Re: Database design, Keys and some other things

"JOG" <jog_at_cs.nott.ac.uk> wrote in message news:1128029620.610189.274300_at_f14g2000cwb.googlegroups.com...

> Proposition:
> I am putting forward for discussion that Codd's Information Principle
> (based on an Liebniz/essentialist viewpoint that an entity is defined
> by its attributes, and is nothing more and nothing less than the
> aggregate of them, [Celko - D&DB]) is not a complete model of the
> nature of information. Rather I am proposing that there is no single
> attribute of an entity that can be gauranteed to act as a permenant
> immutable reference for it (nominalism - there are no universals), and
> so an external one to represent our concept of an entity be provided.

The usual answer to this is: there must be some way to discriminate among any two entities, otherwise how do you know there are two of them.

Immutability in a changing world and absolute guaranties are another matter. :-)

> Difficulties arising from the IP:
> Primary Key non-mutability - often necessary to employ internal
> surrogates that require management [Codd - RM/T paper].
> Lack of Internal Referencing checks - often necessary to explicitly use
> cascade mechanisms to prevent dangling references, orphan records.

It is hard to keep a database up to date. Do you have a solution (even partial) to this problem ?

> Possible Solution:
> provide any encoded proposition with an automated reference that
> guarantees primary key mutability, but is externalised from the data,
> which then remains a pure encoding of the representation of real-world
> attributes (essentially the partial map that is the tuple becomes a
> datatype) [W, Kent - D&R]. Joins may be pivoted on this reference, and
> integrity is continually maintained by the DBMS between these
> references to prevent orphan tuples.

This is an encrypted text for me. Received on Fri Sep 30 2005 - 10:49:53 CEST