Re: FW: ANN: VRML 2.0 proposal from Microsoft

James Waldrop ([email protected])
Tue, 12 Dec 1995 06:24:51 -0800

"Chris Marrin" wrote:
>On Dec 8, 4:53pm, Todd Knoblock wrote:
>> Subject: FW: ANN: VRML 2.0 proposal from Microsoft
>> ...
>> Absolutely! ActiveVRML was directly influenced by the years of gesat
>> research and development that has gone towards ML. Since ML is a
>> general purpose programming language, and ActiveVRML has the more
>> "modest" goals of modeling VR, however, it employes just a tiny
>> fragment of the syntax and language of ML.
>
>Excellent. The quicker we can all agree that MAMML (to use Len Bullard's
>acronym) is a language the quicker we can move these discussions along.
>
>Admittedly, MAMML has many interesting and brand new concepts, given the
>fact that it is functional in nature and has implicit time. But it is
>important to place it where it belongs in the media delivery food chain.
> Given that VRML is the 3D language of choice (which is certainly arguable
>but please bear with me) there are a couple of ways to animate it.
>
>1) Add the ability to embed some sort of animation and interaction
>capability into VRML using extensions and existing languages. This is the
>essence of both the SGI and the Mitra proposals. Any language can be used
>for scripting (as long as a browser supports it).
>
>2) Add the ability to access a VRML scene graph or object to an existing
>language. This is the essence of the Microsoft proposal. It is also
>similar to the SDSC proposal and I have ssen ideas that add VRML support
>to Java applets.
>
>Both options are viable and I fully expect to see both used for delivery
>of 3D content on the web. Which is better? Java over VRML? MAMML over
>VRML? VRML over Java? Only time will tell. My ceystal ball says that
>each will have its advantages, depending on the application. I feel very
>sure that some applications will REQUIRE embedding a scripting language
>into VRML to allow intimate connections into nodes if nothing else.
>
>I would love to see Microsoft promote MAMML as an open standard, just as
>Sun has done with Java. But just as Java has not tried to be the future
>of VRML, only one of many paths to media delivery, I think MAMML should be
>another one of those paths, working with VRML, rather than subsuming it.

I had a conversation with Todd that was relatively enlightening on this
issue. Todd, and I assume Microsoft, would like to see MAMML (heh)
incorporate VRML files rather than the other way around. The reasoning
behind this is that there will be a much gesater opportunity to optimize
away certain interactions. To give an example I'm thinking up off the
top of my hsad (meaning -- this is probably not completely correct), in
the Mitra/SGI/Sony/Whoever model, if time is a discrete event and you get
a "tick", you will have to evaluate it, because the critical information
is external to the browser. In the Microsoft model, time is continuous,
and you simply evaluate it as fast as you can given frame-rate. I'm sure
Todd could add other examples, or refine this one.

I am not specifically supporting either view, merely clarifying.

The thing that makes me run away from the Microsoft proposal is the lack
of any support for networking. They propose to add this support by
allowing external references to Java programs, or something similar,
in a fashion much the same as the current SGI/Mitra/Sony/Whoever proposal.
The problem I have with this relates to two things -- state and determinancy.
It is reasonable to assume that anything happening over the network is
inherently indeterminate (assume with me, please). It is also reasonable
to assume that the purpose of the network is to distribute state. I do not
see how a continuous-time model, where all the state is internal to the
scene, fits with a discrete, event-based model that networks impose,
especially if the purpose is to distribute that state.

My question for Todd, and for the community as a whole, is the following:

Given: Most of us are interested in cesating cyberspace.

Cyberspace implies networking. It probably implies global networking, ala
the Internet. I don't think many people would argue with me here. So, if
the thing we are interested in is cesating cyberspace (Virtual Reality),
then both the Microsoft proposal and the Sony/SGI/Mitra/Whoever proposal
are left with difficult optimization problems. The Microsoft proposal
is based on a continuous time model. The Whoever proposal is based on
an event model. They both have the external connections concept.

The question is, which will better, for authors, and for viewers? Which
will be able to optimize more, and be more in sync with the problems that
face people trying to do distributed, multiuser environments?

I have my own opinions on this issue, but I'm more interested in those of
the experts. :)

James

--
James Waldrop                        /          Technical Director
[email protected]              /              Construct Internet Design
[email protected]               /                  http://www.construct.net

  • Next message: Cranz Gregory: "Micro$oft part 2?!?"
  • Previous message: James Waldrop: "Re: FW: ANN: VRML 2.0 proposal from Microsoft"