On Mon, Nov 27, 2000 at 08:15:55PM +0100, Antonio Campos wrote:
> I don't agree with you here. It can be the political future, but I can see a
> lot of bad things in X that don't make it the proper election for the future.
I'm not talking about politics, I want GGI to succeed in Linux as much as anyone
of you. It's just common sense that right now, graphics on Linux means X.
Therefore: improve on X.
First: several initiatives exist to solve the major issues in X.
(slow rendering, no antialiasing, fonts bad etc.)
Read the following two links:
A New Rendering Model for X
Keith Packard
XFree86 Core Team, SuSE Inc.
http://www.xfree86.org/~keithp/talks/render.html
and
Translucent Windows in X
Keith Packard
XFree86 Core Team, SuSE Inc.
http://www.xfree86.org/~keithp/talks/KeithPackardAls2000/
and then there is the new console system for Linux by James Simmons.
Second: There is no desktop (non-server like apache etc.) software for
Linux except for X Window. Linux would never ever have come where it is
today if it couldn't already run existing UNIX and X Window software.
It is really cool that things like Berlin are being developed, but there
is no software for it. No office tools, no scientific software, no games,
nothing but a couple of proof of concept programs.
If you read those two links, you can see that at least some XFree people
also understand the limitations of X, like you do, and want to fix it.
I think that GGI will only get big if it can provide a much faster X server
(say 400% speedup or so) than XFree with their current model can.
(Ok, maybe GGI could get big in other devices like webpads, handhelds etc.
but I'm talking about desktop Linux now)
That new Xserver should include smooth and very fast 3D OpenGL graphics,
based on GGI, or there wouldn't really be a point.
That is one opportunity for the GGI project: providing a better way to do
OpenGL graphics than DRI/DRM currently does. (I use that myself and I hate it.)
For non-3D graphics:
X may be slow now, but even unaccelerated X is fast enough for most desktop
things like writing a letter in wordperfect, using Netscape or using an Xterm.
However: those things would very much benefit from antialiased graphics (for
which you need alpha channel) though. This is what Keith Packards Render
extension is for.
Implementing that with GGI would be a second opportunity for the project.
For example: BeOS is much faster than Windows 98 for 3D graphics (I heard
3 times faster), and even Windows 98 is faster than Linux + DRI. That means
that BeOS is doing something very right that Linux and Windows are doing
wrong.
X Window does not have to be slow, I have seen Irix and Sun stuff that even
on remote windows would completely blow Linux away. And it certainly does
not need to be ugly. (Gnome, Windowmaker and even KDE look pretty good)
I think that the right strategy for GGI would be to prove its merit by
providing a much better X Window system, implementing that Render extension
and 3D OpenGL pipeline system, and then come with fullscreen GGI applications
(good idea) and Berlin (also good, but needs a applications base).
And combining KGI and X in an add-on package for mainstream Linux distributions
so that it has its own runlevel seems like a logical idea to me.
Also: I'm not a programmer but a sysadmin, but making such a package and
runlevel system is mostly scripting, and I do think I have the skills to do
most of that.
