quaternions was the only way i could get mine to work... like you said, i had gimbal lock with euclidean.  I had an easier time with quaternions cuz my shapes were cubes... so they would always with a surface flat on axis system.

well it was very hard for me to comprehend... and i still don't fully get it, but mine is dif in the fact taht i was using a physics engine that was giving me the results of the physics in quaternions... i would just pass those onto grrlib to draw the resulting dice... the hard part was being able to take the resulting quaternoins when the dice came to a rest, and tell the game what face was up... i was able to figure that out using google (for about three hours) and finding the mathmatical relationship between the four parts of the quaternioin... (angle.w, angle.x, angle.y, angle.z)  i'll be happy to pass those on to you if they help...  if YOU are the one editing your quaternion values, you could just play with the .w, .x, .y, and .z values to see what they do.  it seemed pretty easy, i just kinda skipped that step since the physics engine was doing that work for me.  Check your email for the GRRLIB code i used to mess with the quaternions.