[ietf-dkim] incremental vs. infrastructure adoption

[Dave Crocker]

Charles Lindsey wrote:
>> A whitelist is useful as soon as a single recipient (filter, user, 
>> whatever) can apply it.
> 
> Be careful there. We want people out there to welcome and accept these 
> protocols when they start to be deployed. A huge spate of false 
> positives and false negatives will rapidly cause it to come into 
> disrepute and to wither. This is not a technical problem - it is a 
> social problem, but we ignore it at our peril.

I do not see how your statements relate to the one of mine to which you appear 
to be responding.

Are you suggesting that my statement was incorrect?

If so, how?  Please remember that the topic was as stated in the Subject line 
and my comment was in that context.


> Q. "My cousin John, who uses the same bank as me, gets nice "SAFE" logos 
> on his bank messages. I don't see them. Should I be worried?"
> A. "You need to use a different MUA."

There is a new car that automatically performs parallel parking.  Is there some 
sort of problem creating by observing that you get that feature only by buying a 
different car?

Or a different end-user feature by using a different MUA?

In other words, if your example is intended to be cautionary, please clarify, 
with substantial detail.


> Q. "I bank with Bank of America. I always used to get those "SAFE" 
> logos. Now they have suddenly stopped".
> A. "I see from your headers that you are on holiday in the Netherlands 
> Antilles. The ISP you are using there doesn't use whitelists for 
> American Banks - just for the Bank of the Antilles and ABN/AMRO."

This, I think, is a very interesting example, conceptually.

The detail you supply, however, is confusing.  Your sentence beginging "The 
ISP..." does not make sense to me.

I guess that you are presuming that all of the relevant email functions are 
under control of the local-access ISP and that, therefore, when traveling, I 
might get a different MUA?


>> Although true, there is no requirement that it be "standardized".  
>> Different existing lists have different access methods.
> 
> No implementor of the popular MUAs is going to invest effort in 
> implementing this "SAFE" Logo if there are lots of different protocols 
> which might be used. 

Statements containing firm absolutes, when they involve human decision-making, 
tend to be problematic.  By way of example, note that web page designers 
actually do have a wide variety of pages, with the same content, designed for 
different devices and browsers.  That such a situation is absurd and often does 
not work correctly is not the point.  The point is that I believe it disproves 
your absolute.

That said, of course, simplification is the reason competitors collaborate on 
unified protocols.

But again, I find myself not understanding how your point is responsive to the 
discussion in this thread.  Perhaps my commenting that standardization was not 
required was taken to mean that standardization is not a huge win?  Gosh I hope 
no one thinks I would believe such a thing...


>>> You can't do it in the headers, because Bad Guys can write headers too.
>>
>> Not when the headers are signed.  (eg, <http://goodmailsystems.com>.)
> 
> And there you gave a good answer (and it also brings the thread back on 
> topic :-). OK, such schemes are possible. But the Bad Guy can still 
> insert the "SAFE" header and sign it, so there is still work for the 
> verifying agent to do by way of spotting such already present headers 
> and acting accordingly. .

Typical discussions, about such indications of safety, distinguish between what 
is carried in the message, across the Internet, versus what is generated within 
the trusted Administrative Management Domain (ADMD) of the recipient.  (It might 
be generated at presentation time or it might be added to the message within the 
trust domain, but that is a minor point, I think, for this discussion.

User-visible safety indicators are typically viewed as the realm of the 
delivering ADMD and not of the message as it is received from the open Internet.

In that context, how does your described threat survive?

d/
-- 

   Dave Crocker
   Brandenburg InternetWorking
   bbiw.net