In combination with the Issues section of HashTablesCowan, why the apparent bias against user-supplied hash-functions?

While sets keyed by simple types (numbers, symbols, strings, characters, ...) certainly are extremely useful, I also use complex structures with mutable elements not contributing to the equivalence predicate (or hashfunction), and I'd certainly want also to use them with records (distinguished by contents) too.    

SRFI-113 can't do either of these as far as I can tell, but I can live with that if a separate hashtable api gives me the added generality.