> Recently i saw a running demo from General Magic's "MagicCap" UI
> environment. It seems to do a lot of the stuff mentioned by others on this
> list earlier (assuming i understand the emails correctly ofcourse). MagicCap
> UI shows a downtown view on the desktop, using a mouse one could navigate
> (VR style) around houses, rooms, hallways, libraries, etc...One could
> customize wall colors, wall papers, posters, and other artifacts in and
> outside the rooms. Drag and drop feature is supported. Linking of objects
> is thru drag and drop. However i am not sure if linking of objects over
> distributed networks is supported. They claim everything in their
> environment is an "object" and almost every object could be linked to any
> other object.
I haven't seen Magic in person, but have read the blurb about their
programming language and seen demos in a television programme about it.
They have a neat scripting language in which you can "fork" processes
over the network link so that you end up with the child process at
the other end. They also support named classes so that if both ends
agree that they share some classes, then the objects can be compressed
to just the data, with the code stripped out for sending down the wire.
My problem with this approach (apart from Magic being a proprietary
language) is that it is too procedural and subject to security concerns
about downloading general purpose code into your own system. I believe
we can develop am approach similar in spirit to SGML for a more declarative
approach in which behaviours are specified using scripts which are highly
restricted in their API. I am currently following this approach with HTML+
for adding client-side checking of forms.
We were recently shown a demo of a "VR" system running on a PC. It used
realistic 3D rendering with distancing and neat behaviours for objects,
e.g. you could open a drawer in a desk and take out a working calculator
or pen. It shouldn't be too hard to realise this kind of thing in the
WWW context.
Dave Raggett