Simon Pockley wrote:
> 
> It seems to me that `version', `format' and `based on' -   are 
> dealt with under `Source' even if the relation is one way.

We really need to clear this one up.  It is becoming a FAQ.  
I could quote several messages to this list which followed 
Helsinki which clarified the proposed usage of "Source" with 
respect to "Relation", and my message of Sunday also spoke 
to this, but perhaps I'll try again.  

Yes - "Source" as an element independent of "Relation" is 
semantically ambiguous.  In the full RDF type framework, 
information about all Source's for a resource would be 
retrievable through following a chain of Relation's of 
the types indicated above, thus as Simon Pockley correctly 
observes, Source seems to be a subset of Relation.  

However, it is likely that it will be a long time, maybe never, 
before a complete RDF framework has been put in place with 
DC metadata for everything which would allow this to be a 
complete solution.  And even then, short of elaborate caching 
arrangements, it may not be very efficient.  Thus, we need 
an alternative way to allow _information_about_an_ancestor_, 
which is nevertheless relevant to _*discovery*_of_the_present_resource_, 
to be attached to the _metadata_of_the_present_resource_.  
It has been proposed that the "Source" element be used 
explicitly to provide this facility.  

Thus, metadata in DC-Source is in fact metadata for a  
resource(s) other than the present one, but which is 
privileged because it is considered to be important 
for resource discovery, and is thus _embedded_ in the 
metadata set for the present resource.  The fact that 
the metadata actually refers to a different resource is 
flagged by inserting the key-word "Source" in the 
attribute identifier for the metadata element.  

Thus, any resource can have 
(1) its own metadata, relating primarily to the current instantiation, 
DC.Creator, DC.Date, DC.Format, DC.Identifier, etc
(2) some additional metadata which is related to ancestor(s), 
DC.Source.Creator, DC.Source.Date, DC.Source.Format,
DC.Source.Identifier, etc
- the embedding mechanism.  
(3) a pointer(s) to ancestors
DC.Relation.IsBasedOn, DC.Relation.IsFormatOf, etc.  
- the ideal 1-to-1 mechanism.  
In principle the information stored in (2) could be 
discovered by finding the metadata sets referred to in (3), 
but in some cases, although it is less elegant (it is a 
deviation from the 1-to-1 rule) it may be more practical.  

Note that the syntax used in (2) here implies that Source 
_must_ be qualified.  In the litest version of DC 
- no qualifiers - DC.Source and DC.Relation would sometimes 
have the same value - an identifier for an ancestor, 
though for DC.Source it may be a more distant ancestor - 
but in all other cases the usage would be different.  

Is this clear, and does it accord with other people's 
understanding?  

---------
(This is substantially a repeat of a message already 
seen by subscribers to dc.relation.  Further discussion 
will be on meta2.)
-- 
__________________________________________________
Dr Simon Cox - Australian Geodynamics Cooperative Research Centre
CSIRO Exploration & Mining, PO Box 437, Nedlands, WA 6009 Australia
T:  +61 8 9389 8421   F:  +61 8 9389 1906   [log in to unmask]
http://www.ned.dem.csiro.au/SimonCox/