Day 2 / Common Track / Talk 1 Talk – Keynote: Open Source, the IETF, and You Speakers: Alissa Cooper Report by: Michael Kehoe and Evangelos Haleplidis.
Alissa Cooper, the chair of the IETF, was the keynote speaker. Alissa set out to talk about the relationship between open source and the IETF. She argued, in her keynote, that these are mutually supportive and that the Linux developer community and the IETF should work more closely together.
Alissa begun by discussing how the IETF works and the standard document that it produces, namely the RFC series. Then she mentioned the focus of her talk, creating a standard and an implementation in parallel. To illustrate her talk, she mentioned two prominent examples in the IETF, TLS 1.3 and QUIC.
In regards to TLS, Alissa mentioned that support and implementation for TLS 1.2 took 5 years. However in the case of TLS 1.3, which was finalized last August, which solves issues from the cryptographic algorithms and makes more improvements, chrome and firefox already support 11% of the traffic and 50% of facebook traffic is using TLS 1.3. The reason for this fast support was the parallel implementation. There was a real-time feedback loop between the standardization process and the implementation. One of the reasons this was feasible, is that the IETF has a flexible model. While standards take a lot of time to be completed (TLS 1.3 took 4 years) the total effort is collaborative and provides a better outcome.
Alissa then switched to QUIC. QUIC is an ongoing effort, developed initially at Google and after two years it was brought to the IETF. The IETF begun work in QUIC in 2016, with the first version probably been released by the end of this year. Since 2016 the wire protocol has been changed drastically. The open public standardization would eventually provide a more used implementation. Already there are two dozen QUIC implementations, with two being open sourced.
One topic that Alissa mentioned was that of protocol ossification. When some parts of a protocol changes, it usually can't get through middleboxes. QUIC and TLS were ambitious. They knew that middleboxes couldn't change and so, both took the network reality and made their work around them.
TLS 1.3 saw middleboxes creating issues and therefore masqueraded it as TLS 1.2. The other way is to use encryption, which was the approach QUIC used. QUIC has everything encrypted over UDP.
Alissa then enumerated a couple of points when implementing a new standard: 1. Design for deployed reality, don't expect the network to change. 2. Use encryption. 3. Document well in advance protocol invariances. Decide which part of the protocol won't change. This approach will create more flexibility for the future.
The talk shifted towards how the IETF works. IETF is around more than 30 years. The IETF meets three times a year and starts with a two day hackathon over the weekend which is free to attend, with the focus being towards collaboration rather than competition. Then the following five days are filled with meeting over a number of parallel tracks. The IETF is organized in 6 different areas, Internet, Routing, Ops & Management, Transport, Apps & Real Time and Security. There is also the IRTF, which are research groups, focused on research and experimentation.
To participate is really easy and open to all. You can use the datatracker, mailing list, use collaboration tools or contact the chairs. Before proposing something for the first time, it is advisable to talk first to people, to get the lay of the land. The way to participate is by writing internet drafts. There are two kinds of drafts, individual and working group drafts. Also there are three different types of RFCs, proposed standards, informational and experimental.
IETF supports a number of tools, open source projects, visualization and validation for writing documents.
Site: https://www.netdevconf.info/0x13/session.html?keynote-cooper