Concepts
What Git does for versioning, SciWIn does for provenance.
Every command you run through s4n — or every connection you draw in SciWIn-Studio — is recorded: the artifacts involved, the exact command, and how they connect to the steps before and after them. That record is a graph, not just a log, and it can be re-executed, inspected, or handed to someone else, on their machine or on a remote server. See Provenance & Publishing for what that record actually contains and where it can go from there.
Why not just a script?
Section titled “Why not just a script?”A shell script or a notebook cell also “does the work,” but it doesn’t describe itself: nothing in python analyze.py --input data.csv says what data.csv actually is, what produced it, or what the script depends on. Re-running it correctly six months later, or on someone else’s machine, means reconstructing all of that from memory, a README, or the shell history.
SciWIn’s tool and workflow descriptions are the Common Workflow Language (CWL) instead — an open, YAML-based standard for describing command-line tools and how they connect, portable across plain Docker, HPC schedulers, and Kubernetes, and readable by any CWL-compliant runner, not only SciWIn’s own. That’s the deliberate trade against writing the same pipeline directly in Snakemake or Nextflow: both are capable tools, but a Snakefile only means something to Snakemake. A CWL Workflow document means the same thing to cwltool, to REANA, to s4n execute, or to a colleague using none of the above. See CWL for what’s actually inside those files.
What that buys you
Section titled “What that buys you”- A re-executable record, not just an audit trail. How the provenance graph is built, annotated, and packaged for publishing is covered in Provenance & Publishing.
- A choice of execution backend without rewriting anything — locally, in a container, against a GA4GH TES server, or on a remote REANA instance.
- Two front ends over the same core — SciWIn-Client on the terminal, SciWIn-Studio on a visual canvas, operating on the exact same on-disk project. See Architecture for how the pieces underneath both fit together.