Re: Are we losing users to Gentoo?
On Wed, 20 Nov 2002 14:38:29 -0500
Mark Mealman <mmealman@thebisgroup.com> wrote:
> >
> >
> >Mark Mealman wrote:
> >
> >>> But Debian's bleeding edge really tends to lag. What's KDE up to on
> >>> testing, version 2.2? Mozilla is 1.0? Java's at 1.1?
> >>
> >>
> >
> >I'm missing something here... what does the term "bleeding edge" have to
> >do with testing? Unstable + experimental + packages downloaded from
> >developers' web pages (many of which are apt-gettable) is the "bleeding
> >edge" of Debian. Testing isn't even close.
> >
> >
> If unstable/testing isn't even close to the packages being used in the
> wild, then that pretty much validates my point that the Debian archives
> can be pretty stale.
>
> In fact if you polled around for the biggest complaint about Debian I
> think you'd see "out of date" as the winner.
>
> Although in my experience the above isn't true with smaller apps. Debian
> is often first on the scene with security updates as well. It's the
> larger more complicated packages that lag.
>
> >
> >
> >>> It takes a lot of time
> >>> for developers to gather the sources, compile binaries across all of
> >>> Debian's supported platforms and make sure they play nice with other
> >>> packages.
> >>
> >>
> >
> >True, but are you suggesting that Gentoo is "better" because they don't
> >bother making sure that things play nice with each other? I think that's
> >one of Debian's strongest points.
> >
> >
> It's not about "better" it's about the differences. Debian has
> stability, but the very nature of how it achieves stability means it
> really lags behind other distributions when it comes packages like KDE.
...which either means the developer doesn't have time that week to
get the latest thing packaged, or that in the developer's judgement,
whatever-version-which-happens-to-be-the-latest is unsuitable given
that attempts are always being made to play nice with other packages
even in stable either in its substance or its packaging.
Specific example is xfree86-4.x: Branden would not start the work on
that before woody had a stable version of the xfree that was stable at
the time; he wouldn't have had the time. Once woody was released (which
implied a stable X was done for woody) he turned his attention to later
versions of X, which involved a process of getting to know the all-new
stuff in the source code. Note the difference: In Gentoo, just compile
it; it should work, and might even work with other stuff. In debian,
actual care and time is taken to understand the underlying issues.
Gentoo would lagtoo (sorry :) if that level of care were taken.
Larger packages take more time. That part is fact. If a good job is to
be done, the time must be spent. You don't get the kind of stability
debian is reknowned for without spending the time. If Gentoo doesn't
spend this kind of time, they're probably making a mistake, or they
look at Gentoo as a slackware-with-ports: not much care, but entire
system can be rebuilt from scratch with one command. But one thing
is for sure: some people like Gentoo. Perhaps the bottom-line statement
that can be made on any comparison between the two, is that they are
different, and meant for different purposes.
I think Andrew Suffield does nothing for debian, and harms debian's
image, by using epithets like "moron". He knows my position on this,
and has for years: the thing to do is try to -help- people who will
also shoulder the burden of learning for themselves. It is -only-
when they -refuse- to learn for themselves or at least help when
they are being helped that a process of shunning should occur,
because then you have a person who (1) doesn't know that he doesn't
know, and also (2) won't correct the situation, or even help do so.
But some of his technical points make sense. Many people have been
asking for optimized pentium "ports" (in the debian sense), when the
reality is that any such port (like, compile all packages for 686 with
-fomit-frame-pointer, or other such combination) must be -maintained-.
If you look at the Gentoo thread mentioned elsewhere in this
thread, some messages posted there lend support for this, comments
like "if you try to apply a set of CFLAGS values to all packages
blindly, ..." (one ending is) "... many apps might crash" (another
is) "... you might not get any performance increase" (yet another)
"... any performance increase won't be seen overall in apps that
are peripheral-intensive, i.e., disk-intensive".
Other comments in that thread include comments like "Hey, here's
my CFLAGS, ..." "... why won't half my apps work now (including
even gcc now)?" "... it might help you, george" (and george says
"no, I have a m68k and your CFLAGS has pentium options") "...
now you can all try it out", etc, etc. So there are people who
don't understand optimization issues yet. They are not deserving
of epithets unless they refuse to help themselves grow.
> Gentoo on the other hand uses a build system that allows for rapid
> deployment(KDE 3.1 final is in Gentoo and I don't think 3.1 has even
> been officially announced yet), but it won't ever achieve Debian's
> stability.
(small point on kde 3.1 final existing before announcement disposed of:
it won't be "final" until it's "announced". by definition. also, there
may be current reasons why the announcement has not been made.)
Yes, that's exactly right. But debian's build system has never failed to
work for me, and it has in every other dist I ever ran (which was mostly
SLS, early slackware and redhat). So, I contend here with you because
there's nothing wrong with debian's build system. In fact, the build
depends introduced in woody probably make debian a better choice. Here,
you have a -build- system that enjoys debian's stability. That's not
easy to beat...
> -Mark
-Jim
Reply to: