Client Hello and SNI

[christopher brown]

Hi Folks

Is it safe to assume there will always be a Client Hello and SNI at the start of each new UDP/ QUIC flow?

Thanks

Chris

[christopher brown]

Also is it safe to assume that on a new UDP connection to a server ( new destination ip ) that the QUIC packet number will start at 1. Is it possible it can be non 1?

Thanks Chris

[Ian Swett]

The first version of QUIC will use TLS 1.3 and TLS 1.3 has an SNI, so it's certainly subject to change in future versions, but I believe version 1 will always have a Client Hello and SNI.

It is definitely not safe to assume the QUIC packet number will start at 1.  Even though GQUIC currently starts at 1, the IETF draft specifies randomizing the first packet number.  Even in GQUIC today, sometimes the client hello is lost and retransmitted with a larger packet number.

--
You received this message because you are subscribed to the Google Groups "QUIC Prototype Protocol Discussion group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to proto-quic+unsubscribe@chromium.org.
To post to this group, send email to proto...@chromium.org.
For more options, visit https://groups.google.com/a/chromium.org/d/optout.

[christopher brown]

Ian thanks for the quick response, this makes sense.

Chris

On Monday, 11 December 2017 14:07:51 UTC, Ian Swett wrote:

The first version of QUIC will use TLS 1.3 and TLS 1.3 has an SNI, so it's certainly subject to change in future versions, but I believe version 1 will always have a Client Hello and SNI.

It is definitely not safe to assume the QUIC packet number will start at 1.  Even though GQUIC currently starts at 1, the IETF draft specifies randomizing the first packet number.  Even in GQUIC today, sometimes the client hello is lost and retransmitted with a larger packet number.

On Mon, Dec 11, 2017 at 6:22 AM, christopher brown <chris...@gmail.com> wrote:

Also is it safe to assume that on a new UDP connection to a server ( new destination ip ) that the QUIC packet number will start at 1. Is it possible it can be non 1?

Thanks Chris

On Monday, 11 December 2017 10:35:24 UTC, christopher brown wrote:

Hi Folks

Is it safe to assume there will always be a Client Hello and SNI at the start of each new UDP/ QUIC flow?

Thanks

Chris

--
You received this message because you are subscribed to the Google Groups "QUIC Prototype Protocol Discussion group" group.

To unsubscribe from this group and stop receiving emails from it, send an email to proto-quic+...@chromium.org.

[christopher brown]

Hi Ian

Is there an expected timeline when youtube or google will move from GQUIC to the IETF spec? 

Thanks
Chris

[Jana Iyengar]

Hi Chris,

We're waiting for the IETF spec to stabilize a bit more before moving to it.

- jana

To unsubscribe from this group and stop receiving emails from it, send an email to proto-quic+unsubscribe@chromium.org.

[christopher brown]

Ok sure

Thanks for quic reply 

;)

[christopher brown]

Hi Jana/Ian

Any word on moving over to IETF/TLS 1.3 ?

Thanks

C

[Ian Swett]

QUIC version 44 is now supported by Google servers and uses the IETF QUIC invariants(minus one bug Dmitri noticed and we fixed), so the public header will start looking like IETF QUIC very soon.

QUIC version 99 is a placeholder for other IETF work, is fairly complete in terms of framing, but we've only started implementing the new TLS mapping that came out of the Stream0 design team process and was agreed upon at the Kista interim.

[christopher brown]

Thanks again Ian! Sounds good

[christopher brown]

Hi Ian/Jan

Any update on the move to TLS 1.3 ? By move, I imply enabled out of the box.

Also is there is variant client/server which I could use to validate the existing implementation?

Thanks for the good work

Chris

[Ian Swett]

Version 46, which uses an the IETF invariant header, but still uses QUIC Crypto is now default enabled.

Some more documentation on the gQUIC v46 header is here: https://docs.google.com/document/d/1FcpCJGTDEMblAs-Bm5TYuqhHyUqeWpqrItw2vkMFsdY/edit?usp=sharing

There's not yet a production version that uses TLS 1.3, but the code is implemented and has been used for IETF interop events.

--

You received this message because you are subscribed to the Google Groups "QUIC Prototype Protocol Discussion group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to proto-quic+...@chromium.org.

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/proto-quic/d9f8a8b9-61e6-407d-a49d-bbf4caa79f9e%40chromium.org.

[christopher brown]

Thanks again Ian

Chris