Re: Slink to Potato
On Mon, Oct 04, 1999 at 03:38:05AM +0100, Mark Brown was heard to state:
> > > The stable GNOME packages are actually produced by the Debian
> > > maintainers - they're just distributed from the GNOME site.
>
> > So, why would they not be introduced into slink-proposed-updates?
>
> The only things in proposed-updates should be minimal fixes for very
> important (eg, security) bugs - things that will go into a new point
> release of slink. New upstream releases don't generally make those
> criteria, and the GNOME one certainly doesn't.
I've managed to stay out of this up until now, but this is something I
have been concerned about myself. This last statement is really the crux
of the matter - should stable remain untouched except for security
updates?
I think when the new debian comes out, I'll really have to be called
Debian 3.0. As far as I'm aware, the traditional naming convention for
software projects means that a whole number release signifies a release
which is incompatible with the previous release.
Because of the developments such as glibc2.1, perl, and probably
numerous other things, you can't take something out of Potato and put it
on a Slink system and expect it to work. It's an all-or-nothing
arrangement. They may both be `GNU/Linux', but they are essential
incompatible.
The problem for most of us trapped in slink-land is that, while the
linux and open source community is streaking ahead in leaps and bounds,
with advances in software coming so quickly that it's hard for anyone to
keep up, we are merely marking time.
Let me make clear that I have nothing against the Debain model. I know
that it's completely volunteer, and that a bunch of volunteers who do
the packages, organised via a series of high-traffic mailing lists, is
as close to anarchy as you can really get. I fully appreciate the
dedication that the developers have, and know that they wont release
anything until they believe they have it right. For a production system,
I'd much rather have an slightly rusty debain install than a brand new
RedHat one.
Maybe the problem with the debian model is that everything has to work
together so well before it is released, and there is so much that has
to work. The number of debian packages seems to grow by a dozen or so
every time Debian Weekly News comes out, and I suspect by the time
Potato becomes stable, it'll need at least 3 CDs for Main.
The problem is that, unlike with RPMs, very few people outside the
debian project package DEBs of their software - why would they when they
may as well be offical for the same effort. Yes I know you can use
Alien, but really, that's a hack (if quite a good one at times), not a
solution. So if you want any new software to easily add, you need to
wait for a new debian release.
Luckily for us, there are a few people who package and release stuff for
Slink, such as netgod, and the GNOME and KDE maintainers. In my
experience, all of this stuff has worked quite well, and I have no
complaints. But to keep up to date with anything other than what these
people maintain, you're on your own.
Now, I've been using linux for a few years now, so I'm not scared to try
a `./configure; make; make install', but then you loose the benefits of
dpkg/apt looking after your system. You also run the very real risk of
not having the libraries you need, or not in the right places, and if
you want to update something `significant', like the GIMP, you're
probably going to have to be installing lots of libraries yourself, and
ending up getting into all sorts of confusion.
I've looked after a few RedHat machines, and although I'll take Debian
any day, they did have one advantage. Their base system, like Debian,
was only really updated for security fixes and the kernel differences,
between major releases. But at least RH users have the possibility of
looking in the Conrtib archives (their contrib means `user contributed',
which makes more sense to me than contrib meaning `depends on something
in non-free', but maybe I'm just thick...), and finding nice new
releases of their favourite packaged. Sure, it might not work, it might
hose their system, it might even install a backdoor and have you owned
in half an hour, but it's an informed choice.
Debian stable, on the other hand, works perfect, every time, for a long
time - but it doesn't change for a long time either. Something of a
doble-edged sword, I guess. I'm not necessarily suggesting we go the RH
way, and throw quality control to the wind, but it's an interesting
difference that highlights some issues.
Where does this leave me? I like trying new things, and I'm as
up-to-date as netgod and gnome want me to be. I'm also pretty
adventurous, and I'd like to try some of the newer stuff. I compile a bit
of stuff myself (Lyx was one of the apps mentioned in the original
message in this thread, and I've got that working fine by compiling it
myself). On the other hand, I'm not a programmer - I'm a user. I can fix
some things, apply patches if I know about them, install a new kernel,
but I can't read C code, so in most cases I can't fix a problem myself.
A purist might argue that I shouldn't take on unstable because I wont be
able to fix it myself. Others argue that they have had potato running
for ages, and it's safe as houses. Do I risk it? I have no idea yet.
I'm sorry this has been pretty long, and I'm not sure that I said
everything I wanted to say, but I probably can't think of everything. As
a parting note though, I would like the developers to consider whether
the development cycle is likely to improve.
For potato there seem to have been lots of issues, the main two being
glibc2.1 and Perl. What will they be for the next release? Will it be
glibc3? Will it be Perl6? How long will it take to get all those
packages ready for the next one? Where will debian be when kernel 2.4
and XFree 4.0 come out? Will they make it into the next release? the one
after? How long until it reaches critical mass and we can no longer keep
up?
I'm not trying to be pesimistic - I'm sure we'll find a way somehow.
cheers,
damon
--
Damon Muller (dm-sig6@empire.net.au) / It's not a sense of humor.
* Criminologist / It's a sense of irony
* Webmeister / disguised as one.
* Linux Geek / - Bruce Sterling
Reply to: