[OAI-implementers] Requesting a part of a record possible wit h OAI-OMH?

Hussein Suleman hussein@cs.uct.ac.za
Thu, 22 Jan 2004 17:05:54 +0200


just wanted to reply to one specific issue - where you ask if two new 
"Part" verbs should not be considered for a new OAI-PMH version.

the current OAI-PMH is very popular largely because lots of prospective 
features were thrown out (during development of the protocol) to keep 
things simple - thus making implementation easier, faster and more accurate.

the principle adopted was that PMH transfers data - raw data - and thats 
it. anything more interesting than raw data transfer should be viewed as 
a different "value-added" service provided by a repository or service 
provider. in all likelihood, if there ever is a newer version of PMH, it 
will still do only raw data transfer :)

i should also add that Tim's solution (DC records to describe data, with 
pointers to actual data) is something we implemented at VT a few years 
ago for data similar to yours (large files of XML-encoded numerical data).


Joachim Wackerow wrote:

> Sorry about the late answer.
> Thank you very much for all your comments and the summary, your input 
> helped me very much.
> Here is the summary from my perspective:
> I understand, that the actual version of OAI-PMH can request a whole 
> record (XML document), but not a part of record (a fragment of XML 
> document).
> The proposal from Naomi (1) is a workaround, which will work if it's 
> only necessary to request only a few defined parts of a record.
> Jeffs proposal (2a) has the attractive idea to request a part of a 
> record by the means of XPATH. In my eye's it's a drawback to depend on a 
> external service for the reason, that the "main" protocol (OAI-PMH) is 
> lacking this feature (nothing against external services). I would like 
> to think more about a general solution (for DDI). But nevertheless, 
> Jeff, thank you very much for your offer.
> Hussein's proposal (2b) of a new protocol on top of or parallel to 
> OAI-PMH sounds very good. The proposed way to add this new protocol 
> seems to be very flexible also for other new protocols, it would be a 
> way of a general extensibility of OAI-PMH. The question seems to me: is 
>  "requesting a part of record" not a candidate of a new version of 
> OAI-PMH? "GetRecordPart" and "ListRecordParts" were the tools I'm 
> looking for.
> The actual version of OAI-PMH has the intention of requesting whole 
> records (primary use of harvesting protocol). The question of requesting 
> a part of a record by the means of XPATH will arise with large XML 
> documents. The service provider could be interested only in a part of 
> the information. Large XML documents have usually DTDs/Schemas with a 
> deeply nested structure and a lot of elements or have repeatable 
> elements which could be repeated very often.
> The size of DDI files reach often 500-1000 KB. The description of a 
> social survey consists often in several hundred questions/variables and 
> in an extensive general study description.
> I see one problem with the idea to request fragments of a XML document 
> stored in a record: the client would not be able to validate the 
> response, because itís only a fragment of a XML document. A better 
> solution would be to embed the fragment in a frame of a valid XML 
> document (like a "condensed version" of a XML document). That would be 
> dependent from the XML DTD/schema (the question would be, which elements 
> are required).
> Right now we are just thinking about using OAI-PMH to offer DDI-files. 
> Perhaps there is another way to offer XML files for service providers, 
> Hussein mentioned SOAP.
> I will try to give the DDI community 
> (http://www.icpsr.umich.edu/DDI/users/listserv.html) a hint on this 
> discussion thread.
> Achim
> _______________________________________________
> OAI-implementers mailing list
> List information, archives, preferences and to unsubscribe:
> http://oaisrv.nsdl.cornell.edu/mailman/listinfo/oai-implementers

hussein suleman ~ hussein@cs.uct.ac.za ~ http://www.husseinsspace.com