URI encoding seems to be somewhat ad-hoc. I propose to always encode reserved characters, plus (maybe with an option) encode what is now called ILLEGAL_CHARS and align that with this list.
/cc @obiltschnig