Dear DNSOP,

John Todd and I have put together a draft that adds a client opt-in to ECS:

https://datatracker.ietf.org/doc/draft-farrokhi-dnsop-ecs-opt-in/
https://github.com/Quad9DNS/id-ecs-opt-in (working copy)

We understand that ECS is a controversial topic. RFC 7871 says in its
abstract: "Since it has some known operational and privacy shortcomings,
a revision will be worked through the IETF for improvement." Ten years
later, we would like to introduce an opt-in option. 

Today users have limited control over whether their resolver sends
client subnet information to authoritative servers. When using an
ECS-enabled resolver such as 8.8.8.8 or 9.9.9.11, a client can "opt-out"
with a SOURCE PREFIX-LENGTH of 0, but to ask for a shorter prefix it has
to supply its own address, which it does not know when it is behind NAT
or VPN. 

As a DNS resolver operator, we have to run multiple instances to
accommodate users who want ECS and users who do not. Users choose by
pointing to the right instance. 

This draft defines a new EDNS(0) option. A client can use the option to
ask the resolver to forward its address information, and to limit how
many address bits are forwarded. A resolver that implements it forwards
the client subnet only when asked to do so. A resolver that does not
implement it will ignore the option and keeps its current behavior. We
believe this improves end-user privacy and gives users control over
their own address information, and also allows a single resolver
instance to serve both kinds of users. 

The intended status is Experimental.  We use a new option because the
signal does not fit inside the ECS option without conflicting with RFC
7871 Appendix A. 

We welcome your comments and reviews, and would like to know whether
DNSOP is interested in this work.

BR,

-- 
Babak Farrokhi

Attachment: signature.asc
Description: PGP signature

_______________________________________________
DNSOP mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to