Re: Consensus Calls for Transport/TLS issues, post-Cupertino

[Kazuho Oku <[e-mail]>]

2019年10月22日(火) 9:42 Mark Nottingham <[e-mail]>:

> The following issues have proposals for resolution, and discussion so far
> seems to support consensus to accept them. If you object, please do so on
> the issue or in response to this message (changing the Subject
> appropriately!). Absent any pushback, we'll direct the editors to
> incorporate them next week.
>
> See <https://github.com/quicwg/base-drafts/projects/5> for the current
> state of issues in the Late Stage process, itself defined at <
> https://github.com/quicwg/base-drafts/blob/master/CONTRIBUTING.md>.
>
> * #3097: Is CONNECTION_CLOSE ACK-eliciting?
>    The proposal is <https://github.com/quicwg/base-drafts/issues/3097>
>
> * #3085: Stateless reset detection should be datagram-based
>    The proposal is <https://github.com/quicwg/base-drafts/pull/2993>
>
> * #3054: Label for key updates
>    The proposal is <https://github.com/quicwg/base-drafts/pull/3050>
>
> * #3046: Handling of Retire Prior To field
>    The proposal is <https://github.com/quicwg/base-drafts/pull/3096>
>
> * #3037: Require peers to check if RETIRE_CONNECTION_ID sequence number is
> valid
>    The proposal is <https://github.com/quicwg/base-drafts/pull/3036>
>
> * #3027: Codes for frame encoding errors
>    The proposal is <https://github.com/quicwg/base-drafts/pull/3042>
>
> * #2944: Layout of PreferredAddress
>    The proposal is to close with no action.
>
> * #2928: Lift single-packet ClientHello requirement?
>    The proposal is <https://github.com/quicwg/base-drafts/pull/3045>
>
> * #2863: unrecoverable loss pattern leads to deadlock
>    The proposal is <https://github.com/quicwg/base-drafts/pull/3121>
>

I feel sorry but I object to the proposed approach. As I have pointed out
in my comments in #2863, not discarding the Handshake key forever breaks
our assumption that no Handshake packets will be sent or received once the
handshake is confirmed. We missed that there are these side effects when we
rush to the conclusion at the end of the interim.

Specifically, we'd be required to define how Handshake packets deal with
migration (does the path for Handshake packets change? If it does, how do
we fill in the SCID field of the Handshake packet? If it does not, until
when do the endpoints need to retain the original path?).

I think we should fix the deadlock, rather than changing other parts of the
protocol that currently depend on the Handshake key being discarded at some
point.


>
> * #2823: Do Initial secrets change after Retry packet?
>    The proposal is <https://github.com/quicwg/base-drafts/pull/2870>
>
> * #2741: Re-visit initial keys discard
>    The proposal is to close with no action.
>
> * #2152: Why does stateless reset have to be checked after MAC failure
>    The proposal is <https://github.com/quicwg/base-drafts/pull/2993>
>
> --
> Mark Nottingham   https://www.mnot.net/
>
>

-- 
Kazuho Oku