[ODE] Russ' plans for ODE?

[Jon Watte]

> Having some method of adding a form of "fire and forget" object would 
> certainly help in the utility library. Perhaps registering callbacks and 
> what not. 

You can build "fire and forget" functionality on top of ODE, if you structure your application into separate components. The "floaty bits" component should be structured to do fire-and-forget if that's what the non-floaty-bits part of your application wants. Note that this is likely to require a little bit of wrapping of ODE, rather than having the application call ODE directly, but you typically need to do that anyway to support things like custom mu and contact modes, triggers based on collision, and remembering trimesh data anyway.

> 1. You need a extremely fast way to determine what cloth nodes can 
> interact with which ODE geoms.

You can use, for example, the hash space, and put spheres in that space for each of the verlet particles. However, brute-forcing it may still be faster, because of simplicity.

> 2. You need a way to determine the clostest position to move a cloth 
> node so that it doesn't violate the ODE geoms.

What's wrong with the "normal" and "depth" values that ODE gives out of dCollide() ?

Cheers,

			/ h+