[dns-operations] DNS cookies in a mixed resolver anycast environment

[Tobias S. Josefowitz]

On Fri, May 31, 2019 at 7:13 PM Ondřej Surý <ondrej at sury.org> wrote:
>
>  The whole point of DNS Flag Day is that we don’t have any obligation to resolve domains running on broken DNS server. It have become “fix you sh^Htuff” instead of “fix our stuff” how it have been before.
>
> Also it’s perfectly OK (on technical level) to not support EDNS, but in this particular case (p4.no) it’s the fact that the server returns FORMERR + OPT RR that’s causing the resolution failure as 6891 says:
>
> > Responders that choose not to implement the protocol extensions
>
>    defined in this document MUST respond with a return code (RCODE) of FORMERR to messages containing an OPT record in the additional
>    section and MUST NOT include an OPT record in the response.

As an implementor and operator of multiple authoritatives that choose
to not implement "the protocol extensions", why is that anyway? Never
having even heard of EDNS at some point in history certainly was the
norm for responders, so must have been (and still is) handled
gracefully. I never appreciated that RFC 6891 introduces an additional
RTT for implementations choosing not to implement (in cases where the
querying party is not already sending non-EDNS queries).