Re: does a table always need a PK?

In an attempt to throw the authorities off his trail, "Heikki Tuuri" <Heikki.Tuuri_at_innodb.com> transmitted:
> Yes, the above definition is exact and mathematical, because it
> describes a mathematical relation. It does not describe an
> 'RDBMS'. You cannot find a formal specification of Codd's rules
> below from it.

It describes, with pretty adequate formality, what "relational" is supposed to mean.

A database system claiming to conform to the "relational" model needs to start by conforming to being, well, "relational."

This is quite sufficient to establish that a lot of database systems that claim to be relational simply aren't, simply by considering the "uniqueness" criterion.

The uniqueness criterion demonstrates a big, bad bug of _practical_ importance since duplicate tuples generally are errors, and are a problem to fix when they occur.

-- 
(format nil "~S_at_~S" "aa454" "freenet.carleton.ca")
http://www.ntlug.org/~cbbrowne/emacs.html
Can I get a rubber stamp that says "don't spam me?"
-- Ken London on alt.accounting...