Re: Relations as Repeating Groups & Namespaces

"Dawn M. Wolthuis" <dwolt_at_tincat-group.com> wrote in message news:c9nh8r$i8f$1_at_news.netins.net...

>
> > For example they are usually one and the same on the 1 side
> > of a 1:n relationship, but not always so on the N side. The N side table
> > will take on any attributes of the relationship as well. For example:
> >
> > OWNER - OWNS - REGISTERD_VEHICLE follows the syntax:
> > entity - relationship - entity,
> >
> > or if you wish,
> >
> > relation - relationship - relation
> >
> > In this case, the business rules state that an owner may own one or more
> > vehicles, and for simplicity's sake, a vehicle can be owned by one and
> only
> > one owner. So the logical model (with some attributes) is:
> >
> > OWNER
> > driver_lic_nbr PK
> > driver_lic_state PK
> > name
> > etc
> >
> > REGISTERED_VEHICLE
> > vin PK
> > make
> > model
> > etc
> >
> > OWNS
> > purchase_date
> > lien_or_own
> >
> > Note that in this case, the relationship "OWNS" iteslf has attributes.
> This
> > is not normally the case, but it is not infrequent.
> >
> > So, the tables (physical model) map out as follows. This is done by
> > following established, published, proven rules to go to 3NF:
>
> proven? I think 1NF, on which 3NF is based is an axiom -- I can find not
> mathematical proof for it based on set theory or first-order logic axioms,
> for example. Can you point to the theorm/proof for this?
>

Proven to work, not proven by mathematical or formal logical proof. The algorithm employed transforms a model from the logical world to the physical. I personally don't know of any theorem/proof for this (not that this means there isn't one- I just don't know of any). I do know that the algorithm exists, that it works, and all cases up to ternary relationships (including weak entities) taking cardinality and particpation constraints into consideration have been accounted for (although there has been some argument as to how to handle ternary relationships). Again, I can refer you to Elmasri & Navathe 3rd Ed. This time, chapter 9, though my professor from my M.S.I.S. program states it more clearly in his course materials by using diagrams (unfortunately, not available in electronic form). Anyway, you would love chapter 9. In addition to providing and explaining the ER-to-Relational (more-or-less equates to logical-to-physical for puposes of this discussion, and no, I don't want to get into a discussion about logical vs. physical) mapping algorithm, it gets into superclass/subclass relationships, tuple relational calculus, existential and universal qualifiers, safe vs. unsafe expressions, domain relational calculus... And that's just chapter 9. There are 27 chapters and 4 appendices for your reading pleasure. After you've read this book and/or taken two or three semesters of graduate level relational database theory, we can talk. Note that I am a practitioner, not a theorist, which is why, as you may have noticed, I supply examples rather than enounce theories and discuss proofs. Having studied the theory (in exhausting detail), I am content that it is internally correct, and can be applied properly to most situations. Note that among scholars, there has been little in the way of major revisions to relational theory. There have been extensions and expansions (object-relational), some refinements (EER) as well as some argument as to implementation (SQL compliance, E.g.), but no one has said, "Wait a minute - there has been no definitive discussion. We need to clarify some things for everybody." Except by posters to this newsgroup. Like those who insist that data is code or functional dependencies are functions :) Sorry- didn't mean to open that can 'o worms again, but I couldn't resist a good elbow jab.

But seriously, if you have $90 to spare, you may want to winch up a copy of Elmasri (4th Ed out now). Or maybe hit a campus bookstore and peruse it. Bring coffee- there's not much of a plot, and character development is, well, not very interesting... Received on Thu Jun 03 2004 - 19:04:26 CEST