[ODE] From unstable to trunk...

[Adam D. Moss]

gl wrote:
> Given the state of the community, I'd be tempted to merge promising 
> patches directly into the main codebase. Yes, it will occasionally
> cause problems, but you will certainly get feedback on them : ).  And if 
> something doesn't get any feedback (as now), you can assume it works, at 
> least without detriment.

What we have now is very similar to the situation you describe,
though in theory better.  We have the UNSTABLE branch which lots more
people can check into at will without very much quality assurance,
then the good bits get transferred to the trunk.

We used to check patches into the trunk with (almost) reckless
abandon, but instabilities crept in which look a LOT of time and
effort to trace back to a particular change.  So we moved to a
system whereby the (very few :() patch moderators would try to
understand and test each patch reasonably thoroughly before
accepting it into the trunk.  The problem with that was that
it just took too much time, bottlenecked through too few people.

So what we have now in theory is the best of both worlds, with
the UNSTABLE branch being open to a lot more patch-wranglers
and acting as the first quality filter.  In practise it's
only been a mediocre improvement overall since there isn't much
feedback on the UNSTABLE branch (not that there was ever much
feedback from trunk) to base trunk decisions upon, plus it can
be painful to extract the good stuff from UNSTABLE.  Still an
improvement though.  This can be improved much more by more
people being willing to QA patches as they go into UNSTABLE and
then providing feedback on how those patches seem to be acting in
real life.

--adam
-- 
Adam D. Moss   -   adam at gimp.org