Well, I did my homework...
I'm getting frustrated, because I feel the solution can't be *that* complicated, but I don't have the skills to devise it. Especially considering that The Wikireader has already implemented all the tools to parse, compress and index the WP dumps, so that much is done:
http://github.com/wikireader/wikireader
So what remains to be done is the Sony Reader part (Linux + JScript)
http://www.reghardware.co.uk/2009/08...k_sony_reader/
I so wish I could program in Python!