I agree I think. . .
I currently work on the CosmoWorlds VRML 2.0 authoring tool, and I agree that
VRML is the output of such authoring tools. CosmoWorlds has a little widget
that lets you create a route from an eventOut to and eventIn. I assert that
other VRML 2.0 authoring tools will have similar functionality eventually.
Changing the syntax in the file format to 'look' more like a method call would
be irrelevant in this authoring process. Changing the VRML process model to
'act' more like method calls would be imposing an interative paradigm on VRML,
and I do not yet see why that is justified.
> In terms of how this work, the wrl file is first read into an object
> structure that follows the language specification in a straightforward
> way. This object structure is then transformed into a problem domain
> specific structure. Its this transformation that is affected by the
> file format, so that's why I think we are already discussing this in the
> context of authoring tools.
I agree completely with the above statement.
Perhaps I am stating the obvious in a disagreeable manner ;).
I would like for us to determine how a designer using some tool can create VRML
content in an object-oriented fashion. I think that by doing this, we'll find
giant weaknesses. In my experiments, for example, I found that having
different JAVA classes for the 'Nodes' returned in a JAVA script node and the
External Authoring Interface to be really sad. At the same time, I think it
will minimize issues like "how to specify a route."
VRML was in some ways too academic an exercise. The specification changed so
quickly that nobody tried producing content in the early days. I think that if
it had been, issues like the EAI v. Java Script classes would have been flushed
out. I am worried that the work we do here will be the same if we pursue
"polymorphism in the file format" as a goal instead of really figuring out just
when object-oriented development can be useful for the author. To design an
object oriented VRML that is useful for designers, etc, we should be thinking
about the objects that they deal with first, and then see what VRML drops out
of it. . .
Thanks,
Curt
-- The question of whether a computer can think is no more interesting than whether a submarine can swim. - Edsgar Dijkstra