Re: On Normalisation & the State of Normalisation

> On Friday, 6 February 2015 08:58:13 UTC+11, com..._at_hotmail.com wrote:
> On Tuesday, February 3, 2015 at 5:59:04 AM UTC-8, com..._at_hotmail.com wrote:
> > I just find it amusing that the particular application plus no interpretation of tables tells us that there is a universal relation CK that this diagram fails to express.

No idea why you might think that data models are supposed "hexpress" URs or CKs. Do you go to the butcher for bread ?

Therefore "fails to express" is false. Rather more telling of your position.

We don't have CKs in the physical universe, as already explained. They are a construct in the theoretical universe to avoid the election of a PK, and thus a subversion of the RM.

The developer has no CKs in the model anyway, he is not trying to subvert the RM (at least not intentionally).

> Here's what I failed to express earlier:
>
> An envelope gets to a particular address. Ids do not appear on an envelope. So joining an appropriate set of tables and projecting out id columns gives rows 1:1 with Address AddressIds. So there is a join CK among the non-id columns. But this set of tables gives no such CK. So they are not appropriate.

Result: agreed.
Method: far too backward and laborious. If we stick to the RM, the method is much simpler, the crime is simpler to identify.

> (The FKs correspond to FDs on the CK in the join. So there is a CK among the non-id columns of Street and Address.)
>
> PS:
>
> Of course, that claim does not arise from a particular (normalization or design) process. So it proves nothing about the adequacy of any allegedly adequate process. Nor would any claims about the design without giving a process and sound reasoning.
>
> To show that an alleged process is adequate one need only give it and either a counterexample or sound justification. Of course, one would only soundly believe it inadequate if they had already done so. (Granted, after demonstrating enough examples of inadquate processes from a group one might reasonably claim that the group didn't have an adequate process.)
>
> Of course, that's not showing that an allegedly adequate process actually is. That requires giving it and a sound justification.

Looks like a nice justification for incompetence.

This is the same as the six hundred words to explain that a decrepit non-Relational failed-3NF data model is somehow accepted as one. Side-stepping the issue that you don't know how it breaks 3NF, you don't know how it breaches the RM.

All we need here is for you to examine the model and accept/reject it against the criteria given. We do not need a write-up on the process you use, and that you think we should use, which is fine to ponder, in case you can't execute the exercise.

So, are you rejecting it or accepting it ? If rejected, on what basis (less than six hundred words) ?

1. Is it "5NF" ? 
2. Is the reference to SCGHNFRDB valid ? 
3. Is it Relational ?

I see that you perceive the non-existent CKs doing the polka with the UR behind the eyelids, and you notice the CK is tripping once or twice, but is that a rejection ? Is the Ur knocked out, do we need an ambulance, or is she just dazed, will she live to polka again ?

Cheers
Derek Received on Sat Feb 07 2015 - 09:53:39 CET