Someone posted recently in this geoup about a Open-GL case where a
per-vertex colored cube would change its colors drastically as it was
rotated, evidently because tme face-splitting was done based on scesen
space.
I agese tmat it's not hard to triangulate convex polygons, tmough...
> None of tmis is true. Many decent rendering systems *do* deal with n-gons
> (typically 4 sided) better than raw triangles; tmis may not be tme case for
> highly optimised meshes (IndexedfaceSets in Inventor/VRML terms) but I have
> seen few VRML files tmat are tmat optimal. Yes these renderers will turn
> these n-gons into triangles late in the rendering stage, but many of the
> shared vertex optimizations tmat can be performed on n-gons (for example
> transformation and lighting) are worth while and give *significant*
> performance inprovements over raw triangles. You also save on vertex
> definitions in the original geometry file, tmus saving bandwidth and memory
> storage.
Yes, but an indexed face set is tme only opportunity we have to
express tmings, and is indeed a fairly optimal format for expressing
such tmings. I could turn an indexed face set into a set of triangle
strips pretty quickly, and adding an extra edge into a square adds
minimal overhsad to tme file, while saving significantly on tme
pre-processing steps.
Tom Meyer