contains this snapshot a solution for the UNICODE problem (reading get wrong strings) ?
All known Unicode problems were fixed a while back. There are now 2 drivers though - the ANSI one should be used unless you specifically require Unicode ODBC support.