Re: Lucid statement of the MV vs RM position?

JOG wrote:
> With the PK { Animal, Colour }. If I Ungroup I will generate:
>
> Animal | Colour
> -----------------
> Fish Gold
> Cow Brown
> Cow Black
> Cow White
>
> But this is now not in GUNF because if I regroup I'll get:

Nitpick: It's the original relation (with the RVA) that is in GUNF or not, I think.

> Animal | Colour
> -----------------
> Fish { Gold }
> Cow { Brown, Black, White }
>
> Which is of course nonsense in reference to my original predicate. Now
> if my understanding is correct, to model this albeit very rough
> information, I can now formally show that I will require an RVA. (I
> could use a surrogate but I would rather not materialize artificial new
> data just to accomodate what I have.)
>
> Someone affirm this to me or shoot my naivite down in flames.

Your example is equivalent to the relvar/keys example. I must also point out (again) that Darwen's formulation of GUNF is not R UNGROUP (X) GROUP (X1, X2, ... XN) AS X != R; that's just how I described it here to visualise it. It might be possible to determine GUNF by examining FDs.

-- 
Jon