Process expressions and Hoare's logic: Showing an irreconcilability of context-free recursion with Scott's induction rule

Abstract

In this paper processes specifiable over a non-uniform language are considered. The language contains constants for a set of atomic actions and constructs for alternative and sequential composition. Furthermore it provides a mechanism for specifying processes recursively (including nested recursion). We consider processes as having a state: atomic actions are to be specified in terms of observable behaviour (relative to initial states) and state transformations. Any process having some initial state can be associated with a transition system representing all possible courses of execution. This leads to an operational semantics in the style of Plotkin. The partial correctness assertion {α} p { β } expresses that for any transition system associated with the process p and having some initial state satisfying α, its final states representing successful execution satisfy β. A logic in the style of Hoare, containing a proof system for deriving partial correctness assertions, is presented. This proof system is sound and relatively complete, so any partial correctness assertion can be evaluated by investigating its derivability. Included is a short discussion about the extension of the process language with “guarded recursion”. It appears that such an extension violates the completeness of the Hoare logic. This reveals a remarkable property of Scott's induction rule in the context of non-determinism: only regular recursion allows a completeness result.