That footnote confirms the traditional position that a floating point zero need NOT have an all-zero bit representation.