> So you see SRW as a query mechanism against the aggregated
> database, not an alternative to OAI to form the aggregated
> database? (That is what I was thinking of too.)
>
> So use OAI for data aggregation, and SRW for distributed
> searching.

Not necessarily - SRW is mute about distributed searching. SRW searches
a single database (defined by the http endpoint). That database could be
a database that has been built by OAI harvesting. The immediate
advantage to putting SRW in front is that you can then mix and match
user interfaces to it (rather than build a UI straight on top of the
database).

That isn't to say someone can't write an client to cross-search a number
of SRW servers, or even a SRW server whose backend cross-searched a
number of SRW servers making them appear as a single "virtual" database
or come up with lots of other exotic architectures - just that there are
other motivations for putting SRW in front of an OAI harvested database.

Matthew