Very roughly. And I think that the examples make it very hard to
understand how this might work in a machine-actionable environment.
That's true with all of the relationships that are defined by FRBR.
It does occur to me that the
expressionA --> expresses --> workX
could possibly be repeatable, now that Thom has weighed in on the VIAF
treatment of some uniform titles. You could have a statement internal to
your database that uses internal linking (and is very efficient for that
reason) but you could have another statement that is based on authority
files that is more useful for sharing. I could also see having one
manifestation identifier using the ISBN where it is available even
though that will not work for all of your data. In other words, we can
do something like we do today and throw in as many identifiers as we
have because they *might* help someone make the connection. While an
identifier that was created internal to a database during record
correction, where a human actually decided "this expresses that" would
be more likely to be accurate, it might be worth having less precise
identifiers (like ISBNs, which have a certain level of error in our
data) because those create hints that others might be able to work with.
This is similar to what we tend to do in de-duping of bibliographic
records: you look at all of the identifiers, but do additional checking
because you know you can't rely on them 100%.
On 5/16/12 6:45 AM, Mark Ehlert wrote:
> Karen Coyle<[log in to unmask]> wrote:
>> If you look at FRBR you see that what connects an expression to a work is a
>> expressionA --> expresses --> workX
>> RDA does not define that link, but I think that link is going to be between
>> an identifier for an expression and an identifier for a work.
> Chapter 17 seems to fill that gap, if roughly.
[log in to unmask] http://kcoyle.net