IETF Technical Plenary Session Monday, 21 July 2014 Toronto, Canada Minutes by Cindy Morgan, IETF Secretariat 1. Welcome Russ Housley welcomed the community to the IETF 90 Technical Plenary. 2. Reporting 2.1. IAB Chair Russ Housley delivered the IAB Chair report: http://www.ietf.org/proceedings/90/slides/slides-90-iab-techplenary-11.pdf Russ Housley noted that since IETF 89, the IAB has: - commented to NIST encouraging transparency and openness for all of their publications - commented to ICANN on principles, mechanisms, and process to develop a proposal to transition NTIA's stewardship of the IANA functions - appointed Sarah Banks and Robert Sparks to RSOC - appointed Sean Turner to ISOC Board of Trustees - appointed Matt Miller as liaison manager to ECMA TC39 - appointed Russ Housley and Lynn St.Amour to the Coordination Group for the transition of the IANA stewardship - published RFC 7288: Reflections on Host Firewalls - published RFC 7295: Report from the IAB/IRTF Workshop on Congestion Control for Interactive Real-Time Communication - published RFC 7305: Report from the IAB Workshop on Internet Technology Adoption and Transition (ITAT) - published RFC 7241: The IEEE 802/IETF Relationship Russ Housley noted that the IAB is currently trying to fill a seat on the 2015 ICANN Nominating Committee, and has extended the call for volunteers through Friday, 25 July 2014. The IAB is currently seeking feedback on the Independent Submissions Editor in order to determine whether to renew the current contract. Russ Housley reported that the 2014 IAB Retreat was held in conjunction with LACNIC. During the retreat, the IAB restructured the current Program framework to allow the IAB to focus on the topics most relevant to the Internet today. One major change was the merger of the existing Privacy Program and Security Program into the Privacy and Security Program. The IAB will focus a significant amount of effort on this Program in 2014. 2.2. IRTF Chair Lars Eggert delivered the IRTF Chair report: http://www.ietf.org/proceedings/90/slides/slides-90-iab-techplenary-2.pdf Lars Eggert reported that four Research Groups will meet at IETF 90, as well as two proposed Research Groups. The IRTF Open Meeting will be held on Tuesday, 22 July 2014 at 1420. Lars Eggert reported that two Research Groups (Routing Research Group and Network Complexity Research Group) have closed since IETF 89. Three IRTF-stream RFCs have been published since IETF 89: - RFC 7122 on Datagram Convergence Layers for the Delay- and Disruption-Tolerant Networking (DTN) Bundle Protocol and Licklider Transmission Protocol (LTP) - RFC 7242 on Delay-Tolerant Networking TCP Convergence-Layer Protocol - RFC 7253 on The OCB Authenticated-Encryption Algorithm The Applied Network Research Prize is awarded jointly by the IRTF and ISOC. The committee received 46 nominations; six prizes were awarded for 2014. The first two prize winners presented at IETF 89; Robert Lychev will present his paper during the IRTF Open Meeting this week. Lars Eggert noted that the nomination period for the 2015 award cycle is open now, and encouraged people to submit papers at https://irtf.org/anrp/2015/. 2.3. RSE and RSOC Chair Heather Flanagan delivered the RSE report: http://www.ietf.org/proceedings/90/slides/slides-90-iab-techplenary-5.pdf Heather Flanagan reported that the RFC Production Center has seen an unusually high number of documents submitted to them so far this year, resulting in the RPC not meeting the SLA of getting 67% of documents in an RFC Editor-controlled state published in six weeks or less. The largest factor was that several abnormally-long documents were sent in clusters, and clusters are more challenging because the cross- references and terminologies must be checked. David Black, chair of the STORM Working Group, noted that the STORM iSCSI documents were responsible for a large portion of the clusters, and also noted that on top of everything else, the STORM WG requested expedited publication of those RFCs in order to coordinate with other standards bodies. He thanked the RFC Editor organization for their help in publishing those documents. In addition to publishing documents, the RFC Editor has made progress on the RFC format change. The Tools Development Team is working on the SoWs required for the new tools, and the Design Team continues to work on the requirements and descriptions. The RFC Style Guide is currently in AUTH48 as RFC-to-be 7322. The draft describing how Digital Object Identifiers will be handled is still in progress as draft-iab-doi. 3. Technical Topic: Network topology and geography Andrew Sullivan introduced the speakers for the evening's technical topic on network topology and geography. Antonio Gamba-Bari is a Ph.D. student at the University of Toronto, working with Professor Andrew Clement on the IXmaps Project, mapping Internet routing and surveillance from a user privacy point of view. Jane Coffin is the Director of Development Strategy at the Internet Society, coordinating collaborative strategies for expanding Internet infrastructure and access among emerging economies. Amogh Dhamdhere is a research scientist at the Center for Applied Internet Data Analysis (CAIDA), doing research in Internet measurement, Internet topology, and Internet economics. All three technical presentations are available at: http://www.ietf.org/proceedings/90/slides/slides-90-iab-techplenary-9.pdf At the end of the presentations, Andrew Sullivan moderated a panel discussion that included questions from the audience. * Andrew Sullivan asked whether, given the evidence on boomerang routing, people would spend money to keep traffic local. Jane Coffin replied that in certain countries, they cannot afford the infrastructure to do so. In those cases, the technical community must be built up, so that they can work with the governments to help them understand the importance of the IX. Jane noted that there are not a lot of companies who are willing to go in on their own to do this, which is why it is important that the Internet technical community, the RIRs, NSRC and others all work together on this. Andrew Sullivan asked why some Canadian traffic between two buildings in Toronto might be sent to New York or Chicago first, when that is clearly not the shortest path, and whether there is a network economic effect at work. Antonio Gamba-Bari replied the IXmaps Project has discovered from their traceroutes that some networks will decide to peer with certain networks, but not others. That raises questions not only of efficiency, but whether the data crossing national borders is subject to other inspections. Andrew Sullivan noted that the routing will not necessarily protect data from inspection, because surveillance happens within Canadian borders as well. He asked whether it makes the network worse to send local traffic outside borders, and how big the cost is. Antonio Gamba-Bari replied that there are deals between big ISPs and small ISPs, so presumably the costs have to do with who owns the networks and what sort of deals are in place between the exchange points. Amogh Dhamdhere added that longitudinal measurements play a role in trying to tease apart some of these issues. There is data about the existing business relationships, but those change over time and can be tracked historically to correlate them with performance data. He noted that CAIDA has a lot of traceroute data, as well as the AS relationships and the historical AS-level topology data. * Jacques Latour of CIRA (one of the sponsors of the IXmaps Project) noted that CIRA has been doing research in Canada for the past three years, and one of the changes he has seen with putting in a new IXP in Winnipeg is that first, you must build a community to put the IXP together. When the cost of bandwidth went down, the incumbents did not want to peer with the local IXPs because they wanted to keep the traffic and the money involved. So in Canada, they are building a "donut architecture" around the incumbent; the challenge is getting them to peer with the IXP. Andrew Sullivan replied that Canada is a big place, although there are not very many people in it. He asked whether the sparsely-populated areas are getting these new IXs. Jacques Latour replied that the idea was to go coast-to-coast, and that the key thing is to work to build a community so that the IXs can be sustained and operational on their own. Andrew Sullivan noted that this is similar to Jane Coffin's work, except that she is dealing with densely-populated areas where there are other incentives that are perhaps not working. For those who are concerned about the IXP being a point of surveillance, the monitoring is already happening. Andrew asked whether we can measure the positive benefits in those localities. Jane Coffin replied that ISOC is working hard to collect more data; the information currently available is linked from her presentation. She noted that it is not easy if the IXs do not already have the software installed. Once the servers are in, they have to work with the locals to understand that it is not a surveillance system, and encourage them to publish their statistics on daily, weekly, monthly, and annual traffic. She noted that it is very difficult to grow these communities of interest; it can take years. The work must be done at a pace that does not look like anything is being forced or has surveillance implications, and it takes a lot of training workshops. * Tim Shepherd asked whether IXmaps has looked into finding a way to show the actual geographic path between routers, noting that just because two routers happen to be in the same country, it does not necessarily mean that the shortest path was used. Antonio Gamba-Bari replied that yes, looking at the lines from Google Earth can be misleading, but that there are tremendous challenges in terms of geolocation and geoprecision. Much of their data relies on crowd-sourcing; they have information at a city level, but not necessarily at a building level. He noted that his colleagues are working to improve that particular aspect of geolocation. * Ruediger Volk asked how much of the full picture was being presented for the more developed parts of the Internet. He said that he was not sure whether it was important what particular side of Toronto a certain router was on. He asked whether the IP routing trace actually shows how the traffic flows; how much of the traffic is directed by users who signed up with an email service across the globe? Ruediger asked how much is controlled at the IP routing layer, and how much is happening on the application side. Amogh Dhamdhere replied that one way to look at that sort of question is to do traceroutes of the kind of content that you care about, and to identify it by looking at the traffic passing through CDNs. Jane Coffin added that in developing countries, there is often not an option for consistent, high-confidence service, so users use external services. The local technical community must be built before more services can be hosted locally. She suggested that one could look at the traffic levels before and after an IX is installed to see the differences in traffic levels, quality of service, and latency. Antonio Gamba-Bari noted that the IXmaps traceroute generation application is crowd-sourced, and runs on a specific port on demand by the user. For the next level of the traceroute generation, IXmaps is looking for an application that runs in the background and monitors all the time. * Carlos Martinez commented that he has an uneasy feeling about the use of the term "efficiency," because "efficiency" is generally tied to a metric or indicator, but making things efficient in terms of delay can lead to hidden costs in terms of red tape. The perceived inefficiencies in how traffic is routed have underlying causes that go beyond the technology, and those underlying causes should be addressed as well, because if the boundary conditions are changed, the system can fix itself. * Joel Jaeggli observed that there are geopolitical and economic considerations that have created the current pathways, and that those pathways are used for more than Internet transit--trucks and ships and cars and airplanes also travel along lines to places with strong economic ties. Andrew Sullivan replied that it would be interesting to take the existing data, and to track new IXs going in and see how the patterns are emerging, and whether they follow the existing lines. * An unidentified member of the IETF community (1) said that one of the more interesting metrics presented was the cost per unit balance, and he wondered whether any effort was underway to map this to geographies and use this as a way to focus attention on particular problem areas. Jane Coffin replied that one of her colleagues who is an economist for the Internet Society has done some mapping and corollary work, as suggested. They have some mapping of the IXs, but are also looking at corollary cost considerations and places where latency has come down so that they can analyze it in greater detail. She noted that some of the reasons are regulatory and policy-based, but that some have to do with having a strong technical community. * An unidentified member of the IETF community (2) noted that some peering points have economic advantages because carriers already have arrangements with each other at certain points to freely exchange data, so it makes financial sense for those carriers to move data through those existing points regardless of the actual distance between the final destinations. He asked whether such peering relationships will be encouraged when trying to create more IXPs. Jane Coffin replied that some of the more advanced networks run by incumbents do not necessarily peer at the IXP. When the incumbent does not peer at the IX, smaller, less-sophisticated networks may level up from a technical perspective as they gain confidence in their own infrastructure; that is a network effect from peering at the IX. * Robert Kisteleki reported that RIPE NCC, along with ISOC, has started collecting data before, during, and after IXP deployment. They have deployment within Africa, and are starting local measurements to figure out the best spots for new IXPs, and how things change when there is an IXP locally. * Andrew Sullivan thanked the panelists for their time. 4. IAB Open Mic The IAB took the stage for the open microphone session: - Jari Arkko, IETF Chair - Mary Barnes - Marc Blanchet - Joel Halpern - Ted Hardie - Joe Hildebrand - Russ Housley, IAB Chair - Eliot Lear - Xing Li - Erik Nordmark - Andrew Sullivan - Dave Thaler - Brian Trammell Lars Eggert (IRTF Chair) and Heather Flanagan (RFC Series Editor) joined the IAB on stage. There were no questions from the community, and the IETF 90 Technical Plenary was adjourned.