Re: History of 1NF
Date: Thu, 13 Mar 2008 01:21:01 GMT
Message-ID: <1k%Bj.77466$w94.11329_at_pd7urf2no>
David Cressey wrote:
> "Marshall" <marshall.spight_at_gmail.com> wrote in message
> news:97b499ef-f27d-4c45-aae0-8510a5650dbe_at_i29g2000prf.googlegroups.com...
>> Back in the day, Codd originally defined First Normal Form >> (or maybe he was just calling it Normal Form back then) >> and proved something about it. Something about an >> isomorphism between nested relations and unnested >> relations? Could someone state specifically what he >> proved, please? And if anyone has a reference, that'd >> be good too. >> >> Thanks, >>
>
> I don't know anything about the 1969 paper, other than its existence.
>
> In the 1970 paper, he claimed and demonstrated that for any collection of
> nested relations, there is a collection of unnested relations that can
> express the same information. I'm not sure whether "can express the same
> information" is an isomorphism or not, but I think that is what you are
> driving at. He didn't use the terms "nested" and "unnested", but again I'm
> pretty sure we're talking about the same thing here.
>
> He called the collection of relations with no nesting the "normal form" for
> that data. This term later became "1st normal form", when other normal
> forms were discovered. What he called a collection of relations is, I
> think, what you and I call a "schema". It's clear from the paper that
> normalization is not an absolute requirement for using the relational model
> of data.
>
> I read a lot between the lines about what Codd's intent might have been in
> including and excluding various topics from the 1970 paper. This is thin
> ice. It is fair to assume that he was quite judicious about what to leave
> in and what to leave out. The paper covers an enormous amount of ground in
> relatively few words.
>
> Here, I think that it might have been quite a lot easier to build the first
> relational database management system if one took the simplifying assumption
> that all attributes would be "simple" datatypes. (There have been a lot of
> arguments in c.d.t. about what makes a datatype "simple", but in the
> context of the paper, the term should be clear enough.) I think that Codd
> may have been raising the point about normalization in order to forestall
> arguments against building the first RDBMS based on the complexity of
> supporting nested relations.
>
> His point is that even if you don't support nested relations, the schema
> designer is not limited by that in the expressive power of the system.
>
> Anyway, this is my take on it. I never actually read the 1970 paper until
> after I started participating in c.d.t. You will remember a certain former
> participant who said that she had read the paper over and over again, and
> still didn't unerstand it. I was naive enough, at that time, to believe
> that I could help explain what Codd might have meant, and this would be
> helpful.
>
> This is just my take on it. HTH.
>
>
My take is pretty much the same, pretty clear to me that Codd was careful on several fronts including putting it in a way that practical ways to implement would be fairly obvious (it seems System R missed a point or two).
I remember reading an interview with Codd in the early 1990's, I think it might have been in the long-gone DBMS magazine where he said that the term "normalization" had come to him when there was talk by the US' President Nixon or Sec. of State Kissinger about normalizing relations with China. Later I read Date saying that wasn't so. One of the few times I disagreed with Date (not to say I always understand him, though).
I think Codd might have been merely putting a stick in the sand with his "normal form", at least as far as structure or precluding rva's is concerned. He could have been thinking only of the calculus, ie., the operators of the algebra when he mentioned that. Regarding any isomorphism I think the essential point in the 1970 demonstration was that if you introduce enough attributes, you can express the same info without nesting. His example didn't exactly introduce attributes, since they were already present, but I didn't think that difference was important. (By the way, on Darwen's newsgroup a couple of years ago, somebody pointed out that there are various typo's in the 1970 paper, not in the text but in the figures. Received on Thu Mar 13 2008 - 02:21:01 CET