Going back a couple of posts, I wouldn't really call bending normals in any sort of bad way cheating. Normals (in this type of instance) are only really used to give some sort of approximation for environmental effects anyway like shading/normal mapping, cubemapping/refraction/reflection which can all be done in varying ways with different results.
imo, the 'I did this by cheating' thing in demos should only really come into it when the machine is made to do something (or looks like it's doing something) it really shouldn't be capable of which back in the amiga days was probably a lot more common.
Bending normals based on an unrelated physical characteristic such as intensity of lighting or direction of viewer is what I mean by cheating.
I agree what you see in those images could be simulated in other ways, maybe at 50 fps, maybe in 1.1k (current exe size). Maybe. Though I really seriously doubt it. I *know* you cant do it with software rendering at those resolutions and I'm 80% sure you cant do it with traditional opengl and dynamic cube mapping, though if anyonme wants to prove me wrong, I'll be happy to see the code!!!
I was trying to point out that raytracing isnt seen as a methodology to cheat and create unusual effects normally and I think this is a shame. Most people spend forever tuning performance or lighting to be fast and CORRECT. People have spent years and years finding ways to do multi-pass, environment mapped, multitextured polygons to get "effects". Nobody seems to be doing this with raytracing yet.
If its just the first images dont convince you then have a look again, and if you can think of *any* way to do a real time wada basin in normal opengl or software rendering...I'll eat my code :-)
No way can *that* be done with environment mapping.http://sizecoding.blogspot.com