RE: Impact of XML on Data Modeling

Michael Kay wrote:
>Scott Tsao wrote:
>> If these observations are correct, my next question would be: Is the
W3C XML Schema 
>> the best choice on the market today for data modeling in the XML
world? (why or why not) 
>
>I think there are two difficulties in using XML Schema for modelling.
One is syntax - a model, 
>by definition, captures the essential qualities of the real thing and
leaves out unnecessary 
>detail. XSDL isn't really a language that allows you to forget about
the detail, or to refine 
>it incrementally. It's a bit like trying to use Java for modelling. The
other is the focus on 
>describing the structure of a particular document type or message type.
This is OK in situations 
>where the document is the data. But in situations where the document is
a message, a transient 
>data flow between components of the application, it works less well,
because it's difficult to 
>capture the areas of commonality between all those messages - there's
some underlying data 
>model and it's only distantly related to the structure of all the
messages that are flying 
>around. Michael Kay http://www.saxonica.com/

First of all, +1 to Michael on this topic.

Many years ago (1999?) I gave a presentation on "UML and XML" and made a
brave attempt to show
how UML could express the ideas that would be realized in an XML Schema.
Turns out it's not
a simple mapping.  Years later, Simon Johnson (of Rational Software) and
I created a UML model 
that expressed XML Schema.   This meta-model featured a lot of OCL code
to express constraints
and a number stereotypes to change standard UML objects into more XMLish
creatures.  In short, 
it was a >great< learning experience, but had little other benefit.
(N.B. Matthew Fuchs also 
did a lot of work in this modeling area for SOX.)

As a second reference point, Paul Biron and his team created a set of
rules (and tools) for 
defining XML languages in UML for use in the documents created at HL7.
The tools map between
UML and XML Schema.  AFAIK this effort has been largely successful, but
it doesn't use XML
Schema for modeling:  it uses UML, and then generates XML Schema
documents to implement the
designs; I don't believe the tools support reverse engineering XSDL into
UML (I could be wrong).

The real point is that XML Schema is a definition language, not a
modeling language.  I had 
to burn several weeks trying to prove to the contrary in order to grok
that. [1] 

HTH

David Ezell

[1] I'm not dismissing the value of this discussion thread.  It's very
interesting and informative.

Received on Wednesday, 30 January 2008 18:14:24 UTC