Grant McLean wrote:
>  From: Robin Berjon [mailto:robin.berjon@[...].fr]
> >I see mostly one reason to prefer clarkian notation over 
> >normalisation, it's an ugly reason but it's unfortunately a real 
> >one: QNames in content. That filter won't be able to update it.
>  
>  Can you give an example of the 'QNames in content' phenomenon?
>  (I'm not trying to propose a fix, I'd just like to understand the
>  problem better).

As Eric pointed out, what's probably the most frequent example is XPath 
within XSLT. <xsl:template match='foo:bar'> , with foo mapped to 
http://foo/, will match any bar in that namespace. If you read the XSLT 
spec carefully, you'll see there are a good number of other places where 
QNames occur in attribute content, including some you might not expect.

Another frequent example people have to live with is XML Schema, with 
many such occurances. The problem is that you can't have a generic 
filter that will know when to change content and when not. Take the 
following example (from a horrible language, but still):

<simpleType name='Foo.ST'> 
   <restriction base='bar:SuperType.ST'> 
     <enumeration value='bar:one'/> 
     <enumeration value='bar:two'/> 
     <enumeration value='bar:three'/> 
   </restriction> 
</simpleType> 

It just so happens here that if you're processing this document to 
normalise that silly "bar" prefix to the obviously better "baz" prefix, 
you will break the document if you *don't* change its occurence in 
'bar:SuperType.ST' and you will break the document if you *do* change it 
in the enumerations.

-- 
Robin Berjon <robin.berjon@[...].fr> 
Research Engineer, Expway
7FC0 6F5F D864 EFB8 08CE  8E74 58E6 D5DB 4889 2488

_______________________________________________
Perl-XML mailing list
Perl-XML@[...].com
To unsubscribe: http://listserv.ActiveState.com/mailman/mysubs