Re: VRML Survey now online

Brian Behlendorf (brian@wired.com)
Wed, 20 Jul 1994 13:49:50 -0700 (PDT)


On Wed, 20 Jul 1994, Mark Waks wrote:
> Paul writes:
> >> Matrix Representations
> >
> >The issue was not the format, but the generality of the supported
> >transformations (all projective transformations vs. just rigid
> >motions).
>
> Just a comment here -- the format is *an* issue. Personally, I want
> to see simpler tranformation commands also available (eg, "shift",
> "scale", etc), so that someone can write basic transforms without
> knowing the relevant math. That wasn't covered in the survey; I
> plan on bringing the point up again later...
>
> However, I suspect that having the general matrix representation is
> *also* quite useful (and arguably should come first)...

Ack - my brain wiring got screwed up, and instead of asking for
matrix representations with simple commands the given, I should
have asked if simple commands were something we should support
with matrix commands the given. I personally think both should
be fully supported and interchangeable.

> >> Two-tier system: object description files and scene layout files?
> >
> >Since we want to be able to inline and embed objects from around the
> >Web into a single scene, we obviously don't want to require every
> >object file to be a complete scene (with lights, camera position,
> >etc). I'm probably misunderstanding the question.
>
> I dunno -- that's how I read it also. I regard it as absolutely
> *critical* that we have the ability to define files that just contain
> objects; without that, I don't see how we could ever hope to
> distribute the object base. I was a little surprised that this was
> even a question; I had figured it obvious that we needed to be able
> to separate objects from scenes. (Fortunately, most (all?) of the
> proposed systems allow that fairly easily...)

Yeah, this is what I meant. The topic was brought up briefly but
not resolved, so I threw it in - seems like it gets resolved in all
the proposals anyways.

> >> Engines
> >
> >What exactly is an "engine"? Are you asking if the language itself
> >should be required to predefine interactive behaviors for specific
> >objects like doors? Or are you asking whether the language should be
> >capable of supporting references to common scripts?
>
> I think the concept here is taken from Inventor. Inventor systems are
> made up entirely of nodes and graphs; one type of node is an "engine",
> which allow you to manipulate the fields in related nodes in various
> simple ways. Basically they allow you to attach simple dynamic
> properties to the graph in a fairly elegant way.
>
> I assume that the question is, should we have something like this
> in VRML?

Again, exactly. It's between no interaction and a full scripting
language - a set of behaviors is defined, and the scene can be described
as a function of those engines. I.e., a bird can fly in a circle defined
as (X=sin(time), Y=cos(time), Z=10) where "time" is the engine. Those
engines could be tied to events, too. Someone with experience with
Inventor might want to give a better explanation...

Brian