There's a claim going around that Mastodon is "centralizing around a few instances" because 1% of instances account for 84% of users. https://mathstodon.xyz/@manlius/109383027753990134

That's misleading. 1% of instances is ~360. That's a lot more than I expected! Impressively decentralized.

Measuring centralization as a % doesn't make sense. Hypothetically, if 100K new people each started their own instance, centralization as measured by this stat would greatly *increase*, even though that's not what actually happened.

Manlio De Domenico (@[email protected])

Attached: 1 image Again, a quick and dirty analysis, but if I have no bugs then the 1% largest #Mastodon instances accounts for 84% of all users. The top 5% accounts for 97% of all users. We can say that the system is effectively centralizing around a few instances, and this might be a problem for the overall stability and sustainability. @[email protected] @[email protected] @[email protected] @[email protected] @[email protected] @[email protected] #MastoStat #ComplexSystems

Mathstodon

The goal of decentralization shouldn't be a uniform distribution of instance sizes. That's unnecessary and unrealistic.

Decentralization prevents any one entity from exerting too much power over the whole ecosystem, allows cultural differences between communities, and enables semi-independent technical experimentation by instances so that successful innovations can percolate. These are the things that made the web successful. Mastodon has all of these, and isn't in danger of losing them.

One question is about how the current level of (de)centralization will change. I think two divergent paths are possible.

1. Instances continue to be volunteer run (and, in my ideal world, organizations like universities and media outlets will run instances for their members/employees). In this case, decentralization will continue and even increase.

2. There's an influx of commercially run instances. If this happens, there are inevitable centralizing trends — see how Gmail dominates email.

@randomwalker Agreed. There's even a certain 'verified' status to some servers that is probably prized. Like if you are on an official university's server, NASA, huge company etc. that only lets actual employees or members sign up then that's basically like having a blue check.
@randomwalker guess it'll depend on how differentiated the experiences between volunteer-run and commercially-run instances become (assuming that "commercial" refers to instances that seek to acquire new users and profit from running an instance, as opposed to just being an official organization instance) . IMO there's lots of room for a variety of instances as people have different kinds of affiliations they might want — eg geographic region / hobby / professional organization

@randomwalker I think you might have under-analyzed the likely #evolution of the volunteers in world 1. They're going to be confronted with increasing hosting and especially moderation costs, which I think will drive them to professionalize and to make it harder for non-professional instances to federate.

That is, existing commercial entities don't have to adopt #ActivityPub for it to centralize; the volunteers could adapt in centralizing ways too.

@jyasskin @randomwalker
I think that's exactly correct, and part of the problem with any approach that relies so much on the server level for social and technical infrastructure, including moderation and #ActivityPub.

Mastodon and all decentralize globally, but they still centralize to the instance, putting scaling costs there, and now with more inefficiencies from the lack of global coordination.

@randomwalker It happened in XMPP world, Facebook and Google refused to federate w each other and so their chat apps became walled gardens.

@randomwalker I will fight to make 1 come true (and I think we're on the right track).

I honestly don't think commercial instances have a real future. Having a huge instance is costly and aside from crowdfunding there is no real way to monetize.

I also think, as instance "owners", we have to try and promote all the instances and celebrate decentralization, instead of putting ego and numbers before the good of the fediverse.

@randomwalker True! The power law effect will likely appear over time but the fact that anyone can start a new instance to facilitate the exchange of ideas makes it akin to BitTorrent.
@abhi24 But a big instance can refuse to federatewith these new upstarts. It has happened before, to wit XMPP.

@randomwalker it feels like the measure for decentralisation is simply "how big is the biggest instance?" (or top 5 instances).

Another metric would also be how easy it is to move your profile to another instance (for any reason). So far I understand it is fairly easy and might prevent any instance to force/capture users. Until custom features are introduced, this should be ok.

@randomwalker And of course if another Elon manages to buy the biggest few instances, migration will be a lot easier!
@randomwalker I think it will change over time as people realize and gain the confidence to create their own instance. I'm considering it and I know quite a few people have started their own. It takes a while for people to understand this is different and we don't need to all stay on the same big instance we joined on day 1

@randomwalker same thought. There is always going to be some „preferential attachment“ when it comes to those kind of systems.

The largest instances are currently closed for signups, which actually makes it more complex for new people to join, but will actually lead to more decentralization than what would „naturally“ have happened.

Imho mastodon is currently way to concerned about decentralization.

@randomwalker But totally agreed that % of instances is the wrong way to put this.
@eckles Sorry about that! I didn't see that comment thread and got the 36K figure from instances.social.
@randomwalker But yeah totally agree with your point that this is a good number of instances and that's what's important

@randomwalker Wow, yeah, that is a terrible analysis. It's disheartening how many enthusiastic replies it has, with little pushback on this.

Small correction: the author clarifies in one of the replies that they looked only at the 12k active instances, not the 36k total. So 1% is about 120 instances, not 360. Still a lot!

(And the author should of course have been clear about the absolute number from the beginning.)

@randomwalker The underlying data are also incomplete. The source https://instances.social maintains a self-reported list of open Mastodon instances. The analysis excludes many open and almost all closed instances.
Mastodon instances

@randomwalker It's true that it's not a good quantitative value to assess stability to make a pass/fail judgment on the current state of things, but it does say something that we already know: That there is a handful of very large instances that have gotten most of the attention from the influx.

Not exactly news, I know.

Whether one thinks this is a real issue depends on where one stands on things like moderation, and defederation with un(der)moderated spaces.

Ultimately, it's probably bad for the social health of both the instance and the networked communities to have people so densely packed, but it's really good for outreach to have marquee hosts, particularly with the Mastodon name in the URL.

@randomwalker *waves from his 2-person instance*
@randomwalker What about the Gini index or Theil index?

@randomwalker I don’t see it really taking off any other way. No one wants to be given too many choices.

In an ideal world where it takes off, it stays open like email but also centralized to some good large hosts…like email. But always with the option.