Maybe opengl could benefit from implement a streamlined ray tracing engine into hardware, allowing for near video quality realtime graphics, and accomodating lighting/shadows,ect at the same time. Using pov-ray, on a p2 300, I can software render scenes at about 1 s a scene for low resolutions (as in 100x50)… in a hardware enviroment, it may be easily possible to improve this by several orders of magnitude. The integration could also allow for more complex types of object, beyond triangle, quads, ect. I am interested in hearing what people think of this.
Raytracing has been discussed some times before. Raytracing belongs to a scene graph API, which OpenGL isn’t. To do raytracing, you have to know about the entire scene, which is not the case in OpenGL, it only knows about the current primitive being drawn.