[
Lists Home |
Date Index |
Thread Index
]
- From: Robert Worden <rworden@dial.pipex.com>
- To: 'Mike Brown' <mbrown@corp.webb.net>,"'xml-dev@lists.xml.org'" <xml-dev@lists.xml.org>
- Date: Sun, 20 Aug 2000 22:03:34 +0100
Mike Brown [SMTP:mbrown@corp.webb.net] wrote:
> I want to use XML to express relationships between instances of the
> conceptual/abstract objects that are each modeled by separate XML
documents
> (or separate collections of documents).
Do you always have a separate XML document to model each object? Isn't it
more common to have several different object classes in your model, and
typically many object instances per document, with relationships both
within and across documents?
Expressing relationships (UML: associations) between objects even within
one XML document is not that simple, because XML gives you so many ways to
do it. XML can express a relationship between two object instances in any
of the following ways:
- as an idref/id pair
- by shared values of attributes or elements (just as relational databases
express relations by 'foreign keys')
- by nesting of elements (element representing object A inside element
representing object B)
- by 'denormalisation', attaching the attributes of both objects to the
same element
- separately, outside the elements representing each object
There are many variants on these and they are not just academic; you can
find all of them in commonly used XML languages. So there are lots of
'standard' ways to express object relations; although I believe initiatives
like ebXML would like to encourage some rather than others (shutting the
stable door after the horse has bolted...?). So did the BizTalk 'canonical
form', before BizTalk started accepting everybody's schemas.
This plethora of ways to express relations is what makes XML translation
hard - preserving the relationships while translating them from one
representation to another. We have built a tool that can understand and
translate all the known relation representations within a document, but not
yet relations across documents.
Robert Worden
http://www.charteris.com
|