I have read somewhere that boost provides a better typedef. Something that allows to use the 'underlying class' but it is not a macro for example I want to define typedef std::wstring cvarchar_t to use for ODBC/Unicode db access (on win32) and then specialize some templates for this type that would behave differently then wstring I cannot remember what that mechanism was called. Also, sorry, for mixing this question in: is UTF-8 a more 'standard' encoding if I want to interface from my C++/boost programs to ODBC and to Boost.Python on Windows and FreeBSD (I have problems figuring out how to support the difference between UTF-16 on windows/wstring and Unix UTF-32 wstring ... while the unicode ODBC drivers are all 2-byte strings) thanks in advance -- V S P toreason@fastmail.fm -- http://www.fastmail.fm - Same, same, but different...