Maintaining Rocker: Sustainability for Containerized Reproducible Analyses
Carl Boettiger (University of California, Berkeley)
To put Rocker, the de facto standard for reproducible, containerized R analyses, on a path to sustainable maintenance through refactoring, improving the quality of documentation, expanding the community, and targeting new hardware platforms.
The Rocker project provides a suite of standardized, containerized computing environments that enable computational reproducibility for scientific analyses in R. These carefully curated and version-harmonized containers assure that scientific results can easily be moved across platforms, reproduced across operating systems, scaled up to HPC systems, and re-run years later even as software becomes out of date. With over 6 million downloads to date, Rocker containers have become the de-facto standard environments for work in the R language. Rocker images are widely relied upon by large public scientific infrastructure and individual research teams, by universities, instructors and private companies.