Re: does a table always need a PK?

Heikki Tuuri wrote:
> "Lee Fesperman" <firstsql_at_ix.netcom.com> kirjoitti viestissä
> news:3F4A7F1D.291C_at_ix.netcom.com..
> > Bob Badour wrote:
> > > "Heikki Tuuri" <Heikki.Tuuri_at_innodb.com> wrote in message
> > > news:GEg2b.11$G37.0_at_read3.inet.fi.
> > > > 2 years ago we discussed the 'correct' definition of 'relational' in
> > > > comp.databases.theory with several people. I think the concept is
> > > > vague. For example, Codd's 12 principles are not formulated
> > > > as mathematical axioms.
> > >
> > > Read the 1970 ACM paper instead of 12 rules of thumb.
> >
> > Forget it, Bob. He has a commercial interest in claiming that the
> > Relational Model is vague.
>
> if I can recall, the 1970 paper is not formulated as mathematical axioms
> either. Or is it? Do you remember?

Frankly, it is you that is vague about the Relational Model.

> Codd accepts NULLs while Date disapproves. How can 'the relational model' be
> clear if there are differences like this?

Date disputes one of the 12 rules. That hardy makes the Relational Model 'vague'. Codd is the inventor of the Relational Model, and his vision is clear.

I disagree with Date on the subject of nulls and have expressed these views publicly, including during the "Great Nulls Debate". In my view, Date has not proven his case nor provided a viable solution to the very real problem of representing "missing information".

The Relational Model is not some mathematical toy. It's very practical purpose is to model the Real World in a database. It does not shy from the difficult problems this entails. One of these problems is the representation of missing information. No rational person would claim that any solution to missing information is clean (at best, it is quite dirty). I happen to think that Codd made an excellent compromise in including nulls in the Relational Model.

> Lee, of course, has a commercial interest in claiming that 'a/the relational
> model' is crystal clear. He tries to market FirstSQL on that basis.
>
> "* FirstSQL/J Object/Relational DBMS "
>
> By the way, FirstSQL probably is not Codd-12-relational? Why do you claim it
> to be an 'object/relational' database then? Is it 'Lee
> Fesperman -relational'?

FirstSQL/J is not yet compliant with all of Codd's 12 Rules. However, we are improving in that area and are the only DBMS vendor dedicated to relational fidelity. It's on our to-do list!

Codd's 12 Rules do not define the relational model; they are guidelines for users to evaluate the relational compliance of their DBMSs. For more definitive information, see Codd's papers beginning with the 1970 article through his 1979 "Extending the Database Relational Model to Capture More Meaning".

The online white paper "Relational Databases, Do Users Benefit?" (http://www.firstsql.com/ireldb.htm) provides an assessment of FirstSQL/J's relational compliance. It's a little out of date, partial domain support has been added since.

> > He has stubbornly refused to educate himself, clinging to 'thumbnail'
> > descriptions of the relational model.
>
> No, not at all. I looked at the 1970 paper 1.5 years ago.

That's good, but the Relational Model needs much more than a 'glance'.

> > See Database Debunkings (www.dbdebunk.com) for a glaring example.
>
> Well, that quote was taken from a discussion at comp.databases or .theory. I
> did not remember then that Codd had mentioned integrity constraints in his
> 1970 paper.

Integrity constraints is just one area where I faulted your implementation. Here is a link to the actual article -- http://www.dbdebunk.com/page/page/622534.htm.

-- 
Lee Fesperman, FirstSQL, Inc. (http://www.firstsql.com)
==============================================================
* The Ultimate DBMS is here!
* FirstSQL/J Object/Relational DBMS  (http://www.firstsql.com)