If I accept that BIBFRAME won't use existing vocabularies such as FOAF (While I'm sceptical about this as a reason I've started to reconcile myself to this being a firm decision in BIBFRAME that is unlikely to change) then these two approaches look, superficially, pretty close to each other.
What isn't clear to me is:
1) Does it make sense for Person, Organisation, Topic, Place etc. to be subclasses of Authority? Are there properties that a BIBFRAME Authority has that span across all the types that are currently regarded as BIBFRAME Authorities?
2) Is a BIBFRAME Agent equivalent to a foaf:Agent, or is it intended as a more conceptual entity? (a similar question can be asked of Place)
3) Is it intended that each BIBFRAME Authority is linked to only one BIBFRAME Work or Instance? (this is a question directly or indirectly raised by some of the other discussion around this)
This questions are linked I think.
One other thought. One of the things that http://bibframe.org/vocab/Authority.html
suggests is that the $0 control field from existing MARC records will be used in the creation of BIBFRAME Authorities. So consider two (highly abbreviated) MARC records:
245 00 $aBarchester Towers
245 04 $aThe Warden
Under the current description (as far as I can understand it) of BIBFRAME this would convert in one of the following ways:
The first representation is what I'd naturally hope for. The second is what I think would be required (under the current proposals) to roundtrip between MARC and BIBFRAME.
If we wanted to achieve something between these two options - i.e. a single entity for Trollope, but the ability to roundtrip with MARC, then we would need to but the information about 'authority used in relation to this work' into a different structure (possibly along the lines that Tom has already proposed in terms of linking a creator to a work...?)
Sorry to go on
Owen Stephens Consulting
I had to do some diagrams to explain what I see as the possible relations between BIBFRAME "authorities" and "library authorities, and since email doesn't work well for that I did a blog post instead:
I apologize if this splits the conversation, and can echo back here any significant comments.
The main gist is exploring options for the positioning of the BIBFRAME authority thing in relation to the traditional library authority "record."
It needs some example code, but I'm struggling with that because code isn't very readable, and the more readable types tend to mask the triple-ness of the underlying data.
Owen Stephens has given me an idea for a fourth diagram, which I will add if I get a chance.
[log in to unmask] http://kcoyle.net