Report from the IAB Workshop on Analyzing IETF Data (AID), 2021

2.1. Tools, data, and methods

The IETF holds a wide range of data sources. The main ones used are the mailinglist archives, RFCs, and the datatracker. The latter provides information on participants, authors, meeting proceedings, minutes and more. Furthermore there are statistics for the IETF websites, working group Github repositories, IETF survey data and there was discussion about the utility of download statistics for the RFCs itself from different repos.¶

There are a wide range of tools to analyze this data, produced by IETF participants or researchers interestested in the work of the IETF. Two projects that presented their work at the workshop were BigBang and Sodestream's IETFdata library; the RFC Prolog Database was described in a submitted paper (see Section Section 4 below). These projects could be used to add additional insights to the existing IETF statistics page and the datatracker statistics, e.g., related to gender questions, however, privacy issues andd implication of making such data publicly available were discussed as well.¶

The datatracker itself is a community tool that welcomes contributions, e.g. for additions to the existing interfaces or the statistics page directly (see https://notes.ietf.org/iab-aid-datatracker-database-overview). Instructions how to set up a local development environment can be found, at the time of the workshop, at https://notes.ietf.org/iab-aid-data-resources. Questions or any discussion can be issued to tools-discuss@ietf.org.¶

2.2. Observations on affiliation and industry control

A large portion of the submitted position papers indicated interest in researching questions about industry control in the standardization process (vs. individual contributions in personal capacity), where industry control covers both, technical contribution and the ability to successfully standardize these contribution as well as competition on leadership roles. To assess these question it has ben discussed to investigate participant's affiliations including "indirect" affiliation e.g. by funding and changes in affiliation as well as the nessecarity to model company characteristics or stakeholder groups.¶

Discussions about the analysis of IETF data shows that affiliation dynamics are hard to capture, due to the specifics of how the data is entered but also because of larger social dynamics. On the side of IETF data capture, affiliation is an open text field, which causes people to write their affiliation down in different ways (capitilization, space, word seperation, etc). A common data format could contribute to analyses that compare SDO performance and behavior of actors inside and across standards bodies. To help this a draft data model has been developed during hackathon portion of the workshop which can found as Annex 1 - Data Taxonomy.¶

Furthermore, there is the issue of mergers and acquisitions and subsidiary companies. There is no authorotative exogenous source of variation for affiliation changes, so hand-collected and curated data is used to analyze changes in affiliation over time. While this approach is imperfect, conclusions can be drawn from the data. For example, in the case of mergers or acquisition where a small organizations joins a large organization, this results in a statistically significant increase in liklihood of an individual being put in a working group chair position (see BaronKanevskaia).¶

2.3. Community and diversity

High interest from the workshop participants was also on using existing data to better understand who the current IETF community is, especially in terms of diversity, and how to potentially increase diversity and thereby inclusivity, e.g. understanding if are there certain groups or lists that "drive people away" and why. Inclusivity and transparency about the standardization process are generally important to keep the Internet and its development process viable. As commented during the workshop discussion, when measuring and evaluating different angles of diversity it is also important to understand the actual goals that are intended when increasing diversity, e.g. in order to increase competence (mainly technical diversity from different companies and stakeholder groups) or relevance (also regional diversity and international footprint).¶

The discussion on community and diversity spanned from methods that draw from novel text mining, time series clustering, graph mining and psycholinguistic approaches to understand the consensus mechanism to more speculative approaches about what it would take to build a feminist Internet. The discussion also covered the data needed to measure who is in the community and how diverse it is.¶

The discussion highlighted that part of the challenge is defining what diversity means, how to measure it, or as one participant highlighted to define "who the average IETF is". The question was also raised what to do about missing data or non-participating or underrepresented communities, like women, individuals from the African continent and network operators. In terms of how IETF data is structured, various researchers mentioned that it is hard track conversations as mail threads, split, merge and change. The ICANN-at-large model came up as an example of how to involve relevant stakeholders in the IETF that are currently not present. Vice versa, it is also interesting for outside communities (especially policy makers) to get a sense of who the IETF community is and keep them updated.¶

The human element of the community and diversity was stressed, in order to understand the IETF community's diversity it is important to talk to people (beyond text analysis) and in order to ensure inclusivity individual participants must make an effort to, as one participant recounted, tell them their participation is valuable.¶

2.4. Publications, process, and decision-making

A number of submissions focussed on the RFC publication process, on the development of standards and other RFCs in the IETF, and on how the IETF makes decisions. This included work on both technical decisions about the content of the standards, but also procedural and process decisions, and questions around how we can understand, model, and perhaps improve the standards process. Some of the work considered what makes a successful RFC, how are RFCs used and referenced, and what we can learn about importance of a topic by studying the RFCs, drafts, and email discussion.¶

There were three sets of questions to consider in this area. The first related to success and failure of standards, and considered what makes a successful/good RFC? What makes the process of RFC making successful? And how are RFCs used and referenced once published? Discussion considered how to better understand the path from an internet draft to an RFC, to see if there are specific factors that lead to successful development of a draft into an RFC. Participants explored the extent to which this depends on the seniority and experience of the authors, on the topic and IETF area, extent and scope of mailing list discussion, and other factors, to understand whether success of a draft can be predicted, and whether interventions can be developed to increase the likelihood of success for work.¶

The second question was around decision making. How does the IETF make design decisions? What are the bottlenecks in effective decision making? When is a decision made? And what is the decision? Difficulties here lie in capturing decisions and the results of consensus calls early in the process, and understanding the factors that lead to effective decision making.¶

Finally, there were questions around what can be learn about protocols by studying IETF publications, processes, and decision making? For example, are there insights to be gained around how security concerns are discussed and considered in the development of standards? Is it possible to verify correctness of protocols and/or detect ambiguities? What can be learnt by extracting insights from implementations and activities on implementation efforts?¶

Answers to these questions come from analysis of IETF emails, RFCs and Internet-Drafts, meeting minutes, recordings, Github data, and external data such as surveys, etc.¶

2.5. Environmental Sustainability

The final discussion session considered environmental sustainability. It discussed what is the IETF's role with respect to climate change both in terms on what is the environmental impact of the way the IETF develops standards, and in terms of what is the environmental impact of the standards the IETF develops?¶

Discussion started by considering how sustainable are IETF meetings, focussing on how much CO2 emissions are IETF meetings responsible for and how can we make the IETF more sustainable. Analysis looked at the home locations of participants, meeting locations, and carbon footprint of air travel and remote attendance, to estimate the carbon costs of an IETF meeting. Initial results suggest that the costs of holding multiple in-person IETF meetings per year are likely unsustainable in terms of carbon emission, although the analysis is ongoing.¶

Discussion also considered to what extent are climate impacts considered in the development and standardization of Internet protocols? It reviewed the text of RFCs and active working group drafts, looking for relevant keywords to highlight the extent to which climate change, energy efficiency, and related topics are considered in the design of Internet protocols, to show the limited extent to which these topics have been considered. Ongoing work is considering meeting minutes and mail archives, to get a fuller picture, but initial results show only limited consideration of these important issues.¶

4.1. Tools, data, and methods

Sebastian Benthall Using Complex Systems Analysis to Identify Organizational Interventions¶

Stephen McQuistin, Colin Perkins The ietfdata Library¶

Marc Petit-Huguenin The RFC Prolog Database¶

Jari Arkko Observations about IETF process measurements¶

4.2. Observations on affiliation and industry control

Justus Baron, Olia Kanevskaia Competition for Leadership Positions in Standards Development Organizations¶

Nick Doty Analyzing IETF Data: Changing affiliations¶

Don Le Position Paper¶

Elizaveta Yachmeneva Research Proposal¶

4.3. Community and diversity

Priyanka Sinha, Michael Ackermann, Pabitra Mitra, Arvind Singh, Amit Kumar Agrawal Characterizing the IETF through its consensus mechanisms¶

Mallory Knodel Would feminists have built a better internet?¶

Wes Hardaker, Genevieve Bartlett Identifying temporal trends in IETF participation¶

Lars Eggert Who is the Average IETF Participant?¶

Emanuele Tarantino, Justus Baron, Bernhard Ganglmair, Nicola Persico, Timothy Simcoe Representation is Not Sufficient for Selecting Gender Diversity¶

4.4. Publications, process, and decision-making

Michael Welzl, Carsten Griwodz, Safiqul Islam Understanding Internet Protocol Design Decisions¶

Ignacio Castro et al Characterising the IETF through the lens of RFC deployment¶

Carsten Griwodz, Safiqul Islam, Michael Welzl The Impact of Continuity¶

Paul Hoffman RFCs Change¶

Xue Li, Sara Magliacane, Paul Groth The Challenges of Cross-Document Coreference Resolution in Email¶

Amelia Andersdotter Project in time series analysis: e-mailing lists¶

Mark McFadden Position Paper¶

4.5. Environmental Sustainability

Christoph Becker Towards Environmental Sustainability with the IETF¶

Daniel Migault CO2eq: Estimating Meetings' Air Flight CO2 Equivalent Emissions: An Illustrative Example with IETF meetings¶

8.1. Annex 1 - Data Taxonomy

A Draft Data Taxonomy for SDO Data:

Organization:
  Organization Subsidiary
  Time
  Email domain
  Website domain
  Size
          Revenue, annual
          Number of employees
  Org - Affiliation Category (Labels) ; 1 : N
    Association
    Advertising Company
    Chipmaker
    Content Distribution Network
    Content Providers
    Consulting
    Cloud Provider
    Cybersecurity
    Financial Institution
    Hardware vendor
    Internet Registry
    Infrastructure Company
    Networking Equipment Vendor
    Network Service Provider
    Regional Standards Body
    Regulatory Body
    Research and Development Institution
    Software Provider
    Testing and Certification
    Telecommunications Provider
    Satellite Operator

Org - Stakeholder Group : 1 - 1
    Academia
    Civil Society
    Private Sector -- including industry consortia and associations; state-owned and government-funded businesses
    Government
    Technical Community (IETF, ICANN, ETSI, 3GPP, oneM2M, etc)
    Intergovernmental organization

SDO:
  Membership Types (SDO)
  Members (Organizations for some, individuals for others…)
  Membership organization
    Regional SDO
      ARIB
      ATIS
      CCSA
      ETSI
      TSDSI
      TTA
      TTC
    Consortia

Country of Origin:
  Country Code

Number of Participants

Patents
  Organization
  Authors - 1 : N - Persons/Participants
  Time
  Region
  Patent Pool
  Standard Essential Patent
    If so, for which standard

Participant (An individual person)
  Name
  1: N - Emails
    Time start / time end

  1 : N : Affiliation
    Organization
    Position
          Time start / end

  1 : N : Affiliation - SDO
    Position
    SDO
    Time

  Email Domain (personal domain)

  (Contribution data is in other tables)

Document
  Status of Document
          Internet Draft
          Work Item
    Standard
  Author -
    Name
          Affiliation - Organization
    Person/Participant
        (Affiliation from Authors only?)

Data Source - Provenance for any data imported from an external data set

Meeting
  Time
  Place
  Agenda
  Registrations
    Name
    Email
    Affiliation