gnomonlearn
Docs/RFC-0053 view-mapping claim review

RFC-0053 view-mapping claim review

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.

Current adoption status (2026-10-16)

All 78 claims have current source-bound reviews: 57 match and 21 mismatch. The approved combined adoption repairs 58 owner references and four index-policy literals. It closes 45 mapping-only findings and, after substantive review against §4.1 and Lean, the two separately approved index-policy findings. The eight other affected semantic holds and 13 language-type identity mismatches remain.

No domain class, authored RFC, Lean catalog, proof, or runtime contract is verified by these source matches. The complete corpus has 4,703 matches and 2,259 mismatches; RFC-0053 retains 52 mismatches including its 31 domain findings. Earlier review observations below retain their historical scope.

Historical claim review (2026-10-05)

Status: all 78 claims reviewed on 2026-10-05; ten match and 68 mismatch. Owners: Compiler, Storage, RDF ontology maintainers. No normative facts changed.

The text of these claims agrees with the corresponding narrative sections or structured Lean diagnostics. However, 23 subjects have no definition in any authored domain source. They use identities such as EntityTypeView, ExpressionIndex, InsteadOfTrigger, MaterializedView, ModuleSchema, and ViewMappingManifest, whereas the domain source uses hyphenated catalog names or the emitted manifest declaration. The naming invariant similarly uses view-naming-snake-case instead of the domain catalog’s snake-case-naming.

Case and spelling changes produce different RDF IRIs. Text agreement cannot establish the identity mapping. Inventory every incoming use, identify the canonical owner, prepare explicit source substitutions, and get owner approval before changing links. The view/index/trigger catalog classes also need the separate correction described in the domain review.

The index naming invariant and generation-input claim have defined subjects and match their authored sections. These two source matches do not establish compiler output or SQL conformance.

The PostgreSQL evidence separately reproduces failures in authored security assumptions. The security proposal defines a pending replacement contract. Source-bound claim reviews must not be presented as evidence that those assumptions pass on PostgreSQL.

Acceptance: canonical subject/object references resolve, generated RDF preserves approved links and provenance, source-bound review hashes stay current, and query consumers observe the intended identity changes. Review index-emission summaries with §4.1’s conditional policy rather than claiming an index for every typed field.

Type mapping and security claims

The next 28 nodes contain two source matches and 26 identity mismatches. The lowercase language-type references are undefined in the authored domain corpus; their SQL cast mappings do not resolve canonical language identities. The other undefined references repeat the catalog aliases inventoried in the identity proposal, including the two permission-function operation aliases.

The KP29 security-model claim and materialize-slot claim have defined subjects and text matching §5. That source agreement does not prove correctness: the database probe demonstrates that security_barrier alone does not establish caller RLS context. The domain review also records the materialize slot’s incorrect Operation classification.

Materialized-view policy claims both have unresolved identity links and rely on SQL rejected by PostgreSQL 14.24. These are separate failures: repairing the links cannot make native materialized-view RLS executable. The normative security proposal must be resolved and tested in the supported version matrix.

Write-through, joins, and regeneration

The final 25 nodes contain six matches and 19 unresolved-identity mismatches. Their text follows the authored pipeline, provenance, join, and lifecycle sections. Undefined trigger/view/index aliases and invariant aliases account for the unresolved structured links; the bounded identity proposal inventories them across the full claim source.

M53030 also exposes an authored conflict: Lean treats column removal without a RemoveField migration as an error, while §9.2 explicitly allows the JSON key to remain after a view column is removed. Define when actual data deletion is required and narrow the diagnostic accordingly, or change the optional retention rule with explicit owner approval.

The Phase-1 read-only claim relies on absent write-through triggers, which the database probe shows is insufficient for a simple view’s DELETE path. Join RLS claims retain the same unresolved owner/caller context as entity views. These runtime findings survive any RDF alias repair.

The concurrent-index recommendation must be reconciled with the pending RFC-0054 transaction contract. Generated SQL, migration diagnostics, and dependency regeneration require executable tests beyond this source review.