{"slug": "same-data-different-verdicts-canonicalization-and-receipt-lessons-from-ietf-127", "title": "Same data, different verdicts: canonicalization and receipt lessons from IETF 127", "summary": "At the IETF 127 Hackathon in San Francisco, projects working on verifiable evidence of AI agent actions asked independent implementers to write verifiers from their documents alone, and found that each implementation passed its own tests while failing between implementations. The Action Evidence Packages project's Python and JavaScript reference verifiers serialized the same number differently (1e-07 vs 1e-7) and returned ALLOW and DENY verdicts, while the Agent Evidence Conformance Matrix found four JSON canonicalization implementations agreed on 20,121 of 20,140 cases. The post, checked against slides and drafts, draws lessons for VAP, the metaframework maintained by VSO, noting that VAP does not define how a transparency service serializes entries and cannot make a receipt sign its tree size.", "body_md": "At the IETF 127 Hackathon (San Francisco, 14–15 November 2026), several projects working on verifiable evidence of AI agent actions asked other people to write verifiers from their documents alone. Each implementation passed its own tests. The failures appeared between implementations. This post collects the concrete ones, checked against the slides and drafts, and notes what they mean for VAP, the metaframework VSO maintains.\n\nThe Action Evidence Packages project reported that its own two reference verifiers disagreed on a single number: the Python one produced `1e-07`, the JavaScript one `1e-7`, and the verdicts came back ALLOW and DENY.\n\nYou can reproduce the serialization difference in a second:\n\n``` python\n>>> import json\n>>> json.dumps(1e-7), json.dumps(1e20), json.dumps(2**53 + 1)\n('1e-07', '1e+20', '9007199254740993')\n> [JSON.stringify(1e-7), JSON.stringify(1e20), JSON.stringify(2**53 + 1)]\n[ '1e-7', '100000000000000000000', '9007199254740992' ]\n```\n\nRFC 8785 (JCS) requires the ECMAScript number serialization, so the Python defaults above are not JCS output. The last pair is worse than a formatting difference: the integer cannot be represented exactly as an IEEE 754 double, so a JavaScript reader has already changed the value before any hash is computed.\n\nThe Agent Evidence Conformance Matrix (maintained by Sankalp Gilda) ran four JSON canonicalization implementations over the same corpus. They wrote identical bytes on **20,121 of 20,140** cases. The eight inputs that split them:\n\nThe slide's conclusion: a JCS profile has to say \"refuse\" for these. RFC 8785 limits its input to I-JSON (RFC 7493), which already rules out duplicate property names, but a verifier still has to decide what it *reports* when it meets one.\n\nTwo operators registered the same four SCITT Signed Statements on their two Transparency Services and checked each other's receipts. The same statement produced a different leaf on each log: one hashes the statement's bytes, the other a digest of its `Sig_structure`. RFC 9943 and RFC 9942 do not name the entry serialization, so you cannot check a receipt without knowing the service's rule.\n\nThe authors list their limits: the operators co-champion the same project, the second log's signed tree head was not checked, and one side's checking script is unpublished.\n\nIn an RFC 9942 inclusion receipt (`RFC9162_SHA256`), the signature covers the Merkle root; tree size and leaf index are in the unprotected header. A receipt read alone proves inclusion under a signed root, not position or log size. One service edited 8,255 (tree size, leaf index) pairs: a plain receipt was accepted for the issued pair and 31 others; with the signed tree head, only the issued pair. An erratum was announced on the SCITT list (filing not confirmed by us).\n\nAt the SCITT working group on 18 November, one slide summarised an experiment: a single Signed Statement re-encoded 64 ways with CBOR's encoding freedoms gave 64 different hash values, and a stock decoder rejected none of them. If you reference a record by digest, you have to pin both the bytes and the canonicalization. `draft-mih-sokolov-cose-deterministic-encodings-01` does this for the COSE Hash Envelope preimage, for CBOR only (`\"dtrm\"`); pairing it with `application/json` is forbidden.\n\nFrom the VAP v1.2 text:\n\n`event_hash` is computed over canonical JSON, so attaching a COSE receipt does not change it.\nStill open, and now recorded as **unadopted** candidate questions (CHANGE-002 preparation register, §5):\n\nFix the text first, then build the reproduction kit; otherwise the kit freezes one implementation's behaviour.\n\nNone of these results would have changed. VAP does not define how a transparency service serializes entries, and it cannot make a receipt sign its tree size. None of the projects above is an implementation of VAP or VCP. VAP v1.2 §1.6(c): conformance does not warrant the correctness, fairness, or safety of the underlying AI decisions — only the integrity, completeness (at anchor granularity), and attributability of their records.\n\nZero external implementations of VAP or any VAP profile; zero Evidence Packs accepted in any proceeding (count dated 25 October 2026). No claims about contracts, customers or revenue. First-party work by VSO (operated by VeritasChain Co., Ltd.), prepared with Claude (Anthropic) as an AI agent, checked against the primary sources below; no external review. VSO authors `draft-kamimura-vap-framework` and has corresponded with one of the hackathon champions on IETF coordination.\n\nLonger version: [https://medium.com/@veritaschain/what-ietf-127-showed-about-evidence-someone-else-has-to-check-de2cf37ecb41](https://medium.com/@veritaschain/what-ietf-127-showed-about-evidence-someone-else-has-to-check-de2cf37ecb41)", "url": "https://wpnews.pro/news/same-data-different-verdicts-canonicalization-and-receipt-lessons-from-ietf-127", "canonical_source": "https://dev.to/veritaschain/same-data-different-verdicts-canonicalization-and-receipt-lessons-from-ietf-127-1pac", "published_at": "2026-11-19 23:38:26+00:00", "updated_at": "2026-11-19 23:42:59.249552+00:00", "lang": "en", "topics": ["ai-agents", "agent-protocols", "ai-safety"], "entities": ["IETF", "VSO", "VAP", "Sankalp Gilda", "Agent Evidence Conformance Matrix", "SCITT", "RFC 8785", "RFC 9942"], "also_reported_by": [], "takeaway": "Four JSON canonicalization implementations agreed on 20,121 of 20,140 cases, and Python's default json.dumps emits 1e-07 where JCS requires 1e-7, so verifiers must pin canonicalization and refuse ambiguous inputs rather than trust default serializers.", "recommendation": {"advice": "Pin both the bytes and the canonicalization rule when referencing records by digest, and make verifiers refuse duplicate keys and non-I-JSON numbers instead of silently reinterpreting them.", "playbook": "output-verification", "playbook_url": "https://wpnews.pro/playbooks/output-verification"}, "alternates": {"html": "https://wpnews.pro/news/same-data-different-verdicts-canonicalization-and-receipt-lessons-from-ietf-127", "markdown": "https://wpnews.pro/news/same-data-different-verdicts-canonicalization-and-receipt-lessons-from-ietf-127.md", "text": "https://wpnews.pro/news/same-data-different-verdicts-canonicalization-and-receipt-lessons-from-ietf-127.txt", "jsonld": "https://wpnews.pro/news/same-data-different-verdicts-canonicalization-and-receipt-lessons-from-ietf-127.jsonld"}}