I haven't gone into much detail here, but I wanted to get people thinking
about the two problems seperately. In particular, as Don Brutzman pointed
out at the Symposium in Monterey, this kind of separation allows the
semantic models to be driven my different mechanisms (including (with the
right protocol) MUD/MUSH/MOO's).
This will also make the standard more easily adopted by people who want to
be able to share the geometry and semantics, but use it in different ways.
Again this can be the difference between a VRML browser viewing a "static"
dynamic scene which is moving in a pre-determined fashion, and some kind of
interactive viewer where some objects are being driver by a sophisticated
server process.
(Ducking the flames...)
Chris Hall
NeTpower Inc