Re: The definitive XML series. (XML Modeling)

Chris Olds (colds@nwlink.com)
Fri, 10 Apr 1998 14:22:28 -0700


One of the difficulties with XML as SGML-- is that there are still some
types of relationships and common constraints that are hard or impossible to
represent using XML DTD syntax (which I feel is the place to start, since it
is the schema for an XML document). With all of the hype about how XML is
"mata-data for the web", it seems more natural to many people coming from
the modeling and database world to use XML instance syntax to deliver both
the meta and instance data (and, of course, since the instance under
discussion is metadata, we can always add another meta- to everything and go
around again, but I won't go there). The XML Data effort (
http://www.w3.org/TR/1998/NOTE-XML-data/ ) is just such an attempt to use
XML instance syntax for schema information. Unfortunately, the current
public proposal is explicitly relational in style, and does not provide the
level of information found in a UML model (which I belive is also lacking in
important ways, but I won't got there now either).

I have not tried to do an XML DTD for UML, but it seems like a good idea.
One of the good things abour XML is that there is nothing that prevents my
doing this. One of the bad things about XML is that nothing prevents my
doing this. The reason it is bad hinges on the horrible truth that XML is a
very nice syntax for marking up a document (file) according to its semantic
content, but does not provide a way to convey the semantics of the document
in a standard way. While this is obviously a hard problem, I think that for
XML to live up to its promise there needs to be some more work on the
definition and publication of shared ontologies so that we avoid the problem
Alice had with the Humpty Dumpty (
http://www.bibliomania.com/cgi-bin/mfs/usr/httpd/Bibliomania?file=/usr/httpd
/Bibliomania/Fiction/Caroll/CompleteWorks/p2-ch6.html&line=151#mfs ).

At 06:59 PM 4/10/98 +0200, Dirk Riehle wrote:
>Now with XML, I hope we can simply kick out our proprietory
>parsers, file/repository readers/writers, etc., and use off-the-shelf
>implementations.

I think this is both possible and a Good Idea; I am currently doing this for
the product I work on. I actually wrote an XML processor (parser) of my
own, but only because I wanted a C++ interface and couldn't find one.

>PS: An example would be a UML "formalization" using XML for
>the repository of a modeling tool. Or, what about XML based modeling
>of database schemas?
>
>(PPS: Is it schemata in English?)

American Heritage says that either is correct.

/cco