Re: Treating addresses in 3NF?

From: Mike Sherrill <MSherrill_at_compuserve.com>
Date: Fri, 15 Nov 2002 10:12:24 -0500
Message-ID: <s23atu4r73tdbpujmutac4122ufkirbat7_at_4ax.com>


On Tue, 12 Nov 2002 16:26:40 +0000, Rich Leigh <rtl101_at_ecs.soton.ac.uk> wrote:

>I know that in a lot (i.e. almost all?) databases, a customer list which
>included an address is usually stored as CustomerID, Customer Name,
>Address1, Address2, City, Post/Zip code, blah blah.

Surely not "Customer Name", but "Addressee"? They're logically distinct attributes that sometimes (often?) have the same value.

>But as a purely, uh,
>"theoretical exercise", i.e. I don't have a bet on it with anybody at all
>;), when this is normalised as far as 3NF, I come up with the tables;
>Customers (CustID, CustName, AddressID)
>Addresses (AddressID, Zip/Postcode)
>AddressLines (SequenceNo, AddressID, Data)

Hmmm. Maybe it would help to stop thinking about dependencies for a little while, and think about the simple facts represented by an address. Work on a conceptual model.

-- 
Mike Sherrill
Information Management Systems
Received on Fri Nov 15 2002 - 16:12:24 CET

Original text of this message