On July 28, 2005 at 14:39, "Hector Santos" wrote:
Take for example with two signatures. Is this an example of a 3rd party
signing? What policy controls this?
Using your spoof example:
DKIM-Signature: a=rsa-sha1; s=whatever; d=ispoofyou.org;
c=simple; q=dns;
h=Received : From : To : Subject : Date : Message-ID;
b=dzdVyOfAKCdLXdJOc9G2q8LoXSlEniSbav+yuU4zGeeruD00lszZ
VoG4ZHRNiYzR;
Received: from 10.2.3.4-example.com [10.2.3.4]
by submitserver.example.com with SUBMISSION;
Fri, 11 Jul 2003 21:01:54 -0700 (PDT)
From: Joe User <joe(_dot_)user(_at_)example(_dot_)com>
To: Suzie Q <suzie(_at_)shopping(_dot_)example(_dot_)net>
Subject: I need your help?
Date: Fri, 11 Jul 2003 21:00:37 -0700 (PDT)
Message-ID: <20030712040037(_dot_)46341(_dot_)5F8J(_at_)example(_dot_)com>
A SSP lookup for _policy._domainkey.example.com is done. If order for this
to pass, example.com must allow for a relaxed/neutral policy.
The above example I provided does not provide an i= tag, so it assumes,
according to the DKIM draft, that i= is equivalent to d=. The problem
occurs if the signer wants to set i= to the actual identity of the
entity they are signing for if the signer is acting as a third-party
agent. The DKIM states that i= must be a subdomain of d=, but this
seems restrictive for 3rd-party scenarios where the signer may be in
a different domain.
What it might all come down to is we might run into situations where the
user usage of a domain might be restricted in certain areas of the internet.
For example, a mailing list.
What if the mailing list server began to sign all its distributed mail? or
any MTA signing all outgoing mail?
Well, if the originating address domain is not local (a relay maybe), it
might have to lookup the SSP to determine if it is even allowed to sign the
message.
Another problem is how the end MUA sorts it all out. With multiple
signatures present, which one is relevant? I think MUA display issues
are important to consider because how MUAs support DKIM will determine
the real effectiveness of DKIM being used to help in deal with spam
and phishing.
The DKIM draft punts on the existence of multiple signatures, but I
think this may be a mistake. I think multiple signature semantics
should be defined, especially how a signature is added to a message
already signed. I can envision a signature "chain" where a signer may
want to include an existing signature in their signature to provide
a cryptographic binding of the two signatures.
To help determine how multiple signatures effect things, it would
help to list out some practical usage scenarios where an entity
will want to sign a message that has already contains a signature
(like the mailing list example you mention).
Alternatively, multiple signatures could be explicitly forbidden.
All of this of course needs to take into account of the Originator
Address and how that address is protected from spoofing or from
actions forbidden from the OA's SSP.
--ewh