A formal model for automated software modularity and evolvability analysis

Abstract

Neither the nature of modularity in software design, characterized as a property of the structure of dependencies among design decisions, or its economic value are adequately well understood. One basic problem is that we do not even have a sufficiently clear definition of what it means for one design decision to depend on another. The main contribution of this work is one possible mathematically precise definition of dependency based on an augmented constraint network model. The model provides an end-to-end account of the connection between modularity and its value in terms of options to make adaptive changes in uncertain and changing design spaces. We demonstrate the validity and theoretical utility of the model, showing that it is consistent with, and provides new insights into, several previously published results in design theory.