Rob - thanks for your comments, and just a few more of mine. I'll try to reduce things down a bit.

First, it has occurred to me that, perhaps ironically, the "Functional Requirements for Authority Data", FRAD, [1] is very close to OA. Unfortunately, the English version isn't available online, but if you grab one of the ones whose language you can make out, all that really matters is Figure 2 (p. 7 in the French version, for example). It's not exactly the same, but I sense a similarity in the intent, especially in terms of making clear WHO is making the statement. FRAD adds one more thing, which is under which rules the statement is being made. I could imagine, therefore, BIBFRAME authorities looking much like OA if FRAD is adopted.

On 5/5/13 2:17 AM, Robert Sanderson wrote:
[log in to unmask]" type="cite">

Agreed, however these were simply to demonstrate that annotations expressed in RDF require more than a single reified triple, to refute the claim that we're simply reinventing RDF graphs.  There needs to be an ontology and defined structure.

Yes, and this is bringing up for me some work I'm doing around the Dublin Core Application Profiles [2]. I need to catch some other folks to have a discussion of this, but will get back to you if anything interesting arises.
[log in to unmask]" type="cite">
Yes, definitely.  The cataloguing type of use cases I would see as clearly annotations:

* Reviews / Notes
* Tagging / Subject classification
* Citations

Essentially, when the information is provided by a third party, then it would be good to be expressed as an Annotation such that the provenance information is maintained and is compatible with other systems that use Open Annotation.

I'm still having a hard time with the dividing line between stuff and annotations of stuff. A recent article explaining the OA concept [3] said:

 "Annotations describe resources with additional information, which is valu-
    able to other users, who are searching and browsing resource collections."

If you have "additional information" then there must be "non-additional information" -- and if you have a third party then there must be a first party (the second seems to get skipped over in that analogy). I don't think this is a generalizable question (one person's "additional"... etc.), but at least for any community like libraries some definition is going to be needed if annotations will have some operational meaning. That's what I think is not clear (yet) in the BIBFRAME document. I suspect that the division in library data will be based on provenance, but that's not how BIBFRAME defines annotation.

[log in to unmask]" type="cite">

The point I was trying to make is that the bibframe annotation model has a property bf:annotationBodyLiteral, which has a literal as the object.  The Open Annotation model explicitly does not allow this, for several important and well discussed reasons.  Thus a point where the two models are not compatible.

Right. But then again, DC types doesn't include "literal" as a type. The whole typing thing is something I would need to think about more. If your only interest is literal v. URI, then the BIBFRAME choice is fine. If you have a video or sound file, that would be a URI. Something that is interesting in the OA model is that it allows you to make statements about the thing the URI points to [e.g. annotating a particular coordinate area on a map]. I hadn't thought about that, and now will have to go back to the OA document and see where that fits in. This is one of those areas where the "meta" aspect of library data has an affect. Must think more.....

[log in to unmask]" type="cite">

Note that there is nothing about library data that would prevent anyone from applying an Open Annotation to the data. The question is, do we want this to be the only way to, for example, assign a subject heading to a Work? OrI wouldn't assume that the BIBFRAME data will exist *as is* in the open world.

Then I would seriously question why RDF is being used at all.

I see this differently, perhaps because of my work on Application Profiles. I think that many communities will have an internal view of their data, and another "face" that they present to the open world. The basic structure of RDF, that is, triples, and the use of defined ontologies, is a good way to organize data whether it is open or not. In fact, there are enterprise uses of linked data. Modeling your enterprise data on RDF makes it much more likely that you will be able to extract a public view. In the Dublin Core arena we are looking at Application Profiles as a way to add the constraints that are needed within a community or enterprise that has some non-open needs, while easily allowing the derivation of an open view.

[log in to unmask]" type="cite">
Was the choice of annotation as a model discussed on this list (and hence available in the archives) or was it simply presented fait accompli?

Rob, what you and I are doing here is the full extent of our ability to interact in the BIBFRAME process. If you look at the "Contribute" page [4] of BIBFRAME you will see that the only involvement for the community in the BIBFRAME development is discussion on this list. Discussion that, like this thread, takes place generally between list members with no direct involvement in BIBFRAME development. I would have simply replied to you individually, but I'm of the mind that a few readers on the list might find our musings interesting. In terms of having any influence over the development of BIBFRAME, I have no such conceit.


Karen Coyle
[log in to unmask]
ph: 1-510-540-7596
m: 1-510-435-8234
skype: kcoylenet