With the new wave of users and @Luca working again on a visualization, this reminded me that I wanted to discuss a brief exploration of the I did some months ago.
In 2018, some researchers took a screenshot of Mastodon instances, I wanted to check how these evolved after 4 years.
I also tried to get a more global picture of the network to see how the different platforms interact.

So I'll talk about how many instances died (and other stuff) in a short thread. ⬇️

· · Web · 1 · 5 · 17

First off, a *lot* of them died... 75% to be precise.
75 percent!
So of course, most of these were very small instances that people probably set up just to try things out, etc, but not only: even instances with more than 100 users suffered heavy losses.

The network as a whole, though, is still alive and well, so hopefully most users had enough time to see it coming and migrate, but not all of them did, so backup your data, folks ;)

Besides down instances, I also looked at how interactions between platforms look like today.
(this is probably not a faithful representation of reality because some data may be missing and platforms may not have been sampled well)

Without surprise, though there's no good reason for it, people mostly interact within a same platform (even when they all do microblogging).
However, we do see that there are also significant cross-platform interactions so the Fediverse is still faithful to its name.

I did all that quickly over a workshop and you should take it with a pinch of salt (except for the down instances, this is solid).
But I think it does tell us a few things like:
* we really need nomadic identity so people don't fear instances going down
* cross-platform migration may help break these lock-in tendencies where people only stay and interact within one platform as they can't easily hop-on somewhere else to check things out

Despite these issues, one can see that many (hopefully most) people still manage to enjoy cross-platform interactions that hopefully suit their needs :)
So I'm really happy the Fediverse is working out for them but I do see many ways we could open up more possibilities to fedizens.

If you are interested in the issues I mentioned and more, you can find some of my rants and discussions on these topics here:


@tfardet to quantify your own nuancing: what if you filter the data to only compare instances of a certain minimum size before and after, what is the percentage of lost instances then? Maybe it's interesting to compare various thresholds, like 10+, 50+, 100+, 500+, and 1000+ urers, to see if there's some kind of "critical mass" after which these instances are more likely to become self-sustaining.

@vanderZwan that's what the right side of the figure does, if I understood you correctly (% of instances that went down for each size bin)

@tfardet ah, thanks! Serves me right for not tapping on the image (the preview only showed the left side). So it does seem to taper off in general, right?

@vanderZwan it does (fortunately!) but there's still a non-negligible (and even sometimes a significant) fraction of "large", i.e. > 100 users, instances that went down over 4 years)

@tfardet Interesting analysis. Nice to see a comparison with prehistoric times.

I hope I am allowed to nitpick among scientists and interpret the figures right: When 1 of 3 instances with around 1000 users goes down (left figure), I would say that is 33% and not 50% (right figure).

Thanks :)
As for your question, nitpicking is actually encouraged!
That being said... Mhhh... Not sure it's 1 in 3: it's logscale so hard to say... I'll try to check in the evening ;)

Sign in to participate in the conversation

Fediscience is the social network for scientists.