A colleague who is part of my library's little Linked Data learning community pointed the members of our group to the Failure thread. Karen commented on the inbound application of Linked Data to a given user experience and James mentioned that WorldCat Identities, while cool, is not necessarily genuinely useful to patrons.

We have a functional implementation of enhancing our library catalog in exactly the manner Karen describes at Google (albeit on "full record" pages rather than search results presently). This runs against our full production catalog, but not the production web address [1]. See, for example:

Just for Fun by Linus Torvalds

See the "Data from other sources" section below the MARC bib data. Here we can assert that the author of this book about Linux (according to the subject headings) is also the creator of it (according to the info card). Additionally, he appeared in films that may be in our catalog on the same topic. The MARC data alone does not have that context for the work. To the extent that someone is searching our catalog and we already have their attention, we can help them see our collections in a greater context.

So this is possible and we have made our reference implementation for how to do it available [2]. It is not great code, but it demonstrates one way to accomplish this task (comments welcome!). We also gave the high level overview of this prototype as part of the IGeLU/ELUNA webinar series and in conjunction with Ex Libris' Linked Data Collaboration program [3].

Now, because this is a BIBFRAME list, the one thing to note here is that this is not based on having an RDF representation of our entire catalog. This is possible simply because Ex Libris provides us with the first URI (or at least the name authority IDs for main/added entries and subjects) to kick off the Linked Data crawl. One of the things that seems to get lost in the Linked Data and libraries discussions is the fact that embracing Linked Data does not need to be a stark all-or-nothing proposition.

Personally, my greatest point of skepticism about BIBFRAME in particular is that it fails the killer app test. As Jeff T. noted in the failure thread, MARC actually solved a profoundly important and (near) universally useful problem that libraries faced if you understand it from the perspective of its original design purpose: it was a data *transmission* format primarily and evolved into a cataloging format later. The problem it solved was how to share records to gain efficiencies in the collective, cooperative creation of bib data sets.

While I am eager to embrace the benefits of RDF that Simon pointed out and would love a BIBFRAME'd version of our catalog (assuming it evolves and continues going through phases of progressive improvement), I worry that BIBFRAME has no such practical and pressing need. As our prototype above demonstrates, we can embrace the web of data and try to enhance library discovery from remote sources without necessarily requiring a complete transformation of the catalog.

[1] Our primary catalog drops the "-ld" portion of the URL. So it is running against the full production data set, but only if you know the URL to start. We are doing this so we can fully evaluate it but without prematurely rolling out unexpected consequences to patrons.