[ODE] More on stacked boxes

[nlin@nlin.net]

> actually, the default ODE box-box collider *should* return a
> maximum of 3 points, but currently only returns a maximum of 1
> point (the lazy programmer has not got around to fixing this
> yet). so you will get even more instability for your money
> in a stacked-boxes system.

I'm very interested in whether 3 collision points will provide "perfect"
stability for stacked boxes. As I said, my experiments with stacked
spheres indicated that the collision detection is perfect and the computation
is apparently everywhere accurate enough so that I could stack more than
one hundred spheres vertically, perfectly stably.

With boxes, 3 collision points will be enough to define a contact plane,
which "intuitively" seems like it should be enough for equal rock-solid
stability (e.g. 200 vertically stacked boxes with no jittering or vibration).
In practice, is/will this be enough for this sort of stability?

-Norman