Re: 4NF is Where It Is At! [WAS Re: 1:1 relationships]

In article <vPei6.35402$2X4.107790_at_petpeeve.ziplink.net>, David Cressey says...
>
>"Jan Hidders" <hidders_at_REMOVE.THIS.win.tue.nl> wrote in message
>news:96bjhb$rfh$1_at_news.tue.nl...
>> Interesting example. What you describe is a dynamic database
>> constraint. But normalization usually only deals with static database
>> constraints. Because the constraint does not imply any new functional
>> dependencies (unless you introduce a column CO), it has no impact on
>> the normalization process.
>
>Can't the situation be described as a multivalued dependency? For the
>moment, let's leave the CO
>out of the model. I just included a description so that we who are
>discussing this can have a conceptual
>base for understanding the resulting relationship between phone numbers and
>addresses.
>
>In particular there is a set of addresses that are served by an office, and
>a set of phone numbers controlled by that office.
>That means that, given an address, I can't tell you specifically what the
>phone number is, but the address determines a set of possible phone
>numbers. Likewise, given a phone number, I can't tell you what the address
>is, but I can give you a set of possible addresses.
>
>Isn't this an example of a multivalued dependency?
>

If you leave out CO, then there would only be 2 columns, while MVD requires at least 3.

With CO column there are 2 interpretations. 1. Each CO describes a range of phone numbers and a range of addresses it connected to. It doesnt care if particular address has particular phone number assigned. Here we have MVD:
CO->>Addr|Phone
2. We do care which addresses are assigned to which phone numbers. Cannot deduce any MVD in this case. Received on Fri Feb 16 2001 - 01:48:02 CET