Dave Crocker <dcrocker(_at_)brandenburg(_dot_)com> wrote:
Seems to me there should be a discussion about the implications of
having distributed rule-processing engines. For example, where do we
have related experience, such that we can understand the dynamics and
administration, and have any hope of expecting this stuff to scale?
We know DNS works, and scales. I don't think anyone's arguing that
systems can maintain & publish their own MARID records. So to the
rest of the net, those records in DNS look much like any other records
The main differences between MARID and other kinds of data in DNS
are size and how much work the receipient puts into interpreting them.
These issues have been discussed for recipients, but not for MX
secondaries, because the topic of MX secondaries interpreting MARID
records hasn't come up. It should be discussed in further detail.
Hmm.. other than the "RCPT TO" issue Douglas Otis brought up, MX
secondaries aren't that different from any recipient SMTP server so
far as MARID goes. The main difference is that it's more critical for
the MX secondaries to obtain the MARID information for a domain.
As a related point, I've been part of a system which implements a
kind of distributed "pobox" via hesiod. Each site publishes a list of
users(_at_)common-domain (which maps to a local email address), and each
site is listed as an MX for that domain. They all do cross-queries
for hesiod information, so they all have the complete user list &
forwarding table. Mail can then get queued on any site that's up, and
forwarded to the final destination by whoever received it.