Hacker News new | past | comments | ask | show | jobs | submit login
The importance of decentralisation (ungleich.ch)
300 points by telmich 22 days ago | hide | past | web | favorite | 134 comments



It's sort of an ongoing hobby of mine to study the forces of decentralization vs centralization.

This article about decentralization does what many other evangelism articles do: talk about the ideals and benefits.

However, I believed what's rarely discussed but more important is the economic forces that prevent decentralization from fulfilling the idealists' vision. Yes, decentralization will be in effect for niche groups but I don't see it becoming mainstream.

To purposely be provocative to spur discussion, I will make a bold claim: Decentralization is an unstable equilibrium. It's the centralization that becomes the stable status quo.

If tpcip protocol and http protocol are already decentralized, why do we have centralized services that have "too much power" such as Facebook/Google/Youtube? It's because different actors can spend more money on their particular http node than other http nodes. Those unequal economic forces is what makes decentralization tend towards centralization. There is no technical protocol specification that can prevent that.

E.g. if Git protocol is decentralized, why is there so much concentration on Github? It's because John doesn't want to install a git server on his laptop and punch a DMZ hole through his home router and leave his laptop up & running 24 hours a day to serve up his git repo. He'd rather spend the weekend playing with his children. And Jane doesn't want to spend $30 on a Raspberry Pi and install Gitlab on it to serve up her git repo. Multiply John and Jane's by a million other devs with their own various reasons for not serving up their git repos in a decentralized manner and the emergent phenomenon you get is something like Github.

See the trend? Centralization is a natural outcome of millions of people not wanting to (1) spend money and (2) not wanting to spend extra time -- to fulfill ideals of decentralization.

I wish we would discuss the above factors much more often and there were more articles about it.


> See the trend? Centralization is a natural outcome of millions of people not wanting to (1) spend money and (2) not wanting to spend extra time -- to fulfill ideals of decentralization.

Exactly. It's not a technological phenomenon, it's an economic one. It's actually present in all markets, but traditionally there is a "local" effect that allows variation to exist (if it's not free to ship goods to anywhere in the world), or a negative return to scale.

Technological goods, especially SaaS, can be "sold" anywhere in the world for near-zero marginal cost and tend to have positive returns to scale. Conventional economics dictates that this will tend to a monopoly. https://www.quora.com/Why-is-increasing-return-to-scale-inco...

Decentralists also tend to to recognise that, for most users of communications services, having a central authority to discipline bad actors results in an improvement to the service. This didn't exist for email so the Spamhaus and related services had to be invented.


> Technological goods, especially SaaS,

I think it's more accurate to call them "digital goods", because IMO what enables the less than near-zero marginal cost (I consider them negligible costs) is the digital nature of the "merchandise"

(a jet engine is a technological good, but it is material, not digital)

other than that, I completely agree. The trend towards centralization comes from the economics of the market (not from the nature of technology)


Regarding your point about spam and bad actors: even though it's not the solution we've ended up with (i.e. training spam models, creating rulesets, sharing blacklists etc), this problem space always makes me think of Microsoft's 'Penny Black':

https://en.wikipedia.org/wiki/Penny_Black_(research_project)

There are further economic problem there; people won't want to pay anything for email, for example, now that they're accustomed to having it for free.

But if sending a message came at the cost of a few smartphone GPU cycles -- or some other widely accessible but difficult-to-hoard resource -- and as long as it didn't affect a person's mental perception of their bank balance -- perhaps there'd be opportunities for something like Penny Black again.


Funny, I like the efficiency added by the ticket server, but this solution is actually more centralised the one that came before it -- hashcash: https://en.m.wikipedia.org/wiki/Hashcash

Hashcash has the features you described above, but no mechanism for "returning" PoW to a user.


Nature and Biology gives us a very strong counterexample to this. If centralization was the stable form you would expect to see an overwhelming preponderance of it. Instead you will almost always see a blend of centralized and decentralized forms, with a bias towards decentralization wherever it's a good fit. The vascular system is centralized; the nervous system is mostly federated; most of the rest is highly decentralized. A counter argument is that intelligent design changes the relative value of these different forms but I'm not sure whether there's enough evidence to support this.

My theory on the matter is that the trend towards centralized forms in our current society has a lot to do with Conway's Law (organizations which design systems ... are constrained to produce designs which are copies of the communication structures of these organizations.)

We live in a society that currently has a very hierarchal, centralized and structured systems of power. This means that the subsystems (eg Corporations) and the artifacts of those systems are subtly pushed towards a similar form.


Completely agree.

I think the efficiencies and benefits to centralization OP pointed out are real, but that there are practical benefits to decentralization as well.

Decentralized solutions are not saddled with the same need to appease everyone at once. You can have local solutions extremely well adapted for local situations that know more details about the context than a large company trying to deal with everyone ever could.

I’d argue centralized solutions are only as stable as the problem is uniform. If the problem differs even slightly between actors looking for solutions, those actors are either settling for a less than perfect fit when adopting a centralized solution, or the centralized solution is complex enough to handle those differences. As differences increase, the tradeoffs become unacceptable and/or the complexity becomes too much to handle, and the centralized solution falls apart.

I don’t think a lot of centralized solutions are currently centralized because they’re efficient solutions to uniform problems; I don’t think most everyday, practical problems are all that uniform when you take everything into account (ex: supply availability for physical goods, UI expectations/preferences for digital stuff) although again, there can be/often is at least some gains in efficiency. I think it has more to do with the organizational forces you describe.


> I’d argue centralized solutions are only as stable as the problem is uniform. If the problem differs even slightly between actors looking for solutions, those actors are either settling for a less than perfect fit when adopting a centralized solution, or the centralized solution is complex enough to handle those differences. As differences increase, the tradeoffs become unacceptable and/or the complexity becomes too much to handle, and the centralized solution falls apart.

100%

I've been thinking of how this applies to our world today. For example, it is in Google's interest to train us to only use a subset of queries and words so their centralized solution can work (same with YouTube and any recommendation engine probably).

Predicting or prodding you towards certain items (ads), pages, suggestions and the like - intentionally or not - only serve to reinforce a society of increased uniformity and thus, increased desirability of the centralized system.

Newsfeeds suffer the same problem as does even autocomplete I think.

It scares me about our future sometimes because products will slowly become more hostile to you the less you fit into a centralized model of their ideal consumer


Arguably, the combination of centralized and decentralized parts ace much closer description to what the internet is today than the idea that it's solely centralized. Same is true of society; while the 20th and 21st century are replete with efforts to completely centralize society under one organization (there is the Party, only the Party, and nothing outside of the Party), such structures have consistently failed and collapsed. (Which does not stop people from continuing to propose them seriously, unfortunately.) Society is a mixture of centralized and decentralized elements.

The problem is that in biology, you don't have intelligent entities exploiting the centralization, which is I think where the metaphor breaks down.

Interestingly, the metaphor isn't completely broken; I have seen serious discussion of thing like cancer, and even more interestingly, what can convince a cell to not be cancer, in terms of self-interest and exploitation and such. One of the mysteries of how multi-cellular life ever got started was how one cell convinces another to not seek its own survival at all costs. Being rather complicated multicellular entities ourselves, we take this for granted, but it's a legit research topic.

Still, the cancer abusing the system is not doing it with human-level intelligence, whereas arguably Facebook is a super-human level intelligence, so the metaphor is not great.


This.

The influence of Conway’s law is underappreciated. We are all trained to see hierarchies as the only principle for building structure, it’s a dominant normative ideology. Once we see how to build decentralized systems, we can discover they are often more resilient and more effective. But we have to overcome the current centralization of power and our own tendency to structure the world in hierarchies.


I would argue that there's an incredibly powerful force that drives the consolidation and centralization of systems in modern society: the wholly artificial legal concept of property. We are given a legal right to accumulate control over as large a swathe of material goods and organizational structures as we can within the confines of the law, and this ownership is backed by the promise of state violence should it be violated.

The concept of property, and the fact that our laws and government are built around protecting it, is the sole mechanism through which massive disparities in power are achieved and maintained in society. It centralizes wealth through avarice, making our stomachs unlimited in size. It also creates the dilemma of externalities, as we differentiate the value to us of the things we "own" versus the things we hold in common with others or that are "owned" by the state.

Property is also a entirely artificial concept. It takes the practical concept of use and possession and extends upon it to completely absurd extremes. It ties our society up in knots wasting a massive amount of human energy in the activity of determining and enforcing property ownership, and leads to incredibly convoluted structures in laws and markets, often leading to huge structural failures.

For instance, intellectual property law is completely absurd, on every level. It is an effort to force something that is fundamentally free and infinitely available to everyone to become scarce exclusively so that we can properly ensure it fits into a property-based system. The amount we lose as a society from the fact that all software is not simply open source is incalculable.


>wholly artificial legal concept of property

Property is an informational concept more than a legal one. The concept of "mine" is present at a very young age, and clearly present in various animal groups. The data-structure is a simple association of objects with an owner, and for the substrate animals use only their brains, but humans use paper and, increasingly, bits. Banks and local government agencies keep the most official of these records, but informal records count too. (My son may think my daughter's blanket is his, but she will protest loudly and vehemently, and there is not really a formal reason for her claim).

concept of property...is the sole mechanism through which massive disparities in power are achieved and maintained

Property is a necessary, but insufficient condition for inequality. Property is how you keep score; the reason some players win far more than others is not entirely known, but I'm pretty sure it has more to do with personality, talent, and luck than the jurisprudence of property rights.

It's not clear what a society without property would be like, BTW.


The "natural" concept of property is much more narrow than our abstract notions. Animals, and human societies before they develop the abstract property theory, generally define property through possession or use - your house is yours because you live in it, that sort of thing. You can't establish an association simply by saying, "this is mine". This puts a natural limit on how much property can be accumulated.

Once you have abstract property as a concept, it's all about tracking the association in and of itself, separated from any physical manifestations of it (possession and use). Your house is yours not because you live in it, but because there's a title on it in your name, and a record in the registry. This makes it possible for one person to have ten houses, live in one, and rent the rest out to those who don't have any. If they accumulate the rent, they can eventually buy more houses, and it becomes a positive feedback loop.

But there's no particular reason why any given society has to adopt such a system of property, and not all societies did or do today - especially with respect to land and other immovables. This can often be seen in the colonial history, where the colonizers expected to be able to buy and sell things that the indigenous population didn't even consider property, or at least the kind of property that could be bought and sold. More often than not, the colonizing power would then declare that in the absence of a "proper" abstract property claim on e.g. the land, it's there for the taking of anybody who is willing to make one. And that is how we ended up in the present situation, where that system of abstract property rights is mostly enforced throughout the globe - but I don't see how it follows that it's a system that is inevitable for any human society.


"But do you have a flag?"


> Property is an informational concept more than a legal one. The concept of "mine" is present at a very young age, and clearly present in various animal groups.

All money and contracts indicating ownership of property mean nothing without either the consent of all involved or the backing of violence (mind you, our system very much relies on violence, we have people who are empowered to enact violence to enforce property law).

There's also a huge distinction between possession (physically having and using something) and owning it (having the enforceable legal right to control how it is used).

And while this concept of control/territoriality might be naturally present in human psychological development, so are things like trying to use violence to resolve disputes and other asocial aspects of human psychology that we actively discourage. The idea that we should use property as the basis for how society functions is something we seldom interrogate regarding whether it creates the kind of world we actually want.

> It's not clear what a society without property would be like, BTW.

Try reading The Dispossessed by Ursula K. LeGuin. It really opened up my mind regarding the way our society is shaped by the idea of ownership.


The concept of property is highly important in terms of giving us a framework by which we can resolve conflicts over who has legitimate use or possession of things. Without the idea of property, you devolve into a might makes right scenario where the strong or sneaky can steal from anyone without recourse. That doesn't sounds like a society anyone would feel motivated or happy living in. Why try when someone can just steal away your hard work? It's why countries or communities that have gone down the road to socialism and communism have always invariably collapsed. Every anti-property proponent I have ever asked about where the vague lines of what counts as legitimate "use" or "possession" has never been able to provide a clear and consistent answer. Where does my use or possession end? For example, I trade two sheep for a bicycle which I need for transportation to get to my job. Can someone come along and take my bicycle any time I'm not using it? If so, how is this theft justified?

As a side note I'd like to add that I agree with your assessment of intellectual property. The arguments defending IP law are indeed weak, but when it comes to real things like money, goods, and land, property rights are a necessity to keep humanity from devolving to barbarism.


> Where does my use or possession end? For example, I trade two sheep for a bicycle which I need for transportation to get to my job. Can someone come along and take my bicycle any time I'm not using it? If so, how is this theft justified?

You either work out an agreement with the people who might take it that a bicycle be available for you after work or you just go and find another bicycle. It's funny you say this because bicycle and scooter sharing systems have become incredibly popular, because it's far more efficient than everyone having to keep their own vehicle for the long amount of time that they're not being used. Imagine if we didn't need to have parking for storing so many vehicles. The fact that you need a vehicle to be in a certain place at a certain time is a logistical issue that needs to be solved, but owning it as property that everyone else is excluded from using is not nearly the only solution.

If we structured society such that the accumulation of material wealth was infeasible, there would be far less need to worry about people going around and taking things they didn't legitimately need.

> It's why countries or communities that have gone down the road to socialism and communism have always invariably collapsed.

I'd argue this statement is completely unsupported. There has been no form of communism that hasn't been embattled by attacks from established capitalist systems at their outset, causing them to either consolidate into authoritarian centralized regimes in order to survive or to get wiped out in their infancy, and this centralization is what leads to failure. Socialist nations who have transitioned smoothly from capitalism have done so very successfully (take Norway). This may say bad things about the ability for a property-less system to survive in a capitalism-dominated system, but it doesn't say whether it would be capable of staying stable and successful in a less hostile environment.


> If we structured society such that the accumulation of material wealth was infeasible, there would be far less need to worry about people going around and taking things they didn't legitimately need.

Is there anything to stop people from just moving into and taking over my house when I go out for the day?

Also, I like having a car set up exactly how I like it, with music I like, the mirrors where I left them, and also with a few things in it that I might need even rarely.

Call me selfish but I don't want to deal with a dirty public car that may or may not be available when I want it. All other public things are dirty and unpleasant, I don't see why shared cars or bicycles would be any different.


> Socialist nations who have transitioned smoothly from capitalism have done so very successfully (take Norway).

Do you have a source for that? I'm pretty certain Norway (and other Scandinavian countries) are not socialist and more of a social democracy, where they simply provide a large social safety net through heavy taxation.

I believe they still very much are in favor of a free market economy.


The sovereign wealth fund, which is the real champion of Norway's current situation, is the result of nationalizing their natural oil reserves. Effectively, making the natural resources of Norway the property of all the Norwegian people.


Yes, exactly. Is the human body centralized or decentralized? The ego could be said to be the central "pivot" of the system, but then you have the phenomenon of "multiple personalities", and so on.

Once you start to think of biology as the study of billion-year-old self-optimizing nanotechnology it makes sense to look to life for models for designing the structure of our own artificial systems. There's the question of what do we want to be like? A starfish? A meta-human?

A decentralized system that depends on each or most of its elements to function is a single "point" of failure for the whole. (E.g., a relatively small wound can cause death by exsanguination.) On the other hand, consider the radical indignities that the lowly planarian will endure only to placidly regrow a tail or a head or whatever. (See "What Bodies Think About: Bioelectric Computation Outside the Nervous System" https://www.youtube.com/watch?v=RjD1aLm4Thg https://news.ycombinator.com/item?id=18736698 Which, incidentally, adds a whole layer of complexity and nuance to the concept of intelligent design, since the whole of Nature has now been scientifically shown to think...)

It remains to be seen whether Humans + Internet = Self-Organization or along what lines. (What's the news out of Hong Kong this week? Are the distributed, self-organizing people of HK prevailing against the centralized imperialistic communist state?) Governments everywhere are waking up and smelling the data coffee. The major players already have databases on the bulk of world population and are collecting (exponentially) more every day. What kind of cyborg-state shall we make?


> Are the distributed, self-organizing people of HK prevailing against the centralized imperialistic communist state?

I think you make a good (and very underappreciated) point about the relevance of the ideas in this article to geopolitics, or how humanity ~chooses (or so we think) to organize itself.

[didericis] also makes some good points that are relevant to this perspective:

> Decentralized solutions are not saddled with the same need to appease everyone at once. You can have local solutions extremely well adapted for local situations that know more details about the context than a large company trying to deal with everyone ever could.

> I’d argue centralized solutions are only as stable as the problem is uniform. If the problem differs even slightly between actors looking for solutions, those actors are either settling for a less than perfect fit when adopting a centralized solution, or the centralized solution is complex enough to handle those differences. As differences increase, the tradeoffs become unacceptable and/or the complexity becomes too much to handle, and the centralized solution falls apart.

The dangerous part imho when it comes to a geopolitical application is, while most of us can recognize the complexities and tradeoffs in a computer system, at least when they're pointed out to us, very few seem to be able to see similar complexities in geopolitics (a massively more complex system), even when it's pointed out. It's as if we've developed an intuition that passionately rejects the very suggestion, a behavior which in itself highlights the complexity involved.


> The dangerous part imho when it comes to a geopolitical application is, while most of us can recognize the complexities and tradeoffs in a computer system, at least when they're pointed out to us, very few seem to be able to see similar complexities in geopolitics (a massively more complex system), even when it's pointed out. It's as if we've developed an intuition that passionately rejects the very suggestion, a behavior which in itself highlights the complexity involved.

Well, one of the fascinating things that anthropologists point out is that every society succeeds by-and-large in getting the children to behave like (that society's version of) adults. In other words, culture is somehow stable even though it gets renewed from generation to generation. So we are programmed in a way, and for that to work as well as it does, parts of the program have to remain stable, and that implies that the conscious ego cannot be allowed direct control over all parts of the internal model. And, in fact, there are only a few autonomous functions that are also (partially) subject to conscious control: breathing, blinking, and a third one that I just forgot... heh

In other words, the rigidity you're talking about is "works as intended" for the original context: small tribal groups living in the wild. What we call civilization is only about 12K years old, eh? With most of the action happening just in the last two or three centuries the complexities in geopolitics are entirely novel.

However, since about the 60's information about reprogramming ourselves has become more and more available (Just for example, "Programming & Metaprogramming in the Human Biocomputer: Theory & Experiments" by John C. Lilly was published in '67 I believe.) Rapid progress has been made, but a concerted effort to apply this information as part of an overall political shift has not arisen.

Bucky Fuller pointed out that we have the technological capability to create a kind of utopia, from my POV it seems we also have the psychological "technology" to get over our problems and actually do it.


Absolutely! I think a huge part of the problem is that we're saddled with all sorts of baggage from the structures of the past that constrain the ways we think and approach problems.

We have very rigid notions of the ways that laws should work, or property ownership.


That which is optimal for one state is not necessarily optimal for all states, but some powerful people wouldn't benefit from a reoptimization, so it won't happen.


> Well, one of the fascinating things that anthropologists point out is that every society succeeds by-and-large in getting the children to behave like (that society's version of) adults. In other words, culture is somehow stable even though it gets renewed from generation to generation. So we are programmed in a way, and for that to work as well as it does, parts of the program have to remain stable, and that implies that the conscious ego cannot be allowed direct control over all parts of the internal model.

Undoubtedly we are programmed to some degree, but is there anyone that claims to know with any sort of precision how much we are programmed, and to what degree that contributes to our success? People like Sam Harris "point out" that rationality should be all we need to continue succeeding into the future as we have in the past, but the problem is, this is largely speculation.

> and for that to work as well as it does, parts of the program have to remain stable, and that implies that the conscious ego cannot be allowed direct control over all parts of the internal model.

It may be beneficial and easier for things to be stable, but how stable do they need to be? The environment (society, politics, technology, etc) that humans live in is certainly not stable, and we continue to be successful, but how can we accurately attribute this result to programming of some sort? Take away our technology for example and cast modern day mankind back into the past a decade or three (say, due to an extremely severe recession/depression) - could we handle it psychologically and societally? To what degree is modern day society psychologically dependent upon the comforts that scientists have delivered to us? How do we proclaim to know such things?

> And, in fact, there are only a few autonomous functions that are also (partially) subject to conscious control: breathing, blinking, and a third one that I just forgot... heh

Getting a short haircut, catching a flight to a far away land, and trying to kill as many of the people your government has told you are your enemy? Working two jobs and sparing yourself even the smallest comforts of life so your children can have a chance at a better life than you, instead of blowing your brains out which would be so much easier? Biting your tongue for the hundredth time when your boss pins another case of his incompetence on you? I speculate that all of these things involve both conscious and unconscious functions, but again we don't know. Some people can pull them off, others can't.

> In other words, the rigidity you're talking about is "works as intended" for the original context: small tribal groups living in the wild. What we call civilization is only about 12K years old, eh? With most of the action happening just in the last two or three centuries the complexities in geopolitics are entirely novel.

Exactly. Might it be foolhardy to rush headlong at maximum possible speed into financially optimizing every last thing we can, with zero thought for whether we are psychologically operating in the shiny new world we are building? Are there any signs of rapidly increasing strange behavior in societies? It seems like we're often told to worry very much about such events, but the remedy is even more change, and faster!

> Rapid progress has been made, but a concerted effort to apply this information as part of an overall political shift has not arisen. Bucky Fuller pointed out that we have the technological capability to create a kind of utopia, from my POV it seems we also have the psychological "technology" to get over our problems and actually do it.

Boom. Well, I would say it has arisen many times, and to varying degrees, but the people who have been appointed to run our societies have tended to not be terribly supportive of such movements, wish makes me feel even more uncomfortable.

I too believe we easily have what it takes to run our societies fairly and properly, and have for quite some time....or at least we used to - I'm start to worry that we might be starting to lose some of the psychological and other less intangible prerequisities to make such a world happen, and I don't share Sam Harris' optimism that rationalism is enough to carry the day. It seems to me there is far more evidence against that theory than supporting it.

On the bright side, it's nice that you and I mostly agree on something for a change! :)


Cheers, well met. :)


>Nature and Biology gives us a very strong counterexample to this. If centralization was the stable form

The context of my claim was about computer information systems becoming centralized. Yes, I understand we do have decentralized washer & dryer machines at home instead of everyone going to centralized laundromat. And building residential houses is still a very local business that resists a hypothetical giant national corporation that builds every house in America. Even Jeff Bezos' Amazon can't conquer 1000 different local building codes and construction techniques to build homes. However, the realm of information technology seems to behave differently.

I like to use the decentralized-Git-to-centralized-Github example because:

- programmers are the ubergeeks that would conceivably be the vanguard of decentralization because they are already predisposed to understand its benefits

- programmers (unlike mass consumers) have the technical know-how to implement their own git server if they wanted to and overcome any challenges of ISP NAT router.

And yet, the vast majority just think: "Hmmm... there's a free service that can host my repo which means I don't have to spend _time_ or _money_ to mess with any of that home setup?!? Sign me up!!!"

There is no "evil capitalist monopoly" that forced Git centralization on to resistant programmers. Instead, the programmers willingly put their code on Github because of the benefits. The economic forces overwhelm the ideals of the decentralized protocol.

As mentioned before, a git distributed protocol can't stop an actor from spending more money on that protocol than others. Github Inc spent extra money on Ruby programmers to create extra features such as issues tracking, web landing page, hosting disk space, reliable backups, DDOS protection, etc. Unequal spending by actors makes one node with better features more desirable than other nodes (such as a personal home node) and this starts a feedback loop of centralization.

To restate the claim earlier: decentralized computer information systems that need to share data is an unstable equilibrium because it depends on having free-will actors not spend extra money on a protocol to gain an advantage which is also an unstable equilibrium.

Even if this thread's article's vision of IPv6 improving direct connectivity were to happen, I still claim the overwhelming majority of programmers would continue to choose something like Github rather than host their repositories on personal home servers. Personally, I wouldn't want my repo to be traced back to my home's ip address. I don't want China DDOS'ing my home ISP connection if they don't like my iPhone app that monitors Hong Kong. If I was a female, I wouldn't want my home git repo to give away my ip address and invite digital stalkers. I'd use Github just for the ability to shield my IPv6 address.

If the programmers -- which we can think of as technical thought leaders -- aren't leading by example (with git) for decentralization, why would we realistically expect other mainstream consumers to adopt decentralized setups? Yes, there will always be decentralized communities but it will always remain niche.


> There is no "evil capitalist" that forced Git centralization on to resistant programmers. Instead, the programmers willingly put their code on Github because of the benefits. The economic forces overwhelm the ideals of the decentralized protocol.

There is no "decentralized protocol" in git. Changes to history are local. You then convince a remote to accept your development history as truth. I feel that you must miss the point of git's distributed and decentralized nature to say that a popular service like GitHub is detrimental to its ideals.

The sense in which git is decentralized is that in order to develop, you need a a full copy of the development history up until the point at which you want to contribute. You then create a history that is unique to you and maybe you push this to a remote at some point. Insofar that your history is shared, there is no difference between the repository at a remote (or an archive etc.) and the repository you and other developers have. Insofar that your history isn't shared, the workflow is entirely decentralized, enabling a distributed development process. Git isn't decentralized to satisfy some lofty ideal, but because this enables a convenient workflow.

A side effect of this is of course that GitHub going down forever isn't the end of the world in terms of git, because anyone interested in the code probably has an somewhat up-to-date copy of the development history that they can push to another remote or continue working on for their own intents uninterrupted by such an event. But that's a side effect IMO and not the most valuable sense in which git is decentralized. The loss will primarily be in git non-goals like discoverability, issue tracking and facilitation of the pull request workflow that GitHub offers. As far as I'm concerned, those are the only things I'm placing bets on that may result in a loss to me if GitHub dies.


I made/make no claim about the relative moral merits of capitalism or a hierarchal society at all. However you only need to look closely at your stated claim to see why Conway's Law is relevant here: your claim about "technology" rely on our concept of money / funding staying the same. Given that our concept of money has been evolving non-stop over the past centuries it seems unlikely that it won't further evolve.

I'm not saying that your claim is wrong, rather the opposite I believe it's holds true in the present. Computer Information Systems are reliant on larger economic forces. In many cases centralized systems are not better on a purely technical basis, but rather are better at attracting capital. However there are undoubtably systems of human organization and resource management that would change the fundamentals assumptions behind this.


>I made/make no claim about the relative moral merits of capitalism

Sorry for the misunderstanding. My reference to capitalists was a general point and not related to anything specific you wrote. Others in the thread mention capitalists so I just added it to my reply for color.

>However there are undoubtably systems of human organization and resource management that would change the fundamentals assumptions behind this.

I appreciate you bringing up Conway's Law as an explanation but I'm not fully convinced. I think there's an even more fundamental force than any artificial human organization such as Conway: if a bunch of humans don't want to do something , it creates an opening in the market for another group to do it instead. If that "something" is information technology related, it tends to create centralization.

A bunch of us don't want to set up a personal git server (for whatever reasons), and therefore, it creates an opening in the marketplace for Github to exist. At this primal level of desires, I don't think we have to bring in Conway's Law.


No worries! I appreciated the opportunity to clarify my thoughts. Yeah, it doesn't help that my theory is a tough one to validate, since by definition the conditions don't exist to experimentally test it, so you are left with mostly thought experiments. One thing I think is in favor of your argument, decentralized technologies do seem inherently harder to reason about.


While decentralized protocols don't entirely prevent centralization, they're still better than if the protocol was centralized. While Facebook and Google exist, we can still make websites. While most people use Gmail, we can still run our own mail servers.

Even if 90% of users use the centralized service, the protocol being open gives enormous value to the remaining 10%. And as long as the 10% keep the protocol open, the _possibility_ of decentralization can keep the centralized services honest.

We only need to keep the googles of the world from completely shutting down the open protocols. I mean, even that can get hard, but it's still entirely reasonable goall.


>as long as the 10% keep the protocol open, the _possibility_ of decentralization can keep the centralized services honest.

This is the avenue we should be traveling. This is what makes the internet a better place.


Yup this is my thinking as well. Also this opens up for smaller niche communities to entirely setup whatever they want. I.e, their own social media, their own youtube, etc. Small rural communities can setup their own email server, chat, without dependencies on Slack or Google.


Dropping into the conversation late, to say that this is right - but - the danger is that Google controls the defacto browser and defacto web standards as well as search, and yes a large chunk of email, the biggest video hosting service and mobile operating system! So in reality I think the open standards will erode in practical terms to be Google standards over time, unless something is actively done to prevent that.


>Even if 90% of users use the centralized service, the protocol being open gives enormous value to the remaining 10%

It's more than that. The 90% benefit greatly from the openess too. All the @gmail.com can talk to @yahoo.com or @hotmail.com or @whatever.com. Try to get your whatsapp to talk to my telegram app or some other chat app.

It's a pain to don't have whatsapp and only a small inconvenience for whatsapp-users to have to deal with somebody like me. But it's my way to point out what is broken about chat apps.


That trend toward centralization and aggregation is mirrored in wealth, influence, and matter. Barring outside intervention, returns on capital will outpace returns on industry, businesses will monopolize, and water will flow downhill.

Because it is so natural, those who would oppose these forces are motivated to do unnatural things like blog, protest, legislate, regulate... Governments redistribute wealth, hold antitrust hearings, and people write blog posts like this one. I agree with you, unless we find a way to fundamentally alter incentives, effort spent toward decentralizing is effort spent swimming upstream. Of course even in streams that eventually end up in the sea, there's a lot of activity in eddies.


> Centralization is a natural outcome of millions of people not wanting to (1) spend money and (2) not wanting to spend extra time -- to fulfill ideals of decentralization.

Ironically enough, people do not live year-round in centralized hotels with cleaning services provided. They live in their own decentralized homes, and make an economic decision whether to take care of their cleaning needs by themselves, or to outsource and hire somebody to clean their domicile for them.

There seems to be an inherent assumption that running your own services necessitates taking care of maintenance yourself. I think this is a mistake - we just haven't been able to develop yet a model where we own decentralized installations and outsource the maintenance, even though in many other instances, this is the case - not just cleaning and houses, but endpoints and OS updates.


"Ironically enough, people do not live year-round in centralized hotels with cleaning services provided. They live in their own decentralized homes, and make an economic decision whether to take care of their cleaning needs by themselves, or to outsource and hire somebody to clean their domicile for them."

I think that this is mostly to do with the fact that you can't simply take advantage of economies of scale and centralize ever one into a single housing block the way you can with a lot of web-based businesses. Despite this, a lot of people do live in apartments complexes with centralized maintenance and utilities. Moving to a house is about gaining space than independence, I think.


This isn't a very good comparison though for looking at technology. The problem of washing dishes, cleaning a kitchen, vacuuming, making a bed etc. Is O(n) in _economic_ complexity. Managing data and technology services is somewhere around O(log(n)). Managing 100 servers is much more difficult than managing 1, but managing 200 is not much more difficult than 100.


It belies the point though that people at large prefer decentralization, just not the costs involved, until a model comes along which reduces the costs to a level which people can afford.


> There seems to be an inherent assumption that running your own services necessitates taking care of maintenance yourself.

An assumption that should be challenged. You can run your own chat server, with end-to-end encryption, thanks to the work being done by Matrix. Check out https://modular.im (I use them and am quite happy with it.)


Love the challenge. Let me start by replying to your DMZ comment: with IPv6, you actually have to rethink and with it you can organise the Internet very different from today.

I agree with you that people don't want to maintain complex systems. But then again, running a git server or even cgit, is a matter of little minutes spent, especially if you use ready to deploy toolchains.

I also think you have a point in terms of equilibrium - people are in general more comfortable with something they know. And due to the long history of IPv4 shortage, people are used to having private IPv4 space only.

It is like animals in a zoo - we don't know anymore, how good the life used to be with public IPv4 addresses, we have grown up without them.

What I have seen from many discussions and visits in the last months is that IT startups are now reconsidering and embracing IPv6, because it gives freedom.

I'd also say the approach is not to directly replace facebook & co., but to think completely different. Why do I need a website to write to you? I can write directly to you, if both of us are on an IPv6 network.

Thinking different is the key for decentralisation.


You're missing the point. Even if running a git server is easy, not everyone will bother to maintain it and setup a regular backup process. Also, where do you back that data up to? Some central server or DHT/BitTorrent. Inherently, AWS is popular because companies would rather pay to have someone deal with the headache of managing hardware and infrastructure.

> Why do I need a website to write to you? I can write directly to you, if both of us are on an IPv6 network.

Because I'm not always online and you might not be when I'm.


> Because I'm not always online and you might not be when I'm.

That's a good thing, because maybe I don't want to be able to reached all the time. If you want to, you can add your IPv6 address on your phone, which is virtually always online (I actually do this on my phone).


>If you want to, you can add your IPv6 address on your phone, which is virtually always online (I actually do this on my phone).

Is there a guide on how to do this? How do you manage it? I'd love to self host from a computer just sitting in my house. Way easier than something like a VPS.


>you can add your IPv6 address on your phone

How does one achieve this while a phone is roaming between cell towers and various wifi access points?


With a wireguard based IPv6 VPN for example. It's exactly what you can get on IPv6VPN.ch


Re IPv6, I don’t expect connectivity to suddenly appear. The NATs will be gone, but I bet you that every single home router will get a firewall with a default “block incoming” policy. A new network protocol does not make all the security bugs disappear overnight.

And for users, it does not really matter if they have to visit router admin page to set up NAT port forwarding or open up the port in firewall.


This.

Setup a raspberry pi with an IPTable rule to log and black hole all incoming traffic, then open it up entirely to the internet through your router.

I did this and, after about 48 hours, had a nice long log of bots hammering the server scanning for vulnerabilities on open ports.

I imagine non-technical users (the ones at a high risk of getting malware just browsing online) and wonder if they are truly better off if we open up their computers to the world without first rethinking the security model.


> And for users, it does not really matter if they have to visit router admin page to set up NAT port forwarding or open up the port in firewall.

It matters when port forwarding is impossible because the user lacks a unique IPv4 address. Earth has 7 billion people and 4 billion IPv4 addresses, so carrier grade NAT is inevitable.


> To purposely be provocative to spur discussion I will make a bold claim: Decentralization is an unstable equilibrium. It's the centralization that becomes the stable status quo.

Spurred. Firstly, this ignores the variance between absolute decentralization and absolute centralization. By reducing the discussion to black and white, we can't recognize a middleground that may not be as decentralized as all-in-home yet not so centralized as owned by a few. I'd argue that much of the internet remains in this middleground, popularity notwithstanding. Secondly, what is and isn't mainstream is the result of ease of use, implementation quality, and improvements over alternatives. Popularity and other knock-on effects may follow, but needn't be the initial goals. That these aren't achieved in decentralized setups is not evidence that they never will. When they are, and the economic forces then come along with adoption (that's the usual order after all), you'll see a shift in focus without necessarily a shift towards centralization. Until then we accept that centralized versions of more popular services are better implemented currently.

One thing everyone agrees on, the vast majority of users don't care (nor should they).


I would say if you assume that markets don't regulate themselves to the best, the argument that users should not care, is not valid.

Or put it the other way: if users don't care, they will be exposed to changes / forces they don't control nor want.

I think this is very similar to political activity: if you don't vote, your life will be more influenced by the opinions of others.

If you vote (care), then you can make a change


> the argument that users should not care, is not valid

I agree from the user's perspective. The point is nuanced but important - users should care, but the reality is they don't and from an implementer's perspective we should neither expect nor require them to. So to rephrase, I mean "users should not have to care in order to use the product".


Got your point - thanks for clarification!


To summarize, centralization allows certain efficiencies that are not permitted in decentralized systems, and those efficiencies have value. Of course, decentralization itself also has value, so the question in any given circumstance is, "what balance between efficiency and decentralized robustness do I need?"


The centralization-decentralization cycle is about efficiency vs robustness trade-off at a system level.

When we have a decentralized system that feels safe, we centralize it to increase efficiency. Eventually we have a disaster that convinces us that decentralization was worth it and we re-create it. Then eventually we grow complacent and repeat the process.

When I design a decentralized system now, I just assume it can only live a decade or two before it gets killed and I need to do it again under new branding.


I've been warning people of a certain 1T$ tech company that abuses customers and developers under the guise of centralization.

Our worst fears have become true. Banning apps, anti consumer technology, anti competitive practices that effect non users.


Abuse is another one.

Note how you almost never see spam on GitHub. And you can effortlessly cross-reference issues across repos and notify other users by mentioning them. That works, because global view across all repos makes spotting and blocking abuse easier, and moderation is possible thanks to the business side subsidizing it.

Contrast that with WordPress trackbacks. They're all spam. And blog comments are full of spam too, because individual operators don't have know-how to fight them, other than outsourcing spam filtering to… a centralized service.


Maybe we don't really need those "trackbacks" and unmoderated comments from random anonymous users.

Maybe decentralized services should function strictly on a web-of-trust basis. In the end, if you have something valuable to contribute in a given domain, you have either joined a relevant organization or have contact with people in the field.

Maybe collaboration between complete strangers with unknown reputation, is not a good idea and explains why it doesn't function like that in the real world?


I see this as a problem with trust. Decentralized systems require less trust for them to work (think consensus mechanisms) but ironically users today don't trust the tech because it is still early IMO, centralized systems need more trust. People these days are using centralized systems because they trust the centralized system (FB, Google) more then the decentralized tech.


I agree and think the solution would be to go back to a web-of-trust scheme where identities are transitively validated and reputation used to ensure well-behaved operation.


>See the trend? Centralization is a natural outcome of millions of people not

Agree

>This article about decentralization does what many other evangelism articles do: talk about the ideals and benefits.

Exactly! This week I tweeted this old clip of Steve Jobs praising decentralization https://twitter.com/maramesque/status/1194136650500255744?s=... then it hit me how Steve worked so hard to make Apple centralized.

Sure, Apple was great during his lifetime, but what happened to this very centralized company after Jobs passed away?

Did Apple live up to Jobs legacy and promises from protecting users privacy to building great products?

I think we will always live in a world full of centralized products used by the mainstream.

The mainstream will eventually get smarter and founders will always face the pressure to improve products, respect users privacy and attention.

The days of launching an app from your college and watching people downloading it from all over the world are long gone.


That trend seems real, yep, and your points ring true to me. In response to the items you mention (in reverse order), decentralization could become rapidly more attractive if & when:

2) The time required to configure, install, and use decentralized services reduces to within a small margin of the equivalent centralized systems

1) The cost of losing custody of your personal data, and the cost of being influenced and constrained within a proprietary system are more apparent and can be stacked onto the 'free' price tag

Work towards this takes a long time, especially while software engineers are in such demand in the profitable private sector (which, although there are exceptions, tends to skew towards building proprietary software).

In general I'd tend to believe that another, second trend is that decentralized and open source software does always eventually catch up (in a kind of asymptotic progress curve to infinity) - and that we'll reach this 'switch margin' eventually, it's just a question of how long it will take.


sort of an ongoing hobby of mine to study the forces of decentralization vs centralization. - Reading list/blogrolls please :)

Idk is one state is necessarily more equalibrium-ish, in general. I tend to think it's more of a khaldun-esqu cycle. Decentralisation is often stronger at creativeness & flexibility, so is necessary for innovation. It's good at finding solutions to unarticulated problems. There were multiple centralised/proprietary attempts at inventing the web, but I don't think anything but a decentralised www could have become what it did.

Ceentralised systems have their own strengths.

I agree that Facebook is easier than html, and that the easiest option will win. The interesting (imo) question is "why?"

One reason is undoubtedly economic. There was/is massive incentive to centralize & own chunks of the web. Decentralised www doesn't have that.

Another reason is (imo) related to the "OSS UI" problem. When programming problems are specific and legible (add multiple language support, fix crashing bugs, etc?), OSS works really well. When the problem is "create a fun and intuitive UI" OSS can really suck. IE, Facebook beats www the way osx beats Linux.

Lastly, what level of centralised or decentralised are we looking at.

The www's uses (sites, users, code...) is decentralised but the protocols, browsers, DNS and such are very centralised.

If you want to invent a way of sending a proprietary way of sending a new type of online wink, go ahead. Let a thousand flowers bloom. Getting something new into the official protocols OTOH..... near impossible.

So... In some senses, Facebook was/is the nimble decentralised actor. One of many. The www (the protocols) is/was the monolithic sloth.


I don't think centralization automatically becomes the status quo.

The reason why we have the current centralization of amazon, facebook, github etc is because the web as we are currently using it has centralization built in: A browser always connect to a fixed server. But that is not god-given, we created it that way 25 years ago, and we could allow additional other ways. A browser could instead also be connecting to a generalized service for selling, communicating etc.

Another way how this could be changed: We could bypass centralization by ensuring that a webpage on ones device can aggregate many services, by emulating visiting other websites in the background, and we make sure that those sites will not be able to find out that are are not being called directly as the main address, but will only be giving out their data. For that, we would have to remove current Cross-domain requests restrictions.


Actually yacy goes this direction: https://yacy.net/


On a semi-related note, IPFS allows your web browser to connect to web sites without necessarily downloading anything/everything from a single source, instead allowing arbitrary content to be distributed to any/all client machines.

https://ipfs.io/


The trend I see is that an enormous amount of effort is made to make "centralisation" easy, while a relatively minute effort goes toward making "decentralisation" easy. Another name for decentralisaton in this context (computers and the internet) is "DIY".

Perhaps we could agree that centralisation in this context is inherently easier than decentralisation to begin with. If true, then what I see is enormous effort to make it even easier. "Frictionless".

Recent history has shown that centralisation in this context is a proven path toward making money, directly or indirectly, at users' expense. As such, decentralisation (DIY) is the path toward helping users save money or avoiding sacraficing other things of value (privacy, control, etc.).

Perhaps time spent on making DIY easier is worth the investment.


https://sandstrom.io solves this problem but I guess they had trouble with traction as they set out to build a two-sided marketplace for a market that wasn't ready. Selling to consumers and SMBs is a huge hurdle, as they'd settle for convenience and cheap when given the choice. And due to SaaS' economies of scale, a decentralised solution is never going to be able to compete with that.

I'm currently building a VPN mesh network for personal use, and the more time I spend thinking abt it, the more https://zerotier.com 's model makes sense. What if consumers, SMBs can create VxLANs over the internet and then host apps and services in that space (Sandstorm Oasis) that are completely private and only accessible to devices enrolled to participate in that VxLAN.


There are very interesting ways around this, eg with some decentralized architecture supporting coins to pay hosters to host people's data, eg Axiom https://axiom.org/

There's no reason github couldn't exist in a truly decentralized environment where people retain ownership of their data, and also retaining its tiered payment structure.

It's just a matter of market forces giving them an incentive to move in such a direction.

I believe most people would choose to retain ownership of their data if they could, as long as it's free and zero effort.


The layer of the stack matters somewhat. You can make it easy to get John to deploy git to AWS. If it's easy enough, cheap enough, and the software were as nice as GitHub, that'd be a reasonable counter-force to centralization. The ship has sailed in the example you mention, but for future categories of internet applications, enabling easy self-hosting seems like a solid aspect to a multi-pronged strategy towards preventing centralization of those platforms. (Incidentally, this is what we're doing with hubs.mozilla.com, enabling self hosting as a means of pre-empting the rush towards centralized avatar communication services)


I like to study this force as well and I agree with your claim. A degree of centralization appears to be quite useful in the real world.

Recently I've been thinking about modelling an economic system after biological ones. It seems they face a very similar problem of resource allocation. Extreme centralization (think big tech) could be thought of as a form of cancer in this frame. Biology solves this through a limited lifespan of cells. So an economy modelled after it might include something like a lifespan for companies.

I'm interested in what you think on this or if anyone knows exsisting resources on this relationship.


Paul Krugman -- Increasing Returns and Economic Geography

> This paper develops a simple model that shows how a country can endogenously become differentiated into an industrialized "core" and an agricultural "periphery." In order to realize scale economies while minimizing transport costs, manufacturing firms tend to locate in the region with larger demand, but the location of demand itself depends on the distribution of manufacturing. Emergence of a core-periphery pattern depends on transportation costs, economies of scale, and the share of manufacturing in national income


This isn't really any different from any other market. In general, unregulated markets in the context of a property framework that allows for unrestricted accumulation of capital tend to be monopolized over time. And, of course, such monopolization will tend to correspond to centralization of services.

With your GitHub example, I think the better question isn't why John doesn't want to run their own server - that much is obvious. It is, rather, why doesn't John want to go to hosted GitLab instead?


> E.g. if Git protocol is decentralized, why is there so much concentration on Github?

If cooking is decentralized, why is there so much concentration of food at the mall?


I have spent a lot of time thinking about this and related market influences since entering the career world 20 years ago.

It started around early 2000's, while working in market research. Got me thinking about all the money dumped into it being a waste... if people could just exchange information freely in some organized fashion.

During 2000s - 2015s the corporations worked hard to form the scaffolding of, and then solidify this organization as they envisioned it.

They too saw the opportunity I was seeing. However, their incentivized profits over morals. The temptation to harness control over this exchange information was to great to overcome. They started off innocently enough. They were just "gardens" without the walls at first. In general, the public was free to admire, as well as take part in, what felt like cultivating meaningful dialogue that theoretically should only enhance life for everyone.

Fast forward 15 years, the novelty of the garden has worn off. We want to go explore something new. But as the time wore on, the garden slowly evolved into something more akin to a maze. Now we are looking for a way out, but cannot even distinguish the walls of the maze from the exterior walls of the garden.

At this point, I probably don't have to say it. But this "walled garden" is a really just a euphemism for intellectual slavery.

I have ideas on what needs to be done, but they seem to fly in the face of conventional wisdom. Especially when it comes to privacy issues.

To me this tells me I am on the right track, because as hard as I look for other answers, I still end up back with decentralization. This problem is hard. It requires faithful means to accurately identify and attribute data to its source creator.

I believe all the tech we need to make it work is readily available. The missing ingredient is incentive. But it's a chicken/egg problem. Until a majority of people want to faithfully cooperate together, it'll not happen. And no single person will agree to work on something that needs cooperation of the majority if they don't believe in it.

Enough ranting... I'm looking further for people to have discussions along these lines. I have ventured out into meetups and whatnot, but I just cannot make connections with like enough minded people to take my ideas further. I know they aren't unique, but they are worthless without others to share with...

can email at [s/.//("L.O.O.M.I.S.5.3")] at google's mail dot com

Even if you can manage to navigate this maze, there is really no way out. The perimeter


guess that should be s/\.// ...


Economy at scale and network effect. The marginal cost, eg the cost for github to host one more repo is very small. But for each new repo they host the value of the service grows.


Here is my counter to these thoughts: Decentralized internet was not created by a myriad of decentralized decisions. A centralized authority came up with the idea and created TCP/IP. Just like centralized decisions were made to create the decentralized/balance of power democratic systems we have today.

Capitalism itself is naturally centralizing. That is why we have anti-thrust laws. We don't just leave it to consumers to make sure centralization does not happen in the market. It is the job of the government to prevent it.

I agree with you that centralization is the natural order or things. But I disagree with your implication that it is somehow the individual users who are responsible for keeping a system decentralized.

In much of Europe, a large number of cell phone service choice and broadband choice was created because government mandated that companies with infrastructure had to let other companies borrow it and offer services on it.

Capitalism and free markets exist because a centralized government maintains it.

The solution IMHO is not to blame the individuals but to combine in collective action and collectively agree/decide we want a different arrangement. In other words one has to bring this to the political level. That could mean anti trust action against google, facebook etc.


> Capitalism itself is naturally centralizing. That is why we have anti-thrust laws. We don't just leave it to consumers to make sure centralization does not happen in the market. It is the job of the government to prevent it.

Centralized social systems such as Governments, Banks etc exist mostly because we need someone we trust (Bank) to arbitrate interactions between two peers, if you remove these kind of trust with the help of tech (eg cryptography) the bank and governments become obsolete because we can now verify things for our self without a 3rd party.


Simply not true.

Faith in cryptography will not help anyone with fraud, theft, and trust.

Naive neo-liberal libertarians and anarchocapitalists attempt to claim deregulation will just work because they've never encountered abuse, fraud, negligence, or read any history.

  One important point: if we actually include all 7 billion 
  people on the earth, most of whom have zero BTC or 
  Ethereum, the Gini coefficient is essentially 0.99+. And  
  if we just include all balances, we include many dust 
  balances which would again put the Gini coefficient at 
  0.99+. Thus, we need some kind of threshold here. The 
  imperfect threshold we picked was the Gini coefficient 
  among accounts with ≥185 BTC per address, and ≥2477 ETH 
  per address. So this is the distribution of ownership 
  among the Bitcoin and Ethereum rich with $500k as of July 
  2017.


  In what kind of situation would a thresholded metric like 
  this be interesting? Perhaps in a scenario similar to the 
  ongoing IRS Coinbase issue, where the IRS is seeking 
  information on all holders with balances >$20,000. 
  Conceptualized in terms of an attack, a high Gini 
  coefficient would mean that a government would only need 
  to round up a few large holders in order to acquire a 
  large percentage of outstanding cryptocurrency — and with 
  it the ability to tank the price.

  With that said, two points. First, while one would not 
  want a Gini coefficient of exactly 1.0 for BTC or ETH (as 
  then only one person would have all of the digital 
  currency, and no one would have an incentive to help boost 
  the network), in practice it appears that a very high 
  level of wealth centralization is still compatible with 
  the operation of a decentralized protocol. Second, as we 
  show below, we think the Nakamoto coefficient is a better 
  metric than the Gini coefficient for measuring holder 
  concentration in particular as it obviates the issue of 
  arbitrarily choosing a threshold.


  ...However, the maximum Gini coefficient has one obvious 
  issue: while a high value tracks with our intuitive notion 
  of a “more centralized” system, the fact that each Gini 
  coefficient is restricted to a 0–1 scale means that it 
  does not directly measure the number of individuals or 
  entities required to compromise a system.


  Specifically, for a given blockchain suppose you have a 
  subsystem of exchanges with 1000 actors with a Gini 
  coefficient of 0.8, and another subsystem of 10 miners 
  with a Gini coefficient of 0.7. It may turn out that 
  compromising only 3 miners rather than 57 exchanges may be 
  sufficient to compromise this system, which would mean the 
  maximum Gini coefficient would have pointed to exchanges 
  rather than miners as the decentralization bottleneck.


  Conversely, if one considers “number of distinct countries 
  with substantial mining capacity” an essential subsystem, 
  then the minimum Nakamoto coefficient for Bitcoin would 
  again be 1, as the compromise of China (in the sense of a 
  Chinese government crackdown on mining) would result in 
  >51% of mining being compromised.
  
  - Balaji S. Srinivasan (the CTO of Coinbase) 
https://news.earn.com/quantifying-decentralization-e39db233c...


>It's the centralization that becomes the stable status quo.

I wouldn't go as far as 'stable'. Centralization is very susceptible to entropy and therefore there needs to be a constant renewing/repairing force so that the structure does not collapse on itself.

Centralization is a big reason why Communist regimes fail after a few decades (even when they get big immediate gains)


I see your trend, but you have missed the answer that is common to all those examples. I have been talking about decentralization before it was cool. Since at least 2011. I want decentralized power generation via solar panels, I want mesh networks to be widely adopted etc. Those networks already exist.

Here is the answer: SOFTWARE. We have amazing hardware - phones, tablets, routers, computers. There is no good software to run on those things.

Facebook, Google, YouTube - why are they so popular?

It’s the software stupid :) (it’s an expression, not calling you stupid).

Facebook has better social networking capabilities with real time chat, videoconference calls etc. than Wordpress or PHPBB. That said, Wordpress powers 30% of the Web. But most of the Web is kinda static, not dynamic like Facebook / Slack / Telegram which took MILLIONS OF DOLLARS TO DEVELOP.

Now, if I ask you, with your amazing hardware, how would you collaborate on a document? Why of course Google Docs or maybe Dropbox etc. WHY? THEY HAVE THE SOFTWARE!

They didn’t give it to you. They have it on their backend servers. Same with YouTube - they do all the transcoding, social networking etc.

Why is github popular? Because you have profiles, ratings etc. The always-on broadband led to an explosion of “software as a service” models, and “the cloud”.

Once upon a time there was AOL, Compuserv and MSN. Steve Case was the Mark Zuckerberg of the day. Then TBL invented the Web Browser which didnt really do much but it let anyone run a web server permissionlessly and HTTP was an open protocol. Fast forward 10 years and aol is just a website. Everyone left the walled garden, starting with the brands. Why have AOL KEYWORD FACEBOOK when you have FACEBOOK.COM. The Web is what in fact allowed facebook, google and amazon to even exist. AOL would have never given them permission!!!!

So today you had those sites and they have the software of Web 2.0 or 3.0 or whatever. Which open software could disrupt all this? SOCIAL NETWORKING AND COLLABORATION. Web 3.0. Meaning user accounts, profiles, realtime updates, notifications, videoconferencing and maybe payments.

My company built exactly this and more (events, group rides etc.) We are like a more modern Wordpress / Drupal codebase. And we are working on a utility token to monetize the ecosystem

See https://qbix.com/platform the video there

Also see https://qbix.com/token (speaks about how to get from a feudal society to a free market on the Web).

Please see this for broader societal implications:

https://qbix.com/blog (latest article)

Look through the articles on that blog for more info.


I think you are confusing anti-competitive control pursuing forces of capital-backed companies and capitalism in general with natural forces. Centralization is unnatural, this is why animals are territorial and people don't all live in the same one country. And why capitalist ideology requires propaganda and force to keep people from revolting. Nobody wants to submit to somebody and give them all the control they want.

And it's not hard to make things decentralized, federation is not that intrusive for example, it's hard to find a way to fund and sustain it when everyone from the system who can fund it also wants centralization, dominance and control.


The article calls ipv6 an “easy way” to decentralize but I don’t see it.

If you click the link on how to get your own ipv6 space you get two options: Ask your ISP for an address or set up a tunnel to someone else who will give you one.

These are the same options we have today for ipv4. I went on my crappy large ISP’s website. They are not handing out static ipv6.

Why would they? The problem has always been more about corporate power than tech.


That'll change with time. I'm behind ISP NAT for IPv4 but have static IPv6 direct to all my devices in the home. My ISP ran out of IPv4 relatively early so had to go this way.

On the other hand, this won't really change much for the firewall issues named above. Computers are too insecure to expose them all to the internet, all the time. We'd need pretty radical changes in how software is written to create a truly flat worldwide network (which has never existed before).


I may be somewhat confused, but how would a different IP address/system prevent centralized services? It seems to me that the same market problems, the strong vendor lock-ins explained in the post, will still persist in IPv6. There's no mention of how/why IPv6 solves those problems, just that they do...


It doesn't prevent them, but with IPv6 everyone can take back control. With IPv4, you are unable to get a decent amount of public addresses to run your service (if you are not having tons of money already).


You are talking about lower layer protocols here that don't matter for decentralization. In fact, a decentralized transport protocol can be made that doesn't require your ISP to give you an internet routable IP address nor put you behind NAT.

Think e-mail, for example. It never required you to have an internet routable IP addresses to communicate with anyone. There were and are plenty of local networks where people run local SMTP servers that communicate with upstream SMTP servers over local network and only those have public IP addresses to communicate with SMTP servers over public internet.


> with IPv6 everyone can take back control

Any service provider can just decide not to route consumer-facing IPv6 addresses. It's not like we all had tons of our own free routable static IPv4 addresses two decades ago, you still had to get them allocated from a service provider and have them route to you, and be allowed to host services. They're "giving away" IPv6 routing and allocation right now, but there is absolutely nothing stopping them from ending that practice.

You only have the control they allow you to have.


Like, I would feel like "I have my own dedicated global address nobody can take from me" is a decentralizing factor... But at the end of the day I still need my ISP to get me to the Internet, so not really.


The difference is, you can take any ISP and continue your business. 1, 2, 3 or even move somewhere else.


We have DNS for just that purpose. You don't need provider independent IP addresses and shouldn't rely on them not changing.


So just use a domain name instead of an IP address, like everyone else.


So, how does a DNS entry to 10.0.2.2 help to be reachable?


Assuming it is not trolling. Every time you get new IP from the ISP, you update DNS record for that domain. And browsers can connect to it to the exposed 80/443 port.


I believe GP's point is that some providers don't provide a publicly routable ip address.


Ah, haven't experience such thing yet. But if such thing happens, indeed it becomes way more complicated.



If you want to selfhost many of those apps, we have built a tool (Cloudron) to take away most of the deployment hassle for many of those apps mentioned in the link. Also to work around the provider lock-in with ipv4 addresses while hosting from your home or on-premise we have a built-in dynamic DNS feature https://cloudron.io/documentation/networking/#dynamic-dns So in the end it doesn't matter if you change ISP or if you don't even have a static ipv4 as such.


Cloudron is missing apps for most use cases I actually care about.

Can you add a google photos type app? Maybe Piwigo (it has a good mobile app) or OwnPhotos[1] (it has face/object recognition). Lychee doesn't have a mobile app or face/object recognition.

Can you add some communication apps like Matrix Synapse or Ejabberd? Mattermost is walled off from the outside world.

Can you add some social apps like Pleroma or Mastodon?

[1]: https://github.com/hooram/ownphotos


Well, Cloudron provides apps for most use cases most people need. So the fact that you need some obscure apps doesn't invalidate Cloudron as a great solution. I've been using it for a long time and love what it has to offer. Plus the devs are great and keep working on making it better.


I wasn't very much aware of ownphotos looks great and will add it to our backlog. Thanks!

Otherwise, Matrix+Riot is being worked on already. Mastodon is similarly mostly done https://forum.cloudron.io/topic/1136/mastodon-microblogging

Our latest addition is now OnlyOffice, currently in testing mode.


I think something like Sandstorm/Cloudron is the way forward with decentralization. I'm more than happy to pay someone else to manage all my services for me. But I see where my apps are hosted as the least interesting question. The important issue for me is lock-in. How easy would it be to switch if Cloudron does something I don't like, or if they outright ban me for some reason? If it's easy to switch, that keeps incentives aligned and fosters competition and progress. Does Cloudron do anything to mitigate lock-in?


Since the whole stack runs on your server there is no real way to ban anyone as such. Worst case is that you would loose access to updates, however all our app packages are MIT licensed (eg. https://git.cloudron.io/cloudron/nextcloud-app ) as such you could just build the app image with docker locally and push to your server manually.


Does anyone else find it ironic that this is hosted on GitHub?


Even the mode of presentation ("awesome-$FOO") is a very uniform meme.


Very nice one - thanks for the pointer, I'll even add it to the blog!


... and updated.


+1 for Cloudron


Local-first software development[1][2] could also be a force for decentralization, since you don't need a coordination server to hold the authoritative version of a document. The relevant protocol for discussion is the CRDT (conflict-free, replicated data type).

[1] https://www.inkandswitch.com/local-first.html

[2] https://news.ycombinator.com/item?id=19804478 (6 months ago)


Some random thoughts on centralization vs decentralization. In theory, IPv6 allows every human, pet, alien and robot to have an address. Each address can host let's say the following services

1. info - dns

2. data transfer - ftp/http/p2p/...

3. communication - email/voip/...

4. entertainment - game server/media server/...

Right now the biggest problems with decentralisation are

1. ownership of hardware

2. assigning of address and interfacing with the network.

3. configuration and setup of services

4. scalability of the service

5. using decentralized stuff for illegal activities

I think 2 and 3 is the biggest win improving decentralized services and getting rid of facebook and the like. Why are two and 3 still hard? This will also make decentralized services appealing to normal folks. The biggest challenge to making decentralized services mainstream is 5.

Hardware is manufactured by a few monopolies and they impose some restrictions and tracking abilities. The assigning of an address is done by the telecom and they have some rules, regulations and tracking abilities. Each country right now is setting up new rules and regulations for the data that comes into it. As configuring and the setting up of stuff is hard we once again have a few major services. The illegal activity makes decentralized services seem like the wild west and centralized services seem like stable societies. The spirit of legality is to ensure fairness but when that doesn't happen people turn to decentralized mediums to express themselves.


I wrote an article about just the e-commerce aspect of this a few weeks back:

https://battlepenguin.com/tech/the-death-of-the-mom-and-pop-...

There are fewer and fewer websites with things for sale. Everyone puts their stuff on an Amazon, eBay, Newegg, Etsy or Reverb store. The big carriers are the store-of-stores.

Some people setup a Shopify or other site as well, but it's usually secondary, or it's to sell merch for another thing like webcomic or blog.


As someone who works in the domain name industry, I'd like to note that there is ever increasing pressure to police domain name use. When you buy and use a generic TLD such as "com" or "ninja", you are buying a US product and it is subject to sanction laws and other things you may not realize. CC TLDs (two characters long like "ch") are products of that country and have their own laws UNLESS the registry company is located in the US - in which case it is also subject to sanctions.

I point this out because unless you get everyone using your IPv6 address directly, your name is certainly NOT decentralized.


> As someone who works in the domain name industry

Do you have any opinions on Namecoin and/or Ethereum Name Service? I'm not looking for an argument, just genuinely curious what somebody in the industry thinks.


Any thoughts on which TLDs are least susceptible to shenanigans?


Decentralized ones like provided by OpenNic

https://en.wikipedia.org/wiki/OpenNIC


Is there an effective way to suppress denial of service attacks in a decentralized network?

Given that there are DNS root servers, isn't the internet actually centralized?


You have to centralize somewhere—IP address assignments are also centralized.


Not really -- look at CJDNS, Tor, or IPFS for examples of totally decentralised addressing schemes. These can be mapped to human readable addresses through Namecoin or Ethereum Name Service. We do centralize somewhere, but we don't have to.


So long as ISPs are centralized, it doesn’t matter if everyone has their own IPv6 address because ISPs are gatekeepers that can censor requests.


The Internet Archive (parent of the Wayback Machine, Top 300 website in the world) is doing a lot to help re-decentralize the web.

Check out this post for more info:

https://news.ycombinator.com/item?id=17685682


This underlines the positive side of Tall Poppy Syndrome. Envy has a number of serious down sides, including making the sufferer miserable for, usually, very little return. But when people stop using an Amazon, Google, Microsoft, etc., because they are successful, it creates a counter-force to centralization. Like hate, when well directed, envy can be good. Like hate, it can easily get out of control and cause great damage, up to and including genocides.


Envy might be a factor but I feel like wanting freedom and mistrusting authorities/corporations are stronger here.


Sorry for the naive question but can anyone explain why hosting an IPv6-only server can help with decentralization? As a provider of some service, say you have a website where you publish some work, is there any advantage of setting it up as an IPv6 server?


Yes and no.

You can build centralization on every layer of the network stack. Facebook is centralization on the application layer. The problem is that with IP addresses being as rare as they are in the ipv4 space you have a lot of tricks that you need to use and ultimately you go through someone using a real ipv4 address.

Ipv6 doesn't have that, so it lets you decentralize further down the stack. While this is good in its own right, currently we don't have issues with ip address monopolies, unlike application layer monopolies.

And that's the problem we have. Doesn't matter how many open layers you have, someone can always build a closed system on top of them.


So what's the latest deal with IPv6?

IPv4 was supposed to run out and the Internet come crashing down 5 years ago, yet here we are, and everything seems just fine, while a billion smartphones were added.


> while a billion smartphones were added

A huge number of them being on IPv6-only mobile networks using NAT64 to access the old Internet.


That was a really bad miscommunication - especially as we are running out of IPv4 in Europe this month. - see https://ungleich.ch/u/blog/when-does-ripe-run-out-of-ipv4-ad...


Missed a really cool vanity url opportunity :)


The internet will not stop being decentralized just because there are some consumer product monopolies.

When Amazon starts running its own dark fiber, and you can only use that fiber to buy shoes that are only sold on Amazon, then part of the internet will be close to being centralized. But that would still just be a small part of it, and it still won't happen entirely.

There's a very long tail between Amazon and the user. How does Amazon connect to the user? Sure, it starts in their datacenter. But then immediately they need to connect to multiple points of presence, which means multiple bundles of fiber going in different directions. And so you'd say, sure, Amazon has lots of DCs, so they could just run dark fiber between all of them. But they're not _everywhere_, so they need to eventually peer to a more global network.

Eventually you get to the ISP. There's two kinds of ISPs: wired and wireless. While they're increasingly the same company, there is a wealth of technology, expense, competition, and physical infrastructure wrapped up in each. Copper and fiber runs to every home, customer support, billing, management, contracts with public and private entities. There's multiple companies in these industries that are bigger than Amazon.

Say Amazon becomes its own ISP. They can either run fiber to every home (lol) or become their own nationwide wireless ILEC (lol) or they can become an MVNO and rent access to an ISP's gear (possible) or they can just pay internet backbones to peer with them and get access to all ISPs' customers. The first two would basically be like making a brand new Comcast; uh, good luck with that. The third is what Comcast already does: they rent ILEC's networks to provide their own mobile service, capturing more customers. The last is how the internet operates today: the ISP deals with the complexity of getting everyone in the country online, and Amazon just pays to connect to POPs.

In a non-net-neutrality world, any ISP can add a line-item to your bill for you to get access to Amazon. In that case, Amazon becoming an ISP avoids that, capturing more profit in the process. But why in "Bob"'s name go through all that work, when you can just charge people individually to access your website? Aka, Amazon Prime. So in order to completely control your access to shoes, they can either build an ISP, or just..... use existing ISPs. Currently, most ISPs aren't adding line-items to access Amazon, so the latter works fine.

In another bizzaro possibility, Amazon could merge with every ISP in America, creating a hyperconglomerate, so only Americans could access Amazon, and every ISP bill is charged for Prime, and every non-American ISP has to pay to route traffic to Amazon. I think that would just crush Amazon's sales, but it's possible. But still the internet would be decentralized, at least globally.

And one final option that actually already exists in developing nations: Amazon and a handful of other monopolies subsidize ISPs to create a "bare-bones" internet plan, where you literally can only surf to Facebook, Amazon, and Google, but you only pay $10 a month. This would be a consumer-only internet that is totally centralized and monopolized - but it's still not the whole internet.

Then of course there's every other business in the world that is not consumer-oriented, all of whom depend on the internet for their business. They also have an interest in a decentralized internet, because it helps them compete with each other, too. They'd be happy to fund a decentralized internet, if only for themselves.

This is all besides the fact that decentralization is actually an architectural decision made by a central organization - the DoD. They made it decentralized because it just works better, not because they wanted the whole world to hold hands and sing kumbaya. Even in this fantasy world of a centralized internet, with one company managing all the consumer services, b2b services, and internet connections, they'd still keep a decentralized architecture, because they know it's really friggin' robust. The network architecture has nothing to do with who has control.

If your concern is monopolies, an internet protocol does not change this at all. If your concern is being able to host your own services, that's still at the discretion of your ISP.


this article doesn't deal at all with DNS so kinda misses like the biggest centralization flaw on the internet but whatever




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: