Re: Future of GNOME

[Tim Moore <tmoore tembel org>]

Dietmar Maurer wrote:
> 
> I'm a little bit disappointed about gnome. Although many people
> implement programs, it seems that we don't have any new ideas. Most of the 
> initial proposals are ignored, and I don't see what makes gnome better than
> other approaches (Maybe I'm wrong, but if the only advantage of gnome is that
> GTK is under the GPL, it's easier to rewrite QT (and provide C language
> bindings for QT)).

At this point the major advantage of GNOME is that it's under the GPL.
For many people, this usually implies other advantages as well.
Personally, I feel that GNOME is currently at a state where it's playing
"catch up" with other, older desktop environments. However it's also
being developed much more rapidly than any other. Plus some interesting
architecture is being built in right now, CORBA being the biggest.

Also, it's probably not easiest to just rewrite QT for a few reasons:

1) QT is a moving target and nobody outside of Troll can have any
creative role in its development

2) The GTK widget set is pretty stable and we can now work on adding
widgets, themes, etc. Harmony, AFAIK, is still pretty far off.

3) There are other issues in the KDE design which Miguel et al disagreed
with (like the standard WM).

> I think it's time to rethink the gnome design and state more exactly
> what gnome should be. The low-level part (gtk, libgnome..) is OK, but many
> other things are unclear (UI-Guidelines, software component model, ...).
>

I used to think this, too. Then I realized that nothing gets done in a
free software project run that way (e.g. Freedows). It really is best at
this stage in the game to let everybody just hammer stuff out. Then the
good stuff stays around, and the less good stuff (e.g., the *-properties
apps) go away or get improved.
 
> Just some example/thoughts:
> 
> gnome-mdi:
> 
> I've written two programs where it's possible to use
> the gnome-mdi interface. But it seems that there is no common acceptance
> on gnome-mdi, although it's in the core libraries.

Well, I personally dislike it. I agree that it needs to be addressed,
but consider it a low priority.

> baboon:
> 
> What's the state of baboon? Will gnome use a software component model?
> (gwp doesn't use corba/baboon).

This is probably still too experimental to be using yet. :-( But
considering that it's no more than a couple of months old, I think it's
doing well.

> panel:
> 
> OK, the panel works, but it provides the same functionality as one
> module of my window manager, not much more - features like a drawer are useless
> for most people. And the menus are at least uncommon. I can't see any
> new idea.

Well the nice thing about the panel isn't that it's revolutionary, but
that it's standard. Used to be each window manager had its own panel and
its own set of basically equivalent apps which you could dock in the
panel. With GNOME and CORBA, there's a standardized interface to the
panel. This means two things. One, most people will be using the same
panel program. Two, even if alternate panels are developed, they can
still use the same panel CORBA interface. So the result is that applets
only have to be written once. Furthermore, applications can register
their own applets with the panel, so for example GTCD and the cd_player
applet could be merged into one executable. KDE lets you do something
like this, but I don't know much about it. I don't think you can easily
switch panels.

> scripting language:
> 
> Scheme is a nice language, but most people don't like it (see the amount
> of gnome programs written in scheme until now). Tcl/TK would be a much
> better choice because there are so many Tcl/TK programs - and it's very
> easy to learn Tcl/Tk.

GUILE was designed such that scripts in other languages can be
automatically translated into scheme. One of these languages is Tcl. So
instead of using Tcl/Tk, you should use Tcl/GTK+ (if it exists...if not,
write it!) and the Tcl->SCM translator.

More info at:
http://www.gnu.org/software/guile/guile.html

> control-center:
> 
> The control-center uses a new user interface. This is bad because it's
> uncommon. If we want such thing we should make it more general, so that we can 
> use it with other applications too.

Again I dislike the control-center interface, but this is just a growing
pain.

Tim