Re: Some proposed package removals and additions

[Havoc Pennington <hp redhat com>]

Joe Shaw <joe@helixcode.com> writes: 
> I think that, despite my "helixcode.com" email address, Evolution is an
> absolutely essential component for GNOME 1.4. Sure, when we were
> brainstorming the release cycle a few months ago, the GNOME 1.4 release
> was primarily for Nautilus, but now when you look at it, it appears that
> Evolution is going to be ready at about the same time.
> 
> A GNOME 1.4 release with both Nautilus and Evolution will be absolutely
> killer, and the benefits in delaying a GNOME 1.4 release if necessary for
> Evolution drastically outweigh the negative consequences. Evolution is
> more than just a mail client (although it does do that rather well): it's
> a full groupware suite, with an addressbook and calendar that are tightly
> integrated through bonobo with the shell and mail components. It, along
> with Nautilus, represent the next generation of GNOME applications and
> truly showcase the benefits of this infrastructure that we have spent so
> long developing.
> 

My problem with this is that we aren't comparing GNOME with Evolution
to GNOME without Evolution. We're comparing GNOME 1.4 with Evolution,
to GNOME 1.4 with Evolution beta followed quickly by a stable
Evolution release a month or two later, probably as part of GNOME
1.4.x. I don't really see the point in delaying 1.4 given that.


My general approach to releases is, "every X period of time (where X
should be 6-9 months or so) we freeze everything, wait for it to
stabilize, and dump it out as a release." Every other release or so,
you can also include incompatible library revisions.

If you try to rationalize releases ("we should have feature X, Y, and
Z to release") then you end up in the Debian Zone (release every 2
years). The only way to get regular releases is to _just release_ on a
regular schedule. Then any given feature is guaranteed to come out in
a release within 6-9 months of its completion.  (And applications can
of course come out separately from GNOME, well before that.)

There is _always_ one more feature. If you want users to have the most
up-to-date stable stuff most of the time, then you need to just
release often. The "try to get as much as possible in a release"
approach results in inferior testing and stability, and getting things
to users more slowly. It also makes each release like Doomsday, where
developers panic if they can't get stuff into it, and the result is
even longer delays. 

Release early, release often. Then each release is no big deal; not so
much code has changed, and if things don't make the release, it's not
a major issue.

The one caveat is that you should have at least _some_ significant
feature in each release, and Nautilus is the one we decided on many,
many months ago, with the rationale that gmc is the one glaring piece
of crap in the GNOME desktop.


That said, unless Evolution is delayed beyond its planned schedule
this is a non-issue. ;-) So get coding and start your freeze on time
so you have time to stabilize it. ;-)

Havoc