Re: Problems with xmlns parts in XPointer

At 09:19 AM 6/4/01 -0400, Elliotte Rusty Harold wrote:
>Yes, that's the reason. It seems wrong to me that the in-scope namespace 
>bindings suddenly lose their scope inside an XPointer. I'd be more 
>comfortable with this if XML Schemas and XSLT hadn't already established 
>that namespace bindings do apply in attribute values.

Speaking as a bit of a devil's advocate (since I wasn't necessarily opposed 
to the shorthand originally)...  I would say that XML Schema and XSLT have 
established application semantics for *their* attribute values that allow 
for QNames to be interpreted in a certain fashion.  XPointer is allowing 
for QNames in a different fashion, partly dictated by the fact that the 
QNames in question are likely to represent element and attribute names in 
documents other than the current one.

>I do see your point about copying and pasting. However, the same can be 
>said for elements, especially elements in the default namespace with no 
>prefix, but others as well. If you copy and paste something that depends 
>on namespaces into a part of the document or a different document with a 
>different set of namespace bindings, then the namespace of the thing may 
>change. XPointers are hardly unique. I'm not sure XPointers alone should 
>attempt to address this problem. I think the problem is global across XML 
>specs, and local solutions like xmlns() just muddy the waters. I prefer a 
>consistent if flawed approach to half a dozen different better solutions.

I don't really disagree with your point, but this isn't the way the 
decisions broke.  (By the way, even URIs in HTML may also suffer from 
copy/paste context problems because of BASE...)

>Something very clear and to the point such as "If an XPointer is included 
>in an XML document, then the namespace mappings created by xmlns 
>attributes in that document do not apply to namespace prefixes found in 
>XPointers. The default namespace for the element in which the XPointer 
>appears does not apply to element names in the XPointer either."

>I also suggest a couple more examples in which the document does set up a 
>prefix  that seems to be used in the XPointer, and yet it does not apply. 
>Currently you only give positive examples of xmlns() parts. I'd ask for a 
>negative example that shows the xmlns:prefix attribute doesn't work.

Good suggestions.

         Eve
--
Eve Maler                                             +1 781 442 3190
Sun Microsystems XML Technology Development  eve.maler @ east.sun.com

Received on Monday, 4 June 2001 10:46:08 UTC