@lydiaconwell
> How much of a problem is that?

To be honest, I don't know. It's a potential, more than a real, problem. What do I mean by this? Well, think about how the protocol works: when someone posts something, it gets sent to all the other instances that federate with the sending one. Once it reaches an instance, it gets delivered to whoever can read it, but we're not interested about this second part. Inter-instance delievery is what's important, because it costs bandwidth and money to admins.

The way things are now, they seem sustainable enough. But what happens if a billion+ users come in and start posting? Will the system collapse? We don't know. But it could _potentially_ happen. Myself, I'm more worried about hosting costs for small instance admins than unintended DoS attacks.

> If so, that's one hell of a flaw in the Fediverse architecture.

Not really. That's roughly how any decentralised architecture works. AP is not the best protocol ever, but it's not that bad, either.

Relays may help in mitigating this problem. (They already exist. They are like middlepoints between instances. Newly created ones usually subscribe to them to fill their timelines.) If Facebook maintained a relay or a set of them (thus bearing their cost) and they could be configured so they wouldn't flood the rest of the Fediverse, everything would continue as normal. The question is: will they do that or something to that effect?

@josemanuel that's not how it works. It generally only gets sent to an instance that has at least one follower of the person that posted/boosted it. People on an instance would have to follow a huge amount of accounts hosted on meta's server for a "dos" like that to take place.

@hugot How is that different from what I said, apart from the terminology used (i.e., “at least one follower” vs “instances that federate”)?

> People on an instance would have to follow a huge amount of accounts hosted on meta's server for a "dos" like that to take place.

Yes, hence my emphasis on qualifying my assertions with “potential”/“potentially” and the following quote: “I’m more worried about hosting costs for small instance admins than unintended DoS attacks.”

@josemanuel I thought you were under the impression that once someone follows a user of another instance, their instance will start flooding it with every single toot of every single user of that instance. This is not the case, only toots of followed people are federated. So unless people drastically change their following habits for whatever reason, and everyone suddenly starts following millions of people after meta launched their platform, I don't see it happening. It is already possible for small instances to get DoSed like that, but it doesn't happen often, because people simply don't follow that many accounts. The load of an instance usually scales with the size of its user base, not with the size of the AP network. I don't think the meta instance will change much in this regard.

@hugot To be honest, I purposefully tried to keep the use of terminology short and easy to digest because I didn't want to flood the other person with technical info. I do appreciate your clarifications, though.