Timed releases, versions, etc... [was Re: Freeze proto.h?]

Christian Perrier bubulle at debian.org
Tue Apr 22 05:47:26 GMT 2008


Quoting Dan Shearer (dan at shearer.org):
> On Mon, Apr 21, 2008 at 10:56:24AM -0500, Gerald (Jerry) Carter wrote:
> > Jeremy Allison wrote:
> > 
> > > That's a lot of discipline needed to make that work.
> 
> If the six months are approximately aligned with Fedora/Ubuntu there
> should be a natural cycle of people more looking at their/our bugzillas
> coming up for release time.


I had this discussion during SambaXP with Karolin.

I think that 6 month release timeframe is very ambitious....maybe too
ambitious.

Ubuntu can keep up with that release cycle only because they have
Canonical behind the scenes, investing lots of money in making sure
that releases will happen. It surely works...but can Samba afford
this?

I less know the Fedora release management methods but I suspect that
"some" company still invests quite some money and developer time in it...

For sure, Samba is not an entire Linux distro. But, still, the current
direction seems to be maintaining two code branches for suite some
time (not even speaking about samba4) as, from what I understood,
3.0.* will live for quite some time after 3.2 is out.

I'd love to be proven wrong but my general feeling is that, even with
(what I understand to be) the obvious commitment of SerNet to directly
or indirectly pay someone so that Samba releases happen, things can
change during time....and, still, that makes the release management a
huge work....therefore having a 6-month release cycle really hard to
achieve.

Again, nothing wrong here about key participants in a project being
paid to work on it (some around will know that this is a topic that
made huge debates in Debian in 2006)....but that doesn't even solve
the obvious fact that 6-month cycles are very fast for efforts mostly
based on volunteer work.

Martin Michlmayr (former Debian project Leader back in 2004/2005)
included such study or release management in his PhD thesis work about
"QA improvement in Free Software". You'll find details on
http://www.cyrius.com/journal and could bring more arguments in this
discussion.

PS: OK, I agree that my statement arguing for a slower RM process
comes from someone involved in the Debian project that's well know for
its "slow" release management (non) schedule... feel free to flame us
for that!





More information about the samba-technical mailing list