Re: [Doh] [Ext] Re: Associating a DoH server with a resolver

[Ben Schwartz <[e-mail]>]

On Sat, Mar 23, 2019 at 5:41 PM Erik Nygren <[e-mail]> wrote:

> On Tue, Oct 23, 2018 at 11:40 PM Adam Roach <[e-mail]> wrote:
>
>>
>> So, e.g., interested clients would query for a URI record of type
>> _doh._tcp.resolvers.arpa, and get back a full HTTPS URL as a response.
>>
>
> Reviving this suggestion after reading some recent threads, this approach
> (query your local OS-configured resolver for a URI record with a
> well-known name)
> seems like a reasonably cleanish way to opportunitistically get a DoH
> server URI
> associated with the currently configured resolver.   It's arguably less of
> a hack
> than cramming things into a TXT record on a reverse DNS IP.
>

Technically, a DoH configuration is a "URI template", not a URI.
(Typically, this means it contains a substring like "{dns}".)  Do you
consider this similar enough to be good use of the URI RRtype?


> (Martin asked on a more recent thread something along the lines of "why
> don't
> we just do a .well-known lookup against the network-configured resolver,
> and something
> like this seemed like the best equivalent.)
>
> For better or worse it would also work when a caching forwarding resolver
> is present
> in passing through the record from whatever the caching forwarding
> resolver (eg,
> dnsmasq in a home gateway) was pointing to.  It also has the benefit that
> clients don't
> need to know the IP address of their resolver, they just need to be able
> to do a URI record
> lookup against a well-known name.
>
> Especially of the OS helps mediate DoH server selection (which would be
> beneficial
> long-term regardless) the exact OS APIs that exist today are somewhat less
> interesting.
>
>       Erik
>
>
> (Purely for enterrainment value: this well-known name could be in some
> other class
> like CHAOS rather than IN, like some other recursive-resolver-synthesized
> names
> like "version.bind txt chaos".  This is not meant to be a serious
> suggestion.)
> _______________________________________________
> Doh mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/doh
>