Re: GNOME 3.0 in March 2011

[Steve Frécinaux <nudrema gmail com>]

On 07/30/2010 12:13 AM, John Stowers wrote:
> I have no objection to that. My problem is that in the space of one
> minor release, *every* python plugin for a gtk application (e.g [1]) has
> skipped this 'dying' phase and move straight to 'dead'. That is not a
> nice backwards compatibility story for python developers.
>
> I offer a third option. Plugins add 'pygtk.require(3.0)' [3] to their
> code and libpeas support loading legacy plugins.

It's not as simple: you can't use pygtk and pygi at the same time in the 
same program. So you would have to choose between running pygi-based 
plugins or legacy pygtk ones. And those legacy plugins would require 
some changes anyway: Gedit.WindowActivatable is not the same as 
gedit.Plugin...

The only fix would be to allow running both at the same time, for 
instance by reimplementing pygtk as a compatibility layer over pygi, but 
then there would be no need for libpeas to support pygtk explicitely.

> You suggest that you do not want to maintain def files any more. Fair
> enough, but I suspect that the C-api between the minor releases of these
> apps would be retained (as is expected of C-apis). If the C-api has been
> retained then what is the maintenance burden of keeping the defs file
> around which describe this API?

Actually, I don't want to *write* those, and especially the required 
overrides.