how do we deal with metadata vocabularies?

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

A few weeks ago I asked the question here what we should do about the 
existing bibliographic support in ODF 1.2 (text:bibliography-mark and 
such) in favor of the new metadata field and system.

Patrick suggested deprecating it in the future, with a note in 1.2 that 
this will happen.

The question is *how* to do this?

Traditionally, any new XML added to ODF has a fairly high-bar for 
inclusion. It needs to be fully-specified, and preference is given to 
existing standards; e.g. specs or portions of specs that go through some 
formal standards process.

But the new metadata system I think presents us some challenges, and 
that is: how do we think about suggesting -- either normatively or 
informatively -- what vocabularies should be used for particular use cases?

So here's the thing: we have a standard model in RDF. That is 
standardized. That model gives us reliable extension and flexibility. 
Developers can add anything they want to the RDF, and so long as its 
compliant, it can be read and displayed.

It is, I think, precisely the robust distributed flexibility of RDF that 
means vocabularies are rarely formally standardized in the same way that 
an XML language is. This is a feature, not a bug.

Surely for particular kinds of processing (like bibliographies and 
citations) one has to have some expectations about the modeling, and so 
we need to provide this. The question is how?

I think it's clear we should provide the specification for the citation 
field in ODF 1.2.

I think it's also the case that we should not ourselves define the 
vocabulary for the bibliographic source data. I am working on that in an 
independent project that involves developers from different projects.

	<http://bibliontology.com>

So the question is when and how to include or reference this work in 
ODF? We could just wait until ODF 1.3 once it's more stable, and maybe 
have an informative web page that points to it, as well as other 
vocabularies we might suggest (like vard, or Creative Commons, or Dublin 
Core)?

I'm still a little unclear on what the mapping would like in any case; 
whether it's in the spec or not.

Any thoughts on any of this?

Bruce