RE: spin bit in QUIC

[[e-mail]]

Adding to what Roni has below, first, I applaud to the spin bit design for their time since the Prague meeting and many thanks to Ted for presenting those findings.

>From Ted's presentation, it appears that the design team accomplished what they were asked to do, that is, look at geolocation threat and Min RTT. Slide 7 from Ted's presentation lays out negligible threat and at least not any additional adversarial information.  Given this, I would have thought design team to be less tentative and come to a clear consensus.

Also, agree with Marcus on his enumeration on the need for spin bit and as Roni pointed out inclusion of spin but in v1 gives us some real -world deployment experience on performance and bottlenecks.

I ask the chairs to add spin bit for v1.

-Sanjay


From: QUIC [mailto:[e-mail]] On Behalf Of Roni Even
Sent: Tuesday, November 14, 2017 4:41 PM
To: QUIC WG <[e-mail]>
Subject: [E] spin bit in QUIC

Hi,
It is good that the conclusion about the geo location security issue is that the spin bit does not add privacy risk on top of what can be done without it.

As for the spin bit. RTT calculation are used for trouble shooting in TCP see in draft-even-quic-troubleshooting-video-delivery-00.
The spin bit will provide similar functionality even though it is less reliable as was mentioned since it is not part of the QUIC state machine.
We envision that we will see a lot of QUIC traffic in the network and having the spin bit can help with identifying bottle necks in the network and reducing QUIC latency.  It will be important to have the spin bit in V1 and the main text is already in the github

Thanks
Roni Even