Re: Proposal: 6NF

From: David Cressey <dcressey_at_verizon.net>
Date: Fri, 06 Oct 2006 12:47:03 GMT
Message-ID: <bfsVg.77$6h.45_at_trndny01>


"J M Davitt" <jdavitt_at_aeneas.net> wrote in message news:FufVg.4131$Cq3.3594_at_tornado.ohiordc.rr.com...
> David Cressey wrote:
> > "David Cressey" <dcressey_at_verizon.net> wrote in message
> > news:F4uTg.998$rH1.633_at_trndny05...
> >
> >>First, let's find out where we all agree, then work our way towards
where
> >>there are different viewpoints.
> >>
> >>Here's a starter: A schema in DKNF will have no nullable columns.
> >>
> >
> > I got no answers to the first point, but I think we all agree.
>
> (I guess my contribution was lost in the ether.)
>
> >
> > Here's the next point.
> >
> > An outer join between two relational tables will, in general, produce a
> > table with nullable columns.
>
> Wait a minute: what does OUTER JOIN have to do with
> relational theory? Sure, SQL offers an OUTER JOIN,
> but...

Good point. As I pointed out to Bob Badour, I have NOT asserted that outer join is a relational operator or that the result of an outer join is a relational table.

I'm just trying to establish the common ground, before we get to areas of disagreement.

>
> There are *lots* of things I don't like about "outer
> join between two relational tables:"
> - "outer" join
> What's outer about a join?

I'm just using the term I'm given.

> - join "between"
> "On" or "of," sure; but "between?" Seems strange.

I don't get this objection.

> - "relational tables"
> Relations or tables: they're different. Pick one.

I disagree. The term "relational tables" is very widespread in writings on the relational model. I'm not asserting that a relational table IS a relation. But I am asserting that a relational table REPRESENTS a relation. Not all tables are relational.

> - "in general"
> Very vague.

This was just to avoid going down a rat hole. You can come up with an outer join that doesn't produce any nullable columns, but this is the trivial case. The general case of outer join will produce nullable columns.

> - "[A] table with nullable columns..."
> ...is not a relation.

I'm coming to that.

>
> > Note: Brian Selzer has carried this point further in another subthread.
>
> Yes, he has.
Received on Fri Oct 06 2006 - 14:47:03 CEST

Original text of this message