gnomonlearn
Docs/RFC0014 CLAIM REVIEW

RFC0014 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: 70 of 70 claims reviewed on 2026-10-05. Owners: Compiler, Platform. Final totals: 34 source matches and 36 whole-node mismatches. No normative facts changed.

Version format, aliases, and lockfile authority

The reviewed group covers §3 component rules, §4 breaking-change examples, §10 aliases and lockfile invariants, and two §11 failure outcomes. SemVer is defined in RFC-0040’s domain projection with definedIn RFC-0014; RFC-0014 independently defines SemanticVersion. The identities require canonical ownership review, but SemVer is not an absent domain node.

The determinism claim targets absent versioning-determinism, while the authored invariant is semver-determinism. Four §4 claims and the no-protection claim target absent breaking-change-definition.

The lockfile authority summary omits the authored qualification “for version selection during a locked build.” §7.3 requires registry-based re-resolution to detect staleness. A broad no-registry promise contradicts that procedure. The final §11 row also promises no automated protection from breaking MINOR releases, while §8.1 and an earlier §11 row require a blocking M14800 lint gate.

Acceptance obligations

  • Reconcile SemVer and SemanticVersion ownership across RFC-0014/0040.
  • Repair invariant subject mappings with explicit owners and consumer checks.
  • Distinguish locked selection, freshness checking, and explicit updates.
  • Define when semver lint blocks publication and when consumers can still encounter an incorrectly classified upstream release.
  • Specify prerelease opt-in and verify resolver behavior with executable fixtures; source agreement does not prove implementation conformance.

All claim nodes now have source-bound reviews. The earlier 69-node inventory was incorrect; the current authored protobuf contains 70 nodes.

Change classifications and constraints

The next twenty claims have three matches and seventeen whole-node mismatches. Eleven change-classification claims target the absent breaking-change-definition invariant. Four constraint claims target absent CompatibleConstraint, ExactConstraint, or PatchCompatibleConstraint types. Two version-slot claims target field/import/version, while the authored identity is field/imports/version.

Additional semantic issues need explicit contracts:

  • The generic caret rule says same MAJOR, omitting its MAJOR=0 exception.
  • An internal change can preserve public signatures while changing downstream runtime behavior. The blanket internal-change PATCH rule must be reconciled with the opening behavior-based breaking-change definition.
  • The exactly-one version-string claim needs duplicate-key rejection rules. RFC-0008’s module import production still uses FQN : STRING, while RFC-0014 demonstrates a quoted module identifier followed by version: STRING.

The circular-dependency, full-cycle diagnostic, and first-conflict halt claims agree with §6.2/§6.3 and reference the authored resolution operation.

Resolution, lockfiles, diagnostics, and updates

The final thirty claims have eighteen matches and twelve whole-node mismatches. Seven reference absent declaration/module.lock; the existing ModuleLockfile type and Lean declaration do not emit that identity. Freshness, fatal-phase, and range claims also use absent invariant or phase aliases.

M01410’s description says caret on a 0.x module permits MAJOR upgrades, but §5.1 restricts ^0.2.0 to <0.3.0. The warning trigger must be corrected or the range contract changed before implementation.

Two further implementation requirements follow from the authored examples:

  • Whole-lockfile freshness comparison must specify how the generated timestamp is handled; a new wall-clock timestamp must not alone make a lock stale.
  • Transitive imports depend on the version selected for each module. §6.1’s collect-before-select steps need a concrete deterministic resolution procedure and fixtures with version-dependent dependencies and conflicts.

The matching diagnostic and update claims establish source agreement only. No executable module resolver, semver linter, or locked-build behavior was verified by this review.