Friction in collision response of circular mass

[cleverer]

I'm working on a 2D simulation of the linear and angular motion of rigid circular masses. Here are a couple things I'm trying to figure out:

1) I want collisions to have friction so that e.g. when spinning masses bump into each other, they exchange some of their angular momentum. But since collisions are modeled as instantaneous, there is inherently no friction since the masses spend zero time in contact. So, how do I make it do what I want? Do I just choose some arbitrary constant to replace time in the friction calculation or is there a more general way to handle it?

2) I also want friction to affect linear momentum so that e.g. a mass can roll along a flat surface. But frictional force is always tangential and tangent forces on a circle don't affect its linear velocity, according to the math I've been told, so how can anything ever roll? I suspect I'm overlooking something fundamental here.

Thank you