Report on Workshop on Centralization in the Internet (2021-06): Lixia Zhang
Marie-Jose Montpetit: Comes from the IoT world
Already doing distributed applications and systems
Large problem with keeping data local
Two sensors in the same place can only talk to each other through AWS
IoT is a different world
Dino Farinacci: Monopoly companies are offering centralization
For text messaging, monopolies offer security
If users object to the monopolies and are willing to pay more for security, might happen
Vittorio Bertola: Economy of scale allows someone to compete on price or other factors
Wants to align language
Lixia: There are different ways that don’t have to be provided by central monopoly
Need to think wider
The network works; what is missing is the protocols
Sanchin Garg: 80% of people can run their own mail servers
The problem is that individual servers get blocked
Andrew Campling: Would have talked about resilience
It much more about network effects than scale
Loss of privacy
People feel that they have no choice about whether they give it up
Mallory Knodel: There are natural places where things will get centralized
What are the benefits vs. detriments
Balancing implementation ease vs. user value
Survey of effects could help make better choices
Geoff Huston: This is about as bad as it gets
Purpose of the Internet was to get away from a regulated situation into markets
Monopolistic power makes competitive entry suicidal
Markets will not help us; nor will regulation
This is not telephony any more; how do you regulate?
We don’t know
Users won’t get us out of this
They want things that are centralized
Getting out of this needs revolutionary change
Lixia: The Internet broke the telecom monopoly
Can happen again
BGP is not in any single hand yet
Agrees with Geoff about revolution
Christian Huitema: Doesn’t agree with Geoff about revolution
No heads on pikes
The problem is the acquisition of power, some of which comes from acquisition of data
Behind this is personalized advertisment
If you don’t pay, free brings centralization
Internet has to stop being free
Dominique Lazanksi: We need further discussion
Impassioned interest
We need more measurement
Look at different layers
Unintended consequences of regulation
Christian: IETF could work on minimizing data
The more the protocols are open, the easier it is to work up
Lenny Giuliano: The best we can do is solutions that lower the bar to contribute
Running a mail server in my house is a pain in the ass
Consumers move to where they find value
No confidence that regulation will solve the problems
Dirk Kutscher: Open discussion on next steps
Even if we do revolution, need to know the factors and history
Wants to figure out follow-up activities
DINRG started with over-ambitious approach with “blockchain will save us from centralization”
Where can we make a difference?
Looking to the future: Mark McFadden
Lots of people are writing individual drafts on centralization
Give a new home for these items: DINRG
Will take this to the mailing list
Can cause the RG to be more active
Will get more contributions
Chat Log
Andrew Campling
Some standards favour or accelerate centralisation of infrastructure. Conversely, consolidation at the app layer is often driven by so-called network effects.
Christian Huitema
Andrew, what we see is lack of specific standards impedes competition, such as fighting DDOS, authenticating users to each others, etc.
George Michaelson
we stigmatised regulators as “bad” and forgot why we need them
Geoff Huston
Andrew: Much of today’s efforts have been directed at commoditising common infrastructure so that the value and attention have shifted up the stack to applications. The trend to drive through such legacy infrasdtructure using encrypted tunnels is a graphic illustration of these trends to shift the locus of control up into the applications space.
Stéphane Bortzmeyer
This trend is also a consequence of the interference by the network. With a “hands-off” network, we would not need encryption.
Christian Huitema
Geoff: this is not unlike the rise of the centralized monarchy in France – merchants relying on the king to curb the power of local nobility, thus ensuring transfer of more power to the king
Geoff Huston
And the efforts to regulate the application space have been tough to devise. Its hard to apply the old telephone era regulatory framework, or to apply the industrial framework in the production of widgets when there are no physical widgets any more.
Christian Huitema
(with local telcos in the role of the old nobility.)
Andrew Campling
@Stephane I’m more concerned about the pernicious effect of surveillance capitalism undertaken by apps, together with the continuing loss of resilience due to centralisation and homogeniety of Internet infrastructure.
Geoff Huston
The french analogy is frighteningly accurate in my opinion, including the revolutionary nature of its resolution at the end of the 18th century!
Christian Huitema
But Andrew also has a good point re:surveillance capitalism. Or call that big data. Collecting vast amount of data does increase the power of the big actors. Conversely, curbing that might provide space for competition.
Stéphane Bortzmeyer
I like the analogy, too. So, IETF should be the Jacobins, fighting both the old nobility (telcos) and the king (MetaAlphabetAmazon).?
George Michaelson
Marat/Sade
Danton
Christian Huitema
We know how that ended…
George Michaelson
bad end
Glenn Deen
Classically people thought of applications as sitting on the edge and depending on the open global Internet for connectivity. What people did not expect was that those applications would grow to have their own large >private< networks that bypass the open global internet. The result is that the internet withers and becomes nothing but a local edge access network.
Stéphane Bortzmeyer
Not so bad, we got the republic and the human rights. (And the abolition of slavery, which was restored by Napoleon.)
Christian Huitema
I would see Napoleon as the actual outcome of the French Revolution. And the analogy is, would some kind of Napoleaon replace the current big actors?
George Michaelson
He wasn’t as bad as british historians make out
He did after all, liberate the jews of europe and found a law basis before stuffing up in russia
Dirk Trossen
Centralization is also about reduction of choice: there are ample Bluetooth chat programs (I regularly use one to chat with my wife on planes when we are separated in seating due to unfriendly airline policies) but the problem is that I will need to ensure to have the same choice than anybody else who I may want to talk locally in that plane. Centralization removes that trouble of choice, so one simply uses what most use (which simply does not exist in the Bluetooth chat space). To me, this means that discovery and mediation is a key enabler for decentralization, since it aids choice.
Mallory Knodel
Meant to add to my comments that I liked Lixia’s description of the effects over time and a broader survey of the problem at all the layers (across a vast matrix) should also consider temporal trends.
Simon Hicks
So can we learn anything from telecoms and mobile/cellular standardisation? This has both centrality which we aren’t after here, but does offer a high degree of interoperability. This interop is achieved in standards terms by more detailed standards than we see for the internet. So does making internet standards more detailed offer more flexibility? Is there a sweet spot for internet standards, where more is better, without drowning us in standards?
Vicky Risk
I am ok with ‘regular’ network effects, but the thing that burns me up is the extent to which my usage data is fueling centralization. There has to be some way (eg DMA type regulation) to determine that data derived from user behavior / traffic somehow must be publicly owned.
George Michaelson
l’heure du sang est arrive
Thomas Hardjono
The backdrop for the creation of the packet based TCP/IP was the Cold War.
@Geoff: Do we have an equivalent “force” to break centralization?
George Michaelson
I really don’t think Louis Pouzin or Donald Davies were working to cold war goals Thomas. I agree its a backdrop but packets were theoretically interesting to network researchers outside the DARPA walled garden
Vicky Risk
Not maybe relevant here at the IETF, but global corporate tax reform is an essential goal
Dino Farinacci
@geoff, if they know I want TikTok, they will know I want privacy, they can sell it to me but they think they can get more rev from ad-rev vs customer-rev
Andrew Campling
It would be intersting if RFCs had to include a section on “centralisation considerations” in the same way that they have to cover “security considerations” currently.
Dino Farinacci
turn this on its head, where there will be a cross-over point where they will want our data so bad, they will pay for it
Vicky Risk
Because we are centralizing control by not insisting the giants contribute appropriately to underwriting the rest is society
Dino Farinacci
give me BTC, and I’ll tell you what I am searching for today
George Michaelson
you can have as much tether as you like
Dino Farinacci
@george, sorry not until I know what their stable reserve currency is
Dirk Trossen
@Christian are you suggesting a study on pricing models for the Internet towards a transparent (rather than indirect) pricing to be developed?
Geoff Huston
I don’t see “equivalent” social forces that are able to counter the continuing impetus to fuel these digital behemoths. Its well beyond the component technologies and their underlying properities in centrality or otherwise.
Dino Farinacci
40% of gen-c users are using TikTok for search, the change is coming with the youngest generation
Christian Huitema
I don’t think that paying for data works. People have no idea of the value. Banning the trafficking of data would be much more direct.
Mallory Knodel
It’s a mistake to think the counter measures are obvious or simple. That’s not how we got here and it’s not how we go forward.
Dino Farinacci
Roblox young-users know how to use the Internet of Value
they may be buying virtual clothing for their avatars, but they will be designing such that people want. NFTs might be ridiculous, but look at what art collectors pay for art
Christian Huitema
The good point about banning the trafficking of data is that it can start as a guerilla, by blocking ads, etc.
Robert Wilton
I’m not sure that blockchain is saving anyone from anything
Christian Huitema
blockchain = prrof of waste, and the planet does not need that.
Dino Farinacci
make difficulty = 1 and its not wasteful. And if you want to talk about waste, what about training models with billions of parameters/features, that doesn’t waste?
Simon Hicks
Nobody wants regulation as a concept, but is it part of something that can create balance? Regulation that works globally is awkward. This is not a standards issue as such, but standards that dovetail into regulation is a well-tried concept. IETF does little in that sort of space, but is it something we should consider?
George Michaelson
global standards crossing legal boundaries in regulation go to another SDO
Dirk Trossen
@Christian I asked for transparent pricing, which does not imply direct pricing (e.g., for data) towards users. But it may include the pricing (towards users) for their data (as also Dino remarked) - something like an individual user’s balance sheet, if you will.
George Michaelson
the one we don’t like
maybe we need to think about that
Tommy Jensen
If we view protocols preventing the accumulation of raw data as a de-centralization through cutting off a primary power source of the central monopolies, maybe it makes sense to have “data reduction considerations” rather than a less general “centralization considerations in specs.
George Michaelson
moving from freemium to paid, is not a bad model. people will hate it
Jari Arkko
+1 to the active home proposal
Christian Huitema
+1 Tommy
Shigeya Suzuki
+1
Dirk Trossen
+1 Tommy, indeed!
Andrew Campling
@Tommy Minimising data would be good, but I think that we also need to consider whether new protocols etc encourage centralisation, loss of resilience etc
Dino Farinacci
@tommy, interesting idea, lets brainstorm about how
@andrew, I think we share data with each other, as individuals, as we decide, and dont’ inadvertently share with systems and companies
Tommy Jensen
Andrew, I think data reduction is an implementation of the things you mentioned (centralization is less valuable if bulk users don’t provide big data value), not a competitor.