Re: Is supertyping orthadox?

Kristian Damm Jensen wrote:
> JRStern wrote:
> >
> > On Tue, 20 Mar 2001 12:28:04 -0000, "Brett Gerhardi"
> > <brett.gerhardi_at_trinite.co.uk> wrote:
> > >
> > >The solution they came up with was what they termed as a 'generic
> > >table structure' where basically there is a 'schema' table that
> > >stores the Field name, data type, length, (also has a foreign key
> > >to a 'tables' table to group the schemas together). Also there is
> > >a 'values' table that has a foreign key to the 'schema' table that
> > >in the implentation would have a sql_variant field that stored the
> > >data.
> >
> > See if these guys can even recite the definitions of the first three
> > normal forms. They can't. Fire them.
>
> Agreed. What they propose sounds like a Pascal variant-record or a
> C-union. It has no place in relational design.

Not agreed. Yes, it is inefficient. Yes, it makes specifying and maintaining database constraints very difficult if not impossible. But it is a nice solution if you want a very flexible schema that allows you to record extra properties without changing the schema.

So, as usual, it is a solution with pro's and con's and it depends upon the situation if it is appropriate or not. Why you say that it has not place in relational design is beyond me. I don't see anything unrelational about it.

-- 
  Jan Hidders