1.  Ease of computation.
2.  Determinism.  (Same answer each time for the same item.)
3.  Irrelevance.  (To prevent locality from aggravating clustering,
the value of the hash function should not be predictable from any
meaningful characteristics of the data item.)