Differentiators and detectors

Abstract Given two programs, say P and Q, a differentiator is a test suite T such that execution of P and Q on T produces different outcomes; this concept is used in mutation testing, where it is important to highlight semantic differences between non-equivalent mutants. Given a specification R and a program P, a detector is a test suite T such that execution of P on T disproves the correctness of P with respect to R; this concept is clearly useful in program testing. In this note I elaborate on previous research to refine and characterize these concepts.