IETF Plenary Session Wednesday, 4 November 2015 Yokohama, Japan Minutes by Cindy Morgan and Karen O'Donoghue 1. Welcome Jari Arkko welcomed the community to the IETF 94 Plenary. 2. Host Presentation Slides: Hiroshi Esaki from the WIDE Project welcomed the IETF to Japan and thanked all of the meeting co-sponsors. 3. Updates on hot topics 3.1. IETF-wide issues Slides: The IESG noted that the process defined in RFC 6761 was used for the approval of RFC 7686, 'The ".onion" Special-Use Domain Name', and that process is not scalable. That process is now under review. Jari Arkko thanked the General Area Review Team (Gen-ART) for their help in improving the quality of RFCs. The Benchmarking Methodology Working Group (BMWG) celebrates its 25th anniversary this week. The IETF 94 Hackathon had between 61-95 participants, working on DNS privacy, HOMENET, tools built by the YANG team, and more. Another hackathon is planned for April 2-3 in Buenos Aires prior to IETF 95. 3.2. Administrative topics Tobias Gondrom noted that a report had been sent to the IETF-Announce email list prior to the plenary , and that more detailed slides are available in the proceedings. IAOC Slides: LACNIC will co-host IETF 95 in Buenos Aires. The Fairmont Montreal has cancelled the contract for IETF 98; the search for a replacement site in North America is underway. The IASA budget for 2016 has been approved; meeting registration fees will remain the same. IETF Trust Slides: Benson Schliesser reported that the IETF Trust and IAOC Legal committee are currently investigating details related to implementation of an IPR policy for IETF Hackathon events. 3.3. IETF 95 in Buenos Aires Carlos Martinez stated that LACNIC is happy to host IETF 95 in Buenos Aires. 3.4. NomCom Update Slides: Harald Alvestrand reported that the NomCom is filling 7 IESG positions and 6 IAB positions. 43 nominees have accepted nominations, and interviews are currently underway. NomCom requests feedback from the community by 12 November 2015. 3.5. IRTF Chair reports on new research groups Slides: Lars Eggert reported that in addition to the 5 Research Groups meeting at IETF 94, there are 4 proposed Research Groups meeting this week. Two Applied Networking Research Prize (ANRP) winners presented their papers during the IRTF Open Meeting earlier in the week: - Xiao Sophia Wang for a systematic study of web page load times under SPDY - Roland van Rijswijk-Deij for a detailed measurement study on a large dataset of DNSSEC-signed domains. 3.6. RFC Series Editor reports on progress in changing the format Slides: Heather Flanagan reported that the format work is progressing. The RFC Editor and the IESG are planning a staged approach to testing the new features that will be available to RFCs. The format requirements drafts will go through a full review process and be published as RFCs. The coding will be an iterative process, and updated documents will be created for all format-related RFCs to capture the final outcomes and decisions. 4. IAB Chair calls for more attention to architectural issues Slides: Andrew Sullivan noted that he sent a more detailed report to the IETF email list to the plenary. The report is also available on the IAB website at . Andrew Sullivan observed that IAB work is intended to complement IETF and IRTF work; the link between the IAB IP Stack Evolution Program and the proposed IRTF How Ossified is the Protocol Stack Research Group (HOPSRG) is one example of this. The IAB is working on draft-iab-protocol-transitions to summarize some basic principles to enable future technology transitions, and also summarize what makes for a good transition plan. The IAB also actively works to help shepherd new work; sometimes people need help getting BOFs together. Andrew Sullivan encouraged the community to reach out to the IAB early on if they need help with BOFs. 5. Technical Topic: Measurement-Driven Protocol Engineering Engineering decisions about protocols to deploy in the Internet should be based on relevant data about the environment they will face. Protocols should be designed for the common occurrences, with an understanding of the risks of uncommon ones. Measurement can provide the insight to know the difference. The IETF 94 technical plenary talk, given by Brian Trammell of ETH Zürich and the IAB, and Alberto Dainotti of CAIDA, explored the challenges of scaling up Internet measurement to provide information useful for protocol engineers, examining this from two different angles: the measurement of Internet "path transparency" (whether certain protocols and features work unimpeded on certain paths through the Internet) as well as the understanding of interdomain topology and BGP dynamics. Brian Trammell, "Measurement-Driven Protocol Engineering," Alberto Dainotti, "Measuring and Monitoring BGP," At the end of the presentations, Andrew Sullivan moderated a question- and-answer session with the audience. * David Somers-Harris asked if there is enough data to demonstrate that the Internet can work over UDP. Brian Trammell replied that UDP may not work on networks where it is blocked or rate-limited by firewalls; operators often block UDP. * Dave Thaler asked, should you collect only what you need to answer questions you already know, or should you collect what is available just in case you might someday have a question that needs it? Keeping privacy considerations in mind, to what extent should you limit what you report? Alberto Dainotti replied that it would be helpful if there were reports on what experiments were done, and how they were conducted; they can always find a use for the data. Brian Trammell added that the preference is to over-instrument and under-report. When designing protocols that won't change very often, then the instrumentation should provide the ability to collect more information so that you can minimize the impairment to future protocols. That said, having the context for the data collected is also important. * Paul Hoffman noted that he is interested in research about how middleboxes might break things, and asked who is doing such research. Brian Trammell replied that this is a topic in the proposed How Ossified Is the Protocol Stack? (HOPSRG) Research Group, and suggested that they talk offline after the plenary. * Daniel Kahn Gillmor asked to push a little harder on the metadata privacy concerns, noting that the data collected is a liability as well as an asset. If you want informed consent, then you won't be able to collect the data because users won't consent. Brian Trammell replied that telemetry.mozilla.org is an example where anonymity is used. Alberto Dainotti added that the privacy concerns are indeed great, and investment should be made there. * Spencer Dawkins observed that it is shocking to realize how little we understand about the network. Measurement-driven protocol design came out of transport work, and other protocols might benefit from it. He noted that the original HOPS work was really trying to answer one question, about whether everything can be run over UDP. He asked if there was any sense of what the scope of the to-be-renamed HOPSRG would be. Brian Trammell replied that it originally came up in the context of transport evolution, but there are other pieces that could benefit, like DNS and DNSSEC. He referred to the recent ANRP talk, "DNSSEC and its Potential for DDoS Attacks: A Comprehensive Measurement Study," as the sort of the thing that might be in the expanded scope. * Dave Oran asked if the measurement community has given much thought to the custodianship of collected data and the analytical tools that are used against that data, as these tools have evolved significantly. He observed that we might be able to strike a better balance by having strict rules for custodianship and using these more advanced tools. Brian Trammell replied that that was a good point; there are legal frameworks that address some of this, but there is more work that could be done. A great deal could be done by preprocessing data so that it becomes less privacy-sensitive but still useful for a more specific purpose. * Bob Briscoe pointed out that measurement is not enough; the protocol has to work with future Internet as well, and that is not measurable. You have to look at the trends for that. * Al Morton asked if there will be more support from the protocol layer for measurement. He added that he thinks that if more users have the ability to put measurement capabilities into intermediate nodes, then that is where we should go. Will UDP help us there? Brian Trammell replied that UDP won't help, but that you can have explicit measurability built into the protocol. * John Dickinson pointed out that if you collect data now and anonymize it, it does not guarantee that it will remain anonymous in the future; tools evolve that compromise that anonymity. Trying to get informed user consent is impossible. Brian Trammell replied that that is a good point. Researchers can take the insight today and throw the data away, but the tradeoff is between the risk to privacy and the utility of the measurement. * Justin Richer observed that in general science, when you make a measurement, you don't want to impact the underlying system; however, that is perhaps not a desirable characteristic in this case. He wondered if something could be built into the protocol to support informed consent. Brian Trammell replied that ping is the only thing explicitly meant to be used for measurement; everything else is taking existing information and exploiting it to measure the thing you want to measure. Getting informed consent is a problem, but so is explicitly understanding that a measurement is being taken. * Randy Bush reiterated Daniel Kahn Gillmor's point: owning or holding a customer's data is a liability. A lot of work has been done on data and privacy for research. We need to leverage that body of work. We are currently recording and saving for posterity data which violates privacy. There are tools out there now that can be used to infer peering relationships which are under NDA. * Peter Lothberg said that we should be looking at what we can do to improve the future Internet. Measurement along will show very little variation. Brian Trammell replied that it is more than just measuring; we have to be explicit about how things are measured and how to apply the measurements that are available. 6. IAB, IAOC, and IESG Open Mic session The IAB, IAOC, and IESG took the stage for a joint open mic session. * Aaron Falk observed that there was only one BOF at IETF 94, and asked if the IETF should be concerned about the lack of new work. Andrew Sullivan replied that the IAB is always willing to help foster new work, and that there were many BOFs at the previous IETF meeting. Jari Arkko added that 19 new working groups have been chartered in 2015. Additionally, Alissa Cooper noted that the ART Area has created several new Working Groups without running a BOF first. Spencer Dawkins added that the IAB and IESG talked about the IETF week in preview, and noted during that discussion that they expect a lot of new work to be ready for BOFs at future meetings. * Shane Kerr noted that there seems to be a great deal of formality around BOFs; he has been involved in an number of pre-BOF BOFs to see if there is enough to discuss, and it makes the process seem rather scary. Jari Arkko replied that it is not supposed to be scary, although sometimes it can be a long process. Andrew Sullivan said that people should email the IAB if they need help with with new work; the IAB is happy to help. Alia Atlas added that the IESG is also there to help; a BOF isn't necessary to charter a new Working Group, but having a coherent idea is. Pat Thaler noted that the Deterministic Networking (DETNET) Working Group met for the first time at IETF 94; much of the participation there has been from people without a lot of experience in the IETF. There was one pre-BOF before the first BOF, and there was a lot of help and guidance from the IAB and the IESG; she was pleased with how smooth the process was. * Justin Richer observed that many so-called "Bar BOFs" are happening in actual meeting rooms, and asked what can be done to move those meetings back to the bar and encourage informal discussions. Lars Eggert agreed with this sentiment, and noted that RFC 6771, "Considerations for Having a Successful 'Bar BOF' Side Meeting," was written to address this issue. Andrew Sullivan added that the community works best when people interested in a topic just get together to brainstorm; there doesn't need to be any formality around that. * Rod Van Meter noted that there were several events in Japan prior to the IETF meeting (e.g., W3C, IMC, RAIM). He asked if having all of those events together was successful, and whether the IETF would consider doing it again. Scott Bradner replied that there were W3C people attending the IETF who would not have been able to otherwise, and there is some indication that the crossover between the communities was helpful, even if the scheduling was accidental. Barry Leiba added that as someone who would not have otherwise attended the W3C meeting, he did find it helpful. Jari Arkko said that he thought that it was useful, but that organizing the schedule that way in the future could be challenging. Jeff Jaffe (CEO of the W3C) noted that in this instance, the IETF had already planned their dates and location, and the W3C decided to organize their meeting in conjunction with the IETF. However, since the W3C only has one meeting a year, and both organizations tend to rotate around the globe, finding another similar opportunity may be difficult. * David Black stated that he was one of the chairs of the ISS BOF at IETF 94, and that he thinks the suggestion to encourage more "bar BOFs" in the bars is good. However, he sees a gap in the process, as there is a perception that a BOF can only be held twice. ISS was a non-Working Group-forming BOF, so there was not a lot of prep work. He would like to encourage more discussion BOFs. Spencer Dawkins observed that this has evolved in the last several years, and that he has seen an increased willingness from the IESG to hold non-Working Group-forming BOFs. * Lee Howard noted that "BOF" stands for "birds of a feather," which just means people wanting to come together to talk about a common topic. He implored those with ideas for new work to come to those in the IESG and IAB if they are not sure if their idea is good, or if they need help articulating that idea. The rules are in place to help the IETF produce good work. * Aaron Falk said that he thought the co-location with IMC worked out well, particularly with the RAIM workshop. He noted that in general the IAB seems to be sponsoring more workshops, which is good, but they are not as good about sharing what happened at the workshop with the larger community. He suggested that perhaps using a non-Working Group session at the meeting to share lessons learned might be helpful. Joe Hildebrand replied that the IAB is working on a more formal response about how information from workshops is shared, and walking though some of the considerations that the Program Committee should take into account. For the recent Managing Radio Networks in an Encrypted World (MaRNEW) workshop, public minutes have been posted to the website already. Ted Hardie added that Natasha Rooney will be reporting out on the workshop during the Security Area Open Meeting. * Aaron Falk noted that there were difficulties with the hotel room block for this meeting, with only a small number of rooms available at the conference hotel. He asked if there were any lessons learned from this experience. Ray Pelletier replied that the objective is to have 600 rooms in the conference hotel for the peak night, but that sometimes that is not possible. Fortunately in this case, the Yokohama Bay Hotel Tokyu was very nearby and also had the IETF network. Tobias Gondrom added that the number of rooms available in the main hotel is always a consideration, but that sometimes there are competing requirements. Ray Pelletier added that the situation for IETF 95 in Buenos Aires will be similar, with a limited number of rooms available at the main meeting meeting hotel. Leslie Daigle observed that if you don't know how far from the venue the overflow hotels will be, that will cause stress. Aaron Falk agreed, saying that not having the backup hotels announced at the same time caused some confusion. The IAOC will take this under advisement for future meeting announcements. * John Dickinson asked if the IAOC takes accessibility requirements into consideration when choosing venues; he noted that as a person who uses a wheelchair, in several cases he has had to wait for hotel staff to unlock elevators, or has found room doors difficult to get through. Tobias Gondom thanked him for that feedback, and asked to talk with him after the plenary so that they can get more detailed requirements for future meetings. * Kevin Fall said that he came across a paper from 2007 in which the IAB said that routing scalability is the most important issue for the Internet, and asked if that was still the opinion of the IAB. Andrew Sullivan replied that he was not on the IAB at the time, but that the notion of a "most important" issue is a dangerous one. Routing issues are important, but so are privacy concerns, internationalization, etc. They are all important problems in different ways. Brian Trammell added that that if one wants to see the current IAB's thinking on what is important, to look at the work going on in the IAB Programs. * David Somers-Harris noted that people who are not native English speakers can struggle in the IETF environment, as people talk very fast. Jabber scribes can help capture the discussion, but it is up the the chairs to find scribes. He asked if the IETF was doing anything else to help in that regard, Mary Barnes replied that the Diversity design team was trying to address some of those issues. Jari Arkko added that the RFC Editor has held workshops on writing I-Ds for non-native English speakers, but those efforts are difficult to scale. Tobias Gondrom stated that the IAOC does not currently have anything to support automatic scribing, but that they are open to hearing ideas. Robert Sparks replied that while tooling might be able to help, people can improve the situation faster: pay attention in the meeting, and if you see someone struggling, bring the conversation back to a level where everyone can participate. David Somers-Harris suggested that live transcription of sessions would be very helpful. Jari Arkko replied that they have tried that with the plenary sessions in the past, but that scaling that for working group sessions would be more expensive and difficult. However, if the community thinks it would be helpful, it can be reconsidered. * Sean Leonard said that when he is choosing hotels, his top priority is getting one where the IETF network is available. He asked if extending the IETF hotel network was part of the venue selection criteria. Jari Arkko replied that is something that is considered extensively. * Leslie Daigle asked to go back to Kevin Fall's earlier point about routing scalability. The IAB at that time had a workshop with data that indicated that routing scalability was the most important thing. However, this is a cautionary tale about where you go with the data you have; sometimes the problems won't be solved in architecture and engineering, but in operations and other ways. * Rod Van Meter offered to chat with anyone who is interested in exploring solutions to the language problem. He noted that several years ago, he had looked at speech-to-text solutions, but the quality wasn't very good. He thought the technology might have improved in the intervening years; he confirmed that being able to see the words being spoken can be a great help. That said, he agreed with Robert Sparks' earlier advice that people can solve the problems technology can't. Jari Arkko agreed that this was good advice to end the plenary on, and suggested that an auto-transcribing experiment may be possible in the future. * Jari Arkko asked the community if having the IAB, IESG, and IAOC on stage for the open mic at the same time was useful; the community hum indicated that they found it useful. Jari Arkko asked the community if the overall running time for this plenary worked. A couple of people indicated that the slides presented could be shorter, but when a hum was taken, the result was inconclusive. 7. Close of meeting