Re: Examples of SQL anomalies?
Date: Sun, 29 Jun 2008 19:55:12 -0700 (PDT)
On Jun 28, 8:48 pm, Rob <rmpsf..._at_gmail.com> wrote:
> On Jun 26, 4:01 pm, Gene Wirchenko <ge..._at_ocis.net> wrote:> Let me add another item.
> > Joins will often give trouble with ambiguity of names. Never
> > mind that I am joining on the two columns having the same value. If
> > they have the same name, I have to prefix them with the table name.
> > Sincerely,
> > Gene Wirchenko
> On January first of this year, I pointed cdt to
> in which I described the Aggregate-Link schema, a new way to represent
> relationships in relational databases.
> I freely admit that I was intentionally vague about what one could do
> with it, but you (Gene Wirchenko) have given me a lead-in I cannot
> ignore because it leads in exactly to one important value of the
> Aggregate-Link schema: The separation of structure from data in
> relational databases.
Yes and it was twaddle then, as it is twaddle now. Foreign key references provide all the "structure" one would need. Please don't see this as a request for a debate. I am merely pointing out that the OP should not waste his time trying to decipher your sales-pitch. If you have no useful answers to a genuine question on the OP's part then it should not be seen as another opportunity to sell your chocolate teapots.
> If you look about 3/7ths down the page under "Aggregate-Link Structure
> Isolation", you'll see the following quote:
> "Beginning with the introduction of the Relational Model in 1970 [Codd
> all approaches to relational system design have been entity oriented
> -- that
> is, oriented to the representation of entities. Relationships are
> unnamed and
> for the most part, invisible. Yet relationships provide the structure
> associates tuples in one relation to tuples in another. The join
> is the principal means for retrieving related tuples and their
> attribute data.
> Relationships have never been treated as first-class components."
> I was genuinely hoping that someone would pick up the gauntlet and run
> it, but nobody has. People are busy, they have plenty of their own
> My expectations were unrealistic.
> I am working on the next iteration of the website, one that will
> several values in the Aggregate-Link schema, but it is taking alot
> time than I expected and won't be ready much before September'08. In
> meantime, consider the following observation:
> If a relationship were a first-class object in SQL, one would
> have to specify all of the following when declaring a named
> a. the relationship name and cardinality,
> b. 2 relations (I call them "parent" and "child" because the
> relationships are asymmetric),
> c. the join "columns" (I use the terms "primary key" and "foreign
> key"), and
> d. whether childless parents and parentless childs (orphans) are to be
> part of the relationship.
> As a consequence, by specifying a relationship by name in a query (in
> or some other language), the problem you stated ("ambiguity of names")
> disappear. In fact, the join as an explicit operator would practically
> More than one relationship between the parent- and child relations?
> - Not ambiguous because different relationships have different names.
> Multiple relationships (i.e., more than 1 join)?
> - Composition.
> Intersection, union, relative difference between relationships?
> - Think. Can't these be resolved entirely from the relationship
> without access to the entities?
> I cannot get involved in lengthy discussions here -- I'm overwhelmed
> with the effort to create the next iteration of the website
> I am interested to see if anyone else finds any merit in the
Received on Sun Jun 29 2008 - 21:55:12 CDT