PHIL: Miniumum VRML-capable machine

Tom Payne (jlfs@vax.ox.ac.uk)
Wed, 15 Jun 1994 19:11:11 +0100


Surely the creator of the WWW page/room defines the complexity of the scene?
One may choose to use hundreds of texture-mapped polygons to create a
photorealistic landscape, whereas another will use a few cuboids to create a
room? I believe that the system should _support_ hundreds of texture-mapped
polygons, but that facility would not be used by simpler (i.e. less capapable)
machines. I propose that each scene should be described at a number of
different detail levels, for example the Utah teapot:

Lowest: (1) A sphere at a specified location, pure color, and radius.
(2) A polygonal description of the teapot itself.
Highest: (3) (2) and texture maps for each polygon.

The user could then decide what detail to include based on:

(1) The actual rendering capabilities of their browser.
(2) Their bandwidth.
(3) Whether they prefer a fast-moving wireframe environment (at say
25-30 fps) or a slower, more detailed one (e.g. at 10 fps).

The browser can then request only the information it requires from the server,
thus optimising the use of bandwidth. Future hardware will exploit its
capabilities to the full, but the system will still work with existing systems.

Local descriptions of common objects could also be optimised to the browser,
for example my workstation teapot:utah will use 8192 polygons, whereas my 386
PC teapot:utah will be 24 polygons.

Tom Payne