Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(290)

Issue 446283002: Include the socket receive buffer in the connection handshake. (Closed)

Created:
6 years, 4 months ago by ramant (doing other things)
Modified:
6 years, 4 months ago
Reviewers:
wtc, Ryan Hamilton
CC:
chromium-reviews, cbentzel+watch_chromium.org, wtc, rjshade, cyr_google.com
Base URL:
https://chromium.googlesource.com/chromium/src.git@improve_SendAlgorithmSimulator_72332690
Project:
chromium
Visibility:
Public.

Description

Include the socket receive buffer in the connection handshake. PRESENCE_OPTIONAL The goal is to eventually remove the TcpCongestionFeedbackFrame. Currently this sends the same value for every packet. All it contains is the tcp_receive_window (now called socket_receive_buffer) which is fixed at 256000. So it can be sent in the connection establishment and then never retransmitted. Because only one type of CongestionFeedbackFrame can be sent with an ack, this frees up space for a TimestampFrame which is the motivation for removing the TcpCongestionFeedbackFrame. Merge internal change: 72334240 R=rch@chromium.org

Patch Set 1 #

Total comments: 2
Unified diffs Side-by-side diffs Delta from patch set Stats (+52 lines, -1 line) Patch
M net/quic/crypto/crypto_protocol.h View 1 chunk +3 lines, -0 lines 0 comments Download
M net/quic/quic_config.h View 2 chunks +12 lines, -0 lines 0 comments Download
M net/quic/quic_config.cc View 4 chunks +23 lines, -1 line 0 comments Download
M net/quic/quic_config_test.cc View 6 chunks +11 lines, -0 lines 0 comments Download
M net/quic/quic_protocol.h View 1 chunk +3 lines, -0 lines 2 comments Download

Messages

Total messages: 6 (0 generated)
ramant (doing other things)
6 years, 4 months ago (2014-08-07 01:25:07 UTC) #1
Ryan Hamilton
lgtm
6 years, 4 months ago (2014-08-07 17:32:44 UTC) #2
wtc
lgtm https://codereview.chromium.org/446283002/diff/1/net/quic/quic_protocol.h File net/quic/quic_protocol.h (right): https://codereview.chromium.org/446283002/diff/1/net/quic/quic_protocol.h#newcode73 net/quic/quic_protocol.h:73: const QuicByteCount kDefaultSocketReceiveBuffer = 256000; This is 250 ...
6 years, 4 months ago (2014-08-07 22:04:39 UTC) #3
ramant (doing other things)
https://codereview.chromium.org/446283002/diff/1/net/quic/quic_protocol.h File net/quic/quic_protocol.h (right): https://codereview.chromium.org/446283002/diff/1/net/quic/quic_protocol.h#newcode73 net/quic/quic_protocol.h:73: const QuicByteCount kDefaultSocketReceiveBuffer = 256000; On 2014/08/07 22:04:39, wtc ...
6 years, 4 months ago (2014-08-07 22:16:14 UTC) #4
Ian Swett
On 2014/08/07 22:16:14, ramant wrote: > https://codereview.chromium.org/446283002/diff/1/net/quic/quic_protocol.h > File net/quic/quic_protocol.h (right): > > https://codereview.chromium.org/446283002/diff/1/net/quic/quic_protocol.h#newcode73 > ...
6 years, 4 months ago (2014-08-07 22:33:26 UTC) #5
ramant (doing other things)
6 years, 4 months ago (2014-08-07 23:08:12 UTC) #6
Message was sent while issue was closed.
On 2014/08/07 22:33:26, Ian Swett wrote:
> On 2014/08/07 22:16:14, ramant wrote:
> > https://codereview.chromium.org/446283002/diff/1/net/quic/quic_protocol.h
> > File net/quic/quic_protocol.h (right):
> > 
> >
>
https://codereview.chromium.org/446283002/diff/1/net/quic/quic_protocol.h#new...
> > net/quic/quic_protocol.h:73: const QuicByteCount kDefaultSocketReceiveBuffer
=
> > 256000;
> > On 2014/08/07 22:04:39, wtc wrote:
> > > 
> > > This is 250 KB. Just curious why we don't use a power of 2 (256 KB).
> > 
> > Good point. In "spdy", we used to 64 * 1024.
> > 
> > ianswett: WDYT about using power of 2 (or 256 * 1024).
> 
> Sounds good to me.  Let's do 256 * 1024

Sounds great. Will make the change in the internal source tree (when I back port
chromium changes).

thanks
raman

Powered by Google App Engine
This is Rietveld 408576698