CharToByteUnicode generates Big endian Unicode on Win32 platform even if its
default seems to be little endian. (Both Notpad and Word97 on WinNT save
Unicode file in little endian order.)

/**
 * Convert arrays containing Unicode characters into arrays of bytes, using the
 * platform-default byte order.
 *
 * @version 1.3, 96/11/23
 * @author Mark Reinhold
 */

public class CharToByteUnicode extends CharToByteConverter {

    static final char BYTE_ORDER_MARK = (char) 0xfeff;
    protected boolean marked = false;

    static final int UNKNOWN = 0;
    static final int BIG = 1;
    static final int LITTLE = 2;
    protected int byteOrder = UNKNOWN;

    public CharToByteUnicode() {
    String enc = System.getProperty("unicode.encoding", "UnicodeBig");
    if (enc.equals("UnicodeBig"))


See above comment that it says "platform-default byte order".
However current implementation always generates Big endian order because
the system property "unicode.encoding" is not set anywhere.

If I trust what the comment says, then "unicode.encoding" should be set properly. Otherwise the spec should clearly mention that Java Unicode is
UnicodeBig and unnecessary code should be removed.