Re: Latest version of glossary

"JOG" <jog_at_cs.nott.ac.uk> wrote in message news:1140842075.074019.209510_at_u72g2000cwu.googlegroups.com...
> dawn wrote:
> > JOG wrote:
> > > However if forced
> > > I would probably refer to however the term "entity" as defined in E-R
> > > modelling.
> >
> > What would that definition be?
>
> I assumed that given the widespread use of E/R modelling there would
> indeed be a good, easily accessible definition of the term. A quick
> search on google has highlighted to me that this was no the case - my
> mistake. Undeterred I dug up the original paper. In "The
> Entity-Relationship Model-Toward a Unified View of Data", 1976, P.
> Chen., his definition is:
>
> "An entity is a 'thing' which can be distinctly identified. A specific
> person, company, or event is an example of an entity. "
>
> Awful. Note the apologetic quotes around "thing" - he is neither
> introducing a new term or quoting someone, so its awful use of English.

I read exactly the same paper you did, and I had a very different response. I found the lack of formality in the ER to be refreshing, rather than awful.

Maybe that's because it blends well with my experience at the data analysis stage of projects.

At that stage, most people have a poorly formed notion of what they are talking about. Even the subject matter expert is at a loss to understand why data has to be discussed in such formal detail, when he's got the stuff in his mind? The technical people in the room almost always understimate the complexity of the necessary model to make the analysis comprehensive enough.

An analysis that doesn't dig deep enough results in a model that has to be extended too many times during the project, and that in turn has a devastating effect on "bang for the buck".

The fact that "entity" is so poorly defined is related to the way we think. All of us tend to analyze the universe of discourse into parts. Various religions, at one time or another, have commented on the essential unity and indivisibility of the universe. But they don't have to build databases. We do.

But the analysis of the universe of discourse into entities is SUBJECTIVE regardless of how much
effort we take to make it seem objective.

A less formal approach at the analysis stage allows for more fuzzy thinking at that stage. IMO, that's a good thing. There will be time to tighten up later, when we convert the ER model to a relational model (or a codasyl model or a hierachical model, or an MV model or an object model). Sloppiness in the analysis will be uncovered and resolved at that point. And that point is still before coding has begun.

I have a great example from a project long ago, but I'll save it for another topic. Received on Sat Feb 25 2006 - 15:40:19 CET