On Sun February 27 2005 23:18, Keith Moore wrote:
you'll get
an error, which is what we want to happen.
No, what we want to have is either a clear indication of lack
of an author mailbox (preferred, if practical), or a lack of
any indication of an author mailbox.
Yes, the loopback connection is supposed to be internal to the IP
implementation, and not be translated to any physical network. That
still doesn't have semantics of "anonymous" or of "nonexistent"; it
has semantics of "here".
in practice, it has the semantics of "nowhere", or "you can't get there
from here, no matter where 'here' is."
If I repeat Claus Assmann's experiment using 127.0.0.1 (the loopback),
what I get has semantics of "here".
it's a backward compatibility issue. MUAs and MTAs that know about the
convention are not the problem, since it's easy for them to recognize
such addresses and treat them accordingly. the problem is making this
work with MUAs and MTAs that don't know about the convention.
hey, I'd be fine with anonymous(_at_)[ipv6:::0]
or perhaps even anonymous(_at_)[no-such-network:]
And the backwards compatibility of either is ... ?
no, but it's the most important technical constraint on the convention
for an anonymous address. see above.
"anonymous address" is an oxymoron. "<>" (i.e. no mailbox) seems
less problematical (even with HTTP, SIP compatibility issues) than
some of the recent proposals. At least "<>" has some sort of
precedent (in Return-Path).
Too many things extract data from the From field.
If the field isn't there, they can't extract anything from it.
Less of a problem than extracting something and misinterpreting
it (or bombarding DNS with cruft, or misdirecting responses, or
crashing...).