DNS Operations (DNSOP) Working Group IETF 96, Berlin Monday 18 July 2016 Minutes taken Paul Hoffman Text from slides not included here; please read the slides Chairs: Tim Wicinski, Suzanne Woolf Current WG documents: https://svn.tools.ietf.org/svn/wg/dnsop/doclist.html Also see https://datatracker.ietf.org/wg/dnsop/documents/ Agenda Bashing, Blue Sheets, etc, Nothing Updates of Old Work Three RFCs Three drafts in IESG evaluation Waiting on two docs Five drafts will come through the pipeline One draft in call for adoption Matt Larson ZSK size is increasing, KSK is rolling Lots of drafts are being considered for adoption draft-ietf-dnsop-terminology-bis status: Paul Hoffman Aaron Falk: If we can't agree on a term, maybe we should stop using the term Benno Overeinder: Are we also looking for implementations? Paul: No. draft-ietf-dnsop-nsec-aggressiveuse: Warren Kumari Aaron: Your amount dropped below where it started, are legitimate queries being dropped? Warren: We did check, and we're quite sure John Levine: You could instead be mirroring the root Warren: Yes, but this helps for some DDoS John: Also usable for IPv6 reverse DNS blacklists You have old advice Duane Wessels: Did you turn this on for everything? Warren: Yes TLS-TCP-DNS implementation: Sara Dickinson Lots of recent improvement, including this weekend at the hackathon Both authoritative and stubs Ondrej Sury: Can you add padding to the server tests? draft-bellis-dnsop-session-signal: Ray Bellis Ben Schwartz: How does interact with HTTP wire format draft? Ray: Should not be bad Ben: HTTP 2 allows out-of-order Aaron: The PLUS BoF on Thursday might be of interest, this might be a user of PLUS Stewart Cheshire: These sessions defines a session of a particular type, so it can run over any transport that handles it Do we use something small (will surprise tools) or EDNS0? Suzanne: take that to the list draft-wkumari-dnsop-multiple-responses: Wes Hardaker and Warren Kumari Ralf Weber: EXTRA record allows DOSing Wes: Suggest to have policy on when to use Christian Huitema: Browsers already do something like this Is this really worth it? Warren: There is more than the web Wes: This does not help a resolver that has already cached the information, of course Christian: Double-optimization may lead to worse performance Jim Reid: Similar to an ANY amplification attack Can we get some data on what the optimization benefit will before we move forwards on this? Warren: yes, we shoud Sara: How does this affect client-subnet Warren: Has to be from the authoritative server Marc Blanchet: Any problem with DNS64? Warren: Yes, DNSSEC Teddy Hogeborn: Issue with format of record Wes: This will probably change Teddy: This might help because of SRV David Lawrence: With client-subnet, you'll have to give the most specific answer An authoritative server should need to be sure that it was authoritative Wes: You have that issue today with DNSSEC and TTLs Hums: Lots of people want to hear more data first draft-bellis-dnsext-multi-qtypes: Ray Bellis Limited to one QNAME/QCLASS If we decide not to do any mulitple-query proposals, we should write a document why John Levine: What if you're getting a CNAME back? Has to be specified Mark Andrews: We can specify how to do this Kazunori Fujiwara: Wants a summary of why the previous proposals failed Ralf Weber: The only real use case is A/AAAA, the rest are not that useful optimizations Ray: Also MX/A David Lawrence: Likes this for A/AAAA Peter van Dijk: You are copying the QR bit: why? Ray: They certailny still exist. Helps prevent someone just mirroring. Shane Kerr: Can be a huge win with A/AAAA draft-woodworth-bulk-rr: John Woodworth Viktor Dukhovni: Is this record available for client queries? John: Yes Viktor: This will require on-the-fly NSEC John: If it is not on-the-fly, the client needs to be updated Should only use on-the-fly David Lawrence: Likes this. Don't attach this to a wildcard label. Teddy: It is not really a record type. Maybe it should be a directive instead, but then it doesn't work with AXFR. Maybe needs a new form of zone transfer. Ed Lewis: DNAME redirects but doesn't need to be signed. Just needs a signature on redirector. draft-muks-dnsop-dns-catalog-zones Tim: Read the draft because this might come up in the WG Special Names Portion: Suzanne Woolf Note from the minutes-taker: If you are reading this, you *really* should read all the slides, including Suzanne's draft-adpkja-dnsop-special-names-problem: Geoff Huston draft-tldr-sutld-ps: Ted Lemon draft-wkumari-dnsop-alt-tld: Warren Kumari draft-cheshire-sudn-ipv4only-dot-arpa: David Schinazi Discussion starts Paul Hoffman: Doesn't like the history section because it doesn't quote, it retells Alain: Doesn't think that there are preferences in draft-adpkja, please send to the list There is a fair amont of overlap in the problems listed Difference: how to evaluate a particular string Ted: draft-adpkja covers problems with 6761, draft-tldr covers the bigger problem Alain: Might want to publish separately because they are on different topics Joel Jaeggli: Expected an alternative draft, but thinks that adopting two would be bad Ralph Droms: Thinks draft-tldr is a superset of draft-adpkja Some things in draft-adpkja are not problems, but just not as well specified Which covers the larger set Ted: Did not cover problems with solutions Stewart: Thanks to Ted and Ralph Geoff: Part of the problem is what can be unilaterally solvable in the IETF draft-tldr cannot be addressed ourselve Suzanne: Maybe will have an interim soon