Re: Suggestion: Skip Slink!
On Tue, Jan 05, 1999 at 11:17:43PM -0600, John Goerzen wrote:
> On Wed, Jan 06, 1999 at 12:11:05AM -0500, Avery Pennarun wrote:
>
> > > If, for instance, there's a critical bug in some vital package (say,
> > > X) that only manifests itself on big-endian machines, the release
> > > should and would be held up to fix it. I think you may feel
> > > differently if you ponder what would happen if that logic were applied
> > > to the i386 version. Everyone else releases, but no i386 release.
> > > Could be interesting, don't you think?
> >
> > In that case, the i386 version is broken, so don't release it. The
> > others are not, so WHY NOT release them? I wouldn't feel at all
> > surprised or upset
>
> No, that's not proper either. Debian is a set. Hamm was a multi-CD set:
> i386, m68k, contrib, sources, etc. Debian is not Debian if it's missing
> for a particular platform. (Again, exceptions for platforms just starting
> out and truly not ready for release.)
I disagree. If slink/i386 is called "frozen" but is exactly the same
version as it will be when it is released, then it might as well be
released. If slink/sparc isn't ready at that point, it's not a "missing
part", it's just a late one.
You are saying we should inconvenience thousands of people who could be
using slink/i386 today (in our imaginary world) except that the
irrelevant (to them) slink/sparc isn't ready.
Releasing the i386 version without sources would be leaving out a large part
of Debian in almost everyone's eyes. Releasing the i386 version without the
sparc version doesn't impact most i386 users, and sparc users (who don't
need the i386 version) should find it irrelevant.
> Also don't forget that the source CDs contain source for *all* archs. It
> may not be possible to get accurate source CDs if different archs are
> released at different times, putting Debian and CD vendors in the position
> of violating GPL. Bad, bad, bad.
Violating the GPL? If packages are different between i386 and sparc (which
they regularly are, since non-i386 arches have fewer developers yet), then
the FTP site already violates the GPL by not including two different source
archives (or diffs). We know about that, it'll probably get fixed sometime.
Meanwhile, releasing a CD with the same files on it is no better or worse.
Once the problem is fixed, CD's won't violate the GPL if the FTP site
doesn't.
But there are good reasons for synchronizing releases...
>From the point-of-view of an administrator, I can see big advantages in
having exactly the same package versions in slink/i386 and slink/sparc --
that way I can run a heterogeneous network really comfortably, expecting the
same bugs/features on all platforms.
However, if slink/sparc is released with a few changes after slink/i386,
there's nothing stopping the i386 people from rebuilding their packages with
sparc's changes in order to catch up. We just call that Debian 2.1r2, and
sparc never made a 2.1r1 release. If I'm an admin concerned with
consistency, I stayed with hamm until all my required arches were stable.
Have fun,
Avery
Reply to: