Microsoft Windows uses little-endian byte order on all platforms. Thus, on Windows UTF-16 code units are stored in little-endian byte order in memory.
I believe that some linux systems are big-endian and some little-endian. I think linux follows the standard byte order of the CPU. Presumably UTF-16 would be big-endian or little-endian accordingly.
This is somewhat misleading. For internal processing, where we are talking about the UTF-16 encoding form (quite different from the external encoding _scheme_ of the same name), we don't have strings of bytes but strings of 16-bit units (WCHAR in Windows). Program code operating on such strings could not care less what endianness the CPU uses. Endianness is only an issue when the text gets byte-serialized, as is done for the external encoding schemes (and usually by a conversion service).
markus
