There's a conversation going on about this within the OpenNews community, which may have interesting info for you.
Dan Sinker wrote an article on the topic for Source, see http://source.mozillaopennews.org/en-US/articles/us-shutdown-scuttles-data/.
Also, during the October 9 OpenNews call, there was a detailed discussion about access to government open ...
I aggregate datasets from different sources together frequently. I sometimes use a hybrid database/table representation, that follows some adaption of this basic approach:
Split the fields into two table schemas.
A. one for things that are invariants (e.g., place name, ISO codes, etc)
B. one for things that are variants (e.g., population, coordinates, etc).