Re: Non-unique index and normalization

"Daniel Morgan" <dmorgan_at_exesolutions.com> wrote in message news:3CD03D05.6A430B90_at_exesolutions.com...
> John Hunter wrote:
>
> > Is there a normalization rule that indicates that there must be an
unique PK
> > key on every table in the database? I am having a debate with a
co-worker
> > and one of us says that there has to be one and the other
disagrees. Who is
> > right?
> >
> > Thanks,
> > -John
>
> While it may exist in theory, Celko and others can answer that far
better than
> I, no such rule exists in practice.

I agree, no 'rule' exists in practice but all tables, unless there are duplicate values in the table, have a 'default' PK: all the columns in the table.

The 'normalization rule' that defines that a table should have a PK is 3NF. Here's an excerpt from An Introduction to Database Systems, 3rd Edition - C.J. Date (Yes, a very old copy! :)

A relation R is in the third normal form (3NF) if and only if, for all time, each tuple of R consists of a primary key value that identifies some entity, together with a set of mutually independent attribute values that describe that entity in some way.

--
Pablo Sanchez, High-Performance Database Engineering
mailto:pablo_at_hpdbe.com
Available for short-term and long-term contracts