Re: real-i386 (was Re: i386 requalification for etch)
On Sun, Nov 06, 2005 at 10:35:51AM +0100, Bastian Venthur wrote:
> I think we should at least consider to rename, since the current i386 seems
> to cause a lot of confusion. When even DDs confuse the meaning how can we
> expect the user to understand?
> Most people know instantanously what x86 refers to while i386 (like i(4,5,6
> 86) seems to stand for a certain processor for most people.
> I know that i368 seems to be synonymous for x86 for *some* (read: not most)
> people but I think referring to this architecture as x86 directly would
> make everybody happy. Since intel now officialy refers to x86 as IA32, we
> should consider to do so too. This would be far less confusing than i368
> and it would be consistent with the IA64 arch.
> Kind regards
This seems sensible in a limited way but i think that the problem is
not simply in the name. Up until the change in GCC which effectively
changed Debian compatibility to 486 processors and above, Debian
supported the 386 processor. There was a lot of talking on the lists
at the time and it was agreed that this was a bad situation and also
that we shouldn't rename.
GCC has changed again: full support for the Intel 386 is now possible
again but at the cost of some work. I still have three machines (two
laptops and a very old desktop) that have genuine Intel 386 inside -
but they are old. Some of the VIA chips don't play well with 586/686
optimisations (including potentially all the C3 derivatives)
but the main consideration is that the 386 is still very
much alive and well in the embedded devices space. The only two Linux
distributions still supporting 486 are Slackware and Debian - both
are distributions of choice as a basis for embedded devices because
they can be stripped down.
Current embedded devices may still be running on Woody toolchains -
can we make sure that Etch, at least, provides a _real_ 386 possibility?
Low memory requirements are also desirable - potentially an area
where Debian did superbly at one time but is now not so good: is it
worth, in general, revisiting the minimum spec. needed to run Debian
and to try to reduce the memory needed?
[I note that almost every ARM-based distribution for embedded devices
takes Debian as a jumping off point because there is no alternative :) ]
This does not address the "Debian should move to "Pentium optimised""
thread that periodically crops up when comparisons are unfavourably
drawn with Red Hat / Mandrake / SuSE / Gentoo or whoever.
The alternative is, perhaps, that we take on the tentative ulibc port
and port all the appropriate Debian packages to it - but that means
a whole new architecture, effectively, where much of the code is
Just my humble opinion,
> To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
> with a subject of "unsubscribe". Trouble? Contact firstname.lastname@example.org