I downloaded a dump of a smaller language (piwiki, 1.7MB). To make matters worse, it looks like all the data is only available in one single huge XML-file. Notepad++ crashes after 10 minutes while trying to open it. I don't think it's viable to put such a monster on a Sony Reader.
If there was a single file for every wikipedia page it would maybe be possible to compress the site and decompress it when the user wants to read it. There is a ZIP-implementation in JS for example:
http://jszip.stuartk.co.uk/