Re: Proposal: 6NF

J M Davitt wrote:
> Jan Hidders wrote:
> > J M Davitt schreef:
> >
> >
> >>dawn wrote:
> >>
> >>>Given the definition of NULL that I typically use (with non-SQL based
> >>>solutions), NULL is a value and can be modeled mathematically with the
> >>>empty set. In that case, a relation tuple with a NULL is as valid
> >>>mathematically as one without. Agreed? --dawn
> >>
> >>Well, if it's merely a matter of definition, then, "No. Not agreed."
> >
> >
> > Since Dawn defines NULL as the empty set your answer implies that you
> > think the empty set is not a value. Why?
> >
> > -- Jan Hidders
> >
>
> Partly (i) NULL is already overused

I am using it only one of the ways it is used, unless I qualify it as "SQL NULL" which is quite different from most other uses of NULL within the s/w development profession.

> and (ii) the language
> used on this thread is imprecise.

All language is imprecise. We always make assumptions about what the other person be able to intrepret as it is intended. When there are differences of opinion, then it is helpful to increase precision in the definitions.

> For instance, the reference to "relation tuple" in the
> above paragraph: is she distinguishing between tuples
> in relations and tuples otherwise?

I was responding to JOG's statement
"A relation tuple with a NULL in it is no
> > > relation tuple at all. From a mathematical perspective, case closed. "

So I used the same language.

> What, exactly, does
> "valid mathematically" have to do with her point?
>
> [I realize I'm arguing against a point not explicitly made.]

I was responding to JOG

"A relation tuple with a NULL in it is no
> > > relation tuple at all. From a mathematical perspective, case closed. "

indicating that the this depends on the def of NULL. The case is closed from a mathematical perspective on an SQL NULL (which, I think, was JOG's point), but because he did not qualify or define the NULL, I wanted to be clear that the case is not closed on the use of NULL with databases, in general. That depends on the def of NULL.

> Relations can be empty. zero tuples. Tuples can be empty:
> zero attributes. Can attributes be empty? No, they require
> a value. And, without treatment of attribute types, we're
> going to invent a value, name it NULL, and use it as a
> placeholder whenever we don't have handy another value to put
> there?

I can only guess that you entered this profession long after I did, as otherwise you should know that it is SQL that redefined the NULL to be handled logically as the absense of a value. There are many products that use a NULL the way I have described, where NULL = NULL, just as LOW-VALUES = LOW-VALUES or "George McGovern" = "George McGovern."

> We have, so far, been able to discuss empty relations and
> tuples without attributes without having to use the word
> "NULL." Why must we embrace the noxious term just because
> someone wants to talk about attributes without values?

? "Null" is in the OP, so I don't know what your point is.

I was responding to JOG's statement
"A relation tuple with a NULL in it is no
> > > relation tuple at all. From a mathematical perspective, case closed. "

which is accurate for SQL NULL, but not for NULL in general.

--dawn Received on Mon Oct 09 2006 - 08:56:36 CDT