Re: Getting serious about releasing
Havoc Pennington <[email protected]> 23 Apr 2002 21:57:18 -0400
| Newsgroups | gmane.comp.gnome.lib.devel |
|---|---|
| Message-ID | <[email protected]> |
Sander Vesik <[email protected]> writes: > > The issue is very clearly one of "somebody needs to do the work, and we > can't wait forever for people to show up and do it". So yes, unless people > show up and just fix it, it will get punted. > Precisely. Once we are in a shippable state bug-wise, everyone I've talked to agrees we should have time-based releases (vs. feature-based). That is, we release every N weeks, whatever we have ready at that time, then get on with the next release. The reason GNOME 2 has to be feature-based instead of time-based is that we broke everything. Bad idea. Once GNOME 2 is working, it should stay working so that we can ship pretty much anytime on a time-based schedule. Which means often. So that people get the enhancements often. So the goal for GNOME 2.0.0 is the first thing that's usable and has basic bug sanity. Then we start making 2.0.x releases every N weeks to add more stability and maybe re-enable some of the emergency-disabled features. Then at some point we consider adding larger features and moving on to a 2.2 release; but even then, we keep it all pretty much working at all times. I don't think breaking the desktop (or its build) should be OK in the GNOME project, ever again; if the tree breaks, if people can't use CVS HEAD, the change comes out until it works. There shouldn't be another fiasco like the one we had last summer. For GNOME 1.x, people stopped building out of CVS and let bugzilla sit too long; and for GNOME 2, it spent too long in "only two people hacking on it" mode because the whole CVS build was broken. I'm hoping we can avoid that since we have much better process now, have Luis, have jhbuild, all this fun stuff. I want to keep the release team intact and just keep rolling with releases, just as we keep doing betas right now. Think of it this way: we have 300 volunteers available to hack or document or translate, and many more testers interested in trying stuff. If one person breaks the tree, all 300 of those people are stuck doing nothing. So breaking the tree is just bad bad bad. Havoc