Take two differantly shaped prims (ie: a cube and a sphere) and apply blending to them to end up with, say a cube with rounded edges. Blending could be done as a percentage of each shape, so you can have something more cuby or more roundy.
Make it cost two prims for the final blended prim and have a standardized methedology for how each prim shape will blend into each of the other prim shapes.
Do able?
Worthwhile?
Already been suggested..?
Just thought I'd throw it out here in case it hadn't.
-[jaq]-