gnomonlearn
Docs/Known pinned formal input inconsistency

Known pinned formal input inconsistency

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 unchanged external framework at 3725bd9f6d2711aa9bb179deea84f51e4424382a compiles this minimal observation:

theorem pinned_framework_derives_false : False :=
  Spec.kindNonempty (default : Spec.Resource) rfl

Resource structurally permits an empty kind and derives Inhabited; kindOf projects that field. The universal kindNonempty premise excludes the default value. The control imports the actual pinned Spec.Axioms, uses no sorry or new axiom, and #print axioms reports only [Spec.kindNonempty]. Proofs importing these inputs cannot establish sound normative guarantees. Matching proposition reviews and compilation counts remain source observations.

Actual terminal evidence

observe.py compiles the pinned Universe, Predicates, Kernel, and Axioms, then the minimal observation with leanprover/lean4:v4.30.0-rc2. All five invocations exited 0. The actual control’s log records its axiom dependency. Compiled outputs are retained under the owned persistent .work/framework-consistency-20261013 path; the original external checkout and dependency pin were not modified. evidence.textproto uses the existing ValidationEvidence protobuf to bind exact source bytes, terminal commands, outputs, logs, and the source base commit.

The shared known-blocker registry is keyed to this immutable source commit. Strict check_derivations.py and check_release.py both report the blocker and exit 1. --allow-deferred still compiles all 721 current generated entries and the focused RFC-0036 controls, exits 0 as a diagnostic, and explicitly reports the inconsistency. There remain 718 deferred bodies, 715 mismatched citation reviews, and the other release gaps recorded in the full release log. An absent registry entry is not consistency certification for another pin.

Four focused validation controls pass. Their strengthened CLI decision-path control uses a synthetic zero-deferred fixture and mocked source/compilation plumbing to show that strict mode rejects the known pin while allow-deferred remains diagnostic-only. That mock is not Lean/compiler acceptance evidence; the separately recorded actual False compilation is the finding’s authority. The earlier weaker test source and its terminal output are preserved, and the final test exercises main() instead of comparing an imported helper to itself.

The earlier clean spec ed9c3b76a corpus generation completed 55/57 RFC catalogs, 43,205 RFC triples, 57 axiom records, and five inference-rule records. Its original log and generation protobuf are preserved here as historical metadata. That run neither established consistency nor authorized publication. Final source generation and site staging occur separately after the final clean pin.

Upstream repair acceptance

Use a valid Resource refinement or an explicitly guarded validity domain; eliminate the contradictory universal premise. Compile actual source-backed validity and contradiction controls and every affected consumer, then review and adopt a new immutable framework pin through the existing dependency process. This repair does not select or implement an upstream model, add a replacement axiom, change public Resource/identity semantics, certify another pin, or deploy anything. The current strict blocked status is retained explicitly.