Branch structure

[manulera]

Hi @BjornFJohansson,

I was thinking that maybe we could start doing the development like in Biopython, where changes are made in PRs to master, which has all branch protections instead of dev_bjorn. As far as I know, committing to master does not trigger a pypi release, so I think it would be OK to have master be the "next" codebase and pypi version the "stable" one.

What do you think? I mean this mostly because otherwise dev_bjorn is unprotected and can be force-pushed by accident. Putting protection in dev_bjorn would not be great, because then if it diverges from master there is no way to fix by rebasing if it has branch protections.

[BjornFJohansson]

I agree in principle, but I would like to investigate if the Biopython way is the most common way to set up a collaborative project? Is there a "standard way" or workflow akin to "Gitflow" for example?

[manulera]

The "master is dev, release is stable" is what I have seen in the repositories I have contributed:

• Gene ontology and other ontology repos
• BioPython
• OVE and other teselagen tools
• LinkML

I think gitflow might be a bit overkill for a library with few regular contributors. Also becaus most of what we add are incremental improvements and bug fixes, not so much new features. For the big new assembly model change I can keep a feature branch in parallel for the implementation when I get to it.