The Federation Fallacy (2019) | Hacker News Reader

The Federation Fallacy (2019)

rosenzweig.io

49 pointsTomte54 comments

The beauty of federation is that it provides an opt-out. Take a look at Mastodon, for example (FWIW, I am not a user): yes, at the time the article was written the vast majority of folks were on three instances (I am curious what it is like today), but if anyone disagreed with the policies or performance of those instances, he would be free to move to another instance, or set up his own. By contrast, there is no other Twitter which Twitter federates with.

Likewise, a Gmail user is free to move to another email provider (there are large ones, unlikely to have deliverability issues), or run his own (a tricky enterprise, to be sure).

FWIW, I am concerned that democratic rule has its own failure modes, which need to be addressed when designing a system. Do 50.000001% of users really get to permaban another user? Are there any actions which require a higher threshold, or multiple votes over time, or which are permanently forbidden? The U.S. Constitution has the advantage of a Bill of Rights, whose status has so far prevented the outright abrogation of any of them, even if some of them have come under quite a bit of pressure (the First, the Second, the Ninth & Tenth leap immediately to my mind).

Just saying, 'the will of the majority controls' is a recipe for mob rule — which is exactly what 'democracy' means.

Criticizing democracy because of majoritarianism is like criticizing programming itself because of how Perl, Ruby, or any other particular programming language works. Democratic governance can be implemented using consensus, majoritarianism, majoritarianism with a 60% threshold, or whatever other rule or stipulation. A constitution is of course also possible; then the people would decide on some ground rules which will be very hard to change for the foreseeable future. Which might be a blessing or a curse for the future people.

> Just saying, 'the will of the majority controls' is a recipe for mob rule — which is exactly what 'democracy' means.

You are not saying anything. Mob rule is indeed a pejorative term which is often used as a synonym for democracy. What of it? I could call you by some insult or other but I wouldn’t be making a point or an argument.

In a democracy of three, two people can vote to kill the third. That's why democracy is a problem, and that's why the bill of rights is amazing. Call it what you want.

edit: democracy as a governance method must be built on consent, an agreement to abide by the outcome of the vote. No one would consent to a democracy without limitations. They're not even stable; they keep peeling off minorities until there are only two members left.

You’re conflating a more general problem of foundations with democracy. How can a form of governance even be started? On what grounds? How has the mandate to put forth the ground rules? Has nothing to do with democracy per se.

It has something to do with a group of people having votes in order to decide things, where the action that the majority approves of is taken. Democracy may mean something different to you.

> In a democracy of three, two people can vote to kill the third.

It depends on the specific rules of their democracy.

Majority democracy would allow the majority to kill the minority, as in your example. I heard this goes back to Roman legions, who knew that the 51% could kill the 49% if it came to battle.

Democracy by consent would not allow for that, because that third person would (presumably) not consent to being killed. This requires that people respect each other enough not to impose on minorities.

Ideally I prefer a consent model, where each user has a list of other users they approve of, and a list of users whose lists they approve of (which may be identical.) That's what social networks were sold as. Servers should be dumb providers that are primarily worried about bandwidth.

Sure, it cuts down on serendipity, but if you want that, you can subscribe to serendipity345's feed which will add anybody who asks, and will autodelete based on the proportion of bans issued from list subscribers (or a selected council) on a particular user, card-check[0] style..

Federation at the server level is a poor substitute for the lack of finer grained governance tools.

Democracy and auctions are a last resort when you can't find a real solution. They ensure a type of fairness, but have awful failure modes. The ideal situation is when people are only exposed to what they want to be exposed to. The dangers of filter bubbles are 1) who controls them, 2) how they distort the information used for decisionmaking, and 3) who controls them. Absolute, dynamic control of your own filter bubble might make you stupider, but you should have the right to be stupid if you want to be.

-----

[0] Under card-check, when the majority of people in your shop sign a union card, you're a union shop, no need for an election.

> Absolute, dynamic control of your own filter bubble might make you stupider, but you should have the right to be stupid if you want to be.

Individual users are entitled to filter out things they don't want to hear. They own their computer, they're paying for the electricity and network, everything takes place on their hardware, which is their property, presumably confined within their actual home which is also their property. The computer presents itself as a window to the world, but despite literally being billed this way by eg. Microsoft Windows, it is closer to a tapestry placed on a wall.

Not unlike Bitcoin, it's a lot of resources to dedicate to an opt-out. There's a reasonable question to be asked about the cost/benefit ratio.

Not to imply Mastodon shouldn't exist... It is the beauty of the larger open-source ecosystem that in the absence of Mastodon, someone with a chip on their shoulder about Facebook and Twitter would write it. But the market is such that it will likely forever be an also-ran, and the irreducible complexity of hosting a service online is such that even if people use it, almost everyone will be trading one master for another when they leave Facebook for Mastodon, not becoming masters of their own fates.

Which is fine. We live in a society and all.

I think one of the challenges this first wave of federated platforms have is that they are coming in late and still trying to mimic the interaction paradigms of centralized, ad-driven platforms like Facebook and Twitter. But those platforms were designed, from the ground up, to foster platform lock-in and habituate people into giving up data, providing ads, and be goosed for engagement.

I hope once a Fediverse culture takes root we will see a second generation of federated services that start experimenting and doing things in ways that are, for lack of a better term, "natively" federated rather than simply trying to mimic something closed. That might be when it really takes off, when you start seeing things that work best in a federated environment and make awkward fits for a centralized one. A good example of something like this might be the mod scene in gaming, which Steam tries to get a piece of but can't quite seem to corner.

This would be good.

I found myself thinking a few days ago that half of threading on Twitter could be implemented with simple microblogging and a protocol for back-reference URLs (all of which can be done in something as simple as hugo or Jekyll) coupled with a standard for the user-agent to "unroll" the thread. This requires neither a centralized or federated service; it's built on basic web tech.

The only piece of the puzzle missing is that it only works one way; to see the whole 'conversation,' you'd need to find the tail of it (most recent comments). This sort of data-mashing is one of the things a centralized server can do simply and well (though it's certainly implementable on a federated service). But without a federated service, a rule for semantically identifying "topics" and web-crawler that knows how to sort by newest-to-oldest on a given topic identifier gives you the whole story.

Looks similar to what Indieweb proposes for replies[0]: write a post on your site, than "ping" the original via a Webention[1] to let them know there's a reply, rinse repeat. At that point anyone could build a service to unroll those conversations.

[0] https://indieweb.org/reply [1] https://indieweb.org/Webmention

To some extent the main limitation for a truly federated, decentralized and open network like this is that hosting capacity is, for purely practical reasons, not democratized to the extent that internet access is.

Somehow this used to be better, it looks like ISPs used to (and in some cases still do) provide each customer with some way of hosting files and even publishing static webpages. But while this may have been (barely) enough for the early web it's simply not flexible enough to support the current social networks.

Interesting article and I agree with most of it. But the argument that "Democratic freedom is incompatible with an oligarchic authority" is untenable. Ideally, they are incompatible, but in the real world, most so-called democracies today are effectively oligarchies.

It's also amusing that the author chose Wikipedia as the example for this utopia. Wikipedia is awesome (mostly) but the idea that it's something where everyone participates and has the same access is simply absurd. The fact that Wikipedia editing still obeys the 1% rule is a sign. Not to mention the edit restrictions articles have where certain threshold of editing history must be met to circumvent them, the complex rules, tools and culture, the editing wars, etc. If Wikipedia is known for anything, it's that it's hostile to newcomers and that the elite dominates it (just look at the demographics of the editors). Not sure how grandma is supposed to have 500 edits for the privilege to edit an article about some politician. If our utopia is that, then we're back to square one.

The thing about Wikipedia is that users have the right to fork it, per its license, and have historically used it (see Encyclopedia-Libre). It's at least potentially a federated system, in a way that other comparable sites are not. Interestingly, the same is true of Stack Exchange/Stack Overflow: its content is openly licensed, so in principle you could use it to seed your own Q&A site if you objected to how the original was being run.

> its content is openly licensed

Interesting, I know SE/SO itself is not open source: https://meta.stackoverflow.com/questions/313483/is-stack-ove...

Wikipedia is apparently fully open source however: https://www.mediawiki.org/wiki/Download

> ...most so-called democracies today are effectively oligarchies.

I'd like a taxonomy of democracy. Or perhaps even a catalog of traits. Might help limit crosstalk. Like other nebulous ideas -- love, truth, self -- we're all using the same word to talk about different ideas.

Twenty years ago, I reflexively rejected the very notion of sortition as ridiculous. I had spent a decade as an election integrity activist, deep diving into voting rights, election administration, and other arcana. I was utterly committed to the obvious virtue of the Australian Ballot. To the necessity of transitioning to Proportional Representation and Approval Voting. The hills I would happily die on.

Now? Not so much.

After reading David Graeber and rereading Chomsky and Jill Lepore's early history of polling (eg Gallup), I'm newly open to considering alternatives. Thoughtful deliberation sounds pretty fucking great about now. Stuff like citizen's assemblies to tackle thorny policy debates. And maybe polling is superior to voting, at least for some use cases.

Any way. Current me agrees with both you and OC wrt democracy vs oligarchies.

Restated, our current group decision making institutions and strategies are prone to capture, which then transmute democratic arrangements into oligarchies.

We see this over and over again, big and small. Just witness all the coops (like farmers cooperatives) which are vulnerable to hostile takeovers, thereby disempowering the actual members.

We need better technology. Something that is both more democratic and less resistant to capture.

Is that even possible?

Ever play Nomic? It's game which simulates the Paradox of Self-Amendment. Basically the game version of what we call a constitutional government.

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

Or, we can just emulate our ancestors. According to Graeber and Wengrow (Dawn of Everything), many societies used culture and tradition (vs rulesets) to contain would be tyrants. Often thru violence.

Untenable? You kind of resolved it yourself by saying that “democracies” of today are just “so-called” but are effectively oligarchies. So it’s not an untenable statement.

I agree with the fundamental hypothesis, but the proposed "democratic" web suffers from the same fundamental problem as the federated model: how is it paid for in sweat and treasure?

Most people want to use the internet, not run the internet. No one (outside of some die-hards who will no doubt explain below) wants to run their own infrastructure -- it takes too much effort to do correctly and otherwise distracts from the end goal -- actually using the internet.

The "democratic" web will need to be paid for, but everyone is used to getting internet for a flat fee from their ISP, so no one will pay for a VPS they have to manage (again, outside of some die-hards who will no doubt explain below) because why pay for the cow when the milk is free.

That's right: the reason centralization wins is because its "free" to the user, at the cost of privacy and advertising.

The only way to get the "democratic" web envisioned is to use the democratic version of the same behavior that the best centralized companies use: get a guaranteed revenue stream.

That's right, to get a democratic web you will need to raise a tax and have the government run (or otherwise subsidize) the infrastructure.

Maybe we could call this tax a subscription and contrary to nation state that you have to belong to on birth, your could choose your "internet state" by subscribing. You would get some rights to access the infrastructure of this internet state (mail, DNS, micro-blogging, whatever...) and participate to its administration by voting and have some obligations (pay your subscription, vote, validate identity of people, participate to arbitration instances). I think this kind of corporation is called a cooperative, maybe the destiny of the free-internet are cooperatives.

Federation between those virtual states would have a precedent, existing federation of real states.

Voluntarily paying for subscriptions is exactly what people _don't_ want to do when an equivalent service can be had free of monetary charge.

Also, the largest problem with "subscribing to your tribe" is that on the internet your email address is often your identity. If I have a falling-out with Tribe A and need to move to Tribe B, now I have the hassle of changing over all of the accounts using by Tribe A identity to use my Tribe B identity. This only gets worse when I have to move my entire blog and maybe even DNS entries (assuming I wanted the added hassle of managing my own domain in the "democratic" internet).

Changing identity is the "payment in sweat" I mentioned above. Most people will make the a priori decision that changing identity is too much work and instead favor the centralized systems like Google to manage identity.

That's clearly not different you can also change your email from Google to Live but you also have to pay in sweat as you say. The difference is you can have some control on your "tribe" as you call it because it's a democratic human organization where Microsoft and Google are sort of oligarchic human organizations, some people have power over them, you're just unable to be part of those people for actual money/power reasons.

As I said I see no technical/organizational reason why techniques advocated for radical decentralization (like blockhain, dht, name-independant routing, etc..) could not be used to make the necessary sweat to move from a tribe to another easier (as there are political framework like the European Union which makes it easy to move your life from France to Germany or the reverse); move you signed DNS record from tribe 1 to tribe 2 so you move routing of your email, give some capability to tribe 1 to sync your data from tribe 2, and forgot about tribe 2. If tribe 2 is unreliable and still advertise that it should receive your email, you can do nothing, people in tribe 2 will lose contact with you. But if (real material) state A claim you are a criminal, and state B give you asylum that's great, but it won't make the danger of returning to state A disappear.

I think that the case where tribe would be adversarial is an edge case, the same way Live and Google do not block each other emails.

Anyway what I say will only verify if some people try it, make a global "user of internet association" with some kind of constitution for governing it democratically, amass some capital, create some infrastructure a la google-suite and see where it goes. If nobody adhere to it, I was wrong :) It seems that in this setting many current technology which needs some trust anchor to have some kind of performance become simpler (global ledger, secure name system, etc..) and very powerful.

> That's right, to get a democratic web you will need to raise a tax and have the government run (or otherwise subsidize) the infrastructure.

Is this the terrible conclusion that you were building up to? I see no problem with that compared to the alternatives (existing or non-existing).

One unmentioned feature of federation is that it lowers the cost of switching. If your provider begins to misbehave, you can jump to another one. This is especially true in email, where you can move from one host to another by updating your domain's MX records.

Without federation, switching means migrating both your host and your network of participants simultaneously. That is a much harder and more expensive coordination problem.

You are analyzing the wrong thing: email as a hosting mechanism isn't exactly federated, it is the email addressing scheme that is federated. You happen to be self-hosting your domain (which is part of a different system: DNS, which is more hierarchical than federated, though I have often put it in the latter category), which allows you to move around and even delegate to third parties your email host (which is really what you are doing: federation isn't what is allowing you to delegate your hosting provider, as your hosting provider works for you and you alone), but at the email layer that's because you did the super-rare thing of owning your own domain name and knowing how to have an MX record.

The switching cost for the average user of a federated system--one which has experienced lock-in on someone else's naming system, such as by using a gmail address or having a mastodon account on a large instance--is brutal. There is an advantage here that that is even possible, and there is then competition between these inter-compatible providers, but there are still many reasons why users tend to end up wanting to end up attached to larger namespaces (all of technical, political, and economical). Hell: even some of the smartest people I know seem to have an @gmail.com address (or worse: have let their identities become owned by a company they happen to work for) no matter how immediately "amateur" that makes them look to me :(.

(I gave a short "talk" about this at the Distributed Web Summit a few years ago, but the video they took has an over-an-hour audio sync issue and I haven't taken the time to edit my own copy yet.)

Email federation is more than just the addressing system. Even with a consistent address, your host still needs to be able to send and receive SMTP messages to participate in the ecosystem.

Address lock-in is real but doesn't fully negate the benefits of federation. Even if you don't control your address's domain, you can switch providers without losing the ability to send and receive email. There is extra cost in pointing your bank account to your new address, but it is possible. If your bank communicated to you via a non-federated protocol like WhatsApp, switching would be impossible.

There is an assumption here that 'people hosting their own instances' is the ultimate goal and I don't think that's true. The important thing is that someone can always set up a new instance if they don't like the existing ones - competition isn't a perfect system but it tends towards providing the services users want. Without it though, you're locked in and stuck with what you've got.

Also note that someone else providing you a service is not the same as someone else owning your data - a properly designed E2E encryption scheme can mean only you have your data.

She seems to assume that federations are supposed to be small groups where everyone knows each other. Has anyone claimed that?

Email and the web are the perfect examplea of federations, because they are not controlled by a single company; they are interoperable. From Gmail I can send an email to a friend who uses Outlook. That way I'm not forced to use outlook if I want to communicate with him and I can use any email provider I want. The fact that Google and Microsoft are trillion dollar companies is irrelevant here.

Same with the web, I can change my ISP and keep accessing the exact same webpages as before because the web is federated, and not controlled by my ISP. The fact that I use it to visit pages controlled by other large companies is irrelevant.

Facebook or Whatsapp, on the other hand, are not federated. If I want to communicate with my friends who use Facebook or Whatsapp I have no other choice than to use Facebook and Whatsapp.

Few comments here mention that if you don't like your current provider you can easily switch - take your data with you. I think the problem of "not everyone wants to run things, everybody wants to use them" could be solved if we make our data fluid with something like IPFS (note: _something like_). What I mean is you can have data in multiple places a none is central one (discovery could be local, through friends and hubs etc. like SSB). You could have it on some federated server, on your local pc and if your provider misbehaves your data is still accesible from other places and moving it is no longer manual process where you have to click export, import etc. just login to other provider and that new service will pull your data from wherever it is accessible.

I don't know if I am making any sense :D Please be forgiving.

tl;dr; right now you address someone by nick@matrix.server or nick@mastodon.instance - we should be going in direction of adressing people using only _nick@service_ (or just _nick_) and dont care where the actual content is

I don't agree with the author's views on anarchism, and their analysis is missing a key piece of political thought: who has control of violence in their proposed system of a democratic internet?

In the anarchist federated design, the answer is you, the individual. Even if the admin of your particular chosen federation instance chooses to boot you off and delete all your posts, you still hold a copy of your own data which you can take with you to another instance, or to even start your own instance.

In the "democratic internet", the answer is you, via your vote - but only in principle. In reality, it's whoever holds the keys to the server you want to use. The ultimate authority over what goes on on Wikipedia is not its userbase, no matter how much of a voice they have: it's the people who physically control the hardware and software that Wikipedia is running on. The monopoly on violence that this "online state" holds over you is not a consequence of your own participation - it is a result of cold hard fact.

Worse still, unlike a physical democracy, the "democratic internet" puts your experience of democracy under the control of this authority. If the very fabric of how you participate in an online system is manipulateable by the people in charge of that system, you have little-to-no protection against a coup - what do you do when someone deletes the code that lets you vote, or comment, or anything else? In a federated world, the protection you have is federation itself: a compromised instance is immediately abandonable in favour of any other instance.

Personally, I find the idea of empowering tech illiteracy misguided. Computers are too powerful, and too complex, to allow foundations built on them to be written off as trustworthy. Trust in technology can only be based on a constant adversarial effort, participated in (or at least, possible for) every user. How we get there is a much harder problem, and the solution is not centralisation.

> In the "democratic internet", the answer is you, via your vote - but only in principle. In reality, it's whoever holds the keys to the server you want to use. The ultimate authority over what goes on on Wikipedia is not its userbase, no matter how much of a voice they have: it's the people who physically control the hardware and software that Wikipedia is running on.

And in real-life democracy it's the same. Ultimately it's the army that has the tanks.

> the "democratic internet" puts your experience of democracy under the control of this authority. If the very fabric of how you participate in an online system is manipulateable by the people in charge of that system, you have little-to-no protection against a coup - what do you do when someone deletes the code that lets you vote, or comment, or anything else?

What do you mean by that? If wikipedia jumps the shark people can make a new wikipedia. If France turns into autocracy people can emigrate or revolt. It's pretty much the same.

>And in real-life democracy it's the same. Ultimately it's the army that has the tanks.

There is a lot of truth to this, but the online version is far worse. In a physical democracy, you can't be prevented from participating except by the efforts of the other participants. Even in authoritarian states, there must be people who inform on you, people to drive the tanks, etc.

In the online version, one person can do all of those jobs.

>What do you mean by that? If wikipedia jumps the shark people can make a new wikipedia.

This is missing the point of the article. "People" cannot make a new Wikipedia. The average Wikipedia user cannot write a single line of code, or administrate a server, or do anything useful towards creating a new instance. Only those with technical knowhow can actually subvert Wikipedia's centralisation.

A world where "people" are able to spin up new instances when they want to is exactly what federation aims for - but the point the article makes is that such a world needs all users to be technically capable. The author's idea of an "online democracy" involves users who are empowered even though they aren't capable: how can those people avoid a technical coup?

> The author's idea of an "online democracy" involves users who are empowered even though they aren't capable: how can those people avoid a technical coup?

By becoming capable when it's required or hiring someone who is capable? I think this argument becomes more and more theoretical, in practice both in real life and online you don't need 100% of population to be able to run a country, specialization is fine.

What you need is isolation so when one country goes totalitarian - there's somebody to accept refuges and bomb the death camps if needed.

So the main problem with democracy online is that network effects are unchecked. When Facebook or Google goes evil - they take everybody with them.

>So the main problem with democracy online is that network effects are unchecked. When Facebook or Google goes evil - they take everybody with them.

I think the trouble with this discussion is that you don't really agree with the author. They are proposing a massive online community centralised on a single system - and which is kept in check by user democracy, not alternative instances.

If network effects are what keep Facebook and Google in power, imagine the network effect you would get from all the participants in a real-world democracy being on the same platform. How are those users going to start another instance, even "hiring someone who is capable" to make it for them? They won't, simply. In such a system, most users are not able to circumvent a technical coup. Just look at how many "evil" steps Facebook has taken, successfully holding on to their userbase.

I agree the centralization is needed and that democratic centralization is preferable to proprietary walled garden. I disagree that democratic centralization is sufficient - I think we also need self-hosted options even if they are marginalized by the centralized democracy - to keep it in check.

This:

> Just look at how many "evil" steps Facebook has taken, successfully holding on to their userbase.

is orthogonal. Facebook is not democratic and does not offer a self-hosted version. And it never did. So with Facebook the question is not about what is preferable but how do we get from this worldwide tyranny to a centralized democracy with or without self-hosted alternatives.

I think the facebook situation is actually an argument in favor of my position. If Facebook was a centralized democracy with self-hosted alternatives at some point - it would be much easier for people to migrate off it when it started to become evil, and it becoming evil would be significantly less likely.

As for how Facebook loses now - I think eventually it will be taxed into ground/blocked by most big trading blocks/countries. There is already a lot of talk in EU of enforcing some control over social media. It holds too much power with too little control for countries to let it go forever. Facebook can (and sometimes do) change who wins elections. It can't go on like that forever.

>Facebook is not democratic and does not offer a self-hosted version. And it never did.

But Facebook has been around for a long time. Facebook alternatives have had lots of time to spring up, and some have even had a lot of support (e.g. Google+). Why haven't those alternatives kept it honest?

If it's the lack of compatibility, then the same issue can always appear in any centralised system. The fact that you have a "Export to local instance" button today doesn't mean it will be there tomorrow. If there is a technical coup, your ability to easily leave a centralised instance can easily be taken away - and if you're a non-technical user, what will you do then? There is no fighting the network effect if the people in charge are determined to make it difficult to federalise.

Blockchain systems can only really be manipulated by controlling a true majority of their resources (a 51% attack). In this way, they can be said to enforce a quasi-democratic form of "voting" albeit in a way that still reflects real-world constraints on, e.g. resource use.

But blockchains are basically federated - every user is able to hold a complete copy of the chain and participate in the same way, provided they are technically adept enough.

The current state of cryptocurrency proves my point, in a sense: how can a non-technical user be involved in the blockchain, without needing to understand private key cryptography or maintain a wallet, or a copy of the chain? By centralising that effort in an exchange, along with thousands of other users. But exchanges have exactly the issue of the author's "online democracy" - they are easily subverted by the people who actually control the technology. Thus, exchanges get hacked and their users' "decentralised" currency gets taken, because those users tried to bypass the technical effort required to operate the blockchain in the first place.

Indeed, both the takes on anarchism and democracy are ridiculously simplistic and soaked with western dominant thinking (being at the top of current political systems, linear view on progress, masses are dumb, liberal democracy is the only option).

I'm gonna quote this gem:

> In political history, the birth of free nations typically contains three stages: dictatorship, briefly overthrown to anarchy, reformed to democracy.

They don't even seem to realize that both words have so much different acceptions that they have no precise meaning by themselve, yet asserts facts like "Anarchy fantasizes that freedom protects itself", "In any democracy, centralised power is wielded to protect freedom".

Another hard no-no is the plot. This plot should have been log-log ffs! You're already saying it's probably power law and then continue on saying:

> There is a massive spike corresponding to just a few instances, and the rest of the graph is nearly invisible to the naked eye, so tiny and so overshadowed by just a few giants. Frankly, this distribution is closer to the Dirac delta function than a power law.

This is not serious, especially for someone studying math and doing such elite low-level dev. Don't math-wash your opinions (perhaps you know you're just joking/bs-ing but other people might not).

In the light of all these flaws, the actual content is a mix-n-match of unsubstantiated opinions (democracy is good, anarchy is bad) and then at the end a call to build social institutions on the internet.. I'm quite annoyed at this post, on one hand this is all nice and kind, and i can (hope to) see the good intentions, but on the other hand it's really really sloppy and so easy to conter-argue on.

edit: there is also not a single mention of p2p systems, in an article about centralization and federation, which is like the elephant in the room

>In the anarchist federated design, the answer is you, the individual. Even if the admin of your particular chosen federation instance chooses to boot you off and delete all your posts, you still hold a copy of your own data which you can take with you to another instance, or to even start your own instance.

Which is not much consolation if everybody is on the first instance.

The beauty of how e.g. email is decentralized is that I can change email provider and still get to talk to everybody I used to over email - without having to convince them to switch.

>Which is not much consolation if everybody is on the first instance.

Federated systems are typically designed so this isn't an issue. You aren't limited to seeing only content that originates on your Mastodon instance, for example - you can follow users on other instances, and the instances will share their content as needed.

Of course there are other considerations there - instances may choose to share content conditionally e.g. only with other instances that comply with their banlists. But this is true of email as well.

> The beauty of how e.g. email is decentralized is that I can change email provider and still get to talk to everybody I used to over email - without having to convince them to switch.

If Google blocks your email address as spam, how exactly are you going to talk to everybody on Gmail, regardless of actual mail server provider to that address?

My point wasn't about being "blocked as spam" though, but about wanting out of Gmail or having your Gmail account closed.

That said, if Google marks your email as spam, you can create an alias and send email as that, or create a different email account - same as they can if their previous provider throws them out, folds, and so on.

You still get to email your gmail-using friends.

Decentralization (and federation) is about choice.

Choice to host your own data or delegate to a trusted party, not to have to follow the rules of a single company, with a single culture and within a single jurisdiction. In one word: feodalism.

> But do you host your own mail server? Do your friends? Does your grandmother?

No, but that's not the point. Being federated my friends can switch email providers (keeping their address too if they rent a domain) like they can do with gsm sercice and still communicate with their peers.

They can not do that with facebook, teams, slack, etc. And they also end up having 3000 service providers installed on their phones/laptops/whatever. With 3000 logins to keep secure, 3000 software packages to keep up to date, 3000 services sucking up the battery, etc.

I agree, and I think this model - centralized democratic open source server for most people + self-hosted version when you need it - is ideal.

Without the centralized server somebody will take over with a proprietary walled garden that offers better user experience.

Without the self-hosted version people with special use-cases will move to proprietary competition that will eventually distrupt the democratic well meaning version.

Most recently I've seen it play out in virtual tabletop rpgs space. There is the proprietary centralized roll20.com that barely works because for the longest time it had no competition. Some of the premium features were broken for months. And there's foundry vtt which is much better, open source and self-hosted, but is hard to setup correctly on your own.

What's missing is centralized democratic version of foundry vtt that will make it as easy as roll20 for new players and DMs. It's not like money is the problem - people in this hobby are used to pay for customized dies, figurines, books, maps etc. They could easily charge 100 USD a year for 4-5 person group and that will easily pay for the computing needed and bandwidth. Both of the groups in which I play chipped in for cloud server and a license of foundry no problem. The problem was configuring it and dealing with network issues.

Regarding the distribution of users on Mastodon servers, the article seems a little biased. The author complains about how "an average of six hundred users per server" doesn't look like "every user knows the system admin where their data is stored". The median users per server would have been a fairer metric than the mean, specially because it later points out that three servers have over 50% of the users in the Mastodon ecosystem...

It is OK if most EMail user use Google. The important point is that if Google deletes my account or does something else I don't like, I can use another EMail provider, including one I set up on my own. Yes I know these days it still won't work because Google will censor my EMails and make it impossible for me to communicate with its users. But it is the general idea, and I think there are new systems (like Matrix) where censoring stuff would not be as easy.

And if users are frustrated about not getting my EMails, they could also move away from GMail.

Mastodon is also a bad example because it sucks or is weird. In theory users of different servers could communicate with each other, but my impression was that it is too complicated and not really encouraged. It was also invented for the specific purpose of being able to censor things, maybe that shows.

> In theory users of different servers could communicate with each other, but my impression was that it is too complicated and not really encouraged

What do you mean? The rare times I use mastodon I can just reply to any toot from any instance and it just works.

You can communicate p2p, but not really participate in discussions on other servers. At least it is not encouraged. Might only be a client issue. I don't remember the details, but I think it is the way messages from other servers show up in the timeline.

You can, you just don't share the instance-local timeline. If you're mostly talking with people you follow and vice-versa, it all works pretty seamlessly. But people on other instances won't discover your posts via the local timeline, which is where a lot of people start when they first join.

You might be visible in the global timeline depending, but that's a mess. :)

I have somehow not seen this post before. Thanks for posting it. I couldn't agree more. I particularly like differentiating between decentralization and federation.

The key point here is that this is something almost nobody wants. I'm surprised the author didn't bring up what is probably the most successful large-scale federated system: the POTS telephone network.

Spam calls, robocalls, caller ID spoofing, limited controls on blocking people, etc. There's a reason things like WhatsApp, FB Messenger and the like have grown so big. Many people now view the phone part of their smartphone as just a huge annoyance.

Yes, yes, it's hard to self-host federated services, because system administration is hard.

So we have to make it radically easier, then.

Agreed. Sandstorm in particular was and is a great example of simplifying system administration, and holds the most promise of anything in that realm IMO. I'm not sure if it's in a position to host federated services at the moment, but I know it's been talked about - if that can be made to work, it'd be the simplest solution I can think of.

Reply on news.ycombinator.com