Re: Define "flatten database" ?
Date: Sat, 25 Feb 2006 20:54:00 GMT
"Mark Johnson" <102334.12_at_compuserve.com> wrote in message
> "David Cressey" <dcressey_at_verizon.net> wrote:
> >define a "relational table" as the representation of a relation in a
> >Describe an "SQL table" as an approximation to a relational table. Move
> >from there.
> That paper of Codd's that I quoted in a couple of messages showed his
> desire to even replace the idea of the table/relation correspondence
> with that of a table/relationship.
Were you quoting from the 1970 paper? If so, the paper is clear about the difference between a relation and a relationship. In a relationship, the attributes are specified by name rather than by order. And Codd gives a motivation for doing things this way. It's a burden on the user, Codd says, to make the user remember the order in which a tuple of degree 30 or so specifies its values. I agree with that.
> If the tuples were not always in
> order, as they are in picking elements from successive sets in the
> theory, then he perhaps felt uncomfortable still terming a table based
> on that, a relation. Thus, terminology. Once you move from the theory
> to the implementation, everything else can change, as well. And as
> things are changed on the working side of it, perhaps the connection
> with the theory grows more tenuous. If that poses problems, then that
> can be stated.
> If it both solves and causes problems, so too. Of that,
> the use of self-referential tables is pretty much the example of the
The use of self referential tables is explicitly not ruled out in the 1970
and such use of Connect By, or similar.
I don't wish to get involved in the religious wars over the use of "Connect by". I clearly recognize it as an Oracle specific construct. But I don't hesitate to use it whenever I think it convenient. The nested set model is also very clever, and very useful. I don't hesitate to use it when I think it's convenient. Received on Sat Feb 25 2006 - 21:54:00 CET