Re: Data Display & Modeling
Date: Sat, 08 May 2004 02:39:43 +0200
Message-ID: <409c2c51$0$567$e4fe514c_at_news.xs4all.nl>
Dawn M. Wolthuis wrote:
> mAsterdam wrote:
>>Dawn M. Wolthuis wrote:
>>>Let's say that someone likes to model data as functions [e.g.
>>>PERSON(key)=tuple ], in non-1NF and in di-graphs (so you can
>>>navigate your way through it), with a target
>>>implementation in XML or PICK. I have heard
>>>people say that is a good way to "view" the data,
>>>but we need to model it using relational theory.
>>>
>>>Given:
>>>
>>>1. that there is a good non-relational way to view the data, such as one
>>>using di-graphs with functions on strings.
>>
>>Both you and Leandro ask of a model to represent all data.
>>Leandro rejects models that don't, you
>>asume they do when they don't.
> > So your point is that the relational model is > just one such model and it is > useful to use for some sets of data and not for others?
No, that is/was not my point (but I'll answer to it later anyway). First I'll try to bring home the point - or what's left of it.
Models have limitations.
Valid conclusions drawn from observations of a model
do not always lead to valid conclusions about what
that model is modelling.
I got the impression that both you and Leandro had a
similar take on this, both attributing more expressional
power to models in general than I would.
However, Leandro rejected part of my statement and deleted the rest. I don't know what to think of that. Maybe he has no use for other models than a logical model with constraints and relations, maybe he just doesn't want to go into what constitutes a model, and what it's limitations are. Maybe he doesn't like me. So, whatever the point was, your question here has more relevance.
Back to your question.
Do I think relational just one such model?
No, I think it is special.
Not in the sense of the one true way of looking at
data, but special nevertheless.
Throughout history, several models of bookkeeping have been used. Somehow medieval Genovan double-entry bookkeeping is a good fit when several people, not necessarily trusting eachother, have to closely cooperate registering values, both stock and flow. While definitely more expensive (labour intensive) than other (nevertheless surviving) models, it is widely used and tutored more than all others. Evidently preachers of a 'the one true way of looking at data' have a role in this.
I think relational modeling is in the process of conquering a similar place. If your organization has valuable data to share and care for, you will apply the discipline of relational modeling - and you will do a lot more than that.
While I'm at it:
One of the success factors of the Italian way of
bookkeeping must have been the use of arabic
numerals instead of roman numerals.
I'ld parallell this to the typesystems or should I say typemodel? The typesystems currently in use are hopelessly inadequate. Nobody wants to know a CHAR.
> If I understand you correctly, what are the conditions under which the > relational model yields a better bang for the buck solution for a company > (that doesn't just mean cheap -- it includes data quality over time)?
That's easier asked than answered. I can't. Just some thoughts.
Let's assume, for now, that with model R we can model all user data without loss, and with model X (or E, O, U or even P) we lose some(thing) of these user data. What exactly is that loss? How do we value it? bucks(R) vs bucks(X) would be another relevant comparison. Maybe other ways of modeling excel in the "more than that" (see above) area. Similar comparisons could be made there.
Grrmpf_at_#$! I didn't squeeze 'orthogonal' in somewhere. I won't win at buzzwordbingo. Received on Sat May 08 2004 - 02:39:43 CEST