Measurement of change: stable and change-prone constructs in a commercial C++ system

Abstract

Our previous studies of developers' ability to predict software change revealed a great potential for improvement of the change management process. What we consider most beneficial is characterizing and understanding software change by measuring it and identifying what kind of changes take place and how frequent they are. With such knowledge, it is possible to build change models that help developers make better predictions regarding future requirements. An analysis is presented of a commercial object-oriented system-the PMR (Performance Management traffic Recording) system of Ericsson Radio Systems-that was changed significantly due to the implementation of a set of new requirements. Measures are used in order to identify stable and change-prone constructs of the system. The results from the analysis are complemented with results from interviews with developers about what changes most frequently in a C++ system.