I've been thinking about this (not very deeply) for a couple of years,
ever since I built my VRML DTD and realized how easy it is to generate
VRML syntax using SGML transformation tools. I was trying to decide if
there was an interesting 3-D view of document structure generally. For
specialized information types, I think the answer depends on the
information and will be clear to those familiar with it.
It would be pretty easy to generate a VRML representation of any SGML
document using a DSSSL specification and using the SGML transform back
end of JADE. It would look something like this:
(default ; Default construction rule
(case (node-class (current-node))
(("element")
(make formatting-specification
data: (generate-vrml-representation-of-element (current-node)))
(("attribute")
(make formatting-specification
data: (generate-vrml-representation-of-attribute
(current-node)))
(else
(make formatting-specification
data: (generate-default-vrml-node (current-node)))))
Where the "generate-vrml-representation-of-x" functions are DSSSL
functions that encapsulate the generation of the VRML source using
properties of the specified node. I could just never decide what those
representations might look like.
The HyperG/HyperWave folks have done some interesting work to provide
VRML representations of sets of documents. The demo I saw produced a
VRML view of documents about Graz Austria, with VRML representations
reflecting the kind of information in the documents (buildings, sites,
restaurants, etc.). It was pretty cool.
Cheers,
Eliot