That'd be clearer, though I'd personally consider proscribing these. It's
not like if the fourth column is used you can just blithely ignore it and
expect to get anything meaningful on the screen. Imagine if you took that
attitude towards the fourth *row* ("We thought translations too hard, so
ignored them") - the viewed object is extremely different depending if you
interpret the matrix fully or not. Perspective transformations are in the
same class, only moreso (my sphere turns into a hyperboloid on my system,
stays a sphere on yours since you don't interpret perspective). I'm not sure
what you gain by allowing perspective modeling matrices, if they're not
uniformly supported; generally unsupported features will be avoided by most
content providers, browser creators, etc, especially for such an obscure bit of
functionality.
Personally I know of one person who's ever used perspective transformations
when *modeling*; that was Charlie Gunn, who I believe used them in RenderMan
to view hyperboloid 3D spaces - not your average user. Anyone else out there
who has ever used perspective transforms in modeling (not viewing)? If you
have, do you think it's worth supporting in general, and why?
Eric Haines