I propose that there are the following requirements:
 
1) Mechanism for announcing that a service is available to accept reports of 
DKIM verification failures
2) Mechanism for specifying which reporting protocols are supported
3) Mechanism for discovery of the reporting service of a given protocol
4) Specification of the reporting protocol
 
I regard 4 as being out of scope for DKIM, we should not develop a protocol.
 
I regard 3 as being solved by DNS SRV, its what it is for
 
So that leaves 1 and 2. I consider these to be a DKIM responsibility, in 
particular an SSP responsibility.
 
 
In an ideal world there would be an ideal reporting protocol. I don't think 
that we are an ideal world and so I beleive that DKIM must support some means 
of reporting protocol agility just like we would regard protocol versioning a 
good thing. 
 
The difference is not at all complicated, it means writing REPORT=ARF or 
REPORT=ARF|INCH 
 
 
There is a philosophical design issue here, clearly it is a bad thing for DKIM 
to provide more options than is necessary to address DKIM functions. But it is 
also a bad thing for DKIM to force the use of a particular support 
infrastructure. This is why we made sure that we can in fact use DKIM with XKMS 
or with X.509 pki even though we have a key record mechanism.
 
I don't want to get into the design of the reporting protocol here. I am not 
even sure that ARF is what a standard would look like, we already have INCH 
after all. If we decide to design DKIM to only use ARF we create a dependency 
and have to ensure that ARF is at equivalent standards level. Weak coupling is 
what we need, provide a slot to allow interaction with ARF but don't force a 
choice.
 

________________________________

From: Eliot Lear [mailto:[EMAIL PROTECTED]
Sent: Wed 14/11/2007 10:37 AM
To: Hallam-Baker, Phillip
Cc: Damon; O'Reirdan, Michael; [email protected]
Subject: Re: [ietf-dkim] Proposal to amend SSP draft with a 
reportingaddress(fwd)



Hallam-Baker, Phillip wrote:
> I am happy to say that DKIM MUST support the use of ARF as A reporting format
>
> I disagree that it should be the only reporting format supported. What we 
> want today and what we will want in five years time are going to be 
> different. ARF is a decent start but its limited. To go much further it is 
> going to be necessary to use structured data.
>  

Ok, but let's proceed cautiously here, Phillip.  This is an area where
more standards != better.  How would you propose we manage interoperability?

Eliot


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

Reply via email to