Google is planning to [host](http://arstechnica.com/news.ars/post/20080122-rumors-suggest-google-is-set-to-open-scientific-data-store.html) a *lot* of scientific data. Below are the slides from Jon Trowbridge at Google (my favorite part is comparing the transfer rates of FedEx to Ethernet):

Depending on the API Google provides, it would be convenient to have IDL bindings capable of grabbing parts of this data.

Link via [FlowingData](http://flowingdata.com/2008/01/21/google-decides-to-host-a-whole-lot-of-scientific-data-palimpsest-project/).