In general, the reciprocity of relations does not suggest that both should
be recorded, just that either could be. Almost by definition, one relation
has to be established after the thing it relates to already exists, so
typically the pre-existing metadata would not carry the pointer.
David
At 11:55 PM 1/21/99 +0000, Alex Satrapa wrote:
>Explicitly stating the metadata author-agent using the Relation field sounds
>sane to me, though I'd suggest in this light that metadata authorship is
>important enough for validating metadata that it should have a field of its
>own, perhaps DC.MetadataAgent. Whether this actually contains information, or
>merely points to other information, is a point for debate.
>
>For decent quality verification of metadata authorship, the record for a
>particular metadata author-agent would contain links to all the metadata that
>this author-agent created. That, or we should have a standard means for
>recording digital signatures for authored metadata. So the DC.MetadataAgent
>could contain, say, the link to the author's record and a signature based on
>the rest of the metadata. I can tell you right now, it's less trouble to
>implement verification with signatures than cross-links.
>
>Using the DC.Relation of IsMetadataAuthorOf/HasMetadataAuthor is going to
>blow out your data storage. "Is MetadataAuthorOf" is one of those relations
>that can be explored by finding everything that a particular person is listed
>as being "HasMetadataAuthor". Unless you only write 500 metadata resources in
>your life, using two-way linkage is going to be very expensive, in terms of
>initial storage and continual updates when the location of a resource
>changes. You're going to need a fast RDBMS just to manage your
>"IsMetadataAuthorOf" table.
>
>If it's possible to keep a global directory of authors, then digital
>signatures on metadata would be a more sensible option, IMHO (because it
>means less work for the programmers, which means higher quality software).
>
>Regards
>Alex Satrapa
>
>David Bearman wrote:
>
>> Dear Andrew,
>>
>> I thought I was explicitly NOT proposing that DC metadata do this work.
>> Rather DC metadata can have relation link to a metadata record created by
>> another community in which the agents (who might be owners/holders) are
>> documented.
>>
>> I believe that the Recordkeeping community (archives for those readers not
>> aware of the RMK metadata work) is the most likely community to develop a
>> full metadata semantic for agents, just as the rightsholding community
>> (INDECS, for example) is most likely to develop a full metadata semantics
>> for rights agreements or deals.
>>
>> At 08:35 AM 1/22/99 +1100, Andrew Wilson wrote:
>> >David
>> >
>> >Forgive me if this seems naive but aren't custody/ownership and other
>> >questions relation to provenance more properly the domain of
>> >recordkeeping metadata, and nothing per se to do with discovery
>> >metadata? Aren't you trying to get DC to do a lot more than was
>> >originally intended?
>> >
>> >Andrew Wilson
>> >National Archives of Australia
>> >Email: [log in to unmask]
>> >Ph: +61 2 6212 3694
>> >Fax: + 61 2 6212 3997
>> >
>>
>> David Bearman
>> President
>> Archives & Museum Informatics
>> 2008 Murray Ave, Suite D
>> Pittsburgh, PA 15217 USA
>> Phone: +1 412 422 8530
>> Fax: +1 412 422 8594
>> [log in to unmask]
>> http://www.archimuse.com
>
>
>
>--
>Alex Satrapa
>tSA Consulting Group Pty Ltd.
>Canberra, Australia
>
>
>
>
>
David Bearman
President
Archives & Museum Informatics
2008 Murray Ave, Suite D
Pittsburgh, PA 15217 USA
Phone: +1 412 422 8530
Fax: +1 412 422 8594
[log in to unmask]
http://www.archimuse.com
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|