[darcs-users] On Release Management

Guillaume Hoffmann guillaumh at gmail.com
Tue Nov 16 13:01:38 UTC 2010


Hi Reinier,

there's a thesis written in 2007 called "Quality Improvement in
Volunteer. Free and Open Source Software Projects. Exploring the
Impact of Release Management". It studies time-based releases in
volunteer FOSS software. Basically, the idea is that time-based
releases are the best solution for a project like Darcs, where
everyone is a volunteer and there is enough work done in the lapses of
6 months between releases, so that each release is interesting in
itself. Otherwise, feature-based releases come out of the blue, nobody
knows really how to schedule their work, and releases are usually
delayed.

For this process to happen well, the author gives some conditions,
among which: it is easy to rollback a feature that is not ready for
the release.

This is problably here that we had a problem with the 2.5 release. It
should have been released on time without the problematic features,
and the development should have continued normally after that. The
lack of modularity of Darcs's code may explain why we are having
trouble at doing this, so maybe this something we should improve for
2.8.

Guillaume


More information about the darcs-users mailing list