[gentoo-releng] x86 GRP

[gentoo-releng] x86 GRP

[John Davis]

[-- Attachment #1: Type: text/plain, Size: 911 bytes --]

Hi all -
I am back from vacation for now :) I will be gone again come Friday, but
will be back in full force Tuesday. Of course, if there are any urgent
needs, have jforman, beejay, or devrel buzz me on my cellphone.

Ok, on to the real issue. Do we really need to build GRP for all
subarches of x86? I am pondering nixing all of the subarches except x86
and i686. Doing so would save infra space on the mirrors and beejay some
build time ;) No other arch offers all of their subarches, so why does
x86? If there are no compelling reasons to not drop the other subarches,
I will bring this up before #-dev and try to get some user input. Thanks
in advance for your commentary ;)

Cheers,
-- 
John Davis
Gentoo Linux Developer
<http://dev.gentoo.org/~zhen>

----
GnuPG Public Key: <http://dev.gentoo.org/~zhen/zhen_pub.asc>
Fingerprint: 2364 71BD 4BC2 705D F338  FF70 6650 1235 1946 2D47


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-releng] x86 GRP

[Seemant Kulleen]

[-- Attachment #1: Type: text/plain, Size: 1149 bytes --]


> Ok, on to the real issue. Do we really need to build GRP for all
> subarches of x86? I am pondering nixing all of the subarches except x86
> and i686. Doing so would save infra space on the mirrors and beejay some
> build time ;) No other arch offers all of their subarches, so why does
> x86? If there are no compelling reasons to not drop the other subarches,
> I will bring this up before #-dev and try to get some user input. Thanks
> in advance for your commentary ;)

If I may pipe in here -- I also see no need for *all* the sub-arches we
offer.  So I'm glad you're thinking of this route -- I think it might be
still ok to offer x86, 486, 586 and 686.  Honestly, 686 is the sub-arch
that least needs GRP -- provided we keep in mind that GRP's sole raison
d'etre is fast install.  The slower sub-arches are what need the boost
more than p4, ath-xp, ath-mp, p3, and god knows what else we have on the
mirrors.

Thanks,

-- 
Seemant Kulleen
http://dev.gentoo.org/~seemant

Public Key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x3458780E
Key fingerprint = 23A9 7CB5 9BBB 4F8D 549B 6593 EDA2 65D8 3458 780E


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-releng] x86 GRP

[John Davis]

[-- Attachment #1: Type: text/plain, Size: 1439 bytes --]

On Wed, 2004-05-26 at 22:14, Seemant Kulleen wrote:
> If I may pipe in here -- I also see no need for *all* the sub-arches we
> offer.  So I'm glad you're thinking of this route -- I think it might be
> still ok to offer x86, 486, 586 and 686.  Honestly, 686 is the sub-arch
> that least needs GRP -- provided we keep in mind that GRP's sole raison
> d'etre is fast install.  The slower sub-arches are what need the boost
> more than p4, ath-xp, ath-mp, p3, and god knows what else we have on the
> mirrors.
> 
> Thanks,

Seemant -
You are always welcome to "pipe in" :) I agree with your idea that the
slower arches do benefit from GRP more so than the i686 class machines,
but I would like to keep GRP sets at a minimum (magic number 2 or 3) for
space concerns. We offer 5 sets for x86 as is(x86, i686, p3, p4,
athlon-xp).

Is there a better choice for subarches to keep? Can x86 cover the i486
and i586 people, or do we need to ditch i686 and add in a i586 set?
Should we do three sets and have ix86, i586, and i686? 

My argument for keeping i686 is the sheer number of machines out there
that are i686 and above. I think that x86 and i686 offer the most
flexibility for what we can offer.

Cheers,
-- 
John Davis
Gentoo Linux Developer
<http://dev.gentoo.org/~zhen>

----
GnuPG Public Key: <http://dev.gentoo.org/~zhen/zhen_pub.asc>
Fingerprint: 2364 71BD 4BC2 705D F338  FF70 6650 1235 1946 2D47


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-releng] x86 GRP

[Seemant Kulleen]

[-- Attachment #1: Type: text/plain, Size: 747 bytes --]


> Is there a better choice for subarches to keep? Can x86 cover the i486
> and i586 people, or do we need to ditch i686 and add in a i586 set?
> Should we do three sets and have ix86, i586, and i686? 
> 
> My argument for keeping i686 is the sheer number of machines out there
> that are i686 and above. I think that x86 and i686 offer the most
> flexibility for what we can offer.

Come to think of it, actually, I agree with you -- x86 and 686 probably
cover the gambit.  If we felt like going out on a limb, a 586 might not
hurt.
-- 
Seemant Kulleen
http://dev.gentoo.org/~seemant

Public Key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x3458780E
Key fingerprint = 23A9 7CB5 9BBB 4F8D 549B 6593 EDA2 65D8 3458 780E


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-releng] x86 GRP

[Donnie Berkholz]

[-- Attachment #1: Type: text/plain, Size: 658 bytes --]

On Wed, 2004-05-26 at 22:56, Kumba wrote:
> Seemant Kulleen wrote:
> 
> > Come to think of it, actually, I agree with you -- x86 and 686 probably
> > cover the gambit.  If we felt like going out on a limb, a 586 might not
> > hurt.
> 
> Only reason I can think of for i586 GRP is for the faster 
> Cyrix/K6-{1,2,3} machines, which are decently powered, but weigh in as 
> i586 class (Cyrix because it lacks that CMOV instruction and such).

That actually doesn't technically make it 586 if you read the spec, but
still issues come up with programs written to expect all 686s to have
the missing instruction.
-- 
Donnie Berkholz
Gentoo Linux

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-releng] x86 GRP

[Kumba]

Seemant Kulleen wrote:

> Come to think of it, actually, I agree with you -- x86 and 686 probably
> cover the gambit.  If we felt like going out on a limb, a 586 might not
> hurt.

Only reason I can think of for i586 GRP is for the faster 
Cyrix/K6-{1,2,3} machines, which are decently powered, but weigh in as 
i586 class (Cyrix because it lacks that CMOV instruction and such).


--Kumba

-- 
"Such is oft the course of deeds that move the wheels of the world: 
small hands do them because they must, while the eyes of the great are 
elsewhere."  --Elrond

--
gentoo-releng@gentoo.org mailing list

Re: [gentoo-releng] x86 GRP

[Benjamin Judas]

This would make sense to me, however we would have to prepare for a lot of
complaints about this decision from our users. What about an other
decision? If a user wants to put up a box fast, what does he really need?
I think the packages mostly needed are KDE and OpenOffice. So why not
simply provide a much smaller GRP-set for the fast archs and keep the
complete set for the slower ones?

Regards,
   Benni

>
>> Ok, on to the real issue. Do we really need to build GRP for all
>> subarches of x86? I am pondering nixing all of the subarches except x86
>> and i686. Doing so would save infra space on the mirrors and beejay some
>> build time ;) No other arch offers all of their subarches, so why does
>> x86? If there are no compelling reasons to not drop the other subarches,
>> I will bring this up before #-dev and try to get some user input. Thanks
>> in advance for your commentary ;)
>
> If I may pipe in here -- I also see no need for *all* the sub-arches we
> offer.  So I'm glad you're thinking of this route -- I think it might be
> still ok to offer x86, 486, 586 and 686.  Honestly, 686 is the sub-arch
> that least needs GRP -- provided we keep in mind that GRP's sole raison
> d'etre is fast install.  The slower sub-arches are what need the boost
> more than p4, ath-xp, ath-mp, p3, and god knows what else we have on the
> mirrors.
>
> Thanks,
>
> --
> Seemant Kulleen
> http://dev.gentoo.org/~seemant
>
> Public Key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x3458780E
> Key fingerprint = 23A9 7CB5 9BBB 4F8D 549B 6593 EDA2 65D8 3458 780E
>
>


Benjamin Judas                                  http://dev.gentoo.org/~beejay
Gentoo-developer                                        http://www.gentoo.org
Germany

GPG-Public-Key : http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xC31DEDD8
Key-Fingerprint : 4E65 AAFE 785B 61D8 E4D9  1671 E017 87B7 C31D EDD8


--
gentoo-releng@gentoo.org mailing list

Re: [gentoo-releng] x86 GRP

[Sven Vermeulen]

[-- Attachment #1: Type: text/plain, Size: 895 bytes --]

On Wed, May 26, 2004 at 10:56:08PM -0400, Kumba wrote:
> Only reason I can think of for i586 GRP is for the faster 
> Cyrix/K6-{1,2,3} machines, which are decently powered, but weigh in as 
> i586 class (Cyrix because it lacks that CMOV instruction and such).

One of the reasons I think more in favor of i686 is to be 'better' than
Mandrake, which optimizes for i586 iirc. And if I'm mistaken and they
compile for i686, then the reason is to 'not be lower' ;)

Okay, that's a marketing PoV, but hey - we need that as well, don't we?

Wkr,
      Sven Vermeulen

-- 
 Bent Hindrup Andersen, Danish MEP, about the Software Patent Directive:
    The approach of the Commission and Council in this directive is shocking. 
    They are making full use of all the possibilities of evading democracy that 
    the current Community Law provides.     <http://lwn.net/Articles/84009/>

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-releng] x86 GRP

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 1227 bytes --]

On Thu, 2004-05-27 at 03:05, Sven Vermeulen wrote:
> One of the reasons I think more in favor of i686 is to be 'better' than
> Mandrake, which optimizes for i586 iirc. And if I'm mistaken and they
> compile for i686, then the reason is to 'not be lower' ;)
> 
> Okay, that's a marketing PoV, but hey - we need that as well, don't we?

I was always under the impression that the GRP was designed for one
purpose, and that is to get a Gentoo machine up quickly.  I never did
understand the reason for the optimized GRP sets.  In fact, I only ever
downloaded the "i686" set simply because it would work on both AMD and
Intel hardware, from a Pentium Pro on up.

I definitely agree that we need 2 GRP sets for x86, an i686 set for PPro
and faster, and an x86 set for those below.  There is little to no
advantage to the lower sets, and it does take up a massive amount of
mirror space.

This is going to be a hot issue with the users.  Perhaps we could get
something posted on the next GWN with a link to a survey of some sort? 
Perhaps have a week later for an "end date" and see what the users
decide.

-- 
Chris Gianelloni
Developer
Games/LiveCD Teams
Gentoo Linux

Is your power animal a penguin?

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-releng] x86 GRP

[Jason Huebel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

John Davis wrote:
| Hi all -
| I am back from vacation for now :) I will be gone again come Friday, but
| will be back in full force Tuesday. Of course, if there are any urgent
| needs, have jforman, beejay, or devrel buzz me on my cellphone.
|
| Ok, on to the real issue. Do we really need to build GRP for all
| subarches of x86? I am pondering nixing all of the subarches except x86
| and i686. Doing so would save infra space on the mirrors and beejay some
| build time ;) No other arch offers all of their subarches, so why does
| x86? If there are no compelling reasons to not drop the other subarches,
| I will bring this up before #-dev and try to get some user input. Thanks
| in advance for your commentary ;)
|
| Cheers,

K, time to chime in... Since I'm amd64 (and there's no subarch), I don't
really have much to do with this decision. But here are my thoughts for
x86.  I think i386, i686 and *athlon* are the three subarches that
should be supported for x86.  athlon would cover the original athlon,
athlon-xp and athlon-mp.

But I like the idea of supporting the faster arches with a smaller GRP
set.  Perhaps support them with just the following:

OpenOffice
KDE
Gnome
Firefox
Thunderbird
Evolution

Those apps probably cover 90% of user's needs.  You might throw KOffice
or Mozilla in there as well, but the idea isn't to give them a
/complete/ system, just a /usable/ system until they finish compiling
the remainder of their apps.  And Mozilla is a relatively quick compile
on high-end machines any (30 minutes on a low-end amd64).

- ---
Jason Huebel
Gentoo/amd64 Strategic Lead
Gentoo/amd64 Release Coordinator
Gentoo Developer Relations/Recruiter

GPG Public Key:
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x9BA9E230
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFAtfjhbNgbbJup4jARAuLOAJ9/KS5zLZqP/sE88nQXsD3ALYAM9gCfW1FU
LQd1/VwnTGd9WqgDgENQDaI=
=ztmc
-----END PGP SIGNATURE-----

--
gentoo-releng@gentoo.org mailing list

Re: [gentoo-releng] x86 GRP

[John Davis]

[-- Attachment #1: Type: text/plain, Size: 1601 bytes --]

Hi all -
Thanks for the discussion, I will try to hit on all of the major points
;)

It seems that the general consensus is to keep i686 and x86 if we do in
fact decide to drop some other subarches. Am I right in saying that?

Although I see where you all are coming from with regard to the minimal
GRP set, I would rather that we just stick to a monolithic GRP set for a
couple of reasons. One (big) reason is that GRP not only accomodates
rapid installs through reducing application install time, but it also
enables networkless machines to install a full gambit of common apps. A
user could have a dual processor P4 HT with 3 GB of ram, which would
compile mozilla in about 20 mins, but if they don't have a net
connection at install time, they are SOL. Also, do we really need to be
maintaining two sets of GRP? I would rather focus our efforts on LiveCDs
and other future projects, such as X LiveCDs, GameCDs, etc.

When I first made my decision about GameCDs, I was rather naive ;) There
really is no reason that they should provide install functionality
outside of basic hardware detection. Chris and Livewire are right - it
is a waste of space otherwise. At some point, I would like to make
GameCDs a full blown releng sub-project, provided that we have the
manpower to run with it. First priority though is to stabilize our
release media.

Cheers,
//zhen
-- 
John Davis
Gentoo Linux Developer
<http://dev.gentoo.org/~zhen>

----
GnuPG Public Key: <http://dev.gentoo.org/~zhen/zhen_pub.asc>
Fingerprint: 2364 71BD 4BC2 705D F338  FF70 6650 1235 1946 2D47


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-releng] x86 GRP

[Aron Griffis]

[-- Attachment #1: Type: text/plain, Size: 686 bytes --]

Jason Huebel wrote:	[Thu May 27 2004, 10:19:14AM EDT]
> But here are my thoughts for
> x86.  I think i386, i686 and *athlon* are the three subarches that
> should be supported for x86.  athlon would cover the original athlon,
> athlon-xp and athlon-mp.

Since this is turning into a poll on the releng list, here is my
opinion: If I'm installing from GRP, the point is to get a machine
working quickly, not have an instantly optimized installation.  I
would be totally satisfied with x86 *only* and skip the others
entirely.

From the sound of it, that's a really different view from most
developers...  ;-)

Regards,
Aron

--
Aron Griffis
Gentoo Linux Developer


[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-releng] x86 GRP

[Grant Goodyear]

[-- Attachment #1: Type: text/plain, Size: 484 bytes --]

> [I] would be totally satisfied with x86 *only* and skip the others
> entirely.
> 
> From the sound of it, that's a really different view from most
> developers...  ;-)

Well, not from that of this dev.  I agree completely that for x86 is
likely to suffice for the vast majority who want GRP.

Best,
g2boojum
-- 
Grant Goodyear	
Gentoo Developer
g2boojum@gentoo.org
http://www.gentoo.org/~g2boojum
GPG Fingerprint: D706 9802 1663 DEF5 81B0  9573 A6DC 7152 E0F6 5B76

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]