15.30 - 15.45: Chairs, welcome and agenda

• note taking: Kristina Yasuda

15.45 - 16.00: Carolina Caeiro (onsite), Nathan Alan Internet Standards Tracker

Presentation

• from DNS research foundation
• motivation of the research: prevent internet fragmentation and identify potential standards that lead to it
• Why Standards?
  • Standards shape future of the internet
  • participating in standards has barriers (time, etc)
  • SDOs are the opportunity to make Internet better
• Internet Standards Obervatory
  • funded by ISOC foundation and RIPE community Fund
  • current SDOs focus: ITU-T, IETF, ETSI
  • supporting academic research with the help of internet society
• Standards Tracker
  • can filter by proposed use
  • helps users to decide which standards proposals to prioritize
• Dashboards for the analytics platform outputs
  • steps to get insights
  • 1. document upload. select options to process the doc
  • 2. extracts phrases/words in the system. and tell the system what key words one is looking for
  • 3. can create one’s own scoring system
  • raw output is indexed documents - can filter the data, join datasets and perform calculations
  • 4. can produce visualization of the data: most frequently occuring phrases and data, can search those.
  • system is very flexible and lets you determine your phrases/words of interest
• resources: ISTO knowledge center

Q&A

• Dirk Kutscher: what is the color-coding?
• Carolina: a way to signal, based on the dictionary, which ones require more attention and participation from the community. Would like to incorporate more civil society input.
• Dirk: a lot of contributions from China. how do you ensure this is not a political tool?
• Carolina: analyzed proposers in ITU WG 95% are Chinese or Korean, so yes, there is this bias. Hoping to diversify when expanding to IETF.
• Peter Koch: who is the target audience? how can one access?
• Carolina: started with 3 groups, it can grow. To access ITU data, you need to be a member state. For IETF tracker, we plan to open up more. Indexing is public
• Colin Perkins (individual): how much of this needs custom data and custom indexing vs for IETF how much of keywords can be pre-programmed to expand the usability?
• Nathan: system needs to be fed information. Any information we can get to improve search results would help.
• Carolina: ITU-T tracker is used as a basis for IETF tracker, so would like feedback how to tailor it better for IETF.
• Mallory Knodel: in ITU-T useful for member states where participants cannot track all the work. but categorizing engagement by color is concerning. amount of engagement vs how far the draft is in the process - the latter matters a lot. suggest asking WG chairs which drafts they want to prioritize would be more helpful than computer says “read this”. please treat IETF differently from other SDOs.
• Carolina: a lot of thoughts on this topic, will follow up

16.00 - 16.15: Sebastian Benthall (remote), Bigbang update

Presentation

• BigBang 0.5 Goffman. Update to the work that started in ietf 116 [connectivity problems during presentation]
• next steps: mapping actor dominance in WGs
  • actors: individuals, organizations (affiliations, institutions)
  • activities: leadership (chairs), authorship (RFC authors), influence (ML activity)
• BigBang 0.6 roadmap
  • data ingest, analysis, other SDOs (ITU-R, ITU-T, 3GPP)
• https://github.com/datactive/bigbang

Q&A
none

16.15 - 16.30: Susan Hares (on site), Consensus decision-making in the IESG

Presentation

• looking at what happens after MLs, WGs. genre of social studies
  • IESG mission. RFC2026 describes ietf process.
  • IESG review can take from weeks to years. insteresting to understand why.
  • Big news: time that it takes through IESG depends on the decisions the chairs make on the drafts. 50% result in draft becoming an RFC, WG forming, etc.
  • no way to record how many decisions chairs make. Hidden in IESG, need to dig out.
• three phase mixed-study; models
  • what percentage of decisions chairs make? if their decision matters, how can we predict the decision?
  • could not get 1:1 predictions. went to solidarity.
  • solidarity def in leadership: where you will put an extra effort.
  • looked how conflict works with solidarity. end result: conflict matters.
  • please go read definitions.
  • what matters is amount of data had to dug through. amount of data in data tracker is legendary, collecting it is difficult.
  • reliability 90% is good - anything below 80 is not acceptable.
  • take away. perfect match with RFCs. no interviews.
• three conclusions
  • quantity of quality data matters. going from 10% to 100%.
  • triangulation is critical
  • solidarity appears to be better than OCB for IESG.

Q&A

• (sorry could not hear the name): Amount of data in datatracker about IESG decisions has steadily increased overtime. Is this new data what you need or still gaps with what you need?
• Susan: for drafts, yes. for WGs, what is not there is the management cycle. 2015-16, repetitive management on IPR, was just repeating. had to create a stream of management actions. 1/3 of time chairs spend is on their management items.
• Nick Doty: following up on the data being exportable.
• Susan: data analysis uses IPML analysis aka string analysis with human being behind saying what is important. leadership analysis. Accepted tool. hope it’s getting better to export. mechodology of coding: sent a survey to iesg and used those questions to encode it. ex: are you willing to go extra mile to do xxx and then went back to the minutes to check if they have checked it.
• Andrew Campling: based on what you looked at what stands out that should be done differently?
• Susan: give time to socialize. solidarity is based on trust. trust is based on social interactions. please do not cut the funding on iesg getting to know people. ietf chairs need to work on it. you will get more documents in the end.
• Ignicio: how hard is it to identify solidarity? hard to reflect in minutes?
• Susan: first, look at the survey. than look at the minutes if it matches with the survey results. social science technique. did you know only 11/15 people ballot? reason: busy, etc.
• Susan: this is out of leadership scope too. best innovation occurs inbetween.

16.30 - 16.45: Matthew Russell Barnes (on site), Communication Patterns in the IETF. Queen Mary University of London

Presentation

• built a social graph of Mail-List (ML) interactions. Stats with 10,000 nodes, 560,000 nodes.
• ML have become significantly less active over time
• three layers of leadership: participants, chairs, Area directors
• terminology
  • degree: amount of activity of an individual
  • average neighbor degree: activity of indiv’s connections
• positive and negative correlation
  • whether neighbors are helped or stifled by active individuals?
  • positive correlation for both for philanthropy and community
  • if WG chairs talk a lot, participants talk more.
• who people talk to
• temporal three-edge motifs
  • owtward star: makes announcements
  • inward star: gets questions
  • mixed star: both makes and gets
• results
  • similar level of all 3 stars in 3 groups
  • more inward and mixed stars among chairs/ADs: participants like to talk upwards and have discussions
• WG chairs are facilitators

Q&A

• what happened in 2016? areas changes and chairs got more questions than usual
• Carolina Caeiro: did you take into account “usual suspect” - active contributors?
• Matthew: not really. Just distinguished three levels of leadership.
• Jay Daley: when looking at philanthopy - have you correlated the length of mails and contribution? does that disencourage or encourage others?
• Matthew: did not look into the content of the messages
• Kristina Yasuda: curious how the ML discussions decreased. assume Git Hub (GH) is the replacement. any plans to look into this?
• Matthew: plan to look into GH next
• (sorry did not catch the name): +1 to the previous speaker
• Marek Blachut: question on the focus
• Matthew: looked IETF broadly
• Alexander Railean: in the long distance, GH is a proprietary platform that might be taken away. we might lose something transitioning there. ML does not disappear. would encourage to consider collectively what we lose moving to GH

16.45 - 17.00: Nick Merrill (on site), CDNs and States.

Presentation

• most internet traffic is from offnet. Tier-1 network is much less relevant than it used to be.
• curiously few research on CDNs and state.
• Internet is geopolitics. bidirectional relationship of causality
  • Iran has selective Internet regime.
  • East Ukraine’s routing dependency on Russia. will be easier for East Ukraine to deny internet access
• Questions
  • 1. how do states want the Internet to be?
  • 2. how do CDNs want internet to be?
  • know less about 2 and even less about relationship btw 1 and 2.
  • why care?
• David Clark control point analysis 2012: it is very complicated what happens when users use internet, BUT there are few control points that are run by handful of companies with only one jurisdictions behind them which is USA.
• US gov consents how CDNs work. but not all states do.
• stuxnet explanation. sohpisticated zeroday that attacked Iranian nuclear plant.
• will states even be motivated to stuxnet like attack on CDN?
• some states will suffer less from such attacks because they rely less on CDNs than others.
• if the goal is to disrupt internet, CDN is a good way to do it.
• Q to the group on question 2. if we can put 2 against 1, can identify tensions and collaboration opportunities.

Q&A

• Brian Trammell: post let’s encrypt data. Open infrastructure - if can find data centers, it is a choke point. Analyze the replaceability - how much open standards helps with that
• Mallory Knodel: case not present. Neither state nor CDN that creates this iland. ex: Armenia small proportion of the internet runs through Armenia more through the neighbors like Iran - as the result very censored even if the state does not necessarily want it to be that way.
• Dirk Kutscher: 2 ways of looking at CDN: centralized points, but also shields to DDoS attacks. Hard to say abolishing CDNs will help.
• Matthew: did not talk about abolishing CDN. Only relationship with the state.
• Dirk: you mentioned censorship, but it happens at multiple layers. Connectivity, but probably more effectively at the application layer.
• Andrew Campling: scope of CDN is pretty broad. it’s homogenuity is easy to crack. Like how Cloudflare manages endpoints.
• Chair: most of the users in the Internet belong to the West.
• Matthew: data is skewed because of the source
• Ramakrishnan Raman: what is the definition of US-based company? (Merrill, Pulse Centralization Dashboard, Internet Society)
• Matthew: if you are listed on NY stock exchange, congress can question you in a hearing. if you are not on NY stock exchange, you are free from this.
• different goals from one another, hypergiants like google have different goals from akamai. akamai and cloudflare also have different goals because serve different market
• Vasna Manojlovic: what do the endusers want rfc8090. does not matter if you are colonized by state or a private corps. gaia as a good alternative