[ODE] Re: A (hopefully) simple collision response question

[Tim Schmidt]

> Could that hierarchy be automatically optimized just prior to the start of
> the simulation, by 'flattening' it, so that each object was exactly one
> transform below the root transform?

Probably for 'deep' scenegraphs this would reduce the amount of computation time
that must otherwise be used to calculate the final transformation of each leaf
object.
As far as I know the BranchGroup.compile() function in Java3D does this, too.

Well, at the moment such a behavior is not included in my TransformGroup,
because I primarily needed a kind of 'twin' class to the Java3D TransformGroup.
Connecting both worlds have become easier with it. 
Actually I have just combined the existing dGeomGroup and GeomTransform and
reinterpreted the vector and the matrix in the latter as a relative offset for
all underlying geometry in the corresponding collideTransformGroup(...)
function.

Maybe I will add the 'flattening' at a later time, if I have to care about
performance issues. But currently I have to focus on other things, so that I can
not make any promises.

Surely, if somebody has the time ... (just thinking out loud, too *g*)

 - Tim