I tend to agree with the sense of what Stu is saying here.
For the (Agent) elements, people seem to want to know the
type of the thing (party, process) which is the agent.
The concept is not new in discussions around qDC and the DC datamodel,
it was just a bit alarming the way it was introduced here, complete
with new tokens, in a kind of fait-accompli data-model in the context
of the qualifier voting form!
What really concerns me, however, is that there are important implications
for the data-model, and the 1:1 principle, if we include Object Type as an
explicit class of qualifier for the (Agents). We must understand these
implications properly before we go ahead with these qualifiers.
IMO the root of the problem here is that there has been a tendency
to use the term "agent" (creator, contributor,publisher) to refer to
both the nature of the thing identified by the value,
and the nature of the relationship between the thing identified by the value
and the DC-resource being described.
I believe that a cleaner model recognises that there are a class of
things (parties+processes) which exist in their own right, and which
may act as "agents" in respect to dc-resources. The dc-property
"agent" takes values which have datatype "party|process", just as the
dc-property "coverage" takes values which have datatype "place".
The nature of the thing (person, organisation, instrument, event,
if you like), is quite distinct from the nature of its agency in
this instance (creator, contributor, publisher, all the other roles).
The things are not intrinsically authors, hairdessers, etc, they are
persons who acted as authors or hairdressers in these instances.
In fact, since DCT1 includes party and event, these "things" may be
dc-resources in their own right, and the information which is given
in the voting proposal as the agent's Object-Type is its DC.Type.
The 1:1 principle applies, and the situation can be disentangled quite neatly.
The (Agent) elements are clearly a special class of *relators*, which
connect members of a special class of resources (parties+processes) with
dc-resources.
But even in cases where the "thing" which has an agent relationship
to the dc-resource does not have its own dc-description, the same
principle should apply - the information called Object-Type here is
information associated with the thing identified by the value of the
agent element, and NOT with this instance of the (Agent) element itself.
For these reasons, in the ballot I've voted to reject all the Object Type
qualifiers (and this model has also influenced my vote on Role and its encoding
schemes). But for the third time I find myself in a minority of almost one,
so I'm trying to understand why I'm so far out of the mainstream. Can anyone help?
--
Best Simon
|