W3C home > Mailing lists > Public > public-lld@w3.org > March 2011

Re: reconciliation of disparate models - Karen, Ross

From: Thomas Baker <tbaker@tbaker.de>
Date: Sun, 13 Mar 2011 19:35:49 -0400
To: public-lld@w3.org
Message-ID: <20110313233549.GA2464@octavius>

    Jeff, it made sense to me up until this part: What part of it are
    you agreeing with? That they sub-class books to Expression? Or are
    you referring to something else?

    I ask because most interpretations of FRBR have "published things" as
    Manifestations, which puts FABiO in a distinct minority. However, it
    does make them an interesting case for interoperability of different
    interpretations of FRBR.

    I also think it isn't just FRBRer in OWL that is rigid. FRBRer
    in OWL is reflecting the will of the FRBR working group, and RDA
    is aligned rigidly with FRBR because of the insistence of the RDA
    working group. It's not the technology that is causing the rigidity --
    it is the will of the developers of these standards that is causing
    the rigidity. While the FRBR report appears to be more open, the FRBR
    committee's interpretation of the FRBR report (and some of them were
    among the authors of the report) is what has guided FRBRer.

    I guess what it comes down to is that I don't see this as a clash
    of technologies but as a clash of communities. I'm assuming that
    the technology either exists or could be developed to do whatever
    we want -- the difficulty is that we don't all want the same thing.


    I obviously feel pretty strongly about this issue, but I also often
    find myself taking seemingly contradictory positions on it, as well.

    One extremely grave mistake, I think, is that FRBR(er) will offer the
    world all these relationships that people are going to be really
    interested in using ("the resources are different editions of the same
    thing", "this book is a translation of resource in dbpedia", "this
    musical was inspired by this book", etc.) but only available on
    FRBRer's (very strict) terms.

    To have to build up all this complex W-E-M and W-E-M-to-W-E-M
    scaffolding simply to say that this MP3 of the original cast recording
    of "Les Miserables" has some relationship to this copy of a book (in
    English) by Victor Hugo is going to be a non-starter.  Even *within*
    libraries this is going to be a non-starter, much less the outside
    world.  Add to the mix that it's uncertain we have enough data, at
    hand, to make all of the required entities.  Top it all off with the
    fact that only a cataloger is going to know if "Les Mis" the book and
    "Les Mis" the musical are two separate works (and even they might
    disagree about whether or not the translation of the book should be a
    considered a separate work or not).

    All of this is to say, producing extremely constrained (and highly
    complex) data models that require a high degree of specialization to
    use will counteract any advantages of adopting RDF in the first place.
     I don't buy the argument of powerful inferencing and whatnot, because
    it would still just be our same old data (albeit in a different
    package) since we can't include or mesh anybody else's data.

    I would also argue that FRBRer (and RDA) place FAR too much importance
    on the Group 1 entities.  The (absolutist) emphasis on WEMI begins to
    overshadow the thing being described.  The model forces itself to the
    forefront, even when it's not necessarily needed.

    This is not to say that I don't think FRBR is valuable (and good!), I
    absolutely do.  However, it should not get in the way or be
    prioritized over any other aspect of description.  If you want to
    describe this resource /as a book/, you should be able to link it to
    the generalized ideal of the story and the mini-series adaptation
    without needing to explicit build out the WEMI models.  At the end of
    the day, I'm describing a book, not a Manifestation and Expression
    that are related to a Work.

    I mentioned my concept of "implied FRBR" on the list and I think
    Jeff's superclass proposal also serves the same purpose:  we should be
    able to make Group 1-ish sorts of assertions about things that
    (implicitly) contain Group 1 entities because /we know they're in
    there somewhere/.  We shouldn't have to know where or what, exactly,
    they are.  What you have to give up with this, however, are
    entailments that make, say, a BIBO citation both a Manifestation and

    I think only by addressing FRBR as an abstract entity (either by a
    generic "Endeavo(u)r"-style super class or through implication)
    instead as the foundation can we possibly hope to achieve any
    non-library interoperability/adoption.

    Also, I completely agree with Jon, re: blank nodes:  if you are so
    uncertain of the identity of data you are sharing that you cannot
    provide a reusable identifier, why are you sharing the data at all?

Tom Baker <tbaker@tbaker.de>
Received on Sunday, 13 March 2011 23:36:29 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:27:43 UTC