[tor-talk] exit node cache

jiang song luolisongjiang at gmail.com
Mon Oct 29 16:10:00 UTC 2012


>
> The second one is much closer to reality. :-) The exit nodes stops
> reading from the connection to the web server when it can't forward
> the data to the client fast enough.
>
> But the web server isn't slowed down/stopped by not sending ACKs but
> rather by reducing the window size to zero. ACKs are still sent as
> data is received (moduly shortly delaying acks for other reasons).
> This is all done within the TCP stack which does all this ACK and
> window size stuff; the tor process just reads from the socket.
>
>
--------------------------------------------------------------------------------------------------------
after reading some articles, my thought is:
for example, if you are watching youtube html5 videos using firefox, the
webserver will continue sending packets.
There is no server-side rate control, unless there is congestion inbetween,
in this case the firefox browser client will
delay or not send ack to the web server and so the web server will decrease
its congestion window size due to ack timeout,
and thus the sending rate is throttled.

In the case of tor exit node, I think If the exit node continues to send
ack to the webserver when its buffer is already full,
the webserver will continue to send packets to the exit node, then what
will happen? I think the exit node will disgard the packet
and never send ack, so in this way, the webserver will reduce its
congestion window and reduce the sending rate


More information about the tor-talk mailing list