Re: Views for denomalizing

"Alfredo Novoa" <alfredo_novoa_at_hotmail.com> wrote in message news:t407015adk2hcvr56sh04an51rv85fjgfv_at_4ax.com...
> On 2 Feb 2005 22:34:47 -0800, lauri.pietarinen_at_atbusiness.com wrote:
>
>>1NF is just another way to say that all data must be perceived as
>>relations,
>
> That is the Information Principle, not 1NF.
>
> The Information Principle is a prerequsite to normalization.
>
> IMO 1NF is supefluous and only exists due to historical reasons.
> Normalization starts with 2NF.

This is the new way to define 1NF which then permits those who have always thought 1NF was imperative and have defined every other NF in terms of it, to still hold everything that have held to in the words that are used, but change the underlying meaning of those words so that the majority of practitioners are unaware that 1NF now means something different.

I'm interested in getting the word out on how what-use-to-be-called-1NF is now NOT NECESSARY -- AT ALL! By folks like Date and you simply changing the meaning of 1NF, no announcement is going out to professors in any big way that the ought not teach 1NF like they used to, nor to data model practitioners that they don't have to remove those relation-valued attributes when modeling for any reason than tools not supporting their efforts, etc

I want to spread the word -- what used to be referred to as 1NF, that I now have to find some other term for because otherwise I keep calling it 1NF, is DEAD! I don't find it helpful at all that the term has been redefined. I can see that helps with the slight of hand that is being done in relational circles. How can we get the word out that 1NF, as understood in the 70's, 80's and by many sense that time, is no longer?

>>and relations, by definition do NOT have duplicate tuples (or rows, in
>>SQL-speak).
>
> Nor nulls.

This always throws me off since it is SQL NULLS -- three-valued logic NULLs -- that are disallowed. For any tools where NULL is a value, rather than the absense of value, this is not an issue. I think I am correct in saying that there is no such rule required when working in a two-valued logic rule (which is where I prefer to work). Is that right? I also work in an environment where every relation is a function (that is, has not only a candidate key, but a single declared key). Thus the issue of no duplicate rows isn't even brought up. So, if you DBMS tools use a two-valued logic and all relations are functions, then the only 1NF issue is the one that has just been defined away, thereby forcing every relation in this environment to necessary be in 1NF. However, the model defines itself (and IBM defines it) as being a NF^2 (NF squared) or Non-First Normal Form database. So, you can see how this redefinition of 1NF, while it works for Date and for you, is terribly confusing to some of us.

Please advise. Cheers! --dawn

> Regards
Received on Fri Feb 04 2005 - 17:32:01 CET