Bug#240896: not pending anymore
Michael Banck said:
> On Fri, Jun 04, 2004 at 03:26:05PM +0200, Wichert Akkerman wrote:
> > We should just pick a name that will not confuse people
> [...]
> > which includes it should be one that is used elsewhere (think
> > toolchain and kernel) as well.
>
> The question is: Do we want to pick a name that does not confuse users
> or developers?
>
> Sure, the toolchain uses x86-64, but only developers get in touch with
> this.
Or x86_64.
> On the other hand, amd64 seems to be well introduced in public now so
> that potential users will recognize it as the port they should install.
I keep hearing this argument but I am not sure it is correct. Just as
people do not buy i686 processors (they buy Pentium IV processors), they
do not buy AMD64 processors (they but Opteron or Athlon64). So it really
seems the only people who know what AMD64 refers to are the same people
who know what x86_64 refers to.
Therefore, the decision should like be made for technical reasons. The
current technical considerations I have heard are the following:
* what works with current Debian tools?
This is amd64 since (1) it is already in use (doesn't break current
builds) and (2) it does not contain a dash or an underscore.
* will other similar architectures, e.g., Intel Pentium EM64T (also known
as IA32e), be on this port?
I have no idea the answer to this question. Multiarch may make this
question irrelevant. One can envision someone setting up a single
multiarch system that can be installed on i486, i586, i686, amd64, and
ia32e. This would be similar, although much more powerful, than using
i386 today on an Athlon but also installing and running the kernel-image
for Athlon. That is, the bulk of the system would be vanilla i486, while
the compiler, linker, and other toolchain programs would be capable of
producing all the types of binaries and certain programs which require,
for example, more memory would have arch specific versions installed.
If this comes to fruition, it would make sense to keep amd64 and ia32e
separate. Indeed, gcc will likely need to support both independently so
their choice of x86_64 may need to be revised. So perhaps the answer to
this question also points to amd64. Time will tell.
dd
--
David Dooling
Reply to: