Hi, > On 2 Jul 2018, at 11:53, Tony Finch <d...@dotat.at> wrote: > > Paul Vixie <p...@redbarn.org> wrote: >> >> for QTYPE=A, add AAAA as a desired additional data type. >> >> for QTYPE=AAAA, add A as a desired additional data type. > > How does the server signal to a client that made an A query that there > are no AAAA records so the client does not need to make a followup AAAA > query? My answer: use DNSSEC :-) > > What are the incentives to implement this? The current state of the art is > happy eyeballs version 2, which specifies concurrent AAAA and A queries; > the client has already made both queries before it finds out it only > needed to make one. I'm not sure how a client could know in advance that > it only needs to make one query.
This really isn’t about incentives, but not making the situation worse. A single query without signalling (opportunistic) and without state will double the latency compared to dual A + AAAA query fired at the same time. > I think there would be some benefit to this between auth and recursive, > provided the recursive server eagerly validates additional records and > promotes them to the authoritative answer level of RFC 2181 trust. So, instead of adding additional complexity of validating and storing the unsolicited information in additional section, the “prefetch” code in resolvers could be enhanced to pre-fill the cache with additional records when specific RTYPE is requested… Ondrej -- Ondřej Surý ond...@isc.org _______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
