Long before Perl and C++ were invented, a hashing algorithm meant mapping the first so many bytes of the key to a starting address, these being designed to be spread evenly across the total available addressing space, and then adding the remainder of the key and the value in sequentially (this needs an end-pointer to be stored at the landing point). Old school hash lookup expects therefore to land immediately on the full match value most of the time and worst case have to trawl along sequentially to find it from there, but not too far to have performance impact. Somehow it looks like this problem would tempt me into going that far old school to achieve all the theoretical benefits, especially if straight C is an option.