Oops... I hit the wrong key, sending a partial post. Sorry :) Again: I have a problem with strings and unicode chars, when writing slq statements on Oracle ODBC driver through HSQL. I'm writing here because I suspect that a fix could come from writing 8 bit strings, if it is possible somehow. I'm sorry, the problem is very deep into it's context, and I know it's not easy to follow the reasoning. These are the facts: - I execute an SQL statement, an INSERT, where some values come from a web browser, and are UTF-8 encoded - Oracle ODBC expects UCS-2, so it converts the values into UTF-8, of course scrambling the data - when I convert, in Haskell, UTF-8 to UCS-2, I see that someone (perhaps HSQL) strips one byte from the two Char's bytes, so again Oracle receives bad data As I haven't found how to force the driver not to strip the byte, and also I don't like to convert data two times, I'd try to send 8 bit strings, but don't know how. Is Word8 a solution? If it is, what is the contstructor?, as w = W8# 1 doesn't compile. I'm weak on low level Haskell. Where do I find references? On the Report? Thanks, Enrico