It strikes me that harmonization of XML standards (i.e. where you have different XML standards covering much the same ground and you want a workable strategy for converging them) needs to be as much concerned with granularity issues as it is about field-level reconciliation.
When people talk about metadata, they very frequently are talking about properties of chunks of information bigger than the field level. In XML there are now four major divisions where such metadata pools: the element level (i.e. attributes or structured properties), the namespace chunk (often just an ID), the document level (usually in the head of a head-body pattern), and the document set level (usually in discrete XML file(s) in the archive or directory.
So I think that a general-purpose harmonization system would have to work at each of those grains, at least.
There are dribs and drabs around. DSRL provides a simple remapping of constants at the element level, DTLL allows parsing of data items, MCE works at namespace chunk level, RDF allows statements like "this is a that", but we don't have anything at the document or document set levels.
Now that XML is in its second decade, harmonization issues are increasingly come to the fore and being met by embarassed silence by gurus old and new. Standards consortia have provided opportunities for new kinds of market-enhancing cooperative behaviour by document stakeholders, and the XML technologies have served fairly well the flowering half of the schema-development-cycle but do not have good mechanisms for serving the other half of the cycle, which is where the focus of groups turns to the consolidation or culling of schemas.