ODBC / SQL Native Client doc: UNICODE description is out of date by FLAESCH Sebastien

I believe the ODBC manual section about "UNICODE" support needs some review:

http://msdn.microsoft.com/en-us/library/windows/desktop/ms709439%28v=vs.85%29.aspx

The first sentence is wrong: UNICODE is described as a two-byte encoding (i.e. UCS-2), which is false. UNICODE refers to a mapping between characters and integers, it's not a specific encoding of integers, as a sequence of bytes in memory...

I am not sure that in the ODBC world (including other platforms as Microsoft Windows), Wide-Char API functions expect only the UCS-2 character set... So far I could not find any clear documentation about that, and reading this section did not help much.

What encoding should I use when writing a "UNICODE" ODBC application, that is supposed to be portable over different platforms, like Windows and UNIX? Can it be UTF-16 today?

To me this section needs some clarification, even in the Microsoft-only world...