Removing these layers between the subject and the description would simplify use of the data, IMO. In particular, anyplace where one can NOT use a blank node there is a gain in ease of use.

Unfortunately, this is possible input from the AACRs and RDA:

260 ##$aParis :$bGauthier-Villars ;$aChicago :$bUniversity of Chicago Press,$c1955.

e.g. Paris : Gauthier-Villars; Chicago : University of Chicago Press, 1955.

That could logically be expressed as two separate publication statements (I assume the current display form harks back to limited space on cards):

Paris : Gauthier-Villars, 1955.
Chicago : University of Chicago Press, 1955.

As simple triples, though, there is nothing to retain the connection between the individual places and the publishers since there is no order inherent between triples.

x bf:publisher "Gauther-Villars" .
x bf:publisher "University of Chicago Press" .
x bf:date "1955" .
x bf:pubPlace "Chicago" .
x bf:pubPlace "Paris" .

(NB: this is not the only place in our data that repetition of data elements and the reliance on the order of elements creates difficulties. That is another reason to do some re-thinking of our data in light of new technologies.)

We struggled with this when developing the RDF for RDA, and I don't think that there is (yet) a solution that works well in practice. RDA retains the concept of "publisher statement" that is conceived of as a single multi-part description. Thus, in RDF the RDA "publisher statement" would be a node with place, publisher and date. The complex statement above could be two different nodes and that would solve the issue of connecting places and publishers (while relying on UI developers to provide the traditional display), but means creating a publisher node like the one BIBFRAME has today. Assigning those nodes identifiers doesn't really make them re-usable, though, as you point out.

Given that BIBFRAME has a property for the display form of the "publisher statement," it may come down to a question of purpose: apart from display, what do we anticipate doing with places and providers, and do those functions (e.g. search, linking to maps, creating timelines) require us to maintain the proper place/provider relationship when there is more than one?

My 2 cents is that this is one of those areas where separating display from data could have some practical advantages. Your solution provides both.


p.s. This gives me additional respect for the document + data method, which relies on the document for structure and display, and still surfaces useful data. See how WorldCat does this with -

On 7/31/14, 9:50 AM, Ford, Kevin wrote:
[log in to unmask]" type="cite">
Dear All,

Recording "Provider" information, such as who published, produced, 
manufactured, or distributed something, where that happened, and when, 
is presently modelled in such a way that a resource is devoted to this 
information.  An example:

<> a bf:Instance,
     bf:publication [
         a bf:Provider ;
         bf:providerDate "1966" ;
         bf:providerName [ a bf:Organization ; bf:label "Hamlyn" ] ;
         bf:providerPlace [ a bf:Place ; bf:label "London" ]
     ] .

In the above, the resource employs a blank node, but it would not need 
to.  Regardless, this approach has a couple of significant problems:

1) Semantically, "providerDate" is unclear because it is actually 
supposed to convey the "publication date."  And the (publication) date, 
in fact, is an attribute of the Instance (the manifestation basically) 
and not the "Provider" resource. (And simply bf:provider would be better 
than bf:providerName, but that is a small point.)

2) It is not very reusable.  The above bf:Provider is only applicable to 
things published by Hamlyn in London in 1966.

We'd like to explore simplifying how this information is handled in 
bibframe by eliminating the bf:Provider resource altogether and creating 
12 properties, 3 each for publisher, manufacturer, distributor, and 
producer, all of which represent the major use cases as has long been 
expressible in MARC.  These properties would be associated directly with 
the Instance.  As an example, the above would become:

<> a bf:Instance,
     bf:publishedBy [ a bf:Organization ; bf:label "Hamlyn" ] ;
     bf:publishedAt [ a bf:Place ; bf:label "London" ] ;
     bf:publishedOn "1966" .

You can imagine 3 each for manufactured*, distributed*, produced*.

This would clarify the semantics and do away with a resource that would 
probably often be identified via a blank node because it is reusable in 
only fairly specific circumstances.  (The above solution does not 
preclude being able determine all the things published by Hamlyn in 
London in 1966, if that is of specific interest.)

FYI: There has been no discussion whether bf:providerStatement would 
change in any way, and I see no reason for it to change (except, 
perhaps, to add publisherStatement, distributorStatement, etc. for 
clarity and parity purposes, versus the one catch-all 
providerStatement).  bf:providerStatement is really designed to address 
the transcription aspect expected in RDA whereas the proposed properties 
are designed to capture more structured data.  It's an undesirable 
duplication, but it is what it is.

Can anyone foresee issues with this approach?


Kevin Ford
Network Development and MARC Standards Office
Library of Congress
Washington, DC

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