XHTML2, RDFa, and MetaData

During the vF2F today, an issue came up that I wanted to try to capture 
more cogently than I did at the time ;-)

Basically, there has been this underlying thread in the Working Group 
for ages that we have never looked at too closely: some elements / 
attributes are capable of setting metadata "properties" that are the 
/same/ as other elements.  E.g. the title element establishes a document 
title - but we claim that setting a meta element with a property of 
dc:title also sets the document title, and that these are the same 
property.  You can have link elements with a rel of next, but you can 
also set rels of "next" anywhere with any element.  You can have a span 
with an about of itself, a resource of a CSS file, and a rel of 
stylesheet. 

Today I tried to argue that this was mistaken, and that we should not be 
co-mingling the metadata from Meta-information elements and attributes 
with those other sources.  My argument seemed to fail.  Steven seemed to 
say that one focal point of XHTML 2 is a harmonization of the metadata 
story so that it is possible to co-mingle metadata and content more 
easily.  That seems consistent with the RDFa approach, so I guess I 
agree that for metadata that is what we are going for.  In the spirit of 
collaboration, I wanted to try to capture what I think are some of the 
ramifications of this harmonized metadata model.

   1. There are many ways to set the same "property" via the
      Meta-information Attributes (e.g. RDFa).  Such settings can happen
      anywhere in a document stream, and can of course change as a
      result of document mutation.  This doesn't matter to an RDFa
      processor, but may matter to an XHTML 2 User Agent.  If we want
      the user agent to care about all of these properties, we need to
      specify conformance requirements for it.
   2. Some XHTML elements and attributes clearly can be used to set
      things that I would consider metadata. INS and DEL, for example. 
      @cite, @datetime, @edit, @layout, @title... Probably others.
   3. There are a finite number of base properties defined in the XHTML
      vocabulary document [1], but many of those properties have
      corresponding elements / attributes in XHTML2.  When those
      elements or attributes are used (e.g., cite, title, role), an RDFa
      processor should probably be extracting triples for those
      properties, just as it would for the traditional RDFa attributes. 
      Correspondingly, to the extend that the user agent cares about any
      of this metadata, setting these base properties via the
      meta-information attributes should be reflected in the user agent
      (@rel values on link, cite, title, others?)
   4. There are potentials for conflict between these states, and we
      need to carefully document how such conflicts are resolved.

I think this is incredibly complicated, and I continue to believe it is 
an unreasonable burden on the user agents to try to harmonize this 
information - but if that's what we want to do, we need to think it all 
the way through.

[1] http://www.w3.org/1999/xhtml/vocab

-- 
Shane P. McCarron                          Phone: +1 763 786-8160 x120
Managing Director                            Fax: +1 763 786-8180
ApTest Minnesota                            Inet: shane@aptest.com

Received on Tuesday, 10 March 2009 21:58:59 UTC