Oracle FAQ Your Portal to the Oracle Knowledge Grid

Home -> Community -> Usenet -> comp.databases.theory -> NIAM (was: Just one more anecdote)

NIAM (was: Just one more anecdote)

From: mAsterdam <>
Date: Fri, 12 Aug 2005 02:33:14 +0200
Message-ID: <42fbee0a$0$11072$>

Hugo Kornelis wrote:

> mAsterdam wrote:

>>Hugo Kornelis wrote:
>>>... the symbols used in NIAM and ORM are so much alike that they are
>>>almost interchangeble. The main difference is in the methods described
>>>in NIAM to make sure that all relevant questions are asked in an
>>>interview, that all answers are correct and that therefore the model is
>>>correct and complete. Unfortunately, that is exactly the part of NIAM
>>>that I've never seen described anywhere on the 'net...
>>Maybe a little explanation from an old Nijssen-fan helps here.
>>The closest thing I read is in english is in Date's
>>material on external predicates. One could think
>>of the interviews as investing in getting the
>>external predicates right, like so:
>>check example tables against real business data and
>>busines cases (not unlike UML use cases) with domain experts.
>>They are called "population diagrams: the table heading (including the
>>predicates) would be called the "intension", the body (including the
>>expansion of the values - by means of the predicate - into the
>>propositions) would be called the "extension".
>>I've used it in practise, some of the systems built on these
>>analyses are still in heavy use.
>>Well - I'm not going to write the book I'ld like to see here,
>>but this is the core of the NIAM interviews you are referring
>>to as I understood them.
> ...Yes, it is - combined with a complete set of
> prescriptions (is that the correct English word?) that will
> tell you exactly which combinations of values in the extension 
> you'll have to present to the domain expert, and
> what conclusion to draw from the expert's answer.

I was so lucky to have been attending a session where professor Nijssen convincingly argued that the result would necesarily be a set of relations in 5NF without ever mentioning normalization in the process which lead to the set.

I wonder: if we would introduce intervals as first class scalars would we wind up with Date, Darwen & Lorentzos' 6NF without the hassle? I gave it some thought. I think so, but I can't prove it.

> It's too bad that Universele Informatiekunde (Nijssen, 1993; published
> by PNA Publishing BV) was never translated. Despite the many errors it
> admittedly has, it's still a very good book.

Is it still available in dutch? Received on Thu Aug 11 2005 - 19:33:14 CDT

Original text of this message