[ietf-clear] more on no callbacks, please

On Mon, 2004-10-04 at 09:34, James Couzens wrote:
<snip>
> SPF is a perfect example. Its a broken protocol which abuses the DNS
> system and has serious flaw as regards forwarding yet by my count
> there are in excess of 600,000-800,000 domains publishing records for
> this.  People are latching on to whatever they can do to stop the
> spam.  We should be paying attention to this and using it to our
> advantage!
Following your harangue of Bind, the trend regarding SPF is to increase
feature bloat beyond already absurd levels.  SPF demands lookups of as
many as 10 different scripts containing perhaps hundreds of records,
with as yet undocumented features added by the hour.  Until fully
engaged, the danger falls to the recipient and hence the laissez faire
publishing of text records.  Limiting SPF to being an occasional
white-listing tool seems the only safe advantage offered.

CSV does not add to the DNS burden.  CSV should actually reduce the
burden once widely deployed and offers a validation of the sender (MTA)
by name.  Financial institutions would be well advised to be consistent
in their HELO naming conventions.  BATV is _not_ required to validate
the sender in the general case.  With other standards in place, a Public
Key BATV would be helpful, should the nominal mail channel of a mailbox
domain be expressed, perhaps by way of a name list, to handle the
exception imposed by forwarding.  The burden upon the mail system to
support BATV should be small when this is an ancillary check.  CSV does
the sender check normally, and is perhaps assisted in the future with a
name list.

BATV does not require a PK mode to be effective against techniques using
the bounce as a means to evade reputation checks.  Private methods of
the signing allow a unique application to each message.  A record of
bounce counts per signature within an active time period will offer a
means to defeat sizable replay attacks.  If there is a replay attack,
accept initial bounces within reasonable limits as constituting
sufficient notice to the sender.  Using this technique also identifies
the message being exploited.  For absolute protections, a meld of IIM
and DK would be a better approach.  Using such a scheme, the public key
is included in the message, and the hash of the public key is used to
access a record in DNS as a means of sender validation.

-Doug