Re: foundations of relational theory?

From: Tony Gravagno <g6q3x9lu53001_at_sneakemail.com.invalid>
Date: Tue, 21 Oct 2003 18:06:42 -0700
Message-ID: <7phbpv8g1h7m9g4csckol25ukoebjftvsv_at_4ax.com>


In my mind the answer to your questions is that a developer shouldn't need to plan for every possible question that could be asked of his/her database. If we create an application where it's function is (now or perhaps in the future) to identify a parent based on the name of one of her children, then the database should be constructed to account for that eventuality. In reality any good application developer can predict the nature of most of the queries that will be asked of the database within a specific business context. That drasically limits the permutations, and allows the developer to create only those structures necessary to support the application. When we have needs to extend out of the box it's no problem to reformat data into new tables, or simply create extended definitions that point to existing data.

Yes, we can in fact identify a parent using the name of a child. A single attribute would contain the names of all children, separated by multivalued delimiters. We can do selects on the file for any parent with a child by a specific name, sort parents by child name with a single parent appearing multiple times in the report as in relational reports, and we can use controlling/dependent fields to pull correlated data like the birthdate of specific children - all of the data for a given parent and her children stored in a single record. If the children are to be considered "people" rather than as mere data units associated with a parent, then yes, we'd create a separate file, point to children from the parent, and have the children point back to the parent and perhaps siblings as well - but these relations aren't necessary unless there is a valid business case for it.

Tony

"D Guntermann" <guntermann_at_hotmail.com> wrote:
>Just for my own edification, are you saying that if I were to ask, what is
>the name of the youngest child of that women driving the white Mercedes,
>it's an invalid question because you can only ask, "what cars does that
>woman own and to which children is she the parent? "
>
>It seems somewhat hard to swallow that you would state that a relationship
>must be uni-directional. That is what you are saying, right?
>
>If the cop that wants to find that the crook with red hair and a tattoo on
>his right arm, driving a white Ford Mustang with a partially seen license
>plate of P11..... from Kentucky, should he never ask the questions because
>they don't match your real-world requirement? Questions such as "To whom
>does a white Ford Mustang with a Kentucky plate beginning with 'P11' belong
>to? And does that person have red hair and a tattoo? Or is he related to
>someone who does?
>
>You have me perplexed here.
>
>- Dan
>
Received on Wed Oct 22 2003 - 03:06:42 CEST

Original text of this message