763cc6d7f4988cea6452e3351240be200178a16626be053bc761607b3c92","split":"train"} +{"domain":"finance","full_message_tokens":68613,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1104,"sample_id":"sample:9f2ba605151c34a1f446f18960c165bbc474f9ec5bc1ee86a60d640fa7ed9572","semantic_task_id":"fd0326f74829732e4a2673630ed6d8004b7291bd0f8f3a45d2509f790718c16b","split":"train"} +{"domain":"finance","full_message_tokens":80220,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1105,"sample_id":"sample:967ef30715658ebedb6fad2983d2246fa52e691c847a39c34a965c30003e378e","semantic_task_id":"fe60e6876885f1b41bf1f865a9e4729a3186fc05faf1a02134f336df61ccde93","split":"train"} +{"domain":"finance","full_message_tokens":72585,"group_id":"0000723125","output_file":"finance_train.jsonl","row_index":1106,"sample_id":"sample:9a1a85fe672b7d8df72390999ef8b70b7b30ac69aeac89e9d09d7d71e508e699","semantic_task_id":"fef511f5e8ad6c986214061d2060cec99162987f8f6134cf8af1e5a7efc3b1b9","split":"train"} +{"domain":"finance","full_message_tokens":125889,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1107,"sample_id":"sample:c0658fe98f1509bf1570f8814e16f3f377e4bf998ae8bece1edb3b72cd4d0227","semantic_task_id":"011743fd5f77dfce5f905399bcfa5ae7ca94f592337ebe12262f6e78c4c757d5","split":"train"} +{"domain":"finance","full_message_tokens":125662,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1108,"sample_id":"sample:a4680dddf6d13822335b50c7330f0e2e40bf5af68ac00285071d6a8d1b5bad1a","semantic_task_id":"0169dd3130da4abd9498fc43dd68dc93249c8d705fe5e38fea5e4809d35ed728","split":"train"} +{"domain":"finance","full_message_tokens":61697,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1109,"sample_id":"sample:009b2d81297a5c1489b45f2a311a5917918748c0455132d675d54c3297f65f8f","semantic_task_id":"022ad6e3f4076cbc776e1615c9c4dbf76e0e9b715bae6825de458b8eb95670db","split":"train"} +{"domain":"finance","full_message_tokens":125890,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1110,"sample_id":"sample:9c2d368c1e16e3da314e247562499f507a5f2435cf953615fca5d8eb0fb527d2","semantic_task_id":"02e6f7806b98265ade9e1b66265d6ab11515bfd0bbb4b92ab26eb9fc398139ef","split":"train"} +{"domain":"finance","full_message_tokens":187261,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1111,"sample_id":"sample:fa5876794b3fec1cf265b956ea4528cc36dbf24a74cfc870b98db051c738fbde","semantic_task_id":"0830791114eecc05ca47b7790c85b88475c2bb03e9604b13c7a4422bab41677c","split":"train"} +{"domain":"finance","full_message_tokens":61695,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1112,"sample_id":"sample:b48e30f1651e826e3da68caab720ff0753188008118b1ba590bc8c4b6aae3c9b","semantic_task_id":"09fff398a8024318586e6ba509f6c8b0233e18794af69657b4ebdcc77e3c533b","split":"train"} +{"domain":"finance","full_message_tokens":120967,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1113,"sample_id":"sample:18d0db1181100d08631f5cc70278cbb0d999207c6b63a79000269f910241399a","semantic_task_id":"0be549c0cc1a26b85e503be474bcc536d1bffc0cb9d9e845d30578dfdad1d324","split":"train"} +{"domain":"finance","full_message_tokens":182340,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1114,"sample_id":"sample:c6bc219de59a4e65a0e435a871512cede7ea398154c96ac30c4e01852fd94514","semantic_task_id":"0c112aa654462f7f69f41e90e2abb744ea85d349317b798e0e2368f139c0f99a","split":"train"} +{"domain":"finance","full_message_tokens":120968,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1115,"sample_id":"sample:e64f64a8a301abb47abd928ca8e5e379eafaf5318483301ea40993f13006dd43","semantic_task_id":"0c82f5d85fddda107d2b79699d457a7cb34324becb94617517c9c454a4d0bb44","split":"train"} +{"domain":"finance","full_message_tokens":123079,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1116,"sample_id":"sample:e3fd0844ca7dcdccebbc2708d819d834e9adcd37c7358d6a648ba92ae9580522","semantic_task_id":"0d42fc2b207f4224385b19cfaa733da31a9ab5ced05aad855f2ae03394f812e3","split":"train"} +{"domain":"finance","full_message_tokens":64273,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1117,"sample_id":"sample:7f366d773fd22e9bffcf7efe4460fa6d48287230f5093704e186cd9a373a5690","semantic_task_id":"0e986e9d5cf7dc39ee2a1fc0a1dbae16baa6c5624288caaa60c73390741e33c1","split":"train"} +{"domain":"finance","full_message_tokens":246489,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1118,"sample_id":"sample:08df0fd1b498646d6add481c83c19adb0426859103421db2e38bb24b91af1a9e","semantic_task_id":"0eef072bbae9324cf831d57094526fb90f16951f4546882ffb386bf9b9320ec2","split":"train"} +{"domain":"finance","full_message_tokens":120737,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1119,"sample_id":"sample:6f6877fcdd574729e608f4579b31e543aadef6d85299aa83e45256bab75b5264","semantic_task_id":"1047143dbe98a7d1593c471cce89bd325a0bd1dc83bff855a058be31e407f6c8","split":"train"} +{"domain":"finance","full_message_tokens":120969,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1120,"sample_id":"sample:b1b1c2594aa210534c009b4c0b9d67adadf75abc4c4d7323852383d843e2f3dc","semantic_task_id":"10aef5032164ad4dc6b96100e74b7c305c35c824ee13d05332ac45b533f5758f","split":"train"} +{"domain":"finance","full_message_tokens":246522,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1121,"sample_id":"sample:341f25193870a64fa2aebb174248d3376f26a3e877ea1918139838127e1acc30","semantic_task_id":"1297167b4e660484b4290a2890fc0fc268d6d1d6f0dec8845eef248c7c55d71f","split":"train"} +{"domain":"finance","full_message_tokens":246489,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1122,"sample_id":"sample:934daf57205c69965bdb76f948f97409a9a5319be467baac913aba4ed4541a17","semantic_task_id":"13060c57325e3d43b6997ecd85c70e8c61341d71d9bc9518e35769a92169b62c","split":"train"} +{"domain":"finance","full_message_tokens":246512,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1123,"sample_id":"sample:7fd9416634bf5a1c73e47a7e35d029ddb911735de797879efd2f13516ffb07f7","semantic_task_id":"1306334c0c0ee3eb8b620e066ae3e994833653b79626af0eff5e3fa1f347f859","split":"train"} +{"domain":"finance","full_message_tokens":64274,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1124,"sample_id":"sample:b27756747bbfa7070d30087fdd18a82034b964e982404847c0edd588ee1e5e98","semantic_task_id":"146d8a6a0120acf30098f95b2341a20d34a4813761a260068496837921a4f2ce","split":"train"} +{"domain":"finance","full_message_tokens":61468,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1125,"sample_id":"sample:6d9111e8f8dd08790186c6d1358274db205297253cca29e09f092103bbbf7186","semantic_task_id":"148912fb7ebb4585dd184e9343bd645d8a69e7c81b3a436125f495d97607712c","split":"train"} +{"domain":"finance","full_message_tokens":182318,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1126,"sample_id":"sample:92d1553d2fa7611b0ca678d0a1234b3442e4f0cd70e29e3e22f436a181c1eafd","semantic_task_id":"148b70e93b14d9eb9c2253d0ca3c2c4552377ff6452a3fd2ae951d6e53ab9de7","split":"train"} +{"domain":"finance","full_message_tokens":246523,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1127,"sample_id":"sample:3fe0ee22ab020fbb142cc22ad02ed6854d0aadbaa604b2fd048c7eb30613298d","semantic_task_id":"149e67a8d425c4cc5e244cb2fbfa7e78ccb0538e10c5be503ccad40798f80879","split":"train"} +{"domain":"finance","full_message_tokens":125888,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1128,"sample_id":"sample:dc6a442f999d3f96084fbeee1cd44ff6bbe4c366067f1a482467a85bfc421965","semantic_task_id":"15ccee58d59915e5999d5e088924ad07163e2ea73c217282405eedb8ba282c42","split":"train"} +{"domain":"finance","full_message_tokens":246489,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1129,"sample_id":"sample:c5923737db763d3a9de993764d05c6983bc8332290198e2adc7cdcd4fcdba9f2","semantic_task_id":"17ff7be65b8a1235d8a3eef485a02a644b7c26c93c37bec1eeb6b4da84118410","split":"train"} +{"domain":"finance","full_message_tokens":120727,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1130,"sample_id":"sample:8b89229ba8bc038c5dafcafd492693e048140c869072390d14fbe94d9e182ce1","semantic_task_id":"1922de130c04567b1be5fc7fc6d86f8c0376061d1a995bedd86d80e688030c75","split":"train"} +{"domain":"finance","full_message_tokens":64321,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1131,"sample_id":"sample:07e3d715ac30b309ce48ea81ef421071781f47089e75fd4de23ffe5c0437f7ba","semantic_task_id":"1985fd86860613f9ee9089e6ee14e19c75060f28e0c78492ff3566d69d25baca","split":"train"} +{"domain":"finance","full_message_tokens":246522,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1132,"sample_id":"sample:b44e82756a8384e2400ab02ebaa21c5e7a6f909cb3e0479223575424af8af02e","semantic_task_id":"1ad0ef2fbc737f04b446e03a4a73daa25e0ffe4488951a67c000f29657ec1cc8","split":"train"} +{"domain":"finance","full_message_tokens":120736,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1133,"sample_id":"sample:0cc8d47fb788484177885f0fbaf5966e273a01d1c3cce6ba80e7d50392b05174","semantic_task_id":"1b1c81dccd72c867d6694fc26a83ef6250f29b53fa9a9ee7e2f3402d86465f30","split":"train"} +{"domain":"finance","full_message_tokens":125889,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1134,"sample_id":"sample:d3621d5627459a0de4b5d6a36aa5ddb700a7c850e7b6148aad71a8a988051b71","semantic_task_id":"1c77df3a26e1d6d1e186f96fe72cb990dfaf243e57a015c9b24950c62e2985c4","split":"train"} +{"domain":"finance","full_message_tokens":61469,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1135,"sample_id":"sample:325468624729a9bb0b97fe31c76e49718b50fa532a35059840d2eb0d314941a6","semantic_task_id":"1cb8202e05c0995487edcb0f100f2b5c347e2317207f566531fc5197f4e2f71c","split":"train"} +{"domain":"finance","full_message_tokens":182379,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1136,"sample_id":"sample:6230ab4c41bc6bba512e7ebe939361726fc54c018e0cb126a1fe357e36a9034f","semantic_task_id":"1e06fd99c9debaa541d8b7e2ce88d5c9de387be598e27e9083df288d7cdc36e1","split":"train"} +{"domain":"finance","full_message_tokens":59399,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1137,"sample_id":"sample:16dbae060eb6d1e0b9ef1d8391f37dee5a52fcc830d04ad6749102c273f175a4","semantic_task_id":"1ea4d7b9a7652c1151713fd6fbeb14bcbb3a6022d7108cdda1b420032a48fe25","split":"train"} +{"domain":"finance","full_message_tokens":59405,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1138,"sample_id":"sample:134711389bae6ae95063aa912c59f3d3adb09f0010982f948f32f5f10d0753e0","semantic_task_id":"209401af817f7b4474ff7fdf3af72bb4769cd8f21d6419702f2e091016d283c2","split":"train"} +{"domain":"finance","full_message_tokens":125663,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1139,"sample_id":"sample:b912a6e832fda3b557857c97e2f010b15745af1950b13acdf4a573f9ceb1c860","semantic_task_id":"2130cf3edf2c3cd7b6a2bdee9e44d3f8e93831d5c2a6500acfee84bb830a66c2","split":"train"} +{"domain":"finance","full_message_tokens":125890,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1140,"sample_id":"sample:4c0aef55d1212cfe819c986f1852c9f33604a20bfca32660e8af559137ed1983","semantic_task_id":"221e6fe72d954fdc05edfde819578450170e19f594ce6508f9777feabe20f0fa","split":"train"} +{"domain":"finance","full_message_tokens":187263,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1141,"sample_id":"sample:aaf3ef2d483c2791c0d686001c2652605d46969ae4072618fd9a173351c0509e","semantic_task_id":"2666524bc95f2bb09dad6874fcc25027e9ca1631bb149a5d93d3c676ffcf8bd1","split":"train"} +{"domain":"finance","full_message_tokens":187240,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1142,"sample_id":"sample:41e68fbdfbcda990e191e4ab56a05e053fc10912233cdc7a7b52f63e88e9d1af","semantic_task_id":"28bf1663a36255a8666c4ad66db0b79eef77a8f083ad5fd4781b75b8a006cf75","split":"train"} +{"domain":"finance","full_message_tokens":120736,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1143,"sample_id":"sample:c812e0e62ce4b38a5632292427497d6c5244447f7c659becd33278f7fee5a44a","semantic_task_id":"29f4578bbb90f00e17c284ab981d54905b51caa1c2232ba7d7f9ac565ac0be53","split":"train"} +{"domain":"finance","full_message_tokens":187261,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1144,"sample_id":"sample:c2e35b2785c836996c954fd917169d1a1637ba671da5325284127d119088bda0","semantic_task_id":"2a9c8a1d8b34a40e6aa907b90ecc597a9ddeb966278353247d2f4016d5876935","split":"train"} +{"domain":"finance","full_message_tokens":64274,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1145,"sample_id":"sample:3cda4a155e3e1e5af567778adfcecb0024f24ef1742fde9d0cfb5ec48ada4780","semantic_task_id":"2ac091ae5f8d0aafcb2d70da6a0b9d083d3d957e6a085bb9da06ff86a098f890","split":"train"} +{"domain":"finance","full_message_tokens":182331,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1146,"sample_id":"sample:1b987ba06eccca9727e3fb01c0c78bee1a7d3aea98ce3a6369851be02ef39ed9","semantic_task_id":"2ee317e267993436fe368393fb293a7762941608912bb63131211eb48862de09","split":"train"} +{"domain":"finance","full_message_tokens":187240,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1147,"sample_id":"sample:ac7ae41c01d90b73bf0d5d41ceafb3fb431b11a62624007dfd74bd422e718542","semantic_task_id":"305f3a78d9441f29f1480881f6260f87ab550c6ef2e4e9b7f1adf88dca2f9f64","split":"train"} +{"domain":"finance","full_message_tokens":182319,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1148,"sample_id":"sample:c60442ccbdb47d2b659e7373387b06eca68f1212a9a3db9aa4c2693895b9da1a","semantic_task_id":"3179887e6ad7d57928679c2b046501322362e86d7ecaf402c8fae1b8a5b843b5","split":"train"} +{"domain":"finance","full_message_tokens":182319,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1149,"sample_id":"sample:fae3b25b5754508b25e286b48509c55f9d0b91a67c82cb57ccf0208ef9e28630","semantic_task_id":"344aac83239e4b7cc987dc8a2587bff98cbc777e2f0db29dca1bd1795d43e296","split":"train"} +{"domain":"finance","full_message_tokens":123546,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1150,"sample_id":"sample:179b4f768c5657cb58b48554ad376b58545cd984153d63ba55d80386743fb0c6","semantic_task_id":"34fc2aee302e8dfb03a6ffc7e4b4167c40da44c0a8c2f145694b7366804b1b69","split":"train"} +{"domain":"finance","full_message_tokens":182341,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1151,"sample_id":"sample:58664431e7c890c713a468e5c845cb88cd1b6253ba41e795b25de6d257dbe215","semantic_task_id":"3722b91ba9ab475203fd95b71b991d26aa8d235a4cd2d12af40a1cc6fbc14bb9","split":"train"} +{"domain":"finance","full_message_tokens":64275,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1152,"sample_id":"sample:6caf57adb18a5e7e566ea71a1712fd5e5b642f69e9c6783e5293eb8f68347a6d","semantic_task_id":"38ce845f3a4482db4f11a92029454ee479e233053696379c6f8c62ac5e8cce49","split":"train"} +{"domain":"finance","full_message_tokens":123545,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1153,"sample_id":"sample:5bdf91d81c760cb5f8312c53d29a3b82b5930e5f5f95a7a341a1acf4efb5798f","semantic_task_id":"3ab711437f1d436dd625add487c895be2ab9daa64d4acfc57844893f30b9888c","split":"train"} +{"domain":"finance","full_message_tokens":59352,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1154,"sample_id":"sample:122e6160c2e102ff3a9ccc259ca3771eb43dd6fd0d9aedfb3b52a594b7506c82","semantic_task_id":"3d8452eaac9cd853374bba0e2e200c10da93e41a99402226abf8de0edf9092cd","split":"train"} +{"domain":"finance","full_message_tokens":187240,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1155,"sample_id":"sample:3a3f8546780ab448e030a9c55a264d0204c438e2cfb8aa55ff676ff5f69beaf0","semantic_task_id":"3e75627d14a691b6b2825d9a01744ee943794cae704f12453eebaa1ba433c45a","split":"train"} +{"domain":"finance","full_message_tokens":123085,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1156,"sample_id":"sample:768be8f394a2d767fb0431b572030cd3128c0e355a288de2ca2d2b236d46a27e","semantic_task_id":"3ed728f3729ba0ca5209b7645d5bee62c5da2165ef6b9d4be8067dab251a943b","split":"train"} +{"domain":"finance","full_message_tokens":120740,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1157,"sample_id":"sample:de0853b245615ff9510186e802fc2a707a6d94d8a845ef4c69424f42d28ad1f3","semantic_task_id":"3edf67c3e1a503685ef7f367ef6a19db9dd03c1015d3ad8c7c3f6874734f2b60","split":"train"} +{"domain":"finance","full_message_tokens":123545,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1158,"sample_id":"sample:b36fc057e46dc3f01dd8105e36d2a2fe2c034556e8f85d0c309f51fc980268a9","semantic_task_id":"3ef790cc9652b27e487208262dd474877601844b1b38e88ba4710661fdf41033","split":"train"} +{"domain":"finance","full_message_tokens":123546,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1159,"sample_id":"sample:1ce67f2ef1bbcc5eae9051266660a51cadbed468cf2df1ae62ef20ffbc51a283","semantic_task_id":"3f7c37920dd94617f82226599bd485a5888f803a960042add68fc66eb9a5f0fe","split":"train"} +{"domain":"finance","full_message_tokens":182341,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1160,"sample_id":"sample:0114db2381ed3f8bf13a4aa62d4b8662d188cc11fbcf2f977f75c170605f09b7","semantic_task_id":"3fcf75896f6662dc59c6ba15f8f4abcce27bbf19c3aef5106acbd6a505e96dbd","split":"train"} +{"domain":"finance","full_message_tokens":125885,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1161,"sample_id":"sample:cfbd1d394174dff845d00a75bb876b99f2d89771eaff45ddf45cc4a62921206d","semantic_task_id":"40c6c9ead3b4c3a618ceeec4c141b75364dda3e6f1cde5843a81c54af0b4cb03","split":"train"} +{"domain":"finance","full_message_tokens":120740,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1162,"sample_id":"sample:62f16d2e83ec2fc71f02906b6e83b7a4cd01c0d1f1b37d480bdb2a45b64a0230","semantic_task_id":"4642079a7ab996ee69079f4db8359756fed88e76133aee5d7c06c72f19e41ff1","split":"train"} +{"domain":"finance","full_message_tokens":182342,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1163,"sample_id":"sample:3df76071a89ed953e96bf48bd2e0251d5a5d5d10763635aa522639b648769c7b","semantic_task_id":"46842ed52e81e50f12d0bc64b61df3fa4ba2c20ad16a6a87c7b80d77a2fdc6b6","split":"train"} +{"domain":"finance","full_message_tokens":187262,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1164,"sample_id":"sample:9835dc4398de577095b95c3818117abe2c1661299aca5c4c40d7159046ab6021","semantic_task_id":"488f18c116ff47228a4d6aaf89c83aff23e83396048ae756da89e9fdb68ff46c","split":"train"} +{"domain":"finance","full_message_tokens":120726,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1165,"sample_id":"sample:13ce1c64fe7d9887e1ea8e0d1857da3172341e7b7bcb1da3ba8aa179c288da26","semantic_task_id":"48b47247efe9b32c30e33e4feffb2cc2152c6e8f18238ba081ed4acd9c8210d4","split":"train"} +{"domain":"finance","full_message_tokens":246489,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1166,"sample_id":"sample:3048757a377aa2c5bf5a6ee9895b8ca7865bb446af65fd6904ac2bbbeec50bf5","semantic_task_id":"4932b39d6f852d8d6d553a6d0277b58b92a8e89256a40952794028be13dfe713","split":"train"} +{"domain":"finance","full_message_tokens":125663,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1167,"sample_id":"sample:65d97996ed197400cc17be5575bfbdda4831d27104a42443b0dcf1e3afe7a186","semantic_task_id":"4a0cefb0a463efaa73faa6f563259420f30170093ba97094607374561e566058","split":"train"} +{"domain":"finance","full_message_tokens":125663,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1168,"sample_id":"sample:f298bae62abe318042e734356e0a26b1d69d97611ff8ff8329e44a56c1a99832","semantic_task_id":"4a6675088ca75e68e018ffb0f56c045ecace68df8da88cb24a4f1339a73486e3","split":"train"} +{"domain":"finance","full_message_tokens":125885,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1169,"sample_id":"sample:9b4c06590264312ede52193f75d732614602ee3c7216589397c507a459088e1d","semantic_task_id":"4c58c227e6743638d727eae556b3ee4d453229ffa7aacccd1350682fc6df7089","split":"train"} +{"domain":"finance","full_message_tokens":182341,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1170,"sample_id":"sample:126edc427b74af2421eb26a800b3530445a525028aaed7de5ec5593c4390e036","semantic_task_id":"4cee56662403daf8efe1824d4973b107e7ce1a4815bb6722af39b65067710228","split":"train"} +{"domain":"finance","full_message_tokens":64273,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1171,"sample_id":"sample:cfaed06830165c1b0f1f356d614e966efa7b7d5d638befcd6cfaf8a57c148d8e","semantic_task_id":"4f2256bb9c51d70bfc8eb128b16e463d9ad0a1884c3bfe83f41cf8b075b0c157","split":"train"} +{"domain":"finance","full_message_tokens":120726,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1172,"sample_id":"sample:8c9d7ef6ffa4ed651815097730ed3da3c6ec081dcfaaa146e03a21804e8a4bcd","semantic_task_id":"51c14eb52c0a4a7bcec6ac17551da345daf638e52fa0f618c2fe6eb1429d0c53","split":"train"} +{"domain":"finance","full_message_tokens":125661,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1173,"sample_id":"sample:1655638eb301f07220bb6087ae34a14f603008c2d239b828fb1dfa752c785763","semantic_task_id":"53c4297f510c58bb06e71e07272705084fbfe743565377b072442796b6b8f29a","split":"train"} +{"domain":"finance","full_message_tokens":246525,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1174,"sample_id":"sample:189a55e9dd1c385850d11d1068d97d7d9601edb354f4c9281d7939837b35fe82","semantic_task_id":"5724d8532589fc2985dc971082821813c5c7942d933102d0f36e39630c8c7c78","split":"train"} +{"domain":"finance","full_message_tokens":125888,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1175,"sample_id":"sample:608c2ec6958bcfb62e4a6539448b6fda2b4678a7b2645e9961ddc16f5e956e49","semantic_task_id":"585c29429948f1a596f560e4e97609c962d92d82ed10b36670d61269088eb70f","split":"train"} +{"domain":"finance","full_message_tokens":120735,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1176,"sample_id":"sample:fa08bce0cdf5da567559e2d3d85d5be47158a88759fce8156140668ea3fe4c22","semantic_task_id":"5e0b6f766c879427042e1c8406d8ce8eb0c572edd9178d2c64c8e0d2cb13c0ea","split":"train"} +{"domain":"finance","full_message_tokens":246572,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1177,"sample_id":"sample:53e00bfc304abcfae54fbab19b3bfb814224099ca328a615c45dd739123caa8d","semantic_task_id":"5e8a646e40281219f8bfc45accc236ae22ba593584728a240ae3f6736a0ed827","split":"train"} +{"domain":"finance","full_message_tokens":187240,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1178,"sample_id":"sample:196550d06b79541720eb433fd8cfda918cb7b7984ef9f3b18ab4f069dc00925e","semantic_task_id":"5f28f658dd8249f4278192105735dbd45dfb55bcd9ecd930ed559d69ba3fca2a","split":"train"} +{"domain":"finance","full_message_tokens":125890,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1179,"sample_id":"sample:01c55d7625bd3857a035b05d70e970e894b1dfc27cb12d8b8a74e2c4004dd5c9","semantic_task_id":"60c6809de0685e5e2f67514dc16a7a9b7a55e602f508c2433dad31004ed5dcb8","split":"train"} +{"domain":"finance","full_message_tokens":246512,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1180,"sample_id":"sample:fe97abc8b3406c1d3c7555c07f27e1f829a81452e36c66cbdd70d1b70b794f75","semantic_task_id":"6102b976a2bdadae3beaa305e6ca4a36638eebadb18f4060ec64a7c2f28d7695","split":"train"} +{"domain":"finance","full_message_tokens":120967,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1181,"sample_id":"sample:96267ab658efb8d92a8fac201cb9180017fde85b35d9d7873fac009ca61db455","semantic_task_id":"627a20ff2230037413c9c71b49a3cedacb26c12fb14cc0f1a329001d10a413f0","split":"train"} +{"domain":"finance","full_message_tokens":123086,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1182,"sample_id":"sample:723cd6b8317d96b35d267aca63af8a9988e9b40afb4793ffd6ebb20e0cc4bc42","semantic_task_id":"6366d0281290cce8e29ad332aa815568f6d52f9d16196715edbc7309b46270f7","split":"train"} +{"domain":"finance","full_message_tokens":246522,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1183,"sample_id":"sample:a43516affc7a39651b93a51e490c36e16d1b5225e17714e11a8f3cbdc287b5d6","semantic_task_id":"66748f64d934449c152acb0d1624844c93ed5518d6278ebab2c2e551bd3d2c85","split":"train"} +{"domain":"finance","full_message_tokens":187262,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1184,"sample_id":"sample:eac413b793833edde07515b3ad7ea603a5a6e0a12052e95d88592d29d054a830","semantic_task_id":"679b27a35b4750810ed45ff853fe4897321774f71b67ee19d136909ba210a6fa","split":"train"} +{"domain":"finance","full_message_tokens":120736,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1185,"sample_id":"sample:5844e4ce8163d71e171906de87029b3a35b8388b16b4968b983878bad123e1e8","semantic_task_id":"690577a6766e6bf9a6ad52ec6d5b09315bee616b7696c78fe4efca59f9df1e70","split":"train"} +{"domain":"finance","full_message_tokens":61517,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1186,"sample_id":"sample:a087bd5e0bc8807708eb6bf65d2ed5b2492fa87a6b69f769399ef12de4980641","semantic_task_id":"6ac80d1c00d14fb7ea1c71bbbd74f1f6dce165647bc129fc6cc83a0721adb2cb","split":"train"} +{"domain":"finance","full_message_tokens":123546,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1187,"sample_id":"sample:0d191f986ae9b85514e5eb416582f2e214a66fb60b24a1851fe70cdf60d80a08","semantic_task_id":"6f84f47d642e8021e4fb94d225d8f66707559183d2335d13d632a58f36597aaa","split":"train"} +{"domain":"finance","full_message_tokens":123084,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1188,"sample_id":"sample:be9a92e800361c614e3979c702da114f913d0a573c9b69b40342c7caa2591789","semantic_task_id":"710b9e26964bc86f316aeb9552a8b0973906ec7e3488cde8f3dd91462a3f186a","split":"train"} +{"domain":"finance","full_message_tokens":182340,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1189,"sample_id":"sample:0811df12b5c07ea468a95c3d6d08ec0cd44ef967e64881735d1b08f9817e1404","semantic_task_id":"72e92a197c4b78e6f7576908b4c8b83ccdc43ec8ed4c4cc7e2590e18362e3019","split":"train"} +{"domain":"finance","full_message_tokens":120967,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1190,"sample_id":"sample:374f9f9e22bbcdc96767ed8a671fd929fea4d189d94b76ab84d4b7c3633d7fd9","semantic_task_id":"7427159bbe78ab98170e7d8144028f5ce740dcb1a67a83c9353310d0fbff211c","split":"train"} +{"domain":"finance","full_message_tokens":246533,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1191,"sample_id":"sample:18efcd0118b440f01485e1fa1c25b0ecd0fd90a041ea9b63cdad17329d719cff","semantic_task_id":"768d92cc9e8b0ed4dfcf42cb0dda8241b2d70f398b5cc952b35410052b89279a","split":"train"} +{"domain":"finance","full_message_tokens":61736,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1192,"sample_id":"sample:46077912a385d5d699d4b6c498ed1eb3ea25ff53eb666bae93971bfe5b7bf801","semantic_task_id":"78a159c47bcc9c067310d407c3c51b7bbbf0f4d6303c5d64ad528ac26bfd7a84","split":"train"} +{"domain":"finance","full_message_tokens":187242,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1193,"sample_id":"sample:efb0d4fd8e0d5e04ad4a667ed9284fa9fbb87b7e118ade3ed1d469b642ddf0a3","semantic_task_id":"78fd6608c08a089916ecf09cd39eb8bdf5ebeb6ca7537fa4c6c91c42b6e0b487","split":"train"} +{"domain":"finance","full_message_tokens":120739,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1194,"sample_id":"sample:93a3d875bb92275b794e3775a380821bb0cfe551126a0f08e8e28c57162c5945","semantic_task_id":"7a6416bbb320e483c677851c3d10b9173cda3780fa8ddcd493a59bc1cba0487b","split":"train"} +{"domain":"finance","full_message_tokens":64358,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1195,"sample_id":"sample:9e1539a3d4be611e3c66d21caaea731240dccb56562142a82a538a14698440c1","semantic_task_id":"7a7571ca60af204fcd035f209fb5d8e10eca6353af8bf6126235a5289fe24cf7","split":"train"} +{"domain":"finance","full_message_tokens":61468,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1196,"sample_id":"sample:6e777dcf8488f85130d05a00a4b839f89a9594d0ffdd92ef07d4ddbdfdd43965","semantic_task_id":"7c58a2e39afb68af7943f4ca8250aac137de653a400121fa3f593aae122f7549","split":"train"} +{"domain":"finance","full_message_tokens":187241,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1197,"sample_id":"sample:954bc8dd751159fac350cfbb475a943c5e1c3e1aea887e6e132b94afe05540e8","semantic_task_id":"7c7a176836a6b5a4443182cb3b8b39526c27260cd7fdcb1cb0cfbf8ee8deafdf","split":"train"} +{"domain":"finance","full_message_tokens":123080,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1198,"sample_id":"sample:211f07f92f2b511ab9b257ff7d5688132abe2903caf1b893db721c0c6fb766de","semantic_task_id":"80188683b32b83f86f3ba7f1f6bf6fc542419dbb045c4568265f9af935a00d0c","split":"train"} +{"domain":"finance","full_message_tokens":246571,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1199,"sample_id":"sample:5d390974d0fa8018e818a3490245ffc81048706fcfbd261118b81eeca233f736","semantic_task_id":"8134365699c2488ded829cafdd8f09e95b72dce811346c9a871d4a835c2b2e9d","split":"train"} +{"domain":"finance","full_message_tokens":125885,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1200,"sample_id":"sample:383475a26602f81e1a3eac569e285ce9fd5eda80887aa8898673400dad231f78","semantic_task_id":"81efde13937e20daad49d8a07cbbc0ea461c9f6462e74c19c1e29582e0290d76","split":"train"} +{"domain":"finance","full_message_tokens":246489,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1201,"sample_id":"sample:7562e05169c3b818aaf9bf34ac737d2d714a16bbf0af10f452f21b39d276bfd6","semantic_task_id":"81f4f8a28ce83c739a3a40bde16047c8fac00248f55e4159fb4cdafde6dbb743","split":"train"} +{"domain":"finance","full_message_tokens":125660,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1202,"sample_id":"sample:283d3985930616d1465ab53895c057771c6c7441805d7a573a9eddf607e48018","semantic_task_id":"821214c4992a359bdde46ebc486186fb7ea29d9a5e68f9462439501c20693f1b","split":"train"} +{"domain":"finance","full_message_tokens":61470,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1203,"sample_id":"sample:fddb410a931f2b399c946b364b987ce6f301030d83d096c921073f09f4e30c09","semantic_task_id":"82937879b676816faade4f45360331592f964398a75a0d2896a0cb7ffb1f73ad","split":"train"} +{"domain":"finance","full_message_tokens":123080,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1204,"sample_id":"sample:dd203d80a8a51ebab699b7789f083071ea02e5c16adeb937b5975fcb5678df7e","semantic_task_id":"8369d37aa16425cfcbc0adbdbbd0265dcf9314027f06944f000c166aa3524544","split":"train"} +{"domain":"finance","full_message_tokens":123544,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1205,"sample_id":"sample:b52ad63a0465454a3a7d2f941f02466fc3e69467e47b78530f3c5122f19f6484","semantic_task_id":"84a7af866fbe4e595465d873caca405af22dd224a0aa0828c05e2bfa64931519","split":"train"} +{"domain":"finance","full_message_tokens":59352,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1206,"sample_id":"sample:001249ad5c8b0d8b7d5dc1bfe03288bc8dd29649a314181155328003af8db133","semantic_task_id":"85ace02c70bdbd9f6bbc73a97ff18693fc4619e36c92d6d1925bb38f9e42cc98","split":"train"} +{"domain":"finance","full_message_tokens":182320,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1207,"sample_id":"sample:579833ae534337175e9549f104f5bb866555a9c983796ec60e99b9c02631428b","semantic_task_id":"86172f75045f3a6304c546bd4475866cc0fe55277fa705d22ac8498a1becc5aa","split":"train"} +{"domain":"finance","full_message_tokens":123086,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1208,"sample_id":"sample:5fbbaad53f02e4cdff78a187bb29b9dcacc2f0204401dc2989a2713ca89aade6","semantic_task_id":"868ae121c2363d810c50c937465f2e0e0dfd5d3c22c2f63ecb96faca74614247","split":"train"} +{"domain":"finance","full_message_tokens":120726,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1209,"sample_id":"sample:488a8cdaa6b81725b4c0a5813dc315b53a7b0bc0aa0da7c844820ad965717c13","semantic_task_id":"868d5e0c5eac682aa92ce743a547b45960f048fd821677e0a9ee9284873f05fa","split":"train"} +{"domain":"finance","full_message_tokens":123085,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1210,"sample_id":"sample:b071eb0ef8d90dc1338f0c6a7f3027c4226b717c08811e220dcc693e12c3f876","semantic_task_id":"883a59391cbc7474dd994d263e934788124ba2f031852a7099e7db3bac090751","split":"train"} +{"domain":"finance","full_message_tokens":125889,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1211,"sample_id":"sample:f3bbefbe02d6c9b027b9e797d2c8df36af26f18a0509195e3f620bcf473c77c3","semantic_task_id":"88746b5b68ac5a21ad124205536692aab8e6712e8f098fd4323dbffafb86dbec","split":"train"} +{"domain":"finance","full_message_tokens":61743,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1212,"sample_id":"sample:2f975930474281bd842f9bdc22e9aaf1a71d05618a6efd10855f73411633d40c","semantic_task_id":"8896572ef1d48b3ae00f4da000d71f2b8201db36e6d40edb56e952e7968fd264","split":"train"} +{"domain":"finance","full_message_tokens":187276,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1213,"sample_id":"sample:bf44186edf006c7b8561f9e74c0f55e41621ce19f1e0002945bbdb1dca841b3d","semantic_task_id":"8a59bdd22ba0a6128fc6f61d0c82e7754d60d29123c3c5e8df573aa96ace91e1","split":"train"} +{"domain":"finance","full_message_tokens":123546,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1214,"sample_id":"sample:1732885cde6d7869a0e6a282012a58bd0f1f346080256515855146f49a762f4d","semantic_task_id":"8df6cd75b94e92f811f11cac74b66a25abf112a9032fdd637298d0adbbacd75e","split":"train"} +{"domain":"finance","full_message_tokens":246521,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1215,"sample_id":"sample:2e371f1f6927b4addd746a7869a068e18b2b5cc6792150dedc09030644284cd6","semantic_task_id":"8f9fc46f59a844329f2ea201edbdbaedf9aee8886fa5b6b2fcec757cc8da7d1c","split":"train"} +{"domain":"finance","full_message_tokens":123544,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1216,"sample_id":"sample:254198982e677972a3b253e04e9b2452fa2afb2f954ee6f080204e84925b14cf","semantic_task_id":"92218d58bf250cf8982a487d21447168f8c6d05932f372b936eeb4347fd1d2e6","split":"train"} +{"domain":"finance","full_message_tokens":182331,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1217,"sample_id":"sample:0253a45385e28cf8c9d51becc55572e28f261ca506d79ac17cd2b92e399c348a","semantic_task_id":"93a73c093544cf6a8e953ae717ba29dea16ec0599f1616599aa5ae5bf30bf0d4","split":"train"} +{"domain":"finance","full_message_tokens":125885,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1218,"sample_id":"sample:c1c6ac0084a360c92a4043d508c55be3620078af66cea8c0bb5075def05834ac","semantic_task_id":"94493d983bbd3ebb14faa61b0b915f001ec7c293de170d2238bdfd0f2bbd5d68","split":"train"} +{"domain":"finance","full_message_tokens":61515,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1219,"sample_id":"sample:dd76e796ccfbb25154d8bd06a91b2d9c09e9337e6ab7e4b0aff4ed4f3ec49bb5","semantic_task_id":"9487e4dba5e90988aa92d823223d7dc5614f99207332f0aa4ce523325d5e04d9","split":"train"} +{"domain":"finance","full_message_tokens":59352,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1220,"sample_id":"sample:acb5b266e1b1ae3c67cddf746d1c6f3381008d1484e44d674d159aa35422e9f1","semantic_task_id":"971ab7c0c37ae4b140cce2528460edaec047ebceb2c5a607dd80aa2cbc282e55","split":"train"} +{"domain":"finance","full_message_tokens":61523,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1221,"sample_id":"sample:faa50e89329ab6d5e1a966e06f63168fbb804a7a01a986f9f3dd76be078d644e","semantic_task_id":"9725950af7b89533fcfa0f77f11eb3c9c2a1ae2eaa807806fea8a1aa6bbddf16","split":"train"} +{"domain":"finance","full_message_tokens":182341,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1222,"sample_id":"sample:6db1904478ef0305340e7bcb1c9d233aa0519024926d5dac07072d4294bd8ab7","semantic_task_id":"99642caf73daa8f6609ce317fa68639b856bff4bfb8af99afe28244d3124d0d1","split":"train"} +{"domain":"finance","full_message_tokens":59352,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1223,"sample_id":"sample:a9b8934c1035c515ce68245a4e492fb823882c0b6ba3f8606cbaf10d34cb52fb","semantic_task_id":"9ba9c39f5cfb2657f51d698f571bd1b6f08c3d2be622c1e00a0a313022d3ec49","split":"train"} +{"domain":"finance","full_message_tokens":120969,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1224,"sample_id":"sample:8b2a2426decd75972159e1ded43ee42cc096b3f572b9ff763613058d1a7da6ca","semantic_task_id":"9ea32d940677078ac2d7b66aa8ae8824a421c5e5bf027eac717390b196d4d706","split":"train"} +{"domain":"finance","full_message_tokens":123084,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1225,"sample_id":"sample:cf38a2878669c332ebd677af3b8fabeb0d369440b3d5829b9bf29986207c8c07","semantic_task_id":"a02231ad88a1ee4700b00a5b52a2f900d9b05371b2d56abc77a4dd81a6168f53","split":"train"} +{"domain":"finance","full_message_tokens":120742,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1226,"sample_id":"sample:c80ac0811c1c0a0e725b73c7053e0f092cd8a41e764f8b604330cc2d865b3341","semantic_task_id":"a0a1af07a4206265f61b06a51dfbda69c3791627f9197725ac536d70ec0dbcab","split":"train"} +{"domain":"finance","full_message_tokens":246489,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1227,"sample_id":"sample:0915a6b7c8caddc9c3b6cce92622bb69f06d703b5bf04198cc20cb834dc3e504","semantic_task_id":"a1e652c1aad1d68118ad083241c6bb814c94096e9222fa05cfb742b797205a5a","split":"train"} +{"domain":"finance","full_message_tokens":61468,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1228,"sample_id":"sample:953bb4a28ad6dfe3af0c1501e704eca9b42541edb48d4e8580e23f7a5e1821ae","semantic_task_id":"a55c7847d27f51b01b873b22e00a4f666ecec809a4843f7caf1a576dd4256590","split":"train"} +{"domain":"finance","full_message_tokens":61469,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1229,"sample_id":"sample:71456719cfac0a06d42816f3bc6d471f744a5e629c5aaf57413a217ffea09747","semantic_task_id":"a5986e190967536e9e5c09808a700e865ab7e25282c7fc7b2f23931e48599f4d","split":"train"} +{"domain":"finance","full_message_tokens":187240,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1230,"sample_id":"sample:8f054d641c51c820c5f7229c82189e882b0edcf4c12658feb4f1a2232ee2a6d3","semantic_task_id":"a667a4530bff61e7a2019b398689fa736a42ae967c6ff7a07291b3b2f60e116f","split":"train"} +{"domain":"finance","full_message_tokens":61697,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1231,"sample_id":"sample:ff2cac7ba864bb1b5f40858f559df295117f02e7dc90fd52eb520b41f22ce0bd","semantic_task_id":"a6dd7b00b28b54442ef0f0f3b3408f52752338668cae90b87c082a7f3e611c27","split":"train"} +{"domain":"finance","full_message_tokens":61470,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1232,"sample_id":"sample:826c173cb391e3d83872fef3bb93a3637db1d40b43a2017a38240c35b227895a","semantic_task_id":"a73db8dd52b2aaf80a66bfa226cc1e3e5ef2d73561b18587e6abe5b23c7f72ea","split":"train"} +{"domain":"finance","full_message_tokens":246521,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1233,"sample_id":"sample:d270e0cb99c5eb549e32a0bbd82a8165db63e95f9e21181d2a193ab11d1d4a41","semantic_task_id":"a854d0860860ad069e606ed058c97854e087531c3c0e3b384b503e24e1a9af9e","split":"train"} +{"domain":"finance","full_message_tokens":182319,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1234,"sample_id":"sample:fab41e386eabff441199751524205b6078ae708abbca738a70ce9c2a5b8192e5","semantic_task_id":"a8ceef64b940551d702be039518b40a864c35ebd81db410acbba549059da3dbc","split":"train"} +{"domain":"finance","full_message_tokens":61781,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1235,"sample_id":"sample:eba8e02a99a9c2c86dc1ae42870bbfb173292d5aaefebf79da83ada3861086cf","semantic_task_id":"aaaa037e6c8cae102e1cd832b2378f8b803902d0e2d002dec7069cf19055af69","split":"train"} +{"domain":"finance","full_message_tokens":64313,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1236,"sample_id":"sample:03eeab3f2c96bf036b8b1a665423d54286d16d686d506d3fbb29fbb0b843b964","semantic_task_id":"ac983fa3a7e8fbbc1c7630565109f7ca47ac9085cd667f3179e7e80df6b29333","split":"train"} +{"domain":"finance","full_message_tokens":182355,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1237,"sample_id":"sample:104950c357588117bda79bfdbaa12e9e31d2827daf21c13dd0e95d538dd1826d","semantic_task_id":"ace679742fd038bae5b41b3b5922b3fe07d565cdc897d6fa29078c5125c54761","split":"train"} +{"domain":"finance","full_message_tokens":61697,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1238,"sample_id":"sample:4d1ad14ae4e090e3a28733bf6423eccc81182a4f3c9a673c2aec097527d535bc","semantic_task_id":"ae9bd387bc61cb5425f653bd32e0773f896b3486622182322184c7521da210a7","split":"train"} +{"domain":"finance","full_message_tokens":182319,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1239,"sample_id":"sample:ab0b3678038dbcfdf70d4557d6c6686dbb74c22fe757ae93ac0c09d77cf0f080","semantic_task_id":"b02a8dea5d9b4b3ab997bd0e368d4d4e3d39a46ea73130555128a4dc473f987a","split":"train"} +{"domain":"finance","full_message_tokens":246489,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1240,"sample_id":"sample:816e988a32491fd52e1b86971eddb26d6197a7f0d502ff7b97b8dae498168320","semantic_task_id":"b1f91b922a990de9493e4ee244bc259f0968279c115277be21c31d5b5d90e863","split":"train"} +{"domain":"finance","full_message_tokens":182319,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1241,"sample_id":"sample:065b1754bd91b931813972fc037e32b659b8af026bbcfd84448aca10d75c3a70","semantic_task_id":"b2d3d29ab0681f680c4281a234c800d4c1f458b27393766235ea674ccb6dbcca","split":"train"} +{"domain":"finance","full_message_tokens":120741,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1242,"sample_id":"sample:37f90c7e5a8e7501e996f596f4b252714b8a956809bb0b69e1c37fae3dfd57c7","semantic_task_id":"b515d6ff5ec6b38960cde3f94afefaf09aba84b351f55a8286e581dc35843a5c","split":"train"} +{"domain":"finance","full_message_tokens":61507,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1243,"sample_id":"sample:7f833cf87a9596e2cd86a4bbb481dd38d1f2807eaf156976c15fc8034b66dd0f","semantic_task_id":"b5f00254756ed9165d31eec20a7f90063d564f888f192d4b831c2c6186860a6e","split":"train"} +{"domain":"finance","full_message_tokens":61752,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1244,"sample_id":"sample:529dbd80ca95d5dde5d37ebe32e5ecc83c1d37bdb543117bc7e9571e32886402","semantic_task_id":"b8c672ae4b6640b31c74446710b0c187388234d023ebfe5cb53a461d373f2829","split":"train"} +{"domain":"finance","full_message_tokens":123080,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1245,"sample_id":"sample:f21de83b70d0728d80f585eb008b44e189a7a398629b5d0b8b1e84888f18af6c","semantic_task_id":"b99831e1f97c1ecc9288ca7abe98b6be3961e37008808d4cb314c0f38bb7d72b","split":"train"} +{"domain":"finance","full_message_tokens":120741,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1246,"sample_id":"sample:190ada1b663d89b1d9d9345a287193fdc73ec12355c2c2c911d48328db415f72","semantic_task_id":"bd4447ba4ba781208251707f56afbc7bfcdb77775cfc6800e903324a35e4ea7a","split":"train"} +{"domain":"finance","full_message_tokens":120968,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1247,"sample_id":"sample:718fa62a99b46edadb4f347291ddef1ce8f77a278a9c75bdd9fd6460bfd20541","semantic_task_id":"be6c48a9f8584502174c4a64091e178bc99b6381405a9f758f929ea751b6c9eb","split":"train"} +{"domain":"finance","full_message_tokens":123071,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1248,"sample_id":"sample:efe07868b9ec84fe9295129751083ed516b06040a422be573e7b84086ce271d8","semantic_task_id":"c31314bdce295cb3a76c97d1d6609165742dc5269a7a0cd1e9e2cd8bd8aa5c7b","split":"train"} +{"domain":"finance","full_message_tokens":59351,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1249,"sample_id":"sample:8b596d3c28139f9349af44f4fafe0551893430d1cbfb6cd77df989b025ac4677","semantic_task_id":"c41c89d6a8a2bdc3861ccb9789d8f806ea87a2678e6659f21eee82efbcaeb7d3","split":"train"} +{"domain":"finance","full_message_tokens":182319,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1250,"sample_id":"sample:90ed1687a3ef1dc5c805931ca2eac0c504406551e4f9e69823ab4a0903170d55","semantic_task_id":"c4bd7763d2e359ae6dcada47a89530af163b9fe522ad8c3b7ba67903eb25c418","split":"train"} +{"domain":"finance","full_message_tokens":182379,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1251,"sample_id":"sample:3db8e20aae0b71a42b15b164c01c69fda35ece6c7810f5e389e618d3e98c4c57","semantic_task_id":"c55eff44904c003d83ccd6523d452a30c354a0625e73d6a508dee96dcd4da814","split":"train"} +{"domain":"finance","full_message_tokens":120737,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1252,"sample_id":"sample:3956db8d524824ce6f1ab6aa39d7ee722f2100c9120b1baed6339862c797dcd7","semantic_task_id":"c5849a32de3cf1fe4dca7e5e16fb911a451807b47528c5b2651c7d58a18cc1c1","split":"train"} +{"domain":"finance","full_message_tokens":123081,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1253,"sample_id":"sample:6a6b1639d67d237a8d56cd5749e12f497dcc2ed2a7d700d5f964900ed6b117f7","semantic_task_id":"c7127e3aac28f78c69b05eb90286cbe6bb03961e5bcc63b416e31e09bfac84c3","split":"train"} +{"domain":"finance","full_message_tokens":61552,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1254,"sample_id":"sample:8e096cb1eb9e494aa683e18c98ea5d70c0bc2801eec0c70ca70fb67949572dd5","semantic_task_id":"cb09754d3fbd3c68978a2ab17b8ebf8f5638bb6c1adfc1d805bcd251ea853cd6","split":"train"} +{"domain":"finance","full_message_tokens":187301,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1255,"sample_id":"sample:595be8fe50b74d7fa47f410169540653f634e286ddf06ef06d4fdab40b64a747","semantic_task_id":"cb6319080516a445f7799a311767aeea2ba2609c52e39570c1f4a1bda97b8ace","split":"train"} +{"domain":"finance","full_message_tokens":187241,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1256,"sample_id":"sample:c8a8a8ed66cb6bca6f9e23b8a773546527d0edf066d2b5c427bc406341b402b6","semantic_task_id":"cb8064c7c6505b33bf7199d6f8bf26e62562947db8c1a182fa35951436ea2ad6","split":"train"} +{"domain":"finance","full_message_tokens":64273,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1257,"sample_id":"sample:690436c7a32327df3366fdc302991d91cb52b583b4755c075d6ae26d0656708c","semantic_task_id":"cba68d4590df3c4e495686e7280030aba29c68588cd465b2bdc2738d154dc585","split":"train"} +{"domain":"finance","full_message_tokens":125884,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1258,"sample_id":"sample:b6e3ef6614b9b4791cf27088e17e97247b6b0a90118548c53691862b3a4b252e","semantic_task_id":"cbf6c810cc03f9063cb888abcb9661074b059329dc930d3ee7fb5fcbc668e286","split":"train"} +{"domain":"finance","full_message_tokens":59434,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1259,"sample_id":"sample:620405c2365a5ada63de76ab5fa767b10b2b0c21c01a20065c4aec96fb4e8830","semantic_task_id":"cd68a0048ddd6321c8025c583e5d5bcd6412ab9a88741e643f62d2d184ea01f2","split":"train"} +{"domain":"finance","full_message_tokens":246490,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1260,"sample_id":"sample:10918252363127282611bd0a90ecd098ac301ce7911c339a9e3c61eff7d4e984","semantic_task_id":"ce6aadb8e5ea4951d7c585938057435014ee2979ef8c5ea5d61580cc42edda44","split":"train"} +{"domain":"finance","full_message_tokens":123081,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1261,"sample_id":"sample:79623e88d5f03bbcbe919ddd4e385f5e3d21d277263cf83b2b6f7a92b9f00b2f","semantic_task_id":"cf03b4760fb9211e249cca456fee00ac954418b991a2ed39d0d9dda77451f305","split":"train"} +{"domain":"finance","full_message_tokens":64273,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1262,"sample_id":"sample:e0d94b7f39781d5edbc5e331ce5ccd829a09558541c92dd85800b297966760f5","semantic_task_id":"d615c8e6040285c9e5db7cff968a7cc8f7c973ccdc6ffe897f8535516ca3c476","split":"train"} +{"domain":"finance","full_message_tokens":64274,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1263,"sample_id":"sample:5eb66c829e30dbeed7b4336336a3c87035fdcf77f7f8080ff2337fd4672a91da","semantic_task_id":"d6162db586b3357eb0df1a8c0d7e5092d7b61aa3635fa95f1984655f549c0c31","split":"train"} +{"domain":"finance","full_message_tokens":120736,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1264,"sample_id":"sample:b685a0a5dd8a5f21bcb3a8413894017bf9cd1571af99e083831743455f842ba8","semantic_task_id":"d6bf6c32b2acbe8cce94c359068a36b5d0708d1baee580ed6235f7c6a3b64bca","split":"train"} +{"domain":"finance","full_message_tokens":61696,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1265,"sample_id":"sample:d2476e72c93b6b246feedc587f4f3f0fdf8d313e9314078305277fb6d7784495","semantic_task_id":"d7fd979a9078cde15ac5b7d3ceedc1de201ffd7260fe5697413e14a8bbf7e64e","split":"train"} +{"domain":"finance","full_message_tokens":182340,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1266,"sample_id":"sample:4f9f8f5f8195caa4127f0ed46aa9f940e24f23ba6155e3d31475e126dab887dd","semantic_task_id":"d87cc31908c152f66ef1c52569d3c0404551ab149617dbda274be5acdfdb5913","split":"train"} +{"domain":"finance","full_message_tokens":61697,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1267,"sample_id":"sample:c1f49ac074425e2491c8c7b2dd8805f5e61907e2ff2748a7f5c5ee285cc82626","semantic_task_id":"d8dfaca4636bb1da539b4565e4bb62d737534afb0fc408273585ef91fbe395aa","split":"train"} +{"domain":"finance","full_message_tokens":182320,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1268,"sample_id":"sample:8fc8d801b0ee48316e40c8d8b13b5ac475189d76b85bab293d5adbfd1e2f5fbf","semantic_task_id":"d8ee797e8e4f27c99ac9db5955965869d94355158edd1473b6d6b24d7b4cc658","split":"train"} +{"domain":"finance","full_message_tokens":182318,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1269,"sample_id":"sample:54aa3cad48e606cfbc0e55f0aa8d1df5c97b6c76847d1cba332189f97ee688f4","semantic_task_id":"d9096b2ae397b31a7c48aaf716bd8080e90c7fff4dcea479541b9ae366054935","split":"train"} +{"domain":"finance","full_message_tokens":187240,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1270,"sample_id":"sample:5eedf709948673cc8e90624697aa34b43762dab8486003ce9f6e028f63d58682","semantic_task_id":"dba620d032cb662d9880572f3d7b00502119822305489ec723a013b6f238ea00","split":"train"} +{"domain":"finance","full_message_tokens":123086,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1271,"sample_id":"sample:16a07540aa2c1a1750117b45ba5c26bf8404c59a6d80f837cf9d4d05a7908421","semantic_task_id":"dc5d08db190d08798ef94f4eff989fee33a167b1b6e6ad69a8a365e05b6430c9","split":"train"} +{"domain":"finance","full_message_tokens":59351,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1272,"sample_id":"sample:cc2a0170752de2e1b938693e9acdea2d38b08976a393358f99e4f64ccc8f7d7c","semantic_task_id":"dcd9bee1829d93be967b66ee7fdfe171fa639d923be7064e93d6a66e86dbb032","split":"train"} +{"domain":"finance","full_message_tokens":64329,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1273,"sample_id":"sample:e5a41523cafe50f1364d3e05c484891d0b899594f0439be250a3acf6b7ddd3e2","semantic_task_id":"dfd52d5da328ec6644c1b32dcc874c965fc014b581139a227f56f808f616a35d","split":"train"} +{"domain":"finance","full_message_tokens":61469,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1274,"sample_id":"sample:957b7a5fa80ce1b004edac102ae934d9d26661070d0a4e2dae95a3e6446ee05e","semantic_task_id":"e0f8f45695b3a557afc43fd5730dc085f3e07101542570e2fbfc07401bbbec32","split":"train"} +{"domain":"finance","full_message_tokens":187252,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1275,"sample_id":"sample:97b4727759a649d0abd5d4e4c92df79e3c784f2d5a5b2a1f935b667e9d0f56c5","semantic_task_id":"e2715087ae1bc1433ed6b122842c2e171109bde6ec0b28d73f1b0650ba7283be","split":"train"} +{"domain":"finance","full_message_tokens":61695,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1276,"sample_id":"sample:58b95c932b8d864ed1b024c5bb66f4a74bd8ddb1ef39eee468ffc411a71e8d76","semantic_task_id":"e3ca7e9d173aeda88a456406b55493b1ceb70ab32bb52eebe01307da0bcc7050","split":"train"} +{"domain":"finance","full_message_tokens":61745,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1277,"sample_id":"sample:f100ec2274cae6418ae16a0a5ff636d0a9a4f41bbd919cae581785caa0b5461d","semantic_task_id":"e3f5ff11081a390df88fd1e8e40a0b2e299984c37fc32d7191d7b08e194da565","split":"train"} +{"domain":"finance","full_message_tokens":59388,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1278,"sample_id":"sample:59e837bc1b0fa2f9820934a59dc0247d5178c95265338a340bdf244a2243f7da","semantic_task_id":"e58aad733ef7090f8fd163cb4577724c677bd74bc652559ed59410db0378a145","split":"train"} +{"domain":"finance","full_message_tokens":246489,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1279,"sample_id":"sample:75cbc8703d165c1634574c7d06ba16eb9987c3ea91dde2db9d249839249f33de","semantic_task_id":"e612cb9142091f95f51c566084516185a999877085a6fb3ebf83478a3c064781","split":"train"} +{"domain":"finance","full_message_tokens":123085,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1280,"sample_id":"sample:332e8281ac7603c964f58dce93fcb54bbd505e4189207dc53bac154510b19b54","semantic_task_id":"e6ed1c290a038dfcbf4827cd6ea94f98321e5430f75914ca082ea69c3509ab49","split":"train"} +{"domain":"finance","full_message_tokens":120969,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1281,"sample_id":"sample:00210507dfb0b99844e36f691f3f9bc3930bea4a6696e336611c278f06370418","semantic_task_id":"e6f1d40c7a3de7e880d17f59b2b431e94987a572efd015fdf4282de6dee61de4","split":"train"} +{"domain":"finance","full_message_tokens":187262,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1282,"sample_id":"sample:8531edb4c0665a0ab66106ebfdbdfb9a4432180499ac64633a05ea5770d8d91d","semantic_task_id":"e9e388fb6c50b78cae2dc0b2ee012e60fcdd446f159f9411ecfcd36c55c95b93","split":"train"} +{"domain":"finance","full_message_tokens":246534,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1283,"sample_id":"sample:ba906380ce90acbf5c8ca20ea813486bf3d9d5ad8ac28e9ae7012e9858d6ccb0","semantic_task_id":"ece321660145774cf8498459c6f18f011a40a0b90f5e03aa95b7b5f28eebcf7c","split":"train"} +{"domain":"finance","full_message_tokens":125884,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1284,"sample_id":"sample:7682366e7fa00e9302ef4f80a467839097cefe62f1848ce129cc700c4d3e64d6","semantic_task_id":"ee16f63cf44388d92043ca059533cf3d7032aa51f84d92fc07cb4a689d07a510","split":"train"} +{"domain":"finance","full_message_tokens":59353,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1285,"sample_id":"sample:cd3b5f92eea8e7015104835a692067e2c45103d32634de50757a22ea42ce9d40","semantic_task_id":"f182c05025e5491101733b1bf6acbb69ccbac67706634e224b661b8655b7fcd6","split":"train"} +{"domain":"finance","full_message_tokens":123070,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1286,"sample_id":"sample:aa21e1b906fd32449add1da759c3d0206ffffa027d92d8f3c075313dba7a9bec","semantic_task_id":"f3491f70aec65112bec504fa741a21f2bcc17afffd75cbc0910790da5d45fb4d","split":"train"} +{"domain":"finance","full_message_tokens":187262,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1287,"sample_id":"sample:0354bcd559d2f2cbf1d5857e5b1e510adf03f105a125f17270dce43b8cfd3425","semantic_task_id":"f5ec791750c1193c6fcc2eff164455aae6b17d924541f42a71e717e9f28b37c2","split":"train"} +{"domain":"finance","full_message_tokens":64323,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1288,"sample_id":"sample:ea3b51edba0aa6190baab93cc477e18ebb60f4c6ce70388cad866aaf0484a252","semantic_task_id":"f5f8fea488f26d5cd1bc03b7f7d82ace05535c99a112713a6978b175ac6ea9c9","split":"train"} +{"domain":"finance","full_message_tokens":59400,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1289,"sample_id":"sample:3944eb144355fede9cb8d24b3ad5989ae235762c43e353b9a1b6f6f934a3382d","semantic_task_id":"f769093cb992913e4ed3e5076cd8ced5a0a756594443f7eef69b86e5433aa316","split":"train"} +{"domain":"finance","full_message_tokens":125662,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1290,"sample_id":"sample:2f976e5e87112230767837c9fa1d627893d95a77f5364c436ac01e4b348c1cbd","semantic_task_id":"f8d0436d119cba8a3bd1bc58cbb77c428c03013e753b8b598947b3daa2302eb3","split":"train"} +{"domain":"finance","full_message_tokens":246491,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1291,"sample_id":"sample:d04400c2d95dbcede01e22443c426e91577c7ed34391c9ba6bc2ba17e5727974","semantic_task_id":"f97777831da06fb94f0c96fd887e828d4cf0de891ce3655c74b119cf28364bd4","split":"train"} +{"domain":"finance","full_message_tokens":125663,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1292,"sample_id":"sample:6d493ce3e0b11df59bf3e227d4f68eedee1a4a32e174581b7804f5f44ce1be6d","semantic_task_id":"fa3bbc4022d2954060e87bebc7b7c247d5d47428d6d7324cdaa0401ac372c738","split":"train"} +{"domain":"finance","full_message_tokens":59352,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1293,"sample_id":"sample:3a665ca459777be10e3dffd306c2ec8c850cdf1d814aec0bb167e6b0cadaa0ea","semantic_task_id":"fb7425378d1f738dec1904ee288600d16ddf399cfef372a434604f0c199cc94a","split":"train"} +{"domain":"finance","full_message_tokens":61696,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1294,"sample_id":"sample:a968d693d50c751f94e3d05052a07b4dee9d7c38591a1afed947a4c84b1b2957","semantic_task_id":"fd5ccf64845c9c93fa3a1a6a57e28de100da13e6bf1facbbe2bbf037660623b4","split":"train"} +{"domain":"finance","full_message_tokens":120742,"group_id":"0001045810","output_file":"finance_train.jsonl","row_index":1295,"sample_id":"sample:c27650deaf901608ef3dfae37ea30cb98c15ca57474dbdc3dc0fb7e23c4bb6e4","semantic_task_id":"ff7fdb4339758e8c0cd3831bb07e753a8ada497e5429e91430779ad80a1d9d3c","split":"train"} +{"domain":"finance","full_message_tokens":95111,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":194,"sample_id":"sample:2d2803f2fa1266d60a5ba0299a4f0cc9b3b24746bf49cf13b371674f72b09965","semantic_task_id":"00d9be7b081ab9b2251499d41a74214a87fca9b17f71bcc8b7d68a928551f6dc","split":"eval"} +{"domain":"finance","full_message_tokens":206419,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":195,"sample_id":"sample:a327fa6587342b0f512433346f0c0a9c72c0181ccba65e297ee037b1f12855a7","semantic_task_id":"02f9e2519282e9b2870df3b22e07115d081faf6a088ca633679371f4adbac26a","split":"eval"} +{"domain":"finance","full_message_tokens":109833,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":196,"sample_id":"sample:7a593373aa76339e61daa4b77675b9c4b5fa3bffc36488b04b7e009ae761d07a","semantic_task_id":"048a9fd419ad9dc5fccdc7ce5ff7cfed83ad96b802d40693268d173440368157","split":"eval"} +{"domain":"finance","full_message_tokens":201707,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":197,"sample_id":"sample:5bffc9f2f411fa09599e027b101b1d0240a5d22ad06336d748cb9cccb20654f0","semantic_task_id":"05acd48bb4dc45fa75a368f099143c3048568aeb7cea7e05ce69d9845305e904","split":"eval"} +{"domain":"finance","full_message_tokens":95115,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":198,"sample_id":"sample:afe6ebd7abf27920634f3a9ff1635da885f41da954eb4a4f27e47ba7c4b0cbc7","semantic_task_id":"05c1fa2f09f6f74bb14e9aeec9fe99ab2ef861856020c792aec25d437abfc00e","split":"eval"} +{"domain":"finance","full_message_tokens":201671,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":199,"sample_id":"sample:afe39b559b8b8216670384f0478494d212a8cf46f099539e081321f147297eb2","semantic_task_id":"077562f237dd7f5fb5709cb530e5c5af3d9a72e62c84c2369764d9ec1682f38f","split":"eval"} +{"domain":"finance","full_message_tokens":96723,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":200,"sample_id":"sample:c6fed15233f98b6b3f5c57e34d6f63787320fe20efe5e0106579e4bc46266334","semantic_task_id":"086b4f8593ff35a740542aaf15556dee54469fa734710087a8253115ee9f8903","split":"eval"} +{"domain":"finance","full_message_tokens":109838,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":201,"sample_id":"sample:0ba507e8bd191ec1e340609af9014ba6413bd153136d85425cb43ec33d50eb72","semantic_task_id":"086e93625cd5b5382826d37657434b9b1ea3b9b34de316780c2c4c4eaff21c11","split":"eval"} +{"domain":"finance","full_message_tokens":206451,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":202,"sample_id":"sample:c42c3051a29160b08a27c28420d5549f66a5fd08f484acd31ef73e1553544982","semantic_task_id":"08a62f8f1d75a967dfb3bffa14ac020cc48327c15326b3d5a7a35b4056188217","split":"eval"} +{"domain":"finance","full_message_tokens":206441,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":203,"sample_id":"sample:4e041a6fd28a99d788709890caf856e5c7790dbc7fcd059b10bcdff38d46d2d6","semantic_task_id":"0a510ff9e3e61b32492dc5dfbca11fe57939bda0a03388b9a13aafaa0baae027","split":"eval"} +{"domain":"finance","full_message_tokens":109834,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":204,"sample_id":"sample:f0609f96b96183a68a0db790c0635530ef556d64915af5d034c099d683756316","semantic_task_id":"0d0782b4d5bc2c428f312b2ce793a50f45c406fa202e7d9b9e1ff51ac7a8d3ad","split":"eval"} +{"domain":"finance","full_message_tokens":95111,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":205,"sample_id":"sample:3ed4738d1e9b241e00b8fd84d0ec7cfdc4fcc4baa57d02f785c0767c0afa3980","semantic_task_id":"0f8f809df7f1a5df77e4db0e43af669a4a4de795bafc72a4031ee776e6b91b4c","split":"eval"} +{"domain":"finance","full_message_tokens":93567,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":206,"sample_id":"sample:aa84524e75feb8182cb1241fda9910ddb2d8050a77f1f5386b8da854bdeb1185","semantic_task_id":"107e01b9b7a5432f94735ccdf8cd802852d5b093adf14f70fe3c58c1d2414933","split":"eval"} +{"domain":"finance","full_message_tokens":91568,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":207,"sample_id":"sample:9fcfb81efb73945b4d89d185d511a08aadf0f275dae420ce180a4b37d41303dd","semantic_task_id":"10b421e263bc25af05391534cc716ec061f9c7182b00744545b294af8aa150c5","split":"eval"} +{"domain":"finance","full_message_tokens":188515,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":208,"sample_id":"sample:f9eb58b16eb553247e86beaafa5b0f2b47282ac5c64c987fc3b965c43ace67b1","semantic_task_id":"11221af8c487cb759cd640dd9a71ffa605da247438f912b2c3ddb77b40e7a190","split":"eval"} +{"domain":"finance","full_message_tokens":95113,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":209,"sample_id":"sample:b2812a0dc519036b3e4bbadb4a5560ab1751954b800bb4c454c7ab09eda4cb5e","semantic_task_id":"1217adbf71e16b452060685e628be24bb8fa53f9f925ad20b5623e3f2ca5aafa","split":"eval"} +{"domain":"finance","full_message_tokens":95115,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":210,"sample_id":"sample:fc49abc466392458b00178b35f7f97cd5564f261f8f8eda2e9894eb29958493f","semantic_task_id":"1437e13de01cfcf7180588b4350a5b961521e4d90ce15d045fda318ed19c817a","split":"eval"} +{"domain":"finance","full_message_tokens":96682,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":211,"sample_id":"sample:55450fb7b935579853e6c4b5c6fda5006237aecacb6b2e83dcc0da4217123a87","semantic_task_id":"1713ee2cdb3153b4f9039cd830a2c4176587c8044fb1576e13801c98c76ace5a","split":"eval"} +{"domain":"finance","full_message_tokens":206452,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":212,"sample_id":"sample:ee70cf27014266803c4cb1b5dd8a6fb2989484d476e33487208a1cb2e9acca48","semantic_task_id":"172586f7e0eda6846ebd79f295f2a5576f9a4e5db7248f822d1e5255afe7458d","split":"eval"} +{"domain":"finance","full_message_tokens":206421,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":213,"sample_id":"sample:28b6e224f5ecde75e289f1c40240cf020ff320ed21971f16eb0a63631a5845fd","semantic_task_id":"17329b294894767724727a0edca98956757079f35f14bc3dd8492423c3481c20","split":"eval"} +{"domain":"finance","full_message_tokens":91487,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":214,"sample_id":"sample:b898253fdace25a0a55e4c503e2e188c70a256e3e8a9118c7250f085896ee91f","semantic_task_id":"183fc113bae2be2d9e2f6fbd381736e22fbe29bcc8f500d848ad923d9caa7895","split":"eval"} +{"domain":"finance","full_message_tokens":109880,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":215,"sample_id":"sample:9f49e37381550245e8f746be8182de5dadacd660958776ff3cc1c3d81c2bc598","semantic_task_id":"1952c5cbe0766441822fa003961be917e6232df1e30ada22dfb4f96eb1fc2e48","split":"eval"} +{"domain":"finance","full_message_tokens":96710,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":216,"sample_id":"sample:01cd07ecd6aa14d9005c34413f6c5bf2ff02fa2e9d8168017c5c62a2ef7cf427","semantic_task_id":"1a1e37f16c0f6ef82bd703043e35e90b15c4f1f1b78f51d408ad0faf3198d231","split":"eval"} +{"domain":"finance","full_message_tokens":95115,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":217,"sample_id":"sample:e66c33d6507f176390d3c2c3ee558c4d2ae46d742bdffd1fefa60a99c7d23fbe","semantic_task_id":"1b59ea1c414a53e5f348fa35d7740ec7d3cc5dcf235069de16d64b7471c50ea7","split":"eval"} +{"domain":"finance","full_message_tokens":206419,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":218,"sample_id":"sample:2910c4bcb195f30895277a19ae0683fdfb4c4556f764ed4e081f3b7f91deeb26","semantic_task_id":"1e64af313305550c22fb0223a23b5d7b6f1ba471575c6f3d81b266287ea51e06","split":"eval"} +{"domain":"finance","full_message_tokens":109872,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":219,"sample_id":"sample:4b389529b0fc37e5353b6f8432c72f46a6b55e7cf31e6a25c1108451791ed110","semantic_task_id":"1e6b91cd23149b7ad61da7f55b9417c03768575e6e8b87f86cacab455c6d482c","split":"eval"} +{"domain":"finance","full_message_tokens":206418,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":220,"sample_id":"sample:abb0a157f584ca555d3abe9c584418fd8dcbff4c623358e4994e0ad724809a8e","semantic_task_id":"1eee04495322b2de69a57edb9e61cae5a94c3582e4df116fd91dc6b7dd91ce70","split":"eval"} +{"domain":"finance","full_message_tokens":91948,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":221,"sample_id":"sample:fc47ffa212bec4c11d87b4a8d7b00ccb1f551ad160b6c3c8b4fc6930d38df7e6","semantic_task_id":"1f469be13447dc641e0828e9fb9c24d18025fe556cb9a28ea40abb1804d3b41d","split":"eval"} +{"domain":"finance","full_message_tokens":188537,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":222,"sample_id":"sample:c29dcc01ab94aebd3b51aa216c3a0254897281a122d1bbbeb8805f62fdd53e71","semantic_task_id":"2146d11728c8d00f9ec92a0d4bf43a614c8adb63624cd5ba26684c9381bed1af","split":"eval"} +{"domain":"finance","full_message_tokens":109838,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":223,"sample_id":"sample:b8c3557b51920f55fa61ef1ce8d193c6ddf469397d6fe3233e0cd22ad0963ee9","semantic_task_id":"21a8162e5ababbe60fc108fea6b44008d29df12fafebbf5aca1219aa29e6ed6e","split":"eval"} +{"domain":"finance","full_message_tokens":109872,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":224,"sample_id":"sample:f3d4873ec90f95aa3ae8f5dafafea8ee1f13573e28c37aedea5a60cd493a3799","semantic_task_id":"24bf78039f3a6873662dba83f5760729128c8819cc59e99e13ba8777f0c1f69d","split":"eval"} +{"domain":"finance","full_message_tokens":93521,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":225,"sample_id":"sample:cb13e047d4a6ac52243af529b8808bb0fc3e9baab283e0c89ceb91200d6bbf2f","semantic_task_id":"251d4b4e000ee9fc6a981058cfbf01c467adddf2648646251b77045c126d3de4","split":"eval"} +{"domain":"finance","full_message_tokens":206419,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":226,"sample_id":"sample:076c0e5e75d65e963aedc50d847671b1f8ed53b30ce0de7a3ebacd45c847ab38","semantic_task_id":"27c2298279d773e7e7b3d41cc3813b09a46c259a8aa7fb3a7f74520ef91e06b5","split":"eval"} +{"domain":"finance","full_message_tokens":91911,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":227,"sample_id":"sample:e5876258be762e37f4e46662963854ab4190007117188320d36c97a9c9fc40f9","semantic_task_id":"2910ef7b9b74208d0a6621518fd0100e29ad19c0cbdda8d1ea1d72f5ff429d3e","split":"eval"} +{"domain":"finance","full_message_tokens":106677,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":228,"sample_id":"sample:f340d9f38a5f5c45bb35822dd1c97d1321ec89a587e4f398e351fdec2b310acd","semantic_task_id":"29538c22a35f10f26c0f09262397acdd36bacefe19f075b796b9f23481db52c7","split":"eval"} +{"domain":"finance","full_message_tokens":188517,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":229,"sample_id":"sample:f2052cae73d90ac92569575d998c0910aec74a41258ecd6f043608805644b2d5","semantic_task_id":"29b2e97eb0e97ddde7a61c9d99015cc418716bda6943628859195a451e01c7ee","split":"eval"} +{"domain":"finance","full_message_tokens":201725,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":230,"sample_id":"sample:6443e79915f21b5826208ed9ee67068dcf25b6598a496bea405243ebf7415809","semantic_task_id":"2a7ea77f0eb5963168df5da75bd71e8b25a3657df346a4c8715630117c2834be","split":"eval"} +{"domain":"finance","full_message_tokens":91540,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":231,"sample_id":"sample:74ccc8b106290709582402529659425fb6135b494e8555a9bcce237461f03d02","semantic_task_id":"2d681d5e7941f972d694e72e7ec40c0ffb1aec7e0bd797eb86e1af445e32de10","split":"eval"} +{"domain":"finance","full_message_tokens":95111,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":232,"sample_id":"sample:95f67a0f4944d938dc0ab4ef24aff7813fcc1252c2f641629d2f742de099d0b6","semantic_task_id":"2dc672249ae3f53737a158e2376b21c2225742ec85181a11e0b2ae64b56a28a8","split":"eval"} +{"domain":"finance","full_message_tokens":109823,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":233,"sample_id":"sample:2074995ccaa6639bd974f33d6f8f7b0ae4c7f457c7a45c6e7489e02aa207cefd","semantic_task_id":"3067d00986e08f55cefac2a1a52bbfb6fd34fe2317d9068d6abf2a5f4a983a36","split":"eval"} +{"domain":"finance","full_message_tokens":91489,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":234,"sample_id":"sample:81076a90e9f75c18ae7fc6fc33ae3007f1d95ba22a2ec11251f44831ef92b399","semantic_task_id":"33031de0fa997dd86f944c76654945612c817a3d7abf22483eba7df7f5d438d8","split":"eval"} +{"domain":"finance","full_message_tokens":96725,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":235,"sample_id":"sample:902d5b07d7f56ce30153740669cea438111c8ae74628f0db3f0bc8a3c20cae10","semantic_task_id":"3668d46bbd2fb92e78f1f04427a04abb3ea46eba3b9c908b56c701efbde3beea","split":"eval"} +{"domain":"finance","full_message_tokens":95100,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":236,"sample_id":"sample:6e7d3f4e36df20a3cb7a1c3c9d20f047a5c323d167401a9e9f8428f7cd46cd17","semantic_task_id":"382eea11ae45db5b06ee60553862f691f2e3c91482e1b6993de79bf0ac1a46ad","split":"eval"} +{"domain":"finance","full_message_tokens":109834,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":237,"sample_id":"sample:dd523be5af6ac1cd380d1d55c388879852b097275332ba7128aa45961dbf39c8","semantic_task_id":"3a45b82f85a4253b5a24bd5578421435e49dcc03dd9ad30d0b29b70ec0b62de8","split":"eval"} +{"domain":"finance","full_message_tokens":91488,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":238,"sample_id":"sample:5c950b61a28cc4f9fe22a8effdc84fa7171f72cf5da48248b22dc0ea9d1ae385","semantic_task_id":"3cf175957b620745ffab95a86b8fbf3f6f5bcbeb1476f29577c78aca63d55fac","split":"eval"} +{"domain":"finance","full_message_tokens":206494,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":239,"sample_id":"sample:003082f17e031e1897c30fd14b0e0d977ecf07310989f1fa3a2c28bb7b41c7f5","semantic_task_id":"3d10e550896912c3caca20f4272b101e9f2c07eb29c1401fbc7adbb8e35586e2","split":"eval"} +{"domain":"finance","full_message_tokens":96723,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":240,"sample_id":"sample:68cbd400611209ac86247959cdeba17ea96780cf11c856f6d9b818759e06a2a1","semantic_task_id":"3d1abff5770c128fe227a9abbc1bdebfc2e2c3a955b262c943519ee664dfcc94","split":"eval"} +{"domain":"finance","full_message_tokens":206419,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":241,"sample_id":"sample:c061fc77a6f4839f9e35cac9ec6b1a4dc93177ebee40a0092734f83b4f124e8a","semantic_task_id":"3f2ba2d66c6d28023720a2fa9062139b4ac9f4afed0c2ee35a7fc25e5b382778","split":"eval"} +{"domain":"finance","full_message_tokens":93558,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":242,"sample_id":"sample:51d48993ee55e0f10abe7e57864be410807f96fdc28e627cdfe3610d7ac0c010","semantic_task_id":"41361b6cc37e3edd4d8fc7b50393c9fce4c0a9abb1cfc6a5b12f52d54ff773e3","split":"eval"} +{"domain":"finance","full_message_tokens":95109,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":243,"sample_id":"sample:97f61ac3613a857d1f9270069c671d1ac6c16143b7e94779e95129bedc873c8f","semantic_task_id":"4136f252cd5f3fb39d940f900d564573a4205752d783640668dd6502d0912a43","split":"eval"} +{"domain":"finance","full_message_tokens":91525,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":244,"sample_id":"sample:e13d4f43183b4a0a5c7594bc75d3ee9438b4bc15bc42c943ab40376da73c13cf","semantic_task_id":"423d55566c8f9f9f42cfe3215f76b830fa74d58c77c89193e56313eb783c8c18","split":"eval"} +{"domain":"finance","full_message_tokens":96720,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":245,"sample_id":"sample:d567c3a451e53b625fbd9c77db7ad4bd0253ed5886bf8669607a3969da531b9c","semantic_task_id":"4301e6767955320f18a861d1c4dbcc1173ef3e8f16e8eec017ac1185d6de6c1e","split":"eval"} +{"domain":"finance","full_message_tokens":206439,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":246,"sample_id":"sample:5f72d389be77dd69b37aba901661ee5792d21acd2aba88f81e1aa4219c96f035","semantic_task_id":"448da1b335aa9c31cd9855bf61f2a6e7670edac50e745af3c466116ed761b41d","split":"eval"} +{"domain":"finance","full_message_tokens":201671,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":247,"sample_id":"sample:2109c476545dbd72de08aa7ba4da8ba0a1584ef4cacf2946f10035043e89f693","semantic_task_id":"46217cb5fcf2102aaba9e7b8b71692a2f9bdd95273d9ef634d7c23437864b3f2","split":"eval"} +{"domain":"finance","full_message_tokens":106678,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":248,"sample_id":"sample:3ac7ee4f7c4767e902f8806aa7e0ec8f9e11f2d753aea1ec5a3df58b8bf5d89a","semantic_task_id":"4867858d6544f11db52427a7a74823b68460ab7d6b9d186767b89373d48a4813","split":"eval"} +{"domain":"finance","full_message_tokens":106713,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":249,"sample_id":"sample:18a6d3fdc4c873819d798b43e394b718585ebb2804f4d7a2920e61b66575a2ad","semantic_task_id":"48961e2fcb941edb789b17839d7189e21b1dc3f3df92ead768b440bf20fd326b","split":"eval"} +{"domain":"finance","full_message_tokens":109881,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":250,"sample_id":"sample:21096f98b984227843c3f523d931b0088e52ab445e7e7cfa92e6783cdef6e0e5","semantic_task_id":"4908783cfc7d7ae2128df1f01ca66b864723b3363e40a8a0dbe8e089f7e36cc0","split":"eval"} +{"domain":"finance","full_message_tokens":96720,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":251,"sample_id":"sample:11e34f0a6d2fbabda4b566dc29c4a8dea4f5f5c940a4c9a0ab4b0b14aac29f2d","semantic_task_id":"4bb553a33188fcb252202ea4e83ea00233f6d18ba205f8f59df84017ad0245e6","split":"eval"} +{"domain":"finance","full_message_tokens":106756,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":252,"sample_id":"sample:57432439f47b8de84fae3c832ce7fd6790fe2d66dbfbd76261d8d87d517a56f7","semantic_task_id":"4c3aec8567038ff2a720f69212d006543da960206ca73c43d934f0d885e730da","split":"eval"} +{"domain":"finance","full_message_tokens":96682,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":253,"sample_id":"sample:03d1f0111f7339b6006e88bcac2b930cdb32a59ebd23d22295daedc6711e1f33","semantic_task_id":"4e53ffc7e622eb9dc279328e979ad5b0f96081c1f3327ab6579f1c981979c2ff","split":"eval"} +{"domain":"finance","full_message_tokens":201692,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":254,"sample_id":"sample:37467700dce2a0b0033d1e609ea65cbd56ebb12070a915746636ea3dc38c085c","semantic_task_id":"4e5bc8d6da887d23c1aa1f82bf5822fba628d359372240617f1e57841a94431c","split":"eval"} +{"domain":"finance","full_message_tokens":109840,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":255,"sample_id":"sample:2163e787e8e2fe628c01f9981f5b2b0a2696b923044eef808d88b1828492c58f","semantic_task_id":"5062b317c55ed833800bc4325e2b668bf1fea3551f9e8f09de08b14b9cc1242b","split":"eval"} +{"domain":"finance","full_message_tokens":95115,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":256,"sample_id":"sample:c516148c9ebc2a63ecc1bc49811728918e06db61a43b5016cad537a1550c0a8d","semantic_task_id":"509c91a1322214bfa920a50f4afeedf8fa66eed5a918d86ba6f366fecb79203c","split":"eval"} +{"domain":"finance","full_message_tokens":96680,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":257,"sample_id":"sample:b8bbd07296abf4e3604b704f18fb72f432ab2fcaf736164db6d2be02d92c113f","semantic_task_id":"53a8a1c2a2da5244c63d8f639ae52ecc38667b5609179719da7be4305a475a1f","split":"eval"} +{"domain":"finance","full_message_tokens":188516,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":258,"sample_id":"sample:9b1d3f72a5f905c7f0ee5b95ddaffe2449a87a8c4cc5863faaba9d7aedb40a2e","semantic_task_id":"54431040144b48a57afe1ecd18c02715a576b10732a722ed5c3a051d571ab0ae","split":"eval"} +{"domain":"finance","full_message_tokens":93521,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":259,"sample_id":"sample:c2bb02ee944599899931bcfcb814477c8c551506c22e3851a0832296e2e1a4bd","semantic_task_id":"557a33b1c2284fd1ee3c403ea944cd9952865e362dbf4e128a62483d898fff5e","split":"eval"} +{"domain":"finance","full_message_tokens":109881,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":260,"sample_id":"sample:8f5b336db4dfa3d8a4ffcb3080fd6cbce331ef9a3f0ae85c13c500cd033399c4","semantic_task_id":"58196797341ba168759345550c365e9ac7948738562617ac2426f4cccf27bba2","split":"eval"} +{"domain":"finance","full_message_tokens":93569,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":261,"sample_id":"sample:299c46ec3d7b500f580d5cc2002a4cb6f8bc47d58e382cf2ea680703192d066e","semantic_task_id":"586159644a88df6683b086939e29bfbd57074b8a240420e6c1a6bc233a7531f6","split":"eval"} +{"domain":"finance","full_message_tokens":109838,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":262,"sample_id":"sample:a4bdcde983dad208c637098eab6e036ec0de5822ea002928dedb7f76a23aed79","semantic_task_id":"59e0c555f7a64be9fab4a37fbab85c5370221a92cf16111d84e06f824a7ee4c7","split":"eval"} +{"domain":"finance","full_message_tokens":91536,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":263,"sample_id":"sample:531eb0819290e610e10120c8c592f76b7493624ebd69df289d55d992ac377b50","semantic_task_id":"5bb3854b4cecdee1ae4ae1bec658d373a3cf0e980bcf2c46bc7696ba9edff688","split":"eval"} +{"domain":"finance","full_message_tokens":109872,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":264,"sample_id":"sample:ceddc0dc7435a8f302162aca41f4f1e7146a1f7942b4ab181c13b2fbc0a5522d","semantic_task_id":"5d567242d021c4b2d69117584296207ab13a091551154964dda76bad9d61011d","split":"eval"} +{"domain":"finance","full_message_tokens":109868,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":265,"sample_id":"sample:842197113ec973f30f76595c83722c1becf4dc7ef42661076f49d634dd4d357e","semantic_task_id":"5f88a1114df25999b073ceca41e34039c335d2adbb450b31a073f1a5e747faed","split":"eval"} +{"domain":"finance","full_message_tokens":106677,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":266,"sample_id":"sample:97feefb35ade2639179e42275cba77e73de85880be5656f3595e524a9df0df85","semantic_task_id":"602f928c8a68535f5825cac3be1c7b81ca9af662f928c325c619fa985cbbde1b","split":"eval"} +{"domain":"finance","full_message_tokens":206419,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":267,"sample_id":"sample:e66742481a383db548044988c5c4fb008c6a45728dd234af30013bf4841ce6a6","semantic_task_id":"615a3dc6548e5c9c073f2705f130c0114f3d52bf77920a03859439711fcfe21f","split":"eval"} +{"domain":"finance","full_message_tokens":188515,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":268,"sample_id":"sample:014065e48638b318d6385d1fa9ab551d9753fc3a40b48047feb918dfa2a88027","semantic_task_id":"628cb4a481ca3e67bcd7e252c431c6a1891065b18dbf006102cb6d4448d55d7c","split":"eval"} +{"domain":"finance","full_message_tokens":96725,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":269,"sample_id":"sample:99e5fc03a3fb643ee8f30781417370a2b37419dcae8816e0ff3d385a012127e4","semantic_task_id":"6441c5d3ce04e99ede68b34840379517adc21483da9e11069bed7aa7df968c43","split":"eval"} +{"domain":"finance","full_message_tokens":95113,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":270,"sample_id":"sample:b6855397dbf7dad9a13e448829ff0c583cab7a75e78b66c0b841c2a55fd699a8","semantic_task_id":"652b8f50566ff6421eb32bc413b3da41e4413a0bc242d696411cab76da2d0a3f","split":"eval"} +{"domain":"finance","full_message_tokens":93522,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":271,"sample_id":"sample:4190e2c0ea035d6e31a83c243c7c8a3899e80e83399dee0ddf58ed022f64b234","semantic_task_id":"671aac2dd01fcd1b0bb383f4e6e6775cf08667e2c4cb9ad6115fa3d079cd883a","split":"eval"} +{"domain":"finance","full_message_tokens":106677,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":272,"sample_id":"sample:f0eb0efcf895f228695e69641411ea1f4f48ee4017ca6a81a57093e53cc21f84","semantic_task_id":"673c581dfd707ae71fd805f3b28823ddfe05f36c1304e820b7095500b33c5099","split":"eval"} +{"domain":"finance","full_message_tokens":201693,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":273,"sample_id":"sample:d4f442fdcc6168739e7b936e8754c375c442c33ddfeccde2d663ed54dbf2d182","semantic_task_id":"6745ab0b70a4266cc6aa674d3d98f8e3c2c74b821cedaf57d733483e8e6f0b85","split":"eval"} +{"domain":"finance","full_message_tokens":206452,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":274,"sample_id":"sample:a2acf8706710f143ef78ba102deaf7670a49c8371f0905e61799bb366dc43bc8","semantic_task_id":"6b03bd6a7f8d289b83ec6eacc2f79be418323e42cc66b22278722dd7c0e934f8","split":"eval"} +{"domain":"finance","full_message_tokens":91488,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":275,"sample_id":"sample:1f77ff9756ad671c95d1834036f697cb2bae9dfe4e573d520017d0894a6c31cc","semantic_task_id":"6b5a461b0ad33dd085bd03d81c22c89bdc01c832cb3b1a36c88e3b7d0fe107d4","split":"eval"} +{"domain":"finance","full_message_tokens":206430,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":276,"sample_id":"sample:c82afb3cb372d9caefaff6bbdf1e85ed5af8a1da161b1bd699c5ec4a9d3c79ce","semantic_task_id":"6cd96ab1730d6c70089220e5d60b8122b1bab22beec87e6141e008dfc670dc8b","split":"eval"} +{"domain":"finance","full_message_tokens":96724,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":277,"sample_id":"sample:8ced2dec537617ecee9950dae5ee7d38df8c2b64760d3caeffed877d6a178904","semantic_task_id":"6ef68326df6432d37bf7deb87e6a6ec5a761464bf83449c2c38a0ecbc7dc4048","split":"eval"} +{"domain":"finance","full_message_tokens":93521,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":278,"sample_id":"sample:8901f2cec035160b85bc000dc9cb5f42d23de3df33719b4950007ed20170f0e0","semantic_task_id":"6fc39e2c732a1912491b7a9c5b85c019f744e42cffcb312f2464456482e2deb8","split":"eval"} +{"domain":"finance","full_message_tokens":109832,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":279,"sample_id":"sample:c27533d1c874e4df3ea7b188e9a54d03b15c8392a9cc4f8eda19130d1df8d13b","semantic_task_id":"7152ff74cc5ac991965e8d62e2f218824a1905c4d6d188df772a2b6ccde70de5","split":"eval"} +{"domain":"finance","full_message_tokens":91911,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":280,"sample_id":"sample:1d3d3ca66533283ce9e82a4faa2afbe31f7b02d1802303d3a88dd4d4830742ff","semantic_task_id":"71a80801ffe60236477094d828e41fb1e44b6f2a7093ad1763eb3340c7821d76","split":"eval"} +{"domain":"finance","full_message_tokens":96682,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":281,"sample_id":"sample:0d7efd9a0e3cc1c2079b6e9196a72c73b0b19e30063fed797aabc6cc85829996","semantic_task_id":"71c67d11146382f8b334c6d19e0e6bdf593b5894146d1ecafbe2bedbb02ba816","split":"eval"} +{"domain":"finance","full_message_tokens":106728,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":282,"sample_id":"sample:38235e2a0435304174e6b785bf7d4dd5482f108ec4e63c41523e094f021eced3","semantic_task_id":"760c7aff60327e76f65f85d3d3cc0a032524f57b508c4a98c0bbc00c945c62f5","split":"eval"} +{"domain":"finance","full_message_tokens":96719,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":283,"sample_id":"sample:7c658290ae9e0f500141826442fdd191b946b325ae437315b83d14d51c71178d","semantic_task_id":"76201947d9b6c4b57d97998b64de8958a69ca29b700c3ecaa018f6000fc3a527","split":"eval"} +{"domain":"finance","full_message_tokens":188536,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":284,"sample_id":"sample:69742ef53852a9e3b5a8d695ccffede62ec5b48301c2936550d5ffbc5910f3e9","semantic_task_id":"7813758360e7db0dae5e883fde16e364f50909a20d67f12006ecf90c28babdb8","split":"eval"} +{"domain":"finance","full_message_tokens":106676,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":285,"sample_id":"sample:309b6d5e4afd4468c292a79582c6b1a79c8187360f212d6fbc76e0a20781b234","semantic_task_id":"7c125e50a60bdcbdccd55a51919bb17cf80ea256ac822b2d4f22c577445407e8","split":"eval"} +{"domain":"finance","full_message_tokens":96725,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":286,"sample_id":"sample:bd6b73c18d639b6c49b236c7ecd7aac0e538761e7f05bc2155e46437206d721c","semantic_task_id":"7f4d1110849060cce66086f8279172cb298a56931b5e3c6b3502be4d3a6ddbfc","split":"eval"} +{"domain":"finance","full_message_tokens":109881,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":287,"sample_id":"sample:c23781a1ad4921aa1ef7439ec48c1e9d373d8d618424379fc9ed8efb476b1e4f","semantic_task_id":"809338ffd98faeed5266f1582115c7492d42afee24743a56e29edb4995bed8a5","split":"eval"} +{"domain":"finance","full_message_tokens":109834,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":288,"sample_id":"sample:8113f25040f946d5839e66f94a25530f586f89071dfcf3417c5713c8c8570082","semantic_task_id":"816b215e64772f761b42ee8469c51aa4cbbd534387a71eafef99c1b2a941198e","split":"eval"} +{"domain":"finance","full_message_tokens":91911,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":289,"sample_id":"sample:44cc250910a85f99fb299dedf8c18321a3d98ef843c22b6f61b7f23fc628562e","semantic_task_id":"82d3f577c51a58a3c096ac1f6b0475ce901b47ee9ffd938b758606d09d044c4b","split":"eval"} +{"domain":"finance","full_message_tokens":109881,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":290,"sample_id":"sample:44d92acbb36416e909ff1c55b2407890e7bc02b1f801a01e8b41bb48d1f8123a","semantic_task_id":"83320450fd73a2c1085be8b585ea8383687dc12107af38b6390563db75126e5a","split":"eval"} +{"domain":"finance","full_message_tokens":206455,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":291,"sample_id":"sample:813cbe829a2c46877f147deed7fb081d372613a621b33dcac3af1518f8c3d024","semantic_task_id":"854cfb186339e76d71845ee90a9b2601a2f3618b1f9ee3e7f03e0d5b600a1804","split":"eval"} +{"domain":"finance","full_message_tokens":96721,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":292,"sample_id":"sample:00e7e81d2c225669d7731bfafcd638342fe4004acb88730067d1dc09500ba287","semantic_task_id":"886eb7643e9546100fc348d9a624be18570484682b13e290ef4c908aecb17d3a","split":"eval"} +{"domain":"finance","full_message_tokens":188537,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":293,"sample_id":"sample:d23b1b18f59a7549884618fe1fe579d00b6695b158fa36949c9ef8dc36080400","semantic_task_id":"8cda8ee02d274108cb3538857c734c24494167c2add5fb6b04da9aba3b41b5c1","split":"eval"} +{"domain":"finance","full_message_tokens":96681,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":294,"sample_id":"sample:40ef121f1b853d97c7d6e98ca66d9d175107a794726569d9f14486e92cb51089","semantic_task_id":"915a5dbcc32d6ae217186c605f009212292c51c51334f08d68b4c4bac97234b1","split":"eval"} +{"domain":"finance","full_message_tokens":106677,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":295,"sample_id":"sample:ec40d88bd2b8c72950a49f199bcc48b7b5b8d9a72f385d9228ccbdd219a15ade","semantic_task_id":"93696bcd406838aa2e4737e189067e94e554849ccb5a7ae3b9c8fe5d3f1f4861","split":"eval"} +{"domain":"finance","full_message_tokens":109879,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":296,"sample_id":"sample:1bd77ff86444e70064cef86b6ec3c248a32b5d09b8673952f04dcb3baf053d88","semantic_task_id":"93e8a2520d6092ad85a1cde25e4de1bc59fb7d6705bbb2613fcf5e69b7bf4470","split":"eval"} +{"domain":"finance","full_message_tokens":91959,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":297,"sample_id":"sample:adc94a5dadf8726903e27499d8d1d14f2ee18121d54b5e232d55fbbd188e18d3","semantic_task_id":"93fa7eb8342a2b4ab7a4f6ec6aa7c4b9d477168420577bfd676bfb8858cd3f1d","split":"eval"} +{"domain":"finance","full_message_tokens":109872,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":298,"sample_id":"sample:0195d1c01af11b9ae53a9e1832b93a32e60c4f5a760ddf6d182b249cc8ff1db0","semantic_task_id":"95a62dbd6d3243b83fc8b83a0ad8071e6526ffbebc3c241c78f450a43a0da900","split":"eval"} +{"domain":"finance","full_message_tokens":95115,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":299,"sample_id":"sample:417ec8087768cf6557f1f7ce91e4127976dcb66554c811f9ab714639d9a9f63f","semantic_task_id":"9664fa0243254e794b3cb03733fd7456a49c494f773c207adc04ec7c274004e5","split":"eval"} +{"domain":"finance","full_message_tokens":201682,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":300,"sample_id":"sample:20463fb03536f717ec35d142b3e671e77c1f03bb7c670033a612ea09677511e6","semantic_task_id":"96819e9f314ba19949e111397772a48b1e9cc6b3d929503aea60b556892311ad","split":"eval"} +{"domain":"finance","full_message_tokens":93521,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":301,"sample_id":"sample:e7e9dffd9c995226199c66a7940cf4a86ab7824d08ef8a2badecb79be1569799","semantic_task_id":"98241bfa900da190278d9ff598d9c48312fca02c94d8204eeda5f5851a5bace0","split":"eval"} +{"domain":"finance","full_message_tokens":206495,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":302,"sample_id":"sample:6cdfdfcf2855c58683102c3ed8682dede69352cd9e6b51aa4b3ca224cdf36939","semantic_task_id":"9b680bf7d12171a74db9321ffcaaa5a34960c4785c8411ffc39e94e80276aa79","split":"eval"} +{"domain":"finance","full_message_tokens":106679,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":303,"sample_id":"sample:a6e22782b1938649dd45ca259cbceebe88a58fa6cda5efbb3ce8f11c655afd45","semantic_task_id":"9c2cf99396889f4f1bde3f7a3cad541eb275259fb6d7e9f78226b1552019bc51","split":"eval"} +{"domain":"finance","full_message_tokens":188515,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":304,"sample_id":"sample:8f3a6d5d7cfa280326779fae965a6a21b13386340ec87039f5689045432086f1","semantic_task_id":"9c46dc0389957f76bf4bc41d908abab8a05e0b127b858fb2f0d158dccb3e0d1d","split":"eval"} +{"domain":"finance","full_message_tokens":96723,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":305,"sample_id":"sample:d16e690fd869882ad0eec1e5a01c4d47897cbf4ef288502e521a6a7199c7a947","semantic_task_id":"9c906066cc01ea4cb2e4264f32a393ed204bcca0034661253c4af037808f0ec3","split":"eval"} +{"domain":"finance","full_message_tokens":95110,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":306,"sample_id":"sample:0fb0a3c67cb4d87d8e0368e550e94a7e508207de98a366af4116652a74a120c2","semantic_task_id":"9da6bf15cc7b4f2e6bd2cee6568c672e29acee5f2ea17727577e8180a05bfeec","split":"eval"} +{"domain":"finance","full_message_tokens":201693,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":307,"sample_id":"sample:6b4ced47ab03c4567a68326cadf4363e5657dd5522b6efa3aeba93673060fe16","semantic_task_id":"9dd8398754d5b4fa8fb2c773b7142d20d074cd7bf80c33aed03a9b8cfff72998","split":"eval"} +{"domain":"finance","full_message_tokens":95110,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":308,"sample_id":"sample:4f81fce36e370c03bfd53a519082fdb5f020655b7a5a060e2fbb5c68fb65a5b0","semantic_task_id":"9f3f7d4ff146d68382c9c9345783fc6ab93c06b163069c141828b82561010891","split":"eval"} +{"domain":"finance","full_message_tokens":188591,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":309,"sample_id":"sample:126a5091912275f051d30c6d81240d58420b324cd475e89aa2f4f1fbe0acb83d","semantic_task_id":"a22349ea57d5a592089fab1d1a582186b28fba4274e81d99d20c0d4008365743","split":"eval"} +{"domain":"finance","full_message_tokens":109836,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":310,"sample_id":"sample:13e99c45c502c63dc35be8651864969dbc9f43d4f4783333aee099b1d9ba74b7","semantic_task_id":"a22aba9c5ab6731013138b322b30084f5d124845b381c338db692780e5145989","split":"eval"} +{"domain":"finance","full_message_tokens":206449,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":311,"sample_id":"sample:0c7a73072f4463f0ebbc5013fdbca2a856c7b1ed272f389d463975a253efc2f6","semantic_task_id":"a359ef12d7f302c1f5f887e0f3e70a9cad77f55b425c74d8b500579de3693ee9","split":"eval"} +{"domain":"finance","full_message_tokens":106723,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":312,"sample_id":"sample:d1667b597b247dbedbf8da6c592653385dee3fe79f298a1b79d52ce6891bd86c","semantic_task_id":"a77d9d33d1822394e65cde047226990e53665c23b8cd2353ae1c5a748ba0559c","split":"eval"} +{"domain":"finance","full_message_tokens":91910,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":313,"sample_id":"sample:2f755099418ff965e437221c9fd775bc99f3df0f9f6a334328de2cc8ac80d603","semantic_task_id":"a82e3fa4c55a92055ba899f883bd98b1f1bfe8f8ed16b194c3e2a5432679c41b","split":"eval"} +{"domain":"finance","full_message_tokens":95112,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":314,"sample_id":"sample:778fa2db16c9b325423aea5f4ee47ccded4a6acc8829cb205d220edf3b03a3e0","semantic_task_id":"a93ce734b73ea6cc25cfe7004fda96cdaa289c0f785a6ed347cd44270a58e07e","split":"eval"} +{"domain":"finance","full_message_tokens":91913,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":315,"sample_id":"sample:f94acb5ff47e725fa79cc5d6b603eef3cda9d58d081d9b97e2c73c68767943ad","semantic_task_id":"a971ce0f5216f78054be87bc4ffc792ff4fa588898bb2ee52a498a2dea74fb6e","split":"eval"} +{"domain":"finance","full_message_tokens":91490,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":316,"sample_id":"sample:ab39fc9fc2a0246842678d044b72331993775dbcc4000650a6bd9e8db1f4cfab","semantic_task_id":"a9e9b078d8e408399f38712c4913e75b6ccc9e5327100c8edc919f3659215570","split":"eval"} +{"domain":"finance","full_message_tokens":109872,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":317,"sample_id":"sample:ce1513844121ccdd1fd476da110a077b59d1b6670a15bb8c48bb448b716d8824","semantic_task_id":"ab64dd79fa7c0a041cca4202bd8a0ac42e4415b10cc934cb2a066c5373c3aee7","split":"eval"} +{"domain":"finance","full_message_tokens":201691,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":318,"sample_id":"sample:d5d1553f9c1f180a10d58300a1659206abf167061dba7f26fa03ed2ca295f9af","semantic_task_id":"acb7a32fa63e831217449265ea93ec898d306d084f48fe37677862481e2cc854","split":"eval"} +{"domain":"finance","full_message_tokens":106677,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":319,"sample_id":"sample:b8b678949a5f0daf6b1c8d822bceaf390425c4327e19a7b6df6cefb0f6cd4fde","semantic_task_id":"b16799cf6997455c25a6cac62999562b07d8f710a6473e5d8e55352dcdbf9649","split":"eval"} +{"domain":"finance","full_message_tokens":109838,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":320,"sample_id":"sample:a43d80c141fe8162b413e6e0d850a07d926c26f46e9792520d46bdca3296e1da","semantic_task_id":"b2144ce2b24743130d5c21e4a67f71506ff22c3f5b1ab77ee67558bb17508ace","split":"eval"} +{"domain":"finance","full_message_tokens":93522,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":321,"sample_id":"sample:0d04a0847af79388d9ec81bbc573fb8f803e99f51e46f8eb24928a58e21e5d6e","semantic_task_id":"b2c12b596025b3aa3abafae00ec51d8a0edd9df552293127124229eca118b1ab","split":"eval"} +{"domain":"finance","full_message_tokens":201671,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":322,"sample_id":"sample:2c1e548b8ead08de8e3f52d9a335f4493fedfecd6d030965f5b2f20ea89e332b","semantic_task_id":"b5e631e70a4d4dc191031097a9a34e9c9ad0f30f07ff54712aa1056d480714df","split":"eval"} +{"domain":"finance","full_message_tokens":91963,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":323,"sample_id":"sample:25a4b7772265084af32d71e63bbbd847b5a838a0d10b1c9c837addd2b37e9353","semantic_task_id":"b6af6ed3976eada1167ebb2bd332b90f6f424a628ce7879bbcf1e4a3395c1298","split":"eval"} +{"domain":"finance","full_message_tokens":201747,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":324,"sample_id":"sample:6c2486054e785bd57b7a5a172f75fd1be5a677f45f1df9491e15d33ccc4c7d2d","semantic_task_id":"b82bbdc03416e5c68030b6a140c16bb9018534438f1e82bb6d473d874709424c","split":"eval"} +{"domain":"finance","full_message_tokens":96680,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":325,"sample_id":"sample:f92d6fc7b4f4c0f2adb51a0d2f0ff719e651f405fb58bfa1edcdbe7f5b6413b8","semantic_task_id":"b84a52bccc47dbead538826f4f66924866efd027645c17e2bfbb4bad7f092ab1","split":"eval"} +{"domain":"finance","full_message_tokens":109838,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":326,"sample_id":"sample:8cb72c14b01434f94f9541851d89b469497c9c5cce128a9bbdf501afb3fc795e","semantic_task_id":"ba697ddc31bc9703772e9db266debc47c8f5d956e0046b48aff87eb02b2d6196","split":"eval"} +{"domain":"finance","full_message_tokens":188537,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":327,"sample_id":"sample:3dc5757e3e87b3c52d6b59f76de99c68b827678fa21ab81d8f779fd8653b7db5","semantic_task_id":"bb9e286adf13b97da85272bced75440cdd638380461fa28c687e0fc65284b2d6","split":"eval"} +{"domain":"finance","full_message_tokens":201671,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":328,"sample_id":"sample:961619ae312aa8da80f3ab621e0ef162c98698b599401286852f876f77368c56","semantic_task_id":"bf7d770d151a860ae70d662881e832dd8edb411d43adba187e5a508059850f2c","split":"eval"} +{"domain":"finance","full_message_tokens":206430,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":329,"sample_id":"sample:264baec4a7000cbddabb06d6dbbd5a82a07b2c396ddf5cd95fa874e2fca33de6","semantic_task_id":"c1ec9a72184c976836a06baa42886269231f20b1b9c092b9e9b06d5574865625","split":"eval"} +{"domain":"finance","full_message_tokens":96680,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":330,"sample_id":"sample:1f9d0f7316b7ad107db10ce74a9076b6afe5f8ac341e6f0ffba9c5245f92f6f2","semantic_task_id":"c26298e509d24aa4117e36e0aab5f4ff65f9b012c68f5654583b3a59af8bf29d","split":"eval"} +{"domain":"finance","full_message_tokens":96720,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":331,"sample_id":"sample:840914e4f37f2802bbbf6bf4d2fc27911dcfc227c602c04a8857e01790728584","semantic_task_id":"c3c169d20af2d7cbf049d4fe95742c28c485935d6cdb55a38eae3ba470f4ced4","split":"eval"} +{"domain":"finance","full_message_tokens":91912,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":332,"sample_id":"sample:382385bf7fb8f1e1f0186eae533d12f60115007a02a125f7a8cf006028ef1b9c","semantic_task_id":"c45f8ad3f70afe4e81148f5ddf0cdb8639b9bcddde50ed0d61082b1862ffe73c","split":"eval"} +{"domain":"finance","full_message_tokens":188536,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":333,"sample_id":"sample:e81412f28d450c39893f27de90455d4412168d82494fa141bb26d134642cbb30","semantic_task_id":"c465f59a43d3b92bb7045b2b6a6e680032e089706f893c7a55c4320ff57642fe","split":"eval"} +{"domain":"finance","full_message_tokens":201692,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":334,"sample_id":"sample:73da86f3d96d25268bc0a134a9cb527f8d6f46808f6fe32aad4f0c9a2b84aa4b","semantic_task_id":"c6b7a7c9c77e79f704f1cd4bac6e3da66ec9fbb0bb5c49e6da419c5bb54d3a21","split":"eval"} +{"domain":"finance","full_message_tokens":201670,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":335,"sample_id":"sample:fbb796fcba31cbf89289f62476cc88d3c4ac97e5129af2949576b73179ff1fc2","semantic_task_id":"c6f32030843896b481babd1c8bab24a7a463a91f66cbca343bf808f2250caad0","split":"eval"} +{"domain":"finance","full_message_tokens":188536,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":336,"sample_id":"sample:b0403d236fc07191ab139592c65ecbffed99e3a2c7e5af3040e385db7ec5a903","semantic_task_id":"c7f34f8309832e071c91097a05de31adbff000710036fd5ee60d64755266a3cf","split":"eval"} +{"domain":"finance","full_message_tokens":206452,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":337,"sample_id":"sample:3973e568bac59df3187cbaacfc535e4be614aeb258f5fc44db23a072f38cb55f","semantic_task_id":"c888965dca957927737844c729e2bf57681aff63671f5b885ceea5335eba415d","split":"eval"} +{"domain":"finance","full_message_tokens":201671,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":338,"sample_id":"sample:da7193fa05843283ba969680da9769c098fedbf7f1e20801ba3943d9c0a7e6e8","semantic_task_id":"ca7d4e7ff5ad9e42810643bd0f8eb0539cff4ed2875db9b0041ece8e375f76c9","split":"eval"} +{"domain":"finance","full_message_tokens":201673,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":339,"sample_id":"sample:31a5509f618fb936e55f9a98bd98e549656eb62af6f5aa844affd27cd4af63c2","semantic_task_id":"caa7fce18ca289391672bef07763f44a9689418c9114be9c2e73c2c9f41302e8","split":"eval"} +{"domain":"finance","full_message_tokens":91957,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":340,"sample_id":"sample:7eade94347782e59fcc16f20245d184a63e3b0d55f98bbd1bd19d33b687f2eb4","semantic_task_id":"cac4849cc13aece6972d7d3d9aaaf929fc396cf4e3844db669d27af54f753e50","split":"eval"} +{"domain":"finance","full_message_tokens":91534,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":341,"sample_id":"sample:24279bc2d76b1b61da8adbe6555a9975260aa1a7d75c3f0bbbc9eac2a22d3a2b","semantic_task_id":"cc76ea7cbc1dc0bc03c67ccbd2464003b6b4169b29951cb8ff0b388c75764991","split":"eval"} +{"domain":"finance","full_message_tokens":91911,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":342,"sample_id":"sample:219724db028ca67cebbec29e47b5733a7c7795fa48a6a2868f9e1d51197b7ebd","semantic_task_id":"ce25b620ff71ab6782fa7f8f7cc1e8d706a6f4dc03271e8745ed74fe6a7b72e7","split":"eval"} +{"domain":"finance","full_message_tokens":93573,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":343,"sample_id":"sample:62834b0210ef24882cbe6705cfd2c35cf8cf84a96db7f9b4c3f546849543b308","semantic_task_id":"ce7be838c82b2343fa0d65f335d7226df5f8fca498497b3d31cd3620e76049a6","split":"eval"} +{"domain":"finance","full_message_tokens":201671,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":344,"sample_id":"sample:3df8e617696eead4576e542c3b2963bc727f8ad424fff74067c8dc1b195ff08b","semantic_task_id":"d15c6318c516e48865b0afc7e749841e91dd0924ba21047a9a31cf377ea70ab6","split":"eval"} +{"domain":"finance","full_message_tokens":93601,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":345,"sample_id":"sample:c0ac91ef4481eaa2f37ce7b864aa88ab8914b4069f0db5c6ebb814fa5d644b44","semantic_task_id":"d15f6a41cd8f862fc452cc987f510ebf8f5ba3db084a722c999bd8e243f028a8","split":"eval"} +{"domain":"finance","full_message_tokens":201682,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":346,"sample_id":"sample:e3df0bd4f4c4a587a6b1e4886caf97695404cfcbaa4fb2720eace4b426e47eb1","semantic_task_id":"d3fcc39dd4213a8db5e6cca22d73933d05283bbca518b101f2d3a43d22b50eca","split":"eval"} +{"domain":"finance","full_message_tokens":188516,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":347,"sample_id":"sample:9b4e3f768d733c9dc0bb2311050249abf7e89f68b1fb324e1127a6ec3ac9e9ab","semantic_task_id":"d49fe387c47a3887a2e821c243e7fe0ee6da82f3551efbc8891ee42cbf4932ad","split":"eval"} +{"domain":"finance","full_message_tokens":91488,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":348,"sample_id":"sample:249e6299bc0dd7e0e4484271469e2de201f883c6c9109df2f68ac650ea51f686","semantic_task_id":"d5616d5332b9e18a1957cf7b7822f02ad1ff7d70f716599d69a9f627efe61939","split":"eval"} +{"domain":"finance","full_message_tokens":188515,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":349,"sample_id":"sample:053460289d9cb3230d80fc54a30c7ba6db7f21db3b9a5c05836428b89c2ad243","semantic_task_id":"d6739612d9144c872bceb35792ec02f179f9df5c0c1dcedd586d6fd9ecb23ba8","split":"eval"} +{"domain":"finance","full_message_tokens":91489,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":350,"sample_id":"sample:6ca1720d5f0ba5e67e67d4e9a50b640e95f0c40e135623c130c0bea205b93e09","semantic_task_id":"d7f480261ebd145ca17f4ce2811ba18d529c1a332b6e95975ba6dc65ff986059","split":"eval"} +{"domain":"finance","full_message_tokens":109832,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":351,"sample_id":"sample:fc042a11c44c829aceae6eeb437a004f94f6f05e32257b2d0abbe2fddab94c09","semantic_task_id":"d927ff20cad32f692942fb04a6a6f47e530755365c117fdc716c33655dddc78a","split":"eval"} +{"domain":"finance","full_message_tokens":93520,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":352,"sample_id":"sample:413f01cc7c8341a64c0e783bfe168cc6da4dfdc4c31429f93f51f4fc87c0cd13","semantic_task_id":"d9ccf43aa28bc660be5b757b42e8d39ae356a92bf23745c3c1ebbc2c53b581f3","split":"eval"} +{"domain":"finance","full_message_tokens":106725,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":353,"sample_id":"sample:ff25b677e2f86bc35e2730626e0625c6e27eae4adee97a7cd90b69faa521a270","semantic_task_id":"da5e1d026e63bd19f2d359fc28d304ae8e064fa17661bc8c40a2003672520795","split":"eval"} +{"domain":"finance","full_message_tokens":188526,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":354,"sample_id":"sample:598ca0cdef5e4ae547d0926f20b7a9306a0216a73a7b0192c4ce508fd41dac9a","semantic_task_id":"dc74387357acea36a271d5600ffe0e67a8e05a443ab9b1309dabdcd6caade851","split":"eval"} +{"domain":"finance","full_message_tokens":109837,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":355,"sample_id":"sample:20cba218a15b88df69f7b1ed2bef88c59556420680d61fa29c795bdb57e97ef0","semantic_task_id":"dedf00934b29dab840d956537d20914e5efc78b121b43c74b052289c78a24b86","split":"eval"} +{"domain":"finance","full_message_tokens":96721,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":356,"sample_id":"sample:587358abaa99d7e285715cde5aef85d0b8c6a662f64dde84f8338bd3ee6a183d","semantic_task_id":"e0317385a61845145b5f95e2c8b95c3fd8fee994706d52fcb3708064572fc586","split":"eval"} +{"domain":"finance","full_message_tokens":188551,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":357,"sample_id":"sample:5e6e981ddf7bef09a427e5e77cc69896b8cbb3713df986475d6343622fd1755c","semantic_task_id":"e040fb24fbf0c34a2ad030dfc2fc97b011f89450cc220acdc3fc0ccc8465bfed","split":"eval"} +{"domain":"finance","full_message_tokens":109823,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":358,"sample_id":"sample:8e4f064a7300209d2651209945b38730c69f48554d847c9749ad6c74b744ed50","semantic_task_id":"e0425c4c3fdcbb389d6fb53ccda78e651b51c8640f09378d7545eceaf9930f9e","split":"eval"} +{"domain":"finance","full_message_tokens":188535,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":359,"sample_id":"sample:decb8bfb12610b9f9039b0457c93817e548e6808d4f9aa78e6f3d80f7f60cf32","semantic_task_id":"e1e6b57b07b6aeba8ff81f139f5b3f4f1c51bac474325c2cc0e3bc1c5e3a820d","split":"eval"} +{"domain":"finance","full_message_tokens":91991,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":360,"sample_id":"sample:1d4d549570171a28c371955a4a69e0f4d49d2c540eb657e84a9954220e00d8e5","semantic_task_id":"e23a4425cca53d8ed70d08ccc4d1d34c1bdeee32910bc99f702bb21b5ba2d48a","split":"eval"} +{"domain":"finance","full_message_tokens":206420,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":361,"sample_id":"sample:97c764c299dab4e8d9b86293b47ff38daead1dff4b0264120dc54a7d319fa18b","semantic_task_id":"e3462f2e36e25214059ecf4040a8c2834fecb8e31634313d51bb8356d305fa76","split":"eval"} +{"domain":"finance","full_message_tokens":91488,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":362,"sample_id":"sample:f1fb469084fa783229fcb4f522a0e8886b56dbc1f5768e2d7e6fe8afdf94351f","semantic_task_id":"e352f1b486e5eeff249dd25c86bdf67e88c0893e5e14a4693a5fda24a6987a39","split":"eval"} +{"domain":"finance","full_message_tokens":206451,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":363,"sample_id":"sample:b5a310cc1653e90b8ef24738220d82697acf13dca525f979f2bbaeb8a3e94a45","semantic_task_id":"e48d7e85149896b299b1931f64f13f277cfa555dea24352ee60c7a1fb75a6edd","split":"eval"} +{"domain":"finance","full_message_tokens":91912,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":364,"sample_id":"sample:1091e65ed76f319c0a43741ced3c079b25bcc671802843e57ec333ae46d14e85","semantic_task_id":"e6527d53f80c35e7af7f9010969e07e3de8263a78aa0747d383c6f0f9e28e9f8","split":"eval"} +{"domain":"finance","full_message_tokens":96710,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":365,"sample_id":"sample:a5422929cca36a06d062d4076b981afd4869c7f6ff31d03590fb023ab17a276f","semantic_task_id":"e7b02664c0b9a553f9dbce1bf48b2de2a25f805315e61e33025734f427d57955","split":"eval"} +{"domain":"finance","full_message_tokens":109883,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":366,"sample_id":"sample:879dc0edbd8844a26be042921eb4e3810da1c8ef799e64c0d8aa172f4988fb6c","semantic_task_id":"e90c831c58d9129ce7c6c5732d6551e1ec6dac93ac09335b8661b3f181dff993","split":"eval"} +{"domain":"finance","full_message_tokens":188515,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":367,"sample_id":"sample:e1194d44b085f7cf994677de82b168bb6bf57aad232c3ee5238232b503ee1f64","semantic_task_id":"e946a24dd9a5aa6c76bfaeb3e9cebb70e76148b9f92692282b447abaf52fbc24","split":"eval"} +{"domain":"finance","full_message_tokens":201693,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":368,"sample_id":"sample:f546986d53036e902f60e95c9533b4c4ba7dcac94bf551300a4caf2dce8fc1b1","semantic_task_id":"e9944b1e6d972f51405cee52bef1b22b62e17600fe7423bc33510018a1453a06","split":"eval"} +{"domain":"finance","full_message_tokens":109872,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":369,"sample_id":"sample:149a73058e69c6944d744c9290a6260e859f7bcdbfae04918e12ae30c17ec5a2","semantic_task_id":"e9e6c434f7f39645132d0e2e8083f723cc8ef39501d93f1774308e980d1c9f0f","split":"eval"} +{"domain":"finance","full_message_tokens":201672,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":370,"sample_id":"sample:fe61c6e9908300f9fa289697cb7294ba255e6afb06b3374df78380aaf55ee6aa","semantic_task_id":"ea2300fd5bb33e41834c9740460ce9650f87c00261f68e835aa04ac2c25c29c2","split":"eval"} +{"domain":"finance","full_message_tokens":109874,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":371,"sample_id":"sample:8c89bb7c37d1371a45c114d804f8d3766e151590384ec1e3ece6863db83b74c9","semantic_task_id":"ee553996821827b8300b1fec4bf7abd2a9f56f8d9e7d05335c9903e3062b9ddd","split":"eval"} +{"domain":"finance","full_message_tokens":206419,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":372,"sample_id":"sample:c1adc7d1939b4bf75155215dd42236b8acaa97abb793e9ce5ce6d1c0b688fc5b","semantic_task_id":"eecad6dedacd97ac4d1b9d64e78ec9d839071dad6660723b775a510395233c92","split":"eval"} +{"domain":"finance","full_message_tokens":109881,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":373,"sample_id":"sample:ece5b9400b58b05d4835645bb97f158c39269f78ee8a51696960452b29077320","semantic_task_id":"ef6a7ef866d66673edaa27a1e1a938ab326179606a35efa13de113570e9f8449","split":"eval"} +{"domain":"finance","full_message_tokens":201671,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":374,"sample_id":"sample:5079f980048227ba450cd02aac0432b9b214b6f3e136bd42ce9f80b1184cacd2","semantic_task_id":"f0119efd01a5bedf08111870820a32172e2f507bb577e1ebe1e6334fe297bd09","split":"eval"} +{"domain":"finance","full_message_tokens":96725,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":375,"sample_id":"sample:463a22a42a3a2f7584c0fe66aa181ce071df5efdb827edf48617b3713398f4cb","semantic_task_id":"f0ba1d96ad75c9f352629faf376fd42e97e17750bf68f37ee631c10cdc0baa43","split":"eval"} +{"domain":"finance","full_message_tokens":188514,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":376,"sample_id":"sample:bbe1171fe4700346e30006563e75c4e56b0faa6a9f6b6ba309421e4c63dc089e","semantic_task_id":"f1d8231e012f4a51639476c6ef411c307c71902bbd62e2446a6dd35e421553e1","split":"eval"} +{"domain":"finance","full_message_tokens":206461,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":377,"sample_id":"sample:a0efd8039a869ed1719794d38355ad04cd4788255f6a8b94009377f76bcfed00","semantic_task_id":"f3d7b1a40e6a37b3f9008ddaf9d708e1a9dcc23ddedde8f8df35dcc70e1ba56c","split":"eval"} +{"domain":"finance","full_message_tokens":109823,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":378,"sample_id":"sample:25e24578147f65665ebf59d4c0e37f674aa3bb2a504f688b206bd7607c4bd53a","semantic_task_id":"f612b96e6cddb39d660ad911e678b5fcd3972a9e7ef6f2919430046abb9fdbfd","split":"eval"} +{"domain":"finance","full_message_tokens":96682,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":379,"sample_id":"sample:bc61a8e7a805190053ff00b24df8f27364c172a88962bf0225ed3d8010b23af1","semantic_task_id":"f6ccb92a209eb5b446a5f82d6ec1b4a309993e4d7d92db96a90cdab15ea3a705","split":"eval"} +{"domain":"finance","full_message_tokens":93523,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":380,"sample_id":"sample:d1fdc7c21dc268ff7958d7104a07e77ea4a90a7658437e3ba97e969e02e1c5f2","semantic_task_id":"f6f462aabe74125548d4e757370be9d0026de0d36779206f63aae12067fb03d3","split":"eval"} +{"domain":"finance","full_message_tokens":109823,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":381,"sample_id":"sample:42de12950b50951008a322efac9eb116dd73eb71fcab55881a3ead4fb07f98d4","semantic_task_id":"f7614aa01addaa1828cae883460bbb1623b5552cfb7212d8fc52753bab579373","split":"eval"} +{"domain":"finance","full_message_tokens":201692,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":382,"sample_id":"sample:f9d3e52e206975478c6be4ae3b60eb76a608e597c4c802efe498582b48aa6ceb","semantic_task_id":"f7f6a2198e44430837dd886ff31d4b576c98d1b3223c990624fab6455cc524fc","split":"eval"} +{"domain":"finance","full_message_tokens":201692,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":383,"sample_id":"sample:9e063186b0b4bf09d360bfa57e17811e37e3c2e2423d94298493269b19ffffb5","semantic_task_id":"f806d436894c4f047dfc1b85df7bd72d6799db35f68f23cd0d07fd6e1208b65c","split":"eval"} +{"domain":"finance","full_message_tokens":206439,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":384,"sample_id":"sample:d5b28268d4067cb4ab08b9a04932817c243bba08a22f200fa43a953edaffce09","semantic_task_id":"f9acabe421c93a3bbf8164ced061fa649b8e43632d361215bef8ebe0836bd07b","split":"eval"} +{"domain":"finance","full_message_tokens":95100,"group_id":"0000908937","output_file":"finance_eval.jsonl","row_index":385,"sample_id":"sample:c4fa7a7d127a8355852d5a3aab79842ac44ef30fccd41dfeb59c26b3f6b6940d","semantic_task_id":"fce1a7d7e433d015f9998ffc0434b468d3ab4155059c3265cedfa53b00b06c61","split":"eval"} +{"domain":"codeforge","full_message_tokens":49440,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":0,"sample_id":"3b9eef5aca45a40b731d3d5d2ebca509a2bd090bd5a35261b6cb2ee0d7682948","semantic_task_id":"9825ba561b8a22db23f096bce414724b59f3318f4586ee3b148fe99c91727206","split":"train"} +{"domain":"codeforge","full_message_tokens":49414,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":1,"sample_id":"61832e08254a508fb4b9bca1208e7025d2ec73d5a3dce4f3e761cc0b6fab3573","semantic_task_id":"fb5510578226ef2c4897c425331373425418702af7882905d46ee51180728a6c","split":"train"} +{"domain":"codeforge","full_message_tokens":49575,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":2,"sample_id":"4d710d5373e32fcf5aabd708cd59d7d14790936342fbf366d7adf498e26569dd","semantic_task_id":"597eceb9607ba903ac9e76158f1c0657a385d14ad18ae5f12b1a42fa70c4a545","split":"train"} +{"domain":"codeforge","full_message_tokens":49452,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":3,"sample_id":"371c5ac46fc6d1a75f5aab314198365b727b45d0785218e9c85a2b94fea57146","semantic_task_id":"824b8968df5a3f39c0444e08011060367917f3a869384c83c1f467e28025fdea","split":"train"} +{"domain":"codeforge","full_message_tokens":112565,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":4,"sample_id":"83c97bb235952b5d8d88621b37f6aa883015516b70c5753a5298902e7af257e3","semantic_task_id":"5721b9bc9ce903b5ee3c04e049631dc60f41ba3f8fff93d4aa86da5d64ae10e7","split":"train"} +{"domain":"codeforge","full_message_tokens":112559,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":5,"sample_id":"20d495dd351f420b4ec4b64f5c9810ecc50972425f28e90acf7a08c5a149db85","semantic_task_id":"d519cd4cb8b889c4f30a17f80fab3abee7afe8a65f0a80fefa5455a695a02820","split":"train"} +{"domain":"codeforge","full_message_tokens":112714,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":6,"sample_id":"11e16e46a7c11f130f6c5bb056a084a8c951aeeb5f3164290c5b73828ff63723","semantic_task_id":"f1e0c490f7224b842443b75073e62cdf843cbe5ae3b6a9da342e3685eb356930","split":"train"} +{"domain":"codeforge","full_message_tokens":112638,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":7,"sample_id":"cd7948e26ef85d3c6dfdee5c8e9123f71217b71af9c0c2bbaf93ba9f5dc37b53","semantic_task_id":"11eb8e7d7a8c138dac5317dde28e86753897f7928af7ba33427637754fb0b496","split":"train"} +{"domain":"codeforge","full_message_tokens":112577,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":8,"sample_id":"fb666a926d483ddf2507085cffd93fee758c053348a9a55ea9ecfa3d40e4795e","semantic_task_id":"f6bc007955b38a432de0aab1f119ef4f9b87179a5d9a6ec66382f034f64c5968","split":"train"} +{"domain":"codeforge","full_message_tokens":104526,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":9,"sample_id":"2ea5f007a4cf3b8dc2192dbc8930086ff984c8e4c54f7c8cecc4d30cb331069c","semantic_task_id":"6c9b2edbf0b46cd2240e49b6d08992b35a9e772a169f3104a3ebf30bffde5a06","split":"train"} +{"domain":"codeforge","full_message_tokens":104500,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":10,"sample_id":"d35b15664b311275541b362ae3afc3490f856e8db69b9d3c72004162144680df","semantic_task_id":"fdd7faa5dcd614a5251548767ed9a12620cfa0692147156ff6d3856dc01af911","split":"train"} +{"domain":"codeforge","full_message_tokens":104678,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":11,"sample_id":"9929549081f85a078c2e2d82d20bbaf85bcdc053fc5a3e9781bf983f79894652","semantic_task_id":"4f204082017af903743f7363930896acd45ad9f16e87f0ff0a323458a75ef96d","split":"train"} +{"domain":"codeforge","full_message_tokens":104585,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":12,"sample_id":"6d0bf444ca1518fedaa4d09febcb65f48cbe446eac3d8bb785e0fe51e7de82b0","semantic_task_id":"c16c39ed5f6919d65a24ae0cdac9e0a1d5fd0011fe4cbc0bc3e6d3353b3aee78","split":"train"} +{"domain":"codeforge","full_message_tokens":104538,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":13,"sample_id":"e45b1080b822d1dfbdb825b5b50773f77e52278dd155d79f9f6a6ce54b311f0a","semantic_task_id":"f4750d0babe322b47e50f2429aa5f06545041b9b857b172f0b17d8c61f8f8f4d","split":"train"} +{"domain":"codeforge","full_message_tokens":39007,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":14,"sample_id":"1ab4699ce668c728b2a8d97c5a61777ef3a0cc0591d30bcc0408ada7c41fb775","semantic_task_id":"c3fbdfb6de8e62449ecfe1343b2aa7547c544808f5b238a46eed26de8fef8fc0","split":"train"} +{"domain":"codeforge","full_message_tokens":38959,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":15,"sample_id":"bc8b97e4cbe8964ecd57d6d1ef92839d8e4a6dc8b57f02b7528b77ba11d23650","semantic_task_id":"68ad312dddde3255f9a4296aca17f16aae0e09a79cd41b69961d6a6869831f58","split":"train"} +{"domain":"codeforge","full_message_tokens":39137,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":16,"sample_id":"b9728dd70caef85cd69b8b15c5c7b67440ba7b8c7d531f3460c599ca3db48c6a","semantic_task_id":"bd2b0d2ca97f30ff390fb433f9b2b151dc4a7ba8e8b1da40743544e0e2a3686a","split":"train"} +{"domain":"codeforge","full_message_tokens":39019,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":17,"sample_id":"a98f5e42d8244bd20dd29cde6ec4e572b42384e6f7b7dc82fea01846ccf2752c","semantic_task_id":"92de541e785e85a5f1a601abeda3bde015b1e268be1ad3b577786c71165d79ab","split":"train"} +{"domain":"codeforge","full_message_tokens":33233,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":18,"sample_id":"5dae9096b047faf16622586bd14bc4a4d46e8cccc9c88572e47936edfec6d160","semantic_task_id":"7b47c1b917d86ef5f5a9d769e1e6b3c18f27336adb5de88cf9c2710b9dcdedd1","split":"train"} +{"domain":"codeforge","full_message_tokens":33147,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":19,"sample_id":"e47ac1afc78a9a733bf403ef4b14314078005a7370b05cab9aa67e258b4d39e0","semantic_task_id":"1c883c982e48ae1f534ae54eb0cc31881d8889b6a418ce1e415e999408765da3","split":"train"} +{"domain":"codeforge","full_message_tokens":33325,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":20,"sample_id":"664580464e3d139bdc8b5faa4f88c66161a3faec003ebe446f4698a877092f32","semantic_task_id":"c2ba501f07aac84bf8c5c6351efdfcd831cf6fe58f07bfa8a6a759100e7ed305","split":"train"} +{"domain":"codeforge","full_message_tokens":33245,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":21,"sample_id":"bf2af1f40e956b45eaaa54846ed365e9d5f93a0191cf5a4c13f21b700ded1776","semantic_task_id":"dc80f273aaa742221a922ea1e4b888bdb8a0479ceadc4cb00541ca438ec24462","split":"train"} +{"domain":"codeforge","full_message_tokens":141831,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":22,"sample_id":"4e9056e8391c388c614473d111d37413c9c3c3d36b2cddfbda641a6a1c3f3250","semantic_task_id":"7ddcbd1b43302bf53b53380f05a1d802f5094d39e1e9ad92d3d585f351288e55","split":"train"} +{"domain":"codeforge","full_message_tokens":141795,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":23,"sample_id":"1421f429001ac0d4d0956e24bb0a73065105e1546fcce679f225f79fd03730f4","semantic_task_id":"4622ac27f877f6d88b6bed4c55bed905236758662e49ed884f5a1f0bb32cdc5d","split":"train"} +{"domain":"codeforge","full_message_tokens":141973,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":24,"sample_id":"7a8681d8122b58d885635a948f524ea4eabeb572c538bdf7b112e62677376552","semantic_task_id":"2faa43e0e5fec953596e805cd37b521507ffd8ee312ee779f2b33e6fc48a21bc","split":"train"} +{"domain":"codeforge","full_message_tokens":141832,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":25,"sample_id":"482396557022d092dd0f4399b458dbbe260639712d45e3ca48b19152a74ce53f","semantic_task_id":"0de34c087647e63e7a207c8bff3f3fa5b1a0d3bd4c165aaeff653537d2eb0d0f","split":"train"} +{"domain":"codeforge","full_message_tokens":154206,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":26,"sample_id":"20206c21eda4661191d4729c7d1163392863f7dc0036f202555be6ab545a86fe","semantic_task_id":"4e89efd009211eb9e1ca327529de03f416e12054b2d2dd1f7ef18f7235dc2249","split":"train"} +{"domain":"codeforge","full_message_tokens":154220,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":27,"sample_id":"d4e4f207349eb5edbc30d36682164e466b8b5123d1936169e88425c38dc59470","semantic_task_id":"31f2b4564a1563d4a45c2b2420fe69079b53905ff8f9f708046dac54dc1b7b6b","split":"train"} +{"domain":"codeforge","full_message_tokens":154398,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":28,"sample_id":"c69e875ba070e2db559be7c6603e736541cf8ee67d1b3208e6f42ceaa2311250","semantic_task_id":"0a7da67c09bd65178daefdaa4b6d469ad0a74ac910a81b42fa5461a28cd5a52b","split":"train"} +{"domain":"codeforge","full_message_tokens":65007,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":29,"sample_id":"a77cb9a2ecb30f2632db83eed41a74ddab5c8c0a57894464d12a84ed6192eab4","semantic_task_id":"c3f3f199296f4aaf796f446c88e447f6ae2a9ef786e79f22054b8ae7f5c8b48a","split":"train"} +{"domain":"codeforge","full_message_tokens":65024,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":30,"sample_id":"babdd5eea42d2af5e0ac92ebf367d04492de57d46c1f41926047408704ba5fe4","semantic_task_id":"c75f047f26d23a67d4cc3e057aab1c6b05d01e723baad626fd0bde91285595f6","split":"train"} +{"domain":"codeforge","full_message_tokens":65202,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":31,"sample_id":"979856fc74226ab6259aae9d63561b1f5c80f6a0fee33771feb64ef1bfd6460c","semantic_task_id":"d13f14cf2166c9e4cc480ef110953eda33987b57ccd4e5979041b923ae4aff66","split":"train"} +{"domain":"codeforge","full_message_tokens":98548,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":32,"sample_id":"d9939aab7375ef6990f5aec638cd95d12a7defb4b572ef03882d762a5ed2d083","semantic_task_id":"e22b326c53db073c8d169cd1e685d10753b082a6ef5b6a87cb11667637eda0a2","split":"train"} +{"domain":"codeforge","full_message_tokens":98549,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":33,"sample_id":"1d7fb49af638306a00f9bb161177929dd3ef3e4c1513c5a13b666252d85c1be4","semantic_task_id":"406edcc4ecf2453abd8afb263873d55b0161f9570d3b78f738312bc2d1c6a136","split":"train"} +{"domain":"codeforge","full_message_tokens":98727,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":34,"sample_id":"7115e41bab4d0860286622ddb92c0a53d1deb54732db73377fbe5e4b2b050df6","semantic_task_id":"e03e6f44d558022a3f0fdd98926a3ba852b35488d564740d8c65292238e694e0","split":"train"} +{"domain":"codeforge","full_message_tokens":127366,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":35,"sample_id":"56d01699a028a663fe4c5bac09e9fe45510d03034e81ef1a34edea02d1e05ca5","semantic_task_id":"310453b1f4d5cda0daa6ae0ace421de9ada3942502682ade05db8d442cfe3c3c","split":"train"} +{"domain":"codeforge","full_message_tokens":127324,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":36,"sample_id":"f7a4019f624d4426274c6522b65bed06cf68c2fab544f8edbb4bbbc52c261996","semantic_task_id":"8dbad518c1561d72a81b4b678179ac225a1953c8be04c7dddcfc61607735d2be","split":"train"} +{"domain":"codeforge","full_message_tokens":127485,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":37,"sample_id":"2b83fc1fcc23436ed8c8caa7d5654dc14993af6a1ec620620631dab58ddacc41","semantic_task_id":"317dc32e183aa88749833320b0381779d6fb21d324616db3b95fbfcf2970f4b8","split":"train"} +{"domain":"codeforge","full_message_tokens":127416,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":38,"sample_id":"833b27415b548c4c953c61fdab9fb2922af101c18ccf50e33887056253562d35","semantic_task_id":"e973aeac77f9c39b23748da7e14bb1008c86d62a6ab224fab39e96ddc7856116","split":"train"} +{"domain":"codeforge","full_message_tokens":127378,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":39,"sample_id":"2dc4f03e5b794251d9d879f45c7937fabd3a722fe28ac02d596f0208b6d20b40","semantic_task_id":"b357c6dfc41674ae798585631de7c727472130b9a577df55ba79e47d9f028d29","split":"train"} +{"domain":"codeforge","full_message_tokens":134690,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":40,"sample_id":"182d73aa9fca6d3da36a562aabf35a7a46326a42b80f7b4b6b050c2282ea0330","semantic_task_id":"648cd05cac44368ba5992d210929815c82c5a96e5a5fa4e3a2c46e40f503c999","split":"train"} +{"domain":"codeforge","full_message_tokens":134644,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":41,"sample_id":"df18b7f3e360660aeabf3dcba9c6fd5757668807a6b996a6ec0fbc798a4ea7e9","semantic_task_id":"54e3c03a5dbd56a9c6be7f7bea8eaeff737aeee33b41e82e47d9badd1b0814fd","split":"train"} +{"domain":"codeforge","full_message_tokens":134805,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":42,"sample_id":"d33d74f1832d460eb69b8041ea9341a35a1ac4f968132803a993337497030512","semantic_task_id":"52595dafca2e06a6e058e4ed54007ad3d1f0e8734b66765ce7f96557b291a6fb","split":"train"} +{"domain":"codeforge","full_message_tokens":134736,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":43,"sample_id":"b696c0c13b5fd67ce626d3514b3220b05d3e693ec631f626a949f936eb259cde","semantic_task_id":"ed050c487ba46e9d0492775129f1258789375abf2039f56a71c060a13b8117c5","split":"train"} +{"domain":"codeforge","full_message_tokens":134702,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":44,"sample_id":"4ad2e16c87c9ff38083828baf08014e2a8569eb5b692a479abf3ca0ad5388b5b","semantic_task_id":"39bb212a21d330c7b602e486f14ceb485cae8c2bace70f8973c6415fabe1b738","split":"train"} +{"domain":"codeforge","full_message_tokens":117021,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":45,"sample_id":"3b641d415e77aa325115eedfd35df5441177e73728e30c2bc3a0ffb4fd8da185","semantic_task_id":"d38bc30e7f796cf89bab027c4aebfe01b5868ae2e929c7ecfaa1b99942eef89c","split":"train"} +{"domain":"codeforge","full_message_tokens":116957,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":46,"sample_id":"6732ef99896ffc56a57c4e458b5255f5fe380a2dc7745313b7b1bbfcfd02f314","semantic_task_id":"e2c4e4af550bdd499e5a31399a1817686a1fe9d1414353428511284daa18b65b","split":"train"} +{"domain":"codeforge","full_message_tokens":117135,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":47,"sample_id":"1019b172b879684b5f1e291ba1a967a57fde8498fbf81c830c44c29cce5b12a9","semantic_task_id":"f3340c5d4ebf6f11a56a18602c2f4f2479bf95e67457c51a052dac746dddb5ca","split":"train"} +{"domain":"codeforge","full_message_tokens":117049,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":48,"sample_id":"11e557b9251d03d7063b01af1d39c563907b74ba3481d42e1296e448689e7d32","semantic_task_id":"98b4e42bad6906a302b6a2e1c5a0870cea510a5bd7be72060d94e40c480540d5","split":"train"} +{"domain":"codeforge","full_message_tokens":117033,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":49,"sample_id":"980a78286dc95d23609f36129eb5b182e89e24ae0b44dc39bf401be13e3d828c","semantic_task_id":"5895314c51db969a6276918d816d773c9afc88b2256f50787b59754839163016","split":"train"} +{"domain":"codeforge","full_message_tokens":55261,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":50,"sample_id":"c6ba27cfd810c0b7765e23bffcf493d788fa6d16973ff06c58357e682762d559","semantic_task_id":"bfbc38cb8a293fd6621052a373c40f627f689bee287e30ccb89df759b7960a86","split":"train"} +{"domain":"codeforge","full_message_tokens":55141,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":51,"sample_id":"dbb9e3e63e21475eb4263f6bb5eb8e07ae206f9b7504205cd34b537d28bf7882","semantic_task_id":"e90cce00d7cbcb1501fdd951013474493355ecc8262cac8a30d784298f31422d","split":"train"} +{"domain":"codeforge","full_message_tokens":55319,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":52,"sample_id":"ef8ed65c2e93ffd1c657b601bf2ccaf88f21403e3ab8019a00a7a9aef9b4c374","semantic_task_id":"aac27898b642d3a45dc38aaaf4070ff2ad9700fb6276b9ba868ff0d1741ccf64","split":"train"} +{"domain":"codeforge","full_message_tokens":55273,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":53,"sample_id":"15a9d09d5b4e3f26de888a59a92c93c1e5c69faed3f15a0e022f375c83fc2b3d","semantic_task_id":"db33ec984294ef0350aecf3b891d137c34fd0c2ce11835e6008fad7d3a797dd9","split":"train"} +{"domain":"codeforge","full_message_tokens":45134,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":54,"sample_id":"40d090b7b090b7ed7e24a824f04ddc669a8de06dbdab9323d4e0ebf517d91cf1","semantic_task_id":"6e408669eb1c10f8b51dc22525c7616c7f8f925bbb21b467d8c4bc907bb7ab3b","split":"train"} +{"domain":"codeforge","full_message_tokens":44999,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":55,"sample_id":"f67e719c68ed32f7a4ea7860c3d5d0fe6c2e03cbdfa7e5e516e29a563f91c6ca","semantic_task_id":"79a710aff4b141f40aad6c5f1015c8e3d4e9e6485ced110eb4333d5d68a8f959","split":"train"} +{"domain":"codeforge","full_message_tokens":45177,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":56,"sample_id":"1ddb885577adf2819a7adc4b68ebc467ec89b68389e15975c0098f42509ab67d","semantic_task_id":"9c1d319dfa454048e479d6b6ae38d2ad8e1e808521430498da68aaf9843d6116","split":"train"} +{"domain":"codeforge","full_message_tokens":45146,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":57,"sample_id":"46856dc9f05a8250c00c7688c430e32266c9c758f07a24f12e2dbfc05d238c97","semantic_task_id":"95645a5603e74f0d56d4efbd4c05d008a39a7d858bcc0fc45fa64bab51f92c89","split":"train"} +{"domain":"codeforge","full_message_tokens":158066,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":58,"sample_id":"38ce02b1621c5e226d18cf824983dc87a463f793904ecf83d794f52fbcd884d0","semantic_task_id":"5a3e82cda527d359b6389b40be0fedc709dbe8d445d8afe24885059595a5fa6c","split":"train"} +{"domain":"codeforge","full_message_tokens":157958,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":59,"sample_id":"d391dc17f4398d5870ed2f59580d2c318ae24d2872c49813dc93705eaddcb17a","semantic_task_id":"731449344424e6cef5d34bfea943edb73dc92a327ee8e08ca72059e721fdc203","split":"train"} +{"domain":"codeforge","full_message_tokens":158136,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":60,"sample_id":"9ce961a5aad7cfe8f40397c7a6e4df88722f9c8f16470ee4c3072e2c8a30e8ab","semantic_task_id":"452768f8b4bd6cfd6fd9a250a54bffcd99da4bc76d5f8b1e531b6ed1f9ef571b","split":"train"} +{"domain":"codeforge","full_message_tokens":158067,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":61,"sample_id":"473c1d29d8b24541f10bdffa424c6ea8f3c1c98b95a46b382bb9934d60b7e07a","semantic_task_id":"938238687155f6d04e657cfcc3756332c7a5ed245332e9a2479ced4f4e913b29","split":"train"} +{"domain":"codeforge","full_message_tokens":166107,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":62,"sample_id":"60bcd2abcdc1495180d8fe3b0dfe45bfface7beddc2cda69a893605f2ad40c79","semantic_task_id":"3f500b66daf3c3823d488b0cfc894944fadfa9ed41175225c759fc3c1ede610b","split":"train"} +{"domain":"codeforge","full_message_tokens":166066,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":63,"sample_id":"4d559b30b89f91e83302451d701c3b6985bcc02338e118356b444ecb82b61a39","semantic_task_id":"d0f3f6f2b923281e01680d8d1b15b9c44ead1eb01f5e4092138da730ee791ce9","split":"train"} +{"domain":"codeforge","full_message_tokens":166244,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":64,"sample_id":"d9ca537695bd0d46a513f7691ac4484e10bf451487585f7bd95f5b3b2a3b532f","semantic_task_id":"54f2f88b8b35a55cc9df186b0e8e0e67ead3168961955d971216438c8f794ac8","split":"train"} +{"domain":"codeforge","full_message_tokens":166103,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":65,"sample_id":"b520d58bb326d1021572de031e41b0672f6403e45142e9db0ddb906624089377","semantic_task_id":"aa948d885ecc5734bf8941664f4d45b3e72cd244b85209eae96f8e87a5c3bb1e","split":"train"} +{"domain":"codeforge","full_message_tokens":190437,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":66,"sample_id":"d47c929913b9f930d433f94d52319bfdd87d5516769dd57cd27c2af0adfb1ae3","semantic_task_id":"35f252e641e483d15558196b4f853ac4e41f8a7fd5675268c72b1ea881ccf201","split":"train"} +{"domain":"codeforge","full_message_tokens":190449,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":67,"sample_id":"d1a097e9471f674c2ff07f3acba0ffca38086930beed3dbd796c8ef95069e83c","semantic_task_id":"0b91e2f8514d52116b9f9dd2dfa58d05820ea5b25a932194f7af722a147073a0","split":"train"} +{"domain":"codeforge","full_message_tokens":190621,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":68,"sample_id":"6e22224b0dfd40addc8ccfc055de7db895331708d9adcd227aeec088f8db9d12","semantic_task_id":"8a28fb542df4400af780d61d57252d68b6abcbd4a166f5fb1aced0f7ae6879ba","split":"train"} +{"domain":"codeforge","full_message_tokens":122728,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":69,"sample_id":"b49338a338ec3cbc6fc5b8fa071bf6af4efbda218c7c64b0140751da78eaee06","semantic_task_id":"4ad8ee509ace57d58413fa57de47f69a975875aeacc4b2b3394dc24e59d775dd","split":"train"} +{"domain":"codeforge","full_message_tokens":122730,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":70,"sample_id":"6f3f61e8f5d840aa8e30f14e47461ff81517144f5832ba485e68504e1be9d850","semantic_task_id":"a31d337d6e15c32fd7a8b1d4292ce1e4233efde04f82dd426bbf6e4d068b88ee","split":"train"} +{"domain":"codeforge","full_message_tokens":122902,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":71,"sample_id":"4c44670abd9e7e58b3a5191d65ade8435822f4784d756155d841d0bdfc451944","semantic_task_id":"48ff84ce17e5468098bb9fb37aca90bbb1afc6e9b8e7ba99149be0dd8523cafe","split":"train"} +{"domain":"codeforge","full_message_tokens":149491,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":72,"sample_id":"714820a17cf360187a20a415c7e179a2d1dc4728f5e379c01f3e04599f7aea9c","semantic_task_id":"f18b952735adacdf56420c66bc11503ff2a813d4331c72bbc1706393ed6c8265","split":"train"} +{"domain":"codeforge","full_message_tokens":149421,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":73,"sample_id":"f58582ceb78f470818d711a68cf30c9864abba5ca8e53ff5fd10499a06283689","semantic_task_id":"ce5e76043a643245fb267407686c96d70b936cd220ee517d486004a1d7a15bdb","split":"train"} +{"domain":"codeforge","full_message_tokens":149576,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":74,"sample_id":"26e3d2ff208aaa12b8eb5be518c1a439aca8e4a2545857e887108e843de530d8","semantic_task_id":"2ffae088bfc3a1046cab3d05eadfd2dc5e907d6efd5240313658a30ef56c26c2","split":"train"} +{"domain":"codeforge","full_message_tokens":149514,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":75,"sample_id":"4bf6acbb3f76b383cd4fd26b781610413788708174d6d043910938080e4daa19","semantic_task_id":"202bd2b5b171be12cbf6db60f3875a27a4372252afb37b226883c795580729a2","split":"train"} +{"domain":"codeforge","full_message_tokens":149503,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":76,"sample_id":"d3047e311b53bf22fcce60295215fd36bca5a4a4ef41278e33377f1ad76e9c74","semantic_task_id":"af12a2d4dda51b2a2a5ca8db63674754f11f5e5f4bfefa9ca2ba92d8f778b841","split":"train"} +{"domain":"codeforge","full_message_tokens":147185,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":77,"sample_id":"b50836b73c579c1af74727e0d301c6827933d3d75a2136b108bb41268e8a3073","semantic_task_id":"3303441377cbaeb73b4c36d854acd42b1890a3e42a11b7b79971b27e7f4feaf4","split":"train"} +{"domain":"codeforge","full_message_tokens":147101,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":78,"sample_id":"0ee4cba17da33dc11e41a31531cd506c64273dae470a775661cada2ffa94d9e3","semantic_task_id":"c06243294241f70cc4a6cdb0c3568837c1bf1f37104aa7cf0f29e7e445e3eeca","split":"train"} +{"domain":"codeforge","full_message_tokens":147262,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":79,"sample_id":"99264757994f7f6f0aa73ef815d68026b2cb62341b466a0f99eab0a8dc555d7b","semantic_task_id":"dfa9f00e6a8e27d60826bd76ba0a72d8d501cf5e0a649c15f355bb86188553cc","split":"train"} +{"domain":"codeforge","full_message_tokens":147200,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":80,"sample_id":"b7b2709d834c3edcea3ac05b4e7af25659ccda509a16abe63f9b839f2d0fc9e3","semantic_task_id":"c6c733a6414b1047dfdac283746889d46bbd768942f0c4c5c96e28b4db13d814","split":"train"} +{"domain":"codeforge","full_message_tokens":147197,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":81,"sample_id":"12be0c5f71e6cc795205a107b4f182554dd85abd4a98d0a5a4ddf58c4b44799c","semantic_task_id":"07e3c060bbfb78df535b8a35505f6daf39f3bb16217f591deea2609ac8fd47a6","split":"train"} +{"domain":"codeforge","full_message_tokens":133275,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":82,"sample_id":"8658c7de67a43a128d4577082e1fa68d3da13b9cda4fbd249aebccef25f8c4cf","semantic_task_id":"7f4d4b53c6541814bf0a02fbf08a0598d94573e44121dd92a7db67b2b0122423","split":"train"} +{"domain":"codeforge","full_message_tokens":133139,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":83,"sample_id":"58652f9bc4de895b8938521ac12cae4a411bb783e27b5759d0e4893b61294360","semantic_task_id":"76bf9ded5146e42792f06a68926a8f07e324a7b80b7858b5339d3fbe9c448cf4","split":"train"} +{"domain":"codeforge","full_message_tokens":133317,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":84,"sample_id":"75640308ec10e4df92db5e00beb6352f11eeb51626faa9db6bddf3ba8ceb5fe0","semantic_task_id":"111183f291d8966b14ab99aba6bf0c6b757044b9306eaee082965764812980b9","split":"train"} +{"domain":"codeforge","full_message_tokens":133238,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":85,"sample_id":"10ef1c4edfb01aa9cbca799cf159aacfe19bde895307a5f7909d637aac113c0a","semantic_task_id":"330dccc05d96159c21ce29f34b21343be9912a479973a18fb7138aab1d9fff68","split":"train"} +{"domain":"codeforge","full_message_tokens":133287,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":86,"sample_id":"d4c524d53f18d0fb79be24b79c11b9f774a1464d2c86bbcd5efe9dfdee4ab5e4","semantic_task_id":"2ec8a22eb8cf22d826b5df39ed50c6874a1a892df5b5a96dbbb0f006f0f26a42","split":"train"} +{"domain":"codeforge","full_message_tokens":67162,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":87,"sample_id":"583cc9c7c188fb4ca62af88d86587a1486a079b8196d5cf1ab6edc01c2eac2a5","semantic_task_id":"02719e4cd8e961a98437687e6a1f0e8ba91cab7161ef6d08c0a9c1bea8852b6e","split":"train"} +{"domain":"codeforge","full_message_tokens":66993,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":88,"sample_id":"bdb472a37b582b514e0476f4d586ce2c7559618f9b8059ec39c4fac2dc3a226b","semantic_task_id":"361fbe7424a650e7f7cb1fb49ce2ec57bab861e7a902798e37630e65cdaacaf6","split":"train"} +{"domain":"codeforge","full_message_tokens":67171,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":89,"sample_id":"1b7ddbf5a3c5ceea80803c7d971daa4047fb09df14109d92203db31b98bb6ee3","semantic_task_id":"e9a40b9d50e86e534ec1f2fbd44a457ce3cbc6d32fd9fcd5f60fd74497b8bbe4","split":"train"} +{"domain":"codeforge","full_message_tokens":67174,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":90,"sample_id":"2893ae672a0de8380616397d54ee68461b01592d04ef0550efdab3e600f1ae3e","semantic_task_id":"b38c563c9c171f612273745b93bc55cf0a815b4e00476a1bc4ae65404e2c320f","split":"train"} +{"domain":"codeforge","full_message_tokens":170580,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":91,"sample_id":"183a07f32a0fd7cfd41122e86e1885b25b04f0805e4a6393176d025aa186cf2d","semantic_task_id":"6953cc2069e40551cede074f5e2451193561b7001b8f28a0eb53654db8161dc9","split":"train"} +{"domain":"codeforge","full_message_tokens":170434,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":92,"sample_id":"65137d36ab1c8c3e9ee57aa56d2ec1ef7eeebe168990b4e052ded235400e2f03","semantic_task_id":"cfe9f6cc89aff3da6369a4ca68bae149511a71b788a50d30eb21bb8fde59153b","split":"train"} +{"domain":"codeforge","full_message_tokens":170612,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":93,"sample_id":"48e856a35865717e3174b60413cc0943aedfec84e6546548fc6ff59f09be1753","semantic_task_id":"26c59145e16b4516dbe6f08a206e1454091f9069bd00b58222b0f5a473689eee","split":"train"} +{"domain":"codeforge","full_message_tokens":170581,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":94,"sample_id":"3cf05e7c397d1fbd4909aeca51b4d088e5e982e7ab92717199d3313b18dbbd45","semantic_task_id":"cf8a15cac5bfdae7066aeba4f68ed4d56f7649b39ba273fedb81c1d2812d7d3c","split":"train"} +{"domain":"codeforge","full_message_tokens":182361,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":95,"sample_id":"ce33e92f74081aef31fbec12ebe70dff617667ce2c2d63267a195ca495927cfa","semantic_task_id":"97d007750789e0d0c5c018093755ed15c9bd433bcce3ac913abed012cca7ef3b","split":"train"} +{"domain":"codeforge","full_message_tokens":182248,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":96,"sample_id":"3777d3ab0e531d42846642de918ea93b15c27fa72a2616d0e229be83d367e728","semantic_task_id":"e1512ac9f093d569a8e644bdb48a3dc26f831447fb67ba207aad482fca775a34","split":"train"} +{"domain":"codeforge","full_message_tokens":182426,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":97,"sample_id":"1f8b9f5c968b78d9237813ddd06152259550d6ce6171ac5d5a80983b98dd12a7","semantic_task_id":"607d695a56048a845745a9251e0c210fda19414bfa4067224276b7c11e257acf","split":"train"} +{"domain":"codeforge","full_message_tokens":182357,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":98,"sample_id":"c04aaced878efbcd8b701dbdc4859cd2f0cc000dc8db4272466d9261a60227c5","semantic_task_id":"6de1d723effbd4e03420fd098aff1e6a996d9e902e276bb8904cc413215d56bf","split":"train"} +{"domain":"codeforge","full_message_tokens":202338,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":99,"sample_id":"8405c33a479b8b9408f637acc647c58b89f1981e1a969402ffafba8f0f762940","semantic_task_id":"f59cd4a16af98d045e0d6dd486351419d64656bda4b814f52b1ecb19f54e9a5a","split":"train"} +{"domain":"codeforge","full_message_tokens":202289,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":100,"sample_id":"01b9ba9ac7b15383fbffb21b0e9f39c13e62f2cb4585c4bbc6114eff51a5f0c3","semantic_task_id":"2109e0c9eaff2082fcdfbc6ea03e0f6bbfeff09f3f3dd2b803b0c36790a4d256","split":"train"} +{"domain":"codeforge","full_message_tokens":202467,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":101,"sample_id":"3164450fc8eac1626ff2b9eaf899178b9c285d743bcbb3ad8579d91c9b1940e2","semantic_task_id":"c68661005edfec71e67684d798ef6e0c900ffb3737bd4c93cc5745f3062f5729","split":"train"} +{"domain":"codeforge","full_message_tokens":202326,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":102,"sample_id":"8c4b4edb736f49633c0fa1cd67bf4f9dd7486228d29adfca9f64b594c85a0c4e","semantic_task_id":"ad1632d39396ec9741298974dea63c8163ba8c4c4dd439434aa6d9adb8257979","split":"train"} +{"domain":"codeforge","full_message_tokens":248266,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":103,"sample_id":"fac4aca7334b0d0c0dcc24d341a3381acd023797687a3be7d5fc784010d548b3","semantic_task_id":"d23d009940565372055956480436537acafcdabb4430ac20327e41017183c164","split":"train"} +{"domain":"codeforge","full_message_tokens":248263,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":104,"sample_id":"e728b1029b4b7d3be6348d91c4f2d0ee64407e4387cec311219a83007f6e7c72","semantic_task_id":"fa8f34354445113c997bd553ba2fc369aa14490ae31238cb588a1fc03238ef54","split":"train"} +{"domain":"codeforge","full_message_tokens":248429,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":105,"sample_id":"bef6d7e21f405b7fcf9df01f0e5c0853fcb90099437825abe8ac9333784dcbf3","semantic_task_id":"2bfc3120f8c6a418d40ee3ec21963057cc9b6978a45a875ec310ccce1bba8d1b","split":"train"} +{"domain":"codeforge","full_message_tokens":178240,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":106,"sample_id":"02e1fe3d12ef1bc44e5bec262057c2ed3c0ca65d5ea4ead7a7941e96afbcdf54","semantic_task_id":"557b817c37e44853496685d2048e0f3fab6f9d4f70a6106a00054e163595ef96","split":"train"} +{"domain":"codeforge","full_message_tokens":178054,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":107,"sample_id":"f7e8c1c1883c86454ce7e1c3ba92659ccd5f09ee5f63e7f0a4413659ea7b6be7","semantic_task_id":"0e4ef3a79e2c6b9b064b95de932aa70bf855a5a232e142888f5c1f00d619a017","split":"train"} +{"domain":"codeforge","full_message_tokens":178215,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":108,"sample_id":"479ec41c7efeb563a5e8583b053eee926219632a65691c2e24f382df046a761e","semantic_task_id":"90aecb16a28162190083e1672650d3158289f3abb40e2771b910ec73aa9b41e7","split":"train"} +{"domain":"codeforge","full_message_tokens":178167,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":109,"sample_id":"370de0b83baaf0b6a9ed8923f7e014b3dd05f98127f3aefe60c4973d2f57f20a","semantic_task_id":"bce7749bace6c18f947d4ee54660a874d578c1f4fd5494906c7ef0871f1d8c86","split":"train"} +{"domain":"codeforge","full_message_tokens":178252,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":110,"sample_id":"d3d31f0c47523d2c74ceb68d87dad12ce85e3ecabe5e0d807ed6e09267ed5e4f","semantic_task_id":"74dc283857935891f604ad60a38e3ceafa95ae1255b03347426d9b72696c42f5","split":"train"} +{"domain":"codeforge","full_message_tokens":175340,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":111,"sample_id":"86803b9770a96a584e9c83bb187f1af80db13d2366943f5022329240f6d9c44b","semantic_task_id":"ec98083b83518f85024d451cf0ff78a9ae9b1549bb9e56266533354da21bd490","split":"train"} +{"domain":"codeforge","full_message_tokens":175135,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":112,"sample_id":"7b3f24d6d573f6f4ea30d44ca46bae58f53ea7f88c0503b93808e7f13aafc438","semantic_task_id":"0e6c437efa29bd67aa304ebcd069fece696c7bff540cd7d38392cc97c04012dc","split":"train"} +{"domain":"codeforge","full_message_tokens":175296,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":113,"sample_id":"dbe732cbb0a9dbc7098e29a31302e6f3273a66ffc2995ce8440b732101511c9e","semantic_task_id":"962b37315726d712a8dbae28b969bedef68b0b01bce914f0db6c69a361e37493","split":"train"} +{"domain":"codeforge","full_message_tokens":175248,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":114,"sample_id":"e8dde72c6b4d0965d47d339b3a70f6552a5ee03e049db9f098cfb7589c34bb17","semantic_task_id":"63ad6849590e98d21defa862393b9d77809b267bf2c8962072187b665d6fd6c6","split":"train"} +{"domain":"codeforge","full_message_tokens":175352,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":115,"sample_id":"1e6531eb21a7e5cf5f59770af9a0250a57cb43f20447e83aff13291cbe82752c","semantic_task_id":"b79e4b76dd538a642d623378e25512451642a93da09d84ea4ff4bcbf27a6ff33","split":"train"} +{"domain":"codeforge","full_message_tokens":216893,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":116,"sample_id":"ebe0b24d7a6a4cf3cf95f0fa9c6a16ab65181f2ddb49def0716387eab3e7fc87","semantic_task_id":"6e60261aa92c7724e4b0c78ae6f738ad787fe9e4a486a83ebe0b8f76ce9461bf","split":"train"} +{"domain":"codeforge","full_message_tokens":216708,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":117,"sample_id":"89a86870a1bcb14ed128273f1c9912ce0a2facaad057c48f408c956f35ee6d8f","semantic_task_id":"becc79d9c5677f4d2877f901725c97263a6d4c206e45eee829145d252384bdc5","split":"train"} +{"domain":"codeforge","full_message_tokens":216886,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":118,"sample_id":"b4b1593e52595f9fa1c858bf42920c145377cf7d826d74ffac74ebcb5429715d","semantic_task_id":"807f0a2a0b2eceed7bb388ef849586a1a47c88d256e21c700a4c0df24861f6cb","split":"train"} +{"domain":"codeforge","full_message_tokens":216889,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":119,"sample_id":"f4ffd23fd2a545a00dab84d6810e00d298e2e4a1cdcca8b71704e6b0b0f29107","semantic_task_id":"c8cb0a63fbac04a488c74ab8856f9e906c58ba5f04fb2c21471d2a6a09b6c931","split":"train"} +{"domain":"codeforge","full_message_tokens":231034,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":120,"sample_id":"2d48973ffe9350b16539d0486541bb7c91c85271a1ef51e97ca4c26f957b6999","semantic_task_id":"afffed816e616c1be4b380495d5975a0c65134175737d1af092da582d6d6496f","split":"train"} +{"domain":"codeforge","full_message_tokens":230875,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":121,"sample_id":"07e2f21303c46e033e844e41b2f6eb769823b03d54c91b7602d569ab233c7e07","semantic_task_id":"c2c4e0678dba4ae15057189373d71e7ef56b2ecb5183c3bea2a06c7bcb1ae499","split":"train"} +{"domain":"codeforge","full_message_tokens":231053,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":122,"sample_id":"e58e75c20e38ab9d9b4fc00459e667ba0bd2f91e623d362f4d17ffb6ba1cfd1b","semantic_task_id":"81dbf4b1d597edb87360037f50d52196ccbff24df1f91b9f9c154e4650f24cb3","split":"train"} +{"domain":"codeforge","full_message_tokens":231022,"group_id":"https://github.com/astral-sh/ruff","output_file":"codeforge_train.jsonl","row_index":123,"sample_id":"bddb63be114434095cb6679af4777822b2dc86e89408a9942e1b685ee7cc65c5","semantic_task_id":"7c09c664310674e7626a5d97d501ff2157371358c1e2e6ec716fcfe1d8703604","split":"train"} +{"domain":"codeforge","full_message_tokens":99075,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":124,"sample_id":"2855aa140ae56d9b2dffd0c5734cacf4dad2bcf8f5c706f9fc5d42dac80ae39e","semantic_task_id":"a82dc52839dc566fbd59b0d673160103dc728f15e594bcc5464323ae7be380ce","split":"train"} +{"domain":"codeforge","full_message_tokens":98229,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":125,"sample_id":"6a8bd2322bfb8dcb385de49c4dcc07817936abe6eb60e4440beed9641675bcf0","semantic_task_id":"d0d85aa7a7a67f2dd570da65016b4d76b98fb938c1f95ebe5ff876b88d3b7ebb","split":"train"} +{"domain":"codeforge","full_message_tokens":98354,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":126,"sample_id":"218344132cdfc4d3936c6ae79fe126cdd7c792cba7dae498709475f9fddcd3ea","semantic_task_id":"bac2bd319fb9f4eef22c4cc6834a80bbba7d31c68dde007dfeeea6701cd4fa10","split":"train"} +{"domain":"codeforge","full_message_tokens":84896,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":127,"sample_id":"78cb8853aaef85304d7422e1b8045f44daf4f9530bfeb8ef9f8dfc138ce333ad","semantic_task_id":"1193a06abe691933954d111ef3bf63ed493493bcc056afde93e7f4dd0abf0494","split":"train"} +{"domain":"codeforge","full_message_tokens":84578,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":128,"sample_id":"5ce7c16fc39e2029bf5455c04d7c73f94525f9735b06f6409725f287fe2d0c04","semantic_task_id":"e4384c46d1b3e10b503aaba98d69876c58f6ef0ee6cc03fc1e229d555fc19fad","split":"train"} +{"domain":"codeforge","full_message_tokens":84709,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":129,"sample_id":"006773d63448a36697ea5dfd592e52c696a893e6f8fdfb4a2a3a3aa47e56e177","semantic_task_id":"e918419b0d5c90837b61c38692c4ff8305a5bfe4befe526e8c838f73625979ab","split":"train"} +{"domain":"codeforge","full_message_tokens":48510,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":130,"sample_id":"11be7c748c673ab626627cbf40f674dcdf9650794f2c96984c7b47aad91f4f45","semantic_task_id":"59f53cead12414e71d22da554076b53939a54a3ac0bc259245a19b3db91309a2","split":"train"} +{"domain":"codeforge","full_message_tokens":48365,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":131,"sample_id":"888172706a78efed287054b0e62c1e1c944f4a9e03393f53f7c82c770f8e8d7f","semantic_task_id":"8b2e899128c5dab264a6619bcc6d818172017624b2ab98dc4bf15f451647f80d","split":"train"} +{"domain":"codeforge","full_message_tokens":48491,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":132,"sample_id":"b5dfd07d85c33204bb8e99457fb4ba25ad469040c3555af23ac688a953f15384","semantic_task_id":"3bc7508cd10b863d634abcc28380696d3ad0358796d4e44758da4d10d3432ac9","split":"train"} +{"domain":"codeforge","full_message_tokens":38056,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":133,"sample_id":"6767be21d56279d8c27258ab0f7853f9e71680d211f5aad0db4f5392901878cb","semantic_task_id":"b516ccd4b1602981091df18fb36d3262c6f92e53d7e20c0fbb8c9f33350ab0ec","split":"train"} +{"domain":"codeforge","full_message_tokens":37937,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":134,"sample_id":"fc31c2b81cdbc0cdf352f5755a3fd63d97793bebd0ac1a25602ed5d5724225ed","semantic_task_id":"f7baad57b2980c06f055fb6a5283633f5be904f1954c4b91ea408c2e2161ed61","split":"train"} +{"domain":"codeforge","full_message_tokens":38052,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":135,"sample_id":"0301546c3c4f259d15d782078f73a562ca9339a55c83af693f76fe8f4538c48c","semantic_task_id":"378ed721813d050c2a270ae35e6bdec2824a4ad0204f0bce8ef4e34b415f1266","split":"train"} +{"domain":"codeforge","full_message_tokens":36700,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":136,"sample_id":"b7463d922512cd83a995c901d41cd2ee8eaca906e0d337a981439fe2e073e77c","semantic_task_id":"a47d390af8504e18d71769cc46692aad06474cae2b95de9a4e85310eb883a684","split":"train"} +{"domain":"codeforge","full_message_tokens":36531,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":137,"sample_id":"58f4f6c02198f817c0f87a8788313c4a37382b05113e62805bc5f31c8fd90e6f","semantic_task_id":"641ee1dd72e96aeb5e0bc3b849725de4b7a802ba243fcc267ba179b9c3179192","split":"train"} +{"domain":"codeforge","full_message_tokens":36646,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":138,"sample_id":"71fee0c9199187b4cc42deab71ea60cb125aac73d8c90b923b25c154b1cd5094","semantic_task_id":"d462aa1ab38c147391bdae1450f83f92340920fdfb3b819177f847d17d568d19","split":"train"} +{"domain":"codeforge","full_message_tokens":57074,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":139,"sample_id":"346d46c789959b023a2407856e296d9321e4ba5916c5cfd833c574eba22ec680","semantic_task_id":"9873c33f0a1c835b3460600b87a60d6e709d8eed866ec12f7b7e3e56d965c0aa","split":"train"} +{"domain":"codeforge","full_message_tokens":56827,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":140,"sample_id":"3bdf1252a26c9f1928b9e2d24f547fa1cda39afe7cfea6068ca4b8027bbc620c","semantic_task_id":"07ce8a76541000ebd938912c81c084c6f0b81970dab0a0dabe069cf178d68c8a","split":"train"} +{"domain":"codeforge","full_message_tokens":56958,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":141,"sample_id":"f9e92124d2b9452906c4b1dae1e5d0c0b940ce8d8791518c457d44b7d5586556","semantic_task_id":"0c8b361da5247e81a070e25c3aa809671aa54f7a18bd26041dcc4d533ac9cd2d","split":"train"} +{"domain":"codeforge","full_message_tokens":76097,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":142,"sample_id":"2bcbb45eecd5740c5b5ab49ab279e7ea8cb570314d67700df6a3495d13b76ff6","semantic_task_id":"96d7f1d47db6a810426ca53e7fad8c4d2b5f3249554576fc4224c12708034de3","split":"train"} +{"domain":"codeforge","full_message_tokens":76079,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":143,"sample_id":"766766edd8a2f70da4bb0026a5cd870a58da425075623cfbff2dc3212d4e4779","semantic_task_id":"1f860f5c9bf77f3c20082824a0ab587cee725bb0accbaa09e8689e3789984caa","split":"train"} +{"domain":"codeforge","full_message_tokens":75964,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":144,"sample_id":"76cefdab1a618afa1b9c2e4fc124a22c0e50a0bdca8bc7434ea69dc504796108","semantic_task_id":"38fd685c7779153fe24b3de690a8a36f6b69b741b3bd4515567ca1ea03bb5e14","split":"train"} +{"domain":"codeforge","full_message_tokens":76089,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":145,"sample_id":"d719e325bb481c1700e4ef0b76fc6200d502e3d4a50c93acaf183a869f66427a","semantic_task_id":"8a6f70e2fe84d28036eed0811fc8d3c94235ad5f5cfdb6c06c9ce1701e91f90a","split":"train"} +{"domain":"codeforge","full_message_tokens":58545,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":146,"sample_id":"8e55ab640757bc2570781020eef56ebb899e534332b32bef977dc9baa3bc6b97","semantic_task_id":"095e0c7d8c7cb9d3d5ec36436adf0f53968c36802189ebbda8821fe1a0624c89","split":"train"} +{"domain":"codeforge","full_message_tokens":58393,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":147,"sample_id":"5411edc101f8c48ff06f16ef5f2a9d077cde4aee25c39850ece16a1242c5f3df","semantic_task_id":"55191a9a97d9b21e401179f3a60ec22d7ca955340d180c67ed0db71ae0931f1b","split":"train"} +{"domain":"codeforge","full_message_tokens":58524,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":148,"sample_id":"a6710ce7c9706b314208bd6f93b928c58719e71562b59fddda5c6827b89a6aa4","semantic_task_id":"67425778df924ff2e8310f74e03b69f513be8e1f23dc3988402834997203b117","split":"train"} +{"domain":"codeforge","full_message_tokens":39823,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":149,"sample_id":"5d4865f491b3efa3f4fb9ff0f6f71e513cde62c2c9aefb4defe9d013092a8849","semantic_task_id":"9c8eb06849755febd277fed6222321486a1dd5243c040e1d29a9c33756fd1b19","split":"train"} +{"domain":"codeforge","full_message_tokens":39805,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":150,"sample_id":"1db06f8f27986e170f80e059d633831bb8f3db79874bb50b6b6ce1971ef71fbe","semantic_task_id":"65e8ca8579c82d387bed1ddaa5fd97cfbe2e4de7f24983ba3a29ce0180dd2853","split":"train"} +{"domain":"codeforge","full_message_tokens":39931,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":151,"sample_id":"d0010978217e8585ae768d81dfa2e2a85e14147b44c7db5c4921791fc2d99ed7","semantic_task_id":"f57ba4d82bcab33beeb0939fae91050f03cde846de41f83cc6ad796509c4d5c9","split":"train"} +{"domain":"codeforge","full_message_tokens":42350,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":152,"sample_id":"6883d286fd5849380fe945a99265f1be849d31d2873fe8ad06c0ac119f0f84d7","semantic_task_id":"d53889a4e203b67a2096a9449b0dfa68b07047dc508c11bc1ca15fe87ce6eecd","split":"train"} +{"domain":"codeforge","full_message_tokens":42328,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":153,"sample_id":"2edf3fbb0a7e30d51c7516b6b7b03e25eee4ec876473603183169af58f5a1c1f","semantic_task_id":"29bc6d5cc85eca7b87165769b75c6bde7b57ef4ec30c29e803de44ca95186101","split":"train"} +{"domain":"codeforge","full_message_tokens":42459,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":154,"sample_id":"f3d01d757695220b284c3d1eac225957c5ecd0c2479c305c4764858b6c88d327","semantic_task_id":"dbecb7d08537e0253a06713afe0fa6c6a14cf5c5ce6d4b49ba5113d13b6c0fec","split":"train"} +{"domain":"codeforge","full_message_tokens":51526,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":155,"sample_id":"d40a8ee41934264ee43f1d41a0005182b4ddc7b0fa6f2f3602c9ff8736f495e9","semantic_task_id":"6eb289d6b5ed9d97ad4fdb69b5dad63f6751f3b8c76a3d08b6a3a6f6f1a86da2","split":"train"} +{"domain":"codeforge","full_message_tokens":51462,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":156,"sample_id":"ba430ba38c34630b265786f42c1e42dd4cc9ad4250e1ffbd558566f11d3da993","semantic_task_id":"17f5bc5731459ba1c8178e340715ce73d45b1fb6e65d99ef6122392fccb5fe09","split":"train"} +{"domain":"codeforge","full_message_tokens":51599,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":157,"sample_id":"7ccc3fd4eb100ac0abe5e67874829e5b8a968762bf828384a6a78a3c31b1c73f","semantic_task_id":"ad1fde8a1b2d590182cd7c26af1694a497667a64668c26e5bc86c52e29555b45","split":"train"} +{"domain":"codeforge","full_message_tokens":72665,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":158,"sample_id":"eee36b2c6f0761a71ab1bbd97b61fecbbd5cff8a759450437810a6acfe468bf6","semantic_task_id":"35152e9d788d6b893aa139540fd7a449248bc85e613b214e18aa53f697dd1f72","split":"train"} +{"domain":"codeforge","full_message_tokens":71608,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":159,"sample_id":"d331b307f7351876158d3d4bca96b4295076f4db1b6a81812c8c47babf3f7081","semantic_task_id":"8a8f39dbaa7bfbf0db25eea4bdd3bbd1e54640bb45f5a36d0bbda723def8ae5e","split":"train"} +{"domain":"codeforge","full_message_tokens":71739,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":160,"sample_id":"2f4a2b456f3c13274fb02d4e6cc9e86133b01dfce80766673b9cc558ebdc558e","semantic_task_id":"7458880b6ec736322e34ca5e02e38eb281aa8f36831f13b8b98e0c81f8fe7dc9","split":"train"} +{"domain":"codeforge","full_message_tokens":69525,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":161,"sample_id":"0e3246e84dad6e992307b03491d5f7850e448d0d1fb381233d581ff3bf2c5503","semantic_task_id":"fc1ffe9889fd7ad6fd5eae45c0dd208f2cd8bfbf2e819b4af972e923370c6326","split":"train"} +{"domain":"codeforge","full_message_tokens":68433,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":162,"sample_id":"a8741db16ef0871a6a2b9bf6c4a9c86f3f0bd14ba74f114f5866853df78dbf94","semantic_task_id":"dc2ad41a6c9c0094d320a4128a9dfea176637075058016567fbc879d69699df4","split":"train"} +{"domain":"codeforge","full_message_tokens":68564,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":163,"sample_id":"810dd5faa74c3ed563f1ae07864afca5929fdf19dcb00fcb118570bf78756e6b","semantic_task_id":"856b41b350afe35a1c6edf0a7c18ba2d6c266c606a50610f7ce62e85a240520f","split":"train"} +{"domain":"codeforge","full_message_tokens":54874,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":164,"sample_id":"0c54dc80ab309d520daa35dd114e5eb717df44f210db80f7a2bf43b0b67da955","semantic_task_id":"ae2707535c209451124f72d45c0cd8cc8ffc3590d6196f98e3581e2b7a5873c3","split":"train"} +{"domain":"codeforge","full_message_tokens":54646,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":165,"sample_id":"457edbca0b19d286dbb0422d710dfcb8724e43fe36aa48dfc630cbf12927d572","semantic_task_id":"b2b752877f96c232dcae15ce5f0446c928c6f0803d8056e83398969c3496fc30","split":"train"} +{"domain":"codeforge","full_message_tokens":54777,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":166,"sample_id":"96fe957398e1b3d9449d428e97ca0bd9f0ed10a93e5c339ac6208b0e13f2d161","semantic_task_id":"4b2ee6f8917e4f7c070949b5868cde5a6e5a8fd29fa77b99407d7355fd637d01","split":"train"} +{"domain":"codeforge","full_message_tokens":61012,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":167,"sample_id":"5a406b0a885427bfac5c8003e79cdd90836819c22a9e67e8e25f8e2eb98685dc","semantic_task_id":"e2bac46ec0cfa5b17cac11645451af85d98f9270bcca7167447790ecdf437864","split":"train"} +{"domain":"codeforge","full_message_tokens":60863,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":168,"sample_id":"ead9682acdaa458ddc31126d04bd321add6a87ced34e9ab0d23ee9ff0fa04dfa","semantic_task_id":"f122f00af1904fbdd61768efb726a24b44b87db90083e646dc0817a6d4be6a80","split":"train"} +{"domain":"codeforge","full_message_tokens":60977,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":169,"sample_id":"8d3da621ed77d8537ea6d50edc59a20f5e6505d2dfdcfb7b30f3fdc54ab56ab1","semantic_task_id":"93ae51f698b4721b0c466b8234bef07a908654f202f4555052268c5c14e6a7ed","split":"train"} +{"domain":"codeforge","full_message_tokens":60879,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":170,"sample_id":"62ab277d8f219d4d1615b178472a8a96b1adc3e38927b3618b8ecbdb54da88cd","semantic_task_id":"a1cef69f3a7551cc4e91c4026006841e57e30644aa36e35ac88948975e987a41","split":"train"} +{"domain":"codeforge","full_message_tokens":93544,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":171,"sample_id":"939f6f3479c21df2c992c7c428124144a89a4208030cd50e2f442c93a64edd45","semantic_task_id":"26b39c0aa358e511ae0a9fb11d389e7379a8592c91493af1bacdc31fafe6b3e5","split":"train"} +{"domain":"codeforge","full_message_tokens":93512,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":172,"sample_id":"47a99d3e910f25f8ce3a0f108b4364ac14452564c967785975445252bf65b285","semantic_task_id":"faf70b894a68780f3dc5c9c780b9292cca8aa332b8df31ba3d3c2a7ce0f5b960","split":"train"} +{"domain":"codeforge","full_message_tokens":93693,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":173,"sample_id":"c65ecd4a3d417f6100637b556f62074724c3aa5bfdd84d1e7d65d20a576eb5e0","semantic_task_id":"8c26718385688695d6664c003e842ca40c16b534e89c86354387df50c6d474be","split":"train"} +{"domain":"codeforge","full_message_tokens":93529,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":174,"sample_id":"5d965c90d2abd8ba8bfbe670d12f010cbfa12c8f875015a24f5b47e93ab826f5","semantic_task_id":"17da0f305a08090661c8699a5b676267ca4c39735edba9665fb108b0b5f23f9e","split":"train"} +{"domain":"codeforge","full_message_tokens":93538,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":175,"sample_id":"83058f4fe52049765e3af51ab26949d7674bef6b372a5098b10849d2647d11f8","semantic_task_id":"f8a449feb304afd2bfe1a72f04a7b1cc9afb2b4540ef8689980c9ba2c4748d76","split":"train"} +{"domain":"codeforge","full_message_tokens":70938,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":176,"sample_id":"9964674c94534d36c447782971d4088e0343e164f93a5b0c390edac1f6d8f7d1","semantic_task_id":"70c6662c77a839e6e75445e64e8d8ad638e03e9e636c2851b7762448c4b0752f","split":"train"} +{"domain":"codeforge","full_message_tokens":70790,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":177,"sample_id":"c133d6c587421984331d75180087181f36db46eed74e26fa9c3bff18e3534bb7","semantic_task_id":"26ac6060021206f58267631a37513514236bafdcd66bf0042b33bdf0d06bbc6b","split":"train"} +{"domain":"codeforge","full_message_tokens":70800,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":178,"sample_id":"1ff8b1cfc6186ff9768f926d076a5cf719155caca5c34ebe0f86cef7d702448b","semantic_task_id":"aae4f7a6078dec7fce1225894976ec9e2afddd9eab103d29aad4cc5ad9dff6f6","split":"train"} +{"domain":"codeforge","full_message_tokens":70981,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":179,"sample_id":"c59e94214a75f77e9492dc12476ebb6b0fa99490687183811ec65a32e7d227e0","semantic_task_id":"280fc7a607488757d5dfe197cac15ce46c0f10ad8f7b5fcdedb8f00c7db3f504","split":"train"} +{"domain":"codeforge","full_message_tokens":70950,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":180,"sample_id":"2948972b7f29edc22362b77f6d7403a2e87b94d652c300f4697b22855974775f","semantic_task_id":"7d6100d35eee8b50b1e166d7d15ef7d512b4ea2b5f713b91322bab3257abd2a4","split":"train"} +{"domain":"codeforge","full_message_tokens":63786,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":181,"sample_id":"636131ce78d3e2ffff4376cb4f9dd488209c9161227f9d1f30f64abf5a6936f5","semantic_task_id":"8f14a2058c402c28588ece6f9b7b239d1720a6ade4fae56a50a543431ddd27ed","split":"train"} +{"domain":"codeforge","full_message_tokens":63617,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":182,"sample_id":"ba9390d88c9a714cba8d099dded77196beca74bf39d34b0eb053c96eca4df19a","semantic_task_id":"a3535c72c35cd06ef88d4aeb51feff854170bc8bead9f4e8bf7646a87f531215","split":"train"} +{"domain":"codeforge","full_message_tokens":63798,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":183,"sample_id":"ef4f88f1afe7e9de7de860da72dec2b6e57cf54e89b5e5c3165ff2efca30ac65","semantic_task_id":"41442e1a3c7013155e10c3c848028ccacc0efeb96e431bb8083f3360b55b8d26","split":"train"} +{"domain":"codeforge","full_message_tokens":63636,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":184,"sample_id":"b51f3297f2cf335306abf1a8f6e19f2a879273b17fc125d4004d23d359675049","semantic_task_id":"c66157cf6b9791683906205ac2161d9abd45e6630b62a1abf0040c6f635d236e","split":"train"} +{"domain":"codeforge","full_message_tokens":63798,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":185,"sample_id":"d0abedeed491a2c875db5a88df76ec7a9fc6958b62c8c40bdc569da67d4deb0b","semantic_task_id":"b240c76468d1765a5637d8547239f6ff1b65178dd7c5a921ff02be9ddf758f38","split":"train"} +{"domain":"codeforge","full_message_tokens":70831,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":186,"sample_id":"b8ebd42c1f7a327248ca2c7f543db08236a0fc6b83a507b94c54916354566cb0","semantic_task_id":"6e9c18732373493cd31ebac6398019bbf2a808786c0b0d9482937c9cf5e8c077","split":"train"} +{"domain":"codeforge","full_message_tokens":70669,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":187,"sample_id":"40906a06e3f038e3761c2933627c878f59656f9e51b2dfd32ea541dcc010fdd5","semantic_task_id":"cd5a0a7e292776949c31a92f784ac63134ab0753a0ca0f97be368dfdcd8591c5","split":"train"} +{"domain":"codeforge","full_message_tokens":70784,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":188,"sample_id":"670e3445c9aa6a559b8d5dedb33894daf36b611926c48cd2f116296a182f239a","semantic_task_id":"1fc92b0c80ece6193b3ce3ba6542ac669a421ec297de5dd7430ba77e19b47541","split":"train"} +{"domain":"codeforge","full_message_tokens":70688,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":189,"sample_id":"ee126f5ff735364fddd25bd5988668eb58e281cc7da97e742f773ef543e6bbd3","semantic_task_id":"962f626e30f66438570492640ed1b1be0dc74c1b32635aeae5d094d1e1f2b27b","split":"train"} +{"domain":"codeforge","full_message_tokens":70713,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":190,"sample_id":"f2a6378a7c874a0b9186c41d5c344002237bb0a161ce8abe3dd5fe82dc1e31d5","semantic_task_id":"dd7b354dd196ffa2acf7919c927a91da4ede686686172f1c1d1a2f283ab60169","split":"train"} +{"domain":"codeforge","full_message_tokens":123505,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":191,"sample_id":"0a8238c6e878686c7fbe7dc392e14fb32a805c0d2cb549b845c196bdc4a5a262","semantic_task_id":"bd6c96cc89f2cadacc6f49a420d20ea73fe6d9ab3e11d9b339bf3cc8d5eb84f3","split":"train"} +{"domain":"codeforge","full_message_tokens":122552,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":192,"sample_id":"2df6e3233e5b019a0cd6cf17f76cbb272e407300071b92b006d80fa7f190a6f5","semantic_task_id":"a2a3ffcf5c71be8baa629d86a272dbcfdfc0c4cc5b7e57efe1847756d3eb0eae","split":"train"} +{"domain":"codeforge","full_message_tokens":122677,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":193,"sample_id":"a97f6efab594af0d2b68303a7a00e3de07a85b8e186fb62d834a98950226e95a","semantic_task_id":"0de4d4180fe8f17bb863e76f1a6b3dc40e9e5f26912e8d0ba1446160620eb9e9","split":"train"} +{"domain":"codeforge","full_message_tokens":109054,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":194,"sample_id":"1a48b3239c36f24ac5ace4750aa4149ade27ac662eacbe3d296721cd0d399e7c","semantic_task_id":"f8c3690222ec45c7d08a800edbb3f508d206ac6b3ffd5b38d7f15af75a3f0515","split":"train"} +{"domain":"codeforge","full_message_tokens":108687,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":195,"sample_id":"e25eb9e7ec3d5ca380191f83317f3ed48c49e1c2a0cc9b924cd7d59e77ba7cea","semantic_task_id":"cdb936772610be8577dfffb8b53e54b2be122ae4a443744678618f96467ea04a","split":"train"} +{"domain":"codeforge","full_message_tokens":108818,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":196,"sample_id":"9ad2ce0e81fee4db137e46fc039abc4d916cc6bf461f85926a41928637809f8b","semantic_task_id":"f42d01ac6abf74d7fb16aecba95129cfd01f814157f4ae7b2a52cec8e37d7176","split":"train"} +{"domain":"codeforge","full_message_tokens":62460,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":197,"sample_id":"69519a90800ae543598dc4da6ca279db531a3072c7e9c63d6100fa700333541a","semantic_task_id":"d90765bd76a191840a0f8b38ceba1261a6e39ed23d9177b6af6a008e63ec901d","split":"train"} +{"domain":"codeforge","full_message_tokens":62240,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":198,"sample_id":"9ec44cec0040361e079e9bfb4200371f634ae2e7dece822ee98818d7cd3f71a7","semantic_task_id":"20d355576abf8ca00de8bca8912380f57e49c76fa9bdadc0d7ae530077dbfb9d","split":"train"} +{"domain":"codeforge","full_message_tokens":62349,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":199,"sample_id":"7dc68b6ad30f99bccc82dd621d98e162870d93b82c3c7e95371da9312434f8b0","semantic_task_id":"62c36a4ad95ad7d1d6f704296f4de600016fcb238197c9ecca143524c18d38b8","split":"train"} +{"domain":"codeforge","full_message_tokens":60905,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":200,"sample_id":"a8d0d94b510390041639d63ee601eccee335b2a833b41a1d13dd1f4bbad2e45c","semantic_task_id":"07e243d6904f0bf4744b27048e6250fdd048e67cd86baa1b89f0acb06cef5d69","split":"train"} +{"domain":"codeforge","full_message_tokens":60678,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":201,"sample_id":"998e72fae489f597753c4ff520c7409768c81db0f9fd20b11905617ef0f9881f","semantic_task_id":"fadea57c21a3d292e77b9f278f007e0690b30942b66b4f4a7f1ddbe312f08663","split":"train"} +{"domain":"codeforge","full_message_tokens":60793,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":202,"sample_id":"339b1944d8a8c951d312503ee4aa357ff2318d87d5c3ec44406972932d0fa54c","semantic_task_id":"cc72562ee6e5d537163a3168ca77cf9b20f78ba21656de11cda73f5f7fa00399","split":"train"} +{"domain":"codeforge","full_message_tokens":70838,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":203,"sample_id":"b6c1dd197a4c8bac5808e3c38dbb483fc11470eb6b6f4197932728f5ac533987","semantic_task_id":"0a8cad533c77e56d32a05733b6d0767f83717b9343737259d74fd004b455a712","split":"train"} +{"domain":"codeforge","full_message_tokens":70516,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":204,"sample_id":"4b17c8e54747986f574f53635e2f885f7e04b1f7ec3477d00a0d826118f4f192","semantic_task_id":"66fdc43676f445711354802919618e240acee3371f843ff02c00298bb6e04187","split":"train"} +{"domain":"codeforge","full_message_tokens":70630,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":205,"sample_id":"c4617db8bd59503ae605d36f4db25765d0cae04867204505d4f4c156bf435391","semantic_task_id":"9698ab15b5fe3503e406203687c83a1b79abaf71f628263fa66dc2acda8d111a","split":"train"} +{"domain":"codeforge","full_message_tokens":99007,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":206,"sample_id":"45ecb9c543a0f0c523329be5fd24abaeb8f9edb5188fb68697b0eba1e9c74af7","semantic_task_id":"a969ac33bac09cbfc0b447fcbcb1a872580e2be184cf174208022fd5a643a8ca","split":"train"} +{"domain":"codeforge","full_message_tokens":98756,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":207,"sample_id":"f76dc4f2188f5390ce99dee8d103bbea0c1b807bbc6870fcf021c2d23fec0509","semantic_task_id":"50361736ef4c801ba818e8895417a8f15e53e4cb77b6b011e0019d21abbf5ebe","split":"train"} +{"domain":"codeforge","full_message_tokens":98881,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":208,"sample_id":"6d22de1e2f3a102f0b662e8d598fae75d450474044545950e61898dea6de4a4e","semantic_task_id":"bd6fc1766601e45e68ee99b67b27d8018fbd7cd03040e414ed00286257ce0381","split":"train"} +{"domain":"codeforge","full_message_tokens":92571,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":209,"sample_id":"5fcda804f0ed25218102c4c1ac020db23569b849b880ec7b1eb5a23d8c8a8781","semantic_task_id":"dd948db0dc48de158ebd80c21e0881ae262fae7ffdcd48f0a8916682b3a9af48","split":"train"} +{"domain":"codeforge","full_message_tokens":92416,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":210,"sample_id":"a2b54dea423ed1d3cacc72a52b539a664d8fff84f5513a753378b156a0cf0b23","semantic_task_id":"78532de08630d7f764eb51b84ef5f2a4bf1953ac9ebaa554ba9f9431cd41db96","split":"train"} +{"domain":"codeforge","full_message_tokens":92541,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":211,"sample_id":"9cf1206f53bbb5470db62e22c8183103f380c222a7235d6e575d9f8ca4765bc4","semantic_task_id":"b0514646333c900a4f8bd22e547e63ea654eea14b2c1c917a5ee6ad0bbeb5a9c","split":"train"} +{"domain":"codeforge","full_message_tokens":81928,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":212,"sample_id":"b7adf3025b3ad1eee0424a7a5bbad292a3a48d06ee47d0e64b4bc33777f8fc1a","semantic_task_id":"c6ccae531f1df2d6170337593f3dfa7adaaf4145e31deb479e37073f1ed13c84","split":"train"} +{"domain":"codeforge","full_message_tokens":81750,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":213,"sample_id":"6c85a871e36488371e558a25e3659456dc8694b0a16375332b5d11817d87c711","semantic_task_id":"8d4fc32889d5214d4f2a0c8a61a788ed948fe25a2264b6c79ea4b0185ad1554d","split":"train"} +{"domain":"codeforge","full_message_tokens":81881,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":214,"sample_id":"63dbcf9e90fc39fb613dd242b82ecc9105120ef07fcaafe93ca29d31da38212a","semantic_task_id":"4811ed4850f398d850cd002d5188f000d5660441368d536539ba65ae84a5d0cf","split":"train"} +{"domain":"codeforge","full_message_tokens":58750,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":215,"sample_id":"4a45587b62aa82687cc8e98baf0b2aff1b3ddbbf733a12eceb515928ef9c8d5b","semantic_task_id":"b6aeaac295fee1584c123cbc41f518f5650531c3e5afbf14a59b84bced756990","split":"train"} +{"domain":"codeforge","full_message_tokens":58712,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":216,"sample_id":"b0e13d1a60b5fbd2e4511bd4453f064e4a304bdc7e0c435436b4e55c3032288e","semantic_task_id":"f0d65f4b1dc35bbaba3cb3c0d05796b40c093324cd556c6a046c2861a4dd60c2","split":"train"} +{"domain":"codeforge","full_message_tokens":58832,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":217,"sample_id":"ee85961cc810b301e29465dca5d91f55918728029a4abbebd3066a907b74c52c","semantic_task_id":"036a780610a79bbc7401595b697b1078422a9989201a506b3b3006f9082013a2","split":"train"} +{"domain":"codeforge","full_message_tokens":74984,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":218,"sample_id":"0b0cc99c730301e946f9a95838f9fdeb226043b7c6955cc93e6206a41dfa6a9c","semantic_task_id":"38d4aad6473d35ab25c64ad36127775ca1ba74087bb80ced90d24b92f24c2146","split":"train"} +{"domain":"codeforge","full_message_tokens":74908,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":219,"sample_id":"f25e2e57c8428ad67d3a679f4dcebd13d93eb44c5018236e36794b6390639256","semantic_task_id":"92a035fa4fad6583b270b2c9297afa0751a2ad99cf5ae69c49a2a65d083c9d2e","split":"train"} +{"domain":"codeforge","full_message_tokens":75045,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":220,"sample_id":"081df3e97a37bc51170c16d8da871576ee5d964ae1293ea7152a6a984c1c0896","semantic_task_id":"6dd845e2610f00a149fbe4c8831df323935a2eb8aa1233ea6fe665d01f6147d4","split":"train"} +{"domain":"codeforge","full_message_tokens":91512,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":221,"sample_id":"9633f55f0fc4df2c4f26fa796e5a64595c8957e3cfc1623ba7150353a3a31e6d","semantic_task_id":"7f6c55b38dc9d5d9d307287e1ef357d1cc8068eeca68b64db8ca2ef812f7e85f","split":"train"} +{"domain":"codeforge","full_message_tokens":90429,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":222,"sample_id":"cb13bef37e33870f3ce8406d733aab421981c04bb71e0321122390f29ffb5a2f","semantic_task_id":"a15cbe7170a2e7d54bed5e4d41dbaccf89c395603631bcf8e8c5cf4d1772bfa8","split":"train"} +{"domain":"codeforge","full_message_tokens":90560,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":223,"sample_id":"5b50c399e8f029b3a63c016441e6b359ea735c90f371e94b0648c29c2ee4d9f0","semantic_task_id":"6876b7de2fd3e61a51c18a82ac2f8d25ed7ce29ceb57098602f9c1667701b975","split":"train"} +{"domain":"codeforge","full_message_tokens":102209,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":224,"sample_id":"8b616559dc1b4d77b4d364d04059fbcd5d781c69ac41f3271bab3245bfd92116","semantic_task_id":"3cae4401ed8bd248cfaa49dee85a55b7f18079b2a8b0f8710b6ae552cf8e05c1","split":"train"} +{"domain":"codeforge","full_message_tokens":101055,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":225,"sample_id":"461fe263558b63b72bf3838edd5f254ee782e1370ec56e767f385a3f6d300c9b","semantic_task_id":"8540a05fc9f92017188d4751cabdf9b5aea564e815e80b3edf39d33539e3eb2a","split":"train"} +{"domain":"codeforge","full_message_tokens":101186,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":226,"sample_id":"f139e169e90114697e87d65d7d7cff7811a7e5cbeebeaf993dd687d891575724","semantic_task_id":"13c3b953d1f015e292c6f55b9a3627dd77f5462986f8026123fb8ba60a94b3e7","split":"train"} +{"domain":"codeforge","full_message_tokens":94788,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":227,"sample_id":"096e3d8431f7a0365d074bddb434bb4212b1948bb95fceb3d4b9862f4e804757","semantic_task_id":"c96a21a204256b124028798f851be32e5f5eb6c8b018a26ab59b0a7b4b89f03a","split":"train"} +{"domain":"codeforge","full_message_tokens":93540,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":228,"sample_id":"551cd08034bb4adc0e504a582b2ef722e3b82d959c02e41bf3fbefedc6214914","semantic_task_id":"c2152f0513febee35262c0c7ec2d17383035edd0722c40e55e0e97e0cbe7b67e","split":"train"} +{"domain":"codeforge","full_message_tokens":93671,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":229,"sample_id":"709170a7f328a18ea753ae172329b2976f740cc8a121c999130da9128d10632d","semantic_task_id":"545fc66d3fd3d872b919c343213e2c8c65a7a1fb38c095c8d0f690aacc1600c4","split":"train"} +{"domain":"codeforge","full_message_tokens":90555,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":230,"sample_id":"72445a70d0a2853620481b07dee65da01ec90b164c8697a73ab6ef3e7b9e9610","semantic_task_id":"ff3273ea60fb880d8daedc541f0acb714bc6d72c9ae1f4cb67a01eaf7a29bbeb","split":"train"} +{"domain":"codeforge","full_message_tokens":90309,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":231,"sample_id":"36b557a42fe22735484e4395a9706d482148feb1dd43f0016cd8ad65cdf277b4","semantic_task_id":"112c95fa096680e95cc8f92674203452faa53083430eba96188216948dace252","split":"train"} +{"domain":"codeforge","full_message_tokens":90423,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":232,"sample_id":"1f03ffcb323919ea8c79b91d5161a88723342705ddd8c60e79362136e6b8953e","semantic_task_id":"b2f7f7e065e3f6e067643e081aef14077b60a3cff22ab3e1c737498fe271d151","split":"train"} +{"domain":"codeforge","full_message_tokens":90332,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":233,"sample_id":"20996f80196f9ff765d8fc799a082351a9a76f887e67e2968a8baf60d0786cdd","semantic_task_id":"d2d9288fdfd82340cf0c3ba80eedd4ecbef7fce90a5cc99554836b31a2153b5e","split":"train"} +{"domain":"codeforge","full_message_tokens":118807,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":234,"sample_id":"b98ecb5f6bf6718579a39fdb47d5d856dc8618556de15db1e0cf4995509b3c6a","semantic_task_id":"e924a1b7a611601d96d8dad81d46f8102b7cfde23eaef3325c85de1921091b8d","split":"train"} +{"domain":"codeforge","full_message_tokens":118620,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":235,"sample_id":"bdcba02c271de0ad1876b842430258e509cb2fa107cff99637efe4b3af6149ad","semantic_task_id":"e8ae90eb01c0fbe998ce7f329d9df04a93848da2dbe97d18f0b18cf27e55f4d5","split":"train"} +{"domain":"codeforge","full_message_tokens":118734,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":236,"sample_id":"f94cb67d5a9f1c8085a70b25167417f12caf18e8c20f67c4427025b0934c1d93","semantic_task_id":"a1c90c0cda1afdff1695369b4a68d614cff3c3a0923a492eff82dededb29aac7","split":"train"} +{"domain":"codeforge","full_message_tokens":118643,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":237,"sample_id":"2a95a135e61f92f13925ac7443a64f74c091f870d38dcbd58e37cd9c8bbf070c","semantic_task_id":"402b0f3433194dcabb3a6f1cfe6dacc154c552f837a5788114cb5a2bcebc42a9","split":"train"} +{"domain":"codeforge","full_message_tokens":118645,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":238,"sample_id":"47d083ea9f8558d26778949305daed634fcb9368fa8b7fc14ab5dbe8786ce919","semantic_task_id":"3446868393050fdea177490604e615b597e3a2f43ee9c348414ac196df3f5d8e","split":"train"} +{"domain":"codeforge","full_message_tokens":106657,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":239,"sample_id":"e7b8d5281ad903ff73209c629d625007454f6350d27334eda6809270af8c1c3c","semantic_task_id":"607d2b963dbf7920297553e9a604936e2d24ed9e532e154bd4a4f989d101e916","split":"train"} +{"domain":"codeforge","full_message_tokens":106519,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":240,"sample_id":"d597a3fc0b45c3b520d2e0480ce6c292d54a1de017db1e3d6a272c12686ec7c6","semantic_task_id":"2c7a2b4cb2c1286c34504b9486cfa31e5afe01adff1f82e9f188858b2d43e7ad","split":"train"} +{"domain":"codeforge","full_message_tokens":106700,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":241,"sample_id":"fa752c1f1975f456b8217ab60666cc523a23d1184568299ecde513e058d6587d","semantic_task_id":"1efe46540027298637f1718b793fae8df5fc859b7340599345cc74109987a3bf","split":"train"} +{"domain":"codeforge","full_message_tokens":106543,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":242,"sample_id":"61a9955a0692f9998258d17a48a56c2d337daf9b05f167d60783da233dffeb68","semantic_task_id":"d0f154df90dd99247b6850be0c1b8d2311a9021d2db0884ece261048fa1456ae","split":"train"} +{"domain":"codeforge","full_message_tokens":106669,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":243,"sample_id":"eef909133466a7e53eec684c9009ba9671bd3f61388e886286dab26792e6d4d8","semantic_task_id":"0df205a729a18cf625bcc6972fe27a6c95e55243208885751cb4cffc20045067","split":"train"} +{"domain":"codeforge","full_message_tokens":121585,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":244,"sample_id":"91cb66d369392c82cbff26ca02965588e46ea5b34acd3f5f133256ffa64cc273","semantic_task_id":"59736ec3e2d063a3014ff6442a54b1963c45e5c06e38ff967d886890e18cb44d","split":"train"} +{"domain":"codeforge","full_message_tokens":121391,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":245,"sample_id":"eb7b5fbbe19a7d81a8d013377e1c61cef736597dfd8a095bac2c41e05829d860","semantic_task_id":"423edff62472b0a1379b6d40ee90bb09753359e16a39c138b79576ff3cf9e812","split":"train"} +{"domain":"codeforge","full_message_tokens":121572,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":246,"sample_id":"c27b024ca80a216589172f92b147420c38a1b4c142dcca0d6281fe7fbc103216","semantic_task_id":"cbaf5ace29f3c1094c30fd77545c1ba7be8914a32aee3e80655a10596d83342f","split":"train"} +{"domain":"codeforge","full_message_tokens":121417,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":247,"sample_id":"687fd02a885ec093d2ee078d292e88085aab47ce71762db29b51dd72f6f34993","semantic_task_id":"bba4f88edd050064becdd9d639aa3ad724fb5ccfb26a957bd32cdac05af7fd18","split":"train"} +{"domain":"codeforge","full_message_tokens":121597,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":248,"sample_id":"c4aff2e7ea6f3d0a81398eb18ba2944c73d58a6e16ca5f83316f2066f4dd6fa8","semantic_task_id":"99044c87d12016058981f391e4fca9a74e6357ce4e7cc608ff68ebd921af8044","split":"train"} +{"domain":"codeforge","full_message_tokens":83958,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":249,"sample_id":"42081fac161abbf1f6a04538162808639f56936625438c12438f8de2203ef771","semantic_task_id":"276ce9587da232e36a88724b423c19fef426ae19c263bfc3dc6fe9589feba03b","split":"train"} +{"domain":"codeforge","full_message_tokens":83659,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":250,"sample_id":"28178768cbf178d50c3c72ac3afd45e46ada1a8a46691cfd7c39f164baed66d8","semantic_task_id":"3cee15dd9e57cc270a9e897d24872d4a39fded68afaa29f22af233c1212d08c9","split":"train"} +{"domain":"codeforge","full_message_tokens":83774,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":251,"sample_id":"2f291050482572bcf3c2be0585c2a79221dfce93626aaf3420716a60119ff4e0","semantic_task_id":"34183cd6d91961a27be64ad62b12ec1dfb218fef17d74c9d0f912bec04907daf","split":"train"} +{"domain":"codeforge","full_message_tokens":83685,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":252,"sample_id":"f73eeaa5282c9ad5f956a067d1931c51c2304d5f225318653b65f3a73ca883ae","semantic_task_id":"028d93a6cfa612ed9929dae133713d93957d58507eb76f8b8c851e32f7ea46ef","split":"train"} +{"domain":"codeforge","full_message_tokens":83840,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":253,"sample_id":"c3dc36c694f0666e87e3aecb8570257f9d2bfc181c7954ee6fa9dadd5b3e81fe","semantic_task_id":"ea0cd11b17f34bc45493dd81e90a0ce5c01d70ebc0a46d99101771ee092094ab","split":"train"} +{"domain":"codeforge","full_message_tokens":147653,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":254,"sample_id":"1e630181348376a631ded7d6a97a2c2f2c81b1508ac467859ebb16806547182f","semantic_task_id":"c1dc1e258d2c2a9a94069b5fd2c9f17c0916e089b100ab6f80e20a311ae643c7","split":"train"} +{"domain":"codeforge","full_message_tokens":146661,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":255,"sample_id":"9622a0649f8f042fdf39caa0823651b2606cbc52d54849ab9e73074be4258bca","semantic_task_id":"d0525f2e740f9b8b62f9d9a76d1b5b5c2d0b480dc9bd81da12c393d27572a345","split":"train"} +{"domain":"codeforge","full_message_tokens":146786,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":256,"sample_id":"7451600a47bf62665ab318f59c2d6e01bf8c83134ebf7b44808e446a1c7855f2","semantic_task_id":"e7bd062e8287f2f2391dc1019114dfab0c1aaa9280ead7569c3472ee85c4152b","split":"train"} +{"domain":"codeforge","full_message_tokens":122931,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":257,"sample_id":"392e67a624f1b1c0306e835e5a5b7b7fdd6cb66a9b9b3c06b91e95c301349508","semantic_task_id":"33ef2397c70c2de50a4d0ee01c30082c08acb0100bef14c9ac73d89c695ccd1c","split":"train"} +{"domain":"codeforge","full_message_tokens":122489,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":258,"sample_id":"216af038868901fac9cf45c7a03e54b5bd5f395f0f26e48e8d3614ca143931ca","semantic_task_id":"24641802670fd503be496558eee8f2c488d171c9bdd8b17273cde9f6a6a6152b","split":"train"} +{"domain":"codeforge","full_message_tokens":122603,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":259,"sample_id":"3954012976627e7f3af0137f23df422406241e92a686a67f6ad6cd2ee94019b0","semantic_task_id":"c4301a37abf78af3bc6719eb5a5263aae4872e1e11656fe29bf65450db603c82","split":"train"} +{"domain":"codeforge","full_message_tokens":85309,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":260,"sample_id":"652c22478f22b35c2f12eddb6c915a7a5807c00ff2ba7348192449ee240d23b5","semantic_task_id":"ffa00210d679637cafea172591ba16a5211631f8fca0ee07002d2e4e6632c4db","split":"train"} +{"domain":"codeforge","full_message_tokens":84975,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":261,"sample_id":"24c59327fddc64eae4c9dfe3f4d1fcfee1feccaa6f1880d26088bd13d31534b3","semantic_task_id":"c1e859b52bc538325dcaa87bc4be5c113edfa30f38f83dc80b30a26c9aabedee","split":"train"} +{"domain":"codeforge","full_message_tokens":85090,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":262,"sample_id":"0d463c297cdd03c7cd1ef119a9f87cd2f186af8097d551005f27d248a32f84fe","semantic_task_id":"ed41f42f0419a79d934ec02980098dfb8538abdadd0aa6d4746a999c0936d089","split":"train"} +{"domain":"codeforge","full_message_tokens":95041,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":263,"sample_id":"1b2700e10355a429179bb35932209934e9ec9efc4b5cdd7d8c03740c3b58f0cb","semantic_task_id":"8c1f7fbf4a1ba3c131263c32c55dda9b67e8a937588fd7afc5a0d6d96e0315b6","split":"train"} +{"domain":"codeforge","full_message_tokens":94670,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":264,"sample_id":"c0e8cfefb791838285e52d03a18c845f25135cc9c690b2992a727e3a9e59afe4","semantic_task_id":"3934d64a0567fa6f3470a3c373ece16de72f5953474099b2e1d918575f27057a","split":"train"} +{"domain":"codeforge","full_message_tokens":94784,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":265,"sample_id":"d108a95cf113b2144982c4d146807056ee2af7b7e6e0bdc84c5fef1bb270669e","semantic_task_id":"9def8c8d7a3a8d942e95501cb06fba2c75c44feb33af5849d175bb17ae40813b","split":"train"} +{"domain":"codeforge","full_message_tokens":112811,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":266,"sample_id":"68aef9445b2c11a024f13cd667f33917854fc3a402c780252f8c5b11da156025","semantic_task_id":"6af938508417eef0801a4d758747e09b5763d0ce59c1b888ad19e097c11c998a","split":"train"} +{"domain":"codeforge","full_message_tokens":112485,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":267,"sample_id":"adccaed5dd15b9b3c2196f963bd0929ab528921b17d47ab36c7ca6ffa516cd38","semantic_task_id":"d56781fade4088dd004dee2d5c75c2d063cfca9506b5e6f8455c759f8cc4bd38","split":"train"} +{"domain":"codeforge","full_message_tokens":112593,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":268,"sample_id":"e7556511f8c1f79a88e17e2de8d8cd98a43a631f474ce7670699789eebd68d46","semantic_task_id":"006807e90591221f1ecc389d9c40405c14571e4402dd5b23a8b6adef3479d073","split":"train"} +{"domain":"codeforge","full_message_tokens":115481,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":269,"sample_id":"5451259f38d9357945ffd7084acf127e97338d7717cbca13cd03513616af705b","semantic_task_id":"8a980bb62bd617eb3c20fe48c08d27bac136e1359081ee6270f8693c7bc19749","split":"train"} +{"domain":"codeforge","full_message_tokens":115208,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":270,"sample_id":"b9534bbabb8a94cda4850f0fb640df03c166c89ff43a8b581c2daf0d1723b039","semantic_task_id":"b0c4fe7eae0b58631036c12242be137abbe02bf4393a0e1b29666dd1ce9bd640","split":"train"} +{"domain":"codeforge","full_message_tokens":115333,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":271,"sample_id":"94d03e1c2dc3685957817dbbd41d670695094282ed744164114022b8e247c73c","semantic_task_id":"7c51dab3f018b5253886db24ee9383c43fb69490c80939d67c773952ae88b770","split":"train"} +{"domain":"codeforge","full_message_tokens":115904,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":272,"sample_id":"ee2a987872ac36bc7eb469187bc691618380c09d6bbc245db062147ff808e593","semantic_task_id":"12103fadc1349c79394a79101b5c03f34bf1c26ef8e481b797d760b4e4075fc2","split":"train"} +{"domain":"codeforge","full_message_tokens":115723,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":273,"sample_id":"661606c49eb1fe1a51b0aea92b96c4d06fdc07e41f3a488c77c3cb0b6b843175","semantic_task_id":"027ea045c9eb80a679e92a6f6adef881736834b09a250ccafb18e5bf1576b772","split":"train"} +{"domain":"codeforge","full_message_tokens":115848,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":274,"sample_id":"e1408cd338c022285abdee4cfa911785d5d4e5073d2711d55e7503980bf39c76","semantic_task_id":"4c9217b4e1ce94e93886a548fe2b2ea177b76c1720632b7e91b8b7f4bc7bceaa","split":"train"} +{"domain":"codeforge","full_message_tokens":100855,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":275,"sample_id":"7ed8cc1cf1c98c0a632901dfbf21c4ba98616b281e4c65c0f62755f72153c3c4","semantic_task_id":"1c7b66da1cf04bd5d013fc4cfc61bc49992e5979e41e7437bb972ccaa80ec04d","split":"train"} +{"domain":"codeforge","full_message_tokens":100651,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":276,"sample_id":"7e74d958f97904075c224b5caa341213235cb7b9b938a5bdb04f58db36e76574","semantic_task_id":"5159a5c54a94b3a3ac70fd37895a0317ec205ac75a443ec9813b54c4c112d1fa","split":"train"} +{"domain":"codeforge","full_message_tokens":100782,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":277,"sample_id":"f120aea3fcd1aa836c3444b43316e9739f0778266a3173ca3b3aab630624e241","semantic_task_id":"77cedd419522e8fddccb7fa6a594f0b2384572db268cc142b5de0c42e96a27b6","split":"train"} +{"domain":"codeforge","full_message_tokens":91384,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":278,"sample_id":"543398d3a6ceecb161d51a471b5a433cb35bf2d21fd69bcc42477d17d400abe8","semantic_task_id":"fa7da6fea70f1359c1e3a938fdc076d1fad548d0df30fe084ee7fb715c65b16e","split":"train"} +{"domain":"codeforge","full_message_tokens":91286,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":279,"sample_id":"9c7bb67c7e806ebf9daa169bc5934cd1ecb175b6265ac3f3cfd48148b9e94921","semantic_task_id":"72298ef7b073eb0d4f7059ba3f990de30810bc6a3281783d5db6bafafe7f3f8b","split":"train"} +{"domain":"codeforge","full_message_tokens":91418,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":280,"sample_id":"87c07ef9e65b2e8d7389541e16dc6fbb555e82faf293c9860e77662f0803036d","semantic_task_id":"6a8b40af6da22ee54443920fb4dc7943971aec24294dbdcf128cb4c0ee6f7e09","split":"train"} +{"domain":"codeforge","full_message_tokens":114970,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":281,"sample_id":"2583231606510d1d771e0f57959801619a170b90119ae77e462dc81ef39dcce9","semantic_task_id":"a567bae2d922702e0be8a9fa168565b901b8579db54abdfc139fc4b49ecee51a","split":"train"} +{"domain":"codeforge","full_message_tokens":113875,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":282,"sample_id":"ca431ec2c2b7716325887c81ea90f25abdb5edae100eaa94e8b571c0050b7aec","semantic_task_id":"b8d6ae3f7476a9b455e83b18caab716942a6c3349ddb1d0573d516bb1abd57a8","split":"train"} +{"domain":"codeforge","full_message_tokens":114006,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":283,"sample_id":"f3a0e5a978d7628da322697fe908444bf1dd2d4df67481feed2eb01c7dc10c88","semantic_task_id":"2db473ce071d7384b98ba07dc3c743f2ca012969a86374538e47deca5586361b","split":"train"} +{"domain":"codeforge","full_message_tokens":121068,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":284,"sample_id":"fd7a0d63eadf408f4899bb03fa7f35fae0bc6510919c2d3242a891e4c1e68480","semantic_task_id":"9acad07e03489551cb23cae3df7d64bf1d6493b5ac8283ae49f5b6ef5f5c8159","split":"train"} +{"domain":"codeforge","full_message_tokens":119888,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":285,"sample_id":"bf85b9c8f54dc35127772c4d968089f9b83f6593c5ff60797a552b7685d94585","semantic_task_id":"b988a1620c532561b7be4c17ee1c492b2d1fc3e3adb50fe362b2bc474e2daf86","split":"train"} +{"domain":"codeforge","full_message_tokens":120019,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":286,"sample_id":"060578903ff9892c0ee865952f8ad7cf444fe1dd3eefafb6258a96998ed0bb74","semantic_task_id":"1b49ec738bb583bfa5f670fbe0dd603d2608d4e8b87f57c4545e1eb2290fe0f0","split":"train"} +{"domain":"codeforge","full_message_tokens":127472,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":287,"sample_id":"fa601546e4060fe477d6647f4a10b7d1b7c23b6bec7df4a98f506c08682d7ee8","semantic_task_id":"ffd3501753e4eb19d563e000e7824746dd553d58f4b4f480e7edef7b2066564c","split":"train"} +{"domain":"codeforge","full_message_tokens":126162,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":288,"sample_id":"8998de05cf40370a37b6c36f9c035aefd9274e820fcf1c14303d4ff584fde74d","semantic_task_id":"4c71a85c05a129f15c8f6b4685c48eeccc5f00a856b058946a28b66832c7447b","split":"train"} +{"domain":"codeforge","full_message_tokens":126293,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":289,"sample_id":"b20a9366926d2c62791ebf1f232084a70948eb876c75e077b07a1ec7de3b560b","semantic_task_id":"9970a112b23d95bb832e2cd4ea5c566d8faa1e57c9bc9ec7b044ce04456fc42f","split":"train"} +{"domain":"codeforge","full_message_tokens":130541,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":290,"sample_id":"a8cee807c6c2eb700426e192127e66dfa350acb6bf82c6396b54df856933a195","semantic_task_id":"8ccb3719219596ae6e7ec025f7895a63cf07f388046494c571672503eefa7f54","split":"train"} +{"domain":"codeforge","full_message_tokens":129275,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":291,"sample_id":"36f9ddc90195443b9018874aa862430c956a72f4fda5808164d1889ca7139e58","semantic_task_id":"f68b6965f0fa71801d1846979988c0370e2504ec0cf3805278966c922bb5bfb7","split":"train"} +{"domain":"codeforge","full_message_tokens":129389,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":292,"sample_id":"5d76397b75cda63617d5244828cb5ed18096db73858fab0f1ba96cb916406ba5","semantic_task_id":"88f2b233cbc9f4dc6a2f4963c8f781b51f42a56c0a6f822887cb74be2b5b432e","split":"train"} +{"domain":"codeforge","full_message_tokens":129305,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":293,"sample_id":"8bde8d989356dc625bd7e8e6af471410a36b49c8d2f422e4948a94050bf35249","semantic_task_id":"f8100cbca67d6c618557a673a5160a7876e3192204c0eb09a297b775bc7a848a","split":"train"} +{"domain":"codeforge","full_message_tokens":148350,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":294,"sample_id":"fb4ac20c3cc06952531aca64a30aaa5286b97d21497b0e5266d80d6cce04c074","semantic_task_id":"647113ffe0e89312ac30f5f261a888a6f08987f17e19bc6b31bb2c5ab4acf5e6","split":"train"} +{"domain":"codeforge","full_message_tokens":148066,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":295,"sample_id":"77f9db3cc6a8fcea33b74e31c335c073898015ffe0ba99be4c7b7939c1921775","semantic_task_id":"356cafa5d12e842c4861dfb9945e869141fd798608e2af76f705f0ce42718e5d","split":"train"} +{"domain":"codeforge","full_message_tokens":148180,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":296,"sample_id":"bcb419bcc7afdc4da8d58b0c211dffb8142484b3999313f3b64daea536f2c41c","semantic_task_id":"dde222e47f65dd46b3b193669379f014d629df0eb2fef6aac4e90af23c4e2f75","split":"train"} +{"domain":"codeforge","full_message_tokens":148096,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":297,"sample_id":"6ce973ec04a593a80861d1d4980679532e146d9142ee128a45f8ac9c4414dae9","semantic_task_id":"717a9dc850fdbe35abe3b4c520bd6cbc20fd4cee687c1612b6748cf44dd09d3e","split":"train"} +{"domain":"codeforge","full_message_tokens":148091,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":298,"sample_id":"0cf56f6167ae69dfc5831899d5c8e24e00bb2a07ffbf6b2b9940296b67bd5014","semantic_task_id":"9134aeb70ba45ee6e5cb7179006b6d687c79184002d9e8ae58f64da590b8e2e0","split":"train"} +{"domain":"codeforge","full_message_tokens":131920,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":299,"sample_id":"f1c24ca0c2ae111004e130ecc2595db00a9bf8c1eda777232db002b1dd8c8412","semantic_task_id":"28db92d42deadbeb4c4ce3e1494fd9b0c916e85e23c2ee1a1230444029cddc05","split":"train"} +{"domain":"codeforge","full_message_tokens":131627,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":300,"sample_id":"e1aabe67d8ac2d0470d837cb8a06fec3a1a700588e5325e27b9716ca8fbcc395","semantic_task_id":"21f81a5d2a62901709c60858babd8ea21de6c1e5da377e8912eae7f235269a52","split":"train"} +{"domain":"codeforge","full_message_tokens":131741,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":301,"sample_id":"f8123b2186b3e2f192996cef7a28cf1f67badff4583ed4b59efe9011548f5bd5","semantic_task_id":"1fefa7f22555c2ad5603f32eab9f0dbfb180521e9ad346e648fced5661cabf61","split":"train"} +{"domain":"codeforge","full_message_tokens":131657,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":302,"sample_id":"7cad9d0cc9be0ddca305101ce4a67fe56a4000ac5a3d0588262752c895140683","semantic_task_id":"bac81643b693908002bc10e92b3c4f6b8c85d1ded32117168361fafba6d9bc38","split":"train"} +{"domain":"codeforge","full_message_tokens":131776,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":303,"sample_id":"71e01fd1b175a78eb8ba4d5dbc10a265a2c4791fd9062e4270726c56e7976784","semantic_task_id":"ee0c793143b501dd437f011d885b762e6e86897fd283b612d168aaeb964c2ebb","split":"train"} +{"domain":"codeforge","full_message_tokens":157304,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":304,"sample_id":"1582657daef1553b992cfc1bae2be261c545f238a16fe2e27fd63c7f4b8c18ba","semantic_task_id":"24938e7697f66ae689f10367f6577be6a73fc2700f6f2bc699f4a65774061cb0","split":"train"} +{"domain":"codeforge","full_message_tokens":157110,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":305,"sample_id":"faca482bccc0accc4209aa398b671d6f0e2ae6a6abf92b7ee5d30d73b6bb3078","semantic_task_id":"3cd7b309bb3eb5460da8678800b035e84fcb2dfc1b32f3434e515b516d13b871","split":"train"} +{"domain":"codeforge","full_message_tokens":157291,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":306,"sample_id":"8ff15f6c12ba68823d4240307f7d128bf57f13cb11857584417851c2c41ea0a3","semantic_task_id":"d7b58c29fb8a0f705288f3bbbcd56379a09fc109ab6a2e3486976782f12a53d5","split":"train"} +{"domain":"codeforge","full_message_tokens":157147,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":307,"sample_id":"3707f0e0f92e079856b86a04cce4d88b2572b8eff7b05ca05d08015ea9513e1f","semantic_task_id":"f836bd1186af22a560ab6698e545150e1318c84bbbdf3ddd194343af900dcfbb","split":"train"} +{"domain":"codeforge","full_message_tokens":157316,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":308,"sample_id":"30ceee78d25472026c0b1b65bd80a2ec6bb7ffc28cf4f5dee1fc82e7d1fe2709","semantic_task_id":"29e69298133aa68a47084202420be2bf6e184f7ec820424fdd5d513475146d5e","split":"train"} +{"domain":"codeforge","full_message_tokens":141758,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":309,"sample_id":"d0433b0d2dc4b90e73f03b9371ec86cf6f6d9961e493c5a662aac6adb74d8b0d","semantic_task_id":"0f6987756f4118781b95599dccd24e479fc0e22e872edb993b3614b053ce8bff","split":"train"} +{"domain":"codeforge","full_message_tokens":141434,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":310,"sample_id":"0b64e06e817bf643bf072b54e0afd1a81a28a0d3643f37cb95f58bcfb50244ab","semantic_task_id":"bfe82e576678d4139a9414dc2ff2ae18c33cec729800356943aba92bb2a53939","split":"train"} +{"domain":"codeforge","full_message_tokens":141549,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":311,"sample_id":"1a85a3ca95fc3870afebb124b308b42c544053fa964f0f83d4f2ae35b37aa0e8","semantic_task_id":"d45a18ed56d8511bcfd2f9bbe2791f17d23420930ad34a64e5e24d83e2d477b3","split":"train"} +{"domain":"codeforge","full_message_tokens":141467,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":312,"sample_id":"83a3de8f9e118bf2794d53576febaf8261850255078cec07a1055195018ea575","semantic_task_id":"b9effb1243995fffe65e77a3827cb127cd72e0e7899df9bac2c4df06cfacf58d","split":"train"} +{"domain":"codeforge","full_message_tokens":141640,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":313,"sample_id":"9cb2b6d56d2155fc64ade18cb686547716d66c15f968aa844c80f997e1eac82e","semantic_task_id":"c4092b91c4fde3db796904b68ff04ead814e7ab8cd67760bc41702f2bc72972d","split":"train"} +{"domain":"codeforge","full_message_tokens":184318,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":314,"sample_id":"75cdb1d8de42fdc4ad45ffa0c47268f7376dee5817f4daa535b975367e8f171e","semantic_task_id":"7bcb4715df3b84bec8bb2c59dbeeac370d1dcd2693e0b5e8a656efb28a916fea","split":"train"} +{"domain":"codeforge","full_message_tokens":183207,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":315,"sample_id":"023755b8eb92767c5a4420eb44255150f34c074b8dfa716c86b6c0d316c508ef","semantic_task_id":"2ab5e77ef7d3afab11402289c3f7620fd1a7304b75519c0295e287429d17c55f","split":"train"} +{"domain":"codeforge","full_message_tokens":183315,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":316,"sample_id":"c79b34059b93c69c078b612d8876f87a2f78a814107f42f21950c5612dceb88d","semantic_task_id":"fa426c5981b3937cd377a756de9163889bb6465038eb57481befb7de9481b31a","split":"train"} +{"domain":"codeforge","full_message_tokens":179800,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":317,"sample_id":"bb6cf5c9e5f2747d248eb1591eb86c35ac7ab91877b88a61aeae5ecebbeba8ac","semantic_task_id":"97b7d072eee537ff807f83e10c90bc34201c3cf88c9d9bb59e4e31bb6c8cf072","split":"train"} +{"domain":"codeforge","full_message_tokens":179223,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":318,"sample_id":"a5b7825ef1a3dcb3267b7f148cc2f633637fceece994b5c58dbacbb8a3ccb298","semantic_task_id":"c59f20b955cdda860e0be59351e946447992fc3dc8a4d1d245a63b3bcd041bce","split":"train"} +{"domain":"codeforge","full_message_tokens":179337,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":319,"sample_id":"6de862b4628f00ccbfcc88a0acd85254e2ef310eb04835e6488db2a26c1b75c1","semantic_task_id":"8d3c1a395cfd17418f9d34e943f11374922c2fe285c8b4370a4ff6b2a045004e","split":"train"} +{"domain":"codeforge","full_message_tokens":161386,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":320,"sample_id":"ff26248d77c71daef7cf75852ab1b7d6d47d88571c8ada745b9ccca5f6f6a9cc","semantic_task_id":"498e44eccd6759b42156c25fa0da722fc44e86d5719a8f032f0c0bae0538cff0","split":"train"} +{"domain":"codeforge","full_message_tokens":160904,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":321,"sample_id":"0e7fd16c545ec5a93df1d315be82bb46e90bd1d741f14b829d135c59ac807977","semantic_task_id":"0dc7ae0e6ab9ac6b4fe49bcf2b64860857d148f23e367a3e8d09b8fae89d53c1","split":"train"} +{"domain":"codeforge","full_message_tokens":161012,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":322,"sample_id":"0daa97c3a30733d28ce631bc21d22efc574d554e21f8998cfd22d2031758146f","semantic_task_id":"1ef1ce019e1eaaa111d4ffafb12b0581b7d6fc85367bf9345d202b25b1b2d514","split":"train"} +{"domain":"codeforge","full_message_tokens":153453,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":323,"sample_id":"4e4206c677ef9944c0c775e705fb3fec4e8c1a8e25e68fa0f6fb13a386b099db","semantic_task_id":"16aee7107e541fe15aa4b01beffa2498ccf4c2b7cfb001bd0b35a1f01f58de40","split":"train"} +{"domain":"codeforge","full_message_tokens":153056,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":324,"sample_id":"00d91d39342efcd6cdfcc7449ad7530224e65141e0a848b06541fcb04f67ce0a","semantic_task_id":"6703811780ec363b1e1a0a79019100e1595af35cab180ac303dec082ac85cad6","split":"train"} +{"domain":"codeforge","full_message_tokens":153164,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":325,"sample_id":"896b90cb17ddfef4674200bfa449a336a37adf3a622cc0006446a7538ca40089","semantic_task_id":"6e14d7c014f187a88a74ad3987b66d9fa0d6efb46a8aa80c4eead984863a8f59","split":"train"} +{"domain":"codeforge","full_message_tokens":152617,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":326,"sample_id":"629fa2c2d9162ff5f2fd6b54fa02706c051462ad8d58cb9fa131e54c3276906d","semantic_task_id":"bd25183b3dcf7b4c06b3c92e8c46522d281481ad1546ec5a7f89ea8c1ff24c44","split":"train"} +{"domain":"codeforge","full_message_tokens":152243,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":327,"sample_id":"746614fb3f140f18a63c31acfa3fc40d4a3ef7a2114a72b6a17fbf6437ff33c8","semantic_task_id":"d68d7adb36529b9c2a9265fd06805f7820c6688ff15a951df85786302f403376","split":"train"} +{"domain":"codeforge","full_message_tokens":152351,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":328,"sample_id":"24ce64da9f84ed88a31475878024cf8d3dabd3ac96f18345dc39c709a458a97e","semantic_task_id":"a07829ee8f2283a4dcdff776d30ba65597bf667e0c6c2a711ba6699aa51fdb40","split":"train"} +{"domain":"codeforge","full_message_tokens":157746,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":329,"sample_id":"08a4553b463c122496910602032fc3e5929afa39e9db751cbd2a02ac546fd5df","semantic_task_id":"bd062202c8f8164910f0f2f7c08d1943d323fda7fdb983f8baef9ee2d18e46e7","split":"train"} +{"domain":"codeforge","full_message_tokens":157421,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":330,"sample_id":"50f9d8dacf4085a5fa8e893d3ded23acd328b074686398c6e0b91cfa8878eb91","semantic_task_id":"c45a13a87ca896533ff7cafc599f2128cab8edd322b67f47a3537740b1d40d63","split":"train"} +{"domain":"codeforge","full_message_tokens":157546,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":331,"sample_id":"7c83519eb1c428f101c1360cb416fabe973c619307935bb3ba0fe886fa6bb4c4","semantic_task_id":"a1871fe7ec6d023b9ab31c4552cacdda50fb566e53717c1e05a2b8495f440d1a","split":"train"} +{"domain":"codeforge","full_message_tokens":167465,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":332,"sample_id":"043e37e633817ca0f0b0b1879ee0c4a8646d9642d0d573c3f0de2b755cafde96","semantic_task_id":"43312c1bc7b85ca4fd0742c81a7af1f5a7a0ee60ec907cd2db85bdd79f0966f3","split":"train"} +{"domain":"codeforge","full_message_tokens":167210,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":333,"sample_id":"858e0ff4c6bd13731eeb5e9ce294580c023d7244052369e940f51a6cadc41a2f","semantic_task_id":"b39c63fd11f406c9904f9f88e6aab44b9e995fca3cec0806e8e51f695781584a","split":"train"} +{"domain":"codeforge","full_message_tokens":167335,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":334,"sample_id":"bc3e2f60ff134168ecda43a7f999b04972c9db4f3b72fd18aca9492e8c7e7de1","semantic_task_id":"4c51ede88163f0596d65f349b72bd5870a43929a851a17940bb429a5ed0deb84","split":"train"} +{"domain":"codeforge","full_message_tokens":173475,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":335,"sample_id":"76307e764167328428e531cc6345308a8c3bba9d6184cd15915251e48a693f84","semantic_task_id":"c71f7cfd770c9377b989c81c07005a51e3d17f4173f987ab1fe8fc90de1f1e2b","split":"train"} +{"domain":"codeforge","full_message_tokens":172203,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":336,"sample_id":"dc9ac367bfa06b89f6d944455bb21e2581152632fbeba3f68f63281e07d4db1b","semantic_task_id":"850961981a0406b1fc87597d7012385a1cf825a4aeea259145faa7002da0a6d4","split":"train"} +{"domain":"codeforge","full_message_tokens":172334,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":337,"sample_id":"6ebdc7088a61e897b2cf26f5a8447d81a22f8d238ec1cb0f0c6d57cb18d5301e","semantic_task_id":"d6c649ea96b7ee0bb1dd68e16018085af954bd8bf2298a504cadfdd508084e50","split":"train"} +{"domain":"codeforge","full_message_tokens":160926,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":338,"sample_id":"d125a310b825f0c5f6c74898b725fc37c85d891ef45a43d45e32f05f24bd2112","semantic_task_id":"171d530811594268892c845229a30a50402539d6200c2a68bc83353805fa22d0","split":"train"} +{"domain":"codeforge","full_message_tokens":159712,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":339,"sample_id":"07b585e1fb7b6411883c0569b45b3b8db82a05fdcfe9dfe35a8024715bd9e978","semantic_task_id":"bf4207e6bada8e6138367a837459f56297f41fb3f816d9096f9327b6d9308af5","split":"train"} +{"domain":"codeforge","full_message_tokens":159843,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":340,"sample_id":"dffc1e91073b5f7dc5159b029400d2562ea07ee0768a941029e9c77a57a75710","semantic_task_id":"d169ee49bb7701efc91a40464f142150ddb382618aa2fbad15f358d4170fb764","split":"train"} +{"domain":"codeforge","full_message_tokens":169789,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":341,"sample_id":"7cfa9f7e891fec1fee4ea85db5a5d220bdad02fab7c9589db745986042a33727","semantic_task_id":"a1ccbddb02b731fbe9b67447addfbec89aea6b4f9a8a5cf2b3ee16f2267e8be5","split":"train"} +{"domain":"codeforge","full_message_tokens":168441,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":342,"sample_id":"b959d02cc5adcf9cb4fa57ffbf645fc06ec824aa08148933085187841a0855ca","semantic_task_id":"b1f00208b31cdbbae16969fd02d29bfc77b731ce120763ab8abbd008c2b8427e","split":"train"} +{"domain":"codeforge","full_message_tokens":168572,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":343,"sample_id":"2da92baeab51945bfaf8fdbf38b9c07565c5d811f6db4defab3fee43dc09fa48","semantic_task_id":"e06e810276380ad274b7a2d2ee90e884cb321e6b4a2399cb0b80d78c1a419426","split":"train"} +{"domain":"codeforge","full_message_tokens":182051,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":344,"sample_id":"024b4db652964d8ca5129a52f8ab764e29575a091d2d317a58b992bdd5c5174e","semantic_task_id":"90d1a217f4f06483f6c14b6153e09a226db24f4f582308472ede9df1a8b3c148","split":"train"} +{"domain":"codeforge","full_message_tokens":180697,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":345,"sample_id":"6dfbcecbd4c4553b3ff00e43ab78fbc76d7cc8f74ac70f317501211d663d70f5","semantic_task_id":"ab65d8775ce8f06ae9da608d5f7d367f1943ad1b062c245c8a8449586617ecd9","split":"train"} +{"domain":"codeforge","full_message_tokens":180811,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":346,"sample_id":"3f94563e1c5ece03c281ec2f4e36830131c6589336b1771714b0fcb27728dc4b","semantic_task_id":"3c6ccd36e522a183722c990817092ba91d702f5c896d264bf4d45419ebbe7d3c","split":"train"} +{"domain":"codeforge","full_message_tokens":180741,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":347,"sample_id":"f1a51c468602c1cfd546bd0aaa13310706036fa596642d0b79e664623d843f1e","semantic_task_id":"ffb29dcf2259bb91fec666be8b1e2767e7c0db289c7612d7854502ea1a3d1637","split":"train"} +{"domain":"codeforge","full_message_tokens":220901,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":348,"sample_id":"bbaf8ca85d6264165cb97e3f0f13c4d619360e0f45bd60bf47f483d658f288b2","semantic_task_id":"53b6bb360576029923c3925568406bac6e5f8e81b52f620d9cfa2b7b47498038","split":"train"} +{"domain":"codeforge","full_message_tokens":219535,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":349,"sample_id":"81148adb536c392ecfedbf23621f7b39d906cd1eb780a36a4f8c2125af98342e","semantic_task_id":"a6628f22a8576ca82ece345e40646b97828fe3b0ff622ff1dc3f2799de2b9228","split":"train"} +{"domain":"codeforge","full_message_tokens":219649,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":350,"sample_id":"c8b3deca172ef415dd34257c858997f4da5931614f4d252a0b52ac4e7436ddfa","semantic_task_id":"9bbb63f2d702cdcd72106155ce70400ac58827413d0bc55fa8bdb9e1af33a191","split":"train"} +{"domain":"codeforge","full_message_tokens":219579,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":351,"sample_id":"56e055d6d22d69bb2c48e2fe93286cf5f2f9dc86d836c8f085fdd2d87836e2b5","semantic_task_id":"025d5d8090f78a38685c2936dca25fb8127f6568cf3892a10afea47919a1d876","split":"train"} +{"domain":"codeforge","full_message_tokens":219560,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":352,"sample_id":"a1ee8e02e863afe84527ecb705f202c8c8f7de44e151413edf664e659db58392","semantic_task_id":"a302b8d2dc7da42809250c2c13d0d9de38e410d4415f17c9e36de80d0b679e20","split":"train"} +{"domain":"codeforge","full_message_tokens":201449,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":353,"sample_id":"2fdcecb4b10760bd644403c66ae110be56eb5673a26c38f575f00c1a48e67f3c","semantic_task_id":"116b5431c60b7e7f2faf60f1eb43746e760d1de52b6a7c6edb484ef682d7bc08","split":"train"} +{"domain":"codeforge","full_message_tokens":200039,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":354,"sample_id":"db703ff4fa37923e56af1b3f18457a99f4d1404ac28f7f02e7f1feac647220a3","semantic_task_id":"e39d24e6ce3ff1930e3bfee234fe5e9bbdcd90d6c4ef03b4fdf6a83eb57d55ad","split":"train"} +{"domain":"codeforge","full_message_tokens":200153,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":355,"sample_id":"42cc0aa4c274ce21d63d20b7889a73a99cb0075d94e00f861c5f6f8c7fed52e1","semantic_task_id":"a3df5c54cd7307c331165e34518db0a2a18364d5d76ccded8cf7cdb44ef87d33","split":"train"} +{"domain":"codeforge","full_message_tokens":200083,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":356,"sample_id":"6403a5355a3cb0007dd089b032e36eaf73aa57fae733a4835a90ad7e5649af59","semantic_task_id":"7b306543b7e3dac3357ea5ace2af0f6f29811b4500e83663b6d57bc410e3f426","split":"train"} +{"domain":"codeforge","full_message_tokens":200188,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":357,"sample_id":"0feeecf5a48b3e51281a7240d16db00ddc6cb459bdd43a6c07f97aa610972b8d","semantic_task_id":"0186247dcacddfb13ccc4779ce2c9a4e2641acb9149a7cd9edce699cac1bd173","split":"train"} +{"domain":"codeforge","full_message_tokens":212110,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":358,"sample_id":"0adee6777df7bd8fbbdadc567ac30c6138b7f26fad568cf61611a701f14fd4bd","semantic_task_id":"ade53cfb5dd78a87929698e165ffa2127a704e9e7e0f573f62bfac705ecc8f36","split":"train"} +{"domain":"codeforge","full_message_tokens":211664,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":359,"sample_id":"d4d0ae785905a3ff629b8cd1a23570e9359e3239742fb70c9670b3379ee3e841","semantic_task_id":"fbbb95776d77ecf81002acef484fc4789424ca2845d8ccaea75030acc67f1b21","split":"train"} +{"domain":"codeforge","full_message_tokens":211778,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":360,"sample_id":"c4af21f85af9e7d1114cc5da7741230f335926222d877ed7af2b7b11e97798f9","semantic_task_id":"ee42010504b2bcec08c047f7aa7ec61140e73a1c6232b54213665554ece388e8","split":"train"} +{"domain":"codeforge","full_message_tokens":211714,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":361,"sample_id":"2c037c8308e91e59b738f8263a4ac6ff9914f7116b8478a9e6355308ab66b689","semantic_task_id":"1cc814b5cf36bba3cbdbf874d5f17fff01dbe319735ff0931ecb66419c781ab0","split":"train"} +{"domain":"codeforge","full_message_tokens":211869,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":362,"sample_id":"b17674fc2fa63f4b8eb7ef980a55c14256a6025ab90c4f14b30baa471b62cf39","semantic_task_id":"7be52b6ffc1215790063307697b960142b555177d2aaaff0af5be984dbeeccb2","split":"train"} +{"domain":"codeforge","full_message_tokens":202740,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":363,"sample_id":"d8d91b542050a985739c492af07d0e521f2237bfdcb4fae43e338f2c6da0f5d5","semantic_task_id":"a49027da9ffb799b2243b03a3c76556c1bdd42c12b6697d4a75816cf7b768aff","split":"train"} +{"domain":"codeforge","full_message_tokens":202261,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":364,"sample_id":"d157dfb7da185b02c6faeda8f4e0cc3005ed910ea732a3590246e3e32d0a421d","semantic_task_id":"ba4d2e2492887267d27b6160973fb29cafca3f9de8013e6719b567ab90c74583","split":"train"} +{"domain":"codeforge","full_message_tokens":202375,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":365,"sample_id":"05445f9f87122ea836ab32a8fdad9713adf0a460227c83dca44debbcff066674","semantic_task_id":"3e9a20e682bf37f3dabe7d4a96d0890a46e086ac3da46fbc1ed0b276e49fb1cd","split":"train"} +{"domain":"codeforge","full_message_tokens":202311,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":366,"sample_id":"5911414bba2457e4aa808f7611d19925b8bd0e7e3c49ce9ce2fc20ad51c61101","semantic_task_id":"e3696c91afab6511f6332fa4e540e59e695f88b500d78b668b9eaa7a3ea589bb","split":"train"} +{"domain":"codeforge","full_message_tokens":202466,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":367,"sample_id":"9952e9edc254b351c764ece337a92ad9c4d45126d374de9762968f7963839b80","semantic_task_id":"8b6dc8cb8e6b5e0fdc4ebd8d45ebe6c1394255a441906830036f00199681dfaf","split":"train"} +{"domain":"codeforge","full_message_tokens":260277,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":368,"sample_id":"934bf4f068b1ed9c315968b59b0408ee2b18db5847e708787eb8d54975804317","semantic_task_id":"c004bde503f37bfdfc26baee58608eeeeb34851d09e99c26a2244a32d5de90f9","split":"train"} +{"domain":"codeforge","full_message_tokens":259129,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":369,"sample_id":"7973826ab26b8e90a39add0da58b44181336a60660f14cb7eae0ba5c8894c3a1","semantic_task_id":"ef6a4538027c0f5d79dbaa3da420af8183f9dec8dbdd420499b409d52a9493a6","split":"train"} +{"domain":"codeforge","full_message_tokens":259237,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":370,"sample_id":"3d0c3f0f73173a85ae85326c520732caf247da9fcbf7611924b45799ebe21e22","semantic_task_id":"2db3cc5be6b02d506c820ce8cc4289e2aa055202840cfc386a078af646c3d13a","split":"train"} +{"domain":"codeforge","full_message_tokens":238213,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":371,"sample_id":"06af833b0ff5b802320c9ea01ba7779d64fcec4c6055681839efc6f90a0af1e8","semantic_task_id":"8dc63c2e1b9748739995694fa53c341b02518e697dd48929980907bbd362d969","split":"train"} +{"domain":"codeforge","full_message_tokens":237604,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":372,"sample_id":"51c4e851c4bca9ce14a06275c7e40968acce3c8a6065797bd2b0f45fbc1196f3","semantic_task_id":"195d2a1af3cd148430944d91845363e48fb72613a4ab2f11e47ff58f99c3352c","split":"train"} +{"domain":"codeforge","full_message_tokens":237718,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":373,"sample_id":"8ebee01855e9e26a66c8f731284f4234c04fea9247a001b8eb3b8f35d1d9132e","semantic_task_id":"93fcded429ed3112273045525b3808d884d11dac3365b8f41a4433efd24ebce3","split":"train"} +{"domain":"codeforge","full_message_tokens":201270,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":374,"sample_id":"7d6c9f64941e746be0d1f51a97b440598f11787920e48470e8a06839ea5c1acf","semantic_task_id":"fec4f3dbc129fd05439870d3376ba5761fd91aba6c3184809d63ba90b5218abc","split":"train"} +{"domain":"codeforge","full_message_tokens":200740,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":375,"sample_id":"99405e1f9a1baf8cd68562be32d5d6135b81615090626015cccc7c5ad6b98af2","semantic_task_id":"877ce633fb65dc1392d258b37da9dc61506fc31025d9ada504d8c07ddadb9373","split":"train"} +{"domain":"codeforge","full_message_tokens":200848,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":376,"sample_id":"6a8668004acaeeb5f8e8ac80a5c86cae4b06396a8f4e9de5278f758673442373","semantic_task_id":"ba230c28127e7cda793fae7192561ae0f28a74f07d1ee61bd169dc7b10a2bc1b","split":"train"} +{"domain":"codeforge","full_message_tokens":195769,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":377,"sample_id":"f6db08eb227aace390ae0165dfde045058583a2f7baf1a7bc51e5162972e9c88","semantic_task_id":"7f464cb011c7394a8771895787a0f906ad7a15c66e0be24d5b83d6d555cbbe66","split":"train"} +{"domain":"codeforge","full_message_tokens":195320,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":378,"sample_id":"219d2a49c14ea0cd82f2f4b1698a4d330e9d7091314f1c56a105fc004f2b3b7b","semantic_task_id":"1ce7fb97f64fe37c7b55e75021329a6c03417dc17440de9b01bfcbee25709bf8","split":"train"} +{"domain":"codeforge","full_message_tokens":195428,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":379,"sample_id":"d0038f870a1519853bd15aad6871334698f5a82c8ce594b9543611551717a943","semantic_task_id":"5e8380e789494ce3988e18abd0e2885d75b9d458661786bbdf0d8b80914f1178","split":"train"} +{"domain":"codeforge","full_message_tokens":204178,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":380,"sample_id":"ffa9822dbe13be8d177f55c44b3fac8931ca19fd0683f62c4f63310bb1c4e75f","semantic_task_id":"48d95df8aee28c9531809cce3f758b8bd50a23be475c0d139720fb39e78ff927","split":"train"} +{"domain":"codeforge","full_message_tokens":203730,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":381,"sample_id":"f59e6694d38dd2f02bfd39931a316694888659ae03b9eab4034927fbefb807a4","semantic_task_id":"e2d7ed5f054845b4b8f662bb2c675fdbcad54bc668f2f95ca9bc853fe5d2a1e8","split":"train"} +{"domain":"codeforge","full_message_tokens":203838,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":382,"sample_id":"d1ff0832c27429a1fd54fbec1dbd7c353ffde04ef7e803b8c93a86a4723d50b8","semantic_task_id":"ce1f93e96dadcbecf5abaaad150a5783059eb871263ecbcffc7e073ae4a2a78e","split":"train"} +{"domain":"codeforge","full_message_tokens":230366,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":383,"sample_id":"65e4856a1050b01f8908801e3bde4ae8a61ff1b4a8501b371901b515aa49b573","semantic_task_id":"1b187e211489e43896a7907845e324bd5c1793ef8f03b60305e83b6f625b2688","split":"train"} +{"domain":"codeforge","full_message_tokens":228967,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":384,"sample_id":"391627e1cf44107326c4150ccb12cb93ed09b5e6720f84d3e04089e1168352a4","semantic_task_id":"d0703003ab7cfdb56e3324e14fa832df8edf72d17c58e28b318afd98133b345a","split":"train"} +{"domain":"codeforge","full_message_tokens":229098,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":385,"sample_id":"fc1b79f3c81d92b31ec097bbd9a498533de630cd30725de0a0df129da0d36375","semantic_task_id":"6a01035510b8115ceb688233a8a9af6b1df50aee4df07984b85deb1c117c864d","split":"train"} +{"domain":"codeforge","full_message_tokens":237007,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":386,"sample_id":"b6fc5a49aa9a08d9060ef0b6dffe7ff6fe71fcbf9df76175134befed6e658892","semantic_task_id":"8132d7fb96f5898faff4b49e4a248b45868e369413dfb3a7b6ff3f89da56e1f6","split":"train"} +{"domain":"codeforge","full_message_tokens":235629,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":387,"sample_id":"3de5ce19eb508005e95e8bc478fe994906ef4c00bbdec111cd8e5e3aaac3422b","semantic_task_id":"5272bf8b63ac30441905a3d295b34f7d6bfe64eeb80f4b27d2589c4940343044","split":"train"} +{"domain":"codeforge","full_message_tokens":235760,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":388,"sample_id":"e1cb9327dc277b31ce5399ae2e37dc72cd05ac57d4b463524c9e460044eb2c2a","semantic_task_id":"0748619ab0ac731b9f4be6e1a28358fac3d4dd3732dc3edc62d2ce73811cb574","split":"train"} +{"domain":"codeforge","full_message_tokens":228294,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":389,"sample_id":"507f4c0edbc149f81d2209526c4292669eb43a5c297214ee4a072fce483350ea","semantic_task_id":"d4094116d201bacfc354608b43cee20af4d39361db7f6bf250bf49b86c38c66f","split":"train"} +{"domain":"codeforge","full_message_tokens":226769,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":390,"sample_id":"95d1c4603a28fcf4b8b17aa438be4b115b2bdbce557909f5d1aa141eaf0c4681","semantic_task_id":"ac4285408426fd9774b3a26c29461bbd33154aef5719d5f172e98f3fdf98cd04","split":"train"} +{"domain":"codeforge","full_message_tokens":226900,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":391,"sample_id":"d760009c874a2ab54fe8a0f61e9aaffe7a3df7c335c653de04234eda909df3fb","semantic_task_id":"db02fdf27ff5955206048f9c1f0fc9ed2fb597f0bc50e1cc3878cff617572054","split":"train"} +{"domain":"codeforge","full_message_tokens":221929,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":392,"sample_id":"18a8630d4cb4d5729760b4ddd80d1324731ba54acd4c8aa62829c500e693c882","semantic_task_id":"c2d4021ae5ec42509077028bb7fddca4be55eebfa2b01da87f889e21f3600bd4","split":"train"} +{"domain":"codeforge","full_message_tokens":220541,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":393,"sample_id":"69c6f1e0ab18f7ed1649f376a8143b2e4a304e3a2cd4ac5538e77141d01d3d32","semantic_task_id":"0770ab50a17b3fe77decc7a210bc4202e575cc55e5f1a50e5aeaabb2858b2f99","split":"train"} +{"domain":"codeforge","full_message_tokens":220655,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":394,"sample_id":"f9be8577b9dc2cf0997f9e8ec2670b48ead7a02ab1341a47a441261818b0c3c4","semantic_task_id":"6b366cf78dc66f1593dff00d5f5811d272b3adc8f5d73921f7387913487bf5ef","split":"train"} +{"domain":"codeforge","full_message_tokens":220599,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":395,"sample_id":"71d0a6250bee8162b0efc210b84a36a9d6ac326a552bb55f500b7d3e42cad00e","semantic_task_id":"3b1056d6ab4c568c5fe9442c4654bb4ab669544f93907aed7a2627abb0ee83df","split":"train"} +{"domain":"codeforge","full_message_tokens":261840,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":396,"sample_id":"1a8cdbd34fe4ea0dff86d8917a702e011d8802d43ec178766dbf75bf2ad8da9f","semantic_task_id":"91bc5b52254808e05170a6730eb51042a567de1e74dcec18306bb3ce2638cf0f","split":"train"} +{"domain":"codeforge","full_message_tokens":261954,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":397,"sample_id":"e39e02d094fd66c6c2a9934f689a19ee0be03bed7f3a91c8e7f35226fd839d79","semantic_task_id":"d203ac88105497bc2e70f89ad991417d14645f7d5ab72b70429722868168570d","split":"train"} +{"domain":"codeforge","full_message_tokens":261898,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":398,"sample_id":"7f94997503bfd800786e9fe9d09664e5c43c28045f9ead4c0254fbadeabc3ce4","semantic_task_id":"022e273f6de599ed49974e55b3fb02a71fba82df4677bfd0caaa09f3a43890e2","split":"train"} +{"domain":"codeforge","full_message_tokens":261865,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":399,"sample_id":"fb8c5f0e3b559c28ed5bd060916a08e9a768c0e051492730e99098e2ed68b60e","semantic_task_id":"00043cf2c5632a04f4271f46c4a99bf09f4afbbf98015ad2aec2d2eac5fd86ac","split":"train"} +{"domain":"codeforge","full_message_tokens":252959,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":400,"sample_id":"15c2ed402d80bf026b025b90d45185c8dab753c79d5f3adef68b342ff76d1945","semantic_task_id":"1eefdda518136646c84da682862b1c8f5cefa518369f6db356fa3316a281abf2","split":"train"} +{"domain":"codeforge","full_message_tokens":251461,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":401,"sample_id":"f8cf61bbdcfac309296a312a271bff50b6cd834f65067a2d5a91c5f16b587e59","semantic_task_id":"b7e4101e37dd37867c34362f8053e56b369fb27a02e2c7a88ae0ce3f7cd1182a","split":"train"} +{"domain":"codeforge","full_message_tokens":251575,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":402,"sample_id":"16c4a046545feca1ab32291462726592c7196af418171fac40a581b3dd7e7959","semantic_task_id":"96914921b4fa9fcd9d2f942976a6971707739d8cfa9b8cdb8cbe22ddbe7ec487","split":"train"} +{"domain":"codeforge","full_message_tokens":251519,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":403,"sample_id":"a7d8be2d3aa81803f085ca8c343c8501087e23341b72c133e8fc1543ed155a29","semantic_task_id":"81d53c525a28e9bf6998db73e2d60f54e848fb7ce4bc42b5768217eb22757183","split":"train"} +{"domain":"codeforge","full_message_tokens":251610,"group_id":"https://github.com/oxc-project/oxc","output_file":"codeforge_train.jsonl","row_index":404,"sample_id":"c5f001c248f7490653917f937fa969441918d914f52c437baab08b23052f704e","semantic_task_id":"3820e0b664a43d7d27f6c2fc769589fd40d3afa2fb01c0b13a5116091fd81c19","split":"train"} +{"domain":"codeforge","full_message_tokens":75590,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":405,"sample_id":"9b7d8530e80bbdeccdfbf100a6c910d5ec355639f422322debbc87432be7b083","semantic_task_id":"0e910dcbaa41d196d03a51e4bea705a09562682b35a8e69fd970a48cae02ced2","split":"train"} +{"domain":"codeforge","full_message_tokens":75598,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":406,"sample_id":"535d4c32b9fed7ce1d7d38d26fb843dc1b12e7b00bc6fa5c427dab06b931854b","semantic_task_id":"14e794057193f546041825e3994aefda2f9f1d0bacffbf1f59e71245ba4b4ad8","split":"train"} +{"domain":"codeforge","full_message_tokens":75703,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":407,"sample_id":"4907f9c2b14f61d6a0ad3983a2d79fd78761d6d3597a5168ce9d33092ef80a0b","semantic_task_id":"08d628a32858d66e8c4a7c57d01fa9d490548649688ec49ff81482fb3afefa9d","split":"train"} +{"domain":"codeforge","full_message_tokens":75642,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":408,"sample_id":"accc359b6aae998c1e5028024c6c5c47119b0d918b73d6301556ed0139cc6ede","semantic_task_id":"426a92f197ad6cf1670e1d7784248f083615552606549b52b07484da1d45b75b","split":"train"} +{"domain":"codeforge","full_message_tokens":94280,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":409,"sample_id":"6dd96481bfa818cc4a6de9e1dc01185fb1d8a22de0cb026cdd410614314ed802","semantic_task_id":"699acb212162b2db97df545b528a2ac49f55961a97fc80f4813daae000f45903","split":"train"} +{"domain":"codeforge","full_message_tokens":94197,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":410,"sample_id":"dbb72822ac16787e47ab365201f3b280f9065ada7a575c35f3ba3f2dd911a647","semantic_task_id":"ab0986af8bdbc5db72c17239178ea429c0cf01805a0ce19aab598d31a9a752b9","split":"train"} +{"domain":"codeforge","full_message_tokens":94341,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":411,"sample_id":"18c5e9b0ffd2fdebca5abc8b628a1c67b7455c0916b1dab1d13e053d8cd5dc8a","semantic_task_id":"44470221ca0be6d72c591ada837b77b18a8c6779ce6c1e24c5ac5e4a46905a26","split":"train"} +{"domain":"codeforge","full_message_tokens":94221,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":412,"sample_id":"0ece3d8015cc2849ba4d5c92a019cd7044c0a0bf47eb7d0116d04c15edfd9169","semantic_task_id":"f38085134d2d3ae90f4fb4356e7a5c3b1f50ca89dd510ddab0d315491f2f9c2c","split":"train"} +{"domain":"codeforge","full_message_tokens":116159,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":413,"sample_id":"211f66e18297ce7279e35653c02b8aaaff5864d6ac0ad5afdd0cd20529018ac4","semantic_task_id":"3551f5b2c3913aae08f58c39124f0c3b8b0566c2a2d76a3286bd926ac7e01b75","split":"train"} +{"domain":"codeforge","full_message_tokens":116084,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":414,"sample_id":"c59c75978f49a1db4ae7dbd6d365136a1fc06ab1b472f43bf59442cab9f7ea08","semantic_task_id":"1e863cc2992b546369a74f2d7522e05f30040ecdc6860edff635d39632dbfad3","split":"train"} +{"domain":"codeforge","full_message_tokens":116228,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":415,"sample_id":"67d12697779201eb8d930bee6601ad9227092f8c55ea60e52e554fc6e108e504","semantic_task_id":"b63e31b435492a7da5bcc114cee7c8b112e72c23327e880af7ef47c0f24144bb","split":"train"} +{"domain":"codeforge","full_message_tokens":96248,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":416,"sample_id":"47d5b48f44c8b5f8a937cd8ed70bd2f200e4d3cd15be2d49a234d50c115e2bf3","semantic_task_id":"2edf11e95fa698d0cbf33b3271d19c585c60cefda3bcaee7ea4e54f0e4a63077","split":"train"} +{"domain":"codeforge","full_message_tokens":96250,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":417,"sample_id":"f609678b0fb20e720b193a2d851728fc140c79eaf30c423e5b6ca0389f309734","semantic_task_id":"7fe71807b8cfc714b4cb0a3c587605a1b910e575c9843da9aa3d181a1bfc6cfa","split":"train"} +{"domain":"codeforge","full_message_tokens":96433,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":418,"sample_id":"a681151fc8e55d5d3ee9a91f4c2da2fa8a9a2a5b5eeb8978914f237a76932840","semantic_task_id":"e238ad619f456306c26351a7629be63f9b9142cc304d1f7f14bb541f9c469bf9","split":"train"} +{"domain":"codeforge","full_message_tokens":96274,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":419,"sample_id":"ca2470f54db7eb75244f351300525f5defe7866b3e8b3ca550ee6bc78a9f9a33","semantic_task_id":"7b6df394c138ef78d11efebc96022670b2cc6c8ae0ceb9575eb6ac65702bec2b","split":"train"} +{"domain":"codeforge","full_message_tokens":142581,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":420,"sample_id":"9ebe467c3162315a458471bf4f0874910ace366afcef310b751914862766a3f7","semantic_task_id":"3d6a413ad6d00217cd36737c67ed721b503f428ec9fcbbd9512818192abef125","split":"train"} +{"domain":"codeforge","full_message_tokens":142481,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":421,"sample_id":"7213b04159f036013c77dccdece11c572411ab1b80f879a54f11dc9084682d42","semantic_task_id":"71be2bcd3a7f94fb43f4f78e6a57cff74e05cb2e739ae97b47cee17c9f54a8b5","split":"train"} +{"domain":"codeforge","full_message_tokens":142663,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":422,"sample_id":"48c7e254334b39a0243389cbf3efa3ac4d1edc88470789a6440a57d04b666c90","semantic_task_id":"3380462c2b95900309411a709b06179dc8999068940b9b833f29fbb4b6a38468","split":"train"} +{"domain":"codeforge","full_message_tokens":142504,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":423,"sample_id":"3373c21d02dec4b2e64c223d001a8800e2c1cefb263c44121f16baad2c7a9e9d","semantic_task_id":"963d0336350e8e8ee332fbf71beaed683ed2c1f76e1d6f3b4cb4c7cd23dfc341","split":"train"} +{"domain":"codeforge","full_message_tokens":146572,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":424,"sample_id":"7c33fc1e124fad4c5e2dc12216351b27299df9687d233a4d1e2a6135c72ca3e2","semantic_task_id":"4a3060ed1f7c256766da9a5d87771d8107bc07c11a82fa82afb896705af3a7cf","split":"train"} +{"domain":"codeforge","full_message_tokens":146482,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":425,"sample_id":"32d3e7681deaa038d09cd28d57f8c6a3a5f0b1671b9392657281c439329873e3","semantic_task_id":"9705de33365e2fcdb9d3449dd31448bb9c6a804e173ce6996c50dddb7b6e8778","split":"train"} +{"domain":"codeforge","full_message_tokens":146592,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":426,"sample_id":"57639fbe507222affaf62f8af31313f69fa5adea4c287ca0a02493501bd92fe5","semantic_task_id":"bc971065c698e99622c7dabe151dbc2ed39cbb1f7e71b08a3e6b96a1c54bc634","split":"train"} +{"domain":"codeforge","full_message_tokens":146537,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":427,"sample_id":"dbaea4a880447ca3379114260880d8f78800f7b1cbfbe25ba2e23c1e6f5df0f9","semantic_task_id":"143aaefe3b254d6a745d87269f2e90d3ddbe72f3cbafc4cc907b24357540554f","split":"train"} +{"domain":"codeforge","full_message_tokens":139336,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":428,"sample_id":"c40f8ec43456ddc9f081e091819cdc4250069acae4951b6aed954772a72ded85","semantic_task_id":"21632ca3f8afd573b3e14c689eea3b45bc2cf1986b84b1292652a20dac57eb95","split":"train"} +{"domain":"codeforge","full_message_tokens":139248,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":429,"sample_id":"794768480eff06bdcb810e7c861a5c02983d1142ce71ead75222421f02c4103c","semantic_task_id":"ce56d4c4f4fab370d293c82cc56b668ef184073d042d5212032aa645989174e7","split":"train"} +{"domain":"codeforge","full_message_tokens":139392,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":430,"sample_id":"e80a8ae76439cba833e26d902f66414fb9aaf18c70c4f03aa896e02dd25d9c7d","semantic_task_id":"e3395a6a1ea2419317f1cd0f732a4623974afc176e1fe06ea95ae64628ab3dec","split":"train"} +{"domain":"codeforge","full_message_tokens":139278,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":431,"sample_id":"30e4d352d90eb8a7bc08133d75aad819b1df6268e3d256b6e73a86a3711cf267","semantic_task_id":"68220df058a66cecc60f0cba67a7d7df37085dc8dbd034379cd673bf771ea320","split":"train"} +{"domain":"codeforge","full_message_tokens":167232,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":432,"sample_id":"174af02119dedb941b52ddd5b5cba207ddcdff75b5c4d1424f1bf4ed45a02c62","semantic_task_id":"a55c1d8b64396a2e312e52cdbdd33fed691138c4baaef0041fa740f5b0ef5923","split":"train"} +{"domain":"codeforge","full_message_tokens":167141,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":433,"sample_id":"1296e282bc149aaa2d09df6ab31645c7b64cce4f6f4d893d93db5f602d05ce50","semantic_task_id":"0af10b58553494f487aa8153aa6e2c7f4db0b7590ce3fe1808999419bdcefbb6","split":"train"} +{"domain":"codeforge","full_message_tokens":167285,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":434,"sample_id":"012a6f05ad5ab41e8f724c42f87702a6ef724bce90257412098b80c178bd9444","semantic_task_id":"2202fdd0b6f8050c01086cd267f8c8015da419ff3e3b96cc14f99585fd9db646","split":"train"} +{"domain":"codeforge","full_message_tokens":167171,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":435,"sample_id":"1410419fcf3cdae65a57b755520ca2bdce01ec8344380ea7bb0c9198450410d6","semantic_task_id":"575096cb64707403feaec5a0b543aa470eac98bd2dd8f3a7777ab4618d5647b0","split":"train"} +{"domain":"codeforge","full_message_tokens":187731,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":436,"sample_id":"5dd718e201067c087ad71f994a3c42d41b76442eff488148506a446f13c6ffc8","semantic_task_id":"9a53eaffc310cbf65b859b382f47eec05a7e2071eb3fe4f3f5cab7c5e781dd8c","split":"train"} +{"domain":"codeforge","full_message_tokens":187626,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":437,"sample_id":"ed220aa04201de64020c77bc129a1ae058843bd8103d3af70a5615662bacf8c4","semantic_task_id":"5b970510e0f08d164dcf6021da36c630ec4b9a3e17c8864ef4606cd45e5a8a5b","split":"train"} +{"domain":"codeforge","full_message_tokens":187808,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":438,"sample_id":"a7aa36a09e2d41def3653e1924ffeef8ade7dd6523daef70080839b434a38c1a","semantic_task_id":"09123742241eb250de8608ca248e6612c8d76891e0a2c33c743172ce31e293e3","split":"train"} +{"domain":"codeforge","full_message_tokens":187655,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":439,"sample_id":"cf2c283f53cd4872d5505cd899c383300be09ccb4ace44e6119c622e01f67b47","semantic_task_id":"34d5560fbe934ef38a8ba7e147650a9b12b37698ff31d9c03572973108e58cc5","split":"train"} +{"domain":"codeforge","full_message_tokens":191713,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":440,"sample_id":"9ea8674bae2d0ffa52769ec982eb6a538e0b8d97f8652eef0041f899b1a4b15b","semantic_task_id":"1a125df327d8600713e29b8e2742f57e407c58879c0798d4db675080f528ed4a","split":"train"} +{"domain":"codeforge","full_message_tokens":191618,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":441,"sample_id":"e3f63fc45684f9a4a38f6473ae3b82afec571e0162094b156add42b618d08d76","semantic_task_id":"26393ce17e05aadfea8553471991c5bd4d9b7b579d941820322d42a8afd74874","split":"train"} +{"domain":"codeforge","full_message_tokens":191728,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":442,"sample_id":"e4d92b85c4652a2ab938bafd2c643360db7149c845fcbd676c33536586181fab","semantic_task_id":"e9de800547d9f12f76af5141236b4e88ef73625853da7e6afa2ab8de1eafaba1","split":"train"} +{"domain":"codeforge","full_message_tokens":191679,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":443,"sample_id":"9bf5579fb0d6dcb228aa01d03dd27b66757ef9316cd37a4ec16614f19df6145a","semantic_task_id":"45c0d6958534ab5450187d11c30b552ab56246fab14ff54b721d2fbbbb004880","split":"train"} +{"domain":"codeforge","full_message_tokens":190409,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":444,"sample_id":"f950875e412581aef787b85a361070bc83c8db3ca8695907640c3b716179e33e","semantic_task_id":"93700fddb6827b8d35b26c72d174c7648cc433a511c5d9506ad9c03e529db865","split":"train"} +{"domain":"codeforge","full_message_tokens":190305,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":445,"sample_id":"23759b0d4a4943ad6f6392a6ced06996c52731b1eb0c524e3fe7c64ed09c3ca1","semantic_task_id":"dcd0353ce0dd2976f6ebfd6f8e9ae9028fde18ace67e804aec479d98bf662086","split":"train"} +{"domain":"codeforge","full_message_tokens":190449,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":446,"sample_id":"9bc47a86417e92f6de81cc9250de01937b670914e9c77c128aca14bf06321ddd","semantic_task_id":"08a96f422d8e2001884a86d2e830ebb7ee5054e4b1780e3373982434150e9030","split":"train"} +{"domain":"codeforge","full_message_tokens":190353,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":447,"sample_id":"bba4afab528044177c6c87760302d0931cb1cb34e129dcfa00a43e8e8104b697","semantic_task_id":"4d8aa105aa78861b23eeaa5b5ff21bb4dccfce1b18d9b6a5dcfb81d9fffb1526","split":"train"} +{"domain":"codeforge","full_message_tokens":258694,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":448,"sample_id":"2eeae9b1a375752b15e72637e074b2e50bed55955744d8f54a1b1096d04a59ca","semantic_task_id":"c654e1c75084aa425cfa4de27c3a03f8f05b39d14a094b8324136ac12d10eb33","split":"train"} +{"domain":"codeforge","full_message_tokens":258517,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":449,"sample_id":"03f9e84a5d2d189d2439253801497722c2be5f1c28a6a60f06a3019da726bfc7","semantic_task_id":"7edcb14710a99c06a883b492410696acf1b490acb135b296e11c2ec9995073a9","split":"train"} +{"domain":"codeforge","full_message_tokens":258660,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":450,"sample_id":"5e392df960d4bca14d14905373b072431dd48d108264c10a9c89803bdb2a17b7","semantic_task_id":"8ff32b5ba0d47a42291ed94876a395ec0837a8aad01125bdb469c6f98b6ba85f","split":"train"} +{"domain":"codeforge","full_message_tokens":258552,"group_id":"https://github.com/dprint/dprint","output_file":"codeforge_train.jsonl","row_index":451,"sample_id":"df1caab3a54270723e028754ce691c58716e4bb613756e08b7c34ec6cf030f5d","semantic_task_id":"c25fd635997f9f35df7939af2ed00ca74644cd036780280108aba91c3c2ef69b","split":"train"} +{"domain":"codeforge","full_message_tokens":117363,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":452,"sample_id":"6a9d4d26bae0bfa712d36a727eadbd3e17f60b5b45d219929e47c40a63641420","semantic_task_id":"755a67cd4e30e5d634179cf06eec405b38001e5ad3bf5f9aad86baab4b7a2976","split":"train"} +{"domain":"codeforge","full_message_tokens":117330,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":453,"sample_id":"a899e6a188c10027733a6cbbc5087dc81fe1731316cc0e49e1e7371f9598ed2c","semantic_task_id":"dbad280a4d5ca6e50b9765da5596e293c26f714d741b406b92c5f08929714e72","split":"train"} +{"domain":"codeforge","full_message_tokens":117329,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":454,"sample_id":"e7e9bdfbfe480854cb5e74a5a3eb0fe3460f4a736f28a25618591e5eb0388db4","semantic_task_id":"ffb6b6589b523170b2dafd3356c1b122bbc7dccd93289ccb3ff2a04d8e88a7b2","split":"train"} +{"domain":"codeforge","full_message_tokens":117391,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":455,"sample_id":"7c2f8abfee1a3ca29a8de3386fc065047690f0fd0a174d51e29989ba114bb1f7","semantic_task_id":"67e9cf72a4cde1f3a5eff2c67c80b4929280c0a034acacb994d43acfa2526dd0","split":"train"} +{"domain":"codeforge","full_message_tokens":117375,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":456,"sample_id":"545c88d8de990032334368c9e9e7478211b70b86144ad64973d3b6960385609c","semantic_task_id":"35f1e16eb81a4d12878ce1ca9ead598db9e34fc278ce4f853bade88d73975469","split":"train"} +{"domain":"codeforge","full_message_tokens":49804,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":457,"sample_id":"e7b18d53d1a6a0babf5d0376468a127a87c84603b0cbddf48a35a0c79d7c4067","semantic_task_id":"c795877f6a2e0c9ea355d1a0a498e778c22eb3dc06a2ba886cc9651d3de5b259","split":"train"} +{"domain":"codeforge","full_message_tokens":49765,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":458,"sample_id":"67101fcf004407c2fbcadd6655f3944dbc691167d477e1a68dac080b01fe64bf","semantic_task_id":"e48da3ff785ecb0bba3de2e0dcee1aab96bf1b2e31d8113db27fb0c9a886ed21","split":"train"} +{"domain":"codeforge","full_message_tokens":49778,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":459,"sample_id":"077fd13716485a129033e8ede8ae9cd86c1417ca0bdb646dff2556d0b332122e","semantic_task_id":"b4774dfb4a6c5533cc7d4bd84c6b7b4dcdb97ba477cc949526a53c817c31bf27","split":"train"} +{"domain":"codeforge","full_message_tokens":49872,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":460,"sample_id":"910aaa9a3cbd2589559136aaceb9f51536e8ec4ad2d70c8eb3e59d6ba44a7b17","semantic_task_id":"909309649711698bba1600c7e05d49a00a3b47e2a9f6e8006c101fce260f4e4f","split":"train"} +{"domain":"codeforge","full_message_tokens":49812,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":461,"sample_id":"9aa94a86f60622923d0da86ec5d050decc938ee440a8a74dde6b1f7cddf3c919","semantic_task_id":"6f403c924ad877efd3431982d9e20a29e05fe30b81068cd4a9f77b38417c2832","split":"train"} +{"domain":"codeforge","full_message_tokens":49816,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":462,"sample_id":"4676e9f59f47140dc0354a3c482a979767546ab52d2e6ec11e8112e3f788970a","semantic_task_id":"f150788a30266fd876f615e0840ba43cc6f549896231c17f558ddc5d40f88df8","split":"train"} +{"domain":"codeforge","full_message_tokens":125697,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":463,"sample_id":"638a2f63b2d97033e72179c2d4ebbb79c1989241ff9d395197639d4d04a19439","semantic_task_id":"f7baf1a0af570c271631b4bc1569865374d555c66d0dbe0514d86a699cdfcbe1","split":"train"} +{"domain":"codeforge","full_message_tokens":125680,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":464,"sample_id":"0f3ece2d1f4bed2ca3e26c2abe603bae330fd987645b392be226d067a24ced26","semantic_task_id":"234705acd33cac3480ed036f182f1c2be4f15c5030d3a4562e506b733ec5c210","split":"train"} +{"domain":"codeforge","full_message_tokens":125679,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":465,"sample_id":"38809c8f4a4b9640cbc7c17e5a921dc92f68e6ca15ec0b85ef0de33fae70995a","semantic_task_id":"1583be50dcea1af97e7c0cd40110b2b42b94eb15715a462ecf021d7360061e8e","split":"train"} +{"domain":"codeforge","full_message_tokens":125762,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":466,"sample_id":"fa9737621c763613ce76d92bfd5828168dac082b36437d835fa5ff6b74f0c3c8","semantic_task_id":"4239064eb666e74087b776197dc1149bdf8cfac4b367df479bcd38ebf0ff2548","split":"train"} +{"domain":"codeforge","full_message_tokens":125699,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":467,"sample_id":"4e73966cbbb27323f076bdb696d9fbcf5ac4edda8cf4d00d1732a0f21232bf32","semantic_task_id":"1680f4f9dd71b17603ab7bda1103529142ed3cef7aec3609800c72c9f864aa86","split":"train"} +{"domain":"codeforge","full_message_tokens":126878,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":468,"sample_id":"6e541886ac95138e5474fc533b17e68c611f91fa6648af2d89a1240b51d65c4a","semantic_task_id":"6e7fc943391e11a29f096835798920e695fb8f220e498b9a90b5dd403f288068","split":"train"} +{"domain":"codeforge","full_message_tokens":126861,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":469,"sample_id":"13b231da395affdfdcfb3026ecd04dee7047da95a81f16c93e10422c0e86049f","semantic_task_id":"cdf1ca0d97a4b7a7e72a9e69645a89fac5e86020dcd4820f560c7a60ce91646c","split":"train"} +{"domain":"codeforge","full_message_tokens":126860,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":470,"sample_id":"c3975538646cd5e5b858d404db71b152db41e2187f5e60c4298ab0b1ada6524b","semantic_task_id":"e1740b02b0bdeeaade13a8b3e307d31101effc0998fbda356ce981efd68ef0ae","split":"train"} +{"domain":"codeforge","full_message_tokens":126932,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":471,"sample_id":"accd725e6f8994515ef877a507cb0e2ad1eaa6a031328790f9acc4ea5d1e6aef","semantic_task_id":"2d8f84965225004ed7f348c42fe70bd5ab521460194da7714f8fa0da14a9bfdb","split":"train"} +{"domain":"codeforge","full_message_tokens":126880,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":472,"sample_id":"db55826a8b965405dd1d4c3f5842a5bcaa7c955db7d733e8397755e168090f82","semantic_task_id":"79f54cae6ce95b9ab05a8f174209d950fd1ebbf448b9072b8e5ec5c55d9c4e7b","split":"train"} +{"domain":"codeforge","full_message_tokens":38910,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":473,"sample_id":"994fb5cc094d64d2be83218ab0ccbb0572c3d9cb2a2c5aaa0f43609a0ae61469","semantic_task_id":"33c408f68d4076b4b4ccdbaa95b4395e94e50c5e366d0ae3601c27fe33e887b5","split":"train"} +{"domain":"codeforge","full_message_tokens":38903,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":474,"sample_id":"62873846c729ef071bd1fae7431ffc10b9b3cccbb28dce9f46020fda869ac931","semantic_task_id":"5b73a3861d137f6e26e5241865ea9a21e2bdde656c125b7369003c0d7b216b50","split":"train"} +{"domain":"codeforge","full_message_tokens":39008,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":475,"sample_id":"6766e2df9580869f52adaf2cdceb3c1004da8aecd6866917ed03133a8e9ef458","semantic_task_id":"4f651d4d47ee17df6043b67812a746aa4ba3f6fcdaa3be5e3b88631b2d5ecdc4","split":"train"} +{"domain":"codeforge","full_message_tokens":38926,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":476,"sample_id":"006d6cc39d50ca5ac4337b0598bb23bb445e7085dae5cf95323612e911d6778d","semantic_task_id":"1712fc48fb8d66c92c6e18c1825377042c30b1638d396b305d039474cb8ae20a","split":"train"} +{"domain":"codeforge","full_message_tokens":38912,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":477,"sample_id":"a8e6800256a749feffed83c84a7aa8794703ad2c030d4e2091c13972ac0cbc21","semantic_task_id":"cebadbc4a3dc92cf893b2c66e52fa95955fc7b5930d575a394ddddaa3a49fea8","split":"train"} +{"domain":"codeforge","full_message_tokens":41884,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":478,"sample_id":"f9305163e3d356faa6cbdba62b5976c820b4e4f92b8302f0cdfbcfcf648b45c0","semantic_task_id":"ad51c2034dfbd9df838376001bb43233a041c9af659b15b88d5a8dc3a86feb54","split":"train"} +{"domain":"codeforge","full_message_tokens":41872,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":479,"sample_id":"d9fe11f446f8211cdce36143387d940110387a8e1f8c5b19e1d4c1bfc22286da","semantic_task_id":"5480d393f6dbed2daf6e0d3be4eba7563776486549df3a78993dbfbd5940761a","split":"train"} +{"domain":"codeforge","full_message_tokens":41977,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":480,"sample_id":"011fbae30feb9df6c0fe50377c89d599f035f6a61593112d318dd89296e1a633","semantic_task_id":"d6d840c09a6ffee108fb0fef932dae18db0b6711224cbbf8f3689e818293c701","split":"train"} +{"domain":"codeforge","full_message_tokens":41904,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":481,"sample_id":"42b35f0c155955e496d143ca31b71c995d4658ff0699b9362692a87cd7c3d45a","semantic_task_id":"375b72b51eaeec022457e1a8121b5e65b51e67c51e8bd6fb7d39f08e05a576c3","split":"train"} +{"domain":"codeforge","full_message_tokens":41896,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":482,"sample_id":"6a05a1724550883da809fadf9a0466c9bd4c645c873606f9cfd0ca5eeedbbd1c","semantic_task_id":"42601cbe75e021ce3edd51254fdf79992d795543a680611ea81e5230d46d11c9","split":"train"} +{"domain":"codeforge","full_message_tokens":42456,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":483,"sample_id":"c9b44fa8491a6a70250f7159b3c463b882e805e7270ea524fe734b5481c50529","semantic_task_id":"d9c284c312af7c82161828829e9342c0e7bd5fd2ee1dcafa96e87a7caa78a652","split":"train"} +{"domain":"codeforge","full_message_tokens":42444,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":484,"sample_id":"8140494ca2a65d265f45b012d3147b90254fea2650b52d2da1a513c5236a053d","semantic_task_id":"3f87a4f7915f420be1bf2dd45ea07c5eaa1c4f234b5a7b7bca150e5ee9e1231a","split":"train"} +{"domain":"codeforge","full_message_tokens":42549,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":485,"sample_id":"6febdf1fb095c253d9a38a604424339cb309fe4b70a5f704ebf0cacc4283c215","semantic_task_id":"93867b92d6f5341da1fedc3f144e02a0a93bcf889e63cef7ec568980259d12f9","split":"train"} +{"domain":"codeforge","full_message_tokens":42476,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":486,"sample_id":"870ef3d8ccd7677d9e96bbf42f195f43bb283402e4b4ecf0062755326368a294","semantic_task_id":"a07c1f3fb261074ecbbdf8bb7ab0e53d3a7f16ff1ef80babcac56addc152d03c","split":"train"} +{"domain":"codeforge","full_message_tokens":42468,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":487,"sample_id":"f3b8af25835af28f6aa74eb8c5943ea70c8c149e5898b8a69d665c07b1753b9d","semantic_task_id":"2bde76b1f43fb2353b34c6ebd1c54373b00aa8a44fab0387591d5bc7c95277e3","split":"train"} +{"domain":"codeforge","full_message_tokens":139525,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":488,"sample_id":"f46fa8f9dd620ac75755db91dfd288bad04b940a77e116fbf435ac920f259e59","semantic_task_id":"e98d70b02002511cd57d9f3445f47bb92e2335fb6a9715f7e292538d1b3d53e3","split":"train"} +{"domain":"codeforge","full_message_tokens":139492,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":489,"sample_id":"e05090376afd416120799093d02b28663ec1c9b372d388d95ceb16952d71acd8","semantic_task_id":"c52f746d3e8f13a9cb08976f42455582236f809c073ff3bdfdc4f45edeecda92","split":"train"} +{"domain":"codeforge","full_message_tokens":139491,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":490,"sample_id":"af65a0553b35e2618624528c9c114a6f8b481856ef175c125cbd1e0817c7e3d5","semantic_task_id":"824d709c363165f6c4c4255393d6a1dc05d625c570be78f48fb4731d19964a9e","split":"train"} +{"domain":"codeforge","full_message_tokens":139576,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":491,"sample_id":"df617aea975f28247158689d3995609afa2080736a96a6c6039703f4eec6210b","semantic_task_id":"14b740f64430eadd05ca4f8f07387afb1714a3bda3885c723eaf578818b8f903","split":"train"} +{"domain":"codeforge","full_message_tokens":139537,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":492,"sample_id":"2efb237b0650739e91bcee0cebb1850e96e3a6726e1a1e820314daf57d6f2e23","semantic_task_id":"ca4a23e6805d6655d2e0cd9b1ab8ec6a76373452f213f0f7625360051d763442","split":"train"} +{"domain":"codeforge","full_message_tokens":151605,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":493,"sample_id":"986549ab26c7254821599e53f498cd9e1188a0ae2c10c1d326b69f95f2d32d0d","semantic_task_id":"0895fc1a628a66ddf7ae65a9ad732b2e7a2967024fb39c3687199a9a47a15698","split":"train"} +{"domain":"codeforge","full_message_tokens":151548,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":494,"sample_id":"d786f2593bd085e357c0f57d3181d28f5892edd3805dec36072cf21ae51f6bd5","semantic_task_id":"6809f0c519d87b56d9ca813bfb731fd9a0a0da229eb06c48f5b0d379f2d4c871","split":"train"} +{"domain":"codeforge","full_message_tokens":151547,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":495,"sample_id":"a0e7e9675002db0619688ca25a11a4589fd309c8e8ebfa8621cd534c7c01e153","semantic_task_id":"bc60c4a82b8aa5805c0e58c3d2d6cabe6d7a7204b063f4e5d15c320013cc9b25","split":"train"} +{"domain":"codeforge","full_message_tokens":151642,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":496,"sample_id":"21a4094b1aac1f09cc0f9c11e2e50450b9c9e0908c4f1f89b45363721958fa28","semantic_task_id":"38ec5b6cd94fa54588adcab207c6ede939e234c4b35b75e8b0f7ab88d00448fd","split":"train"} +{"domain":"codeforge","full_message_tokens":151617,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":497,"sample_id":"cc0075c1f0ddf4b7e926dfc05f8ad12be3d07151fcf41be7eac9176269525fbe","semantic_task_id":"baf3b9afa9311e7b994cbc0b03d56993d2b923697b6758601336d798b6e5e57a","split":"train"} +{"domain":"codeforge","full_message_tokens":149201,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":498,"sample_id":"676f974b2227c3f6ee52f8e3a4b3dce2c230fc3b584497e047e3d7dcb9c0f383","semantic_task_id":"dca08f362313b200aa9b546c840da7459d60f11f8399cabc008ab5cc7ad47823","split":"train"} +{"domain":"codeforge","full_message_tokens":149174,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":499,"sample_id":"abd3d53fb4749ff5b8d09ab432164deb80268b0d1c9425d437fbe026bb75acfb","semantic_task_id":"650b30b62eeeaf1fd10745f8a13d00b2bf4a566e11a1bc947b017aee7e2d98ef","split":"train"} +{"domain":"codeforge","full_message_tokens":149173,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":500,"sample_id":"30c5a3b2df9fada4422e192098dec2ab2d4e50cd5738bc9608d6d234d78ca498","semantic_task_id":"d2e3853192195d126d98f449a2124b21432e4113a1dd246b54055af07604890e","split":"train"} +{"domain":"codeforge","full_message_tokens":149268,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":501,"sample_id":"6d1e668a8ca9145515d7e40338ed203db85df48ed1de7b12226f045323e92a81","semantic_task_id":"c54bb4595cfcc7a9c3dd9f56e97935970a4edab8cb8ae4540072c3b88aa8ebb9","split":"train"} +{"domain":"codeforge","full_message_tokens":149193,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":502,"sample_id":"85e988764cc5a8131e5877202e4a70f89c26f42dfa5b6a99fef66ae3874771d2","semantic_task_id":"dffae570370e54b310eefb6027ea2b4a394e019ffee448b45369f6fcde58f36c","split":"train"} +{"domain":"codeforge","full_message_tokens":140408,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":503,"sample_id":"98e716b4a281e08fddb138f0effaed5f36b7d43bf1e200c7a138009d44f1fc1c","semantic_task_id":"56cc66234ba41f0d5202ae9ceb277df4092bccb6b87a2397d6eb31ff81e5a70d","split":"train"} +{"domain":"codeforge","full_message_tokens":140376,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":504,"sample_id":"d1577d1e07bc0f6583496e0ad8d147e09202a3bf35c5ad245f365cea64700005","semantic_task_id":"70e5ada8ee1b569427a1d626c66d3de5a9a12e75ebdeb4a1f1a4aa1b04c83499","split":"train"} +{"domain":"codeforge","full_message_tokens":140369,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":505,"sample_id":"c3d3fb134f31798735a10bf8e20d8a1c3754fbcf837b90339e70a1dab7da6ce5","semantic_task_id":"884e9e1e20290c6988ea25b16f446a72fce3d47ec2ce827c38a018764d38b373","split":"train"} +{"domain":"codeforge","full_message_tokens":140453,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":506,"sample_id":"a7183ed451e1b656c79fb0cd55cebdd7488f9495e4a1a71792c502fd0a3b8145","semantic_task_id":"0714d536468409f81863b86ac350d941949411b06a5d66ced61c2e47299cd0c8","split":"train"} +{"domain":"codeforge","full_message_tokens":140410,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":507,"sample_id":"63abf38ff6e99f7f45f3300ea7e687ef6edc89e9ed715a48891109d5b72bba50","semantic_task_id":"f8062f02f4d2bdd196fbbaf1ead8e711b56b64d57ab29e255b656bb951065577","split":"train"} +{"domain":"codeforge","full_message_tokens":65388,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":508,"sample_id":"8a3ae1df34bed1d329fa8bb6b1c8c11b54aec42da7776e408988540e58b1baa8","semantic_task_id":"c2d945d7302d3e4a5baee5866fc98b03f44925930a4423d3d7c84780f098f405","split":"train"} +{"domain":"codeforge","full_message_tokens":65366,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":509,"sample_id":"22e0672db022bd37b12b75f024a61e202b5b4d5d061be068fd1f5a6608d94236","semantic_task_id":"53188bde7e700c6b449916d1d56924799de8377aa3b107388860c29a5b98e192","split":"train"} +{"domain":"codeforge","full_message_tokens":65471,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":510,"sample_id":"d08bffaf18c94bb57d13abf5fbd9310e59d7b006b825f15c7d55a12ebc5ac827","semantic_task_id":"63168d8c23eaab3ebed3e1c1331bbbcb99deffd392f7d64538f9a272c82dea24","split":"train"} +{"domain":"codeforge","full_message_tokens":65410,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":511,"sample_id":"648ea717f6867c7b3f343d0dc65e82bfad902c23b9ff720b384105f76892db18","semantic_task_id":"1d96cf21e7d7c77b5cb7f939157a8914e97aee16f6698d2d2c7c658dd06e8a14","split":"train"} +{"domain":"codeforge","full_message_tokens":65390,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":512,"sample_id":"f2375366e3e1e613d6d0bb214775e1433541b7676c21b5121ec7eb2ed9a20da9","semantic_task_id":"6f651d420675801c46e3e64db2d4a75df95bd4019a9a4dd6f64f4115f3b15dc0","split":"train"} +{"domain":"codeforge","full_message_tokens":55997,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":513,"sample_id":"c2ca22f4383762a5695b18bc557101de63de2f15391be25bc2e06948acd71239","semantic_task_id":"62d33e08cd3bc6827b1560d4200ef292ffdc257a57b2669490397954be72171d","split":"train"} +{"domain":"codeforge","full_message_tokens":55991,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":514,"sample_id":"84323689ed7dd3cda049625046d7aadeed1dc34c9e808f48c6f222301ef44a01","semantic_task_id":"9792e6b0e56e1f2482db5063db6da17c022c754629159d1296b80897bc4b20e2","split":"train"} +{"domain":"codeforge","full_message_tokens":56090,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":515,"sample_id":"29b935d3970d1cbceb694042b992c1bbd20845338061ceef525b2792ad223361","semantic_task_id":"b19b9f99e5d8506ec4b7c6b3ff5abdccb787ce5b35503cbd37ce50eb0e557a47","split":"train"} +{"domain":"codeforge","full_message_tokens":56029,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":516,"sample_id":"ab0da7ee4dedf0d0fdb9c35129b1692a41cdc55c7d53fcdeccca838e7e8adc4c","semantic_task_id":"088eaedbe243fe73ac040cf62e10266ca623e11308f3dbe9506324d7c1ef0446","split":"train"} +{"domain":"codeforge","full_message_tokens":56009,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":517,"sample_id":"dd8f98bd0ae87cbefd54d7525197ad0b51833120da31589c274b955f8416c237","semantic_task_id":"a671f33f3769dabc733060c946d4e5093dce4819ee859624b73128be594bec6c","split":"train"} +{"domain":"codeforge","full_message_tokens":241326,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":518,"sample_id":"4b2f868e4fd78c16fe69be50e45107cbfafd886e42673d0c7d20cab801dc1177","semantic_task_id":"664238c8ad18d89509388a4a00f2a73f9c5740561fc02700e1e29724262c1a03","split":"train"} +{"domain":"codeforge","full_message_tokens":241262,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":519,"sample_id":"bd0f7550b7e3a1010214e5fc5dd6bfdb64b5c933d1e3676c2a332707918df21a","semantic_task_id":"30c76c1a706708e42889a7121a38def7c35c53bcfcf39e839add0c1456f8265b","split":"train"} +{"domain":"codeforge","full_message_tokens":241261,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":520,"sample_id":"6d89f4b6f8c5ebbcc2de4d938049844c230b64af0df7cf3ec61f1e583ee26035","semantic_task_id":"f53c3163b715fe715cb004207e695d619fff776c7aa14469717c3561933a0f91","split":"train"} +{"domain":"codeforge","full_message_tokens":241406,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":521,"sample_id":"7d96a3ec9e998df22f53c9955e3a50d037802ceb104666cebc3bf44317cfda91","semantic_task_id":"314f4759e65ab61ef947edcd23bbf35cc06c98b4738163fd93859561cd49900a","split":"train"} +{"domain":"codeforge","full_message_tokens":241338,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":522,"sample_id":"bf10efe7980573a2093f64dbb3743187915cb54e1dee5d0c9f7a1b345c1317ea","semantic_task_id":"4167e7b4c492590385963a1bca17dacc5da079423eb0613759c21adf65f80e1e","split":"train"} +{"domain":"codeforge","full_message_tokens":175109,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":523,"sample_id":"06d4b4a80923375d597b881c6271a09818e3297fa9228f9734a4d859b631f805","semantic_task_id":"35507eb83b538bad415efd9f846644cde7693ea4e3fa12ad7d05da7623b940b5","split":"train"} +{"domain":"codeforge","full_message_tokens":175042,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":524,"sample_id":"96e49355f6a05a43f5cf8b946f211abb4c0a405e63e95acd5a28ea0440f29239","semantic_task_id":"d8e0787e5f8565912c2efcc65350e4d90922690d91042ec0f8ba8dddd5806f69","split":"train"} +{"domain":"codeforge","full_message_tokens":175041,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":525,"sample_id":"fb61a3526b699c6576898e764f8649a2b669551599e27002cce407b219575657","semantic_task_id":"0081fe871840f362df7b01d4d640d2266cc253908b10244f0770dc9a4edb0028","split":"train"} +{"domain":"codeforge","full_message_tokens":175148,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":526,"sample_id":"3ea0840830594281f91f6a0600ce970849871241d8f6c2bd0ca152a026b823b4","semantic_task_id":"8d70ce7ef9f1e8101cc05c874a5035b4b8d0814be8379068d56ce6dd3d7b4103","split":"train"} +{"domain":"codeforge","full_message_tokens":175111,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":527,"sample_id":"850aa80875cdf7b881cc84b0aeb76cf8a7a4354d2c6a08c8bc63d4bbdf862e5a","semantic_task_id":"73fe1aef07afef747fef2103a5b954a5558638d4ce581a8baa2d2361140a756c","split":"train"} +{"domain":"codeforge","full_message_tokens":162731,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":528,"sample_id":"f4d2a502e29dfccd964b8c87f3020162ea95f58d600df791f8789d92bac50db2","semantic_task_id":"59e7f7c3255b2e5e59561fb131228b856a30a5fcae07dc82a6493f62006502ba","split":"train"} +{"domain":"codeforge","full_message_tokens":162689,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":529,"sample_id":"6f0967e4dea44bbe241efef52c31f6e7ba4eaefcc6edd8cc1e0c5b7f6e4d341e","semantic_task_id":"28df31a3b71a0a16f2bae3e331b294a886df889d5f876e475ea0a2bcaead210e","split":"train"} +{"domain":"codeforge","full_message_tokens":162682,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":530,"sample_id":"dca76bcd93af0702bf9e42f98424c3d132d2de688d6c412cfe3d024e3ec9f7b6","semantic_task_id":"59647d01985df33958447baf94ec7ffe1bc30de3f0698e939350cb11afe02562","split":"train"} +{"domain":"codeforge","full_message_tokens":162789,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":531,"sample_id":"8c96e43a0c47e4ef45e00a8629731da75cb99c3471870cbbaaa95cab4d3cd39d","semantic_task_id":"ceea47fe98c5086ce0b3c808bf3147790c5b683f8fe0758d714713b9defa695c","split":"train"} +{"domain":"codeforge","full_message_tokens":162723,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":532,"sample_id":"92546d38bef8089080d5428aa69413955932a24b0ee5cdc21d74c897f1fa873e","semantic_task_id":"dd7a6c4722d26718d003ea9c9724ab0b14ab19cfb546a6abc65bd42737ad2224","split":"train"} +{"domain":"codeforge","full_message_tokens":166886,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":533,"sample_id":"33c972a963b3f1c13b2503da202202eea9edbe38869c605e429c82147239d797","semantic_task_id":"4a439cac45f0fcc5c5dd85657d03e8b1b9e6b7d6d621e2f30df91f03859216d7","split":"train"} +{"domain":"codeforge","full_message_tokens":166839,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":534,"sample_id":"c321b6a45a3b33593b27f4cd7d62479fa37de680ed53415dadebb0042b80a16d","semantic_task_id":"8751b25549f21a4f0538255db5626d1f49673bbfb180cdda3b010462bd053b4e","split":"train"} +{"domain":"codeforge","full_message_tokens":166832,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":535,"sample_id":"b94d753084f10e6a1a08c8a0f1430496c80c5e7673dd77a75aeed61670af33ae","semantic_task_id":"bb98f089addce3fa06fad4a4ae4cb90a502e74947bf7a48a6fd581b3256e2393","split":"train"} +{"domain":"codeforge","full_message_tokens":166937,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":536,"sample_id":"585ff1346bdf33722c318fad4db9dffef61692f2a5732b667e715278bf3182a8","semantic_task_id":"bfbe07daa9ba3e689e9cd7952f6fd61d664d94fcf5f1b19b101d0cf4790f8671","split":"train"} +{"domain":"codeforge","full_message_tokens":166888,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":537,"sample_id":"dec0474b1e7722663152c9777878ef134c4aa32f7fb737d4e3b8d9c7f3a0d08e","semantic_task_id":"f5bddcf903aa51d566e4b0a60faa6dd6e5dce23e3233b6245d65e0fde11886a1","split":"train"} +{"domain":"codeforge","full_message_tokens":79501,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":538,"sample_id":"fd11f546a7ee45272c9b057a01f378baf29865313a25445b0dfe0bdfcd1d51b8","semantic_task_id":"2a93c3e795a9b19c96cd2c9a4eb0031b425b59e809ff8f411ae2f1794d66d52f","split":"train"} +{"domain":"codeforge","full_message_tokens":79485,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":539,"sample_id":"05438d9991467213d11b889dd698975bbadb7708449831322a7cb83773c2d947","semantic_task_id":"bc23e004cbe88f3d3d476f6ab349cfcdc701ed465e3f81ca7b9953a7e8abf30f","split":"train"} +{"domain":"codeforge","full_message_tokens":79584,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":540,"sample_id":"e581e3138edf7f0d9cb0ce17270b6ad55d187b17e52a35244b534e62dad83a1c","semantic_task_id":"211124a85db754464b0aad218094ff4c8ef7f1fad0e61426e4857fdd1ad7ab34","split":"train"} +{"domain":"codeforge","full_message_tokens":79535,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":541,"sample_id":"7b257356a13cf7177973c09344852d9e98027186f8a3df2f82994ef45ebb198d","semantic_task_id":"1d4ab08735095fb895b315010d08b55376d50a36d32c1c330a5166cd5f6b4646","split":"train"} +{"domain":"codeforge","full_message_tokens":79503,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":542,"sample_id":"e8fd885dc0a6b1596029ee7056107d705ce2b84e82c224f9394c0415567cd903","semantic_task_id":"49a77f9090ac839bc97de02451f8287f7873000988403277e6c4fd43459cc1b2","split":"train"} +{"domain":"codeforge","full_message_tokens":215096,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":543,"sample_id":"496a707d1789d48d7a7ba9d8d111062f97794e62370ada24929d4680972e4a5c","semantic_task_id":"66d1b6dcaf58fe2ed77fd37aebe6bd7a166ced01f8a5223a3301ff5b05e1c2f0","split":"train"} +{"domain":"codeforge","full_message_tokens":215014,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":544,"sample_id":"ccc9855b12adb77cc47b6ccd49fd005008b69a0e96b41f193aeed37ad2dc9756","semantic_task_id":"486fd053dc005ebeef8473c86e2fb2de04455e9c529c3ed90b9ff61805a6b5c3","split":"train"} +{"domain":"codeforge","full_message_tokens":215013,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":545,"sample_id":"a19d3e653f15b807f942ad2b4069e393f060edd301f9427fbfaec594b5f1b8c5","semantic_task_id":"e519a4d1616413a1dce1d22f945bc1a9c851ac54836d8ce306bd77674818bc1a","split":"train"} +{"domain":"codeforge","full_message_tokens":215153,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":546,"sample_id":"96992233408468ef3455e52c22e589126d27e481831fd7e820276d953625cd11","semantic_task_id":"2e796a46b200a7c06379be0bbc0ac800632d5e4bd64eec99c7d6b6cea42b8b24","split":"train"} +{"domain":"codeforge","full_message_tokens":215098,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":547,"sample_id":"5852ba4846ef885fb4c87691640f8dbe17ccaa1db7c347e261af29b47a5ba169","semantic_task_id":"d816e4f0f9828e54b4c2945ae5aeae8bdfabcf1f935e53888ce2165ca46c70cb","split":"train"} +{"domain":"codeforge","full_message_tokens":203322,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":548,"sample_id":"55745cd11de2e0cb99e15f2d2b1b9ce84d940d24387f073cf67095a1fbef3af1","semantic_task_id":"ec02f6dd19bc84e2e516450e6a5d2cf9542834eb7ed74bfb0fe22e99a58f9096","split":"train"} +{"domain":"codeforge","full_message_tokens":203271,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":549,"sample_id":"cf39ad34a7c7406a14cdc0b47a96b146581feb5790b014d3cccc4e71acbde252","semantic_task_id":"7d8b5dbe608b0368d59220d9aadf82261b73e1c74320ad529f2282de34ee900d","split":"train"} +{"domain":"codeforge","full_message_tokens":203258,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":550,"sample_id":"e37c543684b24789590ddbd30f590a031e63004bd125db29f720004295215924","semantic_task_id":"90674899e4d3104ed0794ccd3d887319a98922a76a0062dc9c19a38bd71f84f1","split":"train"} +{"domain":"codeforge","full_message_tokens":203398,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":551,"sample_id":"55eff9605a6df5cec2c758fdab8af9777b628e0dc556454f228d5694f19fdf88","semantic_task_id":"97bdefdcd730e57e83cac59003514faa4c9a89f0d86f299115542338a805a8b6","split":"train"} +{"domain":"codeforge","full_message_tokens":203314,"group_id":"https://github.com/pulumi/pulumi","output_file":"codeforge_train.jsonl","row_index":552,"sample_id":"e85d8e581dd8dc8d090daa1de4c9bb9ed31e13b7d414a1915b8574a4875bb293","semantic_task_id":"ae27f3332cca2a007907cacc8f355c0a926ed29ea087889930b1842a65d2d0a9","split":"train"} +{"domain":"codeforge","full_message_tokens":40329,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":553,"sample_id":"60ee39cb4f160dc4843e0259f56c9c8da43f293ba9e6b93faea0cdcad941b4ff","semantic_task_id":"56382de6063d1eb3a5443535d80ed9de876fa5fa3d073c83b21ad948cf0c0e65","split":"train"} +{"domain":"codeforge","full_message_tokens":40258,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":554,"sample_id":"0cd62bb969d01182d423d6831dfb8552dc5a8790f21502402387facf94acd82b","semantic_task_id":"5fbc4fe5f0acc0418e5409a6474245014948409e19982acf8a3d76e92895b59a","split":"train"} +{"domain":"codeforge","full_message_tokens":40306,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":555,"sample_id":"76497c2d845a127657061bf5373925a9161356ffd6b3e0157d61b4d06370a5bf","semantic_task_id":"ef6d4c8f0e12cc75101a7cb85b6077754e910c6a4ee4bd9c677955a8c62c8e5a","split":"train"} +{"domain":"codeforge","full_message_tokens":40285,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":556,"sample_id":"25a907eb0f183fb7098eabbeff69e81daab8f03f37111a2aeeff8c60ff61fb91","semantic_task_id":"08f00e9edb072c34be4337fb7e9f875bcb035949b611d961980ffcb794f1e5eb","split":"train"} +{"domain":"codeforge","full_message_tokens":40341,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":557,"sample_id":"d4fa0b3199e19e39c9ce65062b1d94afeeaf5674b63d07242b7f4c4234ce3472","semantic_task_id":"98701d7342db05e515c477ad121af2ca067e5a4dbca5183207f7927cdf9a36bb","split":"train"} +{"domain":"codeforge","full_message_tokens":65027,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":558,"sample_id":"5075fe31ffaceb69e7c23f1efd80650254a96a4298ccc219c5ebff1156644148","semantic_task_id":"0d558cbc031c674623f82415807b4c6dfa0ac4a4fcc5389b85758dbf7bde542a","split":"train"} +{"domain":"codeforge","full_message_tokens":64845,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":559,"sample_id":"15701a80efb5a0d155f82ebd56885ea674c30d16e30d44c8fb3ab133229bd347","semantic_task_id":"6f73417f8c85e8f1d0d562f0b32193fc7ac8311f7d4b58c24287c8a61279f397","split":"train"} +{"domain":"codeforge","full_message_tokens":64900,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":560,"sample_id":"9ea63075536135ea999e796a704d26f9dccee92311a562d62f5b76e81ede1a62","semantic_task_id":"56ab1987e2f4fe3789978a83b48f1582d69f6d0e86fb859f3748dee4dd42feb6","split":"train"} +{"domain":"codeforge","full_message_tokens":64877,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":561,"sample_id":"22655e8af7a7400a81a43ab4bffdd28816ae63de37b48de4c7cefd4c991d1ea4","semantic_task_id":"790163b472513353490e30ae3eb1878283f29f72fd69c2b148327503f2a20e49","split":"train"} +{"domain":"codeforge","full_message_tokens":72740,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":562,"sample_id":"cf88760c7b3c2c34de0573dbbe12a7b7110c503515938dcd66cb09ebb8476f5e","semantic_task_id":"cbe6febde4cd422e54995d9ea143eaa47d3325d9d3bd28e59e484458a8ba77be","split":"train"} +{"domain":"codeforge","full_message_tokens":72707,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":563,"sample_id":"e31dea481a8e8264999489c4e6e5af0b1d236c2a525830a65ad45994a88b531c","semantic_task_id":"a4ac909e51de1f2117d1f3e55e6199087600322b3c1a94faadabbea53e307845","split":"train"} +{"domain":"codeforge","full_message_tokens":72766,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":564,"sample_id":"ba5449bb2fd34191aeeeba8591f261b8053a02c90c0760ca6d8257f6201e6525","semantic_task_id":"ae4f3a89bde311c6ab5c4da76adfbc629ecca22693b947b1ff183f19e32b0b0e","split":"train"} +{"domain":"codeforge","full_message_tokens":72751,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":565,"sample_id":"b7d5afb9c1e4518d8a4e4f9938bdc277ec0be85ceaa8761a0a6ba6d0c73b1d1c","semantic_task_id":"d399d6c20d8d90ebe7e9875e6ddc56553a0bc2ecd1600c5e31f4a1ce07813eb4","split":"train"} +{"domain":"codeforge","full_message_tokens":72710,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":566,"sample_id":"05b69f221ee0fa97723ec97d14a908fdcd7a033ae9e55af0f0f0034d6b3f066c","semantic_task_id":"b8109e67f728296ecb4b261db295479594fe9e29322502ddebb7e662be1899b4","split":"train"} +{"domain":"codeforge","full_message_tokens":70827,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":567,"sample_id":"b102382419d49a6be6b0c56dce17ba3738eec2d12ff1fd6a8ef7f69d90a57cf0","semantic_task_id":"2bdbc38f0a4fade85e4bed8d1ef6bf544f7859722768b26a4313fa60bcf14662","split":"train"} +{"domain":"codeforge","full_message_tokens":70824,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":568,"sample_id":"eef5e63b62afda4614537a0c1ac6626f543259899c917c98a593611307389a0c","semantic_task_id":"0fd71d95fe541219a0398142e21f2b4ea2c8a7299a9685ba0f2d00d3e48aeb7e","split":"train"} +{"domain":"codeforge","full_message_tokens":71132,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":569,"sample_id":"8c4c516b727f185fb1f39dbf7a5fbe2fbfad52fa4b6ffd915ab64dade565fdb4","semantic_task_id":"8b7da446f4e277367107c7908ff99b14b1796758cfc07e590fe9f9b909245340","split":"train"} +{"domain":"codeforge","full_message_tokens":70865,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":570,"sample_id":"7c44ced924e0773baade2c8e8f72d323952497a8e83fe7bfda8f0a1c31eaad29","semantic_task_id":"759c137553b958aa671bfe1942d72bcb0cee4661e71205ccc5eb77b34b49e919","split":"train"} +{"domain":"codeforge","full_message_tokens":70839,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":571,"sample_id":"397073d2d9ca388de65bb0a082f8717145134aa118189fb000ca86d4c6a66c19","semantic_task_id":"4a28113405469562369bad433f2979f5bd474e3aee96e3eb8049369998d70c5b","split":"train"} +{"domain":"codeforge","full_message_tokens":77622,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":572,"sample_id":"6a2d732af283dfc7db872f18eef0689cea7a16e8a7fcd0c03c20e35b30e65fd6","semantic_task_id":"5cbc8d553025e095c74ce6b0057b5f44371c91dc690b0058326e4a0bbedcabaa","split":"train"} +{"domain":"codeforge","full_message_tokens":77620,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":573,"sample_id":"c5df978cc45655318c4c121038bf54d5ede259213117ef08ca874a1452fb4f4b","semantic_task_id":"11f12a1fd59d43d93252977303a3674f44632f1192db0ac163f5ff6c0c78f49f","split":"train"} +{"domain":"codeforge","full_message_tokens":77928,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":574,"sample_id":"f65d994b2c41355a3de478617339db588e34f2bfbf86febc1b6ed4c884681dd8","semantic_task_id":"26e8529b98da8dc909d2ecfacd534766e68f3bb88088e85f5524ea318de35a14","split":"train"} +{"domain":"codeforge","full_message_tokens":77661,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":575,"sample_id":"56427ceadfe0dc26646480e3c479177168e8a1cfc76866737c4506a2a400e5de","semantic_task_id":"a7ad71f2b627a8ad67b32459d376c9b40516f812457f7913c2eefc35097db62e","split":"train"} +{"domain":"codeforge","full_message_tokens":77634,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":576,"sample_id":"f7b0ad49299ea8aecbc2f901533a8b6feb4a273142b828b0fed80c2d2f660982","semantic_task_id":"d06cd8547e0321cd26b799326b1779af9a04d8319e9d27a140be07c333919fd3","split":"train"} +{"domain":"codeforge","full_message_tokens":77141,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":577,"sample_id":"fbe513481446ea63aa143b42dcfaddd65275500f13f9d8b8f061bf8253a298c4","semantic_task_id":"ad8f42e6cdf395dd0756c3bab5be2804179b6b0aa3833bbe6ef5c236d3b1863b","split":"train"} +{"domain":"codeforge","full_message_tokens":77151,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":578,"sample_id":"f1c3ef8e44c90f22ef3a6411097fb74deb4b915c7dc1b7a288d5a4e544928e11","semantic_task_id":"2a30351f3e25c8ec7171cdf2ddde3b0ce0ca764596d2776756ba6382a0b91d43","split":"train"} +{"domain":"codeforge","full_message_tokens":77216,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":579,"sample_id":"3be01d6a4f05201bea117cbbc1edb2e9dfab2415730dcc23d39d4410e0ae2afc","semantic_task_id":"02bbccd19f63585bd801d2319afe4b40c306eaca6c14088a44ba0a0f1a88997d","split":"train"} +{"domain":"codeforge","full_message_tokens":77190,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":580,"sample_id":"d2df69df8629925115e27aca78acf69f0a363ca418a0891a2479bc5ce73f0384","semantic_task_id":"ab7ca9c61d84e20daf8e46b54a8c47f67beedc552bfe816dfcb7c0ab3326482a","split":"train"} +{"domain":"codeforge","full_message_tokens":77153,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":581,"sample_id":"19c6862a4136e408a3802b6acaa522ecc22733e0331041542d3911deb7a84c12","semantic_task_id":"b176a14fa74bc0afdf179cd5998e45eff4b32204e89c1e52595e65356a5c7b52","split":"train"} +{"domain":"codeforge","full_message_tokens":61246,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":582,"sample_id":"043430bba065c6bcda2481e4b077abfc6db396ed67ae79d5ef095871302ec44e","semantic_task_id":"c07c165af846ba2192b554e203bdbac949e20df4fa10b6359784aa9b15478d9f","split":"train"} +{"domain":"codeforge","full_message_tokens":61011,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":583,"sample_id":"3ea4d97001be87cb8139daf1ad415b18829cd359bc30ea385c034b4b08c71b28","semantic_task_id":"0a9db6674f7d14b154263f390464414d03081e1453fd04941267efff9cce3c66","split":"train"} +{"domain":"codeforge","full_message_tokens":61046,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":584,"sample_id":"bdb8b6dfad924bb50ab5b755c3ec317176d9aee9288093e2e88f88f43f4b5eab","semantic_task_id":"2b30cbb70883dfbf0d4fcb2f81fa2fe458bc37e5d866ffc5780a7206ec330266","split":"train"} +{"domain":"codeforge","full_message_tokens":61050,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":585,"sample_id":"8f64de4106802408d28aa6d13ef823ea2a5846840e70db9e2ce58509a00fb375","semantic_task_id":"b6e61e0d48d37167f2035a5b0fcbbbc14d9a9126e23caf4241e9c680d8bddd2a","split":"train"} +{"domain":"codeforge","full_message_tokens":61093,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":586,"sample_id":"99daa977dd673610fbc8cfd3518bc5eac6f8502b8a6db98c7e50d117e3dd518f","semantic_task_id":"07832ba5f4df59a83c48868cef2855da0c92479125e36e028064180e340d4284","split":"train"} +{"domain":"codeforge","full_message_tokens":73644,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":587,"sample_id":"8f63d8e29c845f9a12b40dc6b60f78b1fdcbba2ab36d35f3b9e4fd7618c28ab2","semantic_task_id":"c0150b361400beabe8633d0db1ddfbb0cc1b1b0324b702fe8ca12e28caa9e674","split":"train"} +{"domain":"codeforge","full_message_tokens":73448,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":588,"sample_id":"0013af08be699779897e5c391a3f6517c3d23dacdbec1701aa287a9fd6f86f4a","semantic_task_id":"ae4cdabf144c9704b870b6cdca36a1c3d150a06c90de8450f99b62a0ca1387a5","split":"train"} +{"domain":"codeforge","full_message_tokens":73503,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":589,"sample_id":"a456839977f5571232890557c8fa5b6154290a0098c8d931128bafdbf23a3529","semantic_task_id":"7d31c83e669403cefa8bf0ce034b0938e0b74cb2a888b469cf8f9287807787fb","split":"train"} +{"domain":"codeforge","full_message_tokens":73494,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":590,"sample_id":"45f3b11888f6892e9a48a775c4242de854ae9b10de02e60cc7ebb814f709def0","semantic_task_id":"27e7ef746fb9688f74c5472c14751490bdf482bdc60dd3cbba1894e9a40c33fb","split":"train"} +{"domain":"codeforge","full_message_tokens":73449,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":591,"sample_id":"3cf1987a4bd0cc4c7b68fefea9bdbd10e25b4b3a68853c9973d15a37adea84a6","semantic_task_id":"d19e4ae0527380ad227e78a0487bf8516d9bce0b5e79aa951161e19cd7a6c4bd","split":"train"} +{"domain":"codeforge","full_message_tokens":93816,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":592,"sample_id":"ccffb4f760d23d3d18323e34dc2958eaef32885852ddc85412fb4b01dffd7959","semantic_task_id":"a394b07dc228b11ac39ebe5a0dbe8527c8262ea6df2271df8c2e36f2247cab9a","split":"train"} +{"domain":"codeforge","full_message_tokens":93619,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":593,"sample_id":"d369a554570acecdf4bc322840627e0f55f1710730a0e65d00899968340efbab","semantic_task_id":"c50eff13245b00a540b8e95adc134e8a22d1b54dcb5b86250e3694b88afb9abc","split":"train"} +{"domain":"codeforge","full_message_tokens":93655,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":594,"sample_id":"4c2ca63bbe217c3c258f39e5f1dabe86b91227efc7ea3691cb97a90e28aaab57","semantic_task_id":"662c63aeb263bd60f31f23befa6e7bc919a02c667dbaec19d9b30c0f78d76cb2","split":"train"} +{"domain":"codeforge","full_message_tokens":93675,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":595,"sample_id":"bfa2ee8bfcdf7733d1243d094e71be60f595aebfc08b12671d13689c8379fb3f","semantic_task_id":"a791a483115b9e7ced1224d7a4996d94243eb6d2053895bc4a066a7693642df7","split":"train"} +{"domain":"codeforge","full_message_tokens":93621,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":596,"sample_id":"e1483e6a89d302fc3d07af596e9cf87015ec3c0bd7ff78308634a19e5c409716","semantic_task_id":"7a25f1c8cf58c7412326d2fd2a6e8c0e7d9bc31229a24c86b7a405d0be70d8f4","split":"train"} +{"domain":"codeforge","full_message_tokens":112638,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":597,"sample_id":"4cabff61cfa854e09efdfc1e7bd20c6348ddd3d9bd1dbed137927e5ffa79da9c","semantic_task_id":"df1dda556307424752de5dc3bf540f517291bc4d6b9392c1a1fa7d319746c7c3","split":"train"} +{"domain":"codeforge","full_message_tokens":112587,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":598,"sample_id":"23fcf35eeed37c0ca9b79b60a0a9e24b9860c349689a7ffd48423bbedf32c23a","semantic_task_id":"eb283412ae03db316eff09070dc32cec2af0fe3b99ec15b6d45aa54480467906","split":"train"} +{"domain":"codeforge","full_message_tokens":112630,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":599,"sample_id":"aae7dd54bcd2daa039d2b139d20521b9fa994ee99b09cc2bf7e7ece81d2c3865","semantic_task_id":"8cba466485e1e3a61f70d8fe2baae2f3839394b6a21996c402f5599e61e65caa","split":"train"} +{"domain":"codeforge","full_message_tokens":112655,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":600,"sample_id":"be3d313aecbc23851df2097543a3a069d2d83cb3c9653279e8d620bdd3efe39e","semantic_task_id":"3eb477bf870f24917339298bd1adfe380d9779b50d28f072fdaf959b08e83d54","split":"train"} +{"domain":"codeforge","full_message_tokens":112608,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":601,"sample_id":"c7128279fd5a71bd1ee33e05deb7a4582cc75eef2ca711e38e2b12a830fbface","semantic_task_id":"a0360536bfb151ab77d729fbb77f111d3f6a1be15fe435fbe43dd5bb1d7f97bb","split":"train"} +{"domain":"codeforge","full_message_tokens":106404,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":602,"sample_id":"158be63c849a685c9628ff6fe700dfffce6a65afc874601ee86215404ca11ea5","semantic_task_id":"cd273a42efd3bf48bda947bbfd8090f39e69fe4b1da02aa854070068c427527e","split":"train"} +{"domain":"codeforge","full_message_tokens":106393,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":603,"sample_id":"7abbf570662ca197c249f27a8a2bb451f316b4f33eb92657fc5b2caa641669be","semantic_task_id":"ca49b3773d8da8bd8f9e132a1816a0b843b93ad6c778d46d2d0e5be60eb9ed83","split":"train"} +{"domain":"codeforge","full_message_tokens":106695,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":604,"sample_id":"4834d0aac67f5a9750153a6ad39548b36b3e959445674c3068516ee6a7e0e640","semantic_task_id":"4800520efe9c1d1dd4d3c272ba4576a1f31bb42a575dd62a40f7acf52537787d","split":"train"} +{"domain":"codeforge","full_message_tokens":106452,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":605,"sample_id":"df6e87fbf33aeaaca0ecf4053b8a4cdff3c5c31e477c88f22e73913ca55e0ecf","semantic_task_id":"8cdbebb8b3a4e8f3b1c9710c49e470fa28be2f149b597d59100834508a14c06f","split":"train"} +{"domain":"codeforge","full_message_tokens":106416,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":606,"sample_id":"c91dc182011333a8087c05dc147d76761c4f8ab91bcaab51919af7cdb102dad6","semantic_task_id":"ec31915f7b5ef54f03f394c65a94f77dc11e40c23fb30782f4c66535e5cdad74","split":"train"} +{"domain":"codeforge","full_message_tokens":117068,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":607,"sample_id":"1c0336795339e16c266aba79ad304977ce226bea097aa925220208224b1140e5","semantic_task_id":"e1f6f65d58d284f2fcd0d12b863c03fa328e9210d57dc33a93944fdb952df545","split":"train"} +{"domain":"codeforge","full_message_tokens":117066,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":608,"sample_id":"3bd253083e3b0824e26e71860694869cf49f843c7102662ae900003586a28895","semantic_task_id":"5694dfa54e1596823d939789fc450b71f4125924716de5da4ed88b4935994c0b","split":"train"} +{"domain":"codeforge","full_message_tokens":117125,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":609,"sample_id":"cf6499f049cba2740bcd9949cf671fa4a75abb7c2523c9709e198e509959558d","semantic_task_id":"e9ae0c96555504effb484846b5b0c0e2c48e00568d7c9b461e8577cf4a92205f","split":"train"} +{"domain":"codeforge","full_message_tokens":117123,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":610,"sample_id":"4a94444851e4e52f0a1b15f9fd0203618b3a2fb1c638eee11a251410dd20b35a","semantic_task_id":"50f2defe4e54084bb9843ae6f9516562bbd7e2c4753b91f42210a5150f54a0c5","split":"train"} +{"domain":"codeforge","full_message_tokens":117080,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":611,"sample_id":"929c9b908ec2124ed4fa9f7fed5d098cad4265f7134ce1b126fabba3484a76df","semantic_task_id":"9bd822db6ea02f78f08b080fc110d8ee63e1b161768cbfc25310dbce83c57486","split":"train"} +{"domain":"codeforge","full_message_tokens":103022,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":612,"sample_id":"954fe203372de4bf754e13f308d4b7fbd20fc057d556b8511d78fb8484f6ade4","semantic_task_id":"05e97ce34b23048d0571e2b3a838ec2770f8c4a8e725872c132de9a0975164b8","split":"train"} +{"domain":"codeforge","full_message_tokens":102745,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":613,"sample_id":"5f031056effb3ad596e17c184c01baf0d97c1b2ec12afb1de7d7acebbc7c5f71","semantic_task_id":"e16187b4dab42a7a1c99f26ee91ccd98e25fac7ca6f1e949bb8ab68b84e56b8b","split":"train"} +{"domain":"codeforge","full_message_tokens":102780,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":614,"sample_id":"c65d04c124819049fc135a1e8e8f278414ceed9087f3286c1a910a43683600ac","semantic_task_id":"a9150cb5347ba6e5fd01eab021d9918ae940d0aff8703e55d95f33e154a057b5","split":"train"} +{"domain":"codeforge","full_message_tokens":102805,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":615,"sample_id":"e8d5fd3f255c24acbe24b93ce617458492c33af12dc33a321be439f98e459f9e","semantic_task_id":"3d0d32fd9ba5c2df59aa71c871d5a33057dde3530facf2c8e5765d8164d9f4ac","split":"train"} +{"domain":"codeforge","full_message_tokens":102827,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":616,"sample_id":"8a7140c6c504387880f233cc6e422a560fe109fae0c1abcdcfad098013559e32","semantic_task_id":"50602250606f1873ed03036d353264aa22de1b497ecbd49338f07083149d7bda","split":"train"} +{"domain":"codeforge","full_message_tokens":102433,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":617,"sample_id":"58b80cd48956c463313cf3f33437c488a83965b49dcb392ba96383c3aacdb384","semantic_task_id":"24865eefe798688710da1421c1ec174fcb793a05d837cfded0e4fe772c1134ed","split":"train"} +{"domain":"codeforge","full_message_tokens":102222,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":618,"sample_id":"99c8daae6711d8b11f66ea3b53754daccc332a2f05a3c8c21f83ba8462f3505e","semantic_task_id":"de7d7af45a3350b45285091d6187398bb093b85b4e88370985c8644bb217e584","split":"train"} +{"domain":"codeforge","full_message_tokens":102258,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":619,"sample_id":"efad90b56f2a3c5cbb08d4b86e9fe3f684abae57488f61f6ad18faac9e3fe1d9","semantic_task_id":"348d3673aa2129c0a7696464f9864c4103961bb61f0416d1f7e22732a361c0be","split":"train"} +{"domain":"codeforge","full_message_tokens":102292,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":620,"sample_id":"03530bd45ea3b50c7dce19e4eb4b8dbc28a9f476927cd15b62e0d036fbb3ae95","semantic_task_id":"0b5a91a2ef20b2d0f2e7ed39c5ecd0a2d41d392ed83f0b7f460140640d053f40","split":"train"} +{"domain":"codeforge","full_message_tokens":102238,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":621,"sample_id":"7e6a3acf1a580ab918da4e4476af6b71d4ba5c61aab0cfa491eac17f16ccb038","semantic_task_id":"d21e35991e9aca304c9654c64380141e678d52c6d89b39758e091ff97e53c83c","split":"train"} +{"domain":"codeforge","full_message_tokens":133714,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":622,"sample_id":"17fa3220fbbbc4ad3de9ff55b1b5f520c68a87882d28714732d498b7fb3d06e7","semantic_task_id":"8f4f5869041216b6252c10d3c582ce726893baceb22b8b0bec41cf802e2eb284","split":"train"} +{"domain":"codeforge","full_message_tokens":133499,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":623,"sample_id":"4de35fec32efca55dccb410799abfd311ba99a41753a72ddae05f3920e882a53","semantic_task_id":"9375f0448a75ba5265ff9f99b03b2ef7fbfa34b3a0806bdd846de93366c7135a","split":"train"} +{"domain":"codeforge","full_message_tokens":133519,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":624,"sample_id":"461c398168cc376750d3923b19907ebed2028e527280d2635f2d97972596a1d7","semantic_task_id":"3459de4775e1f1f44e2e8aa6971df8830912cd99ae81047135f0681120b97acd","split":"train"} +{"domain":"codeforge","full_message_tokens":133579,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":625,"sample_id":"e1e6b1f65135f836f09a2c760645a83c885ff30d7e097cad7aea5d376eff1583","semantic_task_id":"8c3ec8424489ba6e1c033e13e11f089227c1842f160672295580180e25ba7e10","split":"train"} +{"domain":"codeforge","full_message_tokens":133519,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":626,"sample_id":"bb13a75ee1f09870c437e2e5b3a42f8898309fb4f32f82022964deaa8cf87efb","semantic_task_id":"cd1357868b656e06f66d192e7f24cec9f1b79c668d2063e519f341f9b39a3741","split":"train"} +{"domain":"codeforge","full_message_tokens":148215,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":627,"sample_id":"2c7335e217dee04b15292cdc04d4f7aa640cb57a2bf4a78084de92a3e005953e","semantic_task_id":"5d9ff1654b73bba21269b8ce7f35ea9e481b0916f76cd5829920a8399266ccad","split":"train"} +{"domain":"codeforge","full_message_tokens":148150,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":628,"sample_id":"93f920a6501e8b2602d78b4ec195dace8b2c1fce84202acaa00820e5581fcefa","semantic_task_id":"8338f2caa93bef7dd04a23013b23de53956ea4a267203b8e877a692afa80d723","split":"train"} +{"domain":"codeforge","full_message_tokens":148193,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":629,"sample_id":"7e41652efedfa715fdc46ddb16a59379d79c5121f99a5b65966eb654a245101c","semantic_task_id":"8bf8f69c8fb6e56ebd7d018958cd1eaba5d227303fbd27f35e790808ffea827c","split":"train"} +{"domain":"codeforge","full_message_tokens":148242,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":630,"sample_id":"278056fa04e6110533e53241b9a401bb0075dfc0f1e5a730c8a1a60d1e43e0dd","semantic_task_id":"a25fcae2e671c3a1a7edbc97ba2cca118086a4d319f743d99ba334c11ab568e3","split":"train"} +{"domain":"codeforge","full_message_tokens":148185,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":631,"sample_id":"8b0ec8c6be663aa021d05ca5b2c7f937408075560804004a82f29fc82656a0d5","semantic_task_id":"c9aaa305e38a5db92bad12a0c8ea9a475a622983c3af61060efedbd5d37142df","split":"train"} +{"domain":"codeforge","full_message_tokens":145850,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":632,"sample_id":"36c685cf013d8101caeae26cf80ce8df58e716f115d07a874d517e8aae8fbc57","semantic_task_id":"c41d943c1cc240253a4e6f9bdbd9e01399e7467227fcb5dac77da6729ac12909","split":"train"} +{"domain":"codeforge","full_message_tokens":145839,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":633,"sample_id":"f8d84e0116f01be4aae875b4b1a45bfe779cc7d724be5274bbf1e1e9e7d286c7","semantic_task_id":"85af31c8e239e690c73ed010b7fdde460e4677ecbbbad2852d2fbb8eb5470f93","split":"train"} +{"domain":"codeforge","full_message_tokens":145892,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":634,"sample_id":"677f16755c98f2b63616abfbd36a86cb7a5df793002c074d24863767e41d5c1f","semantic_task_id":"b4403d2f95ebb6377b9a03c8646d94e9239e6fddaec73c75223d45970cb4b761","split":"train"} +{"domain":"codeforge","full_message_tokens":145914,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":635,"sample_id":"fa185bd593aad90a9960dff1ba0f0a9584e609c34e97030a71d2938568da36b7","semantic_task_id":"dd0cdbd683904500cc1de310fe261a36155fa9146acfc3c0b4cb7a236f2ef9b7","split":"train"} +{"domain":"codeforge","full_message_tokens":145862,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":636,"sample_id":"e78e71b8ee71501d72253147821c3bccc7953f15f57007fc59cafec664f69089","semantic_task_id":"02f2fb6f1d07ae2c75105c1db3ecaf0c607d48744889c234039f346a49612fdf","split":"train"} +{"domain":"codeforge","full_message_tokens":171709,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":637,"sample_id":"291ed7630446d864da997f553c210d163a2a23f52104ab6d5d059af941405b7d","semantic_task_id":"35b6a8888415238b8befbdbb7dd5c25c278c85fbda6e05a89afd6172c4ac6f2a","split":"train"} +{"domain":"codeforge","full_message_tokens":171399,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":638,"sample_id":"30a6afe0a099301fa49efede1c8a8c1d01cd94eaf6244eb72af4a88ea27605fc","semantic_task_id":"34559186a43e2ae1f41cd89c08afeeda18f6ab6f416c8de2950b1ed888e8a4e9","split":"train"} +{"domain":"codeforge","full_message_tokens":171419,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":639,"sample_id":"67bf1f831c5b2eb0b0081542ccdd5c082d99c11facd66b42cc13010deb7654af","semantic_task_id":"1e757b58df05a5260464699e8cd0a097ab4de52c4e233226d4b160892e92743e","split":"train"} +{"domain":"codeforge","full_message_tokens":171507,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":640,"sample_id":"d091968ff5ceeecc243beba09bd3f98e6ac669389f07a2a731c460f4782b3d97","semantic_task_id":"da22cdf18f27b4265ad7df6c661cad60b1f380c35a76c2c170f3f6adaf426966","split":"train"} +{"domain":"codeforge","full_message_tokens":171514,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":641,"sample_id":"00316b179c4af30dc35b1e72e6f1d869379f729fa9dbc373b640a5f9941604fe","semantic_task_id":"6f9b18148b101031c75ebe623451341dd92b1e87552a2b252c6779e6fd0f88be","split":"train"} +{"domain":"codeforge","full_message_tokens":177908,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":642,"sample_id":"f661b5443ed0f79b91e72530317ac94a5064bd0b5406a2ee40634002de8b0143","semantic_task_id":"ffe52a2e4c4561e1d222cd2a9b2e3622d472afb11d2e62f8c19f0f5ae426b28d","split":"train"} +{"domain":"codeforge","full_message_tokens":177665,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":643,"sample_id":"64751204b6dbd2eabcea61ce985a4b350f0d83f82575a478fe1ce83a33ade85a","semantic_task_id":"f6e9be2878acbfd235cac2b91a4bbc4d21a4c8ce4b2f1d0666e71ed863b55fa5","split":"train"} +{"domain":"codeforge","full_message_tokens":177685,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":644,"sample_id":"8242e42fe5b8f90ee3fd9a8435fa308aeae2c35e4cc6c6f17fc75d3ad0f3bc05","semantic_task_id":"f9a6eb7785307e478729b0400e8b6b3c61a3490451e21e191f19a15a9f51c84d","split":"train"} +{"domain":"codeforge","full_message_tokens":177783,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":645,"sample_id":"9ef27143d5982f4e701922ea524007cf2f77875d08b39aefb21b89f8b54ad331","semantic_task_id":"1128fc64ba9e5c6d293f1750bf7079c91abdd09b77ecd50e22c52f756e7908a4","split":"train"} +{"domain":"codeforge","full_message_tokens":177713,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":646,"sample_id":"1b0a26128b97658af35f20cbc49b23916dd94fb939fb6df783bef594bba06e1d","semantic_task_id":"677816c86040c9c5bc95090cc313f025d6d1d41e8e9b34113dca6093a6676e7c","split":"train"} +{"domain":"codeforge","full_message_tokens":208737,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":647,"sample_id":"c21c13e1f4972c1a2ff42eb5c5b07fba95352e0475da6a2fbac2ccd18dee4491","semantic_task_id":"faff7b26d22bd6b95340b330f94dfad0283d2f14322e961ed57aa7f98118f5d2","split":"train"} +{"domain":"codeforge","full_message_tokens":208502,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":648,"sample_id":"319a832e1592c76226335a534567b2d72b8ae40a30db4cacd2e4c3676b09c71f","semantic_task_id":"23c39e8c076283f7ded9c2cb9bf40d6194ec033f81a78ecc9bd6c57b8f7a28fa","split":"train"} +{"domain":"codeforge","full_message_tokens":208522,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":649,"sample_id":"326d15b03d0c46ab49e40fe6f0d91e13fed32b3a93d7c5a581c9dd8385faad6b","semantic_task_id":"b8dec84dba403f878c99933b9e44e8f78bf718e1a4d206831db861b0c2c551b7","split":"train"} +{"domain":"codeforge","full_message_tokens":208628,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":650,"sample_id":"e0f97d3adeedda543054b248ef9b224e3707b78993ee1b10c52f6ba78b602119","semantic_task_id":"b3cd3b2511006492e452e472e11aa464f277da38d9709e13be3a4477bf157525","split":"train"} +{"domain":"codeforge","full_message_tokens":208542,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":651,"sample_id":"c75661cc98d83f311b8242a63502e841fa86eb971d2135d9a1faa6a98bd37cee","semantic_task_id":"f78edf457219b4a2a2ba0834872fa4bf8ee84d14221fd01ec1c7ec7e9f27a15b","split":"train"} +{"domain":"codeforge","full_message_tokens":246732,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":652,"sample_id":"dd92b0c94335d5c70c4f9799bbc6b14a0ba2e7c22dade1d69007d40b17b85765","semantic_task_id":"fe80c15b97019738a8f63ee114fce55da2bf7010976c73c603456f7479c03ec9","split":"train"} +{"domain":"codeforge","full_message_tokens":246402,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":653,"sample_id":"3a3d008adfc59defb53629667f7f9a652d35d452d3a009a1e1390e31fdddb489","semantic_task_id":"7b53da4f00ac94af4e465858669d2a5dd764edcb17d4acd4eb1e3c31d28bd604","split":"train"} +{"domain":"codeforge","full_message_tokens":246422,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":654,"sample_id":"6e3493340a37431ae01b1a99deb010e85db5a4213481c842b707efcc5913fdf1","semantic_task_id":"2b903234f8f5688bb1a5f0836473c2b2c2d060580b22b37d26d720cd8d7b71ee","split":"train"} +{"domain":"codeforge","full_message_tokens":246556,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":655,"sample_id":"6a94a3245706ac8502a3bb4a3e251f6feba10e107441b38c54756a81fc12d6b4","semantic_task_id":"97e0b35c832d130e35592635eb02b32a81534b53725c91ce5157dc31f098d9c0","split":"train"} +{"domain":"codeforge","full_message_tokens":246537,"group_id":"https://github.com/huggingface/transformers","output_file":"codeforge_train.jsonl","row_index":656,"sample_id":"5f65043f1877e4c443d1f7abbe93cab822eb8ea670d71b39da6d5edef7433406","semantic_task_id":"be386adc0030a75068844b3af9dde46607f64b0db959abb2dd237c00df5eafe4","split":"train"} +{"domain":"codeforge","full_message_tokens":85010,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":0,"sample_id":"5ea32d84b91a1dca45dda1ce2272849a393b59ba0d9742bba337dea2ecc788b7","semantic_task_id":"354c92aec6acd0fd9ebcb6f5aea0777306c3f090234a7c7aaa6314a56ed740ee","split":"eval"} +{"domain":"codeforge","full_message_tokens":84977,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":1,"sample_id":"4d506724d763ab988ca9d5b2503953d4fd5c45c1d4da969c490d2374adf927a2","semantic_task_id":"b3adc4a037b97cc723c3338ad96f3bda51180b4ebacf95aefc63d5b4c8fe67c3","split":"eval"} +{"domain":"codeforge","full_message_tokens":85515,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":2,"sample_id":"d906163de6ecd6d12f7a207b843f37c0d2b0ee17794083dc46c4aecb4c5a6f7e","semantic_task_id":"c39a397a986e3868c73ac4a6cb7460020eac97c7f3a0644856f812da316b48e9","split":"eval"} +{"domain":"codeforge","full_message_tokens":84997,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":3,"sample_id":"3760c070c8c0e95e296329c7f7a202adda58171a9b4f6972e5f69b8f258b5a15","semantic_task_id":"f96c3b148aed807fe1140109d8b1bd08895ddff21feebb1b15de2f0985ca6971","split":"eval"} +{"domain":"codeforge","full_message_tokens":84975,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":4,"sample_id":"03c6dd904d7a3e69ead28d420c93b773043cd55db208719dc3e965fb3959436d","semantic_task_id":"7bbbabd6a9ccf6bfad6dfcaf41e32f6dfdb312df6d3de7581b2ef646049a11e6","split":"eval"} +{"domain":"codeforge","full_message_tokens":81759,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":5,"sample_id":"fa431598393eeb4ede1338739ab1728fcb391cec8000f345435b3cb76f3c454d","semantic_task_id":"aee7a2cf474f4f1b5c848402a99533a421664f53410351b45f14197061cfcf65","split":"eval"} +{"domain":"codeforge","full_message_tokens":81681,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":6,"sample_id":"ac2fb98f95b4551a809c34c2d78fc5e5406cff3ded34a3a4e3ff81338f0f4c7b","semantic_task_id":"f3473f89b9126e810cb4574f9e38ef9ea0de0755ae272a1fd674ecba1a1243f9","split":"eval"} +{"domain":"codeforge","full_message_tokens":82204,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":7,"sample_id":"1cd0c2a9ea8fdee4720c2818ff214dc427e1b5a7fcad5c907c936c92de698b1a","semantic_task_id":"17df266c3d06608a5c80c6bcf308b00a1251691b1fabd6b8f520e7a025308f7d","split":"eval"} +{"domain":"codeforge","full_message_tokens":81771,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":8,"sample_id":"82105f82000fb92b58dfd16a0429fab524bf1ecce538924af6f7c8beaadef520","semantic_task_id":"fb91875beaf8df53d362e9a9d73f23824db4e32c38df86aae2479757a78bb1bb","split":"eval"} +{"domain":"codeforge","full_message_tokens":88426,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":9,"sample_id":"cba5bf940409f55cb750d7212dcd779868105b91cb2df436c1ea988909c42b0c","semantic_task_id":"88da10213106667b39393db7516e65b94d48f793c0ee1c9e354225d9bc530062","split":"eval"} +{"domain":"codeforge","full_message_tokens":88348,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":10,"sample_id":"28172e04392700ed110f0731c067a3c0cbd37244eff5c31200f221afe5ec7dcd","semantic_task_id":"da106d0497c79d53b9551c814629c9cf768a4c0b9cfbdfd406b208624e635b39","split":"eval"} +{"domain":"codeforge","full_message_tokens":88871,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":11,"sample_id":"d26a66a5b36ac842281d01ea02d92e3f755b903512fbe0d2221b9ac73c3a34e9","semantic_task_id":"4cbd1354e2ea1024fc4bd6a6495fe508e692eefef70fb5642dc6345b83e13135","split":"eval"} +{"domain":"codeforge","full_message_tokens":88390,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":12,"sample_id":"675407ecb7cf736b2abee2f7385f0a847a48f22c4248bf33c3ea1e2d7b9c2d2e","semantic_task_id":"b0acf42b08400f3a93d261c7f6d103927fa5eb5f0899d39ea5de55ec4587a1f5","split":"eval"} +{"domain":"codeforge","full_message_tokens":88438,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":13,"sample_id":"f4cb3b870ee889ccc33f3a120b39064980cf19b3d4af3d9363bf800915ed9641","semantic_task_id":"3e39800971de06c77df42f1711d6ef21a0588d3c293e5751f37baa5c3de16d2e","split":"eval"} +{"domain":"codeforge","full_message_tokens":51167,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":14,"sample_id":"1f5f128f01671bc180220d94797b1c3655a5ec6e702304f488327e9169d8d1c4","semantic_task_id":"31ca8d43414c9d86ac555fa0528a05ba34165cdce33a7b18086c88a59e52f448","split":"eval"} +{"domain":"codeforge","full_message_tokens":51147,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":15,"sample_id":"e85ff10df9fe4d5ff876deec49d4acc3955c45f5db63073ed7eedd544c5f5dc9","semantic_task_id":"4d636b981f86e1ff51c995837d4656aebc385fe96d7a482db06c16b07b375cc7","split":"eval"} +{"domain":"codeforge","full_message_tokens":51685,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":16,"sample_id":"bbf63f637b50261f7b64b88961b3aadb9f9fb8306a18c3953461154bb424459c","semantic_task_id":"632ba5a72664932caa92399a19b2a403b743ca2ae5f274f6130990d71adbb776","split":"eval"} +{"domain":"codeforge","full_message_tokens":51189,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":17,"sample_id":"245a295aa954359d2d0039a36952928f30144f0c8d8b8bd7b6534763a60b241b","semantic_task_id":"e152d07d531f664b3474f26731837216f57e113bc7cf04a4f9a19981d14aa85e","split":"eval"} +{"domain":"codeforge","full_message_tokens":51179,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":18,"sample_id":"1fdd920d54732de2af311d49f21e8f50eb7cea8639a4fadd89dbc7ea5c90e147","semantic_task_id":"aba8eb622c051de455a3c49ba1dc76239422ad85b74e8481958ffb340a757270","split":"eval"} +{"domain":"codeforge","full_message_tokens":55491,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":19,"sample_id":"456b430e0aaf69cccf1c6c399abe380209bd2b9254e8649b49904ee703ef9039","semantic_task_id":"ee9d1edc992c9a574bf2b486c25bbb46b3e93a75a52b186ff0d799f9d022e2d5","split":"eval"} +{"domain":"codeforge","full_message_tokens":55442,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":20,"sample_id":"7cf402683be1a05e19e624659f6a7e4748b96329cfca21ecd0c432c8d18b1ea0","semantic_task_id":"a2ebe69683650147a59ea465f75026385af83932c5625f3bdd44e505274df2bb","split":"eval"} +{"domain":"codeforge","full_message_tokens":55567,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":21,"sample_id":"4ae2b1ac899b13f65c4d0baec7134c4c13e8fc6b30bfbcd930671d61ef29785a","semantic_task_id":"801fae4cd43485869a9738eb6ed6ea851412d7633b18158a3bf5b8f992c1f4e7","split":"eval"} +{"domain":"codeforge","full_message_tokens":55503,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":22,"sample_id":"5eb31cff6a09a74de35e6c426ac0c97f04f3828222734faf245f7570c178c918","semantic_task_id":"1e3dd45e71a21c0373f4fdccedfdb0684e239ab811715e348101665ead5543d3","split":"eval"} +{"domain":"codeforge","full_message_tokens":93114,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":23,"sample_id":"2a995f3a31417942f7b1b193a7df005166dc504ab1857bea4cdb0f26a9ea2e51","semantic_task_id":"db235f36d379f469e9885816a324d388dcd740caa8c5a78fd910cc9b00a2a524","split":"eval"} +{"domain":"codeforge","full_message_tokens":92966,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":24,"sample_id":"2d2261ca69ba549bd160ecf8ffad66467892a968f2d3b20b211b18e8600efacc","semantic_task_id":"680b422f8a48024ee1a22fa56d24466a958db714580701137f5864c416e18dca","split":"eval"} +{"domain":"codeforge","full_message_tokens":93102,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":25,"sample_id":"494c4fca8eb0d6c23d596735469e44afbd2a4b10ddfa72b86759280e2350865d","semantic_task_id":"ec82d14c21d6bd20e4148ea56f80a5818995a916824a52a0bdf511a82c9e8b7e","split":"eval"} +{"domain":"codeforge","full_message_tokens":92993,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":26,"sample_id":"0ed7778afb6271206bf603f3293f406fb2f5d41171925a815b009108a6956754","semantic_task_id":"6c4790b30c00d13431edfc4f0afbda4cb984595b420eeae8e42fafb8fdcc8b0b","split":"eval"} +{"domain":"codeforge","full_message_tokens":83991,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":27,"sample_id":"815084969ba6b44f1f38dcef50c764cab9f0fe89cc8bf4151fa7a40c10e4aa74","semantic_task_id":"a9508e094298ebb5e1c65bdb1eea0f515d28906b461bd3a5ec34bdabcc92a36b","split":"eval"} +{"domain":"codeforge","full_message_tokens":83843,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":28,"sample_id":"b289cc00abe0be0c314ff9495d2edfc937a34070c7bbbcbc2d5ac097bace4e6a","semantic_task_id":"b26b1518c131ee34256a6f6f4915b1d711cb39c14e7e57f9a2f9050061129d45","split":"eval"} +{"domain":"codeforge","full_message_tokens":84383,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":29,"sample_id":"5f7999567de7c5343f90294e1db63b853c496fe88d93587a2bdc61e9b7713a09","semantic_task_id":"81316c840be5aea762e8189feae20e8936cb6690726c3fad8a48c5ff37ddd51e","split":"eval"} +{"domain":"codeforge","full_message_tokens":83870,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":30,"sample_id":"3fb1409978fb3d0ce18c73f152934fdd5d63bbf552976c787d21d12de60bde3d","semantic_task_id":"54c69182b0b0f11f8af0fb3ff5650c814cb246cdcc9b08d18421a8d09c1db70d","split":"eval"} +{"domain":"codeforge","full_message_tokens":139987,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":31,"sample_id":"af65c943180e6ecfddb0477695ca7d38f0b8bb3709fc17f8210e86445468dcdd","semantic_task_id":"d960e76f954a712bff8f51924076f3f8b4f91e79d5b6c72bc70c95427406d11b","split":"eval"} +{"domain":"codeforge","full_message_tokens":139862,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":32,"sample_id":"dfa28ff2cc32c0259fe716f1f798c3f0a2740f1dedf50433cb05ae5f255eadbb","semantic_task_id":"612f19843850a0343d1631282eac0736f8c0a4455f1084254eba941111163faa","split":"eval"} +{"domain":"codeforge","full_message_tokens":140385,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":33,"sample_id":"f1d2f1fbbb66e738c18689aef8108e25df6f546f196ff2eec14c4f37aba51d46","semantic_task_id":"e6572f7dedec63f1c76379a931dec2be133185bedc79c2d5ac1a0e894dec0152","split":"eval"} +{"domain":"codeforge","full_message_tokens":139889,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":34,"sample_id":"7096e0e6c2dfb90f53c4f86e681b219ae6ac9d1cf372fd761b99dfb5a8f05861","semantic_task_id":"e23c6276ffff24d597e33e35ce35fe62167a915614924dcb9f413b371c6dede4","split":"eval"} +{"domain":"codeforge","full_message_tokens":139952,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":35,"sample_id":"9f5b9d6d1333f9b5df4a0ff5228564522f588ec9a8bfa6f4fd1e290d1b60c916","semantic_task_id":"2352ce563c10a0a539dc811b26127790e24d3b4727bd769b1dd37c874247a44c","split":"eval"} +{"domain":"codeforge","full_message_tokens":113114,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":36,"sample_id":"171400144879b9929cec9d37fc1a20b41cd3ada6e6c5a6eab3e60a9aa841d649","semantic_task_id":"2f311d50d50739da2aec9735ae7d86ad66f21a6866b6d93527aaf319f8d7884e","split":"eval"} +{"domain":"codeforge","full_message_tokens":113026,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":37,"sample_id":"88ac8f523093db34d15e13ffac6b530aaaff4f81fbe35d6960003b5ad8ad2b71","semantic_task_id":"b6fdca76140c0bba09e1a9d307a72d256e26ff8ee76489bc4126fb8c4e3671a4","split":"eval"} +{"domain":"codeforge","full_message_tokens":113549,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":38,"sample_id":"bd0d210cf0a95faccda6578b787dab0ab68d3e2b2453f7f0a63d755a31359e3e","semantic_task_id":"8130b77687d82f3515f7483f338df58c69fa9f638ef67c00601b52dc60af2a13","split":"eval"} +{"domain":"codeforge","full_message_tokens":113075,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":39,"sample_id":"0f8a30a2b8290ae8f92a08bf679c1c707b7027eca28019174700ea0d7548b745","semantic_task_id":"a3e876b3cb9bf1368c1a29baa14a216d30a6112101b58d3fd7bb7e2e9441b19d","split":"eval"} +{"domain":"codeforge","full_message_tokens":113126,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":40,"sample_id":"1c7af03b8317a537642e3fb6804284a010b24b1bf4320396ca10d4a1ead5007b","semantic_task_id":"dfe51734676eb2a99a622b55f772c354697512ad59ebcd950e1cb3444aded896","split":"eval"} +{"domain":"codeforge","full_message_tokens":106184,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":41,"sample_id":"fe9a023920a31e93e713b50c5444cb2c5e4c18822935187d8e5f9e7b11499c65","semantic_task_id":"9990ce7fbe0ecb6f131bdb36a35b49192c08282c2e041f64839313d79c7c3895","split":"eval"} +{"domain":"codeforge","full_message_tokens":106072,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":42,"sample_id":"d517204fc97cf3a98fd6d0cd49c694bf78d6a819a3e34b5a3bb488d2ea756a15","semantic_task_id":"c56a98d75f50eab093de9ceb89bcdfe2b9ac3f621947cea3218d097b7e7c5309","split":"eval"} +{"domain":"codeforge","full_message_tokens":106595,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":43,"sample_id":"71930c115e2abb6d095b0d8c55236688d871706128c97ee75ebc40f3104b9508","semantic_task_id":"29e804d2442442e362febf47297d14d5afdc5e541a9135bede01ae32c149a435","split":"eval"} +{"domain":"codeforge","full_message_tokens":106121,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":44,"sample_id":"1582761219a27b993e27b5078a39dbcc7f621beed451904920f3023fef6853b3","semantic_task_id":"2c8540f678530e49dcf0029154d0fd9bc94febe38c9cecf92cf58c9bb8e24779","split":"eval"} +{"domain":"codeforge","full_message_tokens":106196,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":45,"sample_id":"0175936997d091b435b953e8d5a030a5b74edba54cad571c5f631bf3c66ab1ba","semantic_task_id":"e73cdceb06cc7c87f3ee314574c10b25bde1753bd159d8357073fc3080baab44","split":"eval"} +{"domain":"codeforge","full_message_tokens":86881,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":46,"sample_id":"7f9c766f9e806083eed3bddf54a9dd4d33a08144fa6361ba30fc3576ebe0d378","semantic_task_id":"ed4f9657938fe4616f5e4d170182b5f17630865094403608ec19b16c6e3cd7d4","split":"eval"} +{"domain":"codeforge","full_message_tokens":86822,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":47,"sample_id":"65058fee5a32fd0918cf3d560b0172d3e245eced601bdbd4d4c0f910447c768d","semantic_task_id":"ba91ce46700d2a1ff9ab2875098474843a919e7fb7944dd8645f3518f42e1322","split":"eval"} +{"domain":"codeforge","full_message_tokens":86947,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":48,"sample_id":"5b9c57d007242397044385ed73d96daaead8d63a4fba86202995dced2afddd60","semantic_task_id":"8ea0799e3a3ad1bb56b6afc8e6c705b3807134e8b9b1c67b4583950043584770","split":"eval"} +{"domain":"codeforge","full_message_tokens":86871,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":49,"sample_id":"f5ca287635fcf6f2c9713dc30eafd57676fdd86f97c306a7f9eaca063be6acc1","semantic_task_id":"3e5823bafa860d9c07718f642744c5f678c6d781424a23a85681aa78ebb5155b","split":"eval"} +{"domain":"codeforge","full_message_tokens":86893,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":50,"sample_id":"dc4d34f926167e62539d7e0f0de63dfeb9b5ee417890100f3a92304a3815ab1a","semantic_task_id":"bb1e267353270ee7dbfaa84011a4ac1fc5c0791e72933e45f26f88c400f8483d","split":"eval"} +{"domain":"codeforge","full_message_tokens":110839,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":51,"sample_id":"f255cac801fd67a2f8fbcc2119fd3467ad7c8ffa78b5710f75eea2a693d5b4aa","semantic_task_id":"e9036567c6e7a6799af0ab4bdab0c972a8f75745ba84d754df3956958584dd91","split":"eval"} +{"domain":"codeforge","full_message_tokens":110657,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":52,"sample_id":"06193bdfe91978b7763873924a407b41a8c31c4bc5b10b0ebea61b98b99fa0bb","semantic_task_id":"d720ad1fe5c7fb70b8be21d3619f3bb3ebe37ee6d74c91edab2d792334a3272d","split":"eval"} +{"domain":"codeforge","full_message_tokens":110782,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":53,"sample_id":"eab7fbafcbba8049f4d358ef870a04360fdfa4ddb14949abf54db73eea39b744","semantic_task_id":"f9341394a09aed201720f5457ebeb46ec896bba61146b8bbb5094e754bd5834f","split":"eval"} +{"domain":"codeforge","full_message_tokens":110718,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":54,"sample_id":"1aa98212fb106cd65270b09f76f3a04d22d3ddecdd6c4c91e1968baedaa7f7f1","semantic_task_id":"13fb49672d847de078831a810e8111d9082196fd6ac85002c3a2769bd58a8b2f","split":"eval"} +{"domain":"codeforge","full_message_tokens":119650,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":55,"sample_id":"256cc98c137ea2314fdd65aeb8795dce297ba4218bbc9cd0ba53194cd218937c","semantic_task_id":"27e0818f572eaf303f2bd4ed17da1ff524d58a91218deb6a85411df9b70bf9d1","split":"eval"} +{"domain":"codeforge","full_message_tokens":119463,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":56,"sample_id":"0b9094b54c75033d77d8375032817e3e329b63d09c07479482fbe6aa3c25e75f","semantic_task_id":"ef5cba8f6c78674104ef9c69b2f0b90dbb26481dc689acb4322f2d2ca3d24c9b","split":"eval"} +{"domain":"codeforge","full_message_tokens":119599,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":57,"sample_id":"187d9a99f32b370ca1d50802dce9dea6cab83bdb486fca9b6a3ecc762735d316","semantic_task_id":"a17f65f5eae8e0b6613d6516a8b363a6720fea9966ec0ef5b1e7e20ee77de756","split":"eval"} +{"domain":"codeforge","full_message_tokens":119529,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":58,"sample_id":"fdb318dd9e33a6c628bee502991ffc12d96c968f5ea7dbe3a76226cfef5093e6","semantic_task_id":"266969894e44d05345914999aad577d1fb6015e9908baabaeb4813815f925273","split":"eval"} +{"domain":"codeforge","full_message_tokens":171342,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":59,"sample_id":"31af7c97a216f2ed8bfc3401c3133df0717508186831eabfb5bcf27bea33ddf1","semantic_task_id":"e9863565e025481216a55e5135937355a96af6bed6074f4ea7a9eda65dfbe474","split":"eval"} +{"domain":"codeforge","full_message_tokens":171207,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":60,"sample_id":"c186a224c4cd5362f0775f94e2a2095a3f44d659db58dd7b6f7283c64bda1ee6","semantic_task_id":"29bfebb79304d15558e1ac49cdcf1c8c9d685efde3f25e6b1efcb811eacfe9d5","split":"eval"} +{"domain":"codeforge","full_message_tokens":171730,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":61,"sample_id":"7f9e23669a71ab6fd1ac10da12ec561745db49986109a3c6662cd70b91f56974","semantic_task_id":"e8c59e822aaa0e82f576148cb6e3273df7360cb4289bb29a6707fa371133678b","split":"eval"} +{"domain":"codeforge","full_message_tokens":171270,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":62,"sample_id":"0988eeae9a66815c3e395fe9bc04cfc41e77c2dbf36a3215848590387f196357","semantic_task_id":"5ecd7ffac34482f75264d8d097c0cce5d4bcb6d3984c4c5a0393b33c03c95794","split":"eval"} +{"domain":"codeforge","full_message_tokens":171307,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":63,"sample_id":"51bb6e300b4496dc673b3549220b41e5ebcae005fed668d27f2cd00891708185","semantic_task_id":"8922f53934eb472ebfcca2cefc222464f98fffcd335ad88518045601c0e79c58","split":"eval"} +{"domain":"codeforge","full_message_tokens":130872,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":64,"sample_id":"cc4deee9c6aa0845df2293d1699b4a86e387b49ec2cd6c1330050e3265d16b2e","semantic_task_id":"bb1b22f99f01ddae1c347ae887e3fa5741d6d77337bd2e5f0c5202d4d2103279","split":"eval"} +{"domain":"codeforge","full_message_tokens":130750,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":65,"sample_id":"b19957c3cb9a1ea66d50289e59303094c2707c58ac5a4b3373c06f666d291058","semantic_task_id":"d6e462b7638a76e52cd6f1f77b008ef92cbc74e7a7631ad04fa6b54bdfd89db2","split":"eval"} +{"domain":"codeforge","full_message_tokens":131273,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":66,"sample_id":"90ed24847fac2ee6871eb4915bab32ec417c9b73cecae45b7a3d9c2ebbf1490e","semantic_task_id":"de99f1abee01765c7788035dfd3a82e8088c1669ce700333f694dc66ab061623","split":"eval"} +{"domain":"codeforge","full_message_tokens":130806,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":67,"sample_id":"0081361172d88014655f60c131cc2b984edf3dcfa9f6b4f8173ebdae92c55d2f","semantic_task_id":"80381bd9aeb095b26afd8dbd95b07a2d45f4284e22d1e8e3dea6cc556216f57e","split":"eval"} +{"domain":"codeforge","full_message_tokens":130884,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":68,"sample_id":"a4854f94f5ba8da745fbe72a16d1f1c3c37d495618ec47af1860b99f81a6a4eb","semantic_task_id":"a115adfd330765e30473354fd33d94f5724cb43ce7a6f82e1001c3c6f83012c7","split":"eval"} +{"domain":"codeforge","full_message_tokens":141898,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":69,"sample_id":"e844d27477ed8a8249891fbaa45943f069afd971324e375c75219e91f5bf9277","semantic_task_id":"1ff537332320257651d9e2b1c32137097dfe5b14e5a144d978dc2b8cdbf89f4b","split":"eval"} +{"domain":"codeforge","full_message_tokens":141747,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":70,"sample_id":"be1f17997d8b87f24fd0d5abf46ca1d93d20faf2de5bf470b44a9f9ac5359f04","semantic_task_id":"97b635da361400e3d1da789cd3fe8edb909d96c91b5c64fc9b6d63f45f2d2b57","split":"eval"} +{"domain":"codeforge","full_message_tokens":141872,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":71,"sample_id":"df75650877655f334963f4200a3d10554eeb9d512774a6c101e5d4b95b90f76b","semantic_task_id":"26bf90b25b63c2de7b55b5f4781c0ed2f98020f32e57d29856a53744137cd357","split":"eval"} +{"domain":"codeforge","full_message_tokens":141803,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":72,"sample_id":"d2fc8daa1a8350e89a26f476d61d235ff20e18c97a1d6fdb97667fe5608b41b6","semantic_task_id":"9d0f0fbf9cdea02ee8f748c2cabfa8b65054890926d989362e4d392952d915d9","split":"eval"} +{"domain":"codeforge","full_message_tokens":141910,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":73,"sample_id":"8181e9e0d942db75c543bc34e676544693ea2ec1f485c1d4033d4aa6d5bf58b0","semantic_task_id":"dcc71c0a15316e1a0b987e6632c5b2b100eb788a21e6ea3e9c4de71a9e4117e9","split":"eval"} +{"domain":"codeforge","full_message_tokens":142229,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":74,"sample_id":"8b8c4e05bca5f32660054ad82c3a71751c66c97f4b0ea3e5e02f95cc175f7784","semantic_task_id":"6d619aba344dc5c25fa3929adfa31c86c9fb8891c589f8181aec6433e7202633","split":"eval"} +{"domain":"codeforge","full_message_tokens":142037,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":75,"sample_id":"d53baca1aefcbf7aaf43cf2e75bb755922d3fed1958022dabf05f3d721db948f","semantic_task_id":"00497afd9c2b48984c4230d94dda8da80532ab0ef91bacdbb838a837052e2343","split":"eval"} +{"domain":"codeforge","full_message_tokens":142162,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":76,"sample_id":"444d77054dd0a1ab38de83010178bfc99a0261a539299fc5e8ae8bddb6b6f58e","semantic_task_id":"35091e003a8cd2df78615655feb2eae991076fb58c3942d530ddb9d456c09edd","split":"eval"} +{"domain":"codeforge","full_message_tokens":142093,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":77,"sample_id":"e834140682491ea3d84d85c2fcffea0aa88b0bc396259416f27c803db7e74548","semantic_task_id":"fbd1725c2d08d5abbfcbb930a19b3945545b9c3b7fb4e1473a94651e2a42c75c","split":"eval"} +{"domain":"codeforge","full_message_tokens":142108,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":78,"sample_id":"9e152e29ec70450b30c727dcdf49ed13869118e4424bdb01fe3618894da0fe72","semantic_task_id":"219bc06f7bf2817515b2cb1fea312344b5379d1a28cc0dcab1fefe2d36da6aae","split":"eval"} +{"domain":"codeforge","full_message_tokens":137375,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":79,"sample_id":"c7819c7db8ac056e55bbe05c82722c21c2142abd51948c52cba46e4635eaa90c","semantic_task_id":"b4112b90c495b6c335c0b180d415b5eb521b58fa22c11041b14f80a037c812e2","split":"eval"} +{"domain":"codeforge","full_message_tokens":137154,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":80,"sample_id":"e00cbe245a77111999fa035b8b19c75d3cc85d1ad0efd2214f077c2ebcc2c3c6","semantic_task_id":"0be98d708d169dcfb270c3743170644f09f635da8eb3db9ccd1ce942ebe16b5e","split":"eval"} +{"domain":"codeforge","full_message_tokens":137279,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":81,"sample_id":"9f7f43ff2f4646f50acf2915f67a1c0a94101eab12f633fd250bc917af8bbcd0","semantic_task_id":"a2f1ffa9a07c41a717ea2a43d983eead537a5332cb8f1e6018f0a5e70d968ecf","split":"eval"} +{"domain":"codeforge","full_message_tokens":137254,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":82,"sample_id":"dd2b23ad12223512d8b740f1181ddd25180c2ad9a426f4e513f089372dc49119","semantic_task_id":"d60c53295acefacbf3257834905f71fdfe3a7a187750c204afe8730174de16d9","split":"eval"} +{"domain":"codeforge","full_message_tokens":224814,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":83,"sample_id":"5ab90e5a56330d1cd2650cf4fc2b57f7dcf05444daeae50b383cdcd1cc9055b0","semantic_task_id":"a169b4e9124e494ab5059f899908722a60f887f229257a9f4c5527a14d4907f4","split":"eval"} +{"domain":"codeforge","full_message_tokens":224606,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":84,"sample_id":"2cb1599aeb43d3e43e1bcac952b1a5b5128d0c3f8549bd631a3740dca85d5995","semantic_task_id":"b8c7bdc781af415087305ae83fbdbe607de94d77cc6e3a6c482fc69b0bfb1eb1","split":"eval"} +{"domain":"codeforge","full_message_tokens":224731,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":85,"sample_id":"4b1435280f2ad11f5b2115881c17dd3bf02524456dcb073d03fc2ec8c02008d5","semantic_task_id":"43b8ce5775b1c5a7da6153ac666652f168f586dd7477dc3647ba4b62d138d509","split":"eval"} +{"domain":"codeforge","full_message_tokens":224683,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":86,"sample_id":"08a728f6ba2d298f897c2d6d068f55dfdfcbc9d640ed4a1234099ba520a4c28a","semantic_task_id":"13f485b83323c0087dd62b9d357fe33bae447a0b51caf6e979496f8fbd5452f6","split":"eval"} +{"domain":"codeforge","full_message_tokens":224779,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":87,"sample_id":"85c40c3e365905a20789b9f708b9e463ff9210960b8194d30c90b7b166912380","semantic_task_id":"8d7f3d9e6f9d25ccc5b1a4cb70e0db80192caea2aadc6ffc1eb295fbd89dbb20","split":"eval"} +{"domain":"codeforge","full_message_tokens":221934,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":88,"sample_id":"585070aec685bcb2e78d6a87b9d1a2636345965a90a13f2ff7241adcf960e0ca","semantic_task_id":"b0ade45e89a8a7aeabf61678eb8e285d601accdeb6474ea29c565eee97d3ead5","split":"eval"} +{"domain":"codeforge","full_message_tokens":221640,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":89,"sample_id":"f3ba60c51fd780b65d2a5e1d14e70293d22bd7f814b626ec558f9aa09e816800","semantic_task_id":"dc1d387275e0cc6c5c52ab0bc1146a822b19f7193d8bea0c5191d2d835d0f9df","split":"eval"} +{"domain":"codeforge","full_message_tokens":221765,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":90,"sample_id":"b79976557dd5729350375f98368f6177d856d5dc6bdb9004a89ce168b2180f9a","semantic_task_id":"b85f5665dbdebffc283e7a51c88c94371166f38e987743ca7a8bd56361d0a792","split":"eval"} +{"domain":"codeforge","full_message_tokens":221710,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":91,"sample_id":"9fd4126c99f532e1c89ae5c049d85ae02a7346922b6202c1078beaec1f4da0f9","semantic_task_id":"fdd666a95b21675be612cd202c8d3d8abd551ee756c99eafcd338da32ff1bf6c","split":"eval"} +{"domain":"codeforge","full_message_tokens":221813,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":92,"sample_id":"43e573e2996942232682b1616cb912d1200c3758e8b170f8128b42c2afa46f67","semantic_task_id":"116b7c4dbc79ceab85fe2b642a52d83ced27eff0ea3be7f98c5592901e708eef","split":"eval"} +{"domain":"codeforge","full_message_tokens":223760,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":93,"sample_id":"7d0d8dd007768b3a451a0f6aa4fa77bb06234fe690e929eab68407c6bcd730ab","semantic_task_id":"f6dc9b91d11509a0622dd4593a3dda34aa1d4c14bc91159d14f8e980963a14c4","split":"eval"} +{"domain":"codeforge","full_message_tokens":223459,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":94,"sample_id":"b1f107bb1ea8b46b6c39190d43666ba0f276791d26ef1a6386d36d2672368bb4","semantic_task_id":"73548ccb34016f007678fa2c9582258b86d9e650183bc413ea3bd8784da284c5","split":"eval"} +{"domain":"codeforge","full_message_tokens":223584,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":95,"sample_id":"da64cb65b90538b33ed25950090b7ab6d676f34ad7c6f7f8cdbb3c7ae23b1183","semantic_task_id":"2de3bedf03b3cfcabe9378d879ea0c91970597349964bfba316c15ca0eec7eeb","split":"eval"} +{"domain":"codeforge","full_message_tokens":223529,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":96,"sample_id":"61804fc4e83a4ca20ca926aed316190f21a73710d540b7a44cb96afc2b41842c","semantic_task_id":"47137d28396988f59a9b389d31e1e86fd7e716532f5870dbce7adec2b2890429","split":"eval"} +{"domain":"codeforge","full_message_tokens":223639,"group_id":"https://github.com/astral-sh/uv","output_file":"codeforge_eval.jsonl","row_index":97,"sample_id":"4db0b9ffae0b95e310b4f23d79cb353a77e9266dd5024223a97cb08707954afd","semantic_task_id":"fe566306ebe484f105209367b5abf00a4aa37890d4723f76489b9fe91113357c","split":"eval"} +{"domain":"codeforge","full_message_tokens":247496,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":98,"sample_id":"f88f813c987b41efdae0350b3d999a0c4a6895dea5ffba4ab65c8683238b4f09","semantic_task_id":"c27b2f6033f0481dc90c13e8e20ad5f248c14af384b35b88b35d1901bf098df2","split":"eval"} +{"domain":"codeforge","full_message_tokens":247036,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":99,"sample_id":"4e4a7e2824e05acb376ca94e20f5da234a52e74c7eaedd0a7adb118a94387468","semantic_task_id":"73052c732d07350c3a51509d25130e85d8ce3b451cfc8d744fbda9b0d176e795","split":"eval"} +{"domain":"codeforge","full_message_tokens":247049,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":100,"sample_id":"0e1274a250da8485813a351e08805d4844d1d4829e1b3fd2f8af838d1ad458af","semantic_task_id":"ada5632489a376e2106736eb91b9e54c70f8ce54737d639a099ea56d1c98ed1a","split":"eval"} +{"domain":"codeforge","full_message_tokens":247470,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":101,"sample_id":"19ccf6aa8d506701a6e4c5bd1cfdb39b8036359d496f9110900886fa2fc948fb","semantic_task_id":"e38f19b8eddecd947756ec6af99c4d6419d4f4c0daa26ca9ce15708eba788022","split":"eval"} +{"domain":"codeforge","full_message_tokens":247065,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":102,"sample_id":"2633d286c9edfa39ed106a3dbb03fa40c2daa42a304dc53dc268484c88a1b255","semantic_task_id":"7312a8cc5e14f84c6cb42f36d6ca5113025312bb3b170f3e5bbace888f983519","split":"eval"} +{"domain":"codeforge","full_message_tokens":247508,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":103,"sample_id":"899ffbc30289c5c0e2b0f089eb0249b7f37cfffef944c951b3d9b7dc14cac46d","semantic_task_id":"25371a4c08f771f5f865073822e8d6c4d2c440d020f34d88b78195df82c9e405","split":"eval"} +{"domain":"codeforge","full_message_tokens":97528,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":104,"sample_id":"7252d306b548b7f7ab4acd95261b0c71bc3ded5f0af8a88e9c32d1b89790aeaf","semantic_task_id":"488ce58f3529d46dc478463f6eb6caffc0c0b8fcb88f3e62691344a8d60b47bf","split":"eval"} +{"domain":"codeforge","full_message_tokens":97493,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":105,"sample_id":"bcdfd4d3ba66eb8c71c2aeb17e053658e6ca16eae2b7e02a2c780e2b24f970b2","semantic_task_id":"10f1c8d9480c0af04aacd56833d672b4c2285add1ca3247b6c5748daa44b80c9","split":"eval"} +{"domain":"codeforge","full_message_tokens":97915,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":106,"sample_id":"277a8fe43e3f22752c09708798a07e600c948c553d1d9edb3cfee6d1e896c6b5","semantic_task_id":"101d28439d1a885b0218908704ab9cb5e160069d512d85535d132dd0a7f5d326","split":"eval"} +{"domain":"codeforge","full_message_tokens":97510,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":107,"sample_id":"f6116db41cbe4b8b3f6ba7d96295feef0d0959b259dd0880fb5f25d6eba90a06","semantic_task_id":"4ba3d5222f8b9522e0b85bf2f7ceaf477349440ebd8c9e66f711e7ff8eea95ee","split":"eval"} +{"domain":"codeforge","full_message_tokens":97540,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":108,"sample_id":"63e81ef20aa29e052b98f248bd0b50ff42b5d8322714662b97ac9125c5eb70c2","semantic_task_id":"e312653e746bc3de37aa04ad4454a5e513ac7a9a019c652e4eb3016d600d2633","split":"eval"} +{"domain":"codeforge","full_message_tokens":46017,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":109,"sample_id":"256d3168f8002daec7575907d701d03cbd895f65aa758fb377cea0c235ba5560","semantic_task_id":"4461ce05da95da60a7e161b104b616ad903525c60d1a2523d6ed8fc63dcbefa8","split":"eval"} +{"domain":"codeforge","full_message_tokens":45836,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":110,"sample_id":"1de470786f31496698ed58280e4b2d3a760ca5dd324a0b76b19bedc2b6325a5a","semantic_task_id":"ad9ab6ad8ff9afc54aa06123d9d3491eef8859d6b3d2a6fdedb97e4bf21e751a","split":"eval"} +{"domain":"codeforge","full_message_tokens":45854,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":111,"sample_id":"4a54ee47d65de2facc1247890416ccf4ec2eddf38ac482ed25b0eea1cefb7b91","semantic_task_id":"57afdcd3faaf40c4530183d18e7068da0e5bfddddde2b9df8c98ba44c4d0358b","split":"eval"} +{"domain":"codeforge","full_message_tokens":46180,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":112,"sample_id":"993019c6a129142368f4292a5725b85b61449d4a6dd06ad273736abd77e8a7b1","semantic_task_id":"2102c3d1b65e8a9756a22ae3b1e7ddbf43757c26d83a281ab5466ea888486193","split":"eval"} +{"domain":"codeforge","full_message_tokens":46008,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":113,"sample_id":"95e57845ef0a525723b93a05911cfe5dbcfd5fdb74e4012f2be81127c85414db","semantic_task_id":"e88a28f7c6decd36c78d1f0ed136fda05b4d53b2cf093e5ac7f1757181836c1d","split":"eval"} +{"domain":"codeforge","full_message_tokens":47751,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":114,"sample_id":"dc5c5cb59de66f8e09a436d57a9dd7c0162d3a8f09ab893f8b679d85cd2b2aba","semantic_task_id":"ab4d958514e10714f56a7eee11164d76f8ec4ff64d199d99b02f9998e645362a","split":"eval"} +{"domain":"codeforge","full_message_tokens":47547,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":115,"sample_id":"18780602eb2a06965d066f7eb97f28134ae9d5b67c3ab95a4edc9292d058bc44","semantic_task_id":"6551e8eddd4fa68dbcd837fd086c9ad19bf3bc48e056b5027dfa6a21550c632a","split":"eval"} +{"domain":"codeforge","full_message_tokens":47565,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":116,"sample_id":"c27159792dd271865366ddfbab99c7401c6ab41b728f974599c7e3c48c055cd1","semantic_task_id":"10842b7fef3006e0ed52dd9380de058eeaec3871afa303902d097f83dfc68462","split":"eval"} +{"domain":"codeforge","full_message_tokens":47901,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":117,"sample_id":"82351103bf105370cdd3dea12321e89e30836a8e42ff8c925e1bdee9e72925a6","semantic_task_id":"c3e7ca1c2de3a5d62b5d1d9c48c431bbb955e560b645f873553027f17756864c","split":"eval"} +{"domain":"codeforge","full_message_tokens":47763,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":118,"sample_id":"262a00016dd27fb106a3507e72e21f06e6bbef50346d81c00136a5a680267e2d","semantic_task_id":"d68716764d4eafa3997c2e3ff417ef943c0d8d2fdb6e21dc9e07079665a267d8","split":"eval"} +{"domain":"codeforge","full_message_tokens":258555,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":119,"sample_id":"d689691a1334a81d9dcc8f99f3514a2daa98b9acac825fe6f896df7a33e105f6","semantic_task_id":"29e83bc3ae96e6e21ecf43454c5b270f99db09f378158404b9e7f48bc16f6071","split":"eval"} +{"domain":"codeforge","full_message_tokens":258097,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":120,"sample_id":"4b80e755cc85928c2114cfe75b882d36091677fe35e8ac4b4ca1424cc9f3f1ef","semantic_task_id":"4e9a73be8ad459f47dfe3856c23765da816ae908fa86d8c8a7232cf7eaacb8fc","split":"eval"} +{"domain":"codeforge","full_message_tokens":258518,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":121,"sample_id":"5dd4225eef1abd95681f325cf1e1d1732be888731e8182a0148db0f6733bbdf0","semantic_task_id":"2a3248d79cef978f9f35afe65a80de11a92fb0b34eb5a9d712f8c724607e58bc","split":"eval"} +{"domain":"codeforge","full_message_tokens":258118,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":122,"sample_id":"359fd3b4feb61004bb5d2c8084aed2d533644f432ba0d98e9d7d8c577e10642b","semantic_task_id":"7969f58de6b4a4a22d4297055007a4e1847b757c97b2cd1826155dd28f789eb5","split":"eval"} +{"domain":"codeforge","full_message_tokens":258567,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":123,"sample_id":"69258b5503cb9fdd5741c6f814fb145ddd4854a5538497b2a62c311bc64a0896","semantic_task_id":"68e9c036e77036ef3cfe113b51cd79b1dbeee640509d84d074c11896c95991dd","split":"eval"} +{"domain":"codeforge","full_message_tokens":108507,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":124,"sample_id":"e9765c7cfa12e717e0efd3c3fbcca776240b4ef4fa5de1d013dd93062c719ff5","semantic_task_id":"88732e80f718645bbb631157b5113f0dda94fd97acb069f3044d8fb0ad3d45c9","split":"eval"} +{"domain":"codeforge","full_message_tokens":108447,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":125,"sample_id":"cc14e8ce073c9a3d41cd104d107b5f7297261ef094caa27d50f6a4a79838555a","semantic_task_id":"f45e3316f842498016264243865f2e20554e648b8bd8b4f554cfbe9bc5d59a06","split":"eval"} +{"domain":"codeforge","full_message_tokens":108779,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":126,"sample_id":"f8a6f7a403db89c178be07112adb48a13706899ddfd58930903c92f380e38dae","semantic_task_id":"7271c684e89171d8265c49c95c937817f6d96857c113b70f47b4a45ffbb54eaa","split":"eval"} +{"domain":"codeforge","full_message_tokens":108469,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":127,"sample_id":"74425533f07feb3227b7eca5daf7bd72c92186459f50fe5846152ccaf8af3b48","semantic_task_id":"1df8dd2ef519f1fc08e0b318b51cb53cbb20b9b6b28e500fd0e85cea97967565","split":"eval"} +{"domain":"codeforge","full_message_tokens":108494,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":128,"sample_id":"612a60b9b9b0aef35e2438c16d3b87135dd6bbc3fafda14d6d1155840ab3bc65","semantic_task_id":"94a6bee3f83fe90d1a3ca9f6acb1ef82fd307b5042e846d3816555e66d91f7e5","split":"eval"} +{"domain":"codeforge","full_message_tokens":57011,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":129,"sample_id":"9cfb8d9d01a4bb914d3e397c17a9ee138046abab980e6cd358a602b054cd945c","semantic_task_id":"73252f455dcefdf1979465501e262e372646ad99b58b3ac5dd08ce156dd636ef","split":"eval"} +{"domain":"codeforge","full_message_tokens":56837,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":130,"sample_id":"d5e673db99dc9a5a3c385ce8c03ea691a839f8fe1fe687dc1732981582d5fe50","semantic_task_id":"69690a3e3806682d7a34e6878c7a43988ee02cf1d291209342d7319c88e46a2f","split":"eval"} +{"domain":"codeforge","full_message_tokens":57163,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":131,"sample_id":"9a024503ac5a2c1ec953dc2687b24eef178e6b9c462411baee2171696a23b95d","semantic_task_id":"ab93f78a6d92744effd41310209bd0f78231f6da5d57d3c5e8832421641f923b","split":"eval"} +{"domain":"codeforge","full_message_tokens":57002,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":132,"sample_id":"75579ee2fafea0868a6a537101b366fc6171d54877a31e2a0e832cfe04fd4786","semantic_task_id":"468840544862a813c61474c47cfe407504818053f14f69ba05e8b23b87920212","split":"eval"} +{"domain":"codeforge","full_message_tokens":58719,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":133,"sample_id":"262efc622f66c4f93cf30f4c131e30cce347af24aeadec36821da6d043225a52","semantic_task_id":"28bce25979e092ab0acc8ca5a25ac3338ee54432156846cce1d034384f9af3c9","split":"eval"} +{"domain":"codeforge","full_message_tokens":58494,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":134,"sample_id":"772641ffba5760eae2621c757edb783bdf68409ff96aa93a0e7ced9fd912e193","semantic_task_id":"b036c3559f8c379fba8792d1fb2192b3fe8f9fd0418aebc23b31b4245237e070","split":"eval"} +{"domain":"codeforge","full_message_tokens":58512,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":135,"sample_id":"219eb8daae53c70a6d4af4736da9417d0846c52e23f01cae8c593a5ee8812e36","semantic_task_id":"4e854a5af81b1c90ae1b2d3c2a80d1edc9a606c30690411b86fd6701df9e0c86","split":"eval"} +{"domain":"codeforge","full_message_tokens":58838,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":136,"sample_id":"7f9c33b4614069d28aa19535f3869d3b993cf5b9434071c2bd1f3bedcc1be2d2","semantic_task_id":"ef92bbcf67046f0e556ffcfeca313c985260bdaef96441042c676c3f737efd77","split":"eval"} +{"domain":"codeforge","full_message_tokens":58710,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":137,"sample_id":"2698ab01eb47b4a3461da235995f2017fabc4fb765808a8e1f60b55d8e36795b","semantic_task_id":"ff4051a3dc32bc1052a00d08122ce239da6c279e8b3497fa583ec931d3e42d24","split":"eval"} +{"domain":"codeforge","full_message_tokens":143474,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":138,"sample_id":"98a0cc6c42979f4a746596485cba466f408a2853e16a13ed959dba1c79df479f","semantic_task_id":"e64310f8bd3ce0697eeef349afb90b556287415d8456f5bcd35e279416bfe17c","split":"eval"} +{"domain":"codeforge","full_message_tokens":143254,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":139,"sample_id":"3acbad729a792ef11c5d2e5219e82817408384beef021ca1bb51c87a99c0cf26","semantic_task_id":"957289995a4796c6041966b1d72bd3077760787c2a22a0b31da1379ce03b1813","split":"eval"} +{"domain":"codeforge","full_message_tokens":143580,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":140,"sample_id":"7ca87cabb26f4f88ad853fe830bd310dd4ed55c3a24ba56372cfe4da31caaa0f","semantic_task_id":"f051156f7a92faf1613adc2b0f25b601c40c08c7ff1a47deabfef012c6366fd9","split":"eval"} +{"domain":"codeforge","full_message_tokens":143280,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":141,"sample_id":"b6a280cc3e8e3edfee01163c319adefc7bda520a3853b87265d9ed1183d8e4fd","semantic_task_id":"8508c83fc6d67894b3afb3350b4f2ee5ca0659360bb55ee320bbce969c25fc93","split":"eval"} +{"domain":"codeforge","full_message_tokens":143465,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":142,"sample_id":"c4dbe1ceb290d3ba5555f95e039bebf5ce57db04b6c15cc5178e4ceae30379a5","semantic_task_id":"96e9cc9cad5ebe5fbc308ac0ed13482b783cee452910f45cfa97157e9969eb18","split":"eval"} +{"domain":"codeforge","full_message_tokens":69778,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":143,"sample_id":"f94ca2316ccc67c59cc3557099712a92d65ffc08c5a0f4f0143334a02da80f42","semantic_task_id":"2b70251a488215e079ed985b08a27c1f4963ac715a2ffcc093c0c7006c8865d7","split":"eval"} +{"domain":"codeforge","full_message_tokens":69560,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":144,"sample_id":"4830c6f5e39aea4ead9d89f3468486bf97ddeb4adc518378aa3cf30ef1fa6db5","semantic_task_id":"df37bfa1e2003409bbd8a6330953b255d5e482536311472bb8cdadf1674ee8d1","split":"eval"} +{"domain":"codeforge","full_message_tokens":69886,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":145,"sample_id":"7efc88de77dd7b08c2aae5134fa1e8822f81ace90e20933b5336d8e867e4c091","semantic_task_id":"72682cab54281a0e8b04086b65ede05a972ffa0377f917d67d29691c675e935c","split":"eval"} +{"domain":"codeforge","full_message_tokens":69769,"group_id":"https://github.com/opensearch-project/opensearch-php","output_file":"codeforge_eval.jsonl","row_index":146,"sample_id":"93836e5491a1084555c4537617be5132ebc8d4343750c64f56c41182a954c461","semantic_task_id":"6fdb726c71911c7537b9977f71890399d0ba39bfcfd3211382965d025dc50f4c","split":"eval"} diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/alphabet.stderr.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/alphabet.stderr.txt new file mode 100644 index 0000000000000000000000000000000000000000..f9a22f13d3c2bd30b5099838db584d07c7e0249c --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/alphabet.stderr.txt @@ -0,0 +1 @@ +[transformers] Token indices sequence length is longer than the specified maximum sequence length for this model (313596 > 262144). Running this sequence through the model will result in indexing errors diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/alphabet.stdout.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/alphabet.stdout.txt new file mode 100644 index 0000000000000000000000000000000000000000..29a36bc653444b719cbf483b8d439a2be2c87069 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/alphabet.stdout.txt @@ -0,0 +1,10 @@ +{ + "status": "PASS", + "semantic_tasks": 200, + "capacity_bins": { + "131072": 139, + "262144": 61 + }, + "exact_token_ranges": {}, + "production_eligible": false +} diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/amazon.stderr.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/amazon.stderr.txt new file mode 100644 index 0000000000000000000000000000000000000000..ca8dcb38124ec3b5ec07846b0418144c79f814ee --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/amazon.stderr.txt @@ -0,0 +1 @@ +[transformers] Token indices sequence length is longer than the specified maximum sequence length for this model (299710 > 262144). Running this sequence through the model will result in indexing errors diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/amazon.stdout.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/amazon.stdout.txt new file mode 100644 index 0000000000000000000000000000000000000000..5b72ef660d09edd488a11f12d049ea7deacda833 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/amazon.stdout.txt @@ -0,0 +1,10 @@ +{ + "status": "PASS", + "semantic_tasks": 249, + "capacity_bins": { + "131072": 163, + "262144": 86 + }, + "exact_token_ranges": {}, + "production_eligible": false +} diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/amd.stderr.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/amd.stderr.txt new file mode 100644 index 0000000000000000000000000000000000000000..542316cc980f4c7143eff86aa8b426473d1e931c --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/amd.stderr.txt @@ -0,0 +1 @@ +[transformers] Token indices sequence length is longer than the specified maximum sequence length for this model (265133 > 262144). Running this sequence through the model will result in indexing errors diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/amd.stdout.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/amd.stdout.txt new file mode 100644 index 0000000000000000000000000000000000000000..6bdc75499c6afbeb41bdc9606c790090fdaed179 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/amd.stdout.txt @@ -0,0 +1,12 @@ +{ + "status": "PASS", + "semantic_tasks": 194, + "capacity_bins": { + "131072": 124, + "262144": 70 + }, + "exact_token_ranges": { + "262144": 47 + }, + "production_eligible": false +} diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/codeforge.stderr.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/codeforge.stderr.txt new file mode 100644 index 0000000000000000000000000000000000000000..18acb7570dd74514418d969bf17c92601ed59fae --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/codeforge.stderr.txt @@ -0,0 +1 @@ +[transformers] Token indices sequence length is longer than the specified maximum sequence length for this model (263244 > 262144). Running this sequence through the model will result in indexing errors diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/codeforge.stdout.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/codeforge.stdout.txt new file mode 100644 index 0000000000000000000000000000000000000000..a5ce7a6e767092eda23b8b729e8797e4acda3667 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/codeforge.stdout.txt @@ -0,0 +1,26 @@ +{ + "status": "PASS", + "counts": { + "train": 657, + "eval": 147, + "short": 13, + "rejected_overflow": 1 + }, + "canonical_source_tasks": 818, + "primary_long_samples": 804, + "full_message_capacity_counts": { + "cap64k": 141, + "cap128k": 305, + "cap256k": 358 + }, + "exact_context_token_range_sample_counts": { + "other_long": 767, + "64k": 12, + "128k": 9, + "256k": 16 + }, + "max_full_message_tokens": 261954, + "max_observed_full_message_tokens": 263244, + "production_eligible": false, + "local_training_eligible": false +} diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/intel.stderr.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/intel.stderr.txt new file mode 100644 index 0000000000000000000000000000000000000000..2f29bfde17acd4977ee4ac771d34c48446d77790 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/intel.stderr.txt @@ -0,0 +1 @@ +[transformers] Token indices sequence length is longer than the specified maximum sequence length for this model (332115 > 262144). Running this sequence through the model will result in indexing errors diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/intel.stdout.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/intel.stdout.txt new file mode 100644 index 0000000000000000000000000000000000000000..4dc9a8edaebc362a8151b7414ab68069aa2a5ca1 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/intel.stdout.txt @@ -0,0 +1,10 @@ +{ + "status": "PASS", + "semantic_tasks": 199, + "capacity_bins": { + "131072": 128, + "262144": 71 + }, + "exact_token_ranges": {}, + "production_eligible": false +} diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/meta.stderr.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/meta.stderr.txt new file mode 100644 index 0000000000000000000000000000000000000000..c28207414a5843ef4099222b26c5ea232f57b549 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/meta.stderr.txt @@ -0,0 +1 @@ +[transformers] Token indices sequence length is longer than the specified maximum sequence length for this model (280868 > 262144). Running this sequence through the model will result in indexing errors diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/meta.stdout.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/meta.stdout.txt new file mode 100644 index 0000000000000000000000000000000000000000..c4440e0bc4b29fdebc647a72ba491ce01837c239 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/meta.stdout.txt @@ -0,0 +1,10 @@ +{ + "status": "PASS", + "semantic_tasks": 216, + "capacity_bins": { + "131072": 145, + "262144": 71 + }, + "exact_token_ranges": {}, + "production_eligible": false +} diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/micron.stderr.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/micron.stderr.txt new file mode 100644 index 0000000000000000000000000000000000000000..7c8d976b3a5b0668058b94d837bffcc32db7929a --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/micron.stderr.txt @@ -0,0 +1 @@ +[transformers] Token indices sequence length is longer than the specified maximum sequence length for this model (283529 > 262144). Running this sequence through the model will result in indexing errors diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/micron.stdout.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/micron.stdout.txt new file mode 100644 index 0000000000000000000000000000000000000000..fe112b1d62b7aa62e8ca5dfe42d3655748c1d8c9 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/micron.stdout.txt @@ -0,0 +1,10 @@ +{ + "status": "PASS", + "semantic_tasks": 243, + "capacity_bins": { + "131072": 149, + "262144": 94 + }, + "exact_token_ranges": {}, + "production_eligible": false +} diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/nvidia.stderr.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/nvidia.stderr.txt new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/nvidia.stdout.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/nvidia.stdout.txt new file mode 100644 index 0000000000000000000000000000000000000000..df02404ae0cfb9cb187c09d89d9c91b8ab58f2b2 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/nvidia.stdout.txt @@ -0,0 +1,13 @@ +{ + "status": "PASS", + "semantic_tasks": 189, + "capacity_bins": { + "131072": 72, + "262144": 65, + "65536": 52 + }, + "exact_token_ranges": { + "65536": 13 + }, + "production_eligible": false +} diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/siriusxm.stderr.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/siriusxm.stderr.txt new file mode 100644 index 0000000000000000000000000000000000000000..5f6203a3ea82835862ab8341b42c5c0d75d0a842 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/siriusxm.stderr.txt @@ -0,0 +1 @@ +[transformers] Token indices sequence length is longer than the specified maximum sequence length for this model (382842 > 262144). Running this sequence through the model will result in indexing errors diff --git a/local-probe-train-ready/p64-primary-training-v2/source_validation/siriusxm.stdout.txt b/local-probe-train-ready/p64-primary-training-v2/source_validation/siriusxm.stdout.txt new file mode 100644 index 0000000000000000000000000000000000000000..3eeb139e03f123cd8b36ea7376d7cdeb31e52760 --- /dev/null +++ b/local-probe-train-ready/p64-primary-training-v2/source_validation/siriusxm.stdout.txt @@ -0,0 +1,10 @@ +{ + "status": "PASS", + "semantic_tasks": 192, + "capacity_bins": { + "131072": 126, + "262144": 66 + }, + "exact_token_ranges": {}, + "production_eligible": false +} diff --git a/release_inventory.json b/release_inventory.json new file mode 100644 index 0000000000000000000000000000000000000000..c177cbff404eee331ff76d17586738116c7b39cb --- /dev/null +++ b/release_inventory.json @@ -0,0 +1 @@ +{"attestation": {"digest": "de1565217c20bc621a79092cb5d1d348124466dc43c348482e2b9310c446e8af", "environment": "probe", "key_id": "probe-report-20260824-01", "purpose": "release_inventory", "role": "report", "scheme": "hmac-sha256-v2"}, "files": [{"bytes": 48540761, "path": "05_promoted/eval.jsonl", "role": "promoted_data", "sha256": "cc60fb0c666c593a02a2103721942b2dbe036ef3563e53973aa854cb6885030f"}, {"bytes": 2558, "path": "05_promoted/quality_report.json", "role": "promoted_data", "sha256": "54d2048ae1ab5b67af687235470cc7f25bc451b99eb09968e6c04f2b66015f3d"}, {"bytes": 1064, "path": "05_promoted/release_gate_pass.json", "role": "release_gate_pass", "sha256": "fbab5136b57920281a0917cc287efe17af87581f1bd2a6191a7c9aef44dc4c5b"}, {"bytes": 161851783, "path": "05_promoted/train.jsonl", "role": "promoted_data", "sha256": "6e8dd2b8752b0c71445770119ee0f7295208154d8ae6248a91170d3ee0ce1fab"}, {"bytes": 14210676, "path": "06_training/B1.json", "role": "training_output", "sha256": "c396513770b07cf07fa6c1909bea461fe8c65ee49142ae16a6d73db461749f1b"}, {"bytes": 398, "path": "06_training/B1.meta.json", "role": "training_output", "sha256": "9f2bf4f256c0d3ec2d8bb85392116f25327955cadb77ff601748feeeb58f2dc8"}, {"bytes": 14114164, "path": "06_training/B3.json", "role": "training_output", "sha256": "f78e4d906a5aeacc5dcccc324b7d95fbb5d85d0b20b787c21e285b8d806ead1f"}, {"bytes": 469, "path": "06_training/B3.meta.json", "role": "training_output", "sha256": "eeec97a1bce2cf5f376f0745bc396e6654d1e71e64dd47b1fc6930c76a967c38"}, {"bytes": 14163448, "path": "06_training/B5.json", "role": "training_output", "sha256": "8520d823419cdbe4093a2e5f1b00541ca50e540a4dc93ff313257ccacffaa0f2"}, {"bytes": 622, "path": "06_training/B5.meta.json", "role": "training_output", "sha256": "b95b7a956a9ab828efd403c840656b656880a2d06a509c0d29d0fe94bc306668"}, {"bytes": 88, "path": "06_training/B5w.datasets.yaml", "role": "training_output", "sha256": "25c6f926040f5608b3f95529c5b05a667bd69d16f309e201a9a7aee79b390b33"}, {"bytes": 7057556, "path": "06_training/B5w.json", "role": "training_output", "sha256": "7da381371e7ca74efdba413544fd25e3bbdbd505bcea1ed163712b23af8e96b4"}, {"bytes": 621, "path": "06_training/B5w.meta.json", "role": "training_output", "sha256": "1053655e0bd0be595b7f8a2220e70fa8125cc3617ac5b757b2d30577e9da5a76"}, {"bytes": 915, "path": "06_training/dataset_info.json", "role": "training_output", "sha256": "7bd596940721516a48c6e079950e66a9e0f06224a23f9157f1f98329d1e23220"}, {"bytes": 2853, "path": "06_training/export_summary.json", "role": "training_output", "sha256": "bb7fd8268896188ebcedeb8434428c38c12a5cb89403038475f8bc205c5b8347"}, {"bytes": 2341, "path": "06_training/training_export_manifest.json", "role": "training_manifest", "sha256": "d1a852bfd667dac05d82dac04261ecae005fb86a2d83bbba80779a53639290c9"}, {"bytes": 3756, "path": "README.md", "role": "data_card", "sha256": "90bd891d2b73a24ab087bc607012fbf5c62f705ad0a0e5cdb2d185e8dbf0affe"}], "production_eligible": false, "release_gate_pass": {"bytes": 1064, "path": "05_promoted/release_gate_pass.json", "role": "release_gate_pass", "sha256": "fbab5136b57920281a0917cc287efe17af87581f1bd2a6191a7c9aef44dc4c5b"}, "release_profile_id": "p6-source-dependent-probe-12-v1", "release_profile_sha256": "68601a08ceabc7e6c5dec0a49a398318ebb596a1da65a3ac4bcdc35b5585b495", "schema_version": "longworld-private-dataset-release-v1", "training_manifest": {"bytes": 2341, "path": "06_training/training_export_manifest.json", "role": "training_manifest", "sha256": "d1a852bfd667dac05d82dac04261ecae005fb86a2d83bbba80779a53639290c9"}, "training_manifest_attestation": {"digest": "6d5d4e66161cd8427fdbb864a06302294ed46a84a2c2702d9cbd2689b7b71d51", "environment": "probe", "key_id": "probe-report-20260824-01", "purpose": "training_export_manifest", "role": "report", "scheme": "hmac-sha256-v2"}, "transform_revision": "longworld-llamafactory-sharegpt-v4", "trust_mode": "local_engineering"}