That's not for dots, it's for bodies in 3D with up to six degrees of freedom each

(2D movement would mean three DOF each rigid solid).

I don't know Flash but that can be integrated step by step blazing fast by coding in any generic language. For the integration of the differential equations a coarse first-order finite difference approximation would be enough for credible effects--you only need better approximations if you need to predict the behaviour of a real physical system.