The sustainability of scientific software: Ecosystem context and science policy

摘要: The sustainability of scientific software is a key challenge for science policy. We approach this question by drawing on empirical studies of scientists using software and describe how components are arranged with complements and dependencies into value-‐providing assemblies, periodically revisited by their scientist users. Over time, software declines in scientific usefulness, driven by four factors: a moving scientific frontier and technological change, production friction, use friction and the software ecosystem context. In particular we highlight the impact of the complexity of ecosystem context, in terms of the diversity of use-‐contexts in which a component is used. We identify three broad strategies to address the need for work to sustain the usefulness of scientific software: suppress the drivers, reduce the amount of work needed, or attract sufficient resources able to undertake the work needed to sustain scientific usefulness. We examine three attraction systems: commercial markets, community-‐based peer-‐production and grant-‐making. We describe how these systems bring resources to projects, and particularly highlight how both commercial markets and peer production address the challenges of ecosystem complexity while scientific grant-‐making does not. We conclude by making science policy recommendations to address the challenges of sustainability, by enhancing the grant-‐making system and by facilitating transitions to other resource attraction systems.