On Aug 24, 2006, at 5:39 PM, Jim Fenton wrote:

In addition to other objections I have, if it is not significantly simpler I question the need for an additional mechanism, especially in a security context such as this.

For any 2822.From address signed with DKIM to be trusted, the 2822.From address MUST be restricted to accounts where the 2822.From being signed is known valid and marked as such using the i= syntax. This mechanism MUST exist whether or not any 2822.From policy is asserted about this domain.

Assume that a large ISP finds that validating _all_ 2822.From addresses for a specific d= signing domain generates revenue as a premium service, AND this also reduces support costs related to abuse issues. In addition, this domain can be certified as only signing validated addresses. This certification increases the acceptance rate of messages signed by this domain and increases the value of their service. None of these benefits are dependent upon anyone designating this domain in their policy. Such designation will add to their stature however and can be monitored. : )

A customer of this service can then independently designate this domain in their email-address domain. This designation process does not require the exchange of any keys, key locations, or any special handling of their messages. Just a simple domain name is entered in the DNS they control. Recipients can then identify which trusted 2822.From address are valid when signed by this domain and are designated, without the need of the i= syntax.

The extra effort required would be validating that the holder of the account is also the recipient of the address being used. That effort is not much more than what is needed to assure the 2822.From address in the first place. The prerequisite for DKIM security is knowing who and what to trust. Being able to convey with DKIM what should be trusted is vital for retaining trust and security.

-Doug



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

Reply via email to