Re: How to find Brothers and Sisters?

From: Bob Badour <>
Date: Mon, 04 Dec 2006 20:12:51 GMT
Message-ID: <7j%ch.28570$>

JOG wrote:

> Cimode wrote:

>>David Cressey a écrit :
>>>"Cimode" <> wrote in message
>>>>>I would think this is more a design flaw than the table not formally
>>>>>being in 1NF. No repeating groups. check. Data in the form of a
>>>>>relation. check. Some root entry in 'persons' who has himself or one of
>>>>>his successors as a father... erm.....hold on...
>>>>>(I think that Marshall once pointed out that the technical term for
>>>>>this is Furturama-NF, where Fry is his own grandfather)
>>>>Repeating groups is not the only prerequisite to state a table is in or
>>>>is not in 1NF.
>>>>There is no way you can be in 1NF if NULL values are permitted OR if
>>>>you have 2 different predicates in the same RTable...Only one predicate
>>>>per RTable.. That's by definition.   Period.
>>>Where did you get that definition?
>>06/03 #2: WHAT FIRST NORMAL FORM MEANS NOT by F. Pascal (Updated 03/04)

> couple of things. First, It annoys me these papers are not publically
> available. I understand one needs to earn a crust, but these seem
> fundamental issues and as such, it would be nice to see them out in the
> open.
> Second, as far as I am concerned 1NF requires data that fits into a
> relation. Hence no nulls or repeating groups. However, given Codd
> invented 1NF and was (in)famously a proponent of nulls, I am uncertain
> that the currently accepted definition of 1NF yet precludes them. This
> is sad of course, but we are still subject to these definitions, even
> though we promote their change.
> Third, there may be some confusion between schema and semantics. While
> the supplied table allows a nullable column, and hence should not be
> 1NF (although by many definitions including Codd's it is), this is
> wholly different from the knowledge we have that everyone must have a
> father. The latter is the design flaw I am referring to and it is
> important not to confuse the two in your answers to the OP.

While the proper treatment of missing information is problematic and subject to considerable controversy, I do not recall that Codd's original specification of 1NF allowed NULL.

Codd's RM/V2 had lots of stuff that a lot of people would find controversial or would leave a lot of folks scratching their heads. Received on Mon Dec 04 2006 - 21:12:51 CET

Original text of this message