[ODE] box-box collisions

[Gary R. Van Sickle]

(oops, accidentally sent this only to Russ)
[snip]

> this illustrates a point i've made before: that collision detection for
> dynamics is a more difficult problem that collsion detection alone: the
> contact points that are generated must meet extra conditions so that the
> resulting contact-driven motion will be stable. in particular:
>
>   * contact points should be inside the collison volume.
>

This sounds like it doesn't bode well for any sort of "collision anti-aliasing"
schemes, i.e. preventing a fast-moving small object from going right through a
wall.  Are you saying here that even if a suitable collision point was found,
the fact that it was by definition outside the collision volume (because there
is none) would prevent it from being properly taken into account?

--
Gary R. Van Sickle
Brewer.  Patriot.