[ietf-dkim] ADSP and Discardable (was Re: Lists "BCP" draft review)

[Douglas Otis]

On 6/2/10 10:10 AM, Scott Kitterman wrote:
> "Dave CROCKER"<dhc at dcrocker.net>  wrote:
>
> On 6/2/2010 8:08 AM, Al Iverson wrote:
>    
>>> Agree. "Discard" and "silently discard" mean the same thing, in my
>>> opinion. Though, I am guilty of using the phrase "silently discard."
>>> Maybe in an attempt to be slightly over-specific.
>>>        
>> I do not recall seeing a dictionary or technical definition of "discard" that
>> says anything at all about whether the discarder says anything at all.
>>
>> Taken on its own and without further technical specifications 'discard' does not
>> direct, imply or request that the action be silent or noisy, and if noisy who
>> gets to hear it.
>>      
> IIRC, this is by design since there was no consensus around what exactly to do.  Personally, I tend to favor not having messages silently vanish.
>    
The initial intent was to assert signing domain's policies rather than 
attempting any receiver recommendations.  In this vein, a better an 
alternative to "discardable" could have been 
"no-third-party-services-used" such as N3PS.  The "discardable" 
assertion did not emerge directly from the list.  John explained 
"discardable" as his response to phished domain's concerns of being 
flooded with NDNs.  ADSP did not produce a flood, as seen by the 
phisher's reactions of avoiding exact matches.

Rather than asserting author domain signing policies, ADSP could be 
changed into offering receiver recommendations, which "discardable" 
attempts.  Perhaps "all" could change to "non-deliverable".  However, 
could this result in bounces, in addition to rejections.  After all, not 
all receivers evaluate DKIM prior to acceptance.  Should these 
recommendation be "rejectable" and remain silent about accepted 
messages?  To avoid this quagmire, the WG concluded that ADSP was to 
state the domain's signing policy.   Clearly, "discardable" represents a 
clear departure.  Will "discardable" ultimately prove helpful?  It has 
already caused many to ignore handling for "all", since "all" lacks the 
concise instruction offered by "discardable".  It seems "discardable" is 
in danger of ensuring an unproductive outcome for ADSP.

-Doug