Re: A simple link, it turns out

Alexandr Savinov wrote:
> dawn schrieb:
> >>From August 2005 wired (sometime in the near future) a callout in the
> > article "We are the Web" by Kevin Kelly reads,
> >
> > "A simple link, it turns out, is the most powerful invention of the
> > decade."
> >
> > The RM turned our former foreign key navigating links into set
> > processing joins so that instead of navigating from one "page" to
> > another, we join, restrict and project.
>
> Yes, in the RM we need to do almost everything manually (numerous joins
> and restrictions, hundreds of options) so it is nowadays a low level
> data manipulation and access technique rather than a high level data
> model. The RM knows almost nothing about our data and hence it cannot
> help us in many important tasks. If earlier (say, 10-20 years ago) it
> was used really as a high level data model then now it is used to model
> other data models.

Excellent! Thanks. I had not heard it said that way, but it makes sense. My points are typically about the logical data model and interface between the developer and the dbms api with the "implementation model". I don't care if the RM is below that -- it just doesn't seem to be just right for the API.

> And for this purpose it is very efficient because of
> its powerful record matching, manipulation and transformation
> mechanisms. In particular, complete absence of *logical navigation*
> means (for example, by using links or automatic constraint propagation)
> is actually a verdict and restricts its use to the level of
> implementation of other models. In other words, the RM with respect to
> other (conceptual) data models is like assembly language or machine
> codes for OO languages.

yes, yes -- very good

> Data modelists and designers do not think now in
> terms of relations -- instead they use them (or other means) to
> implement higher level constructs.

Here I will disagree. It seems to me that what we teach in colleges across the US (having read many syllabi) is to get students to think in terms of the RM for application programming.

> Indeed, if you have hundreds or even
> thousands of tables then will you build yourself an access path
> consisting of 10 or 20 tables? Even if you are an expert it will take a
> couple of pages of code and nobody will be able to understand and edit
> it in future. Conceptual models with *logical navigation* allow doing it
> in a couple of lines with absolutely predictable and clear result.

precisely! cheers --dawn

> --
> http://conceptoriented.com
Received on Wed Jul 27 2005 - 18:22:57 CEST