> I can understand the idea > that ADoX trades latency and reliability for privacy and security, > and resolver operators may be hesitant to make that trade without > consulting the clients. Accordingly, I can see the appeal of an > opt-in signal from the client. However, I doubt that this draft > represents the best approach. > > This draft implies that a single resolver instance performs both > ADoX and non-ADoX queries to the same authorities. This creates > a huge number of complications. For example, some insecure records > may be present in cache that were fetched previously over UDP. > Should they be refetched over ADoX to prevent forgery, even though > this reduces privacy? What if they are only the glue records for > a signed child zone? What is the privacy response status of a > query whose recursive stages used a mix of cached and uncached, > secure and insecure, ADoX and non-ADoX queries? What about > QNAME-minimized queries that only reveal a TLD? Should queries > with EDNS Client Subnet be treated as more sensitive in some way?
Fundamentally, this draft is about an option between stub, forwarder and recursive resolver. For production use, the stub requests some level of privacy. The option can also be used for debugging and diagnostics. So I think the question for the working group is whether it is useful to have such and option and whether such an option seems feasible. Currently the request has 4 different values. Extending that to 5 is no proble, but if that would have to become a spearate IANA registry, then the option doesn't seem feasible. Personally, I think this option would be a good idea. We have the problem that we don't really know what privacy means in the context of DNS. Traffic between recursor and authoritative is most often plaintext and at best unauthenticated TLS (with some private interconnects where TLS keys are pinned). Beyond that, privacy is highly sensitive to meta data, the IP addresses used, but also timing of requests and replies. Where confidentiality and integrity are well-studied areas, leaking of data through side channels is hard to model. So in my opinion, it is best to just start. Deploy some code, see what works, what doesn't and iterate. The attraction of this option is that it offers a host fine-grained control over the requested privacy. For example, a host may request pool.ntp.org with no privacy at all. A browser may request privacy with a fallback to unecrypted transports. But a browser's 'Private Window' may (in the future) request strict privacy using an authenticated encrypted channel. This is not something that is easy to set up with different addresses for the host's resolvers. _______________________________________________ DNSOP mailing list -- [email protected] To unsubscribe send an email to [email protected]
