GADT + Newtype deriving = Erroneous errors

[Matthew Brecknell]

Stefan O'Rear:
> They are the same type, and I have Curry-Howard proof of this fact.

If that's the case, then it begs the question why you'd bother defining
Foo in the first place. How would this solve EnumMap performance
concerns?

I am under the impression that newtypes are *defined* to be distinct
types which don't unify. You haven't shown us your proof, but if it
contradicts the definition, then it probably says more about the
definition (or some other construct you've used) than any implementation
derived from it.