Macroscopic characterization of software and its relationship to reliability

This letter presents a model for a large software system that represents it as a statistical ensemble of variables embedded in a logical environment that controls their microscopic behavior. Using a statistical mechanics approach to this model, the macroscopic attributes of software that are analogous to the volume, pressure, and temperature of physical systems are identified. This provides a framework for the conceptual unification of the various prevailing techniques for improving software reliability (e.g., modularity, object orientation, and software reuse), all of which are ways of achieving what we call software hardening.