Here is the same data on EDNS buffer size setting from yesterday, weighed by use: EDNS buffer size for queries over IPv4
4096 38.437% 1232 24.888% 1400 17.489 1472 5.976% 1452 4.427% 1220 3.243% 512 2.277% None 0.665% Old habits die hard - 38% of queries use 4096 Is V6 any different? 4096 47.327% 1232 21.538% 1400 13.304% 1472 1.201% 1452 3.699% 1220 3.62% 512 3.060% None 1.790% Wow - that’s a higher query-weighted measure for IPv6. Seems that out there in DNS land the resolvers that folk use are pretty hung up on 4096 bytes for EDNS in both Ipv4 and Ipv6! Geoiff > On 10 Nov 2023, at 4:43 pm, Ralf Weber <d...@fl1ger.de> wrote: > > Moin! > > On 10 Nov 2023, at 13:33, Geoff Huston wrote: > >> Here’s an analysis of measurement of issues with IPv6 and DNS resolvers from >> a few years >> ago...https://www.potaroo.net/presentations/2017-09-29-xtn-hdrs-dns.pdf >> >> I have not returned to this measurement for some years as there appeared to >> be little interest in the results right up until now! >> >> If there is interest in this measurement work about the DNS, large responses >> and IPv6 failure rates from the working group I’m sure it can be revived! > > That would be fantastic as we since that presentation had a DNS flag day ( > http://www.dnsflagday.net/2020/ ) reducing the DNS buffer size to 1232 which > should no longer make it fragment over IPv6 UDP. I would hope that by now at > least most professionally run DNS resolvers would have implemented this. > > When I looked at your slide on the second more relevant experiment, with > clueless delegation I found that the top 10 bad ASes doing something about > this would eliminate the ~80% of the problem, so there could be some gains > made talking to these people if they still appear in your testing now. > > So long > -Ralf > --- > Ralf Weber > Principal Architect, Carrier Division > > Akamai Technologies GmbH > Parkring 20-22, 85748 Garching > phone: +49.89.9400.6174 > mobile: +49.151.22659325 > > Geschäftsführer: David Matthew McDonald Aitken, Justyna Kalina > Jankowska > Sitz der Gesellschaft: Garching > Amtsgericht München HRB 129886 _______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
