Reply to comment:

Tom Arnfeld
Super interesting. I have a couple of questions; Could you share more about the query traffic sample you used to generate traffic to unbound from the stubs? Also, have you spent any time looking at the real world impact of using TCP for recursive to authoritative? If one of the big sells for using TCP is better RTT measurements and retransmit logic, then this might help for unicast authoritatives hosted in bad networks, or geographically far away from their customer’s resolver. Are you aware of any stubs or resolvers that have options to prefer TCP configuration like this? Do you think it would be safe to enable TCP-first in stubs at scale, without causing resourcing issues on existing resolvers? Same question goes for resolver to auth. Thanks for a well written article!