Re: Another view on analysis and ER

From: paul c <>
Date: Wed, 05 Dec 2007 17:56:59 GMT
Message-ID: <LDB5j.1379$sg.700_at_pd7urf1no>

JOG wrote:
> On Dec 5, 6:09 am, Ruud de Koter <> wrote:

>> There are two troublesome points in your reaction. First of all 'keep
>> the data neutral' doesn't mean no choices are made. Staying neutral is a
>> choice as well. One of the hardest choices I 'd say, because in order to
>> stay neutral, a thorough knowledge of the universes of discourse is
>> necessary. Also, these universes should not be mutually exclusive.

> Thats a fair point. Neutrality is something i've promoted for a long
> time on cdt, and I understand there are issues for processor cycles.
> However one can still take a single conceptual view of data in
> analysis and flatten it out in the logical layer. Take David's
> breakdown of the marriage example for instance - even though it is
> translated from a single conceptual view, by the time it is in the
> logical layer data may be extracted from it via the perspective of
> marriage as an entity, or marriage as a relationship, with equal ease.
>> A second point: we can only keep the data neutral if know all possible
>> perspectives. It is only then that we can consciously model the data to
>> fit all the universes of discourse. Yet, you rightly observe we don't
>> necessarily know all the applications, which amounts to saying we don't
>> know all the universes of discourse. So choices can not be avoided. In
>> that case I 'd much rather make these conscious choices instead of
>> keeping up a pretense of neutrality. At the very least we should be
>> aware that the model resulting from analysis may be biased, and is not
>> the final word on the world out there.

> I think maybe we are referring to a slightly different definition of
> neutrality. I'm suggesting that a logical model should have no bias as
> to whether things are relationships or entities, and leave that to be
> determined by the person generating the queries. Regards, J.

In practice, I think neutrality is fickle, it comes and goes. Surely one useful feature of a decent rdbms would be to give every app the illusion that the db is neutral as far as it is concerned. Received on Wed Dec 05 2007 - 18:56:59 CET

Original text of this message