Will the challenge be that you cannot hook to low enough UI/human interaction events to build an alternative for DOM? E.g. Paste menu is impossible to do (needs CTRL+V in Google docs for Firefox), cannot get information about keyboard typing mode in multilingual setups and so on.