I think the one proposal was very client-specific, which kind of ruled it out for a generic "aname" type. That was Ray's "HTTP" RRtype, that I did a deep dive on.
Basically, you are correct; the easiest path forward would be for client software upgrades to get actual DNS records (rather than rely on getaddrinfo), and do the indirection following (similar to and inclusive of CNAME and DNAME, in addition to the new "HTTP" type.) It avoids requiring recursives do the extra handling (analogous to CNAME chaining and ultimately returning A/AAAA, based on the original QTYPE) It definitely is the case that there is anti-consensus on the original ANAME spec which requires sibling records provisioned/populated/maintained by authority servers -- which is basically fatal to the original ANAME draft. I would definitely be willing to work on a true reset of ANAME that goes in the other direction. I suspect that's what Ray's reset will involve, and I believe that any effort on the current ANAME before the reset, would likely be wasted effort. I really want to avoid the issue where those who have, in good faith, contributed significant effort, want to hold onto that work, even if that work is ultimately counter-productive. Please see the discussion on the list from November 2018 timeframe, on the major issues.... Thanks, Brian On Tue, Mar 26, 2019 at 5:10 PM Tony Finch <d...@dotat.at> wrote: > Matthijs Mekking <matth...@pletterpet.nl> wrote: > > > > I think that would be the wrong direction. I believe there is a need to > > standardize the ANAME resolution process and so my suggestion would be to > > reduce the scope by focusing just on how to do that on the provisioning > side > > (and leave secondary servers and resolvers out of scope for now). > > >From past discussions, I didn't think there was any way we could get > consensus on the provisioning side. > > Dynamic lookups on authoritative servers are out, because it has to be > compatible with traditional secondaries. > > Updates on the primary are out, because that doesn't scale to large > numbers of zones. > > Sometimes a system might have known fallback addresses, but often it won't > (e.g. whether the DNS setup is or isn't coupled to a web provisioning > system). > > But I think it's reasonable to allow whatever provisioning mechanisms > there might be, provided the meaning of answers from auth -> rec have a > consistent meaning that resolvers can use. > > It's also really imortant that ANAME can work in multi-provider setups, so > there needs to be something approaching interoperable semantics for > importing a zone file into a provisioning system. (Though I think the > semantics will have to be very loose in this case, to allow for variations > between existing systems.) > > I haven't seen any objections to support for ANAME in recursive servers > so I'm surprised you think that is problematic enough to be removed. My > understanding was that recursive support is seen as better than trying to > do all the tricks on authoritative servers. > > Tony. > -- > f.anthony.n.finch <d...@dotat.at> http://dotat.at/ > safeguard the balance of nature and the environment > > _______________________________________________ > DNSOP mailing list > DNSOP@ietf.org > https://www.ietf.org/mailman/listinfo/dnsop >
_______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
