Re: [TLS] ECH-HRR Design Team output

[Stephen Farrell <[e-mail]>]

Hiya,

Wasn't sure which email to reply to, but this one'll do...

Until very recently I was of the opinion that ECH split-mode
when one hits HRR with our current design was significantly
problematic.

The main reason was that supporting that scenario with
haproxy seemed like it'd require significant change to the
application, as haproxy is basically designed to only look
at the first ClientHello message when acting as a frontend
that doesn't terminate the TLS session, but just routes
sessions based on the first ClientHello SNI. (Doing ECH
decryption on the 1st ClientHello was ok, but getting at
the 2nd one with all the right state was the problem due
to how the application code was architected.)

With some more delving into things, I've managed to find
a relatively doable (if slightly iccky) way to support ECH
split-mode+HRR in both haproxy and nginx, so I think we can
strike this as an objection to the current design.

This likely affects a few of the issues in github, (in a
good way:-) but I didn't comment on those as it doesn't
quite fit any of 'em exactly. It should in any case help
as the editors set out to close issues,

My proof-of-concept code for nginx [1] and haproxy [2]
is available and if anyone's interested in details I'm
happy to point out specifics. (Note that that code is
far from production-ready but I think it's good enough
to show support is possible without that much change.)

Cheers,
S.

[1] https://github.com/sftcd/nginx/tree/ECH-experimental
[2] https://github.com/sftcd/haproxy/tree/ECH-experimental

On 03/06/2021 22:03, Christopher Wood wrote:
> Hi folks,
> 
> Since the last IETF meeting, the HRR Design Team has worked through 
> the issue of dealing with HRR. A writeup of the team's efforts, 
> discussions, and recommendations is here:
> 
> https://github.com/tlswg/draft-ietf-tls-esni/wiki/HRR-Design-Team
> 
> We believe next steps are to merge the following PRs:
> 
> - https://github.com/tlswg/draft-ietf-tls-esni/pull/423 - 
> https://github.com/tlswg/draft-ietf-tls-esni/pull/422
> 
> To that end, we would appreciate additional review. Hearing no 
> objections, we'll plan to merge these at the end of next week and 
> close out the relevant issues.
> 
> Thanks, all!
> 
> Best, Chris
> 
> _______________________________________________ TLS mailing list 
> [e-mail] https://www.ietf.org/mailman/listinfo/tls