{
  "schema": "csoai.scrapi-ccf-framing-space/0.1",
  "supersedes_nothing": "data-hash-vector.json and data-hash-tag-vector.json are unchanged. This adds a third artifact; published bytes are added to or superseded, never edited.",
  "raised_by": "Extends the tag axis (Nicholas Templeman) and the outer-array axis (Konrad Gruszka, b7n0de) by measuring the SIZE OF THE CLASS both belong to.",
  "derived_from": {
    "file": "data-hash-vector.json",
    "field": "A_signed_statement_as_registered.bytes_hex",
    "sha256_of_A": "8595e4a4c8b93e7b1b7b798dc302a2b7d2890021f7eff372d79b32f78867e4ac",
    "size_bytes": 165
  },
  "method": "Decode A once, then re-emit the identical COSE_Sign1 data item under every combination of six CBOR encoding freedoms. No element value changes; only its encoding.",
  "encoding_freedoms_tested": [
    "tag18",
    "outer_array_indefinite",
    "unprotected_map_indefinite",
    "protected_bstr_chunked",
    "payload_bstr_chunked",
    "signature_bstr_chunked"
  ],
  "finding": {
    "distinct_byte_encodings": 64,
    "distinct_data_hash_values": 64,
    "digest_collisions": 0,
    "rejected_by_stock_parser": 0,
    "silently_reserialising_to_A": 31,
    "reader": "cbor2 6.1.3, CPython 3.13, macOS arm64",
    "plain_english": "Sixty-four legal CBOR encodings of the same COSE_Sign1 produce sixty-four different data-hash values. A stock parser rejects none of them, and thirty-one are repaired into the canonical form by the act of reading them, so an issuer and a verifier can disagree with no error raised anywhere."
  },
  "why_it_settles_the_wording": "Two framing axes were found by accident in one week. Six freedoms give sixty-four encodings, and this does not test integer-width encoding or map key ordering, so the space is larger still. A rule that enumerates framings cannot close a combinatorial class; a rule that binds the octets AS REGISTERED closes all of them, including the ones not yet found.",
  "supports": "Konrad Gruszka's proposed wording on the SCITT list, 2026-09-04, which is better than ours: ours named the tag, his binds the octets as registered and so closes the whole class this artifact measures.",
  "not_a_transparency_service": "Not a Transparency Service. Not a real Receipt. Not a claim about what any deployed implementation does. Vectors derived from A above; the seed, bytes and signature are from data-hash-vector.json.",
  "as_of": "2026-09-04T10:45:37Z",
  "named_variants": [
    {
      "label": "C \u2014 tag removed (raised by Nicholas Templeman)",
      "freedoms_set": [],
      "size_bytes": 164,
      "sha256": "fead5e6058d01360024bcd0d539138d26053c3d7c48475dbccb32b6139e53a00",
      "accepted_by_stock_parser": true,
      "reserialises_to_A": false
    },
    {
      "label": "A \u2014 as registered (definite everywhere, tagged)",
      "freedoms_set": [
        "tag18"
      ],
      "size_bytes": 165,
      "sha256": "8595e4a4c8b93e7b1b7b798dc302a2b7d2890021f7eff372d79b32f78867e4ac",
      "accepted_by_stock_parser": true,
      "reserialises_to_A": true
    },
    {
      "label": "H \u2014 signature byte string chunked",
      "freedoms_set": [
        "tag18",
        "signature_bstr_chunked"
      ],
      "size_bytes": 169,
      "sha256": "3a0b2c7b207af2f6775e76805554ae5e52537d0fd6447d73320260b9fed44603",
      "accepted_by_stock_parser": true,
      "reserialises_to_A": true
    },
    {
      "label": "G \u2014 payload byte string chunked",
      "freedoms_set": [
        "tag18",
        "payload_bstr_chunked"
      ],
      "size_bytes": 170,
      "sha256": "8c5cff1ae618e0a7b70ee6f4eccfc0a261bcdcef258b3406f4ed89b08f5f2ee5",
      "accepted_by_stock_parser": true,
      "reserialises_to_A": true
    },
    {
      "label": "F \u2014 protected byte string chunked",
      "freedoms_set": [
        "tag18",
        "protected_bstr_chunked"
      ],
      "size_bytes": 168,
      "sha256": "31ac0ff6027a00e5ed72a7633e0dc62e4ba0343db9810d51f65bf40de8c97530",
      "accepted_by_stock_parser": true,
      "reserialises_to_A": true
    },
    {
      "label": "E \u2014 unprotected map indefinite",
      "freedoms_set": [
        "tag18",
        "unprotected_map_indefinite"
      ],
      "size_bytes": 166,
      "sha256": "1afbdbb466a80be1c95849c055496f56fab868962b87ba55ad7e2b2b044a4be2",
      "accepted_by_stock_parser": true,
      "reserialises_to_A": true
    },
    {
      "label": "D \u2014 outer array indefinite (raised by Konrad Gruszka)",
      "freedoms_set": [
        "tag18",
        "outer_array_indefinite"
      ],
      "size_bytes": 166,
      "sha256": "79c6215d691cd72ce47feb4e5b26eb047e1880a64c7ef6fbd51f6a0cba092a11",
      "accepted_by_stock_parser": true,
      "reserialises_to_A": true
    }
  ],
  "root_cause": {
    "asymmetry": "The signature is framing-invariant; data-hash is framing-sensitive.",
    "sig_structure_sha256": "60b4c76b84c456ed0604305075f51f6d050476177ef17be70a13a7b9abfb370f",
    "sig_structure_bytes": 109,
    "distinct_sig_structures_across_all_64": 1,
    "distinct_data_hash_values_across_all_64": 64,
    "consequence": "Two parties can agree that a signature is valid and still disagree on the data-hash of the statement they both just verified. Per RFC 9052 Section 4.4 the Sig_structure covers the protected header and the payload and not the container framing, which is exactly why one signature verifies over all 64 encodings. data-hash is taken over the wire octets, which the framing does change.",
    "why_it_matters_for_the_wording": "The fix is not to make data-hash framing-invariant \u2014 the wire octets are what a client fetches and byte-compares, so hashing them is the right choice. The fix is to say WHICH octets. That is Konrad Gruszka's sentence, and it carries a consequence worth stating: a Transparency Service must retain the octets it received rather than a re-serialisation of them, or it cannot reproduce the data-hash it published."
  },
  "credit": {
    "outer_array_axis": "Konrad Gruszka (b7n0de), raised on the SCITT list 2026-09-04 with an independent recomputation on a third platform using his own reader and no COSE library. This artifact measures the class his axis and the tag axis both belong to; it does not discover either.",
    "tag_axis": "Nicholas Templeman (CSOAI), raised on the SCITT list 2026-09-03.",
    "independent_recomputation": "Emek Can Dogru (VERAX Teknoloji / Conarium) recomputed the original vector rather than reading it, which is what surfaced the tag gap in the first place.",
    "prior_art_we_did_not_know": "draft-mih-sokolov-scitt-payload-binding registers `as-transmitted` \u2014 no canonicalization, the pre-image is the exact octet sequence \u2014 and withdrew two canonicalization profiles (jcs-n, cde-n) before it. That is Konrad's proposed wording as a registered algorithm, reached before either of us.",
    "note": "Konrad's recomputation is published at b7n0de/proofbundle PR #178. His proofbundle project is an in-toto predicate for eval receipts with a conformance suite, threat model, formal methods and fuzzing, on PyPI with a DOI. Where his work and ours overlap, his is the more mature and the earlier."
  }
}