Re: testing currently unusable ?
On Mon, Jul 28, 2003 at 12:11:02AM +0100, Richard Lamont wrote:
> On Sunday 27 July 2003 12:58, Björn Stenberg wrote:
>
> > [snip] If debian is only useful for people who
> > can either 1) run unstable, or 2) backport all applications they
> > need, then we fail our social contract.
I like the newest and greatest too, and I don't run stable, but it
really seems much too strong to call stable not useful. It's there
and it works.
> >
> > Just because you and I may be comfortable running unstable or mixed
> > enviroments does not mean this is not a problem for a large number of
> > users.
> >
> > Then what what should they do? Debian currently has nothing to offer
> > non-developers who want to run linux on their desktop. Stable is so
> > out of date as to be virtually useless for desktop use.
>
> Agreed. As an end-user myself, I'm not interested in running anything
> except stable. I'm not a developer and never will be - for me, software is
> a means to an end, not an end in itself.
>
> The ability to just do apt-get update/upgrade to keep the machine
> secure is Debian's biggest strength.
>
> The moment you start mixing in testing or unstable, this breaks,
> because before you know it the damn thing is installing a bleeding-edge libc6
> when you least expect it and you're very soon into rpm-style dependency hell.
> And there's no security support.
It's true that if you mix stable/testing/unstable you tend to get
sucked toward the least "stable" configuration. But I wouldn't call
it dependency hell; the dependencies work to keep things in a
reasonable state.
For best security support, stick with stable. But in practice, the
other versions pick up security fixes too. As far as I know,
unstable gets newer versions with security fixes about the same time
as stable gets the old, patched version (even though official policy
is that security is only for stable). And these in turn migrate to
testing after some time.
>
> The only way for a non-developer to run up-to-date desktop applications
> on stable is to use unofficial packages, such as the Woody KDE-3.1.2 packages
> available from download.kde.org. The same applies to other important desktop
> apps, such as OpenOffice and Mozilla.
>
> For me, the ability to find unofficial, up-to-date .debs at www.apt-get.org
> has been vital. Without these packages, Debian would be useless for me.
>
> Debian is producing a perfectly good stable desktop OS, IMHO. It just isn't
> producing the apps to run on it. I have to get these somewhere else.
>
> The underlying problem is the inflexibility of the release system. Debian
> needs a more fine-grained concept of what 'stable' means. The present policy
> makes sense for the OS, but it seems too restrictive for apps that sit on
> top of the OS and don't have stuff dependent on them.
>
Clearly it's a drag that testing has been in a weird state for quite
awhile. For a couple of months I have held off doing a dist-upgrade
because it would remove a lot of packages I want. But I can continue
to use the packages I have (obviously, for a new install, which
started this thread, things are less pleasant).
While it would be nice if new versions came out faster, or in a more
granular way, I doubt that's technically realistic. The whole
point of having a release is to get a system that is self-consistent.
As recent experience shows, that's non-trivial. In the past, changes
in glibc have been critical. Now, the conversion to gcc 3.3 (more
accurately g++ >= 3.2) means that all C++ stuff needs to be
recompiled. And when a major desktop like KDE or GNOME changes
versions, all the stuff on top has to change. I don't think there's
any getting around the fact that it's a complex and ugly process.
Note that all the examples above involve changes in low-level
components, so I don't think splitting into "slowly change low-level"
and "rapidly changing higher level" would work. Backports can address
the wishes of people on stable to get newer software, but of course
somebody has to do them.
Reply to: