Re: Any new thoughts on OTLT (One True Lookup Table)
Date: Thu, 1 Apr 2004 10:11:05 +0200
Message-ID: <406bce99$0$5068$4d4ebb8e_at_read.news.nl.uu.net>
>
> And this is precisely why Codd wanted the database schema to be
represented
> as metadata, in relational form,
> and packaged in the database, so as to provide context for the data. If
the
> schema makes sense in the subject
> matter language, and provides the context needed to decode the data,
then
> the data is usable by anyone who can access it, and knows the subject
> matter.
>
Describing some aspect in technical terms does not provide (enough) context.
Example :
Describing the metadata in a database does not describe the what the function of the storage is. The example of the book is used because it is so simple and understood by everybody. But in most data, even if the metadata is completely clear, there are often misunderstandings what something stands for. (Example if you record the 'responsible' person to what extend is the responsible true, for a good system this has to be clearified to be used correctly. This is rearely done within the database).
Offcourse I advocate the use of names for identifiers which describe the identifier. (Fieldnames, tablenames). But I think the name of the identifier only explains a very limited amount about the field/identifier/attribute. For me this is the directory aspect of a description. There should be also a dictionary aspect of the 'attribute' which explains in 'human' terms what the attribute actually means. Although this 'context' can be stored in a database, in real life because it has to be 'build' and 'maintained' and there is no 'technical' reason to do this, it is often not done or not done sufficiently.
ben
> If the context is kept out of the schema, then the database is only
usable
> to those who are in on the secret.
>
>
>
Received on Thu Apr 01 2004 - 10:11:05 CEST