Kevin,

Thanks. So in this example NYPL for instance publishes a news item on 
their website (which constitutes the Annotation) linking to the Bibframe 
Work (possibly in an LoC datastore) and the NYTimes article (either 
directly or indirectly). The Annotation Agent constitutes a type of 
provenance data (who said this?), but the Annotation is completely 
outside the framework more or less controlled by the library world. 
Which is perfectly OK with me. Is this what BIBFRAME makes possible? Do 
I interpret this correctly?

About the "target": I would say: in a library framework that is "on the 
web" and "of the web", we should link to the existing entities on the 
web directly. A BIBFRAME Work or Instance could of course be an existing 
online entity with a URI, right? Or do you see the BIBFRAME framework as 
a self contained implementation? In that case we will make a mistake.

About the Annotation links to Work or Instance: yes, in the case of a 
book cover it obviously has to be to an instance. But Annotation links 
to similar entities, like newspaper articles, should be the same, either 
to a Work or an Instance.

Hmm, writing this reply triggers me to think about the concepts and 
implications even more, as you may have noticed.

Lukas Koster
Library Systems Coordinator
Library and Information Systems Department
Library of the University of Amsterdam

On 30-11-2012 17:33, Ford, Kevin wrote:
> Hi Lukas,
>
>>   From the examples given I can't exactly figure out what the meaning of
>> Annotation Agent is.
> -- In short, it is the entity responsible for *creating* the Annotation.  Based on your email, I can see how the heading confounds; I'll be sure to modify it if I use it again.
>
>> the NYPL
>> asserts that the NYTimes article is in fact a review about the bibframe
>> work. Or it identifies that the NYPL publishes information that there
>> is a review about the work in the NYTimes that might be of interest to
>> the NYPL patrons.
> -- It's the second sentence:  NYPL (a source, and a trusted one we presume) says "here is the NYTimes review of this Work."
>
>> Then, the nature of the "target" relation. Does the target point to a
>> BIBFRAME metadata representation of the NYTimes article (containing a
>> URI to the actual article), or directly to the NYTimes article, or can
>> it be both?
> -- We're staring to think it might be best to provide both options: link to metadata resource and/or the review itself.  The metadata resource provides additional information that might aid in the display/comprehension of the "target" but being able to link directly to the target bypasses the need to parse an intemediary metadata resource.  Ultimately, I think community practice will inform actual practice: link to one or the other or both.  Other thoughts?
>
>> Then: there are two reviews, one annotation points to a bibframe:work
>> (NYTimes article), the other to a bibframe:instance (WashingtonPost
>> article). What is the difference? Shouldn't it both be just Work or
>> Instance? Not sure if a newspaper article should be a Work or an
>> Instance.
> -- Every time I saw this variation in my own slides I knew this would raise a question, and understandbly so.  I think both examples present perfectly viable possibilities, which is why I did it.  Community practice or rules may eventually guide whether the targets of Annotation resources should be to Work resources or Instance resources, or whether the linked-to target is dependent on the type of Annotation.  For example, the target of Review annotations might point to a Work while the target of a CoverArt annotation points to an Instance.
>
> Does this help?
>
> Thanks for bringing these questions to the listserv.  Much easier to provide fuller answers without a 140 character restriction (though that can be a good thing too). :)
>
> Warmly,
>
> Kevin
>
>
>
>
>> -----Original Message-----
>> From: Bibliographic Framework Transition Initiative Forum
>> [mailto:[log in to unmask]] On Behalf Of Lukas Koster
>> Sent: Friday, November 30, 2012 10:34 AM
>> To: [log in to unmask]
>> Subject: [BIBFRAME] Annotation issues
>>
>> I attended Kevin Ford's presentation at SWIB12. Unfortunately the sound
>> of the skype session was too bad to understand everything Kevin said.
>> The recording http://3windmills.com/kefo-swib12-bfi/ is much better.
>>
>> After watching the recording I have a couple of questions about the
>> Annotation core class. I already asked Kevin a couple of questions on
>> twitter about this after the SWIB12 Skype session. Fro completeness I
>> reproduce them here, as well as Kevin's answers.
>>
>> - Q: Annotation: isn't it just "relationships"? Why need extra entity?
>> (http://twitter.com/lukask/status/273749651759439873)
>> - A: Annotation can associate two Works (or Instances) not controlled
>> by the agent creating the Annotation. Provides buffer area.
>> Still, you raise a good question. We've debated this a number of times.
>>
>> - Q: can Annotations be used a) to flexibly extend the model? b) for
>> adding provenance data? Etc.
>> (http://twitter.com/lukask/status/273749990894092288)
>> - A: We're wondering about "b," but we've not done any work on it. I
>> think "a" is quite possible, and I've wondered the same.
>>
>> Now, in the presentation Annotations are called "assertions about the
>> other core class elements", holding all kinds of possible additional
>> information (reviews, holdings, book cover images, administrative data,
>> etc.). It is also said: "annotations are additional assertions".
>> There are a couple of examples from which you can see that an
>> Annotation has at least three possible relations: "annotates" (one of
>> the core classes), "target", which points at some external(?) entity
>> (like the book cover, or an article), and an unnamed relation that
>> identifies the "Annotation Agent". Besides that each Annotation
>> probably has a Title or Name property, and hopefully a Timestamp
>> property (to be able to define provenance).
>>   From the examples given I can't exactly figure out what the meaning of
>> Annotation Agent is. It could be provenance information: the NYPL
>> asserts that the NYTimes article is in fact a review about the bibframe
>> work. Or it identifies that the NYPL publishes information that there
>> is a review about the work in the NYTimes that might be of interest to
>> the NYPL patrons.
>> Then, the nature of the "target" relation. Does the target point to a
>> BIBFRAME metadata representation of the NYTimes article (containing a
>> URI to the actual article), or directly to the NYTimes article, or can
>> it be both?
>> Then: there are two reviews, one annotation points to a bibframe:work
>> (NYTimes article), the other to a bibframe:instance (WashingtonPost
>> article). What is the difference? Shouldn't it both be just Work or
>> Instance? Not sure if a newspaper article should be a Work or an
>> Instance.
>>
>>
>>
>> Lukas Koster
>> Library Systems Coordinator
>> Library and Information Systems Department Library of the University of
>> Amsterdam
>> Web: http://uba.uva.nl
>> Mobile site: http://m.uba.uva.nl
>> Digital Library: http://lib.uva.nl