Hey Dave,

On Monday 19 October 2009 12:22:20 Dave CROCKER wrote:
> Barry Leiba wrote:
> >     I suggest that ADSP-compliant mailing lists should be
> > advised to reject "discardable" messages whether or not they will be
> > breaking the signature.
> 
rejection is really only needed if they break the signature which should be 
evident by their settings. The point of DKIM is to preserve integrity however 
the bank statement is a confidentiality issue.

For instance I manage a private maillist of board members that receive paypal 
notices. 100% DKIM validation pass rate. The maillist is configured to not 
modify the message so the final recipients could validate it too if they 
wanted.

> Yes, this is a reasonable idea.
> 
> The question is whether it is the /right/ idea.
> 
> Another reasonable idea is that the mailing list should ignore ADSP, since
>  ADSP is really meant for final recipients;

As the mailing list is probably the last place to see a valid signature 
evaluating the ADSP there is the best idea. The mailing list verifier has a 
greater confidence in rejecting broken signatures there than the final 
recipient. The final recipient could deploy some whitelisting model based on 
the behaviour of the list with minimal risk.

>  note that ADSP only comes into
>  play for recipients who support it.  (Well, that is at least one model.)
>  And there are no doubt lots of other reasonable ideas.
> 
> At this stage, I believe rightness depends entirely on market preferences. 
>  Do we have any empirical data of ADSP use which experiences the problem
>  being covered here,

Some was described here;
http://mipassoc.org/pipermail/ietf-dkim/2009q4/012596.html

>  resolves it in the way being suggested,
dkim=discard is the easy case. 

>  and garners  receiver support?
at the moment receivers who care about dkim whitelist domains/ip or if they 
wish to accept some risk, rely on domain reputation.

> Absent any of that, this discussion is purely academic.
> 
> Each proposal like this is expensive.  It takes time to discussion, run
>  through the process, test, deploy and use.  We should let private
>  experiments determine the preferred handling, before we seek to
>  standardize a solution.
> 
> Particularly since we seem to have only and exactly one market-based
> organization experiencing the problem.

given the level of ADSP deployment it is hardly unexpected that only one 
organisation that this group collectively knows about has disclosed a problem.


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

Reply via email to