Irrlicht Engine

Thoughts on this would be appreciated:
http://people.cs.clemson.edu/~dhouse/co ... l-WK06.pdf
Semi-old paper, 2006, i guess, but it's a top hit for "best scene sorting for raytracing" on Google. I'd like to find out how to put together an optimized scene structure.
I'm interested in putting together a raytracer. I know we have one from sio2, though I don't think it'll fit my needs - I'm looking to do semi-NPR.
But speaking of sio2's, does anyone know if that will work with the changes being made for OpenGL in irrlicht?

You want a raytracer to render NPR stuff? I thought the idea of NPR is, that it doesn't have to be realistic.

And what do you want to do? Do you want to do something realtime? Then what would be the benefit of using raytracing in realtime?

Real-time isn't important - more like running on an (older) machine.
The NPR I'm looking for is more changing the object shading slightly, not what you'd normally think of with NPR. I'd like to set a custom function to change the shading here and there. It's experimental work.

Look at what happened in the 90s in the world of 2D images

analytical (vector) -> pure data mit compression (png,jpeg etc)

take a cue and see why raytracing polygons in pointless (aliasing)

I don't get what you're saying. If you're saying aliasing comes from rendering polygons, fine, but some aliasing can be removed by simply taking more rays per pixel.

But anyways... How about a comment on the paper or sio2's raytracer instead of you-shouldn't-be-doing-this remark?