[vlc-devel] Releasing too slow

Rémi Denis-Courmont remi at remlab.net
Sun Jul 5 20:41:49 CEST 2009


I think I already mentioned it, but I am very dissatisfied with the 1.0 
releasing timeline.

First, there has been way too little informations on what was going on. 
Basically, I have almost never known what was blocking the release at a given 
point in time.

Second, it has been way too slow. Most recently, there have been so few and 
minor bugfixes, that I really wonder why the release was not already done. It 
makes me think we are waiting for things that nobody wants to fix, which is 
utterly pointless. Even the 0.9.0 freeze, lasting 10 weeks, was far shorter:
03-19 1.0.0-pre1
04-17 1.0.0-pre2
05-09 1.0.0-rc1
05-27 1.0.0-rc2
06-05 1.0.0-rc3
06-17 1.0.0-rc4
07-?? 1.0.0??
That is already more than 15 weeks and we're not even done yet, even though 
1.0 has far less disruptive changes than 0.9 had...

Why did we wait four weeks between each of the first four previews? It used to 
be two weeks which makes much more sense to me. Why is 1.0.0 still not out, 
even though even 1.0.0-rc1 (and definitely 1.0.0-rc2) was, as far as the 
source code is concerned, I think, release-grade?

IMHO, the proper time for a freeze is a month or so after forking. The quality 
of backported fixes decreases with time, as the trunk diverges from the frozen 
branch. We do not have enough quality assurance (if any) and dedicated bug 
fixing resources that we could really afford long stabilization periods found 
in more commercially-driven projects.
Also, the new features in trunk start getting stable, such that the earlier 
release makes no sense (c.f. the 0.8.3 unreleased branch).

IMHO also, we could as well have released 1.0.0 two months ago or almost (say 
shortly after 1.0.0-rc2). It seems to me that almost none of the fixes since 
then addressed regressions, which is supposed to be the whole point! That is 
not to say we should not fix old bugs, but they should not block releases 
either. Then, we (as usual) have the contribs. Again, I do not see why that 
should block source code releases. The Linux/BSD people are suffering from 
*BOTH* the Windows releasing delay and the distribution packaging delay. This 
makes no sense. With 0.9, we had the Windows binaries later than the source 
code, which makes a PERFECT SENSE considering the never-ending contribs 
delays. There has now been evidence that I am not the only one being 
frustrated by this.

Frankly, I am getting quite demotivated. The lack of release predictability is 
demotivating, and so are the lack of visibility and timeliness. And it appears 
to hurt our credibility in the open-source community in general. Either we get 
someone to do source code release management seriously, or we might as well 
give up.

Rémi Denis-Courmont

More information about the vlc-devel mailing list