diff --git a/docs/PROOF-NEEDS.adoc b/docs/PROOF-NEEDS.adoc index 34bed899..c331daf2 100644 --- a/docs/PROOF-NEEDS.adoc +++ b/docs/PROOF-NEEDS.adoc @@ -80,7 +80,7 @@ no collisions in kind_to_u8 |I2 |E9 Proof composition soundness |`+meta-checker/src/Echidna/Composition.agda+` |Cut preserves semantic soundness; explosion shows a conflict check is -load-bearing; a sound detector realises every composite, and a sound +load-bearing; a sound detector realises every unflagged composite, and a sound detector that decides its conflict relation returns a conflict witness for every inconsistent one. Theory over an arbitrary logic; no Rust composer exists to verify |Agda diff --git a/docs/PROOF-STATUS.adoc b/docs/PROOF-STATUS.adoc index 65ced1b3..8bd1a297 100644 --- a/docs/PROOF-STATUS.adoc +++ b/docs/PROOF-STATUS.adoc @@ -182,8 +182,9 @@ shipped code. *Verus:* not started; no Verus specification or proof exists. solver in four ways (an `+Uninitialized+` status, a prover missing from the manifest, no checker configured, and a `+verify_all()+` error read as an empty report), and runs the integrity step after proving, not first. -Row stays `+open+` until a Creusot or Verus run on shipped code yields a receipt -with a commit hash. +Row stays `+open+` until a Creusot run *and* a Verus run on shipped code each +yield a receipt with a commit hash. Both are the default Rust toolset (§1.2); one +leg can be dropped only by a recorded §4 exception, not by the other passing. *Trust kernel (2026-11-09).* `+compute_trust_level+` exists once, in `+crates/echidna-core-creusot+`; `+src/rust/verification/confidence.rs+` @@ -281,19 +282,20 @@ re-title them; do not count them as evidence. |R3. Lean 4 legs: E4-L4 `+ConfidenceLattice.lean+`, E10 `+ParetoMaximality.lean+` / `+ParetoStrongMaximality.lean+`, E11 `+IntegrityVerification.lean+`, E13 `+PortfolioCompleteness.lean+` -|`+open+` (the Lean build passes; the verifier placement is unresolved) +|`+open+` (the Lean build passes; owner ruling 2026-11-16 splits it by property) |*Corrected 2026-11-09.* The 2026-05-18 record (all five failing) is stale. https://github.com/hyperpolymath/echidna/actions/runs/42624461538/job/129184535185[Lean 4 (verification)] builds all five with `+lake build+` on the pinned `+lean-toolchain+` (4.13.0), success, on commit `+a2a8643+` (tree identical to `+main+` `+2ab7574+`). A 2026-11-09 audit run of `+#print axioms+` over 77 of the 79 theorems found only `+propext+`, `+Quot.sound+` and `+Classical.choice+` — no `+sorryAx+`. -|*Placement is still open.* These statements are about ECHIDNA's own types -(confidence lattice, Pareto order, integrity, portfolio), so §1 assigns them -to Idris 2 (or Creusot/Verus on the Rust), not Lean. Calling them "§4 +|*Placement decided 2026-11-16 (split by property).* These statements are about +ECHIDNA's own types (confidence lattice, Pareto order, integrity, portfolio), so +§1 does not assign them to Lean. Calling them "§4 exceptions" is not enough: no §4 exception record (reason, alternative -considered) exists beside them. Owner call: write the §4 records, or move the -statements to Idris 2. Being engineering proofs, they carry no PWG (§8.1). +considered) exists beside them. Owner ruling 2026-11-16: each statement is +placed by property (see the classification paragraph under R4), not moved +wholesale. Being engineering proofs, they carry no PWG (§8.1). *Audit 2026-11-09.* What the models say about the code, for R1: `+ConfidenceLattice+` matched the pre-2026-11-09 trust rule branch for branch and now describes the old rule; `+IntegrityVerification+` PI-11 (`+solver_integrity_ok+` iff @@ -335,7 +337,7 @@ here. |R4. Agda meta-checker: trust-level order, axiom safety, portfolio, dispatch, axiom-tracker completeness, integrity, clamp bounds, receipt/warrant (9 modules) -|`+proven+` +|`+proven+` (the modules other than `+Integrity+` and `+ClampTrust+`, which R9 and R10 carry as `+open+`) |First-party Agda in `+meta-checker/src/+`, type-checked with the §2.1 canonical invocation. *Current receipt (exact commit):* the scheduled Agda Meta-Checker run @@ -404,8 +406,26 @@ statement that holds for an arbitrary logic or order, such as E9 composition Creusot/Pearlite and Verus (R1). They are not restated in Idris 2: the strict §1 ruling that #477 applies retires Idris restatements of Rust behaviour (it removes `+TrustKernelMonotonicity.idr+`). Each Agda module retires once its -Rust contracts carry a receipt. Until then these rows keep their `+proven+` -token as statements about a model. They carry no PWG. +Rust contracts carry a receipt. Until then R4 keeps its `+proven+` token, as +statements about a model, for its modules other than `+Integrity+` and +`+ClampTrust+`; R9 and R10 carry those two as `+open+`. That token attests +only that the Agda kernel accepted these statements about a model; it is not +evidence about the Rust. The design review below shows several of them prove +less than their names say. Which statements carry forward, and to which home, +is the per-statement placement; a statement placed `+retire+` counts for nothing +beyond this Agda check. They carry no PWG. + +**Placement by property (owner ruling 2026-11-16).** An earlier ruling moved +R3, R9 and R10 to Idris 2, which contradicted the paragraph above (Codex on +#514). The owner resolved it: each statement goes where its property lives. +A property of one shipped Rust function or type is proved on the Rust with +Creusot/Pearlite and Verus. A fact about an interface, ABI or wire encoding +goes to Idris 2. A statement that holds for any implementation stays Agda +theory. A statement that proves nothing about the system (true by definition, +an instance of `+cong+`, or a model of behaviour the Rust lacks) retires, with +its successor named. This applies to R3, R4's engineering modules, R9, R10 and +the Idris 2 verification corpus. The per-statement placement is recorded +separately. **Design review of the statements (2026-11-09, at `+2ab7574+`, before #477 replaced the trust kernel and retired `+TrustKernelMonotonicity.idr+`):** the kernel accepting a proof @@ -436,7 +456,7 @@ by changing a token. `+clamp-idempotent+`. |R9. E11 solver-integrity soundness -|`+proven+` +|`+open+` (the Agda module checks; owner ruling 2026-11-16 splits it by property) |`+meta-checker/src/Echidna/Integrity.agda+`, inside the R4 receipt. States the obligation that had no Agda record: verification requires an initialised manifest; verification means the recorded digest was matched; a binary with @@ -462,16 +482,19 @@ comparison, not a machine-checked discrepancy or refinement proof. The Rust implementation leg remains open under R1's Creusot obligation. |R10. ClampTrustBounds -|`+proven+` +|`+open+` (the Agda module checks; owner ruling 2026-11-16 splits it by property) |`+meta-checker/src/Echidna/ClampTrust.agda+`, inside the R4 receipt. Restates the four obligations of `+verification/proofs/idris2/ClampTrustBounds.idr+` (bounds, monotonicity, fixed points, and — added — idempotence, i.e. that the clamp lands on a real level rather than between two) over the trust-level type the gate actually checks. |This row establishes finite-order algebra in the Agda model. It does not -validate the Idris2 file or its correspondence to runtime code. Existing -Idris2 interface and engineering checks remain active under §1; their -receipt status is R2. Monotonicity here is by step lemma plus transitivity, +validate the Idris2 file or its correspondence to runtime code. The Idris 2 +file states bounds, monotonicity and fixed points of a `+Nat -> TrustLevel+` +clamp, which the Rust does not have (R2 audit), so those statements are not +evidence for the Rust `+clamp_trust+` and do not count under R2. Under the +2026-11-16 split they retire, with the Rust `+clamp_trust+` contract below as +their successor. Monotonicity here is by step lemma plus transitivity, rather than the Idris2 file's decision table. **Model and code differ:** the Rust has `+echidna_core_creusot::clamp_trust(level, min_level, max_level)+` @@ -666,11 +689,17 @@ mean "ECHIDNA is verified" — it would mean one named obligation was discharged by one named tool with its boundary written down. Nothing here vouches for a prover backend's own soundness. -The `+proven+` rows — R2, R4, R9, R10, R11 — each name the kernel, the -receipt and the limit that qualify them (R2 the receipt-time escape-hatch -audit, R4 the `+UnsupportedIndexedMatch+` caveat, R9 the digest-level scope). -R3 stays `+open+`: its Lean build passes, but Lean is not the verifier §1 -assigns to those statements until the placement is decided. Only theory theorems are graded (§8.1): R11 is at +The `+proven+` rows — R2, R4 (less `+Integrity+` and `+ClampTrust+`), R11 — each name the kernel, the receipt and the +limit that qualify them (R2 the receipt-time escape-hatch audit, and the fresh +receipt that `+NatLte+` totality still needs; R4 the +`+UnsupportedIndexedMatch+` caveat). R9 and R10 are engineering properties of +ECHIDNA's own types. Owner ruling 2026-11-16 places each of their statements by +property (see the classification paragraph under R4), so their Agda checks do +not carry `+proven+`. A row closes when each surviving statement has a receipt +in its home; a statement placed `+retire+` closes through its recorded +retirement and its named successor, never through a receipt. +R3 stays `+open+`: its Lean build passes, but Lean is not a verifier §1 +assigns to those statements; the 2026-11-16 split gives each its home. Only theory theorems are graded (§8.1): R11 is at `+PWG-3+`; the engineering rows carry no PWG. Every *independent* leg is still `+open+`: R5 has never been attempted, which is why nothing here is above `+PWG-3+`. Proven in one kernel is exactly what `+PWG-3+` means, and neither fact diff --git a/docs/reports/2026-11-01-routing-and-rule-engine-recon.adoc b/docs/reports/2026-11-01-routing-and-rule-engine-recon.adoc index e4655fcc..5f033b77 100644 --- a/docs/reports/2026-11-01-routing-and-rule-engine-recon.adoc +++ b/docs/reports/2026-11-01-routing-and-rule-engine-recon.adoc @@ -150,6 +150,10 @@ some other system publishes it. link:../PROOF-STATUS.adoc[`+docs/PROOF-STATUS.adoc+`] records Agda rows R4, R9, R10 and R11 as `+proven+` for the sources identified in link:2026-11-04-agda-meta-checker-receipt.json[the Agda receipt]. + _Note 2026-11-17:_ that record no longer stands for R9 and R10. Owner ruling + 2026-11-16 places their statements by property, and both rows are now + `+open+` (see the classification paragraph under R4 in + link:../PROOF-STATUS.adoc[`+docs/PROOF-STATUS.adoc+`]). That receipt identifies a base commit plus exact checked file digests; it does not certify an arbitrary later checkout or the operational Rust. Other obligation statuses remain governed by their individual ledger rows. @@ -944,7 +948,9 @@ Three new modules, all inside the receipt above: === 10.4 What the proof grade did and did not move -*Home-system leg:* R4, R9, R10, R11 are `+proven+`. *Independent legs:* R5 +*Home-system leg:* R4, R9, R10, R11 are `+proven+`. _(Note 2026-11-17: R9 and R10 +are `+open+` since owner ruling 2026-11-16 placed their statements by property; +R4 keeps `+proven+` less `+Integrity+` and `+ClampTrust+`.)_ *Independent legs:* R5 (Rocq + Lean) is untouched and `+open+`. So the corpus is still **Mathesis Grade 3** — locally checked, untranslated — and this session did not move it, because nothing that establishes Grade 2i/2ii was performed. What moved is the policy