Re: Surrogate Keys: an Implementation Issue

Brian Selzer wrote:
> Your points are well taken. A relation value is a set of statements of
> fact. I think it's important to be able to determine whether or not the
> reality underpinning one of those statements changed from one database state
> to the next, and to a limited extent, how. This requires the ability to
> correlate the premises in one database state to those in the next. This is
> not possible using the operations available in the Relational Model without
> knowing for certain that there is a candidate key on each affected relation
> whose values will remain constant throughout an update. I may be guilty of
> conflating terms when I speak of updating the wrong row, but the problems I
> described in my earlier post are real.
> ...

Six-foot bunny rabbits are real to some people too. It doesn't follow that reality underpins a giant rabbit database.

p Received on Sat Jul 29 2006 - 19:09:33 CEST