FOSDEM 2024 in Brussels

FOSDEM 2024 and I was there!

What is "FOSDEM"? In this blog post, I would like to tell you more about my visit there. FOSDEM (Free and Open Source Software Developers' European Meeting) is the biggest Open Source event in Europe, running in Brussels since 2000, with around 5000 participants each year. No one knows the exact number since entrance is free, and the best guess is only based on WiFi device counts. Since many people are there anyway, several Open Source projects have satellite events scheduled around the conference weekend. For PostgreSQL, that's the FOSDEM PGDay on the Friday before FOSDEM starts. On the Thursday before that, there was a PostgreSQL developer meeting to which I had received an invitation, so I decided to go there for the full set of events.

Not easy to get there...

Travel started with a replacement bus for me: The train from Cologne to Brussels got canceled. So getting there was a bit rough, but I made it despite only being two hours late. When I arrived in Brussels, I met other PostgreSQL developers. We've decided to visit a nice restaurant at the Grande Place in the city center. Some food and some beers later, we were ready to go for the conference.
On Thursday, the Belgian farmers were on strike in the city, and navigating the bus transportation was a bit difficult, but I made it to the conference hotel.

Key Discussions and Presentations

The meeting started with Vik Fearing presenting the new graph features of SQL:2023 and a discussion of how PostgreSQL can support it. After that interesting start, Dave Cramer had a proposal for how to better support binary output format in the PostgreSQL query protocol, but that was deemed to be better solved on a different level.

Devrim Gündüz then brought up the topic of building PostgreSQL with Meson, and how that interacts with packaging. The main problem at the moment is that the Meson build system does not support LLVM yet, so that's still out for actual production use. (But it is already very handy for development since it's really fast).

Jeff Davis then brought up what turned out to be the number one topic for me on most of this weekend: PostgreSQL's handling of locales and collation and the interaction with operating system upgrades. Jeff proposed to have built-in providers for the C and C.UTF-8 locales, which doesn't solve all of the problem, but might be a valuable building block in getting closer to a solution. The discussion then continued over lunch, evaluating different approaches to the problem space.

Another topic was how new contributors get listed on the PostgreSQL contributors pages. Incidentally the whole team dedicated to that question was present (Stephen Frost, Joe Conway, myself), and we explained how the process works. We will likely add an explanation to the web pages since that seems to be not so well explained there yet.

Peter Eisentraut brought up the idea of having better-defined support windows for libraries that PostgreSQL depends on. Currently that information is scattered around several places in the documentation, and doesn't always match the actual values enforced by autoconf or the code. The idea is to track what the oldest RedHat, Ubuntu and Debian releases still in "normal" long-term support are shipping, and cut off things when they drop off that list, as opposed to making ad-hoc decisions in each case on whether code has to support very old library versions. Stephen Frost discussed extensions of PostgreSQL's page format to make room available for per-page extra data like crypto initialization vectors or txid epochs.

Quite some time was spent on triaging long-standing patches that have been in the commitfest for several years. I picked up a few of them that I will do reviews of over the next few weeks. A few ideas of how to improve the patch submission process were poked around. After a long day, we spent the evening together with a group of hackers, still discussing collation problems and LLVM compatibility.

Time for PGDay and FOSDEM conference

On Friday, the PGDay itself took place. Joe Conway presented a compatibility library that AWS has extracted out of glibc to make older collation versions available on newer RedHat versions. The approach is nice, but I think we need to work on making upgrades easier rather than sticking to old versions and just postpone the problem. In the hallway track, Jeff Davis and I were discussing more ideas on what built-in collations could look like to best solve the practical problem. Overall, I didn't attend many talks, but spent most of the time chatting with other hackers.

On Saturday, I was in the audience of the actual FOSDEM conference. In my impression, the hallways were crammed with people as usual, but several others said they felt there were fewer people than in other years. The PostgreSQL booth was always surrounded by people asking questions and picking up merchandise. On the FOSDEM main track, I went to see a talk on the upcoming EU Cyber Resilience Act (CRA). One of the authors of the law was presenting, and I was delighted to hear that most concerns voiced by Open Source projects were actually heard and revised in the text. Another talk was by Bradley M. Kuhn on the GPL, but I didn't take too much away from it.

On Sunday, the PostgreSQL devroom (what FOSDEM calls the conference tracks) took place. I attended talks by Charly Batista (PostgreSQL data types) and Umair Shahid (PostgreSQL clustering), but as the program was more aimed at a general audience, I wandered off to other tracks. One interesting talk I attended was on the Linux kernel keystore that can be used to securely store passwords, which we will use for CYBERTEC's Transparent Data Encryption (TDE) in the PostgreSQL Enterprise Edition.

My conclusion about this long weekend that it was a very nice event. Of course, I will also be taking some thoughts and ideas with me. See you next year in Brussels!