[Mageia-dev] i686 must be Pentium II ?

Giuseppe Ghibò ghibomgx at gmail.com
Sun Sep 26 17:04:19 CEST 2010

2010/9/26 andré <andr55 at laposte.net>

> Thomas Backlund a écrit :
>> Giuseppe Ghibò skrev 26.9.2010 14:59:
>>> 2010/9/26 Thomas Backlund <tmb at iki.fi <mailto:tmb at iki.fi>>
>>>    Giuseppe Ghibò skrev 26.9.2010 02:09:
>>> >
>>>        I don't want to deprive the fun of building a router or a
>>>        firewall from
>>>        an old P133/64 with two ethernet cards, or some mediabox, but
>>>        often you
>>>        can't  (and sometimes you pay of energy power in a year much
>>>        more than
>>>        getting some 30-50E linksys ARM linux based router. And when
>>>        soemone try
>>>        such kind of attempts in the real world with your distro, will
>>>        be very
>>>        disappointed of failures. That's why I in some way asked a survey
>>> of
>>>        oldest hardware based on own experiences.
>>>    You still miss the point that in Mageia community there are many
>>>    users that find 30-50e a _lot_ of money, and we dont want to shut
>>>    them out.
>>> I'm not decreasing the value of the money, but rather I was pointing out
>>> the false assumption that mageia (or the current inherited mandriva)
>>> would work and would work FINE (or at all) on that hardware just because
>>> it was using a compatible instruction set.
>> Well, it depends of what you consider "FINE".
>> I dont expect people using old hardware to try to get KDE or any 3d stuff
>> to work "FINE".
>> But we have lightweight platforms such as lxde and xfce that both works
>> moderate/fast on a 200MHz+ platform with 128MB+ RAM.
>> Then if you want it as a server, its even easier... you dont even need a
>> DE/GUI, as it's manageable through console/shell.
>>  I'm not against this, but if
>>> that we wanna support that kind of hardware there is MUCH MORE work to
>>> do (I suggested a LEGACY section in the wiki, but seems it wasn't
>>> caught) than just keeping the actual flags, because in that way if we
>>> don't change then nobody will complain. Even the simple lzma payload of
>>> rpm packages requires much more memory than in the past with gzip. I'm
>>> not sure with current squashfs for the initial ram disks.
>>> I already cited there are other distro which maybe do a lot better this
>>> job. In many countries there isn't even the broadband, dialup, nor the
>>> electrical power for them. Right now you are almost assuming that a 10
>>> years old instruction set is still a no go, and that our distro is
>>> optimized like the one of the One Laptop Per Child Project.
>> I know we dont optimize for OLPC.
>> Yes, the instruction set is old, but there are many systems that are
>> older. and even if the hw is newer, it still does not enforce full i686
>> spec, as seen for example with either missing CMOV or another broken
>> register. Even Intel got it wrong with some series of the Pentium D wich
>> didn't work with i686 series builds...
>>  Sadly it
>>> isn't. But there is also a 2nd point: on old hardware it is still
>>> possible to run old software and old distros: strange but true. Such old
>>> software is still doing its dirty job. It's not that you get a trojan as
>>> soon as you put the nose out the net. There are still ways of
>>> configuring a distro on a LAN and trust in the people using the
>>> terminals locally. Many schools still use them. In a 2 hours lesson at
>>> school you can't wait half an our just to have your desktop booting...,
>> It does not take half an hour if you use xfce/lxde.
>>  the same if you plan an antispam server with latest antispam tools on a
>>> server of that category (server that was doing it's dirty job with the
>>> distro of 2 or 3 generations ago).
>> Oh, I know several servers out there running on i586 ~200Mhz that has no
>> problem what so ever keeping up with the spam/av filtering.
>>  I also tried such old hardware, but there are much less bloated distro
>>> and less bloated kernels (even non-linux ones) that do the job (or a
>>> specific duty) on such hardware a lot better than ours.
>> Maybe so, but does that mean we should force them to _not_ use Mageia ?
>> --
>> Thomas
>> ____
> To cease support for i586 seems to me to be the height of arrogance.

The height of arrogance? C'mon you are seeing politics (like the apology the
culture of the waste...or a reverse robin hood which stoles CPU cycles to
poor CPU to give to rich ones) where there isn't.

> If new i586-level hardware can still be bought somewhere in the world, it
> is still current hardware.

Also i386 and i486 hardware can still bought somewhere (including ebay) and
maybe one want to run. Why we don't lower to i386 compatatibility set
instead of i586? IIRC the i586 origin was to give something more optimized
than what was the average distro like RH. But maybe this could be changed.

> And look at how many 5-year-old, and even 10-year-old, cars are still in
> use.  Since cars have inherently a much shorter life, computers bought new 5
> years ago, or even 10 years ago, should be still be considered current
> hardware.  It doesn't really matter if most users -

There are many places in the world where old cars can't circulate anymore or
have restrictions in the zone where they are allowed to circulate (because
of laws according to the engine classification they are on). Maybe it's a
lobby of car vendors to sell new cars dunno. BTW, I've an old car. But the
car analogy is not appropriate because you can still use the car to reach a
place at a certain average speed and with certain safety levels (e.g. is
compliant against safety belts) not much different than the one of the
latest shining EURO5 models  (also due to speed limits). Apart this, cars
don't have a shorter life than PCs. While indeed I've seen many motherboard
and hard disks dying after a much shorter period of time (and repairing
would cost much more than buying a new one). Surviving ones on a such long
period of time were just very expensive one (at the time they were new)
which had a very good maintenance. In percentage almost all the MSI
motherboards died after 5-6 years (maybe after 2 or 3). Gigabyte were
similar. ASUS had the lowest percentage of failure. Intel motherboard were
too expensive to buy.

Furthermore many new i586 solution are of much more elitism than newer
hardware. A newer mini-itx with a VIA CPU doesn't cost less than an entry
level CPU of AMD (which arrives at SSE4.1 or more SIMD set) and an
all-in-one motherboard. And even hardware with some particular slot support
(like ISA) doesn't costs less than one having just a PCI + PCIe slot.

concentrated in the richer countries - have much more powerful hardware.  As
> has already been pointed out, there is 64-bit support, and a i686
> compilation of the kernel to satisfy those with newer hardware that can't
> (32-bit processor or not enough memory) or prefer not to use the 64-bit
> compilations.
> Note that the kernel is probably where most of the performance gains are to
> be made with i686, so dropping i586 in favour of i686 would give little in
> performance gains.
> After all, don't we want almost everyone to be able to use Mageia ?

Of course, but it's not a panacea that runs everywhere. My initial post was
that we were still keeping a brake on and keeping compatibility for things
that NOBODY will use or CAN'T USE for technical reasons. NOBODY means ZERO,
NICHT, NADA, NOTHING. If there is at least ONE, then it's not ZERO anymore.

So my post was to keep a BETTER support for old and legacy hardware, not
just CLAIM there is where INDEED there ISN'T or there couldn't be (because
for instance there isn't enough memory to run the installer) or that NOBODY
has TESTED for several reasons.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: </pipermail/mageia-dev/attachments/20100926/a8609cf4/attachment-0001.html>

More information about the Mageia-dev mailing list