Re: MOO and VRML 2.0 ?

• Date: Sun, 18 Feb 1996 10:55:09 PST
• From:
  Tom Meyer <twm@cs.brown.edu>
• Cc: moo-cows@parc.xerox.com
• In-Reply-To: <9602182126.aa28817@usgate.paragraph.com>
• References: <9602182126.aa28817@usgate.paragraph.com>

Dmitri Kondratiev writes:
> At 12:50 PM 2/17/96 PST, Tom Meyer wrote:
> >Well, I've been running a MOO that has been outputing VRML for about a
> >year now (Waxweb, actually that MOO was the first web server of any
> >kind to be serving up VRML), and have been deeply involved in the VRML
> >Architecture Group, which is just now helping to decide what will
> >happen with VRML 2.0.  So I'm fairly immersed in both areas, but have
> >not been that pleased by how suitable MOO is for VRML.
> >
> >I now generate VRML files on the fly from an http-speaking MOO server.
> 
> How you do that ? Looks like you need AI system to generate from text
> descriptions meaningful 3D scenes.

Well, my MOO-based system isn't really being used as a MOO; it's more
of a super-flexible http server at this point.  So the VRML is
generated from hand-built indexes of what 3D objects should be in that
scene. 

> >Before I added in VRML support, the server worked quite well.  Since
> >the VRML files are roughly ten times larger than the HTML files that
> >describe the room, that caused enormous slow-downs in the MOO.  MOOs
> >were never designed for large, sustained amounts of text throughput,
> >so don't handle this extremely well.
> 
> Other people store just URLs poiting to VRML files (maybe on different server)
> in MOO core. In this case, when MOO client recieves URLs attached to rooms,
> volume of data is aproximately the same as in traditional MOO.
> Why don't go that way, and use MOO server/base as a grid partitioning VR space ?

I send out scenes consisting of multiple VRML objects stored on
different servers.  So it's one scene consisting of transformations,
texture maps, and inlined URLs to the actual geometry.  This is still
a big enough file that it can start to overwhelm the MOO fairly
quickly.

Delivering a 200k indexed face set from inside a MOO would definitely
be suicidal; I've made sure not to do that, but to get the flexibility
of having libraries of useful objects combined on the fly (e.g.,
chairs that could be in several positions in the room), you need to
deliver some of the VRML from inside the MOO.

Tom

• Re: MOO and VRML 2.0 ?
• From: Dmitri Kondratiev <dima@paragraph.com>

• Prev: Re: MOO and VRML 2.0 ?
• Next: Re: MOO and VRML 2.0 ?
• Index: MOO-cows
• Thread: MOO-cows