Re: Status of 'sarge' for the amd64 architecture
On Fri, Apr 29, 2005 at 03:50:37PM +0200, Andreas Barth wrote:
> * Martin Waitz (tali@admingilde.org) [050429 15:40]:
> > On Mon, Apr 25, 2005 at 05:22:53PM +0200, Andreas Barth wrote:
> > > Why not? removing arm from testing does not change at all the number of
> > > binary arm packages being pushed each day, as the packages between
> > > testing and unstable are shared (and only few packages go in via t-p-u).
> > > So, the only win is that packages are faster removed - but as unstable
> > > and testing are quite in sync, even this is not so much difference.
> > > Adding a new arch however adds a lot of new binary packages to be pushed
> > > each day
> >
> > well, those should be about as much as are saved by removing another
> > arch -- once the new architecture is uptodate in testing and unstable.
>
> Actually, that is exactly what is planned post-sarge (well, not removing
> an arch, but splitting the archive so that mirrors are only required to
> carry some of our current architectures). There is a simple reason why
> we don't do it now: We prefer to use the ftp-masters time for resolving
> issues we need to release sarge. (And, BTW, of course an architecture
> won't be considered for inclusion in sarge unless we have tested it for
> a decent time in unstable, so even adding amd64 to sid today won't make
> it an sarge architecture, except if we want to delay sarge even more.)
>
>
> > * too much bandwith needed to update all mirrors.
> >
> > do all mirrors sync with ftp-master? would it help to establish
> > a mirror hierarchy where only a few selected mirrors are allowed
> > to connect to our master server?
>
> This is already the case. But there are places where our _mirrors_
> bandwith is too expansive to make the daily pushes even larger.
Sorry, but I still don't understand it:
You could continue to offer the complete archive as it is today, and it
shouldn't be a big amount of work to offer one or more partial archives
(e.g. only stable or only i386) from different locations - and a mirror
with bandwith problems could simply switch to using a partial archive.
This wouldn't be as complicated as the SCC proposal, would have exactly
zero impact on release management and should be implemantable within a
few days.
Considering that this might make it possible to ship amd64 with sarge
which would have a positive effect on the reputation of Debian, could
you please explain which technical problems I do oversee when thinking
that the technical problems of such a solution were small?
If such a solution would e.g. take two weeks and would have been
implemented at the day of the SCC announcement, it was running for one
month today...
Could someone please enlighten me?
> Cheers,
> Andi
cu
Adrian
--
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed
Reply to: