Re: Dreaming About Redesigning SQL
Date: Sat, 01 Nov 2003 16:16:13 GMT
Message-ID: <h%Qob.56961$mZ5.338904_at_attbi_s54>
"Mike Preece" <michael_at_preece.net> wrote in message news:1b0b566c.0310312141.67b3f186_at_posting.google.com...
> In the relational model, as in the Pick model, if everyone had just
> one unique phone number (or none at all), the phone number could be
> considered to be an attribute of the person in exactly the same way
> that their name or date of birth could be.
Yes.
> If it becomes necessary to record more than one phone number against a
> person, does the relational model dictate that the logical schema must
> change?
Yes. Doesn't the Pick model as well? Isn't the size constraint on the number of phone numbers part of the schema?
> I don't know the correct terminology, but would the list of
> phone numbers for a person have to be held separately in its own table
> and "referred to"?
This is standard practice, but I'm not sure if "have to" is the right way to say it.
> This doesn't have to happen in Pick. An attribute can be multivalued.
Are you considering an attribute changing from single-valued to multivalued
> Do you agree that a list of a person's phone numbers can and should be
> considered an attribute of the person - just as a single number is -
> provided an acceptable means exists to access all "rows" containing a
> given phone number?
Roughly speaking, this works for me, but I think it hinges on the definition of the word "attribute."
Marshall Received on Sat Nov 01 2003 - 17:16:13 CET