Though most instances are very small (in user count), the large instances are very large by comparison. The result is that the 20-30 largest instances host around half of all the users of the fediverse.

This graph is a cumulative percentage of all users starting with the largest instance and descending. By 20th largest, we've got 50% of all users. Mastodon.social hosts ~16%! The top 10 get you ~40%. Note that this includes 2 large japanese instances (mstdn.jp+mastodon.cloud)

woo hoo. i'm in the 0-2 user level. 30% of us are like that? really? cool.

@anant @elias
That’s 30% of instances, not users. See the rest of the thread.

yup was considering instances only. Obviously it cant be 30% users although i would love it but with massive instances hosting massive number of users up top thats unrealistic to expect. Good thread indeed

@gorfram
Don't know exactly, but it is a sub-field of science to some extent. Geoffrey West is a former physicist known for studying such things ... look him up and you might find some resources.

His general finding was that power laws (which seems to accurately describe the relationship between the number of instances and their size) pop up all over the place in biology and cities.

The graph x axis labels are confusing. I'm assuming 0 to 2 is actually 0-1 and the next bin is 2-49?

@gorfram I believe they refer to it as a Zipf distribution.

@m @gpollara interesting. Do you follow anyone back?

@m

Very interesting, thank you for explaining. To clarify though, this only applies if you want to be discoverable in Federated timelines.

Presumably your followers will always see any new post from you in their personal timelines, independent of the instance you're on.

One further question though, how about if you use a hashtag someone follows. Will anyone in Fediverse see that, or would they only do so if their instance was connected to yours?

Thank you

@maegul @WestLawns @fediversenews

@dadalo_admin yea absolutely. I haven’t done anything here to account for bots, just basic analysis of the raw data.

How are you determining whether an account is a bot on mastodon.social or other instances? I know some have moderation policies to distinguish them but is that sort of thing typically discoverable?

Content warning: Mastodon structure Click to open/close

@m @gpollara Wow I never knew that's how this worked and I've been here for years! I always struggled to explain why the public timelines looked so different on different instances, that makes so much more sense now. Thank you!

@m @gpollara thanks - it is interesting to better understand these things.

@m @WestLawns awesome explanations to all this, thank you. Learnt something new but can also see how this can feel too much to get head around for the average member of public!

Search is mastodon's biggest weakness. If I search for a hashtag, I would like it to find any post with that hashtag on any mastodon server. But I can't. I know many are aghast at the new onboarding process pushing everyone onto mastodon.social, but it is the only thing which will help new users stay. At least it gives people a chance to find others based on interest. I really wish this was a better experience

@Quentin @gpollara @m the absence of Musk (and hate speech) is one of the greatest strengths.

@Quentin @Gabriele Pollara @m@thias.hellqui.st @maegul @West Lawns The "limitation" of #search to what's known to an instance isn't a political decision. It's a technical limitation that can't be overcome, at least not easily.

Even a search that covers literally 100% of all #Mastodon instances, only Mastodon, as in every last instance that exists the very moment you search for something is technologically impossible. If you search for the hashtag #HelloWorld, and you want to search every last Mastodon instance, you'd also have to be able to search a brand-new, single-user instance that has been started up for the first time two minutes earlier, and where the admin has just dropped a test toot that happens to contain #HelloWorld. At this point, the other Mastodon instances don't even know yet that this new instance even exists.

In order for this to work, every new Mastodon instance must actively push its existence and all of its content to each and every other Mastodon instance that exists at this point. Including one which, in turn, had just been started up two seconds earlier. How is a new Mastodon instance supposed to know about the existence of another instance that's only two seconds older?

But let's take this even further. If you only want to search Mastodon, you'll miss out on a great deal of posts in the #Fediverse. For #MastodonIsNotTheFediverse. The Fediverse isn't only Mastodon. There are many other projects out there, all of which communicate with one another and with Mastodon.

So instead of searching "any Mastodon server", you'd have to search "any Fediverse instance", no matter the project. This would have to include any hypothetical FoundKey instance that has just been spun up for the first time a few seconds ago. And if instances would have to advertise themselves to all other to make Fediverse-wide search possible, this new FoundKey instance would have to know right away about a (streams) instance that was just launched half a second earlier somewhere else, a Mitra instance that's one second older and a /kbin instance that's two seconds older.

I think it's obvious why this will never work. I hope so.

@Quentin @gpollara @m @WestLawns I agree, instances do matter.

@WestLawns @gpollara @m
Oh I'm not doubting that, I'm just saying it's harder than it could be to find the *nice* things you might be specifically looking for 😀

@jupiter_rowland @m interesting explanation, which then implies that basically discoverability on the Fediverse will always be an evolution, rather than instantaneous.

An analogy in my mind is that the Fediverse experience is more like an exploration / expedition through other Mastodon instances & other Fediverse services, along the way picking up people and hashtags of interest that will cast the net wider and wider for you over time.

@Gabriele Pollara @m@thias.hellqui.st I think that fits it quite well.

People start out on the Mastodon instance they've been invited to, believing that this instance is all of Mastodon. Then they discover decentrality and other instances. Then they discover what a wealth of instances Mastodon has. They may actually move for the first time.

After a few months, they stumble upon parts of the Fediverse that aren't Mastodon, down to unexpectedly communicating with someone from entirely elsewhere. Once they follow @Fediverse News, they're really being enlightened, also because the most prolific poster is a driving force behind the success of CalcKey.

@jupiter_rowland @m I gathered it was more technical than political - although, the limit of reach to only your friends' friends, does have the potential to form echo chambers. Ok, so on the web, if I get myself a brand new domain and create a brand new website, no-one knows about it yet, BUT, Google's and Bing's and whoever else's little bots will crawl around and find it eventually, and then anyone who searches for my name on those engines will find me. Could someone create a Fediverse search bot that crawls around and finds new servers, or however it works?

@Quentin @m@thias.hellqui.st Stuff like this has been attempted and promptly been rejected. There are two reasons against it.

One, it has to be centralised. As in, there would have to be exactly one central instance of this search engine. The whole Fediverse which is still decentralised would depend and rely on it. It basically could barely exist without it anymore. Not only Mastodon, but the whole Fediverse, all projects would depend on it.

But decentralised networks depending hard on central infrastructure is dangerous. If that one instance goes down, the whole search goes down. If Jack Dorsey managed to take over the search engine, he would have complete and barely avoidable exclusive control over the whole Fediverse. Ditto Elon Musk, Mark Zuckerberg, Jeff Bezos or whoever.

This is why decentralised networks who have decentrality as a main selling point must never rely on something, anything central.

Also, if you want it to be as powerful as Google, it can't be something created and run by someone in their spare time with zero budget. It would have to be a commercial start-up funded with venture capital which therefore has to find a way to make money because the investors expecte a ROI of 10:1. So it'd either be pay-to-use or spam everyone's timelines with advertising or sell people's data to Google, Facebook and other data brokers without telling them, much less allowing them to opt out.

The other point is heavy resistence coming from the LGBTQIA+ and BIPOC communities who have fled Twitter. They reject Fediverse-wide full-text search altogether, decentralised or not.

Over on Twitter, there are extreme right-wing users who use Twitter's all-encompassing full-text search to find gays, lesbians, trans persons, POC etc. just to harass them. This actually happens. This is why these people have fled Twitter. And this is why they absolutely do NOT want to have this feature in the Fediverse.

For these two reasons, whenever someone tried to introduce some decentralised search engine for the Fediverse, it was promptly and massively fediblocked by countless instances and thus rendered useless.

@m @jupiter_rowland Hey I just want to find other people who share my weird hobbies and interests - and mastodon makes that next to impossible unless said interest is something with several really large instances devoted entirely to it.

@Quentin The Fediverse isn't Twitter. There is no secret-sauce algorithm that automagically (allegedly) presents to you what you're interested in on a silver platter.

The key to discoverability are #hashtags.

If you want to be found, use hashtags. If you want to find someone, look for hashtags. Hashtags in your profile, hashtags in your posts. Don't sit there and wait for Mastodon and the whole Fediverse to become like Twitter so you can continue to use it like Twitter.

Unfortunately, exactly that is what many Twitter refugees still do. They aren't used to using hashtags, so they don't. Instead, they expect some hidden magic to bring them together with like-minded people.

But the first step is to use hashtags yourself so that others can find you more easily.

There are also user directories which you can join if you want to and where you can find other users, for example fediverse.info which has the advantage of not only covering Mastodon.

Generally, in the Fediverse, you can't just consume. You have to partake actively. If you want to be found, do something for it.

@Quentin @jupiter_rowland
Your server should know every other server in the Fediverse, so that is a lot of servers you are searching for a #tag
(I am pro discoverability BTW)

That's all well and good, but if I search for a hashtag, I will still only see posts with that, by people on my server or that my server knows - that does not make discoverability easy.

@Lee 🌏 @Quentin So you say if someone starts up a brand-new FoundKey instance for the first time right now, and two seconds later, someone else starts up a brand-new Hubzilla hub for the first time, these two instances should immediately know each other? And literally all other instances of all other projects in the Fediverse without even a single exception should know both literally instantaneously?

I'm curious about how this should work. Especially since the Hubzilla hub doesn't even have a single ActivityPub connector running yet (ActivityPub is optional per hub and then per channel).

@jupiter_rowland
Hey Juipter
I'm absolutely not the expert on this. I've been asking lots of questions and finding my way.
How Fediverse Servers find each other is something I've been trying to learn. They obviously do, but how quickly and how successfully remains a mystery to me. I plan to install Activity Pub, WordPress plugin, and I am most curious to how people here on Mastodon will find my #tags.

@MrLee @jupiter_rowland It takes two things. Boosts and hashtags.
The boosts will get your posts into the timelines of whoever follows the person boosting your initial WP post. It also gets it into the Federated timeline of the instances all those followers are on.

@MrLee @jupiter_rowland It’s possible (but unlikely) that someone will stumble across the post by browsing the federated timeline, but on most instances that’s just too noisy and is ignored.
Hashtags act as a filter, searching through the federated timeline (they can only match posts that have already made it to your instance through other means - boosting).

@MrLee @jupiter_rowland So you want to get your post boosted by someone well connected with followers on lots of different instances and have a strong hashtag game.

@MrLee @jupiter_rowland For simplicity, I’ve ignored relays. They may help your post propagate, but won’t actually get it in front of anyone’s face. It will help people searching for the webfinger address of your blog as their instance may at least have heard of it and cached it.

@MetalSamurai
If you're subscribing a #hashtag and your server sees a tagged post through a #relay, it will land in your #feed.

@MrLee @jupiter_rowland

So the normal is to have negligible communities
@TheKinrar @fediversenews