String comparison with utf8_bin collation is case sensitive.
  Hence "DELIMITER" did not match with "delimiter".
  The delimiter command matching now uses my_charset_latin1.