Re: [idn] Why we can go directly to UTF-8

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

Eric A. Hall writes:
> This is a massive undertaking.

Every IDN proposal involves massive costs---in particular, upgrading
sendmail. The point of my IDN web page is to collect information on
those costs:

   http://cr.yp.to/proto/idn.html

ACE with slow nameprep requires redeployment of apache, fetchmail,
gethostbyname, lynx, mutt, named, netscape, pine, publicfile, qmail,
sendmail, sshd, tcpclient, tcprules, tinydns, w3m, and many similar
pieces of code.

UTF-8 with fast nameprep requires redeployment of gethostbyname (except
that some versions have been fixed already), mutt, pine, and sendmail.
If anything else needs to be upgraded, please let me know.

> These systems need to know
> that the domain name will have a maximum label length of 63 octets

Wrong again. The 63-byte limit is isolated inside one library routine
that converts textual domain names to DNS packet format. That routine
already works perfectly with UTF-8. I don't know where you got the idea
that applications have to stop users from typing more than 63 bytes.

> If I had kept a list of every device and application that broke when a
> customer of mine thought "_" was legal, I'd gladly give it to you.

In other words, you don't have a single shred of verifiable evidence.   
Your memories are so vague that you can't even tell us what programs you
were using---but you're confident that those programs would have trouble
with UTF-8, and you're making major decisions on this basis.

---Dan