Re: Stored fields ordered left to right
Date: Sat, 27 Dec 2003 10:21:03 -0600
"Marshall Spight" <mspight_at_dnai.com> wrote in message
> "Dawn M. Wolthuis" <dwolt_at_tincat-group.com> wrote in message
> It doesn't affect the semantics of relations or relational operators;
> it just affects how attributes are identified.
OK. Then a model in which there are relations where the tuples are ordered but where values can be retrieved either by name or by position would be different from a relational model where they are not accessible by position. Then Date's point is not that the MultiValue model is not made up of mathematical relations, but that it is a different model based on relations than what Codd and company call the relational model.
Then we are in agreement. Now I just need words to make it clear that a particular model is based on mathematical relations (ORDERED tuples) even though it is not based on all of the other rules imposed within a theory that names itself the "relational theory" (even though it is based on not perceiving the relation's tuples as ordered).
The developers of MultiValue databases (such as IBM) promote them as relational databases and they are mathematically correct. These databases stem from the Nelson-Pick data model, however and not from the work of Codd. If any company is permitted to decide what is relational and what is not, it would be IBM, I would think, so I'll figure they have as much a right as Date does to define what is and is not relational.
So, it seems to me we have a problem with our vocabularly. One group of database theorists have taken a mathematical theory -- that of relations -- and have named their theory after it, even though they do not stick to the mathematical definition and they extend the mathematics with many other rules. They then tell people working with other models that are equally mathematically relational that they do not conform to relational database theory.
To try to straighten this out, I have referred to "the relational database theory" instead as the RDBMS theory or the SQL database theory, but because those are both implementation-based terms they do not sit well with "relational databsae theorists". Is there a term for the "relational database theory" that we (at least I) could use to indicate that it is the relational database theory that does not include all databases that conform to the mathematical theory of relations?
I want to be able to agree wtih IBM that both DB2 and U2 are relational databases (since I DO agree with them) and also agree with Date that the U2 databases are not based on his version of a "relational database theory". Thanks in advance for any help with this vocabulary issue. --dawn Received on Sat Dec 27 2003 - 17:21:03 CET