Sounds like a helpful component would be a dns cache that could
be configured to set a rather high minimum TTL it would honor. I
don't know of one, but dnscache, from djbdns, ought to be easy to
patch. A quick grep for ttl shows a couple of places where ttls are
being clamped to an upper value; the pattern I settled on after a
little looking about was 'ttl > 604800', where djbdns clamps the TTL
to an upper value of 1 week. I expect if each of the two matching
instances were preceeded by something like

        if (ttl < 1800) ttl = 1800;

or thereabouts, that should cause the cache to refuse to honor ttls
shorter than a half hour. Season to match your postfix retry
interval and such a hackup should get deliveries to sites with flaky
DNS to work on the second try.

An alternative view is to say, if someone has DNS sufficiently
poorly configured that many or most initial queries timeout, and
further sets their TTL to shorter than a typical MTA retry interval,
they are the sort of misanthrope who doesn't feel a need to be
reachable via email, so why should you worry on their behalf?

-Bennett

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (GNU/Linux)

iD8DBQE+PrywHZWg9mCTffwRAiubAJ9Wr7v9RP0LXgL6rZ4ihNhiUcKAtACgpsmt
+X+oG13EqOSWRd6+6hXFyR0=
=zQXA
-----END PGP SIGNATURE-----

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]