DINRG Meeting Minutes

DINRG Workshop on Centralization in the Internet – Minutes

• June 3rd, 19:00 UTC to 22:00 UTC
• Meeting material
• Video recording

Agenda

Workshop Submissions

This is the input that was submitted before the workshop. Thanks to all contributors!

1. Measuring Web Centralization, Trinh Viet Doan
2. Centralization is about Control, not Protocols, Henning Schulzrinne
3. Protocol and Engineering Effects of Consolidation, Dominique Lazanski
4. Centrality in the Internet, Geoff Huston
5. We are probably doomed to Internet consolidation, but we should still try do something., Christian Huitema
6. How We Got from There to Here: Searching for the Root Cause, Lixia Zhang
7. Mitigation Options against Centralization in DNS Resolvers, Jari Arkko
8. Decentralization != Equality?, Trinh Viet Doan
9. Decentralized Data Infrastructures for the New Digital Economy, Thomas Hardjono
10. Autonomy is the real goal, Phillip Hallam-Baker
11. Focus on Standards Adoption, Charles Eckels

Workshop

The workshop was organized in four elements: chairs’ presentation, two panels, and an open discussion.

Chairs’ Presentation

Chairs’ Presentation

• motivating workshop
• expecting a longer journey with a series of workshops/meetings on this topic

Panel #1: WHY: Is Big Really Bad?

Topics

• Economic, societal, regulation perspective
• Centralization in the Internet and Elsewhere

Intro Talk: Geoff Huston: Is “Big” Necessarily “Bad”?

video

• “Where it came from?”
• Dominance of a small number necessarily bad?
• Or a particular outcome of this industry?
• Foundations: US in the 1870s
  • dislocation of prviously agrarian workforce by the civil war created opportunity for expansion of industrial workfoce through urbanisation and immigration
  • massive expansion of national railway system allowed efficient distribution of goods throughout the year
  • liberalized capatial markets with offshore investment encourages entrepreneurial ventures
  • telegraph and telephone network allowed the project of power and influence, allowing corporate entities to expand their effecive scope beyond single locale
  • industrialization transformed small scale cottage industries into large scale industrial production
• The Gilded Age
  • 1870 – 1890, total economy transformation
  • industrial innovation and enterprise with elements of greed, corruption and labor exploitation
  • companies overwhelmed regulatory measures of the day
  • companies wrote their own rules
• The Sherman Act
  • attempt to curtail excessive aggregation of power that interferes with trade and competition
  • to protect the public from the perils of failure of a market to sustain competition within the market
  • prohibits anti-compeitive arrangements and conducts that creates market monopolization
  • applied in 1911 to Standard Oil, American Tobacca and General Electric
  • but then not applied again until 1982 to AT&T
• What happened between 1911 and 1982, and after 1982?
  • unintended side-effects of applying the Sherman Act may be a factor – break-up of Standard Oil and American Tobacco caused economic panic of 1910-1911
  • Lobbying by dominant enterprises
  • US assumed a dominant position in international markets
  • US domestic economy was a major beneficiary of large scale US enterprises dominating global market, making them politically untouchable
• Loius Brandeis – “Big is Bad” by definition
  • Brandeis: US Supreme Court – 1916 - 1939
  • big business too big to be managed effectively in all cases
  • growth of these very large enterprises that were at the extreme end of the excesses of monopolies
  • quality of products tended to decline, and prices of their products tended to rise once they had achieved market dominance
  • when large companies can shape their regulatory environment, take advantage of lax regulatory oversicht to take on more risk than they can manage, and transfer downside losses onto taxpayer, we should be concerned
• Counter Argument: Theodore Roosevelt
  • US President 1901 - 1909
  • There are legitimate economies of scale where large enterprises could achieve higher efficiencies and lower prices to consumers in the production of goods and services by virtue of volume of production
  • large enterprises gain access to volumes of capital and resources that are otherwhise inaccessible
  • example: auto and electricity industry
  • Roosevelt: better regular bad bahavior
• Regulation?
  • Brandeis: cannot regular monopolies anymore
  • John Ralston Saul: should never expect private sector to safeguard public interest
  • if we regular, then are we attempting to forestell the condition, or applying regulatory sanctions on the consequent behaviour?
  • what jappens during periods and rapid chance in enterprise landscale?
  • how does regulation remain relevant?
  • how can we prevent regulation from embedding enterprises in the past as institutional norm?
• The Internet’s Gilded Age
  • same today
  • massive lobbying by Internet companies
  • Market Sentiment: capital concentration
  • enough market influence to set their own rules of engagement with users, each other, third party supplies, regulator and governments
  • by taking a leading position with these emergent technologies, these players can amass vast fortunes, with little in the way of accountability to a broader common public good
  • situation as skewed and as corrupted as in the US in the 1880s
• What’s the problem?
• What we want?
  • enterprises have focused very sharply on giving users precisely what they want
  • ability to customize a solution to a market of 1 and still bring economies of scale to that market underlies their success
  • their ability lies in achieving a critical mass in size that allows them sufficient bulk to capitalize as asset that individually is worthless
  • we we all use these services – because they are obtained through the capitalisationb of something that individually we are incapable of capitalizing
• Side-effects
  • Surveillance Capitalism
  • much of the wealth and impact of these activities is built upon a foundation of aggregation of individual user behavior and constructio of personal profiles
  • has benefitted from cavalier attitude towards data security and privacy concerns and the absence of regulatory imposts that attempt to safeguard some basic common aspects of personal privacy
  • US: privacy regulation inside department of trade: you can do whatever, just don’t lie – lobbyism
• What we might want
  • public sector response?
  • perhaps it should focus on user rather than on the market
  • but public sector efforts to so remain largely ineffectual despite the magnitude of the public concern over private sector seizure of the space of public discourse
  • EU & GDPR — fines much too low
  • public sector essential ineffectual
  • social changes on same scale as industrial revolution
• Maybe its more than this
  • if you cannot verbalize the concept – how can you think it?
  • In a world ob abundant content what do we choose to view?
  • Search universal tool for research, library etc.
  • Search arbiter of content selection and assumes a level of utter importance in the world
  • What’s the outcome of search being dominated by a single entity?
• Google Search 89% market share in the US
  • Is it more about what we buy or is this more about what we think?
• Where does all this head?
  • For our society this market-driven digitisation of our world has the potential to be incredibly empowering of incredibly threating
  • or bath at the same time!
  • how do you curb that?
  • is big the problem?
  • or some other aspect that has gotten out of control?
  • if it is out of control, what is the right response?

Panel

• Dirk Kutscher (Moderator)
• Henning Schulzrinne
• Dominique Lazanski
• Geoff Huston

video

Characterizing Internet Centralization

• How would you characterize Internet centralization?

• Examples?

• Henning: Operational Centralization

  • distiniguish several different types of centralization
  • web protocols, DNS
  • not so much that protocol itself encourages or leads to centralization
  • before we do “magic protocol” to “fix centralization”, think about what leads to centralization, even with protocol that were designed with a distributed model in mind
  • CDN as example
  • operational centralization
  • implementation centralization: we seem to converge to only a few one
  • number of mobile operators: three per country
  • CDN similar
  • Search: only one
  • variety of reasons

• Dominique: layers and players

  • agree to notion of operational centralization
  • there are different layer of consolidation, as there are different layers of the Internet
  • not talking about actual protocol itself, but about the actual development of the protocol
  • consolidation also in set of people who are developing protocols
  • consolodidation at network and application layers
  • obvious and long-standing example: e-mail

• Geoff: from dominance of carriage to dominance of platforms

  • locus of value and money shifted up the protocol stack
  • end-to-end
  • but that’s dying
  • now application themselves taking over everything
  • DNS, QUIC just part of the application
  • applications creating their own ecosystem
  • trying to keep the user inside the application as long as possible
  • Paranoid application, ignores libaries and services of the operating system
  • Facebook, Chrome
  • Why is this happening? Money!
  • value of market == aggregate value of consumers’ wealth
  • advertizing changes all of that: access to future money
  • amount of money in future market overwhelm value in the current market
  • newspaper in the past: same dominance of advertizing value
  • folks who are able to sell futures will dominate
  • individuals can’t do that, but Google can
  • distorted market as a result
  • economics of advertizing

• Geoff: hostile environment for protocols

  • protocols were designed open
  • DDoS world
  • everyone outsources services
  • scale of attacks overwhelm capabilities of organizations to protect
  • economics of DDoS and hostility
  • drives consumers into citadels of heavility built castles to defend their assets
  • nothing new

• Dominique: how do we square this?

  • heading towards 18th century newspaper tax (that publishers had to pay)
  • how to we square that and get what we want?
  • Toerless Eckert in chat:
    • There is also a nice opposite example: 12 years ago, i had one unified/monopolist Video service: Comcast. Now i can choose from > 15 competin streaming services. And people do not like this either
  • agree to Toerless’s comment in chat: maybe it’s a human problem?
  • we always want something different than what we have, for example video service & Comcast

• Geoff: video streaming is just an industry in transition

  • Comcast example: industry in transition
  • like few telephone companies that led to hundreds of ISP
  • we got down to three again
  • video streaming is heading into same direction (explosion before aggregation)
  • creates efficiency in the market that other folks cannot compete with

• Henning: Techno-Optimism in the Past and Today

  • Tim Wu: The Master Switch
    • THE RISE AND FALL OF INFORMATION EMPIRES
    • https://www.penguinrandomhouse.com/books/194417/the-master-switch-by-tim-wu/
    • bit older
    • does not account for recent developments
    • but points to lobbying, political influence etc.
  • there is an illusion that occured for earlier technologies where economics actors talked themselves into the notion that “this technology is different”
  • Internet was similar
  • techno-optimism, technology being expected to avoid “bad things”
  • in general, many of us were naive in the first 20 or so years, assuming that the Internet would be resistant against centralization
  • Internet design legend: decentralized ARPANET design
  • implication of decentralization: millions of end systems
  • benign and not so benign illusions
  • many would like to think that technology brings about improvement
  • it’s also good to look back to text books and keynote adresses and the promises that were made
  • and see why those did not quite work out in the way we had anticipated
  • looks at current fascincation with things like Bitcoin which suffers from same illusions as some of the Internet pioneers believed in
  • illusion that technology itself was resistant to centralization because after all it is designed to be a decentralized system
    • routing, adressing, control, delegation whole notion of Internet as interconnected domains etc.

Perceived Benefits and Factors leading to Centralization

• Henning: security, economics, operation
  • because world has gotten more hostile
  • Neither Facebook nor Google has been a victim of a ransomware attack (as far as we know)
  • unlike other (non-Internet) companies
  • tendency to build fortresses, like Amazon, Microsoft, Google etc.
  • promise to protect you from attacks
  • don’t think this trend can be reversed
  • demanding that non-Internet businesses should run their own services decentralized would be unrealistic
  • Antitrust discussion
  • philosophical change in the US
  • Chicago school of economics had the notion that old model was not sufficiently focused on consumer benefit
    • one should look to see whether consolidation would benefit consumer welfare
    • argument for concentration in telco business was always: consumers would benefit ecomonically
    • often hard to argue against
  • Similar argument could be made for Internet services, e.g., Facebook. Users seem to like these services
• Geoff: legislation to protect “national interests”
  • What counters centralization in todays world is actually legislation
  • Governments playing national interest card
  • Rupert Murdoch in Australia, press industry dying
  • Google taking away advertizing business from traditional press
  • Google cannot be banned
  • traditional publishing business forcing legilators to impose taxes on Google, and subsidize incumbent business
  • expecting carriage industry to do the same in the next 5 to 6 years
  • “national interest” argument (critical infrastructure, businesses)
  • expecting rise of ITU-T and international mechanisms to reinforce national regulatory structures
  • as a counter to international centralization (Google is dominant across all markets)
  • one large Internet is over
  • future of a few Google- or Facebook-like ecosystems or lots of little national and regional ones
    • rather expecting many national and regional ones
  • so counter to centralization is actually nationalization rather than decentralization

Potential Outcome of Legislation and Regulation

• Dominique
  • ITU-T may be interesting approach to multi-lateral legislation
  • standardization of processes for regulation on national level
  • agree that it might happen on national/regional level
  • most users don’t care
  • consumers don’t see negative aspectes of centralization

Discussion

• Toerless Eckert: What are top three items to start solving the problems?
  • Geoff
    • some of the issues are deep and difficult
    • it used to be that the amount of capital required to create a national telecom services was beyond the capacity of private capital markets, therefore public investments
    • in many countries (US was exception) railway, road, telephone system was all built by the state – because it was the only party with sufficient money
    • now new world: national interest is quoted again and again, but not with money
    • money is sitting in private investment vehicles
    • those are heavily concentrated
    • VC interest: requires billions of investments, not millions
    • what can be done about this?
    • it’s a capital issue
    • we created this relatively lax and free moving capital markets
    • aggregating capital into free moving vehicles that produce outcomes
    • all other player get massive barriers, can’t aggregate capital to compete
    • real change requires understanding regulation of international movements of capital
    • look at how capital is used for intrastructure investments
    • state and national issues need to play a role in order to regulate to create different market outcomes
    • until this is done, anything we do on a technical level or on a subject-matter regulatory level is never going to work
    • FCC does not have long enough arms to even slightly engage in the scale of the problem over the next 10 years. Wrong agency, wrong remit, not enough money.
    • part of this actually a much bigger question than just this industry
    • Actual question is how do we create large-scale capital movement to further or fuel national communities rather than having a small number of international bodies run their own agenda
    • don’t see clean solution rather than going to the root cause
  • Dominique: capital is one solution
    • in the UK: try to get hardware-based startups instead of just software-based ones
    • historical cycle that we are seeing again and again
  • Henning: avoid fatalism
    • there are some more achievable things you can do
    • privacy laws, e.g.,GDPR
    • making it more difficult for startups to get acquired
    • regulatory mechanisms: data portability, restrictions on end-to-end new markets
    • Sherman Act-like laws: making it illegal for companies like google to dominate adjacent markets (e.g., Google moving from Search to everything else)
    • more local problem: as a community we have been complicit (techno-optimism)
    • illusion that going to back to peer-to-peer protocols will make a difference
    • we have to be a lot more skeptical against those arguments because they will translate into the policy domain (politician adopting techno-optimism)
      • “new Google or new Facebook will come along”
    • consolidation will not be as transient as it was in the past

Panel #2: HOW: How Did We Get Here?

Topics

• Architectural & technical discussion
• Factors enabling centralization?

Intro Talk: Christian Huitema: How Does Centralization Start?

video

• Investment starts with unsolved problems
  • distributed search
  • identity management, contacts, presence
  • spam, quality of e-mail clients
  • DDoS
  • today, no company would dare running their own servers in their basement
• Deal with many platforms
  • motivated by Moxie Marlinspike: centralized architecture design for Signal messaging system
    • https://sorincocorada.ro/signal-messanger-architecture/
    • https://media.ccc.de/v/36c3-11086-the_ecosystem_is_moving
  • decentralized design is really hard today
  • multiple OS
  • multiple versions
    • new releases, countries
  • coordinated developments
    • large team required
  • can standard clients evolve quickly?
    • until a feature is universial?
    • example jabber: several clients, very hard to deploy new features
• Centralized Big Data
  • more customers, more data
    • personalized offerings – perceived useful
    • centralized feedback
  • more data, more solutions
    • invest and fund the deployment
  • more solutions, more customers
    • coordinated deployments
  • more data, less competition
    • barrier to entry
    • example Google Search and Microsoft Bing
    • Google knows so much about you – newcomers don’t
• Centralized and Free
  • make free with advertisements
    • who can compete with free?
    • difficult to compete
  • advertisements require data
    • Bigger means better data
    • surveillance capitalism
    • big issue
  • Better data to tune the service
    • The system knows what you want
    • There is probably a notion here: “the more it is free, the more it is centralized”
• How it ends, sometimes
  • Once big players are installed, they typically stay big
    • there may be some changes
    • think evolutional process
    • until one organism dominates
    • sometimes, exceptions e.g., Windows OS
    • but it took many (government) actions
  • Windows consent decree
    • API published
    • allowed for interop with new systems
    • disentangle app/OS bundles
    • may be a lesson for the future: if you want to break up big players, interoperability and publishing the API may be a big issue
  • Market changed
    • phones: Android and iOS
    • laptops: windows, Mac, Linux
    • cloud: AWS, Azure, Google…
  • Still not distributed
• Decentralized Competition Seems Hard
  • Some problems are political
    • surveillance capitalism vs. privacy
    • acquisitions vs monopoly
    • monopoly vs. APIs
  • Standards take a long time
    • identity, Spam, DDoS
    • really difficult to compete with proprietary development speeds
  • Centralized business model is simpler
    • much simpler
    • address multiple platforms, evolve quickly

Panel:

• Lixia Zhang (Moderator)
• Thomas Hardjono
• Jari Arkko
• Trinh Viet Doan
• Christian Huitema

video

• Jari Arkko
  • contributions was on way to reduce the effects of centralization
  • privacy currently not a priority, too much data centralized
  • looked into discovery mechanisms
  • finding safer services
  • mostly economics, but few other things, including technology
• Thomas Hardjono
  • computation and social science
  • big data analytics
  • EU data governance act, including data cooperatives
  • resource in 19th century was labor
  • today: data
  • need “new deal on data”
  • Internet never had any control of the data that flows through as per end-to-end principle
  • individuals could be co-owners of data that they create
  • how to manage that?
  • what if inidividuals could share data in controlled ways
  • what is infrastructure of the future that would allow data holders make data more accessible (for processing, storing)
  • next Internet revolution is data infrastructure
• Trinh Viet Doan
  • Internet measurements in context of centralization & decentralization
  • how much centralization is out there?
  • do measurement result fit perception
  • what factors are driving centralization?
  • tradeoffs between centralized and decentralized systems that users would have to accept
  • measuring web centralization based on where web sites are hosted
  • increasing trend towards centralization
  • equality considerations for decentralization
  • who has to pay upfront costs?
  • which are typically higher compared to centralized services
  • cannot compete with free
  • decentralization may accidentally exclude people
• Christian Huitema
  • is there space for some kind of competition?
  • example DuckDuckGo – different business model, some small market share
  • are there open space other than decentralization like this?

What has happened?

• Lixia: too much succees or failure?

  • technically may have gotten blinded by success and missed some problems

• Lixia: why is security failing?

• Thomas

  • infrastructure hacks suggest that something is wrong
  • architecture seems to have problems

• Lixia: Why?

• Christian: Security does not pay

  • it’s a market thing
  • you don’t make money by it
  • unless there is regulation or customer movement, better security is not coming

• Lixia: technically community aspects?

• Thomas: OAuth-2.0 example

  • was done in applications area, now being rewritten

• Jari: two reasons why security is failing

  • security is hard
  • some parties don’t care enough, so they lose other people’s data
  • if there is no harm for them, there is motivation to improve security
  • some players are too big to be held responsible
  • not just accidents, also deliberate sale of information

• Christian: security is already centralized

  • building big castles
  • most attacks today (ransomware) are not caused by anything in the Internet
  • mostly exploiting zero-day bugs in implementations
  • typically with support of state-level actors
  • need to get to next-level of software security, but probably not a topic for today
  • In Internet model there is one relevant property: everyone can send data to everyone (unsolicited)
  • Was a design goal
  • also an enable for spam, DDoS
  • we have not worked out a defense against it

• Viet: security is a fast-moving topic

  • smaller players can spend less money in keeping up
  • could also be a factor why security is easier in centralized systems

• Lixia

  • if you don’t let people to freely communicate with each other, would you not end up with a system like today?

• Christian: yes, there is a tension

  • the fact that everybody can talk to everybody is an important property
  • but also led to some of the problems today
  • and we have not teased this apart yet

• Lixia

  • are we lacking an effective solution to enable both free peer-to-peer communication while preventing abuse as well?

• Jari

  • attack question has broader relevance
  • some things are easier when you are big (“attack resistance”)
  • if we had a different design, maybe we wouldn’t have to resort to centralization so much
  • But: we could also be just one indirection away from another level of attacks

What has gone wrong with the web?

• Lixia: focus on content seems incompatible with direct communication
  • Tim Berners-Lee also had a different vision
  • Why has the web developed in different direction?
• Viet
  • traditional broadcast: many readers, only a few writers
  • same models seems to have been applied to the web
• Christian: advertisement-based payment models
  • goes back to a debate in the 90s on how to fund the web
  • “How do I pay the publishers?”
  • One option: no payment
  • Other options:
    1. micro-payment
    2. something else
  • micro payment was discussed, but in the end advertisment model won
• Thomas
  • publishers help authors with copyright issues and enforcements
  • some change is happening (direct funding/compensation)
• Jari
  • much of this in the nature of humans
  • many of us want to consume media
  • but there are other issues
  • centralization of e-mail etc.

Discussion with other Participants

• Henning
  • one reason that is technical, but the solution to it may not be: implementation complexity
    • example: HTTP specification
    • no-longer for an individual to develop a web browser competitively
  • we have increasingly created systems that are too complex to be built in smaller teams
    • barriers to entry
    • Google’s power is to big because they are also in browser market
• Christian
  • there are 18 implementation of all QUIC right now
• Wes Hardaker
  • motivation for creating new things on the Internet is always self-serving
  • only profitable when things are small
  • at some point, it can be more profitable to prevent competition
  • when there are no more users to gain
  • it’s cheaper to take out competition
  • buying other companies, prevent standards
  • question: what are the bigger carrots that can motivate companies to continue standards?
  • interoperable security is harder and more expensive
  • walled secure gardens designed so that people cannot get out (not that nobody can’t get it)
• Viet
  • interoperability also ties to data migration
  • should be easier
• Christian
  • One of the hardest part when doing implementations is to test them
  • scenarios are important (“data”)
  • big players have lots of telemetry, new player don’t.
  • example: QUIC packet loss, latency
  • can reenforce position of big players
• Jari
  • many ways to prevent competition
  • beyond standards and open source: open data and open data models
  • need to have that discussion
• Charles Eckels
  • standards complexity make it hard to use them
  • result: move to platforms that abstract aways the complexities
  • applications just provide what you need
  • there is a lot of time and money invested in improving platforms
  • for standards, we should spend more effort to make standards easier to consume

Open Discussion

video

• Dirk: what is a good way to start this discussion?
  • have we actually made steps towards understanding the situation today?
  • is there anything else needed
  • is there any technical/architectural angle or is it just economics?
• Lixia
  • what have people learned?
  • what are good topics for future workshops on this?
• Henning
  • one lesson: there are things above our paygrade, not easily solvable by engineering
  • there are two things that are solvable though:
    1. we have a broad sense of the technical landscale. Everyone know about Google and Facebook, but the problem is broader. This is browser, routers, local access, e-mail operations, CDNs etc.
       • could be good to document these cases
    2. Even if we cannot solve the big problem, there are some topics that can make a difference
       • data migration
       • reduce complexity for testing
  • these steps could be useful marginally, but are more realistic to tackle
  • don’t just wait for antitrust law changes (which are probably also required) – not our speciality
• Geoff
  • evolution of IETF itself
  • early 1990s: bunch of researchers and research community sharing information
  • public sectors engaging
  • product were open
  • work advantaged all players who bothered to attend
  • not working in the interest of any particular player
  • since then, IETF has become an industry forum
  • like many other industry fora: works to the benefits of the players in a different way
  • standardize minimally and effectively fold in real aspect of innovation
  • companies that invest in innovation whish to reap the rewards exclusively
  • result in the IETF is interesting
  • IRTF like “crippled, lame distant cousin”, whereas the bulk of the activity is on further monetizing and centralizing the Internet
  • outcome of the focus
  • changed from open innovation/research to closed, privatized innovation/research
  • We reaped the rewards of that transformation
  • We wanted IETF to be bigger & relevant, catered to industry needs
  • created a positive feedback loop to further privatization and centralization
  • if you want to create different culture you have to re-embrace the issue of public sector research
  • why don’t we fund universities if it is not to make innovation as a product that is accessible to all
  • once you allow that, incumbents would no longer have access to innovation exclusively
  • current template inside many western economies to make open research a crippled, small brother to private investments, is entirely wrong and leads to these skewed outcomes
  • So IETF should give more focus and emphasis on IRTF and make those processes open and accessible to all
  • enabling folk to use innovation without privatizing it
  • whe are symptomatic to the larger problem
  • mainstream of activity is still in the IETF
• Christian
  • participation in IETF is focused on a few countries – those with big companies
  • suggest to open up to broader participation and open research
  • question: who pays for that?
• Christian Tschudin
  • concerned that this exercise would become a blame-shifting game
  • also have to look at how we have contributed to some developments
  • there are also technical biases that we should try to find:
    • “always online”
    • “client-server” (even for p2p)
  • there are some implicit biases and thinking that lead to centralization
• Toerless Eckert
  • if the core of the problem is above our paygrade, do responsible people understand the technology?
    • regulate interfaces on monopolistic platforms
    • build federated documents
  • informational documents for political decision makers?
  • example: standards that define how to get your own data in and out of systems so that you can move between them
  • or that you can build them in a federated way
  • what are simple starting points?
  • explain to political decision makers what benefits they could get and how IETF and other bodies could help?
• comments in the chat:
  • standards complexity
    • Dirk: not sure it’s a problem
    • community effort days of IETF are long over
    • today companies only concerned with IETF standards if they absolutely need them
    • if they don’t need them – IETF is just too slow
    • SDN as a symptom: “defying standards” – just build your own protocols
    • QUIC as an exception
    • but overall it could be misleading to say that we have to simplify our standards
• Lixia: reflection
  • Henning in first panel: In the early days people had illusion that decentralized nature of Internet would prevent centralization
  • Geoff: if you focus on money making you sacrifice the outcome in other dimensions
  • Currently people show interest in decentralization
  • seems to come from interest in protecting consumers
• Marie-Jose Montpetit
  • autonomous operations
  • networking seems to be associated with centralization and older views (“telecom”) these days
  • In Canada, automation in remote areas
  • decentralization also about disconnected operation at the edge
  • creating subnetwork in autonomous way – redefining what networking means
  • also: energy efficiency (“green networking”)
• Dirk: Could have different design decisions led to a different outcome?
  • different trust systems that web PKI?
  • would that have helped to avoid these economic developments?
  • or does it not matter in the end because it’s all economics?
• Lixia:
  • architecture is the product of the day of the technology
  • example: TCP/IP was once the most advanced technology for the Internet
  • 40 years later, things look different
• Christian
  • big decisions of the architecture of the Internet:
    1. everybody can speak to everybody
    2. network does not know what it is carrying
  • important to enable any kind of (new) application
  • big step forward compared to 1980’s telecom networks
  • in those networks: telecom provider to capture the value of the applications (Minitel etc.)
  • Internet broke that model
  • today, you see many telecom providers trying to capture that value back
  • these two decision get you the current architecture
• Lixia
  • one clarification: there is architecture and there are architecture principles
  • example: “everbody can speak to everybody” more of a principle
• Christian
  • yes, there are architectural design decisions that hurt us later
  • ip addresses can be forged, unwanted traffic
  • we could treat these as problems
  • for areas such as identity, security, Qos: anything that relies on communication between multiple providers and their networks, never succeeds
  • we could accept that as a fact of life and build rest of network around that
  • consider all QoS efforts since the 1980s
• Lixia: APN – application-aware networking
  • fine-granular QoS?
  • Christian: very skeptical
• Jari
  • big fraction of monetary value goes to some parties today
  • maybe they have to be bypassed again like telecoms in the past
  • we have been stuck with server-based communication model but it does have to be like this
  • there are some things we could do
  • not everybody wants that change, but we could potentially do it
• Lixia
  • we tried to collect information for planning next steps
  • there seems to be interest to continue the exploration and do something about this problem
  • please send all questions & comments
  • chairs will think about next steps and discuss with community about next steps
  • might be long process, but this was a great start.
• Dirk
  • will summarize workshop
  • planning for a meeting at IETF-111