eing Suspicious: Criti uing Problem Specifications
Stephen F. Fickas, Pranav Nagarajan · 1988
One should look closely at problem specifications before attempting solutions: we may find that the specifier has only a vague or even erroneous notion of what is required, that the solution of a more general or more specific problem may be of more use, or simply that the problem as given is misstated. Using software development as an example, we present a knowledge-based system for critiquing one form of problem specification, that of a formal software specification. Suppose one were given a problem P to solve. Suppose further that it was known that generating a solution S for P will require a large effort. The question we ask in this paper is a pragmatic one: should we take P on blind faith and forge ahead, or should we scrutinize P carefully before committing resources to its solution? We will argue for the latter approach. More specifically, we propose that three types of specification critiques can account for a useful and interesting set of specification errors: