Debian debates amd64 port [LWN.net]

Debian debates amd64 port

Posted Jul 22, 2004 16:20 UTC (Thu) by markc (guest, #4419)
Parent article: Debian debates amd64 port

It is indeed a shame. I've been using Debian professionally for over 10
years but have grown increasingly uneasy about the lack of forward
movement in general. The lack of respect for Knoppix and how it bolstered
otherwise flagging Debian usage really blew me away. To hold back and deny
a slick technology to 95% of the Debian userbase because Knoppix didn't
cater to less than 5% of it's userbase is just so dumb. The Debian
anti-KDE agenda jacks off more than half it's userbase, another really
dumb stance... and now there is some lame-brained argument (that I neither
know of or want to) against encouraging the development of the next
natural "big thing" in computers where linux, for once, has a seriously
cool edge over M$... a full fledged 64bit OS system NOW, not in another
year when M$ also has some 64bit offering, let alone when other linux
distros already offer 64bit ports. The discussion about when PHP5 will
become available in sid hasn't even started. Talk about dumb and dumber.

I've been using a full 64bit KDE desktop OS for nearly 6 months and will
from now on use this distro for all new servers I maintain, now that I am
comfortable with it. Thank you Gentoo for saving me from the retarded
Debian borg.

(Log in to post comments)

Debian debates amd64 port

Posted Jul 23, 2004 11:57 UTC (Fri) by jondkent (subscriber, #19595) [Link]

I think Debian need to look at how it organises itself and how it releases new versions. As they have inxs if 8k packages and support 11 archs maybe they need to take a look at how Gentoo point releases work. Snapshotting maybe the way forward, i.e. you take all new/updates packages considered stable, merge them with the remainder and release it. This would then become a point release, instead of a r# release. Strict release dates would have to be enforced, say quarterly.

It may be sensible for Debian for review the archs supported to ensure the userbase justifies the workload.

Just my 2 pennies worth

Debian debates amd64 port

Posted Jul 26, 2004 9:52 UTC (Mon) by piman (subscriber, #8957) [Link]

When Gentoo's architecture and package count reach anywhere near Debian's, we'll see how well they scale.

Incidentally, Debian's testing distribution (which is where the releases come from) is snapshotted already. This is how releases happen. However, Debian also tries to actually set milestones between releases (ABI transitions, LSB compliance, new installers, etc). I'm not even sure how Gentoo would manage an ABI transition, aside from just breaking all the packages...

Debian debates amd64 port

Posted Jul 29, 2004 12:29 UTC (Thu) by markc (guest, #4419) [Link]

Your points are very academic and technical according to the Debian
mantra. Fine. However I would suggest that 90% of potential users could
not care less about 90% of the points you raise nor for 90% of the
packages that Debian maintains. The 10% of those 8k packages that appear
on the Knoppix CD are the ones that really count over all the rest by at
least a margin of 10:1. The significant and major advantages of using a
Debian based system are apt-get and packages.debian.org, from a users
perspective, elsewise, most everything else about Debian involves dealing
with a brick wall of formality and condescending developers.

BTW, IMHO, Gentoo has some serious flaws too. They don't have an apt-get
for quick and easily installable binary packages and the python +
file-system based "portage" managing code sucks badly and will not scale,
as you point out... but for the prime substance of an end-users
cutting-edge desktop, it is far more uptodate and managable than Debians
"unstable" branch ever could be. Gentoo needs to be rewritten from scratch
using C/C++ for a single apt-get/emerge interface and using SQLite for all
meta-info (ie; fast and extensible). Both Debians and Gentoos update
systems are painfully slow and neither "emerge sync" nor "apt-get update"
will scale to double their current respective sizes. Both systems need to
look at providing diffs for package updates. Debian solved a certain set
of problems managing a linux system, Gentoo solves a different set of
issues for an uptodate desktop but the ideal or "best" distro system does
not yet exist.

Debian debates amd64 port

Posted Jul 23, 2004 15:30 UTC (Fri) by mmealman (guest, #9223) [Link]

The problem is that Debian actually has a very closed developer process. I once used Debian and ran an app called mirrordir that wasn't updated as it needed to be. I contacted the developer for it and had permission to take the package over(he didn't have the time to keep it updated), but it took 8 months for me to get approved as a developer so I could actually update it. By that time the origional developer had already gotten back to updating it.

Compare that to Gentoo where anybody can submit ebuilds or even build systems and they'll see them incorporated into the core very quickly. This allows it to scale. Under Debian anyone who can actually do anything is always busy, so hardly anything new happens. With more open dists, anyone can easily contribute, so you're not brain draining your core team for things like amd64, KDE or PHP5 support.

Debian debates amd64 port

Posted Jul 23, 2004 19:02 UTC (Fri) by oak (subscriber, #2786) [Link]

Why this is reminding of the recent xfree86 split?

Debian debates amd64 port

Posted Jul 24, 2004 7:39 UTC (Sat) by tzafrir (subscriber, #11501) [Link]

Indeed you didn't have time to keep it up-to-date. Debian wants to have a "father" for each package, not just for one-time update. One of the things that are tested on those 8 monthes is whether or not you're will to /maintain/ a package. The original maintainer eventually did.

Debian debates amd64 port

Posted Jul 26, 2004 9:50 UTC (Mon) by piman (subscriber, #8957) [Link]

Rather than becoming a developer, you could have simply submitted a patch to the BTS, prepared an NMU, and asked another developer to sponsor it for you. That would have taken much less time (probably a month at most waiting for a sponsor; I don't know what kind of patch-writing skills you have).

Given the level of trust and skill required to maintain Debian packages, 8 months seems more than reasonable. Maybe it's not enough, because my NM process took 8 months, and I've had my share of mistakes.

Debian debates amd64 port

Posted Jul 27, 2004 20:18 UTC (Tue) by bronson (subscriber, #4806) [Link]

Yeah, right. That would almost certainly not work. For some weird reason, NMUs are frowned upon. It's as if you're insulting the original packager, even if all you're doing is fixing year old bugs. The end result is a whole lot of discussion about who owns what and very little forward progress. Debian is in serious pain right now.

I'd love to see Debian encourage NMUs, especially if a bug is older than, say, 1 month. This way, packages wouldn't fester quite so bad.

Debian debates amd64 port

Posted Jul 29, 2004 21:04 UTC (Thu) by andrel (guest, #5166) [Link]

The release team have been encouraging NMUs.