Re: nowsp considered harmful

I've already heard several proposed solutions on this list that I think 
would fix the problem adequately, so I'm hesitant to suggest another, 
but...

Another approach would be to bite the bullet on MIME-awareness, but in 
a minimalistic way --  all that a DKIM-aware program would have to know 
is how to recognize the multipart boundaries, i.e. to read the 
"boundary" parameter and detect the boundary lines,  both of which are 
doable as part of the same single pass that computes the signature.  
Signing software would modify the input to the hash algorithm in some 
way (e.g. pretending there's a 255 octet) at each point where it 
recognizes the END of the MIME boundary lines.   Signature checking 
software would do the same, so that moving a MIME boundary would break 
the signature, effectively pre-empting any MIME-based attack I have yet 
heard of.

However, I'm reasonably happy with any of the proposed solutions; it's 
clearly a solvable problem.

A side note:  I absolutely can't bring myself to care if DKIM is 
vulnerable to "vandalism" in the form of corrupting sabotage of real 
messages, because there's no money in it for the vandal.  Nearly every 
protocol on the Internet is subject to horrific vandalism in the form 
of DOS attacks.  Why worry more about it here?  -- Nathaniel