I have had this very same problem for some time. I even sent a nasty-gram
to them about it, I had incorrectly assumed it was a DNS problem at first,
but then just realized that their mail servers AND dns servers where simply
overloaded. I was forced to send this email to them because our mail queue
was growing so large I almost implemented an 'AOL' filter. To this day I
still have mail queueing to AOL but it has not growning too large (yet).
Here is the reply I received for them about this problem. Please feel free
to flame them (not me please!). Notice the date.
Date: Wed, 09 Apr 1997 8:13:27 EDT
From: Michael Runge <firstname.lastname@example.org>
Subject: Re: Mail Problems
> Please resolve your DNS problem. As you can see from the partial listing
> below <deleted>, you are adversely affecting mail service on the internet.
> Otherwise, stop pretending to be an ISP, return to your previous
> role as an "island" on the internet and allow the rest of us to move on.
This is not a DNS problem. It is related to the volume of mail that we
receive. So much mail comes in that the inbound gateways are having
trouble keeping up at some points. We are putting more gateways in place
to keep up with the volume of mail that we receive.
I, too, would not compare us with an ISP. Our connectivity is much better
than any ISP that I know of. Roughly one of five Internet users access
the net through us, according to some numbers I saw a while ago. We do
more volume in email than anyone else in the world. With regard to our mail
gateways backing up, we're just seeing the growing pains much sooner than
anyone else because of the numbers we do.
If one defines an ISP as an access provider that gives you the ability to
run TCP/IP based clients over a TCP connection, then we indeed qualify. As
a matter of fact, every user that logs in is assigned a dynamic IP address
and can run any TCP compliant client over the AOL connection. If you have
AOL 3.0 installed, dial in, shell out, and change directory to where aol
is installed and then cd to net\osr1. In there you will find netstat and
can run it to see your IP address.
With regard to the MXes pointing to CNAMEs... this is not an error, it is
by design. As you know, DNS uses UDP for transport and the UDP packet is
limited in size. Since we have so many gateways to advertise (to keep
folks from getting refused connections) we can't fit all of our gateways
into the packet. The data gets truncated as a result. This is not a problem
for some people, but there are many non RFC compliant MTA implementatations
(imagine that) and when they do an MX query on our truncated data -- instead
of just using what they got, they throw away all of the data. Many SMTP MTAs
not properly understanding how to deal with truncation (misreading RFC 1123),
interpret any kind of truncation as meaning that all the data has to be thrown
away, when in fact you only throw it away if the truncation occurred in the
Authority section. So by having the MX point to a CNAME, the gateways are
broken down into more manageable chunks and aren't all advertised in one go,
which is great for everyone out there with the broken MTAs. After
with Paul Vixie and other DNS gurus, it is our understanding that this is
greatest solution -- it is the only solution presently -- but is not
illegal, and does help us solve the much more severe problem of DNS
If you are running bind 4.9.5, you can configure it so that it doesn't log
the annoying "MX points to a CNAME" messages :)
We're aware of the problem, and are working on a better solution
for it. However, in the meanwhile, people who do query logging will
run into these error messages, even though they're not technically
problematic. As for the problems reaching our mail gateways... I anticipate
that it will not be a problem in the very near future.
AOL Network Operations
-------------------End Of Message---------------------