[ODE] Contacts/constraints?

[nlin@nlin.net]

Russ Smith wrote:

> * contacts may be implemented with non-constraint methods

What do you mean by this? Manually adding externally computed penalty forces
or impulses to the model? 

While we're on this topic, two theoretical questions...

- ODE doesn't use any impulses, right? So collisions and interpenetrations,
  which are violations of the transiently created contact joints, are
  satisfied (within tolerance) by the constraint force solver, which forces
  the system configuration from an illegal configuration to a (nearly) legal
  one on the constraint mainfold? Therefore, is it correct that the 
  specification and enforcing of non-penetration constraints (i.e. contact
  joints) essentially embodies the empirical laws of collision (reversal
  of relative velocity along the contact normal) which need to be explicitly
  coded in an impulse based simulator? What I am really asking is, how does
  all the proper rigid body motion get imparted to the body after collision?
  Is there some magic in the contact joint, or is it inherent in the constraint
  based method that satisfaction of non-penetration constraints automatically
  imparts the correct post-collision momenta to the system state?

- As far as I understand it constraints can be classified as holonomic or
  non-holonomic. Does ODE support both kinds of constraints?

Thanks,
-Norman Lin