Ah, yes, good idea. It does rely on the client to identify the document, though; if the user's Web browser doesn't know what the file is, it will probably be reported to you as application/octet-stream or text/plain if it smells like text.