[ODE] visibility detemination?

[skjold@cistron.nl]

> I could be off my rocker here, but doesn't ODE do CD at two levels of
> precision?  Like, dSpaceCollide looks for potential collisions, and then
> the callback function uses dCollide to get penetration
> points/normals/depths, etc?
> 
> Would the "coarse" CD of dSpaceCollide might make a good FOV test, if a
> frustum geometry class were added?

Yes, that was the general idea being discussed: To take advantage of the callback used by dSpaceCollide in determining potential visibility, but not to actually call dCollide for it.