DomainKey-Signature: a=rsa-sha1; c=nofws; d=github.com; s=mailo; q=dns;
 h=Sender: Date: From: Reply-To: To: Message-ID: Subject: Mime-Version:
 Content-Type: Content-Transfer-Encoding;
 b=OOh3m+pTungwQCooHJWD2VPbm3H5kWBgA8WF2rV9oFroJ1iqrHmvetVnPtkj0YpIHIiRM6
 btXQXmj0tEL4WQIUi1YubhUaaMgFVNwPdqgtOP4nY/RGtCfdNVmGc8N6YSWpBGzundrEgdla
 x5i3lJmzzj2GCPctiavT9F2zX+V4w=
Sender: martin.thomson=gmail.com@github.com
Date: Fri, 09 Jun 2017 06:36:49 -0700
From: Martin Thomson <martin.thomson@gmail.com>
Reply-To: Martin Thomson <martin.thomson@gmail.com>
To: quic-issues@ietf.org
Message-ID: <593aa4719fbf0_57a53f9b3a5d3c3856227@hookshot-fe6-cp1-prd.iad.github.net.mail>
Subject: [quicwg/base-drafts] aa475c: Some minor editorial improvements
Mime-Version: 1.0
Content-Type: multipart/mixed;
 boundary="--==_mimepart_593aa4719f7ec_57a53f9b3a5d3c385610";
 charset=UTF-8
Content-Transfer-Encoding: 7bit
Archived-At: <https://mailarchive.ietf.org/arch/msg/quic-issues/6RH1CmGBexE8kZZfLT0lZnPU0e4>
Precedence: list


----==_mimepart_593aa4719f7ec_57a53f9b3a5d3c385610
Content-Type: text/plain;
 charset=UTF-8
Content-Transfer-Encoding: 7bit

  Branch: refs/heads/invisible_reset_symmetric
  Home:   https://github.com/quicwg/base-drafts
  Commit: aa475cac0471196557a34448ff61680ab737bb0a
      https://github.com/quicwg/base-drafts/commit/aa475cac0471196557a34448ff61680ab737bb0a
  Author: Martin Thomson <martin.thomson@gmail.com>
  Date:   2017-06-09 (Fri, 09 Jun 2017)

  Changed paths:
    M draft-ietf-quic-transport.md

  Log Message:
  -----------
  Some minor editorial improvements


  Commit: 8fc96e85fb5240ad93e7f4a3bbb7d525d000b12f
      https://github.com/quicwg/base-drafts/commit/8fc96e85fb5240ad93e7f4a3bbb7d525d000b12f
  Author: Martin Thomson <martin.thomson@gmail.com>
  Date:   2017-06-09 (Fri, 09 Jun 2017)

  Changed paths:
    M draft-ietf-quic-transport.md

  Log Message:
  -----------
  Symmetric stateless reset

This is disgusting.  It's still not symmetric, because only the server
can propose new connection IDs.  That means that the best option is to
define a new message for providing the client token.

The client can't send a token in transport parameters because those
aren't encrypted.  That means that it needs to send a frame.  I
considered using NEW_CONNECTION_ID, which would also allow for
asymmetric connection IDs, but that doesn't work very well for other
reasons: it makes the stateless reset trivially distinguishable for one.

Thus, we have this.

I would prefer NOT to merge this.  It's complicated and gross and I
still believe we don't need to do this.  The primary reason, which I
don't think was correctly articulated at the interim, was that this
feature (stateless reset) is only really useful if you have are
stationary and therefore likely to receive packets immediately after a
reboot or crash.  Most endpoints, particularly with IPv6, will have new
IP addresses in that state and will not be listening on ports.  I
realize that means for stationary clients that means that they will need
to fall back to ICMP, or the server detecting a broken path, but I think
that is preferable to this mess.


Compare: https://github.com/quicwg/base-drafts/compare/aa475cac0471^...8fc96e85fb52
----==_mimepart_593aa4719f7ec_57a53f9b3a5d3c385610--