ZOË rocks. www.zoe.nu. It's built on Apache Lucene and it actually works. I've just let it spend 3 days importing and indexing over 100k messages (8 years worth, after stripping mailing lists). Now I can search my old emails in seconds, get threaded lists, and easily find all the attachments. It even dealt with duplicates perfectly, which given the state of my mail archives is definately no easy task. 3 days is a long time, but then it was running on a 512Mb 450MHz old machine. I've already found mails I never thought I had and pictures in attachments I didn't remember. Awesome stuff.