On Tue, Sep 18, 2001 at 02:04:25AM +0000, Hans-Joachim Picht wrote:
> > I've decided to ask this question here, in the hope to get feedback
> > from both the developers and the (potential) users of Owl --
>
> As far as it seems not many people seem to be on this list or are likely
> to give feedback :(

Yes, only 60 subscribers here so far and most would leave giving
feedback for others or wouldn't post it to the list. ;-)

> > Do we need an Owl 0.2-prerelease (possible in a few weeks from now)
> > and then its corresponding stable branch (to replace 0.1-stable)?
>
> I would not release a second stable relase, only bugfixes and patches
> and a couple of current-snapshots in order to make extensive testing
> possible, parallel to the direct current tree availabe on the ftp
> mirrors.

Is extensive testing not possible without older current-snapshots?
There're not that many bugs being discovered to require exact version
information in bug reports.

It really makes sense to make a current-snapshot right before likely
breaking current with newer glibc and Linux-PAM, but then it's not
very different from making an 0.2-prerelease. The only difference is
that it wouldn't be maintained, is that what you meant? Then if a
really bad bug is found in it, people who decided to actually use it
(as I expect it to be quite stable in practice) would have to downgrade
that piece to its version in 0.1-stable, or remove it from the system
if there's no 0.1-stable equivalent. Yes, that's a possibility.

BTW, at DataForce ISP we already use post-0.1 Owl-current, including
on two virtual hosting servers (about 600 hosted domains). So we will
actually need at least a pre-broken current-snapshot ourselves. ;-)

> > The situation is this. Right now, Owl-current is binary-compatible
> > with 0.1-prerelease/stable in the sense that both upgrades to -current
> > and downgrades to -stable are possible with "make installworld" and
> > individual packages from -current may be installed on 0.1-stable.
>
> As most of the the major distributions gnu/linux distributions have a
> stable and unstable branch, f.ex running kernel 2.2 and glibc2.1 in in
> stable and kernel 2.4 with glibc 2.2.* in unstable a upgrade is possible
> but a downgrade would be a bloody job.

That's exactly my point, and the question was -- should we bother to
release and maintain a stable branch that has everything we already
managed to get into current before breaking the downgrade possibility
and actually making current unstable for a few months. Right now it
has a few extra packages and extra features in older packages, and I
expect it to have more before the big changes to core components.

-- 
/sd

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]