[Radiance-general] .oct-instances, overlapping octrees?

[Lars O. Grobe]

Hi list.

I have a question regarding the useage of octrees as instances. Are 
problems experienced when the bounding boxes of instances overlap, or 
only if the actual geometry contained in the instances overlap.

E.g. if I have one hollow cube (instance1) and one small cube 
(instance2) inside instance1, the bounding boxes overlap, while the 
geometry doesn't conflict (because instance1 is like an empty room, 
instance2 like a piece of furniture). Should I expect problems in such 
configurations?

The reason for my question is that I am "atomizing" a huge model into 
small objects, which will be instances placed by markers. The objects' 
geometries will all fit nicely and won't overlap, but of course a lot 
of the octrees' bounding boxes will overlap, and I am a bit worried if 
this concept will make trouble later...

Greg Ward wrote in a reply to my ideas about geometrical structured 
surfaces some time ago:
"If we were to minimize code changes, we'd be using instances of some 
kind with clipping, and we'd end up with a lot of overlapping 
suboctrees.  Many on this list can tell you their headaches surrounding 
those if you haven't had enough of your own.  Getting tighter bounds on 
the instanced geometry would require some rather fundamental changes to 
the handling of octrees, and still wouldn't completely remedy the 
problems with side-by-side instances.  They just don't work that well."


TIA+CU Lars.