#ietf102 — Public Fediverse posts

Did the #ietf doom you to this? Yes! #ietf102

Drafts SHOULD be implemented and MUST NOT have bugs #ietf102

Towards a Francophone Internet #ietf102

Is it ethical to post a picture of a cute kitten when you want your work to be adopted by the working group? #IETF102

Back to #DNS stuff, with the second meeting of DNSOP, dedicated to new work.

ATR (sending a truncated response immediately after the large response, in case broken stupid middleboxes block the large - possibly fragmented - response).

#IETF102

#QUIC #HumanRights review by Beatrice Martini. Among the good thing, it reduces the digital divide (improves things on high loss connections), it improves privacy (some transport information is now encrypted BUT there is the contentious issue of the spin bit).

#IETF102

Other #HumanRights review team work: SUIT, firmware update for #IoT devices (mostly #privacy issues). #IETF102

Big discussion at the #IETF : even the Human Rights group talks about meeting venue choice. (gender-neutral restrooms, childcare like at the last #RIPE76, etc). #IETF102

Confession: the #anonymity draft didn't advance because the author is lazy. #HumanRights #IETF102

Funny: in China, a DPI #censorship device examines all the TCP flows (not just HTTP and DNS) so protocol Echo (RFC 862) sees censorship as well, which can be used to detect it. #IETF102

One of the tricks of #Augur is to use "infrstructure" machines (such as routers) to do the #censorship measurements, to reduce the risk for individual users. #IETF102

One of the issues of #censorship measurement is the risk for the people hosting a probe. #OONI, for instance, has a very detailed consent form. #ethics #IETF102

Roya Ensafi on stage to talk about #censorship measurement. #IETF102

Speaking of that, a french senator just proposed automatic censorship on the Internet https://www.nextinpact.com/news/106750-une-senatrice-ps-veut-suspension-automatique-contenus-illicites-en-ligne.htm

Human Rights Protocol Considerations group starting at #IETF102. Agenda : https://datatracker.ietf.org/meeting/102/materials/agenda-102-hrpc-05 #UDHR

Power is out but #iccrg continues #ietf102

The power failure map during #IETF102 on #HydroQuébec Web site

Found. Power is back at #IETF102. #InternetResilienceDrill

Does anyone know where is the power switch in the Fairmont Queen Elizabeth? #IETF102

https://twitter.com/shivan_kaul/status/1020005910574399490 #India #digitalDivide #IETF102

Apparently, the network at #IETF102 runs on battery: traceroute still works. Still no power.

Power failure in the #IETF102 hotel. Will this toot go through?

Lessons: don't use short TTLs. Don't use Amazon #DNS resolvers, which cap TTL. Caching is your friend. #IETF102

In the case of partially successful dDoS attacks, a good part of the requests received by auth. #DNS servers is friendly fire: resolvers keep retrying, adding their requests to the dDoS. #IETF102

First test was the doomsday scenario (auth. servers 100 % down). With a less successful dDoS attack (10 % of requests serviced), experience shows that #DNS robustness works: all clients eventually get an answer (may be after a longer time). "DNS is good engineering" #IETF102

Now, emulating a dDoS with iptables and measuring if #DNS caches really help. With a TTL of 60', 30 to 70 % of the clients still receive a reply during one hour.

After one hour, it does not drop to 0 %: some clients still get stale answers.

Executive summary: trust caching, and stop using ridiculously short TTLs.

#IETF102

Caches protect a lot during dDoS attacks, Even if all authoritative name servers are down, the data are still served. (Actual efficiency of caching measured with RIPE Atlas probes, "reality matches theory".) #DNS #IETF102

Now, #DNS reliability "When the dike breaks"

How does the DNS manage during a dDoS attack? by Giovane Moura During the attack against the root in 2015, no effect for users. During the Dyn attack, a lot. Why?

#IETF102

Measuring MTU on the internet #maprg #IETF102 @mamiproject

Brian Trammel detecting remotely the load of a network. (Send packet, note the RTT, and thanks to bufferbloat, you will know the load from the RTT.) #IETF102

Also, reordering is just a minority of cases, but there is a long tail (old packets suddenly appearing). So, it is probably better to optimize at the start of a #QUIC session for no reordering, then to adjust. #IETF102

Giovane Moura on stage for #Dmap "automating #DNS measurements" Many tools but hard to make them work together. Solution : https://dmap.sidnlabs.nl/ #IETF102

Why not simply using #SASL for #EPP instead of reinventing the wheel with our own login framework? (Disclaimer: I never managed to understand SASL.) #IETF102

Core infrastructure is aging and not designed to be underwater #IETF102 #anrw