Comment 50 for bug 249373

> > It appears that Gnome cannot be trusted to release a new version without
> > major regressions, and Ubuntu needs to rigorously test new versions before
> > committing to them.
>
> But they don't. Ubuntu is committed to releasing on a rigid schedule regardless of functionality or regressions, too.

No need to put the blame on time-based releases. They are definitely preferable to Debian-style releases, which seem to never get ready. The problem is that, in the absence of guaranteed developer commitment (and you never have guaranteed developers in Open Source) and good project planning, it is impossible to make sure that the desired functionality will be ready for the next release deadline. This is an inherent risk of time-based releases: if you're unlucky, you may have very few new features for your next release.

This case was particularly difficult for Ubuntu, I guess, because of the complex dependencies inside of GNOME. Running 2.24 with the 2.22 session manager is no trivial matter (as Gentoo's and Debian's efforts show) so this wouldn't have been possible in the Intrepid time frame. Therefore, the decision was between including 2.24 with the regression, and staying with at least a large portion of 2.22. As we all know, 2.24 won.

As I already said elsewhere in this thread, I regret this decision. I think that, by doing this, Ubuntu sends to wrong message upstream. I'm really surprised that such a large regression was possible in GNOME without any discussion whatsoever. In GNOME, a developer needs approval from three release team members in order to change a single string after string freeze, but he can cut off half of the functionality of a core component and nobody says a word. Strange, to say the least.

But this is getting off-topic. This should be discussed in some higher Ubuntu forum, but I don't know which one is appropriate. If any of you guys have a better idea of how this can be brought to a higher instance in an appropriate way, I'll be glad to know.