> John Barrus wrote:
>
>> However, I can imagine that most parametric models will increase the polygon
>> count (possibly substantially) without adding information.
>
>I'm interested to know why you think that ? (rendering isn't really my special
>areas; I sit more it the multimedia/toolmaker community). I was imagining
>declaring certain points (polygons) in a mesh as parametric and deformable,
>which would be fairly low cost; in addition to texture map information, etc.
> The parametric information needent sit in the model the renderer uses, I
> thought. It can use something more static and hi-performace ..
>
I was thinking mostly about the trees example you gave. Trees are
notorious for creating lots of polygons in simulations. They are
absolutely necessary for out-the-window simulations, like flight simulators
and driving simulators. Today, high-end machines use a single "billboard"
polygon, texture-mapped with transparency and color - far beyond the reach
of the lowly PC or Mac. A tree might make things look nice, but I haven't
been able to think of what I would use a tree for in a VRML scene other
than beauty.
I'm not trashing parametric models in general. I still believe that the
most interesting 3D web pages (scenes) will be ones that are unique and new
- not ones that reuse models from a global cache. Hopefully, it won't be
long before we know if I'm wrong!
Buildings might be the same as trees. We might make a street with a lot of
parameterized buildings. However, unless we put something in those
buildings, they will be ornaments and not be useful except for slowing down
our frame rate. If those buildings represented an actual physical space
(like downtown Boston) they might have some inherent usefulness without
additional data.