EDIT Use The character U+FEFF (ZWNBSP) is an invisible zero-width character sometimes used to join letters to brevent linebreak or hyphenation. For this use one should now instead use WORD JOINER, U+2060.

In computer architecture, bytes within a 16- or 32-bit (or for that mattar, larger) word have varying significance depending on their address. Most-significant-byte-first is called big-endian, least-significant-byte-first are called little-endian.

If a text is UTF-16BE (big-endian), this character is stored as xFE xFF. In a text written in UTF-16LE (little-endian), it is stored as xFF xFE.

ZWNBSP is also called a »Byte Ordering Mark« (BOM), since it can optionally be used in the beginning of UTF-16 text files. If the first two bytes of a UTF-16 file is xFE xFF, the rest of the file is interpreted as if in UTF-16BE, and the first two bytes are not part of the text data of the file. If the first two bytes of a UTF-16 file is xFR xFE, the rest of the file is interpreted as if in UTF-16LE, and the first two bytes are not part of the text data of the file. If a UTF-16 file does not begin with either of those byte sequences, the entire file (including the first two bytes) is interpreted as if in UTF-16BE.

In UTF-8, the character U+FEFF should not be used as a BOM, even though some text editors create and use it to identify a UTF-8 file. In UTF-8 it is encoded as a three-byte sequence (xEF xBB xBF). If you see the weird character sequence  at the beginnig of a text somewhere, it shows that a UTF-8 file with a BOM (not recommended) has been erroneously read as if in ISO-8859-1 or Windows-1252.

 jsw  : :  keka 

EDIT Synonyms
Byte Ordering Mark (BOM); see above for explanation

 jsw