ietf-dkim
[Top] [All Lists]

Re: [ietf-dkim] Straw poll results

2010-08-09 18:12:00
Rolf E. Sonneveld wrote:
Scott Kitterman wrote:

It's not at all clear to me that the answer to that question is 
in any way related to the work of the working group.  What would 
we design differently if the answer was yes (or no)?

Let me try to explain. If the identity of the list contributor is of any 
value to the receiver of an MLM-distributed message, then it is 
important to (try to) preserve the original DKIM signature across an MLM 
redistribution of the message (if at all possible). If however the 
identity of the list contributor is of no value whatsoever, we should 
not bother about preserving the original DKIM signature.

This is why I believe that the current GOOD CITIZEN centric approach 
is very hard to resolve without addressing the "BAD CITIZEN" model 
first or at the very list, the GOOD CITIZEN model should not come at 
the expect at diminishing the high beneficial value of the BAD CITIZEN 
model.

It is a conflict an engineering philosophy and I believe once we 
acknowledge one can not exist without the other, this issue with 
continue in perpetuity.

What you speak of is about trust, something something about the 
message that allows you to believe it is OK?

Consider the contrary question, what is it about the message that 
makes you NOT TRUST the message?

Lets say that we some how invented a way to RESTORE THE INTEGRITY OF 
THE ORIGINAL SIGNATURE so its no longer a factor and I changed my MUA 
or online mail web interface to support and present all four ideas:

      5322.FROM domain
      5322.DKIM signing domain (d=)
      5322.ADSP as a function of 5322.FROM domain
      5322.LIST-ID domain

and I'm going to add a 5th new idea here (data point) called LDSP 
(List Domain Signing Policy)

      5322.LDSP as a function of 5322.LIST-ID domain.

Idea #1:

If the 5322.ADSP result is DISCARDABLE, the DOMAIN is making a public 
policy statement:

       NO ONE CAN SIGN MAIL FOR US. ONLY WE CAN SIGN MAIL.
       This is called a 1st party signing only policy.

The question is why would you (speaking in general) ignore this very 
powerful data point if you see your list message is signed by the list 
and not by the domain?  The original signature is intact. But the 
policy said only they sign mail - no one else.

Doesn't that say there is someone WRONG, not GOOD, about this message? 
even with an original valid signature intact?

Idea #2

Lets say you trust all mail coming from this list because it is signed 
by the list domain.  You feel reasonable secured it is valid GOOD 
message when it is signed by the list and also if any, the original 
domain signature is intact.

But what if the DKIM signature domain is not signed by the list?  What 
it is a 3rd party, for example?


     From: <louie(_at_)FROM-DOMAIN(_dot_)COM>
     DKIM-Signature: d=FROM-DOMAIN.COM
     DKIM-Signature: d=OTHER-RESIGNER.COM
     List-Id: <listadmin(_at_)LIST-DOMAIN(_dot_)COM>

Would you have a lower trust of this message?  Would it still matter 
if the original signature is valid but the list signature is valid but 
resigned by some other unknown domain?

Idea #3:

Let say that we have this new LDSP (List Domain Signing Policy) 
protocol which is a small change to ADSP by instead of anchoring off 
the 5322.From domain, LSDP anchors off the 5322.List-ID domain.

Let say that LIST-DOMAIN.COM has a LDSP policy of:

     DKIM=DISCARDABLE

which says:

     LIST-DOMAIN.COM is the only domain that can sign list messages
     coming from this list.  No one else is expected to sign as a
     3rd party.

Would you still trust this list even when there is an explicit LDSP 
policy exposing to the world that it has an exclusive and  restrictive 
1st party signing policy?

So basically, all I am saying that regardless of how its down, there 
is always the issue of not getting what is expected and before you can 
even trust a list-id or from domain, you need to factor everything all 
the failing about it.

If you trust a LIST-ID domain, thats that mean that only it can sign 
mail? What happens when its some other domain?

-- 
Hector Santos, CTO
http://www.santronics.com
http://santronics.blogspot.com


_______________________________________________
NOTE WELL: This list operates according to 
http://mipassoc.org/dkim/ietf-list-rules.html