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

Re: multiarch?

> On Thu, Jul 07, 2005 at 10:03:09PM +0200, Pierre Habouzit wrote:
>> Le Jeu 7 Juillet 2005 21:17, Josselin Mouette a ?crit :
>> > If we don't start the multiarch effort now, it won't be good for
>> > etch. Are we postponing this to the next release?
>> I hope not ... I'm a quite happy owner of amd64 machines, so happy that
>> I've only amd64 machines for my desktops, and maintaining a chroot to
>> use openoffice is quite annoying (same is true for quake/et but I
>> assume it won't bother debian that much ;p)
> Wouldn't you be better off with a native Oo.org rather than multiarch
> in this case?

World is not pure 64.
There exists software - and will exist in forseable future - which (a) is
needed for users, and (b) is not available 64bit.

Although some hacks are possible to make it run - such as 32bit chroot or
putting stuff in /emul - hacks have never been the Debian way. Debian was
known for years for it's technical quality, and it's a suicide solution to
make this no longer true.

Multiarch, although being some technical challenge, is an elegant solution
for this situation. And it is useful for all archs which have hardware
capability to run both 64 bit and 32 bit binaries (and also for those who
wish to run non-native binaries with emulation techniques such as qemu)

What real reasons (other than "it's cool") are there to run 64bit software?

- in some cases, it gives real (greater than 1%) performance gain.
- in some cases, address space wider than 32bit is useful.

Clearly such software exists. And clearly not all software is such. So what
(non-fanatic) user needs is to be able to easily - in one apt-get command
or similar - install a 64bit version of a particular package, have all
needed environment - 64bit libs etc - installed automatically and
transparantly, and have the complete system stay consistent. Multiarch
could provide exactly this.

Reply to: