Re: Database design, Keys and some other things
dawn wrote:
>
> So, what is your rationale for choosing a single, often meaningless,
> identifier for each relation rather than working with a more natural
> candidate key in a relationship table?
I do not generate only a key; it is also the attribute which is the
name of the relation's instance. In current RM, typically, you do not
have an attribute which is the name of the relation's instance, you
have only the key. The key can be compound - meaning a key can be a set
of the attributes; i.e. the key is not an attribute. So my solution is
the attribute and it can't be meaningless because attribute has
origins in Conceptual Model and meaning in the Real World. A key which
has number as value is annoying, but we always use some additional
information to get meaning in the Real World. It is same case with SSN,
VIN etc.
If we have the Car table with the cars which attributes are all the
time same, it is okay to use existing RM solutions. However the RM's
definition of the key can't support cases which we discussed. In
these cases we need maybe to split the table. This involves a FK and
the data integrity or we need some additional "fields" which are
not the attributes, or we need some additional attributes. If an
entity has an attribute which repeats its value then I prefer my
solution. Relationships repeat their values more frequently then the
entities.
Finally current implementation of the Relational Model causes some
problems.
The TransRelational™ Model is what the database world is waiting. It
provides a completely new approach to implementation. In this Model
columns are stored separately. But the key definition in the RM should
be changed. The logic of the compound key is directly opposite to the
logic of this model.
In 2.4 under 2. on my website I set more general representation of
storing the columns separately. An example 2.5 shows how knowledge can
be represented regarding one data (column Amount) and using the key
which is related to the state.
Vladimir Odrljin
Received on Tue Sep 27 2005 - 03:00:10 CEST
Original text of this message