Re: 1GB Tables as Classes, or Tables as Types, and all that refuted
Date: Fri, 19 Nov 2004 21:44:44 GMT
Message-ID: <gPtnd.29679$uv3.1321090_at_phobos.telenet-ops.be>
Paul wrote:
> Jan Hidders wrote:
>
>> You are so right that it's almost boring. :-) If there had been any >> real logical problems with letting relational variables play the role >> of types then it would have not been possible or very difficult to >> come up with a decent formal data model for that. It wasn't. QED
>
> Isn't Date's "first great blunder" to be seen more as a rule of thumb or
> a software engineering principle rather than as a formal mathematical
> truth?
Very well said. But (as Ja Lar already said) then Date should not claim that there are logical problems but that there are engineering problems, and he should not talk about logical arguments but about engineering arguments.
> i.e. although it is possible to map classes from an object-oriented
> system onto relations in an RDBMS in certain circumstances, as a general
> rule it's best avoided?
Now you make it sound as a design-rule for designing your relational schema, but that's not really what it is about. It is about what the meta-model of the logical model of your DBMS should be, i.e., if you want to build a new type of DBMS and replace the relational model with another type of data model then what should that model look like?
Does that make sense to you?
- Jan Hidders