[dns-operations] Some DNSSEC adoption data points, anyone know of more comprehensive surveys?
Viktor Dukhovni
ietf-dane at dukhovni.org
Wed May 2 05:25:04 UTC 2018
> On May 1, 2018, at 4:54 PM, Ray Bellis <ray at isc.org> wrote:
>
> Have you looked at Bert Hubert's HyperLogLog method for estimating the
> number of signed delegations in a zone by performing statistics on the
> distribution of the NSEC3 hashes?
>
> <https://indico.dns-oarc.net/event/26/contributions/437/attachments/395/671/hyperloglog_1.pdf>
Ah yes, "The German tank problem", I'd seen this a while back and forgot
all about it. The most recent numbers reported at:
https://powerdns.org/dnssec-stats/
are from 2017/02/03, estimating 7,375,455 domains, which is much better
than what I've seen elsewhere. A comparison of that report with my
actual (but often incomplete) counts gives me a clear sense of where
the gaps lie. Note my numbers only include domains where the DNSKEY
RRset validates, excluding delegates where DNSSEC fails). The top 25
suffixes comparison is:
suffix bert's # my number complete zone?
------ -------- --------- --------------
nl 2618262 1409210
com 588136 853800 Y
se 607290 840042 Y
fr 337751 412562 Y 30-day delayed
no 431968 337863
cz 638970 301238
eu 365356 297904
com.br 771183 229670
nu 84151 139901 Y
be 132395 130034
net 113548 121870 Y
hu 111436 99489
org 74039 90460 Y
pl 32193 75955
de 62775 49045
info 28651 35915
ch 16764 31964
ovh 21505 22555 Y
dk 22422 21672
biz 2125 18474
pt 15334 11697
es 7200 11614
io 7502 7436
co.uk 17909 7339
xyz 8548 6671 Y
So, at the moment, the largest obvious coverage gaps are .NL,
.COM.BR, .CZ and .NO. It'd be nice to compare against any more
recent data Bert has collected...
Brazil is clearly my weak spot, if some slightly used, second-hand,
reasonably comprehensive list of .BR domains fell into my lap, I'd
be grateful. I seem to be getting much better fractional coverage
for the other domains with substantial DNSSEC deployments.
DNSSEC adoption under .biz seems to have increased 9-fold or more,
or selse the NSEC estimates from a year ago were particularly
skewed by the non-uniformity of NSEC distributions.
--
Viktor.
More information about the dns-operations
mailing list