Take a look at the file in a win9x ('95, '98, 'ME) version of notepad, or another editor you know doesn't deal with utf8. If the non-low-ascii (IE s-sets and umlouted vowels) show up as two characters, it's utf8 (or some really funky encoding). If they show up as one, then it's not.

From Re: trouble with umlauts - update, "encodes the ö as %C3%B6 that causes the regex to not match, or as %F6 which causes a match there" suggests that it's sometimes being encoded as UTF-8 (0xC3 B6), and sometimes as somthing else (0xF6).

The fact that it's matching when it's F6 but not C3 B6 suggests that your input isn't getting normalized to utf8 at all, but rather to that other encoding.

All in all, it doesn't really matter what encoding it gets into, so long as the encodings match everwhere they need to. (I'd recommend utf8, because it's becomming the clear standard, and because XML is assumed utf8 unless otherwise marked.)

What was the solution you found?