I'm actually applying for an SBIR grant right now to work on the NLP algorithms that power fwdeveryone.com, if you have any interest in writing a letter of support. Basically it would eventually enable someone to mass export something like an MBOX archive onto the web in a cleaned up format with accessible typography. You can play around with prettyfwd.com to get an idea of the current state of the tech, it works well for 95% of (non-commercial) email threads but still needs some more work to support the rest.