Re: some information about anchor modeling

Dana petak, 4. siječnja 2013. 10:29:50 UTC-8, korisnik paul c napisao je:
> On 02/01/2013 1:50 AM, vldm10 wrote:
>
> > First, RM / T is not discipline. There is no industry or theories, which are based on the RM / T.
>
>
>
> Ha! I'll bet many people would agree that RM/T is where Codd first tried
>
> to justify nulls. Today practically the whole known English-speaking
>
> world uses so-called industrial dbms'es that embrace the idea of
>
> recording what we don't know. Maybe the rest of the world, too for all
>
> I know.

In this thread I gave examples, which show that the surrogate key is a bad solution. These examples show that the missing data can not be solved by applying surrogates. One such example, I wrote in this thread on 18/07/2012:

Let R be Relvar, which has the surrogate key K and three properties A,B,C,
Let us suppose that it is somehow possible to decompose the above relvar into binary relvars using the “RM/T discipline”. Let the following example be one possible situation:

K K A K B K C

k3                         k3 c3
k4                         k4 c3

The above decomposition is very bad. For instance, there is the question: how will a user find the real world entity that has the attribute C=c3 and the surrogate key K=k3? Note that a surrogate key is only in the database, it is not in the real world. So, my point here is that the surrogate key makes this table so bad that it becomes not an acceptable design.

--

I did not want to show all the bad consequences of the use of surrogate key in this example, because I think that this example is enough. But apparently it's a good idea to show some other disadvantages that result from the use of surrogate key.

1.  For example, there is the question: A data entry person should enter the attribute b7, into the corresponding binary relation. This binary relation has a surrogate key whose value is k3. How will the data entry person find the corresponding binary relation? Obviously, with a surrogate key, there is no solution to this problem.

2.  How will one make m-n relationships with entities in the above example? Obviously, a surrogate key is not solution to this problem.

3.  How would one apply the update, and delete operation on the data from the mentioned binary relations?

Suppose someone wants to solve the above problems by applying Codd’s Three-Value logic. This approach is not correct because the above problems are not a matter of logic; they are a matter of database design.
For example, if in the above example, we apply VINs (Vehicle Identification Numbers) instead of surrogates, then we will not have the problems mentioned above.

Note that Lukasiewicz and Kleene use Three-Valued Logic before Codd. Codd uses nulls as a method of representing missing data in the relational model.

Vladimir Odrljin