> The question probably is how to break the tyranny of the Relational
> Data Model in general, at least at the application-to-application
> level. Relational thinking is so ubiquitous that I doubt it ever
> occurs to the vast mojority of IT staff that data modeling and data
> access can also be done on the basis of quite different paradigms
> (e.g. XML, RDF or service interfaces).

The next question would be when, and why, would you want to break this
tyranny? The relational model per se has benefits for analysis. XML is
a data format, not a model.

RDF is a data model, but has yet to prove its value in a wider variety
of uses as a logical model. It makes a fine conceptual model, and a
workable logical model in some cases. On the other hand a
well-designed star schema is essentially an assembly of several RDF
schemas into a more appropriate logical model for many kinds of
analysis that business use. i.e. one column of one dimension is
related via the fact to some measurement. All the columns of all the
dimensions of one fact represent all the RDF triples that "belong"
together.

> And worse: usually it is simply a non-option even to think about
> taking that direct SQL access away from them. And often not even if
> they already experienced large- scale coupling on the database
> schema with changes to the scheme being calculated in man years.

These are problems with specific tools, poor design choices, and
improper uses of relational tools. Used in the right places
(e.g. analysis, not OLTP), with good designs (e.g. star schemas,
similar-grain facts, conformed dimensions), and better tools
(e.g. Sybase IQ demands far less resources and far less adminstration
due to column-based storage rather than row-based), the relational
model is effective.

-Patrick








SPONSORED LINKS
Computer software Computer aided design software Computer job
Soa Service-oriented architecture


YAHOO! GROUPS LINKS




Reply via email to