If a VRML viewer reads in a scene with no lights, it can either
add a light by default (for instance a headlight that is always
shining in the direction the camera is pointing, a very handy
default behavior), or it can assume the VRML author meant to
have no lights because the scene contains pre-lit polygons (for
instance the lighting was precalculated using radiosity).
Problems with these assumptions:
Assuming it can add a headlight means the viewer will unneccessarily
add lighting in the case that polygons are pre lit, eliminating the
performance increases that could be had with precalculated lighting.
Assuming the scene is pre-lit means the viewer headlight behavior
can't be used, and authors would be forced to add enough lights to
the scene so that objects in the scene look good no matter what
the camera position and orientation.
A light model node would provide a mechanism for authors to explicitly
tell the viewer whether the scene is lit or not.
My proposal:
FILE FORMAT/DEFAULTS:
MODEL
BASE_COLOR Use only the base (diffuse) object color
PHONG Use Phong lighting model
LightModel {
model PHONG
}
David
--David Mott, Silicon Graphics (mott@sgi.com) <Inventor WWW info - http://www.sgi.com/tech/Inventor.html>