News

Meet you on the NetStream: IETF with all-virtual agenda in desperate times

2020-03-23 Blog

CENTR is trying a new format for reporting on the IETF meetings, and will be publishing a series of blog posts over the next couple of weeks, before publishing a full report at the beginning of April. The first blogpost introduces some of the key topics that will be discussed this week.

By Monika Ermert, eLance Journalist - The 107th meeting of the Internet Engineering Task Force (IETF) has a staggering 1095 registered remote participants after the Internet Engineering Steering Group (IESG) cancelled in-person-meetings due to the Corona pandemic. The Vancouver meeting was cancelled at short notice – with only ten days to go until the meeting was supposed to start. And while initial reactions to the decision led to accusations of fear-mongering, later lockdowns in European countries and around the world emphasises the severity of the situation.

With the immigration blocks for non-nationals to first the US and then Canada, the virtual meeting now set up by the IESG looks like a very reasonable step. It will, additionally, be the largest technical standardization meeting to be organized online.

The IETF is not the only standardization body that has decided to go all-virtual. Similar announcements were made by the IEEE, 3GPP and ISO, leading all the way into May 2020. The ITU postponed a number of meetings that have not yet had the experience of online collaboration.

More new work

Two more slots were given to the so called ‘dispatch’ meetings, one for general matters and the other for security. The dispatch WGs will again allow for new work to be presented and for discussions on where in the IETF these respective topics should be worked on. Even the remaining sessions, added to the Birds of Feather (BoF) and the first remote-only IETF plenary on Wednesday night, cover brand new work. All three, Reliable and Available Wireless (RAW), Drone Remote ID Protocol (DRIP) and Adaptive DNS Discovery (ADD) were only established this year. ADD can again be expected to be controversial (see section below).

A new, virtual agenda

The new agenda is much reduced, with IESG leadership choosing to prioritise new workstreams. Five of the 12 announced sessions are BoF meetings.

Today, the Transactional Authorization and Delegation (TXAuth) BoF meets over its proposed “fine-grained delegation protocol for authorization, identity, and API access”, which is intended to extend on the OAuth and OpenID concepts. TXAuth is already discussing a charter (agenda is here), and this will certainly be a project to keep track of for CENTR’s mojeID proponents.

On Tuesday a BoF (agenda is here) will consider ideas for a Multiplexed Application Substrate over Quic Encryption (MASQUE), with one document from David Schinazi. In his draft, Schinazi describes MASQUE as “a framework that allows concurrently running multiple networking applications inside an HTTP/3 connection”.

Wednesday’s BoF is the Web Packaging planned WG (WPack) whose charter is nearly ready and is currently being reviewed by the World Wide Web Consortium (W3C). It tackles “a way for the publisher to authenticate these resources such that a user agent can trust that they came from their claimed web origins”. The option to use web resources offline looks like another good tool in times when the network is stretched due to everybody working online (use cases for WPack are here) – and ties in nicely with efforts by the IAB in 2019 to cast a light on interactions between the IETF, W3C and Google’s increasingly popular AMP-project.

On Thursday there are two BoFs. Realtime Internet Peering for Telephony (RIPT) wants to update SIP and teleconferencing. To make it compliant with WebRTC, a special extension is already in the pipelines (agenda and drafts are here). Privacy Pass (agenda here) is considering yet another “mechanism for providing privacy-preserving attestation of a previous successful authorization between a human and a server”.

Is virtual the future?

The idea of scheduling this new work was to give a channel for the new groups to form, while long-standing WGs like the DNSOP are expected to self-organize more easily. In any case, a number of the existing WGs have more or less regularly relied on virtual WG meetings in between the regular IETF meetings anyway.

CBOR for example has had regular, at times fortnightly, virtual interims since 2018. The tough schedule bore some fruits according to document author Carsten Borman, such as core IETF RFCs like 7049bis and RFC 8610. With the physical meeting now cancelled, CBOR, according to Borman, will just go back to the heavier virtual meeting schedule, meeting every other week, starting on 8 April (see the growing list of upcoming interims).

It also begs the question: could the IETF go much more virtual in the future, as there have been discussions and even a draft document proposing to reduce physical meetings for ‘green’ reasons?

Yes and no, say engineers. The Quic WG is working on the new transport protocol, which many think will hugely cut into TCP traffic, and meets in person not only three, but six times a year. While the WG is also using virtual interim meetings, face-to-face interactions between those working on such core internet protocols is indispensable, said Quic Chair Lars Eggert. An all-virtual IETF in the future is impossible, he said.

Still, the IETF might have to consider empowering its participants with more robust virtual meeting spaces, as so many other organizations have done. Despite having streamed sessions for many years and using conferencing tools like Meetecho to allow better access for those attending remotely, the IESG decided to keep the agenda for a full-virtual meeting light and one track only. To keep the agenda slender, the Internet Architecture Board met last week already.

The IESG decided to use Cisco’s WebEx conferencing toolset instead of Meetecho for the virtual meeting. Testing was recommended to the 1095 registered remote attendees to make sure the system worked well.

Looking ahead in these Corona-stricken times, several experienced networkers like John Klensin have now recommended participants to be prepared for also virtualising IETF108, originally planned to be in Madrid in July.

ADD: (trying to) work towards consensus

As engineers prepare for a fully remote IETF, some might look forward to a remote ‘popcorn session’. Adaptive DNS Discovery (ADD) has just been chartered after agonizing debates between the ISP and the web fractions. Both sides had seemed to be locked in the fight about who should be in charge of making decisions about whom, and how resolvers should be selected as the first DNS over HTTPS (DoH) implementations shifted control to the platforms.

According to the slashed-out charter of the now officially-established WG, it has to focus on “(1) resolver discovery; (2) means to convey information to the client”. What the client does with the information received, including how he interacts with the user is out of ADD’s scope, ADD Co-Chair Glenn Deen reminded the WG just before the IETF meeting, when yet another round of discussion sparked.

Vittorio Bertola, Engineer at Open-Exchange and one of the engineers in lockdown in Corona-stricken Italy, finds the climate of the DoH/ADD group “much more positive at the moment”, with the technical solutions being focused. One real issue that possibly remains, according to Bertola, is the different conception of ‘discovery’. Bertola ponders whether “the DNS/ISP people use ‘discovery’ to mean ‘a way to ask the local network for which DoH resolver to use and force all applications to use it’, while the Web people say ‘discovery’ means ‘a way to spread DoH traffic automatically among many resolvers’” (though in the end these resolvers would mostly be run by the usual handful of big Internet platforms).

The different drafts that are now on the table speak of this difference in conception of ‘discovery’ and suggest that another intense debate between “those who want discovery through DNS queries, and those who want discovery through HTTPS requests” is highly likely (Bertola).

Four new drafts have been dropped on the ADD table, and the original mechanisms for discovery and privacy-preserving selection proposed by Apple’s Tommy Pauly last autumn have not even been counted. In the new documents, ISP/Network oriented approaches are put forward, and the use scenarios with managed and unmanaged CPEs are meticulously dissected.

How sensible the talk about potential choices to make are was illustrated by the backlash to one new document, put on the agenda of next Tuesday’s meeting by the ADD Chairs. The document “Selecting Resolvers from a Set of Distributed DNS Resolvers” tries to analyse options and potential consequences of resolver selection. The authors, former IETF Chair Jari Akko, IAB Chair Ted Hardie and IAB member Martin Thomson – who in some way constitute the operator and the platform view as being from Ericsson, Google and Mozilla – underline that their “memo aims to discuss the concept and analyse its impacts before dwelling into the technical arrangements for configuring and using this particular approach”. But some on the ADD mailing list criticised this, saying the debate was falling back into policy instead of technical questions.

More drafts for ways to solve the issue will certainly come, but what has to be avoided in the end, Bertola underlined, was a split into two different mechanisms (one DNS and one HTTP originating discovery) forcing “everyone to either implement both or be incompatible with some of the others, creating some fragmentation”.

Does it all matter?

As public life comes to a close in the countries of many IETF participants, one could wonder how much ongoing standardisation work matters at this moment. However, it is safe to say that some of that work is particularly relevant right now as the world moves to streaming. It certainly is engineering time.