fosdem 2026 🗺

So I made it to fosdem for the sixth time. Last year I never wrote anything about it but from 2 years ago there is a nice report. There is also a short guide in there about how to survive the days at fosdem so I won’t repeat that again.

Day 1

What is interesting to me changes from year to year but one thing that has been with me a lot over the last couple of years is the idea to have social networks that are not sponsored by corporations. Similar vibes as to the early days on the internet but more sophisticated. Thus I spent most of the first day in the Social Web room. Many of the talks were mostly about advocacy for the fediverse. Nice in itself but somewhat repetetive over the day.

First out was Matthias Pfefferle - the creator of the Wordpress ActivityPub plugin - with the talk Democratise the fediverse. Large territories of the internet is powered by Wordpress so here is a great opportunity to get critical mass of adopters.

Then came a somewhat similar talk by Hannah Aubry and Andy Piper from the Mastodon organization called Tending the herd: Community at Mastodon. They have received some critic for pushing so much for their own server (mastodon.social) on the mastodon home page. It is a tricky issue since making it easy to sign up and get started is important for the fediverse. Having to choose a server may be a hurdle too high for many normal users. Thus they have kept it. Also - as I understood it - for reasons of having a high load server to make it easier to see how the software behaces under pressure. In this talk they announced that they are working towards pushing for a “best match” server instead based on the users language and location. Opt-in on the servers end. Good news indeed.

I sneaked away from the room after this talk and listened to the Mercurial 20 years and counting talk about the state of mercurial version control. For me it was the first distributed version control systsem I tried back in the days and it still seem to have some advantages over git. Maybe try it out again at some point.

Back to the social web room for three 10 minute talks.

First out Evan Prodromou presenting tags.pub a new take on the problem with tags. As it is now the only posts tagged with something you get to see are those that has passed by the server you live on. With this new approach each tag is a user on the activity pub server tags.pub. So instead of following a tag on your own server you can follow this user that collects all the relevant posts over all (or most) servers.

Then Paul Fuxjäger about Increasing Long Term Stability of Relations Between Fediverse Identities using SSI. Obviously it would be good to have long running identities not directly connected with specific servers. But mostly my take away from this talk is that this is something I would like to learn more about.

Finally in this cluster of talks Django Doucet about the client to server ActivityPub that exists but is barely used. Each platform has built their own API for theis own needs. Normies need their apps. But what if one client would be all that we need? Utilising the standard that was supposed to solve this. Not without concerns this one. I mean this client would need to be able to present and interact with a multitude of different content types.

More talks happened but I will skip on to the Bonfire talk. This is an interesting effort to build a federated space for communities. The design is very modular so the idea is to pick and choose what things to build with and on. Would have been interesting to see a little more hands on.

The last talk of the day for me turned out to be Hong Minhee about Fedify: Building ActivityPub servers without the pain. So he started ouy building his own singel user acivity pub server but realised that there is so much pain building a thing like this. So all the pain was extracted to this little framework called Fedify.

Day 2

I started the day with two talks in the Declarative and Minimalistic Computing room. The first one was about a programming language I have barely heard about before: crystal. A language for humans and computers by Johannes Müller. Turns out that Crystal started with the idea of getting a statically compiled Ruby. So some things added along those lines:

All in all looked pretty cool and is something I may want to try out at some point. Although Gleam is still on top of that list.

The second talk was about Building a minimal cross-platform terminal UI library by Thijs Schreijer. Quite cool how they built something on top of Lua and obviously had some issues with Windows.

Then I tried to get in to the DNS room but it was full.

Plan B can often be a main track and also so this time. I went into La Fontaine to listen to Sal Kimmich about The Hidden Life of Infrastructure: How Control Moves Through Code, Chips, and Nations. This was one of these talks that used to be a couple of hours and was cut down on-the-fly during the talk. The result was quite fragemented although the content was really interesting. I would have loved to hear either the long version or a more curated one hour summary. Still a super interesting topic about the entirety of infrastructure. The talk was organized around a book by the speaker. Perhaps I’ll pick it up.

Then I tried fo the in to the DNS room again but it was full.

Then I tried to get back into La Fontaine for the talk about reverse engineering the spotify protocol. But it was now full.

I then decided to get back to the Declarative and Minimalistic Computing room for an interesting talk about DSLs in Lisp. But it was full.

Three full rooms in a row is my maximum so I retired back to the hang out areas in the middle to write these words.

Another attempt - now I was early to the Retrocomputing room and managed to grab a chair. A nice talk about rewriting 60 year old Eliza in Javascript by Steven Goodwin. He seemed familiar so I have probably attended one of his revious 28 talks at fosdem. The original Eliza was not written in Lisp but in some other old language called MAD something. A really compact thing (that made me think of APL for some reason) that required 5-10 times as much Javascript for a line-by-line port. And the problem with the lack of a GOTO mechanism in Javascript.

Then off to Janson for the two last talks before heading towards a train out of Belgium. The first one was about the geopolitics of code. At a time when open source has made it global - contributors are now from all over the world - the political situation has become less global. It is unfortunate and may result in defragmentation yet again. Is it wise to store open source software on github - a clearly american effort? China is building concurrent infrastructure in order to be independent of the west. What can Europe do? Be a positive force in this mess?

Finally the last talk I attended was about esoteric programming languages called The Hacker Folk Art of Esoteric Code by Daniel Temkin. A really great topic to end the conference with. The usual suspects like brainfuck and piet. But also a pretty new one by the speaker called rivulet.

Ending thoughts

As it is this conference is hectic and it is hard to expect anything from it really. And there is always something unexpected happening. It is both tiresome (all these people just being in the way and occupying the interesting rooms) and energizing (all these people being there with so much energy about open source). This year I managed to attend 18 talks and they were all nice to listen to. Probably reflecting where I am - it turned out to be a bit too much open source advocacy and a bit too little tech stuff. I made up with that by doing some coding in the middle of the second day.

Over and out and see you next year!

written by fredrik at 2026-02-01

Related posts