Matthew Elvey <matthew(_at_)elvey(_dot_)com> wrote:
If being listed with a trusted reputation service is to be a hard
requirement, how do we suggest the recipient deal with the situation
where the reputation service mentioned is unknown?
4xx Unknown Reputation Service error and urgent admin alert?
I think it was Meng that suggested reputation services should rate
other services. I rely on that (figuring that market forces will
drive out those that refuse).
Thus, my expectation is that the receiving MTA would query its
own preferred reputation service, which might return "Never heard of
them".
We probably should specify something about the format of an error,
but in essence it's:
5xx - <my-service> says "Never heard of <your-service>."
THE MARID RR shall return a MARID level as <integer>.<integer> (exact
meanings to be defined in later documents)
???
(My hands were waving.)
We may in fact need a set of flags in addition to a version number;
but I firmly believe we will need a version number in order to keep
straight what features are implemented by the sending MTA. It's critical
that different versions interoperate well.
...
I think all the SHALLs above should be MUSTs.
You're probably right (though I'm open to different opinions).
(Mailers are free to send mail that isn't compliant with the document,
and addressees are free to refuse it.)
Absolutely.
(Mailers that claim to be compliant when they aren't should quickly
acquire a bad reputation.)
"Known-good" shall mean that there is known to be a MTA at that IP
address under the control of the management of this domain; that there is
policy in place to enforce the conditions of the MARID level quoted; that
reports of abuse of that policy are welcomed at <abuse(_at_)subdomain>; and
that
action on violations of the policy will be taken within <blank> business
days.
s/business days/days
because a spammer being allowed free rein over a weekend doesn't make
much sense to me, I'd trust an ISP that acts within 2 days over one that
acts within 1 business day.
We may need to make this a parameter within the MARID RR record; but
I'd hesitate to demand quick response as part of the standard -- better
to leave that issue to the reputation services.
... a separate lookup will be required, to
be accomplished by prepending an encoding of the actual IP address to the
HELO/EHLO string, and the result of that lookup will show whether the
IP address represents an MTA that is 1) known-good, 2) believed-good,
3) believed-bad, or 4) unknown.
I'd like to suggest that in case 3, the mail MUST be refused (unless it
is to accepted for abuse tracking purposes).
I'm very hesitant to put that in the spec. There are particular users
so paranoid about missing an email that they demand _all_ spam-checking
be bypassed.
I suggest that in case 4, perhaps the mail should be refused. In other
words, a valid HELO becomes a MUST.
I see that as a local option. It does smell a lot like a "SHOULD",
though...
If no DNS response is received within a reasonable timeout, the
receiving MTA MAY give a temporary error noting that fact; but it SHOULD
query its preferred reputation service with just the HELO/EHLO string if
the lack of response persists for more than a few hours.
This sounds like a likely good idea, but I'd like more clarity on what
it accomplishes.
Is this is useful if the domain's DNS server is down (perhaps due to
spammer DDoS) ...
(My hands were waving again...)
Yes, I was thinking of DoS attacks, and thinking that reputation
services would become aware of DoS attacks within a few hours. There
are also miscellaneous glitches with similar effects. I couldn't think
of precise language for that, though, thus the waving hands...
--
John Leslie <john(_at_)jlc(_dot_)net>