Might a Hu-Tucker coding of keys truncated to a fixed length be useful as a *partially* order-preserving hash?
Collisions would at least be adjacent to each other in the original ordering, and If a suitably representative set of probabilities were used to calculate the initial coding, I suspect the distribution of hash values would be good in practice. Or, perhaps achieving even distribution in the first N bits of coding could be another constraint on the choice of order-preserving codings. - Gordon _______________________________________________ p2p-hackers mailing list [email protected] http://lists.zooko.com/mailman/listinfo/p2p-hackers
