While there's only one byte order for utf8 (unlike utf16), the BOM might still make sense to easily distinguish utf8 from other encodings.