W3C home > Mailing lists > Public > www-xml-xinclude-comments@w3.org > July 2004

Re: xml:lang implementation report

From: Elliotte Rusty Harold <elharo@metalab.unc.edu>
Date: Thu, 8 Jul 2004 14:27:32 -0400
Message-Id: <p06010204bd133f851c77@[]>
To: Glenn Marcy <gmarcy@us.ibm.com>
Cc: www-xml-xinclude-comments@w3.org

At 9:52 AM -0400 7/7/04, Glenn Marcy wrote:

This situation doesn't seem at all unique to xml:lang.  There
could be other situations where an attribute of an ancestor
of the embedded elements is considered "in-scope" and such
attributes would need to be added to those elements during
inclusion to preserve the semantics those attributes represent.

This is a shocking statement. It looks completely unique to me. I see 
nothing in the XInclude specification that suggests this could be 
done for any attribute except xml:lang. The XInclude specification 
does specify how attributes are handled, and no attributes other than 
xml:lang get this treatment. An implementation that inherited the 
value of any attribute other than xml:lang would be nonconformant to 
the specification.

While the XInclude specification describes behavior related to
the attributes with such properties in the XML specification
that appear in the base Infoset, these are not exhaustive.  It
is certainly possible that other specifications could build
upon the base XInclude and specify additional attributes that
would require such special handling.

Again I see nothing in the XInclude specification or, for that 
matter, the Infoset specification, that in any way suggests this is 
acceptable. If your intent is to allow implementers to add attribute 
inheritance for attributes of their choice, then that's certainly not 
indicated in the spec. It might make sense to do that, though it's a 
radical change. However, it's not what the current public draft 
actually says.

Removing the requirements for xml:lang processing in XInclude
would appear to send a signal that we are precluding such
usage, something that one tries to avoid in producing a core
specification that is intended for use in as wide a range of
XML processing scenarios as possible.

The intent to use this in as wide a range of scenarios as you suggest 
here is by no means obvious. Perhaps it's what the working group 
meant all along. I don't know. But it is certainly not what they have 
written. The specification as written offers no grounds for 
inheriting the value of any attribute except xml:lang.

   Elliotte Rusty Harold
   Effective XML (Addison-Wesley, 2003)
Received on Thursday, 8 July 2004 14:46:59 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 23:09:35 UTC