[Yeti DNS Discuss] Dealing with IPv6 Fragmentation in the DNS

Geoff Huston gih at apnic.net
Tue Sep 5 10:46:11 UTC 2017

> On 5 Sep 2017, at 8:30 pm, Stephane Bortzmeyer <bortzmeyer at nic.fr> wrote:
> On Tue, Sep 05, 2017 at 06:09:02PM +0800,
> Davey Song <ljsong at biigroup.cn> wrote 
> a message of 9 lines which said:
>>> ipv6 operators who don't permit the necessary extension headers
>>> and icmp messages to enter or leave their networks/hosts/apps must
>>> suffer.
>> The ASes in the middle may drop the fragments!
> According to RFC 7872 measurements (a must read, even if it is about
> extension headers, not specifically about fragments), problems occur
> mostly at the edge. Lot of good eradings in its references, too.
> 30.51% of Web servers do not receive fragments. Between 5.08% and
> 6.78% are before the destination AS.

As far as I can tell the experiment used public servers listed in the the "World IPv6 Launch” site and Alexa's list of the Top 1-Million Web Sites. In other words the measurement was header towards the servers rather than towards the clients.

The experiments we’ve recently completed at APNIC performed the IPv6 Fragmentation Header experiment towards the clients. In the fist case we measured from an authoritative name server towards the set of visible resolvers, and the second case measured from a web server towards the end clients. Its looking (roughly) in the opposite direction to RFC 7872


More information about the discuss mailing list