gnomonlearn
Docs/RFC-0053 adopted owner and index-policy repairs

RFC-0053 adopted owner and index-policy repairs

Versioned draft snapshot

RFCs, reference catalogs, and evidence reflect Spec 3b2e3e7a114fb826b1445d39a48756b8fff629cd. Counts and pending/status notes belong to that snapshot. Read the dated project update for later progress.

Source repositories are private during staging. Links marked “repository access required” need access to the linked repository.

This is a trace and gap record from the pinned draft specification. It does not establish proof or implementation conformance.

The owner approved both prepared RFC-0053 batches on October 16, 2026. They were applied once at clean Spec baseline 8e20a7edc92e8c033aded8010542798245bb3d90, using the existing protobuf projection sources and Python emitter.

Exact source and review changes

  • The approved existing-owner.patch is SHA-256 fdaf7c6cd1f3b8aa5c8557e353655b055dff8fda8f632de394dc10536fd0e2ec (15,364 bytes). Its 13 mappings replace 58 references in 55 claims: 54 claimSubject and four claimObject values, all pointing to existing authored domain owners. No target node or class changes.
  • The separately approved index-policy.patch is SHA-256 6a9007aa2165fdd58aa26028ca864555608d411a416735ff587848a11a57cbe6 (1,621 bytes). It changes two claim texts and their two predicate strings to retain authored §4.1’s eligibility and default exclusion. The generation claim retains the limited Phase-1 and expanded Phase-2 query-analysis scope.
  • The exact final-v2 mapping review patch has SHA-256 2e2f74d89a5c5d756f03d0947e40837852c03c71ddda6b03a716c506b9fd621c (74,623 bytes). It proposed 45 closures and ten holds. The two index-policy claims were then substantively reviewed at the combined source, against §4.1 and unchanged Lean eligibility/exclusion statements. They now match. The earlier superseded 47-closure mapping draft was not used.

The actual result is 47 corrected verdicts among 55 affected claims, with eight independent holds retained. The number 47 here follows the approved 45 mapping closures plus the separately approved and reviewed two scope corrections. It does not reuse the superseded mapping-only verdict count.

Current source-bound ledgerReviewedMatchesMismatches
Complete projection corpus6,9624,7032,259
RFC-0053 claims785721
RFC-0053 domain nodes582731
RFC-0053 combined1368452

All 6,907 unrelated review messages and their textual blocks remain exact. Only the 55 owning node hashes/findings and 47 verdicts change. Narrative and Lean fingerprints, all other review fields, every other claim/domain source, and target definitions remain unchanged. The composed source equals the exact owner candidate plus the four approved literals. Claim IRIs, kinds, confidence, and section provenance are preserved.

The regenerated domain-link inventory has 940 unresolved references to 456 identities across 51 RFCs, down from 998 references to 469 identities. RFC-0053 retains 13 references to its 13 unresolved language-type identities. All eleven symbolic dispositions remain unchanged.

Actual combined query comparison

verify_adoption.py (repository access required) recomputes the combined change against the saved actual pre-edit source/Turtle and immutable Git baseline. It reuses the existing actual-adoption query functions and Effects schema (repository access required). The earlier counterfactual effect files remain immutable inputs and are not used as proof of this combined result.

The genuine command a6d460/session21836 ends 1cec76/0. Its complete output is adoption-verification.log (repository access required). The graph retains 89,351 triples, with exactly 62 removed / 62 added triples: 58 owner references and four scope literals.

All 38 displayed SELECT multisets and columns remain unchanged. All fourteen terminal-LIMIT queries were also compared without only their final outer LIMIT. Five complete relations change:

Complete relationBefore → afterRemoved / added rows, including multiplicity
claims/enforcement-claims.rq459 → 4599 / 9
claims/exclusions.rq359 → 3594 / 4
claims/lifecycle-transitions.rq65 → 670 / 2
claims/orphan-claims.rq954 → 90054 / 0
claims/predicate-vocabulary.rq2,274 → 2,2762 / 4

There is zero limit-only page displacement. The other nine complete relations are unchanged, including implicit conflicts. The complete predicate relation retains the new scoped vocabulary rather than treating its strings as aliases of the old predicates. The complete orphan relation loses 54 rows; this is distinct from the 58 repaired references and 47 source-review closures.

Missing-owner references change 998 → 940, undefined identities 469 → 456, and affected claim nodes 982 → 927. Full row differences, multiplicities, multiset hashes, exact RDF deltas, and 362 closing-checked input bindings are in query-effects.textproto (repository access required). comparison-inputs.textproto (repository access required) records the actual comparison inputs using the existing SourceFreeze message. The saved claim source (repository access required), Turtle (repository access required), complete review ledger (repository access required), and domain-link inventory (repository access required) retain the pre-edit bytes captured before patch application.

The existing bounds remain 300 seconds wall, 240 seconds CPU, 250,000 rows per relation, and an 8 MiB effect file. The actual effect is 159,981 bytes. The graph is restored in memory after comparison, and every input binding is rechecked unchanged. Retained generated formal Turtle is existing query context; it was not freshly emitted from Lean.

Closing gates and remaining holds

The ordinary focused gates all return zero:

Generation and initial review checks end e107d1/0. The closing gate process 0d2a09/session96691 ends 0dbbbb/0, verifying the expected results, including the two explicit remaining release holds:

An initial documentation-summary assertion 48f101/1 caught an incorrect sum of overlapping historical domain-review batches. Current counts above come directly from the canonical typed review ledger: 58 domain nodes, with 27 matches and 31 mismatches. The actual adoption and query comparison were already using the correct 52 total RFC-0053 mismatches; neither was rerun or changed to repair this draft documentation error.

The eight affected RFC-0053 semantic holds remain source-bound mismatches:

Claim suffixRemaining issue
sec-5/materialized-view-owns-rlsNative materialized-view RLS assumption is rejected by the retained PostgreSQL probe.
sec-5/materialized-view-uses-cached-permissionSame unsupported materialized-view security model.
sec-5/principal-needs-has-permissionCaller versus view-owner RLS context needs the pending security contract.
sec-5/views-inherit-rls-from-resourceSecurity barrier alone does not establish caller authorization.
sec-6/phase-1-read-onlyAbsent triggers do not enforce the promised read-only behavior.
sec-9/concurrent-index-buildTransaction boundaries remain tied to the pending execution contract.
sec-9/drop-view-cascadesAuthored automatic-cascade claim conflicts with default RESTRICT behavior.
sec-9/orphan-column-removal-emits-M53030RemoveField error rule conflicts with optional retained JSON data.

The other 13 claim mismatches concern language-type identities; all 31 domain mismatches remain. These changes do not resolve security, lifecycle, transaction, type, proof, compiler, or SQL execution conformance. No native Lean/Roma rerun, Site rebuild, consumer pin, release, or public publication is claimed. Installed Python 3.14.7 and retained Python dependencies/protoc were reused; no complete Python import-closure audit is claimed.