Re: [idn] impacted systems investigation

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

Mark Andrews, from the BIND company, writes:
> UTF8 does not require a server upgrade

Right. But Patrik and Paul claim the opposite. This claim is, in fact,
the centerpiece of the IDNA ``design philosophy.''

> gethostbyname() and gethostbyaddr() are clients of the DNS
> and as such are not covered by RFC 2181.

RFC 2181 says that character-set restrictions are forbidden in all
``implementations of the DNS protocols.'' It doesn't matter whether the
implementations are clients, caches, or servers.

Are you saying that your client libraries are ``various applications
that make use of DNS data'' rather than ``implementations of the DNS
protocols''?

Anyway, why don't you make make no-check-names the default, or at least
make 8-bit characters exempt from the test? The specific bugs described
in CA-1996-04 were fixed a long time ago; bugs of this type can't occur
if you use \ddd for unusual ASCII characters in PTR results; your BIND 9
documentation already says that bugs of this type aren't BIND's problem.
Why are you prohibiting lookups of contourcname.cr.yp.to?

---Dan