On specifications, subset types and interpretation of proposition in type theory

Abstract

In this paper we discuss some practical aspects of using type theory as a programming and specification language, where the viewpoint is to use it not only as a basis for program synthesis but also as a programming language with a programming logic allowing us to do ordinary verification. The subset type has been added to type theory in order to avoid irrelevant information in programs. We give an example of a proof which illustrates the problems that may occur if the subset type is used in specifications when we have the standard interpretation of propositions as types. Harrop-formulas and Squash are then discussed as solutions to these problems. It is argued that they are not acceptable from a practical point of view. An extension of the theory to include the two new judgment forms:A is a proposition, andA is true, is then given and explained in terms of the old theory. The logical constants are no longer identified with the corresponding type theoretical constants, but propositions are interpreted as Godel formulas, which allow us to introduce and justify logical rules similar to rules for classical logic. The interpretation is extended to include predicates defined by using reflections of the ordinary definition of Godel formulas in a “type of small propositions”. The programming example is then revisited and stronger elimination rules are discussed.