Research proof index

This theorem covers the "happy path" -- showing that if optimized policy compilation succeeds, then verifyAlwaysAllows and verifyAlwaysAllowsOpt are equivalent.

The verifyAlwaysMatches analysis is complete: if the assertions verifyAlwaysMatches p εnv are satisfiable for the policy p and the strongly well-formed symbolic environment εnv, then there exists a strongly well-formed concrete environment env ∈ᵢ εnv such that the evaluator will not return .ok true when applied to p and env.

Alternate definition of completeness for alwaysMatches: For a singleton policyset, if symcc says the policy does not alwaysMatch, then there exists a concrete environment where the policy does not appear in determiningPolicies.

Concrete version of verifyAlwaysMatches_is_complete.

Concrete version of verifyAlwaysMatches_is_sound.

Alternate definition of soundness for alwaysMatches: For a singleton policyset, if symcc says the policy alwaysMatches, then the spec authorizer should say it always appears in determiningPolicies