ietf-mailsig
[Top] [All Lists]

Re: DKIM o=$ Outbound Signing Policy Suggestion

2005-07-20 16:37:00

Hector,

Thanks; it's good to get some comment on the SSP document as well.

One thing that occurred to me (prompted by a phone call from a colleague) is that signing policy really needs to have two dimensions:

1. What signatures are required? None, third-party, or first-party (From: address) only?

2. How would the From: address domain like the verifier to treat failures? Some possibilities: Reject outright, warn user, send warning to r: address (need to be concerned with privacy issues here), etc.

Does this fit with what you have in mind?

-Jim

Hector Santos wrote:

The document draft-allman-dkim-ssp-00, section 5 defines the Outbound
signing policy tag, "o=".

| o= Outbound signing policy for the entity (plain-text; OPTIONAL,
| default is "~").  Possible values are as follows:
|
|    ~ The entity signs some but not all email.
|
|    - All mail from the entity is signed; unsigned email MUST NOT
|    be accepted, but email signed by a third party SHOULD be
|    accepted.
|
|    !  All mail from the entity is signed; third-party signatures
|    SHOULD NOT be accepted
|
|    .  This entity never sends email.  The "." policy can be used
|    to "short circuit" searches from subdomains; for example, the
|    "ad.jp" domain might use this.  If an initial policy search
|    receives this policy then the email SHOULD NOT be accepted; if
|    found while searching parent domains then the search should
|    terminate as though no policy record was found.
|
|    ^ Repeat query at user level.  This value MUST NOT be used in
|    user-level policy records.

I think that one of the problems with DKIM is how easy (high probability)
the signing process can be broken during transit.

What is missing is a policy on how to handle a broken verification
(non-signed or signed).

I suggest a new policy o=$ tag value to reflect the high value a domain has
on the email.   A  suggested wording could be:

o=$ (High Value Email)

    All mail from the entity is signed; unsigned email
    MUST NOT be accepted, but email signed by a
    third party SHOULD be accepted.  All mail that fails
    verification due to some broken DKIM entity SHOULD be
    reported to company (r=) [before accepting].

Issues to discuss:

- Use this tag or refined o=- to include non-signed or broken signed
verifications.

- Reporting should be a SHOULD not MUST because spammers can use this as a
feedback mechanism as well, in addition,  MTAs should have the final say on
any further reporting overhead.

- Should the email be accepted or rejected (temporary or permanent) while
the reporting takes place?     A company may not want this to occur for
their high valued email.

Point to consider:

In my opinion, MTA will not tolerate infinite failed DKIM transactions.

The same issue has developed with SPF with its relaxed policy provisions
(neutral, softfail).    I would say that DKIM tag o=- is a neutral, and o=$
would be a softfail like concept.   As a result, some SPF implementations
have already began to send reports or verify domains to obtain negative or
positive feedback on the neutral and softfail result.

--
Hector Santos, Santronics Software, Inc.
http://www.santronics.com





<Prev in Thread] Current Thread [Next in Thread>