Re: By The Dawn's Normal Light

> I claim that it's not useful to have one definition for "relation"
for
> mathematics and an incompatible one for IT. On this point I agree
with
> Dawn.

Definitions are difficult to change, though, once they have history behind them. I don't see the mathematical definition interfering with discussions of relational databases, although I'm not a mathematician.

> And it's clear to me that the mathematical definition of "relation"
> does not force the values in the tuples to be atomic.

Nor does relational, for most common usages of the term "atomic." Earlier definitions of 1NF are flawed because the word "atomic" is meaningless.

> And it's clear to me that the definition I always learned for 1NF
does
> force the values in the tuples to be atomic.

That definition was flawed.

> So, if someone wants to adjust the definition of 1NF so that it no
longer
> requires atomic values, I would prefer that they invent a new term,
like
> "Date-Darwen Normal form".

But since the old definition is useless, why not appropriate it? But granted that using the term "relation" implies a correct "version" of the definition of 1NF, so it's somewhat pointless to quibble.

> If someone wants to change the definition of "Relation", I would
prefer
> that they invent a new term, like "normal relation".
> Then they can say that a "normal relation is in 1NF (and may also be
> in other normal forms)".

I don't understand why we'd do this, since the term is so entrenched. Why not take "relation" as implying 1NF?

> If we did this, we could talk about "relations" as such, and talk
about
> "normal relations" when we need to.

Are you saying the mathematical definition of relation has a large bearing on databases topics?

• erk