Developing a Handbook of best practices for research software sustainability at Princeton

In the early 2010s, the Software Sustainability Institute defined sustainability as "software you use today will be available — and continue to be improved and supported — in the future" [1, 2]. Sustainable research software has enabled research that would otherwise be infeasible. The pursuit of sustainability continues, spanning efficiency, longevity over generations of research projects, and spread across research fields. In this context, the Princeton Research Software Engineering (RSE) group, as one of the first and largest centralized RSE groups in US institutions [3], has both the breadth of RSE experience and the team bandwidth to further advance sustainability. In the Princeton RSE group, we are developing a handbook of best practices for research software sustainability. As sustainability is inherently contingent [4], it is not possible to prescribe criteria for it. Instead, the Handbook enumerates characteristics of sustainable research software. Software with these characteristics substantially increases its chances of being sustained. Sustainable research software has reliable tests as its technical foundation; they enable iterative cycles of refactoring and feature addition and also underpin confidence in AI-assisted and agentic development, yielding frequent releases that progressively improve maintainability, reproducibility, and other technical qualities. These technical characteristics are necessary but not sufficient. Long-term sustainable research software also has social and organizational characteristics, including citability, continuous funding, professional maintenance, and active communities. The Handbook collects best practices, based on our experience, that realize these characteristics, from basics to advanced topics such as how to turn unsustainable research code into a testable package. The Handbook facilitates communication with principal investigators and research project managers; the best practices describe what we do and how we identify and resolve software engineering issues within research projects. In this talk, we will present our approach to building the Handbook; discuss what distinguishes it from other best practices guides, for instance, its grounding in the operational practice of a centralized, multidisciplinary RSE group; review the practices at a high level; and suggest ways for others in the community to leverage the Handbook in their organizations. The Handbook will be publicly available under a Creative Commons license. The Handbook is not a static document that we publish and finish; we develop it as we develop sustainable software, with continuous updates as demanded, most frequently by the rapid evolution of AI tools and practices.