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

Hugo Kornelis wrote:
...
> Functional dependencies stem from reality. Whether or not you choose to
> include B in your model does not change the situation where, apparently,
> C depends on A through some intermediary B (that is not in the DB).
>
> In a DB that stores PersonID and EyeColour, one might argue that the
> actual dependency goes back to the parents of the person and their
> genetic patterns - but those will typically not be stored, and yet the
> EyeColour still depends on PersonID.
> ...

Assuming you're saying it's improper to depend on any notion of absolute reality, I think I'd agree. Doesn't a db aimed toward aiding some present function necessarily stand for a very fractional/partial (or even distorted) reality? Eg., if it's not fractional it's probably unwieldy and untoward. Seems to me that the EyeColour dependency hints at this - when the purpose of a particular set of tables isn't concerned with dna, one likely ignores blood lines. Further, I suspect that no db ought to introduce fd's that aren't patently implicit in the user's requirements/biz rules/application intent. In the USA, I gather that an address that is complete enough for a mailman to deliver to, along with a city and a state will determine zipcode, yet I suspect there are many tables in non-postal db's that have a column set such as (Customer, unit, streetaddress, city, state, zip). Received on Wed Oct 29 2008 - 23:59:25 CET