Department of
Applied
Mathematics

Absolute Correctness and Relative Correctness

Ali Mili

NJIT

Abstract

In this talk, we discuss the concept of relative correctness, i.e. the property of a program to be more-correct than another with respect to a specification. Whereas traditional absolute correctness distinguishes between two classes of candidate programs (correct vs incorrect), relative correctness ranks candidate programs on a partial ordering, whose maximal elements are the (absolutely) correct programs. We use the definition of relative correctness to define what is a fault in a program, and what is fault removal, as well as related concepts (fault density, fault depth, fault multiplicity). Also, we show that while the gold standard of program transformation is correctness-preserving transformations, the vast majority of software engineering practice does not involve correctness preservation but correctness enhancement; we show examples where such insights can be used to enhance current software engineering practice.