Re: date as a foreign key

From: David Cressey <>
Date: Mon, 24 Sep 2001 18:42:51 GMT
Message-ID: <L8Lr7.66$>


> Therefore, you insist upon a discrete domain for foreign key constraints
> Then, strings would probably be a poor choice as well, because the

My original comments referred to choices for primary keys. Choices for foreign keys are
driven by the choice already made for primary keys.

Actually, all domains inside a computer are discrete. It is, after all, a finite state machine.

The question is not whether the domain is discrete or not, but whether it's used as an exact representation
of a discrete domain in the real world, or as an approximation to a continuous domain in the real world. If it's an approximation, it makes a poor choice for primary key.

I have no objection to using strings as primary keys. But some strings are better than others. "David Cressey"
is a choice I could object to in a variety of contexts, for a variety of reasons.

    David Cressey
Received on Mon Sep 24 2001 - 20:42:51 CEST

Original text of this message