[Top] [All Lists]

Re: [ietf-smtp] Should we update an RFC if people refuse to implement parts of it ?

2021-06-05 11:56:19
It appears that John C Klensin  <john-ietf(_at_)jck(_dot_)com> said:
No.  I am asking a much more pragmatic question or pair of
questions: (i) whether making a change in this area (see below)
is worth, at this particular point, opening up RFC 6531 and/or
6532 (and reviewing whether other documents in the SMTPUTF8
collection to see if they need similar adjustments) would be
worth the trouble and (ii) whether there is the energy to do
that and to do it well.

At this point, tne answer is clearly no.  I was asking as part of
the EAI testing I'm doing for the UASG, to figure out how hard I
should try to persuade people to fix their code to pass that particular
test.  (Not very, since they all say no anyway.)

FWIW, Gmail and Hotmail/Live/Office365 which I believe are the 
largest EAI-enabled mail systems in the world, also have A labels.
Of course, when I looked at a test message I sent to my Hotmail
account, I found this gem, so A-labels are hardly their biggest issue.
Check out those "with" and "via" clauses:

Received-SPF: None (
 xn--5nq21jyu9d1ta.xn--5nqx41au4nqohsp3axcg.xn--fiqs8s does not designate
 permitted sender hosts)
Received: from xn--5nq21jyu9d1ta.xn--5nqx41au4nqohsp3axcg.xn--fiqs8s
 ( by
 ( with Microsoft SMTP Server (version=TLS1_2,
 cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.4195.21 via Frontend
 Transport; Sat, 5 Jun 2021 16:23:47 +0000

I have found pretty good support for exhanging mail with EAI addresses.  Those
two hosted mail systems do so and most open source MTAs have adequate EAI 
Postfic, Exim, Courier, and now even sendmail (in beta, but it works.)

What I have not found is a lot of support for hosting their own EAI
addresses. You can do it in the open source stuff, but nobody does.
It's fairly painful to do so for a variety of reasons. On most systems
foo@<A-label> and foo@<U-label> are not mapped together so you have to
make two addresses for everyone, and MUA support is spotty at best.
Ned has described some of the reasons that even though Oracle's MTA
has code to support local EAI addresses, nobody uses it.

I have found almost no support at all for EAI in POP and IMAP. The
only open source is Courier IMAP, and the commercial ones with EAI
extensions I've tried are really buggy, to the point of repeatedly
requiring that I patch the python IMAP library to keep it from
crashing during the tests. IMAP already has overcomplex but usable
support for multiple character sets in folder names and search terms,
and I have found that more often than not non-EAI IMAP servers such as
Dovecot work adequately with EAI mail sort of by accident because they
don't try to enforce 7-bit ASCII in their 8-bit data.

So anyway, I expect that EAI and ASCII mail will coexist for a long time even
though there is no downgrade path to smooth the way.  The A-labels are just
a symptom of that.


ietf-smtp mailing list