On Sun, Dec 14, 2003 at 03:58:21PM +0100, Ingo Juergensmann wrote:
>
> Looking at the graphs ti seems obvious that the way how to get buildds
> running again is known for about 5 days now.
> And 5 days are not enough time to inform other archs or give them access as
> well?
> Why should it be easier to get the buildds on mips(sel) and powerpc running
> than to tell others how to do the same? Please give explanation.
> AFAIK the source of buildd is the same for all archs. So, I can´t see any
> difference in setting up the buildd for other archs than setting it up for
> the above mentioned archs.
> And when the source should be different now, why haven´t the other archs be
> informed to build a new buildd from CVS?
I would hazard a (fairly strong) guess that the source code involved in
running the buildds has not changed appreciably. That isn't what has to be
done, to restore a buildd to a trustable status.
Remember, these machines are, behind the archives, perhaps the most
implicity trusted machines in the entire project. Compromise the archives,
and you can silently sprinkle trojans throughout any package on any port.
Compromise a buildd, and you can silently sprinkle trojans throughout any
newly compiled package on one port.
Clearing out and restarting the buildd itself probably takes a nearly
negligible amount of time - at least, that's been my experience, when
experimenting with the entire buildd/wanna-build setup, for the NetBSD
porting work, to figure out which things were actually required, and which
were just nice.
On the other hand, blowing away a machine without losing the *valuble* data
on it, then manually checking that data before it goes onto anything new,
along with a complete reinstall, can be a pretty non-trivial task, and one
that often requires console access - which, in itself, may be a non-trivial
task for a number of these machines.
Why should it be easier to get the machines Ryan works with regularly
running again? Probably because he knows how to arrange any required
access, where there might be data that needs to be copied/inspected, what
that data might be, and what it SHOULD look like, along with probably
having installed the machines in question at least once, and thus being
familiar with any quirks they may have. Oh, and he can probably GET to
them, which may well be physically impossible for him with others.
Thus, he probably has little choice, in some cases, but to depend on others
to deal with some of hte work, and try to coordinate with them (some of
whom may be as much as 10 hours offset from him, which I can tell you
from experience coordinating things between the US and the "Far East",
is no small handicap). And, as has been pointed out to you, it has been
*one* business day since the 12th, assuming that message went out at the
beginning of the 12th and not the end.
Three weeks is a long time to go without a reply. Three days is not. Even
outside of Debian, which I will cheerfully admit has (and often rail
about having) some communication issues, three days just isn't a crisis.
Particularly not when dealing with things that *paid professionals* can, at
times, take a week or more doing, when being paid 8 hours a day.
Let's save it for the really egregious times. So far, the entire recovery
has been suprisingly *well* communicated, compared to a lot of points in
Debian's history.
--
Joel Baker <fenton@debian.org> ,''`.
Debian GNU/TBD**BSD(i386) porter : :' :
`. `'
`-
Attachment:
pgpjaqJZyh8F4.pgp
Description: PGP signature