A knowledge-based approach to scientific software development: position paper

As a relatively mature field, scientific computing has the opportunity to lead other software fields by leveraging its solid, existing knowledge base. Our position is that by following a rational design process, with the right tool support, desirable software qualities such as traceability, verifiability, and reproducibility, can be achieved for scientific software. We have begun development of a framework, Drasil, to put this into practice. Our aims are to ensure complete traceability, to facilitate agility in the face of ever changing scientific computing projects, and ensure that software artifacts can be easily and quickly extracted from Drasil. In particular, we are very interested in certifiable software and in easy re-certification. Using an example-based approach to our prototype implementation, we have already seen many benefits. Drasil keeps all software artifacts (requirements, design, code, tests, build scripts, documentation, etc.) synchronized with each other. This allows for reuse of common concepts across projects, and aids in the verification of software. It is our hope that Drasil will lead to the development of higher quality software at lower cost over the long term.