[ODE] Vehicle with slider joints

[Geoff Carlton]

Hi,
I'm experimenting with a slightly different vehicle model.  Essentially, 
I've got a box, with 4 spheres, each connected via a slider joint.  I 
then manually specify the friction coefficients on the spheres, which I 
work out depending on the speed etc of the vehicle.  This way, I can 
specify essentially 0 friction forwards, and some amount sideways, which 
I can tweak down as the vehicle gains speed, to allow some fun sliding 
around as it takes corners.  After looking at lots of different ways of 
applying the hinge2 joint, and also simulating springs with 4 ray tests, 
I actually like this strange approach the best.

However, I have noticed that the spring like system for sliders seems 
different to a hinge2, even using the same cfm and erp.  While the 
hinge2 seems rock solid, the slider seems a bit out of whack.  Also, the 
slider doesn't have attachment points, unlike the hinge2.  I assume it 
tries to maintain the linear position in the other dimensions, but it 
doesn't explicitly say this (they don't seem to drift so I suppose it must).

I'm considering trying to write my own joint.  This would have anchor 
points, unlike the slider, and internally would call the ball2() 
function, (like a hinge2), but then try to constrain the rotation in the 
same way as an AMotor may do.  Could anybody who knows the internals of 
the joint functions tell me whether this attempt would be worth the effort?

Geoff