This APAC Space web conference was an information sharing session on the Internet’s Root Server System (RSS). RSS Advisory Committee (RSSAC) Caucus Member Anupam Agrawal provided a historical introduction to the RSS and the RSSAC. This was followed by RSS Governance Working Group (GWG) Representative Hiro Hotta, who shared an overview of the discussion on RSS governance.

Edmon Chung (DotAsia) facilitated the open community sharing/discussion. As the discussion could only touch the surface of the topic, the ICANN APAC office will explore further RSS sessions or discussions.

For AOB, Amrita Choudhury (NomCom Delegate) called on the APAC community to apply for the NomCom's available leadership positions, and reminded that submissions must be made by the extended deadline of 29 Mar 2021.


Details of Session

History of Root Servers

Anupam explained that root servers are the authoritative servers at the highest level of the Domain Name System (DNS) which resolve domain names into IP addresses. He then offered a historical timeline to the RSS’ development:

1983: With an increasing number of files and data becoming too many to manage, Jon Postel and Paul Mockapetris published Request for Comments (RFC) 882. Though made obsolete by later RFCs, it first proposed a system — the DNS — to resolve domain names into IP addresses.

1984: The first root server was installed at the Information Sciences Institute in the University of Southern California to test DNS software and further develop the DNS.

1995: Growing to nine root servers, these were renamed following the alphabet letters A to I, such as in A.root-servers.net. Anupam believed that the collective naming at this point in time meant that the root servers had become a root server system.

1997: The number of root servers grew to 13, the same number as today. Of these, only three are managed outside North America – two in Europe and one in APAC:

  1. I-root server — Managed by NORDUnet in Sweden, this was the first non-US root server (NIC.NORDU.NET) added on 28 July 1991.
  2. K-root server — Managed by Réseaux IP Européens Network Coordination Centre (RIPE NCC; the Regional Internet Registry (RIR) for Europe), and hosted in London in May 1997.
  3. M-root server — Managed by the Widely integrated Distributed Environment (WIDE) Project in Japan in August 1997.

1999: L-root server (now known as the ICANN Managed Root Server or IMRS) was transferred to ICANN after Jon Postel’s unfortunate death, a move he had agreed upon earlier. Anupam added that from experience it was not possible to have a root server discussion without mentioning Jon as the system Jon first put into place was still running today. Perspectives on root servers were therefore also extremely important to view in this historical light.

2000: Verisign, Inc., the Root Server Operator (RSO) of A-root server, acquired J-root server to make the number of RSOs 12 (the same number as today). Other major changes since then included the B-root server changing its operating IP addresses, and the E-root server enabling IPv6.

Anupam noted that despite having the largest Internet usage and growth, the Asia Pacific (APAC) region hosted only about 25% of the more than 1,300 root server instances globally (see https://root-servers.org/ for updates). Anupam felt that conscientious effort was needed to increase the number of root server instances in APAC or the number of root servers themselves. For further reference, Anupam shared dns-stats.org to be an excellent source for root server data and DNS traffic (and for IMRS data: stats.dns.icann.org).

RSSAC Formation and Role

The RSSAC was formed as part of ICANN’s creation and bylaws in 1998, with the objective to advise the ICANN community and ICANN Board on the operation, administration, security, and integrity of the RSS. Consisting of the 12 RSOs, its ongoing discussions included the management model of the root servers, which led to the development of the ground-breaking document RSSAC037. Final details and how the RSSAC037 management model would work are currently being discussed by the RSS GWG (more below).

Looking back at the RSSAC’s first meeting in 1999, Anupam noted that an action item had remained open since – the development of operational procedures for the RSS including the formalisation of contractual relationships, and something the RSS GWG had to resolve.

Anupam noted that in addition to its unique position being a RSO and a RSSAC member, ICANN provides oversight and policy development for the Internet’s system of unique identifiers, including the DNS. It also coordinates root zone management through its contractor, Verisign, Inc. Hence, to perform such functions effectively, it is essential that a proper structure is included within RSSAC037 to ensure legal enforceability.

Anupam felt that the creation of the RSSAC Caucus itself was a good start in opening up discussions with the wider community, but more opportunities for RSOs to come together in coordinated efforts, especially for APAC, are still expected. He also shared that in India, more root server instances needed to be deployed especially at local Internet Exchange Points (IXPs), which were increasing, and to this end he called on ICANN and the community for support where possible.

RSS Governance

Hiro shared that while the main business for JPRS (Japan Registry Services Co., Ltd.) is being the ccTLD manager for .jp, it has been jointly operating the M-root server since 2005 with the WIDE Project. JPRS decided to become an RSO because many of its business functions – such as being a registry, registry service provider, registrar, and server certificates issuer – are crucial parts relying on root servers and their importance in resolving domain names. In collaboration with APNIC (Asia Pacific Network Information Centre –  APAC’s RIR), JPRS also supports the deployment of M-root instances throughout APAC. He noted that the number of root server instances was growing fast in the world, standing at a total of 1,370 on the day of the session. Hiro also noted that the RSS operated well and did not have any visible problems so far.

Hiro shared that what led the RSSAC to develop RSSAC037 on the future governance of the RSS were three potential issues identified:

  1. Lack of a mechanism for assuring the technical ability and service quality of the RSS and RSOs;
  2. RSO financial sustainability especially when faced with growing traffic and security threats. Currently, each RSO covers all the cost for its own managed root DNS system and operations; and
  3. Lack of a clear and standard governance structure for the entire RSS, leading to questions of service continuity should a RSO become unable to function.

RSS GWG

The RSS GWG was formed in November 2019, serving as the core of the community-driven process to discuss and develop a final RSS cooperation and governance model based on RSSAC037 by early 2022. RSS GWG’s representatives from APAC were Geoff Huston representing the Internet Architecture Board (IAB), Hanyu Yang for gTLD registries, and Hiro from the RSOs. 

Hiro highlighted the GWG would only discuss and propose the RSS governance model, not the substance of requirements or performance assessments for the RSS and RSOs, such as technical features or an ideal number of RSOs. These issues would be handed over to newly structured organizations proposed by the GWG and community input. Open for public viewing, the GWG’s working sessions could be accessed from the RSS GWG page.

Proposed RSS Governance Model

The RSS governance model aimed to serve three main stakeholders groups — the ICANN community, the Internet Engineering Task Force (IETF)/IAB, and the RSOs. The model has five primary functions:

  1. Strategy, Architecture and Policy Function (SAPF) — Plans for future structure and technology of the RSS, and defines Service Level Agreements (SLAs). This function might discuss topics such as the ideal number of RSOs, and whether the RSS should adopt new technologies.
  2. Performance, Monitoring and Measurements Function (PMMF) — Measures and monitors each RSO’s SLAs.
  3. Secretariat Function (SF) — The communication channel between the RSOs and the Internet community.
  4. Designation and Removal Function (DRF) — Discusses procedures for designating new RSOs and/or removing underperforming RSOs.
  5. Financial Function (FF) — Ensures sustainable financial provision to all RSOs, including reserving emergency funds and Research & Development (R&D) funds for new root DNS technologies. The FF would allow RSOs to request funding support to maintain their service and security level.

After discussing the governance model for a year, the GWG’s current draft was to propose a Public Root Services (PRS) entity as the RSS governance body, which could be an affiliate of ICANN like Public Technical Identifiers (PTI). With the PMMF, SF, and FF functions residing within it, the PRS would coordinate across all entities providing root zone resolution services, carry out SAPF-developed policies, arrange funding for RSOs, and enter into contractual agreements with them. The SAPF, however, would be a community-based group separate from the PRS, managing the strategy and architecture of the RSS. A Multistakeholder Review Panel that would look after the DRF would be formed on request by the PRS Board.


Community Discussion

Edmon observed that ICANN had matured over the last 20 years with much development in TLD management, and it was good to see the governance of the RSS now being formalized and maturing with it as well. Hanyu Yang (Beijing Tele-info Network Technology Co., Ltd.) shared that as the youngest member of the RSS GWG, being in it was a great opportunity to know more technical information about the DNS and RSS while also being able to interact with their technical experts. The GWG’s membership from across various stakeholder groups was broader than the RSSAC’s which was welcomed, but challenges remained in bringing greater awareness of the RSS’s importance to a wider, general audience. She hoped that with increased communication efforts, more from the community would become involved in RSS-related areas such as DNS resolution performance improvements.

Hiro added to Hanyu’s points that the GWG was only focusing on the structure of the RSS’ governance, and could be why it might not seem attractive to many in the larger community. However, he felt that the discussions which would take place after a structure was created would be interesting for the entire community. Edmon commented that the processes of processes were just as important in Internet governance.

In terms of DNS resolution performance, Baojun Liu (Tsinghua University), who joined the RSSAC Caucus in November 2020, shared a root server instance performance study in China that measured the network latency between Chinese Internet users and their queries with DNS root servers/instances. The study found that the geographic location of ISPs had a significant impact on the latency rate. For example, the installation of an F-root instance at China Telecom reduced the latency rate for its own users but had little impact on improving the latency for China Unicom users. Hence, latency optimization was unbalanced and further performance monitoring and assessment was needed (Baojun’s slides are included in the “Appendix” of the session’s presentation deck).

Che-Hoo Cheng (APNIC) asked what an ideal percentage would be for root server instances in APAC. Anupam responded that there could be different models to look at it, and the India Internet Foundation is working on one which considers the number of Internet users served per root server instance. Ideally, the resulting average should be similar across all regions, and based on a quick estimate the target could be approximately 40-45% for APAC. Hiro shared that a RSSAC Caucus Work Party is currently developing a tool to gather local perspectives of the RSS, such as identifying underserved regions in terms of number of instances, but also performance measurement and monitoring.

In response to Che-Hoo, Hanyu shared in the chat room that based on a root server instance analysis done by her organization, the total number of instances had more than doubled in the past five years. North America observed the highest compound annual growth rate of 25.2%, followed by Asia with 23.5%, Africa 21.4%, South America 19.6%, Europe 19.2%, and Oceania with 12.2%. Currently, the US has the highest total number of instances (more than 20%), followed by Germany, Brazil, Canada, and the UK.

Che-Hoo also asked where the PRS’ funding might come from, to which Hiro said it is currently undecided although some could perhaps come from ICANN or revenue from domain name registrations.

  • No labels