Re: [idn] IDNA interoperability failures, once again

["D. J. Bernstein" <djb@cr.yp.to>]

Adam M. Costello flip-flops:
> In that case, IDNA section 3 requirement 1 *prohibits* dig from
> outputing the non-ASCII form of domain names. So there's no problem,

The problem, of course, is that your so-called ``internationalized
domain names'' are being presented to the user as gobbledygook.

Let's try another example: the ``show'' program in the UNIX MH/NMH
mail-handling system. Yes or no: Should this program convert domain
names from your special-purpose 7-bit character set to the local
character set? Again assume LANG=en_US.UTF-8, so this can be done.

Your newly invented ``well-defined pre-existing output format''
criterion says that ``show'' is not allowed to convert. But this means
that MH/NMH users, such as Keith Moore, will see gobbledygook instead of
IDNs.

How is Keith Moore supposed to see IDNs displayed properly in his email?
Let's assume that he has the current version of X, with Unicode fonts
and a UTF-8 xterm. Where exactly are your PunyCode IDNs converted to
UTF-8? Which programs do the conversion?

You keep dodging these questions because you know that any conversion
will trigger interoperability failures. The output that you think is
going to a user may actually be going to another program that does a
domain-name lookup without knowing anything about IDNA. Kaboom.

(To forestall any ``check for a tty'' weaseling: Programs such as Emacs
and Expect will give you a tty anyway; and tools such as copy-and-paste
are on the other side of xterm's tty.)

---D. J. Bernstein, Associate Professor, Department of Mathematics,
Statistics, and Computer Science, University of Illinois at Chicago