On 11/30/12 9:18 AM, Lukas Koster wrote:
> Looking at the described conversion procedures and the RDF example 
> somehow I get the feeling that BIBFRAME Works will be treated as local 
> redundant entities. I hope I'm wrong. Of course the BIBFRAME model can 
> be implemented in a number of ways, both as local silo and as globally 
> shared data. But I had the impression that 'based on linked data' 
> meant abolition of redundancy by linking instead of 
> copying/duplicating, not just using RDF as storage format.

Lukas, I tend to see "one and only one Work description for the whole 
web" as a kind of platonic ideal that we will never achieve. I think the 
LoD community has come to the same conclusion about identifiers for 
things: it would be nice if each thing had only one identifier, but 
we've already built in (and over-used) owl:sameAs because that will 
probably never be the case.

For me, the question of "on the web? or locally stored?" will be decided 
depending on what you are doing with the data. If you have a local 
database that people will search in, you probably need to store locally 
some of the key data you expect those searches to go against. If there 
is information that you will mainly use for display you can simply have 
a link that pulls in the relevant data at the time of display (like the 
BBC links to Wikipedia). Then there is the special case of cataloging 
systems for libraries, which I think will be much like they are today: 
closed systems used by trained specialists with a very high degree of 
sharing and low degree of freedom. Where I think things will be 
significantly different from today is that the reference and user 
"views" will less like the cataloging view than they are today, because 
linking will allow us to add user-relevant data that doesn't interfere 
with the catalog/inventory data that library management (acquisitions, 
collection development, circulation) depends on.

Even in the cataloger view, my suspicion is that the Work will end up 
being a cluster not unlike VIAF. That's my guess.


> Ideally, a Work would be described only once, and as such be similar 
> to an authority record to which links are made from Instances and 
> Annotations. Of course in the real world we will have a limited number 
> of redundant Work data stores. For this we would use "sameAs" 
> relationships.
> In Kevin's RDF example I see however four Instance URIs, which would 
> only make sense if this is a local library 'record' containing links 
> to all local holdings of the work (FRBR Manifestations, originating 
> from local legacy MARC records).
> A globally shared Work entity would of course never be able to hold 
> links to all Instances existing in the world. This would not make any 
> sense at all. A Work would not contain any Instance links at all. 
> Rather, all Instances of a specific Work would obviously have one link 
> to that Work.
> The only use for links to Instances in a Work would be to indicate 
> provenance, which already is taken care of by the 'derivedFrom' and 
> 'consolidates' links in the example.
> Anyway, I would expect libraries only to be dealing with Annotations 
> for their own local holdings, linking, in their own new linked data 
> ready systems, to existing Instances (FRBR Manifestations) made 
> available by authorities such as LoC, OCLC, publishers, etc.
> Of course, Works and Instances might be available for editing by other 
> trusted parties in a wikipedia way.
> Lukas Koster
> Library Systems Coordinator
> Library and Information Systems Department
> Library of the University of Amsterdam
> Web:
> Mobile site:
> Digital Library:

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