Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!uwm.edu!ogicse!cs.uoregon.edu!ns.uoregon.edu!milton!hlab
From: andersen!tsarver@uunet.UU.NET (Tom Sarver)
Newsgroups: sci.virtual-worlds
Subject: Re:  Hardware and VR (Was Re: Cray-Connection etc.)
Message-ID: <1991Apr4.171610.17772@milton.u.washington.edu>
Date: 4 Apr 91 16:28:26 GMT
Sender: hlab@milton.u.washington.edu (Human Int. Technology Lab)
Organization: Human Interface Technology Lab, Univ. of Wash., Seattle.
Lines: 52
Approved: cyberoid@milton.u.washington.edu




It seems that the hardware has limited what we think we should be
able to create in VR.  What about a software architecture which allows
plug-and-play copmonents to be implemented on varying levels of machines
for varying levels of quality?

What I'm suggesting is an analogy to the HDTV strategy of broadcasting
in the highest quality possible, but selling HDTV receivers which can
produce varying levels of quality.  Here, the single, expensive
component (the broadcaster) is sending to another component (the
receiver) implemented on (at least) two different "platforms."  Each
receiver will "implement" some subset of the message, where the
higher quality, more expensive receiver will implement a larger
subset.

The above analogy can be expanded to whole system of components which
are talking to each other (basically in a pipeline configuration).
This strategy can be implemented in, as I see it, at least two
approaches:  the standards approach and the real-time-degradation approach
(I know these are probably not well named, but the names will serve
their purpose).

The standards approach (implemented by HDTV, BTW) sets up some
standards regarding the message and its contents between two components.
One "chooses" the quality level when implementing the standards for
a given platform.  This is similar to X-windows, where the messages
are the same, but each server can have a slightly degraded appearance.
For example, fewer colors (or monochrome, or grey-scale), lower
resolution, and slower updates are all areas in which quality can
degrade without losing the content of the message.

The real-time-degradation approach is similar to a real-time system
which can give an approximation (as opposed to an accurate answer)
when the deadline arrives.  In this case the quality level is dynamic,
based on the circumstances at a given time.  One can see that this
approach is much more difficult to implement because one has to have
a clear understanding of building approximations and refining them.

Do all objects begin life as a collection of spheres and get refined
into something more realistic?  How can we apply this paradigm to all
the components in the architecture?

Implications of the whole plug-and-play goal are if you allow graceful
degradation, you can lose the sense of "being there."  This is where
you make the tradeoff between content and form.  "Content" is traversing
the database, or watching the air flow, or whatever job one is doing
within the VR.  "Form" relates to how real the environment feels.  I
don't think the proponent of _Neuromancer_ ever felt like he was walking
in the park;  the synthetic world felt "real" but still synthetic.  He
could differetiate between being "jacked in" and not.