Re: Surrogate Keys: an Implementation Issue

First of all, a surrogate key is not exposed to the user. That is straight out of Dr. Codd. The system maintains it, like an index and it does not replace the relational key.

Commercial examples include Teradata hashing is one example; Sand compressed bit vectors are another and Sybase SQ Anywhere are been using pointer chains from FKs back to a single occurence of the referenced PK or UNIQUE() constraint set.

IDENTITY is not a surrogate, nor can it be used as a relational key, nor is it practical becuase of the lack of validation and verification. Received on Sat Jul 22 2006 - 15:56:15 CEST