[Fwd: Re: [ietf-dkim] canonicalized null body and dkim]

[Michael Thomas]

Stephen Farrell wrote:
>
>
> Michael Thomas wrote:
>
>> Simply stated, as the draft is currently worded, the simple body 
>> canonicalization
>> will be immune to additions *and* deletions of of CRLF's at the end 
>> of the body
>> in all cases. The proposed change to the normative behavior, on the 
>> other hand,
>> will not be immune to deletions. Deletions are something that happen 
>> in real life,
>> and we have experienced then. Changing the normative sense of the 
>> draft at this
>> point will reduce the number of verified messages.
>
> Thanks for the explanation.
>
> Sounds like a fairly telling argument to me, if people are generally
> seeing such deletions.
>
>> I will also add that finding this has been an extremely maddening 
>> adventure as
>> a developer. If we change it to what Mark and others are advocating, 
>> we are
>> condemning all future DKIM developers to find this problem themselves 
>> just
>> as I have and be faced with unpleasant alternatives of how to work 
>> around it,
>> if they even find it in the first place. I don't wish my pain for 
>> this one on anybody.
>
> There you confused me - aren't we going to make it easy for every
> subsequent coder by telling them about this in advance? But its ok
> that I'm confused, I often am:-)
Not if we change the current sense of the spec, no. If we change rather
than clarify the current draft, every developer is going to have to go 
through
the same detective work on why signatures with lone CRLF's in the body are
being stripped causing signatures to fail. This is *not* intuitive and 
not easy
work. The best outcome is to have a *robust* and *clear* canonicalization.

       Mike