Some hyatt revelations

Functions should be handled by the output domain.  This is because function execution should, in essence, propagate inward: one function has other functions as arguments, and each of those arguments is typed, so that domain is queried if the argument is the result of a function evaluation.

EDIT: actually, that might be the opposite of how I wanted to do it.  I'll need some hypothetical situations to buffer up my position (the issue is overloading).  Either way, good on me for remembering about the existence of vector domains, which make this whole setup much more sensible.

EDITx2: Functions are identified by signature, which means that they should be stored by parameter domain and name.  To determine functions higher up, first find the output domain, and feed that to the function in question.