[Top] [All Lists]

Re: [ietf-smtp] Dotless domains and email

2013-06-21 14:51:31

FWIW and remembering that I didn't do the ABNF for either 2821
or 5321...

Both 5321 and 2821 contain

 Mailbox = Local-part "@" Domain

RFC 2821 contains:

  Domain = (sub-domain 1*("." sub-domain)) / address-literal

While 5321 contains

  Domain = sub-domain *("." sub-domain)

That change, effectively removing the requirement for at least
two components in the domain-part, is consistent with the text
in Section 2.5.5 of 5321 (and the very similar text in the
section of the same number in 2821) but does not appear in any
of my notes about intentional changes between 2821 and 5321
(I've just done a quick check and might have missed something),
including being absent from the change log in the last I-D
version of 5321, so I'm not sure where it came from.    I do
remember some quite explicit discussions during both the 2821
and 5321 work about eliminating the prohibition on trailing
periods in mailbox names.  That change would have both better
aligned SMTP with 1034/1035 and make the distinction between a
TLD name and a possible abbreviation/ alias in need of
completion, but the WG (and group that commented on 5321) wanted
nothing to do with it, at least in part because that distinction
was considered a UI issue and that SMTP should adhere rigidly to
the rule that any <Domain> that appeared in an SMTP Mailbox was
required to be an FQDN.

That is all I'm likely to remember without doing a lot of
time-consuming research, but I do agree with Carl -- while one
would hope that everyone would have abandoned 2821 (and 821,
974, and the mail-specific requirements of 1123) in favor of
5321, we know that hasn't happened in practice and that some
MUAs and MTAs will refuse to resolve single-component domain
names or will, despite years or prohibitions, try to treat them
as aliases.  So there is a predictability problem at best and
might well be some other issues.


ietf-smtp mailing list