Re: debian ppc64
On Wed, Dec 08, 2004 at 07:38:07PM +0100, Andreas Jochens wrote:
> On 04-Dec-08 19:09, Sven Luther wrote:
> > Well, the decision if we will go for a native 64 port, or a bi-arch solution
> > has not yet been taken, so ...
> I am convinced that in the end there will be a native 64 bit port and
> people will use that one. 32 bit binaries will be legacy and only
> used in special cases like old closed source software. It will
> certainly take some time but ultimately this will be the result.
> So why not do it just now?
Because ppc was designed with 64bit in mind, and there is really not su huge
an advantage of using pure 64bit software, and even a disadvantage in the code
creep. Also, full ppc archive is right now around 16-18GB, for binaries only,
the ppc64 will probably grow upto 20GB on top of that. Debian has no ppc64
autobuilders, and the ppc64 userbase is a real minority right now, and
probably in the few years to come.
> > on the glibc situation. The glibc situation is obviously blocked by the sarge
> > release, so it is not time to go steaming ahead, but to consider our options,
> > and see what is best.
> I fully agree. I have no intention to push anything forward before the
> release of sarge.
> > My own plan is to get a biarch compiler, possibly even hacked somehow to be
> > part of sarge, and then build a set of power3 and power4 ppc64 kernels, and
> > maybe a statically linked pstools or whatever they are tools. More cannot be
> > done for sarge.
> I hope that you will succeed with this plan. A ppc64 kernel in sarge
> would of course make things _much_ easier for any 64 bit porting efforts
> regardless of the decisions which have to be made. If I can help by
> testing kernels or tools or otherwise please let me know.
Exact. what hardware do you have anyway ?
> > And have you thought what this would mean for mirror space ? The pure64 amd64
> > got already vetoed by the debian infrastucture admins, since it would kill our
> > mirroring network, so ...
> Is 5 GB extra mirrorspace for a new port really so critical today?
5GB ? You are being extremely optimist on this one, i think. I would say more
like 15-20, for the full archive (stable, testing, unstable, experimental).
Also the problem is not mirrorspace, but replication bandwidth.