Re: A Question on Integrety

From: Marshall Spight <>
Date: Thu, 18 Dec 2003 19:45:59 GMT
Message-ID: <XtnEb.81010$8y1.287268_at_attbi_s52>

"Dawn M. Wolthuis" <> wrote in message news:brsepd$hr3$
> I am not opposed to relational theory, but it is not the entire picture.

It's *an* entire picture.

> There is surely more data stored in non-RDBMS's than in them -- Excel
> spreadsheets, most web pages, and much more.

I'm sorry, but I don't understand the point of this comment.

> Also, many folks working at
> the cutting edge of software development are storing data in structures more

Yes, but I would claim the REASON is that they don't know relational theory.

> Pre-RDBMS and Post-RDBS structures will
> look more similar than the RDBS advocates might like, I'm guessing.

As best I can figure, this statement is saying that circa-1960's hierarchical databases will look like circa-2000 XML hierarchical databases. I would agree.

> But, if this comp.databases.theory newsgroup is just for relational theory,

Pretty much, yeah.

> I'll find other forums for my "stupid" opinions about how storing data in
> di-graphs (at.least logically) is more intuitive and easier to set up and
> manage.

Well it's certainly going to be harder to navigate. But if you want a really nice model for a digraph, I suggest using a relation. One attribute of the relational theory is that data access is content-based, and "navigational" information is the same as content data; there is no distinction. Thus you require only one set of operations to for both structural or relationship info (eg. edges) and non-structural info (eg. nodes.) Both can be queried and manipulated declaratively, by content. I have not seen any other data theory that can manage that trick.

I'm also not clear on how you evaluate how "intuitive" a data theory is. I think once one has been working in software a while, on gets the idea that what one finds easy to understand (after ten years of hard study) is generally easy to understand. One forgets that in the beginning, one had to work hard to understand recursion, pointers, or even simple interation.

> It is the model for the data in the WWW, which happens to be a very
> successful model in spite of broken links (lack of referential integrity).

Also very successful: the x86 architecture, the Chinese language, MacDonald's food. That doesn't mean those things are well designed or examples that we should all strive to emulate.

> And of course, there are many more opinions where that came from.

Here I agree with you.

Marshall Received on Thu Dec 18 2003 - 20:45:59 CET

Original text of this message