IDNAbis News from IETF 72 in Dublin (2008-08-04)

The IDNAbis Working Group held two face-to-face meetings on July 28th and 29th at the 72nd meeting of the IETF in Dublin. A latent problem in the Unicode Consortium's bidirectional algorithm was demonstrated, which causes cross-label leakage of directionality, which in turn makes bidirectional character sequences at the ends of labels to be displayed incorrectly. The root cause is the UTC's treatment of the ASCII "dot" as a period, followed by one (or more) blanks, whether on the right or the left, in a left-to-right, or right-to-left context, respectively, so the "." is, for the purposes of directionality, within a label, rather than external to, and separating, and therefore without directionality, which is the correct semantic for the ASCII DOT in DNS labels.

A work-around is to ban digits at the beginnings and endings of Arabic and Hebrew script strings, affecting domain names in Arabic, Farsi, Urdu, etc., and Hebrew and Yiddish, respectively. This hides the inherent defect in the UTC algorithm, which is probably the wrong thing to communicate to the UTC, even knowing that the UTC won't change its treatment of the ASCII DOT, or re-examine the scope of the UTC's applicability claims for the DNS.