Jonathan- I need to correct you on one point below: I never suggested that one doesn't need records for WEMs; in fact, my intent in this post was to posit an environment where result sets could be electively reified into explicit records defined by the query, which in turn could have explicit statements made about them.
Bradley P. Allen Gmail: bradleypallen Twitter: bradleypallen Web: bradleypallen.org Cell: +1 424 634 0870 On Wed, Sep 15, 2010 at 7:38 AM, Jonathan Rochkind <rochk...@jhu.edu> wrote: > I've written about the set-theoretic view too, also crediting Svenonious: > > http://bibwild.wordpress.com/2007/12/07/frbr-considered-as-set-relationships/ > (that one is short and sweet and recommend it) > > and > http://bibwild.wordpress.com/2010/03/17/notes-frbr-wemi-entities-physicality-interchangeability-merging/ > (less sure about what I wrote here, and it's longer and more confusing, only > if you want some additional reading.) > > > Personally, I don't think it results in what Brad Allen suggests, that if > you use the set theoretic view, you don't ever need to have records for the > more 'abstract' entities/sets. Sometimes you need to make an assertion > about the set itself -- that is, record an attribute for a Work. And then, > sometimes you don't, you can algorithmically extract it from the members. > This is really true whether you think of the entities as sets or not -- > although thinking of them as sets does, I think, make this (and other > things) more clear. > > But doesn't magically make everything simpler, I don't think it really > changes anything fundamental, it's just a more clear way to think about it. > Even thinking as sets, you might need to make assertions about the > higher-level abstract sets, and if so those sets still do need to be > 'entities' (ie, have identifiers and/or records). > Jonathan > > Bernhard Eversberg wrote: >> >> Diane I. Hillmann wrote: >> >>> >>> I'd like to point those of you interested in the more technical side of >>> discussions on RDA and FRBR to a conversation going on in the public Linked >>> Library Data discussion list, starting here with a message from Karen Coyle: >>> http://lists.w3.org/Archives/Public/public-lld/2010Sep/0049.html >>> >>> In one of the responses to that post is a link to a blog post by Brad >>> Allen about "Faceted classification and FRBR" >>> (http://bpa.tumblr.com/post/10814190/faceted-classification-and-frbr). >> >> >>> >>> ... His post is well worth reading. >>> >>> >> >> Indeed it is. Read that, everybody! >> Being a mathematician, the set-theoretic view (based on >> Elaine Svenonius) appeals to me a great deal. Allen's (Svenonius') >> reasoning is very sound, and down to earth when considering our >> biggest predicament, our legacy data. It also goes well with the >> paradigm of all known retrieval systems, based as it is on the idea of >> the "result set", resulting from a query that uses attributes of various >> kinds, and all of them can be viewed as attributes of items. Certain >> combinations of attributes define subsets of items - some of these >> subsets can be called "manifestations", "expressions", or "works". >> >> The identification of the work, however, remains the open question. >> It has to be done somewhere. Traditionally, it was pinned down by the >> "uniform title", and many of our records have this as a distinctive >> attribute. Add to it the language, date, form, medium, numeric >> designation, key, coordinates, etc. - and you single out the >> crucial subsets that FRBR views from the top down. >> >> Certainly, FRBR's entity-relationship paradigm won't get entirely >> lost if we pursue this approach. It remains useful as a kind of >> "Weltanschauung", but the set-theoretic paradigm is, I think, >> more helpful as well as more versatile. It avoids the difficulties >> with the blurred boundaries of the FRBR entities as we encounter >> them in practice. The set-theoretic view can effortlessly describe >> many more entities, all of them subsets of items sharing certain >> common attributes (or "facets", in Allen's terms). >> >> And anyway, Brad's basic point of tying the process of cataloging to the >> item, as we always did, is perhaps the only realistic one under the >> circumstances given. Perhaps it will remain more pragmatic to use >> the item level concept only within the ILS (where it is needed for >> circulation) and retain the level just above it, presently called >> "manifestation", as the focus of attention in cataloging and the >> creation of records (as long as this concept is still necessary and >> useful.) >> >> What does it all mean, in practice? >> It is not a new idea: Extend the "uniform title" element >> by a few subelements (see above), and we're done. It may then >> be useful to build a better uniform title authority data pool, >> ideally an extension of the VIAF pool, but much of the data is already >> there, somewhere. And not every item, as Mac always insists, is actually >> in need of this much attention. >> >> B.Eversberg >> >
