From: cyberoid@stein.u.washington.edu (Robert Jacobson)
Subject: Re: SCI: Imagery and/or Interactivity (Was Re: APPS: Mandala in TV clas
Date: 2 May 1993 18:47:50 GMT
Organization: WORLDESIGN, Seattle



In article <1s11niINNgh0@news.u.washington.edu>
jimh@surreal.asd.sgi.com (Jim Helman) writes:

>cyberoid@u.washington.edu (Bob Jacobson) writes:
>
>   Unfortunately, as long as platform manufacturers continue to stress
>   pretty pictures at the expense of both interactivity and an ability to
>   generate chaos and other lifelike phenomena, we are going to have this
>   interactivity-imagery dilemma with us for far too long.  
>
>I don't see this as a platform issue, except perhaps for old image
>generators which require compilation of the scene's geometric content
>into a static parititions, e.g. BSP tree, and the concomitant
>restrictions on modeling and dynamism.  On most current graphics
>workstations, geometric information is transferred from workstation
>CPU memory to the graphics subsystem every frame and rendering is Z
>buffered.  Hence, everything can be changed by the host CPU every
>frame.  On the highest end machines, you can even transfer video
>texture maps at 30Hz.  So an application can be as "chaotic" and
>"lifelike" as its smarts, the CPU speed and graphics power allow.

I think this undergirds my case, Jim; it certainly doesn't contradict
it.

Perhaps the problem is not that the graphics platform manufacturers
can't handle chaos and uncertainty, but rather that complete virtual
worlds systems, including databases and the platforms that manipulat
them, are not yet optimized to deal with chaos and uncertainty.  Or
perhaps it's the case that these platforms -- I suppose we are now
talking about supercomputers, although the line between the super's
and lesser machines is fading fast -- haven't yet been successfully
integrated into virtual worlds systems.  So the graphics platforms,
capable as they may be of rendering anything, are being fed a diet of
fairly simple, highly regularized data.

There is a meeting taking place this week at Sandia National Labora-
tories to see if there isn't a way to meld virtual worlds and
supercomputing.  (I suppose Sandia, like other federal laboratories,
is looking for new missions to replace using its supercomputing
facilities merely for designing nuclear warheads.)  I'm attending this
meeting and will report as fully as I can on its proceedings and
conclusions.

For now, I'm not convinced that we have all the tools we need to build
virtual worlds that are more than the equivalent of oil-painted
portraits (which, while quite good, are static) or very fine
photography.  I would be curious to hear from database designers and
builders how they are tackling these problems of incorporating chaos
in their models and what technology they believe is necessary to
resolve this issue.

Thank you, Jim, for speaking up on behalf of the graphics platforms
manufacturers, who indeed do a very nice job.

Bob Jacobson