[ODE] Trimesh collision problem

[Jon Watte (ODE)]

John Miles wrote:
> The dTerrain geoms don't have that problem, which is another reason they
> need to be considered as first-class types.  The ODE Wiki page is returning
> 404s at the moment; were the dTerrain contributions made official in 0.6?
>   
I don't think so. I believe dTerrain should be re-structured to support 
configuring "right" "forward" and "up" vectors to support arbitrary 
orientations. Arriving at the sample points for that collider would then 
be a dot product in each direction -- this would allow you to set the 
sample density based on the right and forward vectors, too, as well as 
scale the height based on the up vector! The draw-back is that it would 
squash primitive shapes going into heightfield sample space, so actual 
contact generation would have to be done in global space.

Cheers,

          / h+