Hi,

Viktor Dukhovni wrote:
> I don't think domain-level pinning, rather than per-server pinning
> is workable.  Pins expire, domains change hands, and MTAs unlike
> MUAs and browsers don't have users to "click OK" when stuff breaks.

How would you imagine per-server pinning with workable key-rollover?

I too see problems with expired pins -- see below.

> With STS (which I am not currently endorsing as a universal approach
> beyond the small club of the largest providers behind this design)
> the destination domain owner publishes the policy which essentially
> pins the MX RRset and specifies authenticaiton of same.  The
> webserver that authenticates the policy is at a well-known URI
> associated with the destination domain.  The MX operator just has
> to maintain valid WebPKI certificates from some widely trusted CA,
> that are associated with the MX hostname.  I don't see why CT
> support or lack thereof is pertinent here.

It's not but would be nice to have, no?

> 
>> Currently more than 60% of SMTP MTAs do _not_ have signed certificates,
>> far less even provide valid CN or SAN X.509 information.
> 
> Because why pay for bits everyone is going to ignore anyway.  The
> sysadmins are behaving rationally.

Sure. But then again they need signed certs for the WebPKI reporting and
authentication anyway as far as I understand the current proposal (it's
not submitted to IETF so we're discussing something that the authors
might not even envision).

>  
>> TACK does not
>> require certificates to be CA-signed, this is completely optional. TACK
>> also makes revocation and key-rollover a bit easier than this proposal.
> 
> I'd considered TACK for SMTP, but gave up once it became clear that
> TACK (per-server key pinning) is of no use given MX indirection.

TACK has various problems there, and it's a bit dated by now given
change in some parts of the WebPKI infrastructure and new protocol
proposals within IETF. We could pick it up, dust it off and find a
proper way to use it for e-mail though, this is my idea. There's no
reason in verbatim employing TACK as currently documented.

> FWIW, I don't think that STS is universally applicable either, ...
> That said I'm going to contribute some cycles and text to making
> it as good as it can be, and then time will tell.  Worst case it
> will help the large providers secure their inter-organizational
> connections and protect a large fraction of mail by volume, rather
> than domain diversity.

I see it makes a lot of sense for large hosting providers, and have
noted so repeatedly. The problem is: they're already doing fine with
security deployment and some of them might already have some form of
"pinning" (say Google knows Yahoo's server information and the other way
around - not sure if that's really the case but I'd do it this way on
that scale even without a proper protocol given the Snowden
disclosures). If we end up with a protocol that's only deployable for
large providers we don't get added security for the small self-hosted
MXs on the Internet - and this is my concern. We can't just tell all of
them to use GMail as an MX.

Thanks for your feedback,
Aaron

Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
Uta mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/uta

Reply via email to