Re: The IDS, the EDS and the DBMS

"pstnotpd" <schoenmakers_at_tpd.tno.nl> wrote in message news:ci411b$2sg$1_at_voyager.news.surf.net...
> Laconic2 wrote:
> I mean some sort of engine which tries to solve algebraically. The polar
> to carthesian conversion mentioned by mAmsterdam would not result in
> information loss. Something like square root of 2 or pi stored as such.
>
> I suppose this would not perform very well, but I was just wondering if
> anybody has ever tried it.

I still don't get it. Sorry to be so difficult.

Here's a stab in the dark. Don't know if it's relevant to what you are asking or not.

When Ed Codd wrote his initial paper proposing the use of the relational model of data (RDM) in a database management system (DBMS), he made reference to the general area of work where the RDM had previously been used. These were
deductive systems. Their primary goal was not storage and retrieval as such, or even data integration but something I'll call "inference". I don't know anything more about such applications, but the fact that Codd refers to them must be significant.

The other area of research where "algebra" as such was relevant to the processing of data was artificial intelligence (AI). There were some AI projects in the 1950s and 1960s that stored expressions as expressions, rather than computing the result and storing it. In "Computers and Thought"(1963) by Feldman and Feigenbaum, there is a section devoted to systems that prove mathematical theorems. A lot of this work was done suing vacuum tube computers, but it was still very impressive work. As you suggest, there is a performance trade off here.

Mathematical theorem proving, game playing, and optimization are all kinda linked together. But then, everything is related to everything else, isn't it.

Turning back to the mundane world of the DBMS, there is a section of a modern DBMS that ties into all this: it's the query optimizer. A really good cost based optimizer (CBO) is essentially a game playing machine. Of all the available logically equivalent strategies, which one is, in this situation, the one we "ought to use"?

The generation of multiple logically equivalent strategies is, in essence, algebra. I don't know if you want to call this a "math engine" or not.

Sorry, but I still don't get it. I must be getting dull in my old age. Received on Mon Sep 13 2004 - 15:57:00 CEST