Re: Object oriented database

From: paul c <>
Date: Sun, 02 Nov 2008 19:12:14 GMT
Message-ID: <iYmPk.5714$%%2.437_at_edtnps82>

JOG wrote:
> I have had the opportunity to interact with numerous top-notch
> software engineers - people who have, out of necessity, worked for
> decades now with OO, on large scale projects. The interesting thing is
> that they all seem to develop best practices of using those tools in
> ways that mimic policies that are core to RM (without even realizing
> the correlation). They focus on interfaces and storing underlying data
> as neutrally as possible. They do not rely solely on OID's, and use
> constant identifiers to access objects logically. They decompose data
> organization in ways that echo normalization. And they use invariants
> to mimic predicate constraints.
> These are the people who are worth discussing data models with
> because, well, they realise OO isn't one, and they have built another
> on top as best they could within its constraints.

I'd like to think that's possible but I doubt it. Regarding "out of necessity", IMHO merely accepting the problem and then ducking it isn't enough for progress in general. I can sympathize with anybody who is forced to try to make do after being told by some disinterested CEO to "make it work", but where is the progress in that? We can only get so far with post mortems of systems that are based on code techniques that aren't based on any coherent, logical model.

Regarding "best practices", to some of us, the very term has come to connote desperate compromises when faced with basically futile circumstances.

The fact is, data programming is a field in itself. It doubt if OO fans can progress until they recognize that much of the problem has nothing to do with integrating their favourite imperative programming language.   I suspect that some number of the rest of us would agree this was perhaps Codd's underlying point, that the same few operators and principles can suit different applications. (Personally, I'm still not convinced that Codd was wrong to espouse specialty data "sub-languages" - not sure where I read that, maybe something Date wrote.)

(BTW, regarding OID's, system-generated partial keys always made some sense to me, if only to avoid RVA's.) Received on Sun Nov 02 2008 - 20:12:14 CET

Original text of this message