Big Fedi, Small Fedi - Where I Stand

 

On reading Evan Prodromou's post Big Fedi, Small Fedi my response was a glib "I'm happy with small fedi but I'm fine with big fedi too." But my feelings are rather more nuanced than that. Hence this post.

Maybe some definitions first. By 'fedi', of course, we refer to the fediverse, that is, "an internetwork of social networks," a distributed social network consisting of  'account servers' and the individual accounts that run on them (I like Prodromou's use of the term 'account servers' instead of 'servers or 'instances').

In many ways, a fediverse account server is a lot like an email server. It's a place where people have individual social network accounts just like they have email accounts, and where messages can flow from one account server to another, just like we can send emails from one place to another. We'll come back to that.

So Prodromou sets up a contrast between 'big fedi' and 'small fedi', with the bulk of his article dedicated to defining the two terms. Small fedi is, as the name suggests, small, personal and intimate. Big fedi, by contrast, is the opposite of that. But it's the details that matter, and I don't think we can simply choose between one of these options or the other. That's what my post is about.

What I'll do is go through the elements of Prodromou's definitions and comment on each as we go though them.

Big vs Safe

The 'big fedi' picture begins with the proposal that "everyone on the planet should have an account on the fediverse." By contrast, the 'small fedi' picture begins with the proposal that "The fediverse should be safe. Safe from harassment, safe from privacy violations."

Probably Prodromou would deny he is forcing us to choose between 'big' and 'safe', but it's hard not to see these being presented as the alternatives. This creates the first dissonance of the piece. Why should we be forced to choose between big and safe? Why can't we have both?

We already know the answer from our experiences with email and social media: bad actors. The same people who would send spam and track us and harass us there would try to do the same in a big fediverse. Small fedi, at least in theory, keeps our communities more intimate, which means we can keep the bad actors out.

But we already tried that with things like small private mailing lists and discussion boards (like, say, the Well) and we didn't like it. Small communities grow stale pretty quickly. There's no point continuing to contribute; everybody's heard the same arguments already. The thing descends into a single unending argument thread or a empty bulletin board for conference announcements and calls for papers.

The promise of the fediverse is something different from 'big' or safe'. It is this: a network that is open but where we can be selective. Whether or not the fediverse can deliver on this remains an open question. But it's not forcing the same choice previous technologies were forcing, between 'big, open and unsafe' or 'small, closed and safe'.

Growth

The next distinction concerns growth, with big fedi stressing "we should be doing things to make it bigger; in particular, to bring it to more people," and small fedi responding, "growth is not important." Prodromou expands: "We’ve gotten along this long with a small fediverse. It’s OK how it is, so growth is not important. Growth is a capitalist mindset."

Prodromou is being a bit tricky, conflating growth with capitalism like this. Capitalism depends on growth; without growth, capital doesn't offer a return on investment. Your pile of money doesn't 'work for you'. You are reduced to earning a living the old fashioned way, by producing goods or services. Capitalism works well so long as there is room for growth, but when a limit is reached, it begins to go sour.

This is the cause of what Cory Doctorow has so colourfully labeled "enshittification". The capitalist needs more revenue each successive quarter, and when the market for a product is saturated, and there's no more growth to be had, these increased revenues are derived (for a few quarters at least) by eating the product itself:

Here is how platforms die: first, they are good to their users; then they abuse their users to make things better for their business customers; finally, they abuse those business customers to claw back all the value for themselves. Then, they die.

This doesn't just happen to software platforms; it also happens to doughnut shops and the environment.

Obviously, nobody wants this to happen to the fediverse (except perhaps for a few capitalists who don't care whether the fediverse is destroyed in the process of making money). That is the logic of those arguing that "growth is not important". Because, typically, rapid growth requires venture capitalists, and that's what sets off the process of enshittification.

It's not that people want the fediverse to remain small. I think most supporters of the fediverse would love it to be available to more people. What they argue, though, is that the cost of investment-supported growth is too high. It has to be sustainable without growth; only then can it grow.

Platforms

The two lists appear to diverge at this point; the 'big fedi' perspective asserts that "there should be a lot of different account servers" while 'small fedi' suggests "People who aren’t on the fediverse don’t matter as much as people who are." But we are again being presented with a dichotomy.

It takes us back a few months when the subject of 'quote posts' was being debated. This is a post where you quote what someone else has posted in your own post, and then add your own comment after. The argument against them was that they basically take over someone else's post and (potentially) expose it to a much larger audience than was originally intended. There was a lot of debate about whether Mastodon should allow quote posts the way Twitter allowed quote tweets.

Mastodon founder Eugen Rochko wrote this:

I've made a deliberate choice against a quoting feature because it inevitably adds toxicity to people's behaviours. You are tempted to quote when you should be replying, and so you speak at your audience instead of with the person you are talking to. It becomes performative. Even when doing it for "good" like ridiculing awful comments, you are giving awful comments more eyeballs that way. No quote toots.

Leaving aside the question of whether quote tweets are good or bad, the main question is whether people should be able to access platforms on the fediverse that allow them to have the functionality they want. The alternative, as least as painted by Prodromou here, is that it doesn't matter what these other people want; the people already in the fediverse are happy with the way it is, and there's no need to change just to accommodate new users.

In the end, Rochko recanted, at least on this issue. But of course it was a stupid dichotomy. The fediverse already allows people to make whatever kind of platform they want. If people want quote posts, they can just build or use a platform that supports them. In addition to Mastodon, there are many many other platforms already connected to the fediverse, with more coming. Nobody is saying that there should be only one type of platform on the fediverse. 

This sort of dichotomy is created by people who do not understand that the fediverse is a protocol, not a platform. It is based on a specification, ActivityPub, that defines how different types of platforms can interoperate with each other. Nobody is saying "People who aren’t on the fediverse don’t matter as much as people who are." But they are saying, "Mastodon doesn't have to change; people who don't like it can use a different platform."

Commercial Platforms

The straightforward dichotomy being set up here is between the big fedi "commercial account servers are welcome" and the small fedi "commercial account servers are discouraged." There is of course no direct relation between being big and being commercial, but that's how it's being set up.

Commercial services are being presented here, as they so often are, as the bearers of choice: "This variety includes commercial services. If they provide the right mix of features and trade-offs that certain people want, it’s good to have them, especially if they have a lot of users" (we see this presented in the case of schools as well). But it's false advertising. There is no direct relation between commercialism and choice. Indeed, as commercial services consolidate, they tend to offer less and less choice. We'll come back to that.

The case against commercial platforms is put as follows: "Most commercial services do harm. Even if they’re on the fediverse, they’re going to try to do harm to make more money. So, they should be avoided as much as possible." That's why we have another point added to the 'small fedi' case: "If growth could cause harm, we either should fix the problem before growing, or we shouldn’t grow."

Prodromou isn't specific about the harm caused by commercial servers, but we can be here. We've already pointed to the driver of the harm above - the relentless push for profit that drives enshittification. We should also point to the mechanism, often described as "embrace, extend, extinguish". The idea is that commercial entities embrace an open protocol - like HTML, say, or RSS - and build products around it. Then they add 'features' that go beyond the original specification. These features are not available to open source and non-commercial alternatives; either they are proprietary, or depend on some aspect of a proprietary platform. At a certain point, the features become the product, and support for the original open protocol is discontinued.

Choice and diversity is exactly what commercial platforms so not want. There's no money in that. How often has Apple 'upgraded' its products by removing features? How often has Google discontinued services based on open protocols. The history is well documented. "Anyone still using the open standard loses the ability to communicate with users of the company's software, so people are faced with a decision: abandon the open standard, or abandon their friends who use the company's software."

Nobody expects commercial services to promise not to embrace, extend and extinguish (and nobody would believe them if they did). So the imperative here is to in some way protect or insulate the fediverse from commercial predation.

It's important to be clear here. People are not opposed to commercial products and services - after all, they buy computers and subscribe to commercial internet access providers. People who host their own servers often use commercial products, and even if they don't, they often depend on commercial cloud providers. But there are different models of commercial products and services. Some are investment-based and headed toward enshittification. Others - such as platform cooperatives - are product and service based, and depend on the relation between client and server.

You can't tell ahead of time whether commercial involvement in a protocol will be predatory or not. In any case, corporations are mostly just opportunistic. It doesn't matter whether there's a plan. That's why the response from the fediverse is to create a cost to anti-social behaviour: defederation. 

What 'defederation' means is that owners of individual account servers decide whether or not they will connect with and exchange messages with other account servers. What counts as 'anti-social' is defined by each individual server owner. It doesn't depend on platform: truth.social, for example, runs on Mastodon, but nobody federates with it because it is toxic. Similarly, a commercial platform trying to save money by eschewing moderation could also be defederated by many other servers. Nobody wants their part of the federation to be a Nazi bar. So they defederate without comment or debate.

The real question being considered here is not whether or or not commercial platforms will be allowed. That's always been the case, and is pretty much guaranteed by the open nature of the fediverse. No, the real question is whether non-commercial platforms can interoperate with the commercial platforms. Because the commercial platforms would really rather they didn't. It's like with open content - commercial publishers do not like competition from open educational resources and the like. I talked about all this almost 20 years ago - and the lessons still apply.

Secondary Services

Prodromou is obviously aware of the 'embrace, extend and extinguish' argument because he immediately raises the question of 'secondary services'. These aren't defined, but we can think of them as services over and above the basic services described by the ActivityPub protocol. He mentions a couple of them: automated moderation services (including shared blocklists) and content search.

On this account, 'big fedi' supports these secondary services, while 'small fedi' opposes them. Why? Prodromou describes a few of the potential harms people are worried about: "content search can be used for privacy invasion or harassment. Shared blocklists can be manipulated to cause echo chambers. Machine learning can be biased. Onboarding services favour big account servers."

These of course are problems, but the list misses the main point. None of these is really a problem if there's an alternative, especially is that alternative is a co-op or non-commercial account platform. It's when the alternatives are eliminated that the real harm is caused. And that is the danger of some of these secondary services.

I mentioned above that the fediverse is very similar to email: it's a decentralized network of individual account providers supporting a protocol that allows messages to be sent from one server to another. And I mentioned the problem faced by email providers: bad actors. 

Most readers aren't aware of the current state of email servers. Most people, if they have email, either have it via their employer or have it via a service like Gmail or Outlook. Almost nobody has their own email server, and even smaller companies can't run their own email severs. It is almost impossible to run your own email server today - it takes a lot of expertise and you have to get along with the right (commercial) services.

For example, most email servers use commercial automated spam blocking services like Spamhaus. These services maintain a blocklist - but it's way too complicated to block individual email addresses or even entire domains. So they block ranges of internet protocol (IP) addresses. If you are unlucky enough to be assigned one of these by your service provider, you're out of luck. Maybe you can be unblocked (for a fee) or more likely you'll use a commercial email sending service (like MailChimp or Postman or Mailgun). All of the entities in this value chain - from mail provider to blocklist manager to email sender - are in various stages of enshittification. 

Maybe something like this will happen to the fediverse. Maybe it will become impossible to be an independent non-commercial account server because it will be impossible to stay off the automated blocklists. 

But that's just one example of a general problem . And the general problem is this: if an individual server has to support everybody - even if only blocking them - then it becomes difficult and expensive to run that server, and so only commercial servers can survive (as least, so long as they're growing).

Search is another example (I used to have long arguments about this when we were developing the original eduSource learning object resource network back in the early 2000s). Suppose you run a one-person Mastodon server. If people want to be able to 'search the fediverse' then you need to support not just the one person using your server but everybody in your network who might want to search your posts (among all the others they're searching). That can sort of work in a fediverse of a million people, but becomes impossibly hard in a fediverse of a billion people. 

The right response here isn't that we want or don't want secondary services. Rather, it's this: we don't want secondary services that make the fediverse unusable to all but the largest providers. 

Communities

The next dichotomy is set up as a classic 'individual versus the community' dichotomy. 'Big fedi' asserts "the individual is central" while 'small fedi', we are told, holds that 'the account server is central'. 

Again, this has nothing to do with whether the fediverse is big or small. It's a matter of organization. On the one hand, to use Prodromou's description, "people should be able to set up their environment how they like, including their social environment. They have the tools to do that. The account server may set some parameters around content or software usage, but otherwise it’s mostly a dumb pipe." And on the other hand, "Moderation decisions, cultural decisions, account decisions, most social decisions should happen at the account server level."

I know the 'dumb pipe' their is attractive, and at some level, I even believe it myself, but at another, more important, level, it's wrong: there is no such thing as a dumb pipe. Just as there is no such thing as a neutral technology, and just as there is no such thing as an objective media. All transportation, information and community systems are designed, and these designs have an impact on what happens within the system. 

Moreover, transportation, information and community systems are made up of more than just their physical structure. They also involve users - the people who actually send and receive messages - and that adds an additional dimension to the pipe. Enabling quote-tweets changes how the system behaves, and so does the willingness of a population to use quote-tweets to 'pile-on' individual users. In some cases, moderation decisions have the least effect of all of these.

In the same way, there's no such thing as an environment where 'decisions happen at the account server level'. Server-level decisions have both impacts and inputs from individuals. This may be direct, as in the case of a democratic cooperative server such as cosocial, or it may be indirect, where people in an autocracy vote with their feet, as with X/Twitter. 

But even more importantly that the discussion above is the fact that the fediverse is trying to find some point in the middle, creating what might be described as communities of communities. What distinguishes the fediverse from a traditional social media site is that there is not one degree of separation between each person on the network. Even if it were possible to traverse the network (and as outcasts like truth.social show us, it's not) it would require traversing intermediary servers (in this way the fediverse is much more analogous to DNS than anything else).

Decisions about accounts, culture, moderation and the rest don't happen in any particular place; account servers that have an affinity with each other federate with each other, forming loose (and open) communities of servers; similarly, individuals may have an affinity with other people on the same account server, but they also form communities with people in other federated servers.

I've written about this before, in a very different context, under the heading of 'groups and networks'. At one extreme - the 'big fedi' model, we have individualism (or atomism, etc). At the other extreme - the 'small fedi' model - we have groups. But in the middle - where there is cooperation rather than competition or collaboration - we have networks. And it's this middle point toward which the fediverse is aiming.

The fediverse is a hopeful and idealistic alternative to the models of community we grew up with. Between the alternatives of "being told what to do" and "each deciding for themselves" there is the possibility of "coming to agreement at points of mutual value". We can have something that is neither capitalism nor communism. A cooperative commonwealth federation (CCF), as it were.

Conclusion

Prodromou runs through a few more features of both 'big fedi' and 'small fedi', but they're mostly variations on the themes we've already explored. At a certain point, his characterizations of big fedi and small fedi become caricatures.

Maybe he's describing the case for 'big fedi' accurately. But I don't think he understands the motivations for the arguments he characterizes as 'small fedi'. It's as though he pictures individual account servers as 'tiny Twitters' or something, and doesn't really get how there can be subnetworks of interconnected servers forming communities of communities. It's as though he doesn't understand that the need for 'small' isn't simply to keep things 'human' but to keep things affordable, so that we can actually have some scope for choice and diversity.

And most important, I think, is that he fundamentally misunderstands the logic of choice, which is this: it is orders of magnitude easier to choose what you want, than it is to filter out what you don't want.

In federation, connections are much more important than search. They are much more important than algorithms. It's not that search and algorithms are bad, it's just that they were designed for a different purpose. What we want comes to us; we don't need to find it. It's in many ways similar to having a list of RSS feeds; there's no need to filter out the spam, because there was never any spam to begin with. Federation is about choosing and connecting to what we want, rather than connecting to everything and filtering out what we don't. A community isn't about being in the same place and following all the same rules and reading all the same things - it isn't about sameness at all. It's about being connected to people you feel a connection with, easy person with their own set of connections, where servers are set up to help us in this, rather than creating environments that make this hard.

We're still figuring out how to make this work. That's why people are understandably concerned about being swamped by an influx of new people. But hey, it's OK. We'll be fine.

Image: CBC