People are usually interested in two things - the new cool version
(definitely majority of list participants) and a recent version that's
known good.

The new cool version is easy to locate - just d/l anything, and then
update it. Under the snapshot regime, we would have this available less
of the time. To me the fact that there's always something happening is
very cool.

OTOH, making the current newest a snapshot will not work that well for
getting a stable version - stability comes from testing. Like you say,
we don't have enough feedback from active stable-version users to make
"stable means well-tested" mean anything.

A cheap mans version we could have is this - have a bleeding edge
version, with everything that's out, just like now. Other than that,
alway maintain a pointer to some past version, that experience has shown
to be stable. By this I mean you'd never say "here you go, this new
version is stable, we really got 'em all this time", you'd just
retroactively say, "yup, 7684 was when all the semi-dynamic-synheritance
trouble began. 7683 was pretty stable though...".

OTGH, when we have modules, we might find that things run significantly
differently than now, so making a new scheme might be useless.

Just my 2 cents.
Daniel

Bruce ONeel <beoneel at bluewin.ch> wrote:
> Hi,
> 	I wonder if, for a couple of reasons, we shouldn't rethink 
> the whole concept of Squeak releases.  First, Squeak is quite
> stable, even the alpha/testPilot/unstable versions most of the time.
> Second a lot of cool new features keep popping up in the newer 
> images, whether they are things such as the JPEG movies just
> added or UI enhancements.  Third, while many bugs are fixed, 
> very few of the bug fixes find their way back into old versions,
> even if those versions are yet to be released.  3.1 springs to mind.
> There just doesn't seem to be the interest there.  Most people on
> this list seem to keep up with the current set of updates.  It is just
> too hard to ignore the most recent cool set of changes.  The flip
> side to this is that back ported bug fixes aren't
> tested by 10s or 100s of people daily like the current set of
> updates are tested.
> 
> 	Maybe a better choice would be to do snapshots.  For example,
> 3.2alpha seems fairly stable right now.  Maybe there could be a
> freeze on updates for N units of time, and if no one finds some 
> horrible bug we just call that say 3.2 and be done with it.
> The only change that I would recommend is that in one of the
> welcome windows, near the top since no one reads past the first
> page of readmes, there should be some words along the lines
> that this is just a snapshot and you should join the email list
> and/or keep checking the web site and ftp site for new 
> snapshots.  We should probably think about doing snapshots
> every M units of time, maybe every two months or so.
> If snapshots end up appearing every few months maybe
> the startup code should warn if the image is more than
> 6 months stale.  The snapshots don't have to be hard 
> freezes either.  If some snapshot has some horrible bug then
> it can just be replaced with a more recent one.  One of the
> ideas is to remove the finality of calling things releases.
> 
> 	The download page on the ftp site would just list the most recent
> N snapshots and the VMs for the different systems.
> 
> 	I think the advantage of this are that the "releases" would
> coorespond closer to what the active Squeak community
> is actually using, and would therefore have the chance of
> having the most recent bugs actually fixed.  For those who need
> a reference it would still exist, it just would be the 
> 30/11/2001 snapshot perhaps, rather than 3.2.  I think
> the other advantage is that this moves the Squeak 
> release process closer to the reality, even if it
> isn't exactly the reality we might like.
> 
> 	The disadvantage?  Squeak does not end up with
> nice clean releases.  OTOH, many other software projects
> really don't have clean software releases either.  How
> many patches/service packs/updates/whatever have
> you installed this past year?
> 
> Anyway, this is just an idea.  Thoughts?
> 
> cheers
> 
> bruce
> 
> Dan Ingalls <Dan at SqueakLand.org> wrote:
> > Folks -
> > 
> > I am now "clearing my desk" prefatory to finally getting Henrik's Modules out (sorry, I just got swamped a month ago).  One thing that remains to be completed is to release 3.1, but as I reflect on this, I wonder if we mightn't just put 3.2 into gamma status, let people beat on it for a week, and release THAT instead.  Here are some reasons:
> > 
> > 	It has a bunch of cool new stuff not in 3.1
> > 	It seems pretty stable
> > 	If has if anything more fixes than 3.1 right now
> > 	We are (honest ;-) about to enter a real test-pilot period
> > 
> > I don't care a lot one way or the other, but I think this would be a bit less work for me as well.
> > 
> > How to make the decision?  Well, assume I've decided to do this, and only send mail if you are strongly opposed.  If no one complains in the next day or two, that's what I'll do.
> > 
> > Thanks
> > 	- Dan