On Sun February 27 2005 02:51, Frank Ellermann wrote:
Keith Moore wrote:
[From: <anonymous(_at_)[]>]
since this doesn't define a new domain name (it uses a
domain literal), there's no need to update BCP 32.
It's not clear that it's a domain literal per RFC 822; more
detail separately.
Sure, but you want this idea to replace the "best common
practice" From: <what(_at_)ever(_dot_)invalid> based on BCP 32, and
therefore the best place to document it is an updated BCP.
The scope of BCP 32 is for testing and examples, and ".invalid"
is specifically reserved "for use in online construction of domain
names". There's nothing there about use in mailboxes, email
responses, etc. BCP 32 nowhere recommends widespread of those
reserved names in core Internet protocols, and certainly says
nothing about the Internet Message Format From field; indeed, the
intent is stated as being to reduce "confusion and conflict" which
may occur when names used for testing and in examples escape
onto the Internet -- it certainly doesn't encourage deliberately
bombarding the root nameservers with those reserved names.
It's important that user agents support these conventions,
and don't offer to send mail replies to invalid addresses.
The only reliable way to determine whether the domain part of
a mailbox is valid is to query DNS (see RFC 3696). An off-line
UA (or one on the non-Internet side of a gateway) clearly cannot
do so as DNS is unavailable. So we need a mechanism which does
not depend on domain names.
Moreover, actually querying DNS for large number of messages
containing bogus domain names would place an unwarranted burden
on DNS (specifically, the root nameservers), and might be
construed as a denial-of-service attack.
That was the point of Bruce's draft, and that's also the
reason why Usefor drafts proposed TLD .invalid for about
five years.
The point of the draft is to standardize a mechanism which
avoids inappropriate hacking of UAs and inappropriate
bombardment of DNS root nameservers by inappropriate use of
a TLD reserved for testing and examples. See section 1,
second paragraph labeled "b.". The idea is to have something
more suitable for use in places like the USEFOR WG work than
such inappropriate use of reserved names.