The ARTS Open Framework

The talk starts from a common research problem: papers link to data and code, but those pieces are often hard to run or even reach years later, and some government datasets have disappeared from their hosts. Although computational workflows in many fields look alike, from collecting data to analyzing it with code and publishing figures, there is no widely adopted standard for regenerating those analyses across different machines.

The presenter reviewed existing options, including a commercial reproducibility platform, a neuroscience-specific system and an academic project that has gone quiet, and found gaps in openness, scope or upkeep. The proposed framework requires a license and a configuration file, and recommends version control with merge rather than rebase, container files and built images, and trusted RFC 3161 timestamps to prove when data was collected. A demo used a recording of a cheap mechanical watch to show drift analysis, then opened the same environment in a hosted notebook from a shared link.

In the question period, attendees discussed funding mandates and whether overworked researchers will take on extra packaging work. The answer leaned toward setting good examples in well-resourced labs and sharing standard environments, so newcomers can run someone else's analysis without building it themselves.