[ietf-dkim] SSP security relies upon the visual domain appearance

[Stephen Farrell]

John Levine wrote:
>>>The "From:" header should not be signed if it contains more than one 
>>>sending address. ...
> 
> 
>>Does anyone see such a statement as causing a problem?
> 
> 
> I see it as needless and futile micromanagement.  

Pithy. Kind of a needless and futile sentence though:-)

 > The point of a DKIM
> signature is that the signer is taking responsibility for the message.
> The only semantics that a DKIM signature has is "blame us if you don't
> like this message."  That's it.
> 
> We don't know all of the reasons that a signer might legitimately want
> to sign multiple From: addresses, nor do we know all of the ways that
> a bad guy might try to trick someone into signing his message, with
> multiple From: addresses being rather low on that list.  

That's fair enough.

> I could easily imagine an SSP-like system limiting itself to a subset
> of otherwise syntactically valid messages, e.g. only one address in
> the From: line, sender matches signer, or any of a host of other
> rules.  But for the basic DKIM, a signer can sign anything he's
> willing to, and please leave it at that.

Yes, that's what we're getting at.

So if, during the threat analysis, we identify some such
constraints that make life easier/better when combined with
some ssp options then we could consider standardising them,
or did you mean that any such constraints should be just up
to the individual implementer/signer?

Stephen.