In testing odbc using the postgres jdbc issue I can see that we are not consistently handling the conversion of client encodings to charsets. Also later postgres clients can use the key UNICODE to refer to UTF-8.