[Ietf-dkim] Re: Concerns about the recipe/rollback mechanism and deployment architecture in draft-clayton-dkim2-spec

[Graham Orndorff <[e-mail]>]

> On Mar 21, 2026, at 10:24, Pete Resnick <resnick=[e-mail]> wrote:
> 
> There is also a traffic cost that falls on the entire ecosystem regardless of adoption. Every message that transits DKIM2-aware nodes accumulates Message-Instance headers with base64-encoded JSON, additional DKIM2-Signature headers at each hop and potentially substantial body recipes. These headers travel in the message to all recipients, including those on servers that do not implement DKIM2 and that simply ignore those headers. The recipient receives a larger message with no benefit if their server does not speak DKIM2. This overhead is not optional, it is imposed on the entire ecosystem by the nodes that do implement DKIM2, regardless of whether the downstream infrastructure can use it.
> 
> Of this there's no doubt, but it seems to me that we're going to have to quantify this at some point. Strictly anecdotally, most of the modifications I see from mailing lists are pretty small. Is there any way we can establish how much this will really add and the overall impact? (I'd point out that the real high-volume folks from Google, Yahoo, and Microsoft are in the room, so perhaps they could help.
> 

From Apple’s point of view (if you need some high volume folks to weigh in):  We will of course need to see what the final version will be like.  But in the current drafts, the costs seem reasonable given the benefits.   Memory footprint for this is likely more of a concern than CPU at volume, and both seem okay at the moment.

-g