gnomonlearn
Docs/RFC-0052 platform claim review

RFC-0052 platform 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.

Status: all 80 claim nodes reviewed on 2026-10-05; 71 match and nine mismatch. Owners: Module Platform, Compiler. No normative facts changed.

The reviewed claims cover namespace purpose, platform composition, namespace isolation, install ownership, diagnostic contracts, definitions, and the first audit/auth catalog entries. Agreement with source does not prove operational enforcement, generated SQL equivalence, or implementation completeness.

Two claims preserve contradictions already found in the domain review:

  • sec-1/db-namespace-purpose says every db.* module owns one PostgreSQL schema, while its subject db.pg explicitly owns none in §3.1.
  • sec-10/db-module-single-schema applies dot-to-underscore naming to platform schemas, although §3 explicitly assigns graph to db.graph and ids to db.ids. §5 applies namespace-derived naming to application view schemas.

Acceptance: retain db.pg’s schema-less exception, separate platform schema ownership from application view-schema naming, reconcile narrative and Lean, regenerate projections, and test schema ownership/conflict behavior. Prepare the exact source changes for owner approval before changing the contract.

The M52080 node matches the diagnostic’s Info-level report of phase end-state disagreement. It does not resolve the separate conflict between fresh-install convergence and repeat-install idempotency described in the domain review.

Catalog dependency cycle

The next 30 claims contain 28 matches and two mismatches. Physical schema ownership, module counts, and the listed module responsibilities match their catalog sections. These counts are authored assertions, not independently verified inventories of compiled artifacts.

sec-3/db-pg-depends-on-core and sec-3/db-pg-rank expose an install conflict: §3.1 makes db.pg depend on core, §4.1 makes core depend on db.graph, and §3.4 makes db.graph depend on db.pg. This cycle contradicts §6.2’s cycle rejection. Rank 1 for db.pg also precedes its rank-7 core dependency. Bootstrap SQL may provide storage prerequisites, but the RFC does not define an earlier kind-definition seed that satisfies this dependency.

Acceptance: specify the bootstrap kind seed and its ownership, distinguish bootstrap prerequisites from normal module dependencies, reconcile all catalog ranks and the install DAG, and test cycle rejection plus a fresh two-phase installation. The explicit core-to-db.graph and db.graph-to-db.pg claims match their sections; their agreement does not make the full graph acyclic.

Bootstrap, layering, and visibility

The final 25 claims contain 20 matches and five mismatches:

  • The application schema naming text matches §5, but its structured subject is the db.* single-schema invariant. Give application view-schema naming its own identity instead of attaching it to the platform invariant.
  • The claim that manifest dependencies produce the shown deterministic order conflicts with the cycle above. Also, core.composition depends on core.kind, although both are called rank-8 parallel modules. Define whether ranks are broad groups with internal sequencing or independent parallel levels.
  • The extends call permission is unconditional, while §8 requires template visibility checks. Specify that extends retains export/import and visibility requirements; it cannot grant access to every private template.
  • Both protected create-template claims allow core.* callers even though the same section defines protected as descendant-only. Specify the exception or allowlist explicitly, preserving application restrictions.

Acceptance requires source/Lean reconciliation, corrected structured subjects, regenerated projections, deterministic dependency ordering tests, and access tests for declaring modules, descendants, authorized core callers, unrelated platform callers, and application callers. Owner approval precedes changes to the normative visibility and installation contracts.

The reviewed 92.5% coverage claim agrees with the inherited narrative; this review supplies no current measurement or compiler round-trip evidence.