Re: Something new for the New Year (2008).

From: Rob <>
Date: Tue, 15 Jan 2008 09:06:12 -0800 (PST)
Message-ID: <>

On Jan 12, 6:54 pm, Gene Wirchenko <> wrote:
> Rob <> wrote:
> [snip]
> >I don't know what you mean by "unworkable". Does that mean it is not
> >well-formed, or just that you have a different model?
> Practically everyone would have a different model. Consider
> "family unit". Most would consider that it means something like
> household. You are using a different definition. That would cuase
> problems.
> Now, consider some sticky stuff.
> Three years from now, Bob has graduated from school and and is no
> longer living with his father. What family unit is he a member of
> then?
> A few years later, he marries Allison, and they have two
> children, Samantha and Michael. What family unit is Bob a member of
> then?
> A few years after that, he and Allison divorce, and Allison gets
> custody of Samantha, and Bob gets custody of Michael. What family
> unit is Bob a member of then? What family unit is Samantha a member
> of? What family unit is Michael a member of?
> A few years later, Bob marries Lynn and they have a son Derek.
> About the same time, Allison marries Charles and they have a daughter
> Karen.
> Question: Derek and Karen are not related biologically and have
> never lived together. They might never have met. Are they siblings?
> Are they members of the same family unit?
> Allison and Charles die in a road accident. Bob get custody of
> Samantha. Due to other concerns, he also gets custody of Karen. What
> is the family unit setup then?
> I have not even added examples about adoptees.
> [snip]
> Sincerely,
> Gene Wirchenko
You are 100% correct. I totally agree, the model I proposed would not stand up. I said essentially the same thing in

(and was called to task for saying that 'there is no "clean, efficient way to store this information" in the relational model that satisfies all constraints').

However, the issue for me was not the model per se, only the use of a predicate construction that would allow mapping to the Aggregate-Link representation of relationships discussed here:

I should note also that nowhere in do I advocate the use of the A-L representation for conventional relational databases. Rather, I argue that the best indicated use for A-L is in the use of RDBs and RDBMSs as analytic tools. For example, to ask/answer questions /about/ other RDBs, like:

   Is a given key a candidate key?
   Is a relationship represented by a Junction Table actually 1-to-M?

And the biggie:

   Are these two database A and B the same? (whatever "same" means)

To use A-L for analytic purposes on other RDBs, it is important to know how A-L differs from the more traditional representations. That was the objective in setting forth my primitive model.

Although Marshall says: "I *still* don't know what you find interesting about it", I would argue that in addition to being the first, new relationship representation in 35 years (nested sets notwithstanding), it also demonstrates considerable analytic utility.

Rob Received on Tue Jan 15 2008 - 18:06:12 CET

Original text of this message