There is no way to determine all the sets that an identifier belongs to. This is typically referred to as set orthogonality because the protocol allows a harvester to find out which identifiers belong to a particular set but not vice versa. This is not as much of a problem for a flat space of archives, but organizations like NDLTD and NCSTRL have already started to create hierarchical catalogs based on OAI and existing set information is lost at the very first level. Also, the Internet2 Distributed Storage Initiative wants to work on replication of OAs - this will mean harvesting every set and dealing with duplicates. Can we do this in a way that is more efficient without adding to the complexity?