On Fri, 23 Sep 2011 20:50:27 +0200, Alessandro Vesely said:
Most SMTP servers duly lookup the client's IP and annotate the
resulting name as comment in Received fields. However, I don't
recall
denying SMTP access based on the "iprev" test (as RFC 5451 named it.)
Was it ever à la mode to do so?
At one time, the net was still small enough that it was a safe
assumption that
if you got mail from an IP address that didn't have a valid rDNS, it
was (a) a rare
event because (b) a missing rDNS meant the provider was asleep at the
wheel.
Now-a-days, most providers have automatic provisioning systems that
assign rDNS
for customer addresses, so most of Vint Cerf's famous 140 million
compromised
machines have an rDNS entry, which means it's not that effective
anymore.
(What *is* used a lot today is 'rDNS looks like a customer
cablemodem/adsl connection')
Another name for the iprev test is "Forward Confirmed reverse DNS" (FCrDNS).
With Postfix you configure it with the two commands
reject_unknown_reverse_client_hostname
reject_unknown_client_hostname
We use this check since years as our first defense against botnet spam with
great success. In the last 7 days we rejected emails for nearly 22.000.000
recipients. 49% did not have a PTR record, 29% did not have a matching A
record. Therefore the FCrDNS was responsible for 78% of all rejections. This
means your statement, that this check is not working, is definitely not true.
However you have to live with a moderately false positive rate. Before we
implemented the check, we analyzed out traffic for 3 months and build an
automatic whitelist with 4.000 wrongly configured MTAs. Since the beginning of
the check we get about 1-2 false positives per week reported by our users. This
second whitelist has 230 entries at the moment. This means about 4% of the MTAs
we accept emails from are wrongly configured. We can live with that.
Michael