Re: Relation Schemata vs. Relation Variables

Brian Selzer wrote:
>
> But that assumes that the key value from the current state identifies a
> tuple in the proposed state. That is outside of the scope of a key as
> defined by the model; moreover, it injects meaning into the key that may or
> may not match what the user perceives that meaning to be.

It is the designer's responsibility to choose the right key and to implement the correct transition constraint. That is all.

> Thus, the user
> must specify during the update whether or not tuples in successive database
> states correspond so that transition constraints can be enforced.
>

Exactly. As far as the DBMS is concerned, no such correspondence exists at the logical level. There is only the relation value before the update and the one after it. A constraint based on a notion of tuple-level transitions (identifying tuples by something other than explicit values in columns) would clearly be a dire violation of the Information Principle. It doesn't take much analysis to demonstrate that the consequences would be logical nonsense.

-- 
David Portas