> On Dec 16, 2016, at 5:53 PM, Jim Fenton <[email protected]> wrote:
>
> It seems that Viktor and I have differing views on the usefulness of
> MAYTLS and REQUIRETLS. That said, I agree that it doesn't make sense to
> have two mechanisms for this, so I will try to formulate a hybrid
> mechanism for the next iteration of the draft.
Thanks. With that out of the way, I can elaborate on some of the policy
details that need to be discussed as part of this.
> However, I'm a little concerned that the MAYTLS part of this may
> directly contradict the policy of the SMTP server. RFC 7671 section 4.1
> says, "if usable secure TLSA records are published, authentication MUST
> succeed." There would be a similar conflict for SMTP STS. Wouldn't it be
> presumptuous for this specification to attempt to supersede those
> requirements?
Given that the protocol under discussion is per-message user-specified
policy, by default, it supercedes the policies learned from "strangers"
(namely the destination domain) via DANE, STS, ... And so the new
specification can update RFC7672 to explain that per-message policy
from the user generally takes precedence.
That said, local policy can circumscribe what user policies are accepted,
and so an MTA can refuse to honour either MAY or REQUIRE or both. In the
REQUIRE case, a conformant MTA or MSA must refuse the message. The the
MAY case it could accept the message and hope that delivery will succeed
without a policy downgrade, or it could refuse the message immediately,
leaving the user the option of not setting an unsupported policy.
Since accepting means that the message may be delayed retrying the delivery
for some time, and notification of the delay may not always happen in a
timely fashion, it is tempting to say that the MTA should refuse MAY when
it does not allow MAY, but on the other hand, the mail could well go through
anyway, and the MTA may well have working/timely delay notices. So one might
leave the choice of whether to refuse or do one's best without a downgrade to
local policy.
So the idea is to specify a communication channel from the user to the sending
MTA, and leave it up to local policy to decide which user (and/or receiver)
policies to support.
--
Viktor.
_______________________________________________
Uta mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/uta
