[vlc-devel] Releases plan?

Rémi Denis-Courmont remi.denis-courmont at nokia.com
Mon Mar 10 19:45:30 CET 2008


I cannot say I am at ease with branching now. And I am a bit suspicious about 
the decision process (FOSDEM) that seems to have excluded most of the most 
active developers and builders (xtophe, funman, pdherbemont, FK and me, to 
name a few).

Branches are painful. They just are. git is easier to do backporting with, but 
it's not automagical either. More importantly, stabilization branches SIMPLY 
DO NOT work if they last "too long". The longer they last, the bigger the 
divergence with trunk, the bigger the intricacy of backporting fixes. And the 
longer, the fewer developers actually bother to build and test the stable 
branch, in addition to -trunk, and now in addition to -bugfix.

Empirically, I would say "too long" is a month or five weeks, or so. It 
typically takes 1-2, sometimes 3 weeks to fix large regressions due to the 
certain -trunk changes that stabilization branches are meant to avoid. Once 
the timespan of the branch far exceeds the time it takes to fix a set of 
regressions, it does not make any sense anymore. This is not theoretical. 
It's already happened. 0.8.3 should be a well known example - if you were in 
the project in 2005 that is.

As far as I can see, Qt4 is the only huge blocker at this point, unless we 
release 0.9.0 with WxWidgets (which has been maintained by Gildas, Christophe 
et al inspite of all the API breakages from Pierre and I). I fail to see how 
branching the whole VLC will fix or speed up the fixing of Qt4 problems. 
Sure, I am not involved with Qt4, so I would not know.

We have branched, I assume it means 0.9.0 will have been released by mid-April 
_in_any_case_. And that those most affected, meaning the Qt4 devs, and the 
builders, _WILL_ make sure that does happen.

Rémi Denis-Courmont

More information about the vlc-devel mailing list