Playing catch up with this one too, so apologies if this has been
suggested.
Assuming that the server supports resultSetId's another approach (I'm
not arguing worse or better just putting it on the table) might be to
have a new operation
getRecordMetadata(resultSetId, StartRecord, NumberOfRecords, Schema)
which returns the metadata records of records StartRecord to
StartRecord+NumberOfRecords in schema Schema. Response would look
remarkably like the SearchRetreiveResponse.
Matthew
>
> -----Original Message-----
> From: Robert Sanderson [mailto:[log in to unmask]]
> Sent: Tuesday, June 03, 2003 3:56 PM
> To: [log in to unmask]
>
> > What we wanted was for the metadata elements for the record
> to be mapped
> > into the schema, not the elements of the record. The use
> of a distinguished
> > schema name to accomplish that distinction is bothersome.
> (I'm not saying
> > absolutely unconscionable. I'll leave that to Mike. :-)
>
> I'll leave it to Mike as well, but I'll agree when he does it :)
> I didn't intend rec to be somehow semantically special as a schema. In
> fact, I recall discussing with Adam, possibly not on the
> list, about how
> to include it as a namespace in other schemas and how that could be
> requested ... eg dc+rec for a simple data + metadata schema.
>
> > I think we want a more explicit way to say that we're
> requesting record
> > metadata elements.
>
> Yes.
>
> The main issue, as I see it, is the number of available
> metadata schemas
> (1 or many).
>
> If we limit it to one, all we need to do is include metadata and
> metadataXML (cf recordData and recordXML) in the record
> object, using the
> same rule as recordPacking. There's no added complexity to
> the request,
> no new diagnostics.
>
> But one size never fits all.
>
> So if we have multiple metadata schemas available, then we
> need a way to
> request the one we want. We also need a way to say that it's
> unknown or
> not available, and what schema it actually is.
>
> So it would end up something like:
>
> <record>
> <metadataRecord>
> <metadataSchema>
> <metadataData> or <metadataXML>
> </metadataRecord>
> <recordData> or <recordXML>
> <recordPosition>
> </record>
>
> Plus a <metadataSchema> element in the request. For
> compatability, I'd
> think that if not present it would default to 'don't send me any
> metadata', and present but empty would be default metadata schema.
>
> Not overly complex and certainly better than the alternatives of no
> metadata, one set metadata schema, or server's choice of
> metadata schema.
>
> What's /really/ a problem is if you then want to sort based on the
> metadata rather than the record. For example, retrieving a
> set of records
> sorted by record length or last modified time. I don't have
> a good answer
> for that beyond putting a new boolean typed element into the sort key
> definition.
>
> Rob
>
>
> --
> ,'/:. Rob Sanderson ([log in to unmask])
> ,'-/::::. http://www.o-r-g.org/~azaroth/
> ,'--/::(@)::. 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
>
>
|