two options

["Bruce D'Arcus" <bdarcus@gmail.com>]

I'd like to see if I can cut through some of the recent confusion and  
keep us moving towards a solution.

As Elias and I have been talking about this we really have two  
options to fulfill our requirements and cover our use cases:


Option 1: The all-RDF/XML approach
==================================

This presumes that all metadata is stored apart from the content file  
in the package, and refers from those files to document objects in  
the content.xml file. The sole requirement in the content.xml file is  
that the object to which one would like to attach metadata is  
identified with a local or global IRI.

The citation example I have on the wiki [1] is a complex example of  
this.


Option 2: The Hybrid RDFa-inspired and RDF/XML approach
=======================================================

This would allow a handful of RDFa inspired attributes to be attached  
to content nodes to give them further semantics, and then use the  
separate RDF/XML files to augment those sort of semantic anchors.


Discussion
==========

I favor Option 2 because Option 1 will prove too limited. The only  
reason it works for my citation example is because I rely on the  
implicit semantics associated with a hard-coded element:  
text:bibliography-mark. In essence, this becomes a sort of property  
of the document. But we cannot rely on these implicit semantics,  
precisely because we want to enable custom solutions.

Perhaps we can discuss this for the call next week? I suggest we need  
to make a decision really soon whether we go with Option 1 or 2, or  
simply bring two proposals to the TC.

Perhaps we can discuss this in the context of one or more different  
(but still complex) use cases; say the medical one?

Bruce

[1] <http://wiki.oasis-open.org/office/Metadata_Examples>