Re: foreign key constraint versus referential integrity constraint

Sampo Syreeni wrote:
...
> I think the later talk about domains and e-relations in RM/T (and to a
> lesser extent in RM/V2) makes that a bit easier to understand. As I
> interpret it, Codd started with a very clean and purely syntactic
> model in his seminal article, which will definitely lead to pure
> symmetry in access. But from the very start he also worried deeply
> about the semantics and making them explicit as metadata, which leads
> to the second influential discovery of his: the integral, reflective
> data dictionary. Later on his development of the relational model not
> only added surrounding functionality (which weighs in so heavily in RM/
> V2, what with all of'em T-joins and the lot) but semantics as well.
> ...

Regarding semantics, far be it from me to try to describe Codd's motives. While I think it's certain he later on veered from his original principles, I think no one will ever know if there were personal or more general reasons that drove him to the so-called "semantic model", I suspect it was because of various forces that we are all subject to from time to time. For all I know he might have been bored with the lack of uptake of his ideas and thought a little speculation might stimulate progress. Unfortunately most people, who pretended to read his stuff, including apparently the original system R implementors, chose to fasten on one sentence or paragraph, sometimes taking that as the pre-eminent gospel and ignoring the rest. What he wrote is either full of caveats between the lines or extremely case-based from the get-go. Pretty much every quote of Codd's I see is out-of-context. Received on Thu Oct 22 2009 - 02:47:28 CEST