[open-bibliography] Proposed definition for /book/book

Thu Jul 1 16:24:06 UTC 2010

For what it's worth, I think people can err in setting up a
hard-and-fast system. The library world loves binary systems and
deciding things at the start, not down the road.

For example, a book either does or does not get the LCSH "Man-Woman
Relations." This is strange, as about 90% of literature involves
man-woman relationships to some degree or another. (Tagging, which
allows for degree, is an improvement in this respect.) I would argue
this sort of relationship has more to do with the physical constraints
of catalog cards than anything else.

This idea was extended to FRBR, which involves another set of binary
understandings about relationships. The real world is much more
nuanced. We have, for example, two basic texts of Hamlet, with neither
one being original or the "parent" of the other. The history of works
on LibraryThing is replete with more recent examples. The relationship
between two books is a matter of degree, perspective and intended use.

LibraryThing's solution is to have two layers. The first is a catalog
layer, in which things are stored atomically and almost without
schema. The second is a global layer, which either bubbles up from the
catalog data or is imposed upon it. Either way, the global layer is
understood to be a convenient abstraction, and susceptible to change.
We should remember that classification is largely about imposing
structure for a purpose, not finding structure and telling the truth
about it.

The digital world allows for a richer conception. So it's a shame
library science and the semantic web people are too often drawn to
arrows, trees and buckets.

Tim