Plan, apply and roll back changes to a production ontology, with a blast radius report and a proof an auditor can re-check.
Rust ai-nativedescription-logicsformal-verificationinfrastructure-as-codeisabelle-languageknowledge-graphleanlean4mcpmcp-serverontologyowlowl2reasoningrustshaclsparqltheorem-proving
readme: the limits section said three things that stopped being true an hour ago
FABIOTESS893c83cceaa8
diff --git a/README.md b/README.mdindex bb4b5a6d..8d4aaedd 100644--- a/README.md+++ b/README.md@@ -108,21 +108,27 @@ is named property by property in [docs/trusted-computing-base.md](docs/trusted-cproperty-tested, and partly bounded-model-checked. It is not verified. This work verified thejoint, not the machine.-**The Lean assumes eight of the rule-semantic conditions rather than deriving them.** For those-arms the theorem says the rule is sound because a field of a condition record says so, which is-close to no machine-checked content at all. The independent Isabelle derives them from the-specification tables, which is how we know the assumption was avoidable. Being fixed; not fixed yet.--**Two kernels agreeing is weaker than it sounds.** The Lean and Isabelle model classes have not-been related tightly enough for "both said entailed" to mean what a reader would naturally take it-to mean. They are theorems over model classes nobody has ordered in either direction. Separately,-Lean's model class is LARGER than the specification's, so entailment transfers outward and-NON-entailment does not: the countermodels prove less than they appear to.--**A non-vacuity witness can be vacuous where it matters.** Isabelle's satisfies many of its-conditions only because the relevant extensions are empty, including every condition its own-authors call the sharpest trap. A witness that is vacuous exactly where the conditions are-strongest establishes very little, and it is the failure mode that most resembles success.+**The bridge to the specification is prose, not a theorem.** Fourteen rule arms used to be+assumed rather than derived, and they are now derived, each with an empty axiom footprint, over a+structure carrying the W3C conditions at full strength. What is still assumed is the reading that+turns a conforming interpretation into one of ours, and five facts from the axiomatic-triple+tables that reading needs. Formalising the specification side properly means building its whole+interpretation structure in core Lean, and that has not been done.++**Two kernels agreeing is weaker than it sounds.** They are theorems over model classes nobody has+ordered in either direction, so "both said entailed" means less than a reader would naturally take+it to mean. Lean's class is still larger than the specification's, so entailment transfers outward+and non-entailment transfers only where it has been shown to: four of the negative results now+carry over, and three do not, each of those three carrying the field that fails as a checked+theorem.++**A non-vacuity witness can be vacuous where it matters**, and ours was. It satisfied nine of+twenty-one conditions only because the relevant extensions were empty, and six of the fourteen arms+rested entirely on those, which is the failure mode that most resembles success. It is rebuilt, and+the gate is now stronger than non-emptiness: every one of the fourteen derivations must fire at a+concrete instance. Five conditions remain vacuous, they carry no arm, and one of them cannot be+exercised by any model at all because the specification makes that relation the diagonal. The+Isabelle witness has not had the same treatment.**Negative answers are mostly unchecked opinion.** A refutation cannot be replayed in core Lean, soan "unsatisfiable" from any prover is testimony, not a certificate. One clash rule has a semantic