One of the outcomes of the 2017 European BIBFRAME Workshop  was a paper called “BIBFRAME Expectations for ILS tenders” containing suggestions for “requirements for Integrated Library Systems (ILS) vendors to fulfil Linked Data model, with particular focus on BIBFRAME conformance“. . Over the past few weeks I have pondered this document a bit and wanted to share my reflections with the community.
First of all: Tiziana and the Organiser Group for the 2018 European BIBFRAME Workshop have made a nice good job putting this together; I found the idea to use the Maturity Model particularly interesting. Thank you!
That said, I do have some issues with the general direction this paper takes. The main one it that I find it too focused on technology and too little on functional requirements. The paper suggests a transition from a paradigm (1) where cataloguing is done directly in MARC records and where the Integrated Library Systems (ILSs) use a relational database (RDBMS) to store this and any associated data to another paradigm (2) where cataloguing is done in RDF (using the BIBFRAME data model) and ILSs use a triple store to store the necessary information. I’d say that the first assumption isn’t necessarily true (there are quite some ILSs where data is not stored using an RDBMS and at least some libraries where cataloguing is not done by creating or editing MARC records but using another metadata format that can be converted to MARC if so desired). And I’d also say that the suggestion for a new system paradigm is far too narrow and might even hinder innovation by mandating too strong requirements on which technology to use. There is, for instance, an emerging technology called graph databases that allow for interesting ways of analysing the data in the graph, including finding the shortest path between two nodes, finding “islands” (graphs or subtrees not connected to any other part of the graph) or loosely connected subtrees (e. g. subtrees that are connected by only one edge). If we mandate the use of a triple store, a vendor would not be able to use this technology and thus would lose the possibility to implement interesting statistical functions. In my opinion, a call for tender should be as technology neutral as possible (at least with regard to system internals).
So what should a call for tender contain instead?
My take is, that it should specify the desired functionality. After all, the interesting thing is what we want the system to do (or at least what we want to do with the system). A non-exhaustive list of things I can think of for a LinkedData-based system would be :
- The system must be able to import library data in the following formats:
-- MARC 21 (perhaps in different flavours)
-- RDF using the BIBFRAME data model
-- RDF using the RDA data model
- The system must be able to export data in the following formats:
-- MARC 21 (perhaps specifying a flavour)
-- RDF using the BIBFRAME data model
-- DC-XML (for use with OAI-PMH)
- The system must support the following machine import and export interfaces:
-- OAI-PMH (synchronising both ways)
-- Z39.50 ;-)
-- W3C WebSub (to ensure the system is “webby”)
- The cataloguing module must allow cataloguers to:
-- Connect titles and authorities to other titles and authorities residing inside the local system (e. g. connecting a publication to its successor or to its author; connecting an author to her/his place of birth and the place of birth to the country it’s part of)
-- Connect titles and authorities to other titles and authorities residing in online databases (perhaps mandating a list of search interfaces the system must support)
- The system must allow administrators to
-- Configure the data input forms, e. g. restricting which authority files the cataloguers can use
-- Import and export data input configurations and metadata profiles in a standardised format (possibly stating a list of such formats, e. g. SHACL, ShEx, JSON Schema, XML Schema)
-- Seamlessly include third party databases (IEEE Xplore, EconBiz, PubMed, …) into end user search
- The system must allow end users to:
-- Search in both library internal content and third party databases using a single, united search interface
-- Export bibliographic citations into third party citation management systems
-- Subscribe to an RSS/Atom feed for new content matching a custom search
To me, this approach would have the advantages that
1) The customers need to think of what they really want the system to do
2) The vendors can concentrate on implementing this functionality in a way they are comfortable instead of having to focus on new technology.
What do you think?
 Some of those and much more can be found in the 2013 list of use cases and requirements: http://bibframe.org/documentation/bibframe-usecases/
*** Lesen. Hören. Wissen. Deutsche Nationalbibliothek ***
Dr. Lars G. Svensson
60322 Frankfurt am Main
Telefon: +49 69 1525-1752
Telefax: +49 69 1525-1799
mailto:[log in to unmask]