LP#74179 -- support for i386 machines with >3GB RAM
Andy Whitcroft wrote:
> On Tue, Jan 27, 2009 at 07:38:46AM -0700, Tim Gardner wrote:
>> I am not in favor of another flavor. Despite the reluctance of some
>> folks to use the server flavour, it _does_ solve their use case with no
>> effort on our part. If these users are that sensitive to the
>> optimization differences between generic and server, then they'll have
>> to run a 64 bit kernel.
>> My goal is to actually reduce the i386 flavours for Jaunty+1 to just
>> generic. Hardy should be the last LTS that supports a 32 bit server.
> Then if we plan on having just one flavour in N+1 that would have to be
> PAE enabled.
Quite the opposite IMHO. The only flavour left in i386 would be non-PAE
in order to support those 32 bit CPUs that do not have PAE. While there
are still 32 bit CPUs with PAE support, they are becoming old enough
that I'm no longer interested in supporting them. The number of those
that have more then 4Gb is even fewer.
> I thought the issue there was that some older h/w doesn't support PAE
> necessarily (i want to say lpia is in this situation though am not sure
> why) so we would end up needing an PAE and non-PAE build still?
LPIA is a completely separate arch from i386, so the PAE issues aren't
relevant. We'll do what is needed to satisfy the mobile team.
Remember, I'm trying balance the use cases against maintenance and build
overhead. I'm open to opinions, but I think many of these use cases are
falling below the cut line.
Tim Gardner email@example.com
kernel-team mailing list