My suggestion is that there would be a set of object repositories with
some standard description info available as network object servers. The
descriptions might be searchable ala WAIS and there would be options for
various levels of detail and perhaps file formats. Additionally, there
might be a computation server that would take a parametric object
description (splines, NURBS, etc) and generate a specific level of
detail/poly count on request.
At semi regular intervals, say every six months or once a year, there is
a Collected Objects (or atoms) cdrom created from the libraries and made
available to various cdrom distributors (Walnut Creek, etc)
The VRML client would look first on the local database for named objects,
then ask an object server or query the VRML document as to its source
atoms.
This would greatly reduce bandwidth requirements.
==============================
Jerry Isdale
System Engineer
Illusion, Inc.
2660 Townsgate Rd., Suite 530
Westlake Village, CA 91361
Phone: (805) 371-4530 FAX: (805) 371-4533
Internet: isdale@isx.com
Standards aren't