About This Gallery

This is a gallery of all the statistics I’ve gathered on the Monsterdon watch party.

How does it work?

For all data since December 2024, I collected it myself using the methodology described in detail in the code. All data prior to December 2024 comes from @combatwombat who collects the toots for Monsterdon Replay.

The basic idea

Everyone is on some server. I’m on infosec.exchange. My program logs in as me and pulls all the #monsterdon posts it can find. I look through all the posts I get, and I discover servers of all the other participants. I then go query that server directly to see what it has seen. If I find mastodon.social for example, I go connect to mastodon.social and start pulling down everything that they saw for #monsterdon. And I read all the posts I get from there, and I make a list of all the servers that I have discovered, and I try each one in turn.

For example, there might be someone on jorts.horse that I don’t follow and whose posts don’t make it to my server. But I connect directly to jorts.horse and it gives me the posts, which includes the ones from this person that I don’t normally see. Also, when my server gets a post from a different server during the event on Sunday night, it gets a snapshot in time. Here’s how many boosts and favourites were on it at that moment. It might not fetch an updated copy. Since the post isn’t from my server, my server might have out-of-date numbers for boosts, favourites, replies, etc.. When I connect to jorts.horse on Monday, I get the latest, authoritative version of that post, from the server that knows it. So my Monday morning statistics use the mosts authoritative numbers available to me.

Finally, there are some servers (including Taweret’s timeloop.cafe) which don’t federate with infosec.exchange (for reasons lost to history) and many don’t offer a public API for searching. Mastodon is not the only software out there. Small, personal servers like snac2 and goto social have different APIs, though they speak ActivityPub. Some won’t answer an anonymous query. (That’s fine. Some folks don’t want to be too public, it’s ok. I send them one query every Monday, they say “not allowed” and I leave them alone.) So I might get 20 or 30 copies of a post from @Taweret as seen by various other servers. I keep the version that has the highest numbers, and I throw away all the other copies.

How much data is it?

A typical Monsterdon is anywhere from 2100 to 3300 posts. I capture starting 1 hour before and running to one hour after the event. About 55 servers typically answer the anonymous requests, and It’s not unusual for me to pull down 80,000 posts just to get the 2100 unique posts. So I get TONS of duplicates. It takes anywhere from 35 minutes to 2 hours to pull all the stuff down.

I get rid of fields I don’t use, so I end up retaining about 15Mb of JSON for the whole event. @combatwombat stored pretty much the entire ActivityPub data structure for everything. So that data set is 310,264 toots (from late 2022 to February 2026) and is about 1.3 Gb of JSON.

About the Software

You can see the code that produces these graphs and graphics at mastoscore.

About Paco

I’m a computer geek, cigar smoker, information security person, and bass guitar player. If you want to reach me, I’m on Mastodon at @paco@infosec.exchange.

Stuff I do:

  • My blog. Whatever stuff motivates me to write.
  • John Mastodon. My little tribute to John Mastodon.
  • nova.org. A micro ISP that I run in Northern Virginia, USA.
  • my peertube. A few random videos and things I have posted.
  • www.carecure.net. CareCure: a web forum that provides information and community for people with spinal cord injuries and traumatic brain injuries.