Return-Path: <wave@media.mit.edu>
Received: from aerospace.aero.org by antares.aero.org (4.1/AMS-1.0)
	id AA00174 for  /u/strauss/bin/mail_handler.pl strauss; Sat, 27 Feb 93 20:07:14 PST
Received: from media.mit.edu (media-lab.media.mit.edu) by aerospace.aero.org with SMTP (5.65c/6.0.GT)
	id AA16706 for strauss@antares.aero.org; Sat, 27 Feb 1993 20:07:11 -0800
Posted-Date: Sat, 27 Feb 93 23:07:10 -0500
Received: by media.mit.edu (5.57/DA1.0.4.amt)
	id AA27592; Sat, 27 Feb 93 23:07:10 -0500
Date: Sat, 27 Feb 93 23:07:10 -0500
From: Michael B. Johnson <wave@media.mit.edu>
Message-Id: <9302280407.AA27592@media.mit.edu>
To: rman@aero.org
Subject: qrman on other platforms?
Reply-To: wave@media-lab.media.mit.edu


Hopefully someone on this list will have some good comments for me...

Hi folks.  For some time now, I've been building a system (my PhD) for
experimenting with autonomous animation.  The system runs distributed
over a wide variety of systems (pretty much anything that runs UNIX).
Two parts of the system are inherently limited in their portability,
though: the graphical front end (where the user assembles the the
parts of the animation) and the back end (where the pictures get
made).  I use a NeXT as my front end and am very happy with it.  I've
integrated my distributed stuff into NeXT's Interface Builder, and
everything is hunky dory.  The difficulty is the backend.  Currently,
I've hacked up a version of some software which was written here at
the Media Lab as my rendering back end.  For the most part, it does
exactly what I want in terms of portability (runs in hardware mode on
HPs and SGIs, has a high quality software A-buffer for everywhere
else).  Since it was written here (by Dave Chen), I have all the
source code, which makes me happy.  The problem I have is that I'd
really like to be using RenderMan as my final back end.  I want to use
RenderMan for a lot of different reasons, most of which people here
probably know too well.  The three main reasons are (1) nonpolygonal
surfaces (2) programmable shaders and (3) it has a file format, i.e. a
client/server model.  The first two reasons are why my current
solution is too constraining - our system here is also a client/server
model so I have that covered. 

I'd like to use some off-the-shelf that at least ran on an SGI,
although running on an HP (hardware) and a NeXT (software, mainly
interesting for debugging) would be great.  I've looked at SGI's
Performer, but putting curved surfaces in there looks to be a pain,
and Inventor looks to be young (and not tuned for multi-processor
performance - I have a 4 processor SkyWriter).  Both of these systems
still only give me hardware quality (although it's getting higher),
which means no programmable shaders.  I could stick with my current
system, and take the time to put gl-specific curved surfaces, but
truthfully, modeling has never been my strong point, and I don't know
that I want to spend the time to come up to speed.  Anyway...

Clearly, RenderMan (qrman+prman) would be the perfect solution.  I'd
be happy to put up with (relatively) low quality fast rendering on an
SGI if I could turn around and redirect the byte stream to prman when
I needed to make a production tape.  Unfortunately, qrman (correct me
if I'm wrong, please!) only runs under NeXTSTEP.  Hence my question:
can we reasonably expect qrman to show up on other platforms (HP and
SGI, specifically) running under their native OS's?  This would
certainly make me toss any thoughts of a gl-specific solution, and
happily embrace RenderMan as the be-all and end-all for my 3D rendering
needs.

Any comments for the list from the people in the know?


-->  Michael B. Johnson
-->  MIT Media Lab      --  Computer Graphics & Animation Group
-->  (617) 253-0663     --  wave@media-lab.media.mit.edu

