Talk:OAI Sets Generalisation

From DSpace Wiki

This all looks like a great generalisation, though OAI-PMH sets do have a flaw, specifically related to when an item changes set membership, which in the case of dynamic sets is likely to happen.

If an item X that is part of OAI-PMH set S is changed such that it is no longer a member of set S, future harvests of S will no longer retrieve X. In other words, X has been 'deleted' from the point of view of S, but harvesters will not get this information and will continue to believe that X exists in whatever state it was before it was moved from S.

A similar problem may occur if an item Y becomes a member of set S after its creation, if membership of set S does not occur due to a change in Y.

E.g. say I have a thesis collection and I expose this as a thesis set T. Then I create another thesis collection and add it to the thesis set. However, a harvester doing incremental harvests on T will not see the new arrivals from the new thesis collection, as their OAI datestamp will reflect when they were ingested, not when they became members of T.

i.e. the OAI datestamp of an item needs to reflect its last change in set membership, but this is difficult or intractable for dynamic sets.

This isn't to say that configurable sets aren't worth doing; it's just that these caveats need to be documented and made clear to potential users of this feature.