Re: The Practical Benefits of the Relational Model

> I have been prompted to wonder if relvar names might not best be seen as
> attribute name prefixs (or postfixs etc) for certain *marked* attributes
in a
> relvar. A key problem with relvar names is that they don't really have
any
> place in predicate sentances.

In a textbook I read/studied sometime back, the author discussed basic concepts of the Universe of Discourse (UoD) and the breakdown of the Universal Relation into normalized relation schemas within the scope of a single defined "database". If I remember correctly, he addressed the issue of attribute naming and the fact that in some cases, attributes with differing semantics (and context) might have the same name. In such cases, the author mentioned that some researchers and theoreticians advocated the mangling of attribute names to include and/or concatenate, implicitly, the name of the relvar so that the attribute names would be distinguishable, thereby resolving such conflicts during the normalization and predicate resolution process.

I could be wrong, but I believe I read this in Elmasri and Navathe's textbook, "Fundementals of Database Systems, 3rd Edition. I will double-check to verify.

If I understand you correctly, you seem to advocate the same approach. The major difference being that your arguments stem from discussions concerning Date's principle of Orthoganal Design.

>This is particually interesting in the context
> of relvar projection.
>
Please explain.

Thank-you,

Daniel Guntermann Received on Tue Nov 05 2002 - 03:24:57 CET