Re: [Asrg] E-mail Postmarks

[William Leibzon <william at completewhois.com>]

On Tue, 8 Jun 2004, Eric A. Hall wrote:

> On 6/7/2004 11:12 PM, John Levine wrote:
>
> > Oh, OK, it's modified S/MIME with domain rather than individual
> > signatures.  That's a reasonable idea that's come up from time
> > to time before.
> >
> > Any thoughts about how the verification keys might be distributed.
>
> Back-channel ESMTP verb is my preference. It's the same level of
> reliability as any other channel. Also, it makes the cost of publishing
> the key data equal to the cost of retrieving it, which is an important
> piece of symmetry imo -- most other models put all of the punishment
> (caching, blocking, etc) on the retrieval side, but all the benefits go
> for the publisher, which punishes deployment.

There is problem with that approach. If you let public key be distributed 
as part of the same SMTP session as email itself, it actually means nothing.
Reason are that I can't trust your public key to be really for your
organization just because you say so at the time I'm talking to you. 
For example I could use openssl and easily create self-signed certificate
claming to be aol.com, then if I use it in email session and provide 
public key as part of the process, the other side might actually believe
I'm aol.com.

To be of any use crypto signature based email security proposals must provide
for a way to retreive public keys based on CN (common name) or similar
identity name that is part of the cryptographic signature. Its possible
to do this by making the key available in dns, or by using dns as pointer 
to correct service that holds these keys or its even possible to do it by 
creating new SMTP command and requing recepient site to do SMTP callback
to verify the signature.

But I don't think we'll gain much by extending SMTP to handle such callback 
system for retreival of public keys, in such case I think its better to 
just work on new protocol since it can possibly be of use for protocols 
other then SMTP. I can however mention that SMTP callback for certificate 
retriaval may have small advantage in how it may allow for dns wildcard 
support (see latest MARID discussions) but its somewhat complicated to 
get it done right nor is it really good place for such records as 
incoming mail servers often differ from outgoing.

---
William Leibzon
william at completewhois.com

_______________________________________________
Asrg mailing list
Asrg at ietf.org
https://www1.ietf.org/mailman/listinfo/asrg