Today in Tedium: Recently, a reader asked me to lay out the concerns about centralization and decentralization in relation to the internet, having had questions about it in response to a recent piece of mine on my sister newsletter, MidRange, about Jack Dorsey. It’s a debate that constantly comes up, and if you’re not in the know, it can feel like it’s passing you by. In many ways, it’s about who controls the keys of digital culture, what rooms you can go in, and what you can say. It is the difference between public property and carefully curated environments. It is also the difference between you being completely responsible for your screw-ups and someone else doing something that can negatively affect you—or, depending on the situation, not doing something. Given a decision by a PayPal Mafia alum/rocket man/satellite internet guru/electric car maven/digital edgelord to acquire a significant stake in the social network Twitter this week, this conversation is going to become more important in the coming years. So, let’s have it. Today’s Tedium offers an FAQ on the centralization question. — Ernie @ Tedium
Keep Us Moving! Tedium takes a lot of time to work on and snark wise about. If you want to help us out, we have a Patreon page where you can donate. Keep the issues coming!
We accept advertising, too! Check out this page to learn more.
“The implementation of a protocol must be robust. Each implementation must expect to interoperate with others created by different individuals. While the goal of this specification is to be explicit about the protocol there is the possibility of differing interpretations. In general, an implementation must be conservative in its sending behavior, and liberal in its receiving behavior.”
— Jon Postel, a key figure in the development of the early internet, discussing the nature of protocols in RFC 791, a document about the creation of the Internet Protocol, or IP, system. In its purest form, our networks carry no real bias. They just make it possible to share information based on the protocols given. In many ways, the humans are the ones who complicate all of this.
What did the internet look like at first? Was it a haven of free speech?
In 2016, a software designer named David Newbury came across something interesting in his father’s old papers—a map of the internet, as it looked in 1973. (His dad worked at Carnegie Mellon University, an early school on the ARPANET network.) It was just a handful of nodes, enough to appear on a single sheet of hand-drawn paper, reflecting connections between major research universities (Harvard, Stanford, MIT, and the university that started it all, UCLA) along with some companies (Xerox, Bolt Beranek & Newman) and some government-associated nonprofits (the RAND corporation).
In its early days as ARPANET, the internet was a just a handful of networks that talked to one another using a unified protocol. One network wasn’t dominant over another one.
But there were some significant limitations over how the internet could actually be used that would make any Twitter user lose their mind. The network was not intended for personal use at all, but it still happened sometimes. In one famous story, early ARPANET developer Leonard Kleinrock, who supervised the very first message ever sent through the network in 1969, used the network to locate his electric razor, which he accidentally left in England. While there was no law against it, Kleinrock’s personal request broke formal decorum.
“It was a thrill,” Kleinrock recalled in a 1996 Washington Post interview. “I felt I was stretching the Net.”
The internet—in the raw, unformed version we like to think of it today—didn’t truly emerge until the late 1980s and early 1990s. It was a mashup of publicly developed protocols and services, most of which the average person did not realistically gain access to until the mid-1990s, in part because digital networks were more exclusive in nature and it was harder to distribute the internet to individual homes than it was to large universities.
Of course, there were other, more open digital networks at the time. The University of Illinois’ PLATO system, by the mid-1970s, had become a haven of early game development, of digital publishing, of communication between college students offered access to a new way of interacting with one another.
The internet, in its raw form, was essentially decentralized, with no operating body. When it comes down to it, data is data. If given no rules or regulations, data would fly around the internet with no concern whatsoever for things like laws or freedom of speech or viruses or security. The machines don’t make decisions on content quality unless we tell them to.
But that freewheeling nature created a lot of problems that needed to be accounted for later. One of those problems was explicit content—which could be shared freely online using mediums such as IRC and Usenet, and made the internet unsafe for lots of people early on. Another was the fact that there were no mediators; if someone was to cyberbully you online, there was nothing you could do to stop them.
And while the early internet was seen as an ideal because of the innovations it allowed for, the truth of the matter is, it was an imperfect place, akin to a roadside bar in the middle of nowhere. Sure, you had a lot of freedom to do whatever you wanted because there’s no police in sight, let alone bouncers, but you were also at a significantly higher risk of bar fights as a result.
What is decentralization?
Decentralization refers to a network without a central point, a description that definitely fit the early internet, but also describes any network infrastructure that has no central operator.
An important example of decentralization appeared around the year 2000, when WinAmp developer Justin Frankel, then an AOL employee via the company Nullsoft, developed an application called Gnutella, which allowed for the sharing and distribution of files without any central hub. (Frankel built his software in a centralized environment; AOL almost immediately forced him to give up the innovative project, and he later left over a similar issue.) Later technologies, like BitTorrent, furthered the benefits of decentralization, making it possible to pool the networking capabilities of strangers to download episodes of Doctor Who without having to worry about doing so from a centralized resource.
Decentralization also carries some technical advantages as well—for one thing, it removes single failure points from within the network, especially if those nodes are redundant. When Facebook went down for roughly six hours last year, it caused problems for much of the internet because everyone is so reliant on Facebook being there for reasons beyond its existence as a social network.
If Facebook was a set up like the federated social network Mastodon, one instance of Facebook might have failed, but it might not have taken down the whole network. Mastodon is probably the best-known decentralized social network, as it is operated through a loose network of nodes that talk between one another, each managing its own base of users and their own conversations.
One challenge of decentralization is that if there’s no central body, there’s nobody to pay you. Which is perhaps a big reason why we eventually moved away from decentralized networks.
In many ways, decentralization is at the center of many of the recent discussions around Web3, which, at its basis, sees blockchain technology as a potential solution to the challenge of trying to create a network that is open, offers ways for creators to be compensated, and can’t be limited from a free-speech standpoint.
The year Compu-Serv Network, a business-oriented dial-up network first went online. At first, the network offered remote access to Golden United Life Insurance’s machines in a time-sharing format, but by the late 1970s, the network was expanded to a full consumer-oriented service under the name CompuServe. It represented one of the first online networks for end users, as well as one of the first “walled garden” online networks.
What is centralization?
As I mentioned, the internet in its rawest form is a decentralized network, made up of lots of separate parts, called nodes, that come together and allow for connections to be made anywhere at any time. Protocols, like HTTPS, and domain names put a degree of polish on it, but the truth is, it is a network with no center.
Centralization is in many ways kind of the opposite of all that. It’s intentionally an organizational body with a middle, where everyone is interacting with a single system.
A good example of a centralized network is a bulletin board system. The person who runs that bulletin board system may only have a few lines available and can only support a few users at once. If one person is taking up the phone line for too long, the sysop can easily kick that person offline if they don’t want them there.
In the modern day, the BBS has given way to other types of centralized systems, the most famous of which are social networks, like Facebook or Twitter. You connect with these systems, which are largely a curated experience, rather than a protocol, like the World Wide Web.
But lots of other things can be centralized as well. Many websites are, for example, as they’re built around a single user experience from a single location, rather than something that is consumed in an aggregated form, like email.
Centralized networks are often referred to as “walled gardens,” which refers to the fact that the experience is proprietary, and not externally managed. It can be a safer, highly curated experience (think Disney World and you’re on the right track), but it means that if you find that at all limiting, you’re stuck in a position where you can’t say all the things you want to say.
Why does centralization have an impact on speech? Would decentralization help?
Centralized networks are often controlled by singular entities. For example, in the BBS example I gave earlier, that’s one person in charge of that setup, and the only thing stopping that guy from being a jerk is his own personal compass.
The equation grows more complicated as more entities get involved.
I have another example, and it involves the network that became AOL. (I mentioned this story last year.) In the mid-1980s, AOL was called Quantum Link, and it had a significant user base of LGBTQ users. However, the company’s service primarily used Commodore 64-based computers, and one of the resellers of the popular computer pressured Quantum Link to shut down its alternative lifestyle groups in deference to the reseller, which had Christian beliefs. Quantum Link complied, creating a bad environment for the network’s previously welcomed LGBTQ community.
In a decentralized network, this would never happen. Instead, if you want to block off certain types of information you find controversial, you might need to set up a filter. (Think Net Nanny.) The onus would be on the user—or, say, their ISP—to prevent the connection to the decentralized resource.
Years later, this problem has become more pronounced, as most of our prominent digital resources represent centralized hubs that we interact with on a regular basis. Like our theoretical BBS sysop of yore, the benevolence of the people running it ultimately decides the final result.
However, even they at some point needed some clarification. A key legal measure, at least in the United States, is called Section 230, a portion of the Communications Act of 1996 which specifically carves out room for centralized network operators to prevent liability for the actions of users in most cases. It’s often mistakenly believed by observers that Section 230 prevents websites from being moderated; the truth is, it actually gives them the right to moderate, by allowing for the operators of social networks to manage their networks as they see fit. (The law came about after a decision involving the early online network Prodigy and the company that was featured in the movie The Wolf of Wall Street, Stratton Oakmont.)
The thing is, social networks are generally owned by companies, just as brick-and-mortar retail stores are owned by companies. If someone behaves poorly in the retail location, the retailer is within their right to remove that person from the location—after all, that person could scare away future customers. (Think “no shirt, no shoes, no service,” and you’re on the right track.)
If you wanted to have a totally free network, you would need to basically build a network that was disconnected from any sort of external influence. Any advertiser or publisher or user can influence how Facebook or Twitter operates, limiting just how “free” the network operates in practice. If an advertiser doesn’t want to run their ads next to hate speech, Twitter is compelled to stop hate speech on its network; if users raise concerns about cyberbullying, Facebook is theoretically compelled to create solutions to stop cyberbullying from spreading.
Those who complain about Section 230 or that their freedom is being limited by the social networks they use are effectively disagreeing with the status quo, which generally favors the companies and the existing network effects that have been built. They basically want networks where the people who operate them keep their hands off—which is difficult to promise. After all, social networks have stakeholders.
On the other hand, those who want a more regulated experience—say, because their kids are using it, or because of concerns about cyberbullying or hate speech—face the opposite problem. Social networks have become so complex that no one organization can stop everything; the best they can do is create better algorithmic filters or safety measures to protect users.
The percentage of Twitter’s revenue that comes from advertising, according to Seeking Alpha, which creates a huge pressure point for a centralized network. If Twitter upsets its advertisers through the decisions it makes, it could threaten the long-term viability of the network, because when it comes down to it, it has a single point of failure—Twitter, the corporation.
So where do Jack Dorsey and Elon Musk fit into this conversation?
This week, this conversation around centralization kind of became a hot topic thanks to two Twitter-associated people.
Over the weekend, Twitter co-founder Jack Dorsey publicly tweeted about how he had regrets that the network that made him famous was centralized, not decentralized, which created problems in terms of the network’s infrastructure.
Almost simultaneously, Elon Musk was starting to tweet out a bunch of stuff about Twitter was failing at its end of the bargain, and how he wanted the network to follow more traditional free speech standards. To prove this point, he bought a significant portion of the network essentially overnight.
In some ways, Dorsey and Musk were approaching the same problem from different perspectives. Dorsey (who has decades of experience tackling this problem) sees social media as something that was harmed by centralization, and while that centralization made him rich and famous, it also damaged the essence of the pure internet. Based on his comments, he sees the problem as something a decentralized internet can solve. One initiative he has pushed for on this front is something called Bluesky, a decentralized standard for Twitter. If, say, Truth Social wanted to operate on a decentralized node on Bluesky, it could operate using that protocol—but Twitter wouldn’t have to accept it.
Musk, meanwhile, sees the problem as one of bad leadership, and is pushing for changes to the way Twitter operates so it favors a free network over something more moderated. This might lead to some controversial decisions, like the addition of an edit button, or it might lead the network to bring back controversial figures that were previously banned because of their toxic nature.
Whatever happens with Twitter in the coming months is likely to set the tone for future conversations on centralization and decentralization online. Twitter isn’t by any means the largest social network, but it carries outsized influence, and that makes it worth watching.
What this discussion sometimes feels like.
Which is better, centralization or decentralization?
Ultimately, this is probably down to user preference and use case.
In the case of my newsletter, I prefer using a service called EmailOctopus to actually send the email; it communicates with a tool called Simple Email Services (SES), which is managed by Amazon Web Services (AWS). If EmailOctopus or AWS stopped working tomorrow, I would still be able to send email through another service if I needed to. Because email on its own is a decentralized network, I ultimately have the freedom to decide my own use case and the tools I use to build it.
The thing is, however, setting all that stuff up is a bit of work and might be more technical than you’re willing to get. Simplicity is where centralized networks shine.
But if I were running a newsletter on, say, Substack, and reliant on proprietary features they built, like a comments section, it would probably be harder for me to leave, because I would be forced to give something up.
The problem, when it comes to social media, is that the networks are so broad that they are essentially going to always be at odds with these two questions. It is my view that someone who relies on moderation tools like blocking or muting would probably hate the early internet experiences of Usenet or IRC. While IRC has moderation tools, they are much more limited compared to what Twitter can do, and because there’s no real person in charge, who are you going to report problems to?
But those who raise concerns about free speech, either for reason of political stance or out of a desire to not face any sort of limits on their output, modern social networks will probably never please them, unless they move to a decentralized model. We are all at the whim of an Elon Musk-like figure coming into play and changing most social networks.
The people who own networks can change the rules at any time, and not just from a free-speech standpoint: Recently, Vimeo faced controversy over how it changed its business model to charge heavy consumers for the bandwidth their videos took up.
With centralization, the model can change at any time, and you just have to accept it—think of how Facebook would often redesign its site, disrupting the routines of millions of people.
In effect, decentralization gives you more control over the knobs, but the inconvenience of having to use them. When you hear about things like Web3, what they’re really trying to get at, beyond any financial windfalls, is the idea that a decentralized platform can be as easy to use and safe as a centralized one.
If it can be those things while scaling up to something Twitter-sized, it will be interesting to see.
If you hate what Twitter or Substack or any other centralized platform is doing, you’re kind of stuck with them unless you’re willing to make some big changes to support what you, as a consumer, are willing to do.
It will cost you connections with friends, most likely, because not everyone will be down with such a big change. It’s like moving to another city; you have to be willing to accept the circumstances of that new city.
I will admit I’m not particularly down with what Elon Musk is suggesting he might do to Twitter, because whatever it is, it is ultimately fragile and threatens something that millions of people rely on. The second he leaves, the odds that it breaks grow exponentially. Because it is centralized, it can be regulated far more easily than any decentralized protocol.
But at the same time, there is a legitimate case for centralization in social media. People want moderation, and not because they’re trying to censor people. They also want to not have to worry about putting together their own parts.
At some point, people with money don’t want to invest in a dangerous network—which is not a question of free speech but the tolerance that society has for how free you want to get.
If you want to maximize your freedom, you limit outside inputs. The best way to limit outside inputs? Work from a decentralized starting point.
That’s the side I personally land on all of this.
Find this one an interesting read? Share it with a pal.