[PittleyDunkin]

I wouldn't characterize a forward declaration of a function as a distinct definition of its types—any types in the forward declaration must match those in the implementation; there's no way to implicitly refer to the types used in the forward declaration to ensure a single place of typing.