[OAI-implementers] Resumption Token usage for recovery question

[Xiaoming Liu]

On Wed, 13 Feb 2002, Alan Kent wrote:

> Are resumptionTokens therefore guaranteed to be different for each
> chunk? (I cannot find that in the spec). I thought OAI allowed a
> server to reuse the same resumptionToken for an original request
> and followup requests. That is, the resumption token could identify
> an original request, but not which incomplete-list of records from
> that request to return.

I guess this is an implementation issue, therefore it's not covered in the
OAI-PMH. A data provider may have a distinct resumption token per
request, or not. I have seen both implementations.

> 
> If OAI does not guarantee a distinct resumption token per request,
> then its not safe to use them more than once. If a unique token is
> guaranteed per request, then yes, I guess they could be used to
> recover (if you attempt recovery immediately - before the expiry
> on the token runs out).

At this point I think harvester must trust data providers will do the
right thing. If data provider feel it is unsafe, it will return a http
400, so harvester must restart the orignal request again.

For example, due to the network failure, the harvester may present same
resumptionToken twice, if harvester is OK with it (like stateless
implementation of resumptionToken), it will continue the remaining
records; otherwise it may return a http 400.

Regards,
liu




> 
> Are resumptionToken's implicitly assumed to be different for each
> incomplete-sublist of a complete-list?
> 
> Thanks,
> Alan
> _______________________________________________
> OAI-implementers mailing list
> OAI-implementers@oaisrv.nsdl.cornell.edu
> http://oaisrv.nsdl.cornell.edu/mailman/listinfo/oai-implementers
>