Re: Proposal: 6NF

"Sampo Syreeni" <decoy_at_iki.fi> wrote in message news:Pine.SOL.4.62.0610051557450.18113_at_kruuna.helsinki.fi...
> On 2006-10-05, David Cressey wrote:
>
>> An outer join between two relational tables will, in general, produce a
>> table with nullable columns.
>
> I'd add that the same applies if we want the relational model to be closed
> under arbitrary unions.

I'm not sure I do. Persuade me.

> To me that seems somehow more elementary, and it also ties neatly with
> inheritance modelling: even if you represent entities of subtypes and
> supertypes by different relations, sometimes it's handier to work on their
> union.

Really? It might mean less typing on the keyboard, but "handier"? I always suspect an erroneous intention when I seem to need an outer join. Animal is a supertype of bird. Why should I expect to be able to ask how many eggs a camel lays?

Usually this kind of idiocy appears when using crappy tools that simply assume you're OK about outer joins and their consequent spray of nulls, instead of solving the problem properly. Report generators are a particular object of my loathing in this respect.

Roy Received on Thu Oct 05 2006 - 15:55:51 CEST