referencing IDNA2008 (and IDNA2003?)

Andrew Sullivan ajs at
Thu Oct 21 23:36:48 CEST 2010

On Thu, Oct 21, 2010 at 05:28:40PM -0400, John C Klensin wrote:
> comparisons are among what we now call A-labels.  So an IDNA2003
> implementation trying to compare U-labels directly is simply
> non-conforming or, to use the more precise term, broken.

Just to clear up this remark: yes, but the problem was IIRC that
people sometimes _stored_ the U-label form, and sometimes the A-label
form, and so they'd compare what they had (maybe re-running the
algorithm).  I seem to remember a lot of gnashing of teeth about this
problem partly due to any backward incompatibility introduced in
IDNA2008.  Anyway, even if the implementation is broken, it doesn't
matter: the user experience will undoubtedly be better in a given
application if it uses the same rules all the time than if it tries to
mix and match protocols.  If it weren't so off-topic, I'd now commence
a riff on how trying to make DNS and the host name syntax
interoperable but not quite compatible, and then not telling
everybody, "Please pick one at a time and stick with it," has caused
us no end of trouble.  I'd hate to smell even a whiff of such
encouragement with IDNA2003 and IDNA2008.


Andrew Sullivan
ajs at
Shinkuro, Inc.

More information about the Idna-update mailing list