This is a nice example of federated search. Note that when you input the search term it accesses each repository separately, then provides a link to the search results, if any. Advantages: the search happens in real time. Disadvantage: it's slow, and puts an excessive load on the repositories in the federation. Also, search results are unordered, and there's no value-added (such as third party tagging). Which leads me to really question the value of the federated approach. The authors have just release a poster (in PowerPoint) for the Open Scholarship 2006 Conference.