--On Tuesday, 27 March, 2007 15:21 +0200 Arnt Gulbrandsen
<arnt(_at_)oryx(_dot_)com> wrote:
My two cents:
It's better to disallow localpart(_at_)tld as email address in
2821bis. That is, I prefer to remain compatible with 2821
rather than with 821.
I have two arguments:
1. The change in 2821 hasn't caused a signficant level of
problems, AFAICT, so why change again?
I would venture the hypothesis, with no hard data at all, that
the change has been generally ignored, possibly on the
assumption that it was a stupid mistake.
2. I see quite a few addresses of the form
'localpart(_at_)hostname'. Say between 1% and 1ppm of messages. If
'localpart(_at_)tld' is guaranteed not to be valid, a smarthost's
life is simpler: 'localpart(_at_)something' is either a local
address (not subject to IETF rules) or invalid. A smarthost
has to decide whether it's valid by local rules or not, that's
all. There is no possibility that the address might have two
valid meanings.
Ok course, as others have pointed out, user(_at_)example(_dot_)com could
either be
- a reference to the reserved domain name in "com." or
- a local address, e.g.,
user(_at_)example(_dot_)com(_dot_)foo(_dot_)info[(_dot_)]
So, in principle, one has to be able to deal with the "two valid
meanings" case for all names other than single-parters.
I'm not quite sure what this proves, but it doesn't seem to me
that making addressing mail to a TLD simplifies things all that
much.
best,
john