Re: What is a surrogate identifier

On Mar 16, 4:03 pm, "Roy Hann" <specia..._at_processed.almost.meat> wrote:

> I think you already blew it. There might be an argument for isolating the
> postal code, but it is a very rare business problem that requires one to
> think about the address as N distinct dimensions. In fact, I would suggest
> that you have violated 1NF since almost none of those "dimensions" is
> meaningful except in the company of the others. You have invited us to
> discern spurious internal structure in what should be an atom.
>

I seem to remember us discussing the question of when some data should be a user-defined type and when it should be a set of attributes in a relation. I think this paragraph describes where I had got to but more clearly than I could manage at the time. I think we were both a bit uncomfortable that there was no particular theory to assist with these decisions at the time, but 1NF as now understood might serve nicely. It was a few years ago now, and I think our discussion example was a little more "earthy" than addresses, so memory might not be serving me too well ;)

• Tony