Ben Hutchings wrote:
>On Sun, 2012-05-20 at 11:27 +0200, Goswin von Brederlow wrote:
>> Ben Hutchings writes:
>>> Eventually (wheezy+2? +3?) we would stop building a kernel package
>>> for i386.
>> As in drop the i386 arch?
> No, keep i386 userland only. Though we might consider reducing even
> that to a 'partial architecture' that has only libraries (similar to
> ia32-libs today, only cleaner).
I'd love to see that happen someday, but at the moment, new x86 systems
still get sold that don't support 64-bit. Notably, many low-power Atom
processors still don't support 64-bit. If at some point 64-bit becomes
a required feature on all new x86 processors, with a definite indication
that no new 32-bit-only processors will ever show up, then a few years
after that this change could become reasonable.
The rest of your plan, namely migrating the 64-bit kernel to a multiarch
package instead of an i386 package, seems appropriate as soon as testing
shows that multiarch can smoothly handle it (which probably means after
the wheezy release, for safety). And automatically enabling multiarch
based on processor capabilities seems like a good plan as well;
eventually, that'll start becoming necessary for manageability, once new
partial architectures start showing up for more fine-grained processor
- Josh Triplett