> possible for some sets of records not to have persistent URIs and their
> "persistent" URI would be their position in a result set, but these would
> probably not be useful web resources.
I disagree. The Archives Hub is a useful web resource, but each record
within it has no persistant unique identifier beyond a search on the EADID
field. Value of the content has nothing to do with the design of the
architecture, especially as content providers are only infrequently the
architecture designers as well.
(In my defence, the AH was designed before persistant URIs became a big
thing. A redesign would include them, but hindsight is ever 20/20)
> >B) Why not just use the existing mechanism of CQL?
> ...&query=identifier%="foo1" could be a persistent URI, though in some
> services it might return more than one record.
Sorry, rec.id exact "foo1". This will never return more than one record.
> Allowing any URI to identify a record allows for example the database query
> URI to be different from the full record URI to the extent that they came
> from totally separate servers, or of a different form e.g. an URN, NBN, DOI.
There's nothing preventing the rec.id from taking the form of a URI.
> > C) As record metadata, it belongs with record metadata, however we handle
> I agree, record metadata would need to contain an element that meant "this
> is the persistent URI for this record". I believe it would be better to
> fetch records on the basis of this element rather than record x in result
> set y.
I don't think that anyone disagrees with this position, it's just how it's
implemented within SRW. As I've already argued, we have the ability to
request records by a persistant unique identifier already, so I don't see
a need for additional specific parameters.
What we do need is a section for metadata about the record, which would
include this among other information. Perhaps METS, but that would need
to be investigated as I doubt there is an element for Result set relevance
score available in it so it might need to be extended.
I should also quantify the use of the word 'local' in the description of
the rec.id index. This is a local to the database identifier, rather than
a guaranteed globally useful identifier, though the possibility that it is
this as well is not ruled out.
For example, imagine multiple databases of XHTML files spidered from the
web. One database might have the identifier as a simple incrementing
value, while six others might have the URI it was spidered from as the
identifier. For those six databases, the identifier universally denotes
,'/:. Rob Sanderson ([log in to unmask])
,'--/::(@)::. Special Collections and Archives, extension 3142
,'---/::::::::::. Twin Cathedrals: telnet: liverpool.o-r-g.org 7777
____/:::::::::::::. WWW: http://liverpool.o-r-g.org:8000/
I L L U M I N A T I