Re: We claim that delete anomality is due to table not being in 3NF, but...

Hugo Kornelis wrote:

> You're reading way more into my example than I intended. The point I
> tried to make is that functional dependencies are not determined by what
> is or is not stored in the database, but by how entities and their
> interactions in reality. The DB is a model of reality and changing the
> model won't change reality.

The DB is NOT a model of reality. It is a model of the testimony users give about the bit of reality that interests them. And for the purposes of understanding what the database needs to do, we don't even have to require that the testimony be correct or even given in good faith. (Which is just as well because we can't do that.)

So, bottom line, the content of a database is only very indirectly connected to the real world, and is always suspect.

Furthermore, the only thing we can insist on being able to rely upon is that the derivations we get from the content of the database are correct derivations from that content. To be assured of that we need to model the form of the testimony with proper fidelity. And that is what decides the appropriate functional dependencies.

It's just math, not metaphysics.

-- 
Roy