Re: Removing wm/pager interaction

["Michael Rogers" <mrogers cs ucl ac uk>]

> I don't think we want to go that route; the pager and tasklist belong
> in the panel, there isn't anywhere else sensible to put them (why
> should they be floating around taking up desktop space?).

OK, so we need a way for panels to swallow wm-provided pagers/tasklists. The 
Gnome panel can swallow any app, I'm pretty sure the KDE panel can too, so we 
just need a way to tie this in with the window-manager-chooser so that the 
pager/tasklist gets swallowed automatically when the WM starts.

> From our point of view, we need a pager in the panel. There are two
> ways to get that: WM spec, or require a specific window manager.

And WM spec is the right route. But that doesn't have to mean describing a 
protocol for standalone pagers. It could also mean describing a way for 
wm-provided pagers to be embedded in the panel. Then as you said above 
(oops, snipped) you give new users a sensible default (a wm which provides 
an embedded pager) and let them switch to a less-integrated wm later if they 
want to.

> We could also be window manager independent by not providing any
> functionality, and letting the window manager do all the desktop
> tasks, but in this case we'd have to anoint the official window
> manager that we maintained, and users who wanted a different window
> manager would essentially have to drop all the GNOME
> features. i.e. we'd just be moving GNOME into the window manager, and
> "window manager independent" wouldn't mean anything at that point.

I don't think the wm should take on desktop tasks. But the job the task list 
and the pager do is window management.
 
> The root of the issue is that there are lots of chunks of functionality, and 
> there's lots of variation in what a WM provides.
> If your window manager has a dock/panel thing, then it probably doesn't
> make any sense to run the GNOME one also. There's also file managers,
> pagers, etc., which may be in GNOME and may be in the window manager.

Things like file managers are obviously desktop/application functionality. WM 
hackers who provide them know they are competing with desktop environments. In 
the same way tasklists and pagers are window manager functionality, and 
desktop environments that provide them are competing with window managers.

> There's no particular a priori reason why each functionality chunk
> should be in one place or another. 

Agreed - we just want to keep the interfaces small so that chunks can be 
replaced easily. The pager protocol is a big interface. This indicates we've 
drawn the line in the wrong place. The fact that the window manager has to 
export all of its core functions to another app indicates that.

> If we had had a WM that presented
> the same integrated look and feel you get from the GNOME stuff, we
> would almost certainly have used it. However so far no WM has even
> wanted to link libgtk, so it's impossible to get nice integration with
> GTK apps, write applets with the GNOME libs, etc. This means as a
> practical matter that some of the functionality has to be outside the
> WM.

<vapourware>
I'm going to write a GTK+ window manager when this spec is complete. Since my 
number 1 goal is Gnome integration, the contents of the spec will to a large 
extent determine the design of the WM. Well, that's my excuse for not writing 
anything yet.  :0)
</vapourware>

But you don't need to link with gnome-libs to get swallowed by the panel. And 
it makes as much sense for the pager theme to match the window decorations as 
it does for the pager theme to match the rest of the panel. (Ideally all three 
would match, of course.)

> But, none of the current WM's have a native pager/tasklist that's
> suitable for the GNOME default, at least I haven't seen it.

OK, let's add a pager to sawfish and call it the default Gnome window manager. 
I'd rather spend my time learning Lisp than spend it learning that a camel is 
a racehorse designed by a mailing list.  ;)

> So we need to provide a panel, pager, task list, file manager, control 
> panel, etc. that have integrated look-and-feel. Given that, the only way to
> be window manager independent is this spec.

I agree. But using the spec doesn't have to mean writing a protocol for a 
universal pager. IMNSHO swallowing a wm-provided pager or tasklist is the way 
to go - smaller interface, less restrictions on the wm, better integration 
with the wm.


Michael