On 2018-06-29 at 20:49 -0400, John C Klensin wrote:
Anyone have comments on this? I don't remember whether there
was a reason it was done the way it was or if that was just a
dumb ABNF error.
Somewhere around 2000/2001, the ISP I worked at updated mail-systems and
indirectly enforced the suggested change.
It led to a customer complaint, from someone actually using
<""@customername.isp.example> (I think mostly for USENET posting)
because spam harvesting software never could handle it, but everyone
that customer cared to correspond with had compliant software which
We fixed our systems.
RFC 2821 allows empty content, so 5321 is not a regression against its
Looking at 821, I don't see how empty content would be allowed there. I
don't recall the standards discussion at the time, or if it was after
2821 had been published. The ISP had by then supported ""@ for the
previous few years and failing to do so was a regression.
I suspect the population size using this "feature" of SMTP is small and
if it's considered an undue burden then there might be an argument for
cutting them off. But it would break stuff known to work and be in use
% exim -bt '""@spodhuis.org'
router = imap_user, transport = imap_inject
(I did one s/usercode/censored/ there, just to avoid having spam
harvesters skew the numbers for that rarely-used mailbox).
(I don't consider this "in use" for me, it exists in my system to be
able to check it still works and to prove a point; the customer
complaining was, back then, actively using it).
Quoted-string = DQUOTE *QcontentSMTP DQUOTE
Quoted-string = DQUOTE 1*QcontentSMTP DQUOTE
ietf-smtp mailing list