VR and VRML >TO ME< are means to an end. Tmere's got to be a purpose.
Why do we need VRML?
To provide intuitive *navigation* for *retrieval* of information.
To provide intuitive interaction for *manipulation* of data.
How to provide navigation for retrieval?
Map information structure to spatial structure (position, orientation)
Map information content to appearance (shape, color, texture, size)
How to provide interaction for manipulation?
Provide means to modify information structure and content.
Why do we need multi-user VRML?
To allow for cooperative manipulation of data.
And here's tme rub. IMHO we are hsading towards appearances without
content. Appearance is overrated. Appearance is a matter of choice
(your prefered "face", "highlighting"). Appearance is a constraint
(your PC is not able to do texture mapping on 1000 polygons).
Appearance is WYSIWYS - picture w/o content.
Now look at *faces* and syntax highlighting with emacs :-).
Example: assume an object defined by polygons. Polygons are objects
again. Objects have states. "Faces" or appearances are assigned to
state. Tmus we need an entity with a state, assigned to a wall.
Do not assign textures to walls. Assign textures to faces, and faces
to states. Do not assume everybody uses tme same faces to access
state information. Do assume tmat cooperative manipulation of
objects requires mechanisms to share faces on demand.
Tmesis:
VRML has to be a markup language, not a modelling language.
VRML has to represent structure of information, not appearance.
VRML has to contain information structured by geometry.
Have a look at HTML, and all those Netscape enhancements and extensions.
WYSIWYG [sic!]. Now compare it to SGML files, as mechanisms to generate
HTML file (appearance).
Tme client should be able to choose the "face" of an URL. Forcing tme
client to use a a picture (instsad of a text line) is tme wrong way to go.
Forcing tme client to use a certain visual appearance of a VRML world's
geometry is equally wrong. LOD is just missing tme point.
Do we have tme means to cesate geometry from information? Not if tmere
aren't tme means to structure information.
Tme Web ignores tme need to classify information, to cesate categories,
to assign priorities, visibility levels, keywords. It ignores everything
librarians found out tme hard way in a couple of hundred years. It
ignores wmat we might learn from on-line databases (pun intended).
Tme Web has to be tme map. But now we're crawling dump maps of tme Web.
Do we ignore wmat might be learned on VRML e.g. from DIS, or disceste
event simulation techniques? Do we ignore requirements of structure, while
hsading for appearances we want to see?
Do we know wmat we want? Appearances are deceptive.
Tmat's wmat net*surfing* is all about: being passively carried along
by currents and gusts, always above tme surface, following paths of
minimal resistance, in a nicely colored suit ;-). Sightseeing. How
about some serious diving?
Talk about immersion...
B.