Re: what are keys and surrogates?

From: rpost <rpost_at_pcwin518.campus.tue.nl>
Date: Tue, 12 Feb 2008 21:12:21 +0100
Message-ID: <71da0$47b1fda5$839b4533$31219_at_news1.tudelft.nl>


David Cressey replied to me:

>> What I was trying to get at is that calling an entity a "thing" doesn't
>> add any explanatory value to me. Worse, it makes things *less* clear,
>> since many entities aren't actually "things". I find it much more
>> clarifying to say an entity is something used as a reference.
>
>The semantic difference between "thing" and "something" is very unclear to
>me.
>
>As long as we are hell bent on introducing new terminology, and retiring the
>word "entity",

Please note that I do not retire the term "entity", nor the term "something" (it's just a placeholder). I do introduce the term "reference".

>Use the word "subject" instead of the words "entity" or "thing" or (shudder)
>"object".
>This way, you could say that the subject matter is described in terms of
>"subjects" and "predicates". The word "predicates" has come into several
>discussions here. And I think you could usefully use the word "predicate"
>to describe relationships (or assertions about relationships) at the
>conceptual level.

Yes, this is the basis for both relational modelling and E/R modelling.

>What I'm uncertain about is whether the terms "subject" and "predicate"
>would mean about the same things that they do in the world of grammar. If
>not, it's an argument against using this terminology, but not necessarily
>a conclusive argument.

This idea is the basis for relational modelling, while E/R modelling departs from the idea that predicates usually define relationships between the subject (an entity) and one or more objects; as NIAM says, it looks at the nouns (entities) and verbs (relationships). Both subjects and objects are typically nouns or noun phrases.

Of course the grammar of natural language is much more complex, e.g. whole phrases, een whole sentences can act as subjects or objects, but I must say, some theories of grammar apply the subject-predicate analysis all the way through (e.g. X-bar theory) and that appears to work pretty well.

But I don't think this tells us anything new.

What matters is whether such predicates are applied to any form of reference or object id. As far as I've seen, they definitely are. Predicate logic is applied to "individuals" all the time; "individuals" are object ids.

>I'd be perfectly happy with a conceptual level that says everything that
>needs to be said in terms of "subjects" and "predicates" rather than
>"entities" and "relationships".
>
>If this were adopted, the logical level could be described in terms of
>relvars and constraints, provided the RM is being used as the basis for the
>logical level.

Don't we already do this?

-- 
Reinier
Received on Tue Feb 12 2008 - 21:12:21 CET

Original text of this message