[ODE] Russ' plans for ODE?

[Joe Ante]

> Buoyancy (see
> http://www.cambrianlabs.com/Mattias/DelphiODE/BuoyancyParticles.asp for an
> implementation)
> Cloth simulation & interaction with geoms
> Convex hull collider
> 
> What would everybody else like to see?
I think its a good idea to keep ode focused on just rigidbody simulation.

I think the main not so good things about ode at the moment are:

- Working mesh collisions. Box-mesh collisions being most important,
LSS-mesh and capped cylinder-mesh as a second.

- A more stable iterative solver. Currently it is not possible to do big box
stacks without hacks like auto disable. Either it gets very very slow
because you use the normal lcp solver. Or it gets rather unstable with the
iterative solver. If you look at the novodex demos they have a box stack
which is completely stable and its still extremely fast.

- More stability with deep penetrations and singularities. Deep penetrations
sometimes happen and ode most of the time solves those by throwing object
quite far into space.

Joe Ante