So allocate an instance from an implied set of states instead of explicitly traversing the enumeration. Nice for a business process likely to need change, or where the actual state space is large, but inappropriate for something like utf8 decode where fixed space is required.
I'm not familiar with utf8 decoding, and I focus all of my efforts on thinking about business software. I think it's pretty guaranteed that different representations of anything will make sense for different use cases.
