On Mon, 13 Jun 1994, Jerry Isdale wrote:
> re: object sharing
> Transfering entire geometries across the net whenever you enter a new
> space will absolutely kill your bandwidth. Especially if your VRML (or VR
> Interchange Format - an alternative evolving standard) gives vertex coords
> in ascii text.
> Consider instead that you have at each client machine a library of object
> geometries (and possibly some behaviors). The VRML would then need only
> specify that a "teapot:utah" is needed instead of sending the 500 polys
> needed. The client would look up the symbolic name in its database and pull
> the appropriate level of detail and format required by its image
> generator.
> Object libraries could be distributed on CDROM in some common format,
> with symbolic indexing database. If the object is new to the local client,
> it could request a copy from the server (perhaps not the same one the VRML
> document resides upon, but a different 'geometry server')
This is an *excellent* idea. I was just thinking this before I read your
post. We need to bring in some stuff from the URI working group, I guess.
This is trickier than it might seem - sure, someone can define "teapot" as
a set of 50 polygons, but what makes that the official set of polygons for
the term "teapot"? How is that distributed and officiated? Do we set up
a system similar to InterNIC name service (moderately scalable) where
shapes are "registered"? Or do we instead rely upon HTTP caches and
specify shapes as "http://some.big.machine/objects/teapot.vrml"? This
last possibility seems strongest to me. Anyone here on the URI list?
Brian