Re: Navigation vs Relational operators

"Jan Hidders" wrote in message
news:n11Fc.171714$4P6.8468548_at_phobos.telenet-ops.be...
> D Guntermann wrote:
> >
> > Relational is still navigational, but logically. The relational model
frees
> > us from physical navigation and somewhat from logical navigation. To
> > further provide users the capacity to have even greater freedom from
logical
> > navigation (e.g. specifying an attribute independent of knowledge and
> > application of table names, relationships, and joins), Ullman and others
> > (Fagin,Maier and Valdi) proposed and explored the concept of the
Universal
> > Relation. There was quite an interesting and uncharacteristically blunt
> > public debate between proponents of the universal relaton assumption and
> > William Kent.
> >
> > Refererences:

[snip]
>
> Probably nobody's interested in this anymore, but I just found a nice
> short and accessible on-line presentation of the issues and positions:
>
> http://www-db.stanford.edu/jdu-symposium/talks/mendelzon.pdf
>
> Nice stuff, especially if you know a little the persons involved. If
> only there were heated debates over *this* subject in this newsgroup...
:-)
>
> -- Jan Hidders

Well, I'm interested for one. Thanks for the link Jan.

One way this is of interest is to throw some theoretical light on the way that some query tools (I'm thinking Cognos, Business Objects - that kind of thing), attempt logical data independence. Here end users do not generally specify joins, rather some designer does.

Quote from http://citeseer.ist.psu.edu/vardi88universalrelation.html

    "In a universal-relational system, you must settle for eliminating the need for logical navigation along certain paths - those selected by the designer - while letting the user navigate explicitly in more convoluted ways"

The tools I mentioned above fail (in my opinion) on at least that last matter. They do not provide an easy fall back position for more convoluted queries.

Personally, (I think) I much prefer tools that make logical navigation visible to the users, so that you don't get a discontinuity between 'simple' queries that can be done without logical navigation, and those that require it. In other words, I'm probably with Codd :- "the universal-relation model fails completely as an alternative to the relational model."

P.S. the gzip'ed pdf from off the link above is actually a plain pdf, so just rename the extension to view the article.

Regards
Paul Vernon
Business Intelligence, IBM Global Services Received on Fri Jul 02 2004 - 13:29:56 CEST