Re: Proposing module: PyGTK

[Elijah Newren <newren gmail com>]

Hi,

On Wed, 03 Nov 2004 13:53:01 +0100, Johan Dahlin <johan gnome org> wrote:
> > The Bindings release exists so that applications can depend on the APIs
> > that it offers. I think that we gain nothing if we move a module from
> > Bindings to Desktop, but I think that it would undermine the Bindings
> > release set. I am very much against it.
> 
> I think both Gnome and the Bindings has a lot to gain if we move it to
> Desktop:
>   1) Its now completely acceptable to depend on the bindings in any
>      applications in the desktop, eg for scripts/extensions.
>   2) You can write modules/applications which later has the possibility
>      to be included in the desktop release
>   3) The bindings will be used in more applications, outside or inside
>      of the desktop release thanks to the "official stamp" it get.

All of these could be summarized as "we need an official stamp of
approval that we can point to."  I agree that it appears the previous
consensus isn't enough and that we need something more.  I agree with
Murray that moving pygtk to Desktop or Platform release sets is a bad
way to do that.  I think the core problem is that 
http://developer.gnome.org/dotplan/ spells out the rules for the
platform and the bindings, but not for desktop modules.  So:

I propose writing a page containing the requirements, expectations,
and even non-requirements for Desktop modules.  This will include
things like following HIG, u7y, a11y, i18n; it's okay to break API
(though sparingly, it is hoped); okay to depend on freedesktop.org
modules (?); okay to depend on anything in the platform, etc.  And, of
course, it would be stated that modules in the desktop cannot depend
on items in the Bindings release (unless the dependency is optional,
such as for gnome-games) unless consensus is first attained on
desktop-devel-list.  And of course, the page would state that pygtk
has gained that consensus (and is the only such binding module to have
done that so far).

Comments?  Ideas?  Objections?

> > If you insist on having a precendent, then you might consider pygtk to
> > be like the freedesktop libraries. We don't need them to be in the
> > Desktop release, because they are already part of an existing
> > "platform".
> 
> That's an a idea, but it depends on libraries in the developer platform
> so then it needs to go into a special middle layer, desktop depends on
> it, but itself depends on the platform without being in the platform.
> Seems like putting it in the desktop is an easier solution to me.

Seems like writing up a page clarifying the requirements is an even
easier solution and much cleaner because of:

> > Yes, and that's why there's a Bindings release set. In fact, by moving
> > pygtk from Bindings to Platform, you would no longer be saying that it
> > is an API-stable platform, because we have always said that the
> > libraries in Desktop are not API-stable. Libraries should move _into_ a
> > platform release set, such as Platform or Bindings when they are API-
> > stable, not in the other direction.

Cheers,
Elijah