I need rotational collisions for skeleton/body physics, and also it's alot better if the cubes rotate... And yes Olof, i've read alot so far, and I realize there is alot of work that goes into physics engines, but I find if I make my own engine, it is easier for me to debug and modify. I know how versatile they are to use, but I don't think I need to implement all the features they have. The example i've looked at uses points and tests them in a feild, i'm not sure exactly because the dependencies are a little confusing, but im guessing the way to check basically would be to check all the points on a plane go through another plane? This mathematically makes sense, but would take alot more processing power.

http://www.openprocessing.org/sketch/39230

I wasn't saying anything about the work going into the physics engine, I just said there is a non-trivial amount of work _using_ any physics engine, and that you can learn a lot from it that is useful when you make your own. And most are very modular, so you can use for example the collision detection part of it, and implement your own physics, and thus take it one problem at a time.

There is a certain pride in feeling you have control over _all_ the code in your project, so I totally understand your wish to roll your own