John C. and John R.:
Let me bring up a variation of email address naming that's much more
limited in scope.
Do you see any problems to introducing a new PKIX email address type to
support Unicode in email local part for
Email-Address-Internationalization? This would be an alternative or
refined interpretation of rfc822Name in subject-alternative-name and
issuer-alternative-name PKIX extension. As RFC6532 allows this, but
RFC5280 hasn't been updated to support this.
On Wed, Mar 9, 2016 at 2:15 PM, Wei Chuang <weihaw(_at_)google(_dot_)com> wrote:
On Wed, Mar 9, 2016 at 1:20 PM, John C Klensin <john-ietf(_at_)jck(_dot_)com>
--On Wednesday, March 09, 2016 21:46 +0100 John R Levine
As I said a few messages ago, it is not an accident or a
mistake that there is no canonical form for e-mail addresses.
We understand why some people wish it were otherwise, but the
number of ways that MTAs map e-mail addresses is only slightly
less than the number of MTAs, and the mappings are constantly
The alternative is that for each mapping i.e. email address variant
the sender will need a different set of signing keys and certificates,
which they will have to distribute. Because these email addresss local
have become so ingrained, adopting PKIX for email address authentication
burdensome to point of killing off its feasibility. To quote pull a quote
earlier message, I too worry that "perfect may be the enemy of the good"
And, while smaller than the number of mappings and MTAs, there
are a very large number of specialized formats used for special
purposes like extended tracking or managing of delivery and
non-delivery notifications. Having a canonical form that can be
trusted requires understanding all of those cases, some
deliberately undocumented, and then guessing correctly (every
time) about what one is seeing.
It may be possible to figure out a way to use an SMTP server
or maybe a web server connected to an SMTP server as an
oracle, to ask do these two addresses deliver to the same
place or to ask for a key or a certificate for an address, but
even that is iffy.
Or not, depending on _exactly_ how the operation and selection
function are applied. But the point is the only that server can
definitively answer "do these deliver to the same place" and, as
pointed out for one case below, even that information may not
mean as much as people believe.
We can't even say what it means for two addresses to be "the
same". For example, on my MTA there about a thousand live
addresses that deliver to the same inbox where your message
was delivered, but that doesn't mean I want all of them to
have the same PGP or S/MIME keys.
Nor does it mean that, by the time whatever you are using as an
MUA or local mail routing rules (SIEVE or otherwise) that it
will end up in the same place ("inbox", "folder", etc.) on your
In case it isn't clear, John Levine and I are in agreement about
this and I'm writing only to emphasize his message and that his
position is not unique to him. The problem is really hard,
changing every mail MTA and MUA in the word to match someone's
idea of a desired canonical isn't feasible and probably would
not be desirable even if it were, etc., etc.
Wouldn't the problem size be smaller, and limited to those that want to
the identity of the sender and/or to do a key lookup? At some future
using PKIX asserted identities for sender and recipients for mail delivery
a good idea.
ietf-smtp mailing list