Domain-based change propagation analysis: An enterprise system case study

Abstract

Change propagation has mainly been estimated by maintenance history or source code analysis. However, sometimes history and code are inaccessible, or impractical to analyse, such as for heterogeneous sources. Previously we hypothesised that change propagation from modifying domain level components may be predicted purely from information available to domain users. We proposed domain-based change propagation analysis, enabling analysts and domain experts to predict conceptual coupling independent of implementation. This paper reports on application of domain-based analysis to a significant (enterprise) system. We performed both domainbased analysis and a well known history-based analysis and compared the results. Like history-based approaches, domainbased analysis reveals coupling between software components, can assist to prevent errors in software maintenance, and predict change propagation. We conclude that it may be worth applying to certain kinds of systems where established approaches would be considered impractical.