At 02:46 PM 4/14/2010, Karen Coyle wrote:
Quoting John Attig <[email protected]>:
"Signatory to a treaty, etc." is therefore one of several identifying
elements necessary to distinguish between different treaties (works).
This is independent of the role of the signatories as creators of the
work. Note in the authority record example, that Australia and United
States are also identified as creators -- and presumably, in a linked
data environment, this would be encoded as a relationship to the
corporate bodies.
So is it expected that "signatory to a treaty" will be represented by
an entity or a string? For example, would you expect a signatory to a
treaty to be retrieved by non-preferred (or earlier/later) forms of
the name?
I hesitate to venture into the question of how to retrieve by variant
data; this is a problem that (it seems to me) to exist in any set of
hierarchical data where it is assumed that the variant terminology
that is defined at a higher level in the hierarchy (or, in this case,
simply in a different place in the data) will be inherited by all the
subordinate levels in the hierarchy.
I am also reluctant to try to anticipate how the sort of string-based
retrieval that we currently employ might work out in a linked-data environment.
That said, I would think that the variant terminology would come into
play through the linked entity for the Creator, rather than through
the "signatory" attribute that is there as part of the access point.
Also, if a system could precoordinate the string using an entity and
relationship (thus having it appear as it must in RDA displays and
indexes) would that be acceptable?
I suspect so. The one caveat is that the text strings used to name
the entity might not be the same as the text strings used in the
"signatory" attribute. I don't believe that is the case here, but it
is at least possible (if unfortunate).
I would also note another example of the entity vs. attribute
problem. There are a large number of elements (attributes of many
different entities) for places of various sorts; there is also a
Place entity. The place attributes could have been structured as
relationships to the Place entity; in this case, the reason that they
were not lies in the FRBR model, which deliberately and explicitly
chose to treat them as attributes. This -- along with the other
anomalies that you are identifying -- are candidates for changes in
the underlying models.
John Attig
Authority Control Librarian
Penn State University
[email protected]
