@zithral The model I imagine is: colours are triples (r, y, b), with white = (1,1,1). Mixing colours is taking the minimum of each component.
Is that what you thought?