Title:Software Engineering as Instrumentation for the Long Tail of Scientific Software

Abstract:The vast majority of the long tail of scientific software, the myriads of tools that implement the many analysis and visualization methods for different scientific fields, is highly specialized, purpose-built for a research project, and has to rely on community uptake and reuse for its continued development and maintenance. Although uptake cannot be controlled over even guaranteed, some of the key factors that influence whether new users or developers decide to adopt an existing tool or start a new one are about how easy or difficult it is to use or enhance a tool for a purpose for which it was not originally designed. The science of software engineering has produced techniques and practices that would reduce or remove a variety of barriers to community uptake of software, but for a variety of reasons employing trained software engineers as part of the development of long tail scientific software has proven to be challenging. As a consequence, community uptake of long tail tools is often far more difficult than it would need to be, even though opportunities for reuse abound. We discuss likely reasons why employing software engineering in the long tail is challenging, and propose that many of those obstacles could be addressed in the form of a cross-cutting non-profit center of excellence that makes software engineering broadly accessible as a shared service, conceptually and in its effect similar to shared instrumentation.