Re: Normalization by Composing, not just Decomposing

From: Eric Kaun <ekaun_at_yahoo.com>
Date: Tue, 13 Apr 2004 22:18:28 GMT
Message-ID: <UGZec.146$Aj7.85_at_newssvr15.news.prodigy.com>


"Laconic2" <laconic2_at_comcast.net> wrote in message news:PJmdnXgEZ9L-SObdRVn-sA_at_comcast.com...
> Eric,
>
> I am sure that what I wrote could be worded much better, but I'm going to
> plead not guilty to word molestation. Next thing, you'll want a law
> requiring pedants to register with the local police.
>
> I hesitate to try to fix up the wording, becasue I don't know if the
missed
> communication between you and me is solely a matter of word choice.

I'm not sure whether it's word choice or the difficulties in choosing basic definitions and attendant axioms. Am I the only one confused by the definitions?

> Let me try just the beginning,
>
> A body of data is made up of data items.

So data is the plural of "data item"? Does the body itself have additional attributes, or is it just a collection of some sort?

> Each data item expresses a value.

What does it mean to "express a value" - does that mean a data item has a value, or is a value, or points to a value?

> Each data item also specifies the state of an instance of an attribute.

So a data item has 2 parts: a value that it expresses, plus the state of an instance of an attribute... I'm lost already. I get a hint of a whiff of what you're suggesting, but that's all. I'm not being deliberately overly pedantic here either.

Typically an attribute has a type and a name. Several attributes can have the same type, and an attribute can be referenced in several "places" (e.g. data items - maybe?). Attributes can be attached to something (and here's where it gets tricky), and each occurrence of such a something (which of necessity also has a type) has a value for each of its attributes. Each such value must be a member of the set defined by its attribute's type.

Given all that, I'm not sure how it matches up with your use of "state" and "instance of an attribute." I could guess, but...

> It's unfortunate that the word "attribute" is used in both relational
> modeling and ER modeling. I think both uses of the word are fair, but
they
> sometimes confuse the issue.

Agreed. Guess there are just too few (English) words to go around.

  • erk
Received on Wed Apr 14 2004 - 00:18:28 CEST

Original text of this message