Experimental software engineering should concentrate on software evolution

Abstract

For the future of software engineering it is critical that we devote sufficient energies to software evolution commensurate with its socioeconomic implications. We propose that the area of experimental software engineering focus more on software analysis and understanding to reflect the needs of large, evolving software systems properly. Since the results of evolution experiments do not necessarily scale up, we argue that the experiments should be performed in situ using large systems such as telephone switching systems, banking systems, or health information systems which evolve naturally over decades. We outline two evolution experiments which aim to measure time and effort spent for certain evolution tasks using different software analysis and understanding strategies. Finally, our position is summarized by three theses.