[OAI-implementers] Clarification on deleted records

[Simeon Warner]

On Thu, 11 Apr 2002, Alan Kent wrote:
> On Wed, Apr 10, 2002 at 02:47:46PM -0400, Hussein Suleman wrote:
> > hi
> > ... if deleted records are not tracked... the 
> > service provider has to regularly issue GetRecord with the identifiers 
> > of previously harvested records in order to confirm that each of them 
> > does still exist...
> 
> I am not suggesting it today, but I can see the time when someone
> may therefore want a new command to say 'are these 1,000 identifiers
> still valid?'. Doing a GetRecord per identifier one by one could
> be pretty slow and will unnecessarily return the record (when all
> you want to know is if it exists).
> 
> I was initially thinking 'no-one would ever do GetRecord for every
> record in the database - its too slow! doing ListRecords or ListIdentifiers
> would always be faster', but then I realised you could always do a
> GetRecord on records past a certain age. Even then though, was it
> ETDCAT that had 500,000 records that came into existance one the
> same day? 1/2 million GetRecord requests is not a very serious
> option.
> 
> I am not proposing a new verb though. Sending 1,000 identifiers
> or whatever through in a request I suspect won't be a serious option
> until OAI requests are embedded in something like SOAP. POST would
> work I guess, but GET???!?! (that would be one long URL!).

This is one of the reasons why a multi-item GetRecord was not included in
v2.0. This may be more reasonable if OAI is later embedded in something
like SOAP.

> Periodic complete rehavesting seems the best solution with the
> protocol as is.

That is probably the safest approach with v1.1. 
(In v2.0, this should not be necessary for a repository that claims to
track deleted records forever.)

Cheers,
Simeon.

 
> Alan
> _______________________________________________
> OAI-implementers mailing list
> OAI-implementers@oaisrv.nsdl.cornell.edu
> http://oaisrv.nsdl.cornell.edu/mailman/listinfo/oai-implementers
>