Federation IMO solves problems on the developer side:
1. We want to write software, not to wrangle with the laws of a hundred countries. 2. We don't want to deal with providing infrastructure for millions of users. 3. We don't want to run a community in general. Our passion is software, we don't really care what they talk about.
On the user side, the value proposition is lack of lock-in. That if you like the system, you'll eventually find some place where you fit in.
> What would be the benefit to a user of federation?
If Twitter were federated with other services, basically you'd have a way to reach select Twitter users from a saner place.
> Can I, as a user, set up my own server like I theoretically can with email? How then do we police content in such a world? What about bad actors spamming as we get with POTS and email? Who enforces things like law enforcement requests and content moderation?
People federating with you can decide to stop dealing with you in that case. Federation doesn't mean links can't be broken, or never established in the first place.
> Twitter may well die. Who cares? It'll get replaced by something else. Federation just isn't going to happen with these philosophical arguments given all the practical problems.
No system is perfect, but IMO a federated system, if done well enough is more resilient. In a federated system, the hope is that something like Twitter would be just one of many providers. Thus if the leadership goes nuts one day, it has to content with that there's 20 other nigh identical providers, and that it can over time be marginalized if it keeps being crazy.
Also, ideally federation means each community can be more focused, and have its own culture and local rules that make sense for it, but may not make sense for anyone else.
I couldn't disagree more.
First, nobody cares about this. This has no value proposition to users. It shouldn't even be a topic of conversation.
Second, IMHO it's not even true. It's a bit like how you split a monolithic into microservices. Now you've created a bunch of versioning and orchestration problems where you have to deal with network issues. Federation means talking to third-parties, dealing with them being unreachable or behaving badly. It's a whole new class of problems.
> f Twitter were federated with other services, basically you'd have a way to reach select Twitter users from a saner place.
There's a ton of hand-waving that goes along with this statement. It's like if you tried to cut out talking to Verizon on your POTS exchange. Well, technically you can do that but then you can't talk to the customers on that service and they can't talk to you.
Did you quote the wrong part? Obviously, value proposition to developers isn't a value proposition to users a lot of the time.
Second, this is a dev-centric forum.
> There's a ton of hand-waving that goes along with this statement. It's like if you tried to cut out talking to Verizon on your POTS exchange. Well, technically you can do that but then you can't talk to the customers on that service and they can't talk to you.
It's exactly like that, yes. That's the very point: that if say, Verizon starts doing something unpleasant like spamming, they can find themselves cut off from the rest of the system.
Meanwhile, unlike with POTS, any alternative is accessible to Verizon users, so they could move over with minimal pain.
The ideal is that in a federated system, there's always alternatives, and no instance is too big to fail.
There are ways to make identity floating and separate from the server, at least in theory, but ActivityPub (the protocol under Mastodon and many other things) does not implement anything like this.
We'll see if it gets there, but I find their thinking generally pretty persuasive.
Is it persuasive enough to have multiple implementations and multiple instances running? How many instances of blue sky are there running out there?
Most of the problems you pose exist with every social media platform, and only get worse as scale goes up. We can all agree that as user count goes up, abuse and moderation scales as well.
A centralized platform takes all of that moderation load at once. You need a team of people or some crazy moderation software that doesn't work. Abuse runs rampant as the company doesn't want to invest in proper moderation. We see that on twitter, Facebook, tumblr, all of them.
The fediverse aims to solve this problem by limiting scale. Smaller communities are easier to moderate. We see this with forums and chat rooms. When the community is small, they tend to self-police more. People tend to self-identify with that community and actively protect it.
But small communities are small and don't reach far. So we connect these small communities into larger networks. The implementation details of that are IMO pretty much irrelevant, but to answer your question: posts are duplicated throughout the network. Each server caches copies of all incoming posts to serve them directly to their local users. There are downsides to this approach, but the positive is that it crudely balances server load across the network.
Bad actors aren't much of a concern. If you start a server that just blasts spam into the fediverse, the network will quickly drop you. Individual servers have the option to refuse connections from other servers or domains. Server admins actively network with each other to warn of malicious servers. So far it's been incredibly effective.
As a user, you also have the option to block servers or other users. You also have extensive filtering options so that you can effectively moderate your own timeline.
And of course, there's also the drama between server admins blocking each other and harming users by cutting them off from part of the network. The solution I chose was to start up my own server. It took a weekend and operates exactly the same as the big instances I've been on.
When a server shuts down, users can move elsewhere. You can keep all your follows and followers, but lose your post history. That's obviously not ideal, but what were you able to keep when MySpace closed down? It's not perfect, but you can still keep going. Plus there are limited workarounds, but that's not important.
The fediverse has problems, but not the ones you pose.
Ask yourself how one scales a website to tens of millions of users. It's pretty complicated, right? Lots of servers and load balancers and whatever else.
But what about a website for a thousand users? Ten thousand? That's much more doable. That's just a single machine. The magic of federation is linking ten thousand websites run by volunteers that each host only a thousand users into a more or less cohesive network.
No matter which node in the network your account belongs to, you get roughly the same experience. No one has to spend millions filling up a datacenter. Most servers are run on donations from users or out of their admin's pocket. It doesn't cost too terribly much unless you're serving thousands of users.
When one server goes down, even a very large server, the network stays intact and absorbs the load and the users as they migrate. When a server with tens of thousands of users shuts down, we don't see a dramatic drop in total network users. Nearly all of them just move. This makes the network very resilient. No one person can do any meaningful damage to the network as a whole.
Fediverse is far from perfect. And I agree that it won't be a monetized ad- and celebrity-friendly platform like Facebook or twitter. But that's fundamentally misunderstanding what and why it is.
The fediverse is the next logical step from hundreds of small isolated forums. People built and use federated services because they want small communities driven by and for people who simply want to talk to other people. Nothing more, nothing less. It's not going to grow to encompass billions of users because that's simply not the goal.
The goal of the fediverse is simple: give people a place to talk and share with each other that is controlled exclusively by people who want the same thing. No money, no tracking, no corporations, just people talking to other people.
Everyone I know on mastodon is perfectly happy with the way things are. We see a very stable network with slow and steady growth. That's pretty much all we want or need.
https://mastodon.social/@jwz/109411593248255294 (safe to link to - and its following page via mastodon)
> Every time I do a new blog post, within a second I have over a thousand simultaneous hits of that URL on my web server from unique IPs. Load goes over 100, and mariadb stops responding.
> The server is basically unusable for 30 to 60 seconds until the stampede of Mastodons slows down.
> Presumably each of those IPs is an instance, none of which share any caching infrastructure with each other, and this problem is going to scale with my number of followers (followers' instances).
Is there anything besides the inbox endpoint that actually needs to be real-time dynamic?
> No matter which node [...] you get roughly the same experience.
How do you mean this? Besides HN, I'm on some big guys, a few discords, and some private groups. Some of the big guys delegate moderation to private participants in each sub-forum - of course. NO TWO of these have the same rules. Not even close. Some have fundamentally different basic rules of conduct. Many of the moderators enthusiastically want to block "the bad guys" but they mostly agree to each other's label on that in spite of their rules being all different. Basically if someone was lynched by some other group, that's good enough for them.
I'm willing to be amazed that it works but for now I do not believe that two of these systems actually will agree long term to use the other guy's forum rules.
Of course the local and federated feeds are different, but those are local to the server and will always be unique.
What you're describing is drama, which is pretty easy to ignore.
My previous server got wrapped up in some ridiculously inane large-scale drama that I couldn't avoid. So I started my own server and moved everything over. I still see all the same posts from the same people and they still see mine. It's all exactly the same except for the local and federated feeds.
It's the opposite. We have monolithic services because all of these problems benefit from scale. Is it easier to build one system for site safety and the ability to report illegal content or to build more than one?
> Each server caches copies of all incoming posts to serve them directly to their local users. There are downsides to this approach, but the positive is that it crudely balances server load across the network.
It doesn't load balance anything. It just multiplies the work, particularly given that any such service is going to have a long tail of low-distribution content. Are you going to replicate that preemptively? Or cache-on-demand? If it's the first, it's more work. If it's the second, it's not clear it actually works and certainly adds propagation delay. By this I mean, you can't really produce a feed of content you don't have, for example.
> If you start a server that just blasts spam into the fediverse, the network will quickly drop you.
Famously, POTS has no issue with spam because of this.
> Bad actors aren't much of a concern.
Just like POTS, just like email, just like SMS.
> As a user, you also have the option to block servers or other users.
Blocking servers makes as much sense as blocking a country on Twitter would.
> But what about a website for a thousand users? Ten thousand?
But it's not. You still have to deal with all the traffic from the rest of the network. Otherwise, what are you displaying to your users?
> Fediverse is far from perfect.
It's completely unworkable, offers no value proposition to users, is more expensive to implement and is way more complex.
That's not really the case. Blocking a user is done when the user is mallicious, blocking a server is done when the sysop is malicious, the server is host to bad behaviour, used for spam, etc. Domains and IP's are blocked in SMTP all the time, and the fediverse is just email with social elements attached to it.
And yet we still have thriving communities with tens or hundreds of thousands of people who are all pretty happy and satisfied with the way things are.
Because you don't understand what the value is doesn't change the fact that a ton of people do find it valuable. Calling mastodon a failed platform is laughable when we have so many people who enjoy it. I mean that literally, it's a running joke and we all laugh at takes like this. We even have memes about it.
The value proposition is actually quite simple: it's small, it's decentralized, theres no algorithm, it's wholly independent of corporate or advertiser influence, and it's full of people who value those things highly. It is the antithesis of centralized corporate social media, built by and for people who find the modern corporate web intolerable.
What you're complaining about is ActivityPub. It's bad, everyone knows it's bad, we talk about it all the time. There's a great deal of discussion and effort being put into improving it. It comes with some frustrations and odd behavior, but it's largely irrelevant to most users.
The fediverse in general has quite a lot of active users, and that count is growing slowly and steadily. Clearly these people find value and utility in these platforms. Claiming it's an unworkable platform with no value is objectively incorrect.
> Famously, POTS has no issue with spam
You should do some research into this claim.
Take Twiter as an example. It's rapidly deteriorated under Musk's stewardship [1]. Bias aside, quality is down to support a subscription models with what many now call the "blue check reply guys": low-quality replies from "verified" users being boosted when thos esame people couldn't otherwise get organic engagement.
What would be the benefit to a user of federation? Federation at what level? Is it just other companies? How do you create a feed with federated content? Are you copying it between nodes? Can I, as a user, set up my own server like I theoretically can with email? How then do we police content in such a world? What about bad actors spamming as we get with POTS and email? Who enforces things like law enforcement requests and content moderation?
All of those things are real problems/ Yet the value proposition for users is at best hypothetical. A service can still degrade. It can get swept away by whatever the new social network is. Will that be federated too? Most likely not because it's easier to scale that way.
Twitter may well die. Who cares? It'll get replaced by something else. Federation just isn't going to happen with these philosophical arguments given all the practical problems.
[1]: https://www.theatlantic.com/technology/archive/2023/05/elon-...