🚀 go-pugleaf

Peter Kocks writes
> As was pointed out earlier, OpenGL is a kind of fast procedural
> language for doing 3D graphics, whereas Renderman is object oriented
> and photorealistic.  

Umm, the RenderMan interface is no more object oriented than GL is.  They are  
both procedural, they are both immediate mode (although both have a little bow  
towards display lists - gl has "objects" and RenderMan 4.0 has macros).

The reason it will be a lot easier to write 3D code on the NeXT isn't due to  
RenderMan, it's due to the hard work the folks at NeXT have done on the 3D kit.   
RenderMan buys you a very extensible base, and the photorealistic renderer from  
Pixar.

> Given that in time Renderman will become fast
> enough for realtime use, why not write an OpenGL interface to the
> RenderMan rendering software?  Basically, write procedural type calls
> in the exact same format as OpenGL requires.   This would be a big
> boon for scientist who like OpenGL and the NeXT platform.

It would actually be the other way around.  I.e. you would feed your software  
gl calls, and those would turn into RenderMan calls on the NeXT.  Think of it  
akin to writing an X server on the NeXT.  You receive an X protocol stream over  
the net, translate those into the appropriate DPS calls, and you're done.

Keep in mind that interactive RenderMan was originally developed by Pixar on  
SGIs, although it isn't a product.  On the SGI, the reverse process happens,  
i.e. you get a RIB stream and turn it into the appropriate GL calls.  An  
unnamed friend of mine (who would know) once told me that the only difference  
between a GL stream and a RIB stream was "a sed script".  

--

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