[Python-Dev] [Python-3000] The release process
Stephen J. Turnbull
stephen at xemacs.org
Sun Mar 2 04:29:28 CET 2008
Christian Heimes writes:
> It may sound like a dumb question by why do we need a release tool
> at all? I was involved in the release process of 3.0a2. Almost
> every step of the build process required human interaction.
Interaction, yes, but often it can be reduced to "Abort, retry, fail?"
> I don't want to diminish the effort that was put into welease
> though. But maybe (!) the same time spent for fixing some bugs
> would have helped the RM more.
Which RM? Barry was hoping to get some useful process support at very
low cost; that apparently didn't work out. But welease "is" Anthony's
RM process, and surely it he considered it an investment in on-going
quality or efficiency, or he wouldn't have done it.
The fact that Barry found Anthony's process unusable is IMO not a
reflection on either Barry or Anthony's code. Release processes seem
to be highly personal, even within the same project. My own project
(XEmacs) has 3 concurrent processes going at any one time (stable
core, unstable core, stdlib). In my time with the project, stable
core has seen two RM successions, unstable core has seen four, and
stdlib has seen two. In no case did the new RM adopt the tools of any
of his predecessors, but in two cases one person was a successor
twice, and in both cases they reverted to their old tools. All
processes seem to have been of roughly the same quality (my opinion,
there are no metrics available).
More information about the Python-Dev