Re: 1NF (Marshall)

On Jan 20, 6:56 pm, "Marshall" <marshall.spi..._at_gmail.com> wrote:
> On Jan 20, 6:59 am, "JOG" <j..._at_cs.nott.ac.uk> wrote:
>
> > *bump*Ha! The canoncal term is "ping" from the ICMP protocol.
>
> But as to the actual question...
>
> I would say that my understanding has moved somewhat,
> but not a lot. I still don't feel like 1NF is on anywhere
> near the firm foundation that, say, BCNF has. I've never
> heard a definition of "atomic value" that I find satisfactory.
> However I am finally convinced I know what "flat" means.
>
>
> A guideline is that we wouldn't want to use a nested
> relation for something that we need to address individually.

But then why ever preclude that and so constrict flexibility? We only need to nest the data in a results view if the user so desires. It seems to me that 1NF is far more necessary to base tables than virtual tables.

> So even if invoices are completely contained within
> customers, we probably don't use a nested relation because
> we want to discuss invoices by themselves. Invoice line
> items, not so much. Antsther thing to look at is "on delete
> cascade" as a guideline that nesting might be appropriate.

I suspect Codd intended 1NF as the antithesis to the navigational approach, hence its primary role, and the initial reticence to bend it to allow nested relations. However I've wondered more recently if 1NF goes further than RM and relation tuples, and that there are notions in 1NF that extend to /any/ manipulable representation of a set of propositions.

I've done some work with hypertext systems and arrived at db's with that slant, so to me the network (or hyperstructure) part of RM is the identity/equality links between primary and foreign keys, and 1NF seems intended to restrict it to this alone. As such On Delete Cascade has always appeared to be a feature that one should turn off instead of turn on. Without it you end up with the database equivalent of error 404's.

>
> It does seem to me, though, that despite any relaxing of 1NF,
> we can get ourselves in to trouble with too much nesting.
> However the reverse isn't true; there is no trouble associated
> with unnesting. At worst you are talking inconvenience.
>
> Marshall

Thanks for the response. I'll let you get back to your lettuce theory with Vadim ;) Jim. Received on Sat Jan 27 2007 - 04:01:02 CET