RE: CBV

> From: Mark Shewmaker
> Sent: Friday, May 07, 2004 3:48 AM
>
>
> On Fri, 2004-05-07 at 04:00, [email protected] wrote:
> >
> > Surely, at this point, a simpler and easier solution would be only to
> > accept PGP signed messages from a verifiable source, and
> > possibly include
> > a set of permissible source addresses in the PKI infrastructure
> > with each
> > public key.
> >
> > That's what this starts to sound like.
>
> Hand-waving argument written at 4am here:
>
> Even while suggesting the two types of checksums for SES, I have similar
> misgivings, but:
>
>  o SES has the VERP-ish you-can-reject-bad-bounces feature that SPF by
>    itself doesn't have.
>  o SES has other flaws.
>
> I am curious what happens to SES if all its "other flaws" can be
> reasonably addressed, because right now SES and SPF, in my mind at
> least, mesh with each other but somewhat uncomfortably.

That's what we're all curious about and why I believe it is worth
discussing.

>
> Seeing the way Seth (I think it was Seth) looked at SRS and found that
> doing something similar from the get-go (SES) gets you a lot of SPF-ish
> advantages and the VERP-ish feature, I'm hoping that an SES tweaked
> enough to address its admitted shortcomings can help him or someone else
> figure out a better way it can mesh with spf.

It was David Woodhouse, as far as I know, that first realized that the SRS
signature by itself allowed you to reject bogus bounces.  I share your hope
that by investigating this approach, we can figure out a more optimal
overall solution, whatever it looks like.

>
> To me SES looks like a necessary puzzle piece that just doesn't quite
> fit into the SPF scheme of things for reasons I can neither quite put my
> finger on nor understand how to correct.  I have the possibly irrational
> notion that figuring out a solution to the SES flaws (which seems
> tantalizingly close at hand) can help in understanding the puzzle piece.
>
> That's why I'm hoping we can find good solutions to the SES things
> talked about in this thread.

Lightweight sender authentication really is a puzzle.  Each of the systems
we are discussing solve parts of the problem, but so far, none of them does
the whole thing and each has some undesirable baggage.

>
> Sp, for the moment I'm pushing aside the fact that these suggestions
> sound eerily reminiscent to PK-type solutions, but expect to have to
> come back to it later.
>
> (I hope this makes sense!)

It certainly does to me.

--

Seth Goodman