[F8/multilib] {,/usr}/{,s}bin64

[Jeremy Katz]

On Tue, 2007-05-01 at 07:03 -0400, Matthew Miller wrote:
> On Tue, May 01, 2007 at 12:27:03AM -0400, Bill Nottingham wrote:
> > - installation of third-party software that is only available for the
> >   secondary arch, in a way that allows it to run
> 
> How much of this is there in the real world? I mean, we're talking dozens of
> packages at *most*, right? Matlab, Mathematica, Maple, and Oracle all have
> 64-bit versions. So in our environment at BU, that leaves Splus (rapidly
> being displaced by R) and the Flash plugin.

There's actually a fair bit of legacy software that falls into this
category.  And then also various site-specific custom software that was
built once and continues to be used to this day.  If people didn't want
compatibility to run these legacy applications, then they would have
bought into Itanium ;-)

> And I'm not convinced that the ability to build 32-bit code outside of a
> chroot is a huge deal either. In five years, that's going to be a really
> narrow group of users, and do we really want to design our whole system just
> for that?

There's a lot of lower power devices which are still 32-bit only and are
going to be for quite a while as far as I can see.  And being able to
trivially build for them is one of the reasons why you go for x86 for
them instead of an embedded specific chip like an arm or a mips.

The compatibility and the fact that it's all but entirely transparent is
a huge part of the value of the x86->x86_64 transition.

Jeremy