Steve Thompson-454462 (11/30/2012)
I haven't done this myself, but I've seen discussions about self-managing identities in replicated systems. For example, if you have two offices and want to make sure that identities are unique , you can set it up so that one uses odd numbers (1, 2), the other uses even (2, 2).
Interesting idea... On one of the projects I'm working on now one of my initial ideas was to have unique ID's across every table (didn't want to use GUID due to overhead) so I seeded each table's PK from a unique 8 digit number. So for example UserID would be 10000000, AddressID would be 20000000 and so forth, and the thought was the first two digits of the ID would identify which type it was. It actually became more confusing for everyone, so I just went back to seeding everything from 1 to keep it simple. But I still like the idea of unique ID's across every table...