[Portland] Thoughts on the integration tasks

[Billy Biggs]

Lubos Lunak (l.lunak at suse.cz):

> In Rudi's case, as I see it, the thing to be designed would be only
> the API itself (the IPC), not a library. Then toolkits like Qt and Gtk
> could use it directly, while for others who wouldn't want to do this
> there could be a simple library (that'd be more or less just simple
> IPC<->plain C wrapper).

  If the intention is that the primary consumer is GTK+/Qt, and that a
goal is compatibility with existing infrastructure (such as XSETTINGS,
or xdg-mime, or _NET_WM), then I'm all in :)

> >   - kwin versus metacity behaviour is always fun :)
> 
> If I remember well it's usually just you assuming too much about how a
> WM works, or plain bugs. I don't think there's anything related to WMs
> that needs special care. If you have farther requirements for WM
> functionality, there's the WM spec list.

  I don't doubt that it's just bugs, and we can always argue about
what "assuming too much" is.  All toolkits expose some API that either
isn't quite implementable, or have some mental model that just doesn't
quite fit with some WM, and this causes problems.  I think my point was
more that there's a significant amount of QA work involved in testing an
application under both metacity and kwin just to know whether a bug is
WM specific.  Not sure how to help this besides more general
standardization, but it's hard to go from weird WM-specific bugs to a
requirement for the wm-spec list to discuss.

  For the record, kwin is great.  Lubos, you've done an amazing job (oh,
and Matthias Ettrich rocked too of course).

  -Billy