Building Bodies of Knowledge about Software Development Practices

Опубликовано 6 сентября 2016, 17:02

My experience has been that there are no such things as ΓÇ£best practicesΓÇ¥ in software engineering ΓÇô Although certain practices have proven very effective at reducing cost or improving quality on some types of projects, no practice is a universal best choice for all projects. Selecting practices that are appropriate for the project at hand is a necessary step for achieving project success. Empirical study helps support this endeavor by providing the means to make decisions about which practices to apply in an evidence-based manner, that is, based on the results seen for practices as implemented by real projects, not by gut feeling, fad, or hype. However, the current body of knowledge in empirical software engineering is not yet sufficient to make decision support about software practices an easy endeavor. In many cases such guidance, based on sound experience, is missing or incomplete ΓÇô maybe because the practice is too new, still under study, or the existing experiences do not fit the userΓÇÖs context. Yet techniques do exist that enable the existing data to be leveraged to provide data-driven recommendations to practitioners and help direct further research. In this talk I will present a research approach which we have been applying to develop, evolve, and ultimately transfer into use effective software development techniques, for tasks such as software verification & validation (V&V) and the development of high-performance code for scientific computing. I will emphasize how different types of studies ΓÇô from case studies to tightly focused controlled experiments ΓÇô can produce rigorous results that assist in tech transfer. I will show some examples of how such studies can build upon one another, by design, to create a useful body of knowledge. And, I will discuss some techniques that can be used in the more usual case, which can be applied to studies that have not been designed to be part of such a family, and discuss examples of actionable results have been achieved applying them. Finally, I will present an overview of work I have been conducting for the United States Department of Defense to build a ΓÇ£Best Practices Clearinghouse.ΓÇ¥ The Clearinghouse applies this approach as the basis for a decision support tool designed to improve the DoDΓÇÖs acquisition of software-intensive systems.