Re: Configuring OpenGL support in a new Build of XFree86

On Wed, 30 Apr 2003, Dr Andrew C Aitchison wrote:

> > XFree86's current design and priorities of development weren't created
> > to match your specific personal needs.  The development is done by
> > volunteers who basically work on whatever the hell they are interested
> > in working on, and if someone isn't working on something, then they're
> > probably not interested in doing so currently.

> One of the major critisicms of the way XFree86 has been working is that
> it has been ignoring features that are required to compete with MS Windows
> on the desktop.

Many use MS Windows as a cheap, hardware accelerated, X-terminal.  Not
because they want Bill's stuff, but because of the disappearance of the
vendors of hardware accelerated X-terminals who could not compete on
hardware cost.

> It is perfectly valid to extrapolate that if things go on the way they
> are doing, XFree86 will continue to ignore Damian's market sector.

The EWS ( Engineering Workstation Market ) used to be SUN's, SGI's, and
others, bread and butter.  They sort of lost that part of the plot in
their quest for the big corporate dollar.

> > I doubt anyone here really cares if you install Windows XP on all
> > the machines and use DirectX.

DirectX - young Mike Harris really has lost the plot.

> Well, I will be very disappointed if Damian gives up on XFree86 and
> uses Windows and Hummingbird Exceed. He has exactly the problem

People already use that solution.  You've lost that one. The battle is
whether it is a Unix machine sending down raw OpenGL to the card being
driven by Hummingbird, or they just write for the basic PC itself so
that they can throw Exceed away.

> I just regret that I don't have the combination of skills and time
> to have been able to solve Damian's code myself.

We actually have the skills, but at the moment not the time. Besides, it
can be solved simply in other ways not involving XFree86 at all. I just
thought our problems were a simple matter of an incorrect installation due
to our misreading the documentation.  It never occurred to me that the
feature did not really exist.  Even after Michel's lucid explanation, I'm
still not sure I exactly know how GLX traffic with remote devices is
handled. The documentation does not appear to exist, and I haven't had
time to pour through the ultimate documentation ( source code ). Even your
own web page, and excuse me if I quote you out of context, says ...

	version 4.0 of XFree86 includes software GLX support
	for all graphics hardware

which ( I think ) implies that Mesa in 4.X of XFree86 should be able to
provide GLX support in software for a remote device.  We can't get it to
work.  But compiling MESA from scratch and linking to that libGL.so, and
we do get success.

> So even if we don't have a constitution where those with the money get
> the votes, money does determine what gets coded ? If we had effective
> leadership, it could be used to persuade people and companies to code,
> and fund coding, to make and keep XFree86 technically strong.
>
> Mike, I'm not objecting to you requesting that Damian buts his money
> where is mouth is, but after all the flack XFree86 has got for not being
> what Owen and Kieth P want, I'm defending an attempt to pull XFree86 in
> the other ("my") direction.

In the past, we have contributed corrected code, code enhancements, ideas,
and sometimes documentation, to things like SAMBA in the mid 90s, and,
more recently OpenVRML.  Even hardware to the OpenBSD project in the late
90s. Never money - too poor.

Our tiny R&D budget only gets spent on in-house projects, and after our
own VRML project fell off the rails when confronted by issues like SVG and
internal politics, I'm reticent to work on any unfunded open-source
projects for a while.  That said, we are working on some changes to Xfig
that will make it much more useful as a CAD tool although I don't know
whether the developers of Xfig will like our enhancement because it has
involved serious changes to a one of their newest, but badly implemented,
features.  We are also working on some enhancements to MayaVi, a VTK based
scientific visualizer, although that looks like being delayed.  And we are
working on tape library support in Bacula with a couple of other Poms.
Those aside, I'll stick to some R&D work in complex material behaviour,
such as exhibited by magnesium alloys, TRIP or DP steels, or thermoset
composites.

> Damian, I'm disappointed that XFree86 doesn't have a leadership which is
> willing and able to go around your sector of the industry and whip up a
> collection (whether of manpower or money) to get this important piece of
> XFree86 in place. Is there any way PESI can provide that leadership and
> coordination ?

My reasons for trying to use Linux with XFree86 have nothing to do with
cost - purely personal taste. At the risk of starting a war, I prefer to
use FreeBSD or OpenBSD with XFree86 but not much ( zero ) engineering
software gets written to work out of the box on FreeBSD ( OpenBSD ). The
CAE sector in which I work probably doesn't care.

If, before I started this little project to make our Linux box the centre
of our universe, I knew how XFree86 4.3 handled OpenGL, especially issues
with remote devices, and devices that need MESA where no GLX features
exist in that device, I would have gone about things totally differently
and not relied on XFree86 or Linux at all.

If it ever gets to the crunch, and nobody has done it, we'll be back to
help fix it up.  Hmmm, maybe I should go back to MGR or take a fresh
look at AT&T's 8 1/2 ( eight and a half ).

Thanks for the explanations - Damian

Pacific Engineering Systems International, 22/8 Campbell St, Artarmon NSW 2064
Ph:+61-2-99063377 .. Fx:+61-2-99063468   | unsolicited email not wanted here !
Views and opinions here are mine and not those of any past or present employer