Although Z39.50 is mostly used in the bibliographic community, it would be perfectly suited for a project like this where you have large amounts of distributed data in many formats. You would of cause use XML as the exchange format but the format on the individual servers is not important for interoperability, and the database administrators will therefor still have the freedom to keep there data in any format they like and make most sense to them.
Z39.50 is also a light weight protocol and studies shows that searching many databases in parallel is not a problem, it is usually the database servers that are the bottle neck.
Slashdot Mirror
Who doesn't like Warp?
Z39.50 is also a light weight protocol and studies shows that searching many databases in parallel is not a problem, it is usually the database servers that are the bottle neck.