[Top] [All Lists]

Re: [ietf-dkim] versions of RFC822 mail messages, Where is the formal definition of DKIM-Signature?

2018-02-10 11:26:47
On 2/10/2018 7:50 AM, John Levine wrote:
The idea with DKIM v=2 is that there are things that you cannot say in
a v=1 signature, no matter how many new tags you add, so you need some
way to tell verifiers what they need to understand.  How about this?

We rebrand the v= tag to be a feature list so the syntax is now roughly

   v= word (, word)*

where each word describes a semantic feature.  Feature tag "1" is all
the stuff in RFC6376.  My feature is mandatory to understand tags,
feature name "mandatory", so the signatures start

The listing of 'authorized' features makes sense when the usage may occur later in the session, as it does with ESMTP, for giving the other side permission to use those features. It makes no sense at all for a unilateral exchange where one side uses whatever it feels like and the other side -- getting all this later -- either likes it or doesn't. That is there are no contingent behaviors in the exchange.

In a unilateral activity like DKIM, the mere presence of the usage "featurex=..." serves to flag that featurex is used. There is no incremental benefit into moving the flag elsehwere.


Dave Crocker
Brandenburg InternetWorking
NOTE WELL: This list operates according to

<Prev in Thread] Current Thread [Next in Thread>