[RC5] Re: film rendering

Stefano Cazzulani pinguin at dada.it
Mon Jan 25 20:11:13 EST 1999


> One could also come up with a way to not have to transfer the entire 3D
> model again for each frame.  If it were just made up of objects, you'd
> download the whole object library first, then for each frame you'd just
> need layout data for those objects.  Or maybe a diff method where all you
> download is how the objects in this frame have changed since the previous.
> This would mean sequencial progression through the frames but that's fine.
> And if you fall behind and the next frame is 3 ahead, then you download the
> 3 diffs and now you're on the proper frame.

Unfortunately this can't be done... Rendering software use algorithms
like
ray tracing or radiosity to calculate every pixel of an image; first the
code
to do this is very heavy, and the client would be some megs of
executable.
In the second place, when you render a frame you ***need*** all the
infos
about that frame: this means that you need also all the textures that 
cover any object!! Think about a texture for a commercial productions:
megs
of image only for the surface of a screw in a complex scene, made up by
thousand of objects... This could be avoided only in very particular
situations
and not for every frame... Yes, the client could request the appropriate
pixel
of the texture to the server, when it needs , but in this case you need
a fixed
and fast connection: not a dialup thing!
I think the overhead to do such operations is too heavy and complicated;
you
can simplify an operation, but increase the complexity of all other
tasks...

Definitely , I think that this interesting (yes...it's interesting, no
doubt!)
job can't be done this way.

Regards

Stefano Cazzulani / Pinguin - Bug2Fix

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list