From: broehl@sunee.waterloo.edu (Bernie Roehl)
Subject: Re: TECH: World descriptions
Date: Wed, 15 Jul 1992 14:05:50 GMT
Message-ID: <BrFoHq.1Cu@watserv1.waterloo.edu>
Organization: University of Waterloo



In article <1992Jul15.065912.5580@u.washington.edu> lonachon@anu.edu.au 
(andrew longhorn ) writes:

>OK, I don't quite follow the contiguity problem here. But do we need
>to model boundaries on worlds? The reasons that pop to mind
>immediately are the problems of "being" in one world and looking into
>another,

I would argue that in such a case, they're not really separate worlds at
all -- just different parts of the same world.

> In which case we could model the worlds as "rooms" having
> (inpenatrable?) walls, having only limited doors between them?

Right.  However, I think that in addition to this we need a way of
having multiple worlds ("parallel universes", if you will) that are
*not* connected to each other (and can therefore re-used the
coordinate space an infinite number of times).

>We can't have an amazingly high resolution in a large area
>using a small representation length.

Agreed.  The only point of disagreement here is about what
representation length to use.  32 bits would allow us to implement VR
on a large number of existing platforms, and drastically reduces
bandwidth consumption; these are the main reasons I favor it over 64
or 128 bits.

Of course, there's also the floating-point vs fixed-point debate.

>Could we not agree on an (initially
>experimental) word length for dimensions and use the scaling as
>suggested as a compromise?

I think so.  What I would propose is that we initially use 32 bits,
and make provision for expansion to 64 bits or more when necessary
(and when there's hardware out there to support it).

I'd also suggest we use a fixed-point representation, but include (for
each world) a "world scale" value -- a floating-point number that
indicates the size in meters of the smallest possible fixed-point
value.

In other words, a world whose world scale is 1E-3 would treat the
value 7500 as 7500 millimeters or 7.5 meters.  That way we avoid the
whole issue of floating-point formats and suchlike; people who want to
use floating point can do so, those who don't want to don't have to,
and everybody's happy.

Now, an ascii representation allows us to have variable precision
easily; just use small numbers at first, and use bigger ones when you
can.

However, the data flowing over the network will, in all probability,
be in binary format; how to express a variable-precision value is an
implementation detail that can be addressed later.


	Bernie Roehl, University of Waterloo Electrical Engineering Dept
	Mail: broehl@sunee.waterloo.edu OR broehl@sunee.UWaterloo.ca
	BangPath: uunet!watmath!sunee!broehl
	Voice:  (519) 885-1211 x 2607 [work]