Markus Stumpf wrote:
I am currently trying to contact people in DE. Peter Koch who is doing
the monthly DNS host count for DE domains is willing to add a check for
MX hosts and will provide information about the number of unique (based
on IP address) MX server used for DE domains.
starting with 11291065 MX RRs coming out of this month's DE hostcount (see
http://www.TechFak.Uni-Bielefeld.DE/~pk/dns/hostcount/latest.html or
http://www.ripe.net/ripencc/pub-services/stats/hostcount/ for information on
what it is and how it works), the individual MX targets were identified,
bogus ones deleted and the remaining ones were fed a resolver to find the
number of unique IP addresses:
11291065 MX RRs
7600000 ~ number of 2nd level DE domains
<http://www.denic.de/en/domains/statistiken/>
1058962 unique MX RR targets
1057546 valid MX RR targets (no '.', IP addresses, ...)
146224 unique IP addresses after resolving MX RR target names
144582 non-bogus IP adresses (ignoring RFC 1918 addresses,
unallocated address space, multicast, ...)
Some remarks:
o Targets may reside both inside and outside DE.
o Some larger address space holders ("class B", universities etc) tend
to define MX RRs for almost all of their systems/addresses, sometimes
including dialin lines etc. Since MX is inbound only this may have an
adverse effect on the guesstimate for the proposal mentioned in the
subject line.
o The 144582 addresses belong to 26552 /24 address ranges.
o Approx. 75% of the addresses identified do have a working reverse
mapping (lead to one or more PTR RRs), the rest either yields NXDOMAIN
or has resolving problems (timeout, SERVFAIL). For NXDOMAIN I didn't
further differentiate 'not delegated' vs. 'not named in zone'.
o Viewing /24 zones (and ignoring RFC 2317 for the sake of simplicity)
80% of the /24's covered by the IP addresses found do have a working
reverse mapping.
So, this is just the data, I'm taking no position whether or not these
figures can be used to calculate/estimate the number or percentage of
``legitimate'' outbound SMTP clients.
-Peter