- From: Willy Tarreau <w@1wt.eu>
- Date: Tue, 25 Jul 2017 08:26:29 +0200
- To: Martin Thomson <martin.thomson@gmail.com>
- Cc: Kazuho Oku <kazuhooku@gmail.com>, Subodh Iyengar <subodh@fb.com>, Ilari Liusvaara <ilariliusvaara@welho.com>, Mike Bishop <Michael.Bishop@microsoft.com>, HTTP Working Group <ietf-http-wg@w3.org>
On Tue, Jul 25, 2017 at 03:55:17PM +1000, Martin Thomson wrote: > On 25 July 2017 at 15:22, Willy Tarreau <w@1wt.eu> wrote: > >> There's probably some way that we can describe rules for retrying at the > >> intermediary. It does save a round trip between the user agent and > >> intermediary, and there is some potential value in having intermediaries > >> repair things. But we lose some of the nicer end-to-end properties by > >> doing so. > > > > In fact I think there are some situations where it can safely be done, > > specifically if the request was received as 1-RTT, or if the handshake > > was completed in the mean time because it then becomes equivalent. > > Those are probably both cases where retries are inappropriate. > > If the request arrives over 1-RTT, the intermediary probably shouldn't > be using 0-RTT. I see your point. Some users may consider the intermediary as part of the client, extending it to completely deal with retries, hiding them from the client. > If the connection was completed since the 0-RTT attempt was made, that > doesn't mean that it wasn't replayed already elsewhere. But that's the same if the client retries, which is why clients are encouraged not to take too much risk by sending only apparently idempotent methods over 0-RTT. Willy
Received on Tuesday, 25 July 2017 06:27:00 UTC