Slashdot Mirror
Debian Freeze Process Update
snotty6969 writes: "Freeze Update. Anthony Towns sent in an updated report about the Woody freeze process. We're almost into the last week for uploads of base packages. If there are outstanding bugs you'd like to see fixed, provide patches or upload now. We are also getting into the last days for ensuring that standard and task packages get included in the Woody release. At the moment it looks like a lot of packages will be removed from Woody. Among these are a whole bunch of commonly used programs like gpm, Mutt, CVS, Procmail, Apache and Mozilla. People who can fix bugs in these packages and care about them are encouraged to send in patches or upload fixed packages using Anthony's unofficial NMU guidelines."
I've just switched distros to debian on 3 boxes (home from mandrake, web/mail/cvs/db box at work, and a development machine). I've been really pleased that although it's a bit of a PITA to get set up right, once it's done, it's really done. Yes, apt-get is lovely.
But if things like apache and mozilla (and for me procmail and cvs) are starting to fall, how is the future looking for debian? The thing I love about it is the the fact that almost everything I use I can just apt-get, and it all fits together. If I had to start getting my own packages a lot, it would really dampen debian's best feature.
I really hope this is merely a bit of sabre-rattling done in order to stir up some activity before release.
0.02
Tales from behind the Lagom Curtain
This submission was lifted verbatim from the most recent Debian Weekly News. I just felt someone should point this out, since the submitter didn't seem to consider it worth mentioning.
Don't worry, people. The packages you care about will be in Debian 3.0. (Including mpg321!) We'll make sure of it. :)
It can be a good thing, for sure, but at some point there's a tradeoff that must be made between stability and usability. For most of the basic internet services (web, mail, DNS), development has reached a certain point of maturity, and you really don't lose much by running a 2 year old release of sendmail or BIND (provided you have all necessary security updates, which Debian makes easy).
The problem is, though, once you start leaving that realm of the world, upstream development happens at a really fast pace, and Debian's release cycle does not keep up. I often cite GNOME in slink (Debian 2.1) as an example of this problem. Slink shipped with GNOME 0.3, which you may recall was virtually unusable and was certainly not stable. The most frustrating part of that, though, was that by the time slink actually shipped, GNOME 1.0 had been out for months! How can slink be considered stable when the software that comprises it is an old development snapshot?
A more current example may be apache 2. It is still not available for Debian (even in unstable), which leads one to suspect that it won't ship with woody. If it doesn't, then what happens to those users who need apache 2 functionality on mission critical servers? If they need to run unstable to get the software they need, then that defeats the point of stable. If they need to fetch the source and compile apache 2 outside of the Debian package system, then that defeats the point of apt-get.
IMHO, it is unacceptable for Debian to not currently have a stable release that includes PERL 5.6, XFree86 4.x, Linux 2.4.x, etc. What prevented the release manager from proclaiming a freeze 6 or 8 months ago? Newer versions of key packages were available and reasonably well tested at that point, and a 1 or two month freeze would have left us with a released version of Debian that was both stable and reasonably up to date.
One of the key problems, I believe, is that Debian does not use any notion of release goals. This makes it impossible to say for certain when a freeze should happen. It's entirely up to the release manager. Obviously it's not easy to have release goals for a distribution, since much of the software you want to package is not available when drafting the list of goals, but even some sort of vague, general release goals would help to provide focus.
Or maybe the problem is just that nobody actually wants to do QA debugging so they keep putting it off until the release manager gets fed up and stops allowing new features to be added until some bugs are fixed.
I don't know...I've been a Debian user for 5 years and a developer for about a year. I am very frustrated with the pace of the release cycle. Another OS I use regularly (FreeBSD, which I use at work) shares Debian's reputation for quality and stability, but they release at least two versions of their OS each year. They've released three versions since Debian released potato. Why can't we do that?
noah
Why the hell would they freeze just before emacs21 goes in, just before KDE 2.2.2 goes in, just before ALSA goes good, etc etc?
Because if we applied this criterion, we'd never freeze!
Someone's pet package is always going to be about to be released, and will be left in the cold; IMO, this fear of leaving old software in stable is a large part of what historically contributed to long release cycles. (I think the current one is long mainly because we've completely redone the archive/release infrastructure and we're still working out bugs in the new system. That and, sigh, the installer)
Daniel
Hurry up and jump on the individualist bandwagon!
It's good to see that Debian is maintaining their quality even when rushed. Making threats like this is one way to accomplish that - saying to maintainers with broken patches, "if you don't submit a patch, the release will suck and it will be ALL YOUR FAULT".
And I'm frankly amazed they got Mozilla in in the first place - they hadn't since M18, and with no packaged version Mozilla it was practically impossible to install Galeon.
Win dain a lotica, en vai tu ri silota