Re: TECH Survey comments
Linas Vepstas (firstname.lastname@example.org)
Tue, 2 Aug 1994 19:53:23 -0500
> Date: Mon, 1 Aug 94 11:25:16 EDT
> From: email@example.com (Mark Waks)
> Subject: Re: TECH Survey comments
> John argues in favor of object caching, but only of the exact object
> possessed by the scene designer. Essentially, each object would need
> to be fetched at least once, although it could be locally cached after
> The question is, how do we support both of these models? The answer,
> I believe, is keywords.
> When we specify an object to be fetched in VRML, we may specify one or
> two related lists of keywords. One would be a list of keywords that an
> object *must* satisfy if it is to be considered acceptable for this
> use: "We need a six-foot, left-handed fridge". Another would be a list
> of optional keywords, in priority order; the user's Browser would be
> responsible for deciding whether to fetch the exact object from the
> scene's machine, or to use a close local approximation.
To me, this proposal seems overly complex. I'd rather see a solution
that benefits all URL's, rather than just 3D URL's. That is, have some
scheme where somehow local servers could serve up popular URL's by serving
up a local copy, instead of going over the net. Don't know if anyone
is working on this.
Anyway, if this were implemented, then
(1) create a collection of chairs, tables, etc. that would be
a potentially popular set of URL's
(2) If you, the scene designer, don't really care, then reference
one of these "pop" URL's to get your chair. With a little luck,
it'll be in some local cache.
To me, this seems like the simplest idea.