Re: Is this bad design ?

SPeacock <Xpeacock_at_pacbell.net> wrote in message news:<405A4364.F27A24_at_pacbell.net>...
> A few observations....
>
> 1st, People have relations to other people with a 'type' of relation. (type is
> e.g., spouse and child -- the reverse is the parent/grandparent...etc). Then to
> show legitimacy, the father and the mother would both have a 'ownership'
> relation with a child. Guaranteed that a child needs a ma and pa - at least for
> the near foreseeable future . A 'valid' mother would be one with a 'valid'
> child and father (except for a few rare biblical references).
>
> 2nd, Teachers don't teach children (people). Teachers conduct a class and
> people attend the class. It is only happenstance that the person 'learns' or
> that the teacher 'teaches'. There are few occasions where people 'own' people
> (that went out of style some time ago). But there currently exists special data
> relationships between physicians and patients (people of type patient) and
> perhaps parole officers and people (parolees). At one time the teacher, being a
> mentor, allowed a closer teacher/student relationship. But alas, no longer.
>
> 3rd, the 1st is a bit difficult to do easily or in a straight forward manner
> with an RDMS. This then leads to why not a non RDMS answer if this is a
> newsgroup of 'theory'.

Exactly! I did either read or guess that Ben wanted to do this implementation in an RDBMS and if this were a relational database implementation list then I would not have suggested the non-1NF model that I did. However, from what I have read so far, the two viable models that are accepted as good theory within their realm are Fabian's suggestion for an RDBMS and mine for an XML or PICK implementation.

I'm very happy to discuss theory towards some purpose, but when one model falls short (in this case the relational model) it makes sense to look at other models. If there is a good model that your db provider doesn't handle, then put pressure on them to do so and modify the modeling theories.

If no one had done that, then Date would not have gotten to the point of accepting relation-value attributes, for example. Eventually the relational model might even evolve to where it makes sense to use it ;-)
smiles. --dawn Received on Fri Mar 19 2004 - 16:57:15 CET