Re: database systems and organizational intelligence
Date: Thu, 27 May 2004 09:43:22 -0400
Message-ID: <BdydnYyQb8hHbijdRVn-gg_at_comcast.com>
"Alfredo Novoa" <alfredo_at_ncs.es> wrote in message
news:40b5be6c.187579_at_news.wanadoo.es...
> You can represent a tree with a relation. There is no problem.
You can represent a relation with a tree. I don't understand your point.
> I disagree. Optimizer compilers are here since a while. If you
> translate the optimized syntax trees into code you will have
> normalized code.
The word "normalized" can apparently mean many things.
My first exposure to this word was in the processor manuals description of floating point arithmetic.
In particular, there were synonyms among the floating point representations. That is, different representations for the same number. Of these, one of them was chosen as the "normalized" representation. It was the one whose fraction had a high order bit that was different from the sign bit, or something like that. And there was an operator called the "unnormalized floating add" that would add to floating point operands, but not normalize the result.
I think of the above definition as utterly unconnected to the meaning that Codd gave for "normal form of data" in the original paper. And I have yet to see a definition that legitimately revises that definition. Normal forms beyond 1NF extend 1NF, but do not revise 1NF. Received on Thu May 27 2004 - 15:43:22 CEST