Re: Constraints and Functional Dependencies
Date: 25 Feb 2007 14:59:25 -0800
Message-ID: <1172444365.872334.49350_at_t69g2000cwt.googlegroups.com>
On Feb 25, 9:02 am, paul c <toledobythe..._at_oohay.ac> wrote:
>
> I thought Marshall had retracted "foreign key" and meant it to be
> replaced with "referential integrity".
Yes.
> My only problem with the latter
> is that I haven't been able to find a quote that I can trust where Codd
> indicated whether "referential integrity" involves a "candidate key".
Agreed.
In fact, it seems to me the useful property of a SQL foreign key is exactly the forall ... exists property I described in the OP. As you've said, SQL also has the requirement that the attribute(s) named in the exists clause must be a (primary?) key, but neither of us can see a hard, logical reason for that restriction. Even though it does capture what is probably the 99% use case.
I pulled about 10 definitions of "referential integrity" off the web and about 7-8 of them matched my forall ... exists construct.
The funny thing is, my mentioning of FKs was offhand, and only meant to indicate roughly that we can dispense with "special purpose" constraints as found in SQL. Mostly I wanted to talk about FDs. But this conversation is good too.
Marshall Received on Sun Feb 25 2007 - 23:59:25 CET