My proposal on hotjava-interest was that the extra parameters be
more generally thought of as the "client-side specializer" of the
URL, and so appended to it using '#'. After all, other scripting
languages might wish to store initialization data in something
besides name/value pairs.
(Note: in SGML, because of the moronic field length bounds, this
scheme still requires some kludgery. Java's current <APP> tag is,
of course, totally illegal.)
So for VRML, my proposal would be something like this:
WWWAnchor {
base "http://www/dir/"
name "script.class#x=1;y=2;s=hello;t=0.23324"
... VRML objects ...
}
Here a Java script whose various classes are stored under
"http://www/dir/", and whose main class is "script", is initialized
with parameters x=1, y=2, s=hello, and t=0.23324. It is then given
control of the object hierarchy headed by this WWWAnchor object.
(Note: VRML should adopt HTML3's more sophisticated link
attributes, which include BASE, URN, etc., in addition to the usual
HREF/SRC.)
> However, unlike <app> tags in java documents, in this
> WWWAnchor case, the java applet should not be loaded and
> run immediately
> 4. Extend the vrml WWWInline node similarly, but it
> *would* behave like the <app> tag.
I agree with the principle...but perhaps not the conclusion.
As you're implying, WWWInline is in some sense redundant -- it's
just a WWWAnchor with immediate activation semantics. But now that
we are talking about scripting, why not rely on their much more
sophisticated event handling? (You can load a very minimal class to
begin with, that just sits and waits for the right event.)
Nevertheless, you definitely want LOD to have some effect on
loading of scripts!
--------------------------------------------------------------------
Paul Burchard <[email protected]>
``I'm still learning how to count backwards from infinity...''
--------------------------------------------------------------------