Re: gtk2-perl stable releases

[Murray Cumming <murrayc murrayc com>]

On Wed, 2004-03-31 at 14:55, Ross McFarland wrote:
> On Wed, 2004-03-31 at 07:12, Murray Cumming wrote:
> > On Wed, 2004-03-31 at 14:01, Ross McFarland wrote:
> > > On Wed, 2004-03-31 at 05:56, Murray Cumming wrote:
> > > > On Wed, 2004-03-24 at 04:50, muppet wrote:
> > > > > Now that gtk+-2.4.0 is out, we are able to make our Glib and Gtk2 
> > > > > modules stable.
> > > > 
> > > > I'm not very happy with you declaring total API/ABI stability before the
> > > > date on the GNOME Platform Bindings schedule. Why the rush? Let's just
> > > > hope that you don't need to change anything before then.
> > > 
> > > we decided that it would be a good idea to do a stable release with
> > > enough time left to do another if need before we had to submit to the
> > > language bindings. if that release stands works fine and doesn't have
> > > any major (non-api) problems then it will go into lang bindings. if not
> > > then we'll have enough time to rel another for inclusion. 
> > 
> > You can't call the ABI stable and then break it. How would you feel if
> > GTK+ did that? This is all about being stable and making people feel
> > confident that we are.
> 
> in general what's ABI stability for perl modules?

It is something that would cause existing applications to break if you
change it. It's probably the same as API for you.

>  we do have a few
> things that might be considered ABI, but mostly to perl binding authors,
> not really app developers. that's not the kind of stuff we'd be
> expecting to change anyway, much of that has been stable for months.

OK, that's good. But the point is that the schedule does not require for
you to declare it 100% API stable yet. The schedule has time between the
API freeze and the full-really-100% API-freeze, so that problems can be
worked out. I don't see the point of giving yourself less time. There is
no rush, because the schedule is known.

What about the developer who thinks he still has time to contribute a
last-minute API fix?

> the problems i would expect to see and wanted time to find fix would be
> pkg, build, portability, documentation related, etc. etc. 

Predicting is difficult, but the schedule gives time for this.

> > > a sizable group of people wait around for a stable rel before
> > > downloading and trying out new stuff. if that didn't happen until the
> > > bindings deadline we would of had a whole lot of new people testing (and
> > > potentially finding new problems) with no time/way to fix them. 
> > 
> > That's why we have a schedule - people know when they have their chance.
> 
> if we meet the language-bindings schedule what does it matter if the rel
> we put in it was actually out a week before it had to be?

See above. It's not a big deal in this instance, if you don't actually
plan to break ABI, but I'd like people to follow the schedule, because
the schedule is helpful.

-- 
Murray Cumming
www.murrayc.com
murrayc murrayc com