Re: [idn] An experiment with UTF-8 domain names

["D. J. Bernstein c/o James Seng" <James@Seng.cc>]

Martin Oldfield writes:
> although it would look odd,

If you think that incorrectly displayed IDNs are acceptable, why are you
here? You can move from mail.tc to fifth-word-is-ace-xyzzy-mail.tc, and
we'll pretend that xyzzy has some accents in it. They aren't displayed
as accents, but that doesn't matter to you!

> Am I right in thinking that UTF-8 style IDNs are used then Boris and
> Ivan can't communicate

What you mean here is that Ivan can't send mail to Boris's IDN.

> unless Ivan can persuade his ISP to upgrade his sendmail,

Maybe, maybe not.

The specific problem is that sendmail discards bytes \200 through \237.
This doesn't affect the UTF-8 encodings of the lowercase characters that
you can find in 8859-1. For example, a-circumflex is \303\242.

But I don't find this situation acceptable. sendmail has to be fixed or
replaced. Note that USENET messages now use UTF-8 in the header and are
often gatewayed verbatim to mail, so sendmail has to be fixed no matter
what happens with IDNs.

> but an ACE would only need Ivan to upgrade his mail client ?

By saying ``only,'' you are implying that UTF-8 IDNs require a client
upgrade. In fact, there are already clients that handle UTF-8 IDNs.

Certainly ACE IDNs require a mail client upgrade. When the mail client
sees a From line with an ACE IDN, it has to display some international
characters for the user.

Now, your question is whether ACE IDNs require a sendmail upgrade, if
the system administrator doesn't need IDNs in his configuration files,
and if addresses in outgoing messages are converted to ACE by the MUA.
The answer is still yes, because sendmail often copies domain names into
text that users are supposed to read.

---Dan