Re: Sv: Window Manager configuration utilities

[Havoc Pennington <rhp zirx pair com>]

On Fri, 9 Jul 1999, Matthias Ettrich wrote:
> 
> While I totally agree, I can't resist mentioning that exactly this was one of
> the major points why the early gnome developers flamed KDE and started Gnome.
> Until recently, "complete WM independence" (which reads: no integration at all),
> was one of the major selling points. I'm happy to see Gnome finally growing up
> in this issue, though I can imagine that you have a hard stand trying to
> convince your fellow developers that the KDE approach was actually right :-)    
> 

I think many of us realized long ago that "window manager independence"
was crap if the widnow managers are unmodified. I think the ideal for
"window manager independence" is a set of common interfaces (i.e. this
window manager spec) that allows things to work with any WM.

As things stand now, it's clear that the "dedicated window manager"
approach has been better for getting short-term niceness, but it's still
not quite ideal (as you'll probably agree, since you're on this list).
 
Personally I think some of the more configurable window managers (E and
scwm) and those that work hard to play nice with Gnome (icewm) will end up
being pretty good with Gnome. Scwm in particular can be made to use Gtk
and thus match the theme, etc. E isn't trying hard to work with Gnome but
an appropriate set of themes/config files will make it OK. But none of
these can be nice with Gnome unless we get this spec finished. For that
matter, any of these should also work fine with KDE if we finish this
spec.

FWIW I think the "window manager as desktop environment" approach is
equally stupid; if you don't write apps to work with and match the window
manager, and don't have infrastructure like component technology, widget
set, config files, MIME types, etc. then you are going to end up with
exactly the same problem Gnome has had, only with the opposite end of the
equation missing. WindowMaker is in this position by default (because
GNUStep is moving so slowly) and E certainly seems to be going that way.
However, window managers are becoming desktop environments only because
desktop environments have been lacking, or because they started before the
DEs existed, or because there's no spec for interacting with desktop
environments (there is for Gnome or KDE, but that's not the same - it's
not a standard).

> I still see a huge "market" for standalone windowmanagers with desktop
> functionality, though.
> 

Yes, that's there. I think it's because standalone WMs can still offer a
more integrated environment in many respects. (And also because the
dedicated window managers, like kwm, don't have the same featureset that
the standalone ones do.) However, I think this is a fading software
category, because it's missing half the desktop. The future is in
*interoperable* components, not standalone monoliths. That's the only way
we can take advantage of free software's parallel development model. And
it's better engineering anyway.

I would like to see pluggable desktop shell components, so we still have
choice and room for innovation. For example choice between Gnome/K task
list, icewm-style, WindowMaker-style. Also choice between window managers.
Users should be able to choose regardless of their desktop environment. WM
projects would instead become "desktop shell" projects; they might ship an
integrated set of desktop shell elements, but users would not have to use
the exact shipping set. Alternatively, some WMs might decide to be window
managers only, in the strict sense.

The spec should require enough features to make a nice, integrated
environment while retaining code modularity. That includes setting the
background, either pluggable pagers or a way to embed pagers in the
k/gnome-panels, some standard way to configure common window manager
features, etc. 

The user shouldn't really see the word "window manager" - Windows has a WM
I'm told, but you sure wouldn't know it. Some window manager authors
probably won't implement an extensive spec, but I think it's more
important to get a usable, integrated setup than it is to get all WM
authors on board. As long as we get a few. Some WMs will no doubt continue
along the path of competing with the desktop environments, and that's
probably a good thing to some extent.

However, if no WM authors want their WM to be a desktop component (or set
of components), and all of them want to be a monolithic standalone
environment, then we are going to have to have gnome-wm and kwm and
require their use. There's really no alternative IMHO if we want to get an
acceptable result.

Anyway, not speaking for Gnome; don't post this as an example of some
Gnome buttheadedness or something. I'm sure lots of Gnome people disagree.

My point is simply that Gnome, KDE, and the various window managers are
all lacking some piece of the big picture ideal solution; and the only way
to get the big picture ideal solution is to work together and make this
spec meaty and comprehensive. So, I hope that will happen. Let's work
together to make something *really nice* instead of continuing the free
software tradition of uncoordinated islands of niftiness. After all
everyone here has basically the same goal of making free Unix easier or
nicer or cooler to use.

Havoc