Software development risk model - Applied to data from open-source Mozilla project

Development of large software systems creates many, often thousands, of source code files with complex inter-dependencies. Clusters of mutually dependent files introduce the possibility of a chain of forced consequential changes when a single cluster member file is changed. Our software development risk model shows that density of dependencies within such clusters plays a crucial role in this behavior. We develop a file-rank procedure which orders the entire system's file set by increasing risk. This ranking process should prove to be useful while managing the development of large systems, indicating where attention should be focused to improve Test Risk. We have applied this model to a library from the 1.4.1 release of the open source Mozilla project with interesting results.