[PATCH] set CONFIG_NUMA on PAE kernel, i686 only.

[Arjan van de Ven]

On Tue, 2006-07-25 at 13:46 -0700, keith mannthey wrote:
> On Tue, 2006-07-25 at 15:43 -0400, Dave Jones wrote:
> > On Tue, Jul 25, 2006 at 12:29:34PM -0700, keith mannthey wrote:
> >  > On Tue, 2006-07-25 at 14:34 -0400, Dave Jones wrote:
> >  > > On Tue, Jul 25, 2006 at 11:24:42AM -0700, keith mannthey wrote:
> >  > > 
> >  > >  >  Modern Summit hardware x460 is EMT64e (x440,x445 is not) but as you
> >  > >  > know sometimes customers run 32bit OS's on 64bit boxes.
> >  > > 
> >  > > I hear this from time to time, and every time I ask "why doesn't
> >  > > 32bit emulation in 64bit kernel work for you work?", the only
> >  > > answers that seen to come back are various flavours of
> >  > > "we didn't know about it" or "haven't tried it".
> >  > 
> >  >   Is there a "Guarantee" that 32bit apps will just work with 64bit
> >  > kernels?  I don't think all 32bit apps test (or work) on 64bit OS and a
> >  > such don't fall in the "Supported" configuration. Customers like to have
> >  > ground to stand on when things break.  If they run unsupported in their
> >  > software stack things can get messy (who will fix it?) if problems arise
> >  > or performance is bad compared to 32bit native. 
> > 
> > This is exactly the sort of hand-waving I was talking about.
> > If something doesn't work, lets find out why and _FIX_ it.
> > Scaremongering, and handwaving isn't going to solve a single bug.
> 
>   As noted ISV's move slow and customers even slower.  Some customers
> would still buy AS2.1 if they could.  I have head customers say (this
> year) that the can't move to 2.6 (RHEL 4) because it hasn't been out
> long enough and saw the risk as too high.  It is out of my scope for me
> to say what all the reasons are for why customers don't just run 64-bit
> all the time, but they don't.  I think change takes time. 
>  

but how is that type of customer relevant for FC6 ?