Re: domain questionnaire

vadim Tropashko wrote:
> In article <96n31l$1uf$1_at_news.tue.nl>, Jan Hidders says...
>
> >> 2. weight in lbs vs. weight in kgs
> >
> >The crucial question is again how you are going to denote them. Are you
> >going to write one kilogram as '1 kg' or just '1'. In the first case
> >you can have 1 ADT describing them both, in the second case you need
> >two.
>
> Then, definition of domain becomes dependent on the relation that
> uses it.

I have no idea what you mean here. You can use the same ADT in several relations. Why do you think it becomes dependent on the relation?

> This doesn't look like traditional interpretation of what
> domain is... Or this denotational idea is an additional element that
> we need to bring into relational theory?

No, it's a consequence of the interpretation that Chris Date gives to domains. The denotation is usually not discussed but it gives you a nice 'rule of thumb' to decide if it's one ADT or not.

> >> 3. US voice stationary phone numbers vs. fax numbers vs. cell numbers
> >
> >This is a subtle one. Date allows subdomains (or subtypes, or
> >subclasses, or whatever you wanna call them :-)). So, in some sense
> >these all belong to the domain 'phone number' but are in the same
> >time different subdomains. But these are not really good domains
> >because domains should be static, or, to be more precise, just as
> >dynamic as the rest of the database schema. And since phone numbers
> >can change their role quite fast, this should not be modeled with
> >(sub)domains.
>
> If unified phone domain is a good thing, shouldn't I squeeze email
> addresses into this domain as well? Where do I stop in my
> generalization effort?

Wherever it is appropriate in the universe of discourse that you are modelling. In principle there is no limit just as in OO modelling (that is in fact that we are doing here) there is no reason to stop generalizing, but you usually stop doing it if there is no longer a reason to generalize.

-- 
  Jan Hidders