[dkim-dev] Do you have an interoperable DKIM of the recent spec?

[Arvel Hathcock]

Ok, I'm very sorry for the confusion on this topic.  We did a code 
review and found that what I said before was wrong.

In our implementation, an empty body leads to hashing of nothing (no 
CRLF).  I think this goes against 3.4.3.

I'm going to wait for the final verdict on what I should be doing and 
then make a correction to our code based on this groups collective wisdom.

-- 
Arvel

Michael Thomas wrote:
> That's funny because my null mail is not agreeing at your reflector.
> That is:
> 
> Subject: foo<CRLF>
> <CRLF>
> 
> the body should be canonicalized as <CRLF>, with l=2. Perhaps it's
> the l=2 that tripping you up?
> 
>       Mike
> 
> Arvel Hathcock wrote:
> 
>> > By the way, how do you canonicalize a null body?  Do you calculate the
>> > hash of a null string, or do you canonicalize a null body to <CR><LF>
>> > and calculate the hash of that?
>>
>> Sorry for the delay.  We do the latter.  For both simple and relaxed a 
>> null body canonicalizes to <CR><LF>.
>>
>> The spec makes this clear for simple:
>>
>> "If there is no trailing CRLF on the message, a CRLF is added"
>>
>> but it doesn't specifically say that for relaxed.
>>