Features vs. Time-based [Was: Gnome 2 infinity and beyond]

[Jeff Waugh <jdub perkypants org>]

<quote who="Alan Horkan">

> > Chris Lahey wrote:
> >
> > I think if we are targeting features,
> 
> ... no not really, I wasn't suggesting targeting features any more than is
> done within the current 6 month release cycle.

I call 'Brooooookeeeeen!' 

We hardly even have an agenda for our time-based releases, because after a
lot of discussion about time-based vs. features when we chose this strategy,
we became allergic to creating feature plans. This is broken and wrong - but
fixable. (I discussed this a bit in the 10x10 talk.)

Of course, we won't fix it by saying that we're happy for GNOME 3.0 [1] to
have no agenda as well.

Chris got it right. To fix the lack of agenda, we need to set an agenda that
is independent of the release cycle - particularly for bigger goals that we
think about for Topaz. That doesn't mean dumping the time-based releases. It
means that we do work around and outside the release until it's ready to
ship. Look at Evince - it wasn't created in a six month period with a two
month freeze, it was developed externally and dropped in when it was ready
to ship. Sure, that's a logically separate app, but we can do bigger changes
this way too - shipping an alpha version of a rocking new feature in March
or September with 'technology demo' plastered all over it would be great.

There is *NO PRESSURE* to call something 'GNOME 3.0'. We can do it when
we're ready. We can deprecate the crap out of our platform *RIGHT NOW* and
be happy that it's still API/ABI compatible. We can write amazingly cool new
stuff *RIGHT NOW* and drop it into the 2.x release series when it's ready.
When we decide that GNOME is qualitatively worthy [2] of the 3.0 moniker, we
can purge the deprecated crap, make big changes to the OOTB experience, and
make a fuckload of noise.

Until then, there's ideas to spec and code to write.

- Jeff

[1] Holy shit, just stop talking about version numbers at all. It totally
    doesn't mean anything useful.
[2] There's no point calling something '3.0' based on self-satisfying but
    ultimately boring technical blather. The whole "let's break API/ABI and
    call it 3.0!" thing is a red herring.

-- 
FISL 7.0: Porto Alegre, Brazil        http://fisl.softwarelivre.org/7.0/www/
 
                    The Unix Way: Everything is a file.
                 The Linux Way: Everything is a filesystem.