Simon, et al.:

I like this, particularly the strategy of de-generalizing the agent 
types.  I think if this flies, and includes the recognition that this 
information *is indeed* describing another entity, and can be ignored in 
dumb-down, I might be persuaded to change my vote for these items 
only.  The rest of the agent proposals are another kettle of fish entirely.

Diane



At 04:21 PM 01/26/2000 +0800, you wrote:
>As discussed in Tuesday conf. call, here's a revised proposal for a
>third category to be added to the "Principles" document:
>
>==========
>3.  Agent-type.  These qualifiers specify the type of the entity which is
>identified by a value of the Creator, Contributor, Publisher elements (CCP).
>This is typically a person, organisation, instrument, or other entity
>capable of creative acts or control of real or intellectual property.
>While this type is a property of another resource (given by the value of
>the CCP element), it is considered useful for discovery of the present
>resource (i.e. that one which is the subject of the DCMES description)
>by many communities.  If a value qualified with an Agent-type is encountered
>by a client that does not support this category of qualifier, it can be
>ignored without harming the client application.  The definition of
>each Agent-type must be clear and publicly available.
>==========
>
>A few other non-substantial changes will need to be made to the principles
>document if this is accepted (change "two" to "three" in a couple of 
>places, etc.).
>
>Some additional comments:  using the name Agent-type rather
>than Object-type or Resource-type is deliberately conservative.  Those
>of us who have been involved in datamodel discussions will understand
>that agents are just a subset of "resource", and that, in general, the
>datamodel that underlies DCMES is likely to require more general typing,
>across the values of all the DCMES elements, and covering the full scope
>of DC resources.  But at this stage the only coherent proposal for typing
>concerns the CCP elements, and the concept of typing is only widely
>understood and accepted for agents.
>
>I see no real risk in restricting the category to Agent-types at this
>stage, since any future generalisation will just add a superclass to
>the hierarchy, and Agent-types is a subdivision of this which is likely
>to continue to be useful as it distinguishes a coherent group of resource 
>types.
>
>Of course this would not preclude the possibility of each member of the
>Agent-types also belonging to other sub-classes of Resource-type -
>person and organisation are both legal-entities, person and instrument
>are both physical-objects, etc. - it all depends on why you are asking!
>
>I believe that this concept of each type having membership of multiple
>classes is supported by RDF, though the terminology may be different.
>Carl has waxed eloquent on why we might make it easier on ourselves by
>shifting to RDF terminology real-soon-now, and he may have a point.
>But for the purposes of the present exercise, the prose definition in
>the paragraph above seems to me to be at least as good as the prose
>definitions of Value Encoding Scheme and Element Refinement.
>
>--
>Best                    Simon



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%