Re: [ietf-dkim] Corner cases and loose ends, was , draft-vesely-dkim-joi

On 09/27/2010 11:17 AM, Al Iverson wrote:
> On Mon, Sep 27, 2010 at 1:05 PM, Michael Thomas<mike(_at_)mtcc(_dot_)com>  
> wrote:
> > On 09/27/2010 10:58 AM, Michael Thomas wrote:
> > > On 09/27/2010 10:38 AM, John R. Levine wrote:
> > > > > Ignorance is bliss, I guess, especially when it comes to pontificates.
> > > > > That's what every implementation of DKIM for MTA's, both open source and
> > > > > commercial that I'm aware of does, though some do and don't do the ADSP
> > > > > lookup. News at 11: email is still delivered, with little to no
> > > > > observable
> > > > > impact.
> > > > It is not my impression that they all do the full DKIM validation while
> > > > the SMTP session is open. Mine doesn't.
> > > You would be completely wrong in your impression.
> > >
> > > Source is your friend.
> > Oh, I see John weaseled from "nobody does that" to the unprovable
> > "not everybody does that". In any case, John is completely wrong
> > with his assertion that doing DKIM/ADSP validation at SMTP time
> > is somehow even vaguely untenable. It isn't. It's common as dirt.
> It's hard to imagine a large, DKIM-checking mail provider, like, say,
> Yahoo, doing that mid-transaction.
>
> Do you have any data/insight on how to quantify "common as dirt"? I'm
> doubtful of your claim without it.
Yes, I have a lot of insight. My implementation did it that way, which
along with Murray's (and any other that's based off of milters), do it
that way. I don't have access to Y!'s source to say for sure, but my
impression from Mark and Miles is that they did it in-session and that
all of our experiences were the same: it added very little overhead.

I don't know why this should surprise anybody. Doing DNS lookups in-session
*is* common as dirt for RBL lookups, even if the big boys have the RBL
databases in-house for performance. So the only other issue with DKIM is
the actual computational overhead, and Eric did some calculations that it
was *maybe* 5% overhead, and that was 5 years ago -- Moore's Law only
helps.

So by all means, doubt away. Is there anybody else who was at the Interop
who *doesn't* do DKIM in-session? It was my impression that everybody did
it that way.

Mike
_______________________________________________
NOTE WELL: This list operates according to 
http://mipassoc.org/dkim/ietf-list-rules.html