Software Design Spaces: Logical Modeling and Formal Dependence Analysis

Abstract

We lack a useful, formal theory of modularity in abstract software design. A missing key is a framework for the ab- stract representation of software design spaces that sup- ports analysis of design decision coupling structures. We contribute such a framework. We represent design spaces as constraint networks and develop a concept of design de- cision coupling based on the minimal change sets of a vari- able. This work supports derivation, from logical models, of design structure matrices (DSM�s), for which we have a promising but inadequate theory of modularity. We present complexity results and a brute force algorithm. To test for potential software engineering utility, we analyzed the de- sign spaces of Parnas�s 1972 information hiding paper, with positive results that were surprising in several ways.