Currently SRFI-126 doesn't define the type of the object returned by
the hash-salt procedure.

Surely it should be an exact integer.  But should it be constrained to
being non-negative?

Hash values themselves are non-negative in SRFI-126 (following R6RS),
so I'm considering doing the same for the salt value.

Apologies for noticing this so late.  I believe it's a small enough
amendment to not warrant an extensive discussion though.  Let's say
non-negative is settled if there's no dissent within a week.

Taylan