Requirements for a Collaborative and Efficient Scientific Software Support Architecture

RSEs accelerate research by developing reusable software. Yet the intended wide use of research software also increases the demand for support. To help users of libraries that run on client systems, developers often have to reproduce issues locally, which requires information about system configurations, dependencies, and software versions. This process does not scale well as the number of users grows and can become taxing for commonly small development teams in academic settings.

An example of such a research software is pyGIMLi, a user-side library for modeling and inversion in geophysics that is developed openly on GitHub. Since the publication of version 1.0 [1], the software has been widely adopted and has enabled research resulting in 2-3 scientific publications or theses per week in 2024. The increased use results in an increasing number of GitHub issues, often missing minimal working examples, challenging reproducibility and quick software support. The characteristics of pyGIMLi — a small team of researchers as developers, user-side library or executable, on-demand feature development, and heterogeneous application contexts — are shared by many scientific codes and quite often result in similarly taxing support demands. Therefore, we see a broad demand for an efficient and collaborative support workflow that gives software-developing researchers more time for research.

To meet this need, we introduce CAES³AR (Collaborative And Efficient Scientific Software Support ARchitecture), which aims to provide such an infrastructure. We propose creating specifications for reproducible environments of issues based on repo2docker’s REES. This enables automated tests in the background to check whether the problem occurs only in specific versions or platforms. Furthermore, supporters will be able to directly and collaboratively help fix the problem by running the code in an execution environment like JupyterHub.

To gather the requirements for this infrastructure from both the user and developer perspectives, we will hold a workshop in December 2025. We will present the results of this workshop along with conclusions for the design of the CAES³AR workflow and its technical specifications.

While the initial prototype of CAES³AR will be developed with and for pyGIMLi, we believe that the challenges of supporting scientific software libraries are common across domains and look forward to discussions with the community.

[1] Rücker, C., Günther, T., & Wagner, F. M. (2017). pyGIMLi: An open-source library for modelling and inversion in geophysics. Computers and Geosciences, 109, 106–123. https://doi.org/10.1016/j.cageo.2017.07.011

Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) through the CAES³AR project (Projektnummer 561181781).