[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: amd64 into mainstream

On Monday 18 April 2005 4:35am, Thomas Steffen wrote:
> On 4/18/05, Ed Cogburn <edcogburn@hotpop.com> wrote:
> > Alioth doesn't perform as well as us.debian.org for me for some reason,
> Neither does the core archives server, but that is what mirrors are for :-)

What do you mean by "core archives server"?  Isn't that what (us.)debian.org 
is, and what I'm referring to?  Despite whatever the ping times are with 
debian.org (when I was using i386 I was using a faster mirror too), it 
doesn't have pauses in transfers like alioth has.  Alioth appears overloaded, 
and moving pure64 over to the main debian.org, even while still in an 
unofficial status, would help alioth.  We're talking about work that is going 
to happen eventually anyway, everyone seems to agree pure64 "is ready".  As 
for the mirrors they have always had the choice to not mirror everything from 

> > And don't forget about the rest of the Debian infrastructure, like having
> > non-us support us too,
> Which includes some 8 obscure packages. I think it is time to do away
> with it, anyway. Or create debian-non-EU, debian-non-india and
> debian-non-china for fairness :-)

I'm sure the people who rely on non-us are amused.  :)

> How may of these third parties support PPC, mipsel or S390 because
> they are supported ports? No, I think support is missing because third
> parties can always tell people to go with mainstream, which means
> 32bit userland (and kernel). Redhat and SuSE have already created
> facts: they are (trying to be) fully compatible to 32bit userland.

For Linux, the kernel and userland can be compiled for whichever one you want.  
SUSE, Mandrake, RHEL, Ubuntu and most others have already done that,  The 
*only* problem is with closed-source products that I and many others can, and 
do, happily live without.  64 bit support is not what is keeping AMD64 from 
being an official Debian port.

> > There are a *lot* of Athlon64s out there already, compatible with and
> > competitively priced against existing 32bit X86 chips,
> Sure, but that does not translate into an incentive to install
> debian/amd64. On AMD, you will see a nice performance gain, but on
> Intel you don't even get that.

Does EM64T only add the 64 bitness, and not the extra 8 gp registers?

> [snip]
> (And most of these problems are not 64bit related...)

Right, all of that would be also true if the system had a 32bit CPU.

> The whole testing process seems to be mostly stalled due to
> changing problems with one of the more obscure platforms... no, I
> don't want to restart that debate, but I want to mention the bigger
> picture.

I'm fairly well aware of the bigger picture, as thats what I'm grumbling and 
whining about.  :)

> > But alas, despite the writing on the wall, we still have to
> > wait for Debian to get its act together, all the while losing more people
> > to (K)Ubuntu - I have to admit to having been seriously tempted to switch
> > myself lately - if our AMD64 support is "complete", then Ubuntu's AMD64
> > platform is "polished".
> I full agree here: Ubuntu is more attractive to the average end user.
> But I do not understand why everybody is so upset about this. After
> all, there is no "one size fits all" distribution.

No, I'm not saying Ubuntu will kill Debian or vice-versa, or anything like, 
but I am saying that the more energy and momentum for a desktop system Ubuntu 
takes from Debian, the longer it will be for Debian itself to get its act 
together on the desktop, because everyone who wants to see Debian on the 
desktop are now saying "Why not just use Ubuntu?", and are moving to it.

> Debian can still be the best distribution for servers, for weird hardware
> and for development, but for the average desktop user it never was an
> outstanding choice.

And if Ubuntu takes hold Debian may *never* become a good choice for the 
desktop, that is what I fear, and that would mean abandoning Debian.  :(

> What I would really like to see is binary compatibility between Debian
> and Ubuntu. So far, that seems to be mostly the case, but it is more a
> coincidence of freeze dates than a feature.

Depends on how well Canonical does.  If they continue to gain the momentum of 
"The Desktop Debian", then in a few years they may be even in a position to 
fork Debian all together, or stop supporting compatibility on the package 
level, if Debian itself doesn't respond to the momentum that Ubuntu is 
feeding on.

Sure if they turn evil down the road, we can always come back to Debian, but 
in the meantime Debian itself will not have improved much because most of the 
users and developers interested in the desktop will have spent their energies 
in the Ubuntu camp.  And yes, a resurgent Debian could theoretically pick up 
the last pieces of Ubuntu before they went to the dark side, but that would 
mean making major alterations to Debian to integrate Ubuntu's work which will 
have deviated significantly from base Debian by then, and it means users like 
me will have switch *again*, and I'm fundamentally a lazy person.  :)  I'd 
like to just stay with Debian knowing that they'll be following Ubuntu's lead 
on the desktop in short order.

The Real Problem(TM) is that Debian as an ever larger group of individuals 
with differing goals doesn't know what it wants to be as a single entity 
anymore.  The recent fighting over the "obscure" ports of Debian and whether 
to keep them or not just screams "Identity Crisis!" to me.  And as long as 
Debian remains paralyzed by its identity crisis, they never will be able to 
react to Ubuntu changing the landscape under their feet, heck, just getting 
Sarge out now, after 3+ years, seems to be taking all the energy Debian as an 
organization has right now.  Their dilemma looks like a difficult one to 
solve though, and I'm wondering if the best way to keep the Debian we all 
know and love is for Debian to fork ITSELF, and have "Server Debian" and 
"Desktop Debian" as separate, but loosely cooperating organizations, that 
only keep the deb package format, LSB, DFSG, the Debian Developer process, 
etc., in common.  In other important ways though, like different 
repositories, which platforms they support and the rules controlling that 
determination, release cycle speed, package organization for software 
specific to their focus, etc, they could diverge from one another (or one 
takes over a given area completely, and the other relies on that other's work 
- like Ubuntu relying on Debian's base now).  A "Debian is dead... long live 
Debian" kind of solution to the problem.  Sure, "Desktop Debian" would start 
to look a lot like the Ubuntu effort, but, unlike Ubuntu, and this is the 
crucial difference, it won't be run by a profit-oriented company that might 
one day go to the Dark Side, or more likely, just disappear, and never mind 
the part about the company making arbitrary profit-driven choices about what 
software users will have, e.g. KDE/GNOME, no, it will be run by an open 
non-profit organization just like the Debian we know, like, and trust now.

So is it finally time for "The Mother of All Forks"?  :)  Brother Server 
Debian, meet sister Desktop Debian, your biological twin, she's not as strong 
as you are, but she's a hell of a lot prettier.  :)

(Note the use of fork above is merely for humor, in reality a split of Debian 
would more resemble a "corporate reorganisation" than a "fork".  There would 
I believe be very little, if any, forking of computer code at all.)

Reply to: