Re: Debian-edu/Skolelinux and Edubuntu cooperation

[Tollef Fog Heen <tfheen@err.no>]

* Markus Gamenius 

| > Note that there has been a bit of discussion in the Ubuntu
| > community on doing such [longer-lived] releases, mainly to satisfy
| > enterprise customers.  This would however also benefit us a lot.
| > Combined with the discussion around making Debian-edu a bit less
| > bound to Debian itself, I think we could make a really, really
| > nice platform for schools, whether it be based on Ubuntu or Debian
| > itself.
| 
| When we did spoke at NKUL in Trondheim a few days ago, You did have
| some thoughts on why Debian in the future would be less slow in giving
| out new relases. Could You please share them with the rest of our
| developers. I belive it is wery interessting what you think on this
| with your connection to Debian and Ubuntu

I think we need to go back a fair amount of time, back to 2000, when
potato was released.  Potato was the last version to be released
through a process where unstable was turned into frozen and then RC
bugs were weeded out through stabilisation cycles.  For potato, this
freeze took more than half a year.  It was painful for users and it
was painful for developers, mainly because hand-reviewing all changes
to frozen (which the release manager did) was a huge undertaking.
(Ironically, this process is a lot like what Ubuntu is doing, but they
have more dedicated manpower than Debian did, and it's a subset of
Debian.)

Woody was delayed severly because of (at least) three different
issues:

- the turn to use of testing as a release process rather than freezing
  then stabilising.
- the installer, with first trying to go for debian-installer, then
  switching back to boot-floppies
- infrastructural issues, mainly the security autobuilders.

I don't think releasing woody with the old freeze-based process would
be possible:  Potato was three CDs.  Woody was seven CDs.  Potato was
six architectures (arm, i386, powerpc, sparc, alpha, m68k), Woody was
eleven architectures (potato's + hppa, ia64, mips, mipsel, s390).

For sarge, the one huge thing which has held the release back for a
long time is the installer.  Debian-installer has been under
development for quite some time and has become a really, really nice
framework for installation systems.  I think some of its flexibility
shows in the fact that it's used for not only installing Debian (and
Ubuntu), but it's also used for the Ubuntu live cd by just adding a
couple of packages to the system.  That's something which
boot-floppies could never have done.  There has been a bit of
infrastructure and hardware-based delays for sarge too, but those are
mostly solved now.  Testing per se hasn't been a holdup for etch.  The
release team has done major transitions in testing and learnt how to
do them well.  Examples are the libtiff3g => libtiff4 transition
which, if my memory serves me correctly, took about three weeks.
GNOME 2.6 to GNOME 2.8 was smooth.

What will be the big holdups for etch?

- The installer is in good shape.  It will want some changes, but
  those are, for the most part, not necessary.  Also, we want some
  time to pass, so having some time to hack on the installer is nice
  anyhow.

- Testing as a release process and methology works well.

- I can't really see any big infrastructural things coming up, and
  together with the Vancouver proposal it should be possible to just
  ignore architectures which can't keep up (for one reason or
  another).

We have a bunch of big transitions for etch however:

- GCC4.  This will take some time, but Debian is getting a lot of help
  from the Ubuntu community as well as from Andreas Jochens who has
  been recompiling unstable with gcc 3.4 and later 4.0 for quite some
  time.

- Multiarch.  This won't have to be complete in any way for etch, but
  the base support in the toolchain as well as dpkg would be nice to
  have.

- Xorg, hopefully including X11R7 which means we can get rid of
  /usr/X11R6.

- Newer GNOME, KDE and so on.  Those should be fairly painless, I
  think.

This, together with everybody wanting to get etch out the door fairly
quickly means we stand a good chance of releasing etch not too long
after sarge.  I think a schedule of about 12-18 months is good.  I
think we can reach that.

-- 
Tollef Fog Heen                                                        ,''`.
UNIX is user friendly, it's just picky about who its friends are      : :' :
                                                                      `. `' 
                                                                        `-