PostgreSQL Developers Converge on Brussels for FOSDEM 2024
FOSDEM — the Free and Open Source Software Developers' European Meeting — has been running in Brussels since 2000 and remains the continent's largest Open Source gathering, drawing roughly 5,000 attendees annually. Exact numbers are impossible because entry is free and estimates rely on WiFi device counts. The event's gravitational pull also attracts satellite events: PostgreSQL's FOSDEM PGDay lands on the Friday before the main conference, and this year a PostgreSQL developer meeting was scheduled for the preceding Thursday. That full itinerary is what brought me to Brussels.
Travel Troubles and Developer Meetups
Getting there was the first obstacle. My train from Cologne was canceled, and a replacement bus left me arriving two hours late. Once in Brussels, I connected with other PostgreSQL developers for dinner near the Grande Place. The next morning brought a new challenge — striking Belgian farmers disrupted bus routes — but I made it to the developer meeting hotel.
The developer meeting opened with Vik Fearing presenting SQL:2023's new graph features and discussing how PostgreSQL might support them. Dave Cramer followed with a proposal to improve binary output format support in the query protocol, though the consensus was that this belongs at a different layer.
Devrim Gündüz then raised the topic of building PostgreSQL with Meson and its packaging implications. The main blocker: Meson still lacks LLVM support, keeping it out of production use — though its speed makes it attractive for development workflows.
The meeting's standout topic for me was Jeff Davis's presentation on PostgreSQL's locale and collation handling, particularly around operating system upgrades. He proposed built-in providers for the C and C.UTF-8 locales — not a complete solution, but a useful building block. The discussion continued through lunch and into the evening among smaller groups of hackers.
Other agenda items included how new contributors are listed on the PostgreSQL contributors pages (the entire team responsible was present — Stephen Frost, Joe Conway, and myself — and we'll likely add explanatory text to the web pages) and Peter Eisentraut's idea for better-defined support windows for dependent libraries. Currently this information is scattered across documentation and doesn't always match what autoconf and the code enforce. His proposal: track what the oldest RedHat, Ubuntu, and Debian releases still in normal long-term support ship, and cut off support when libraries drop off that list. Stephen Frost also discussed extending PostgreSQL's page format to free space for per-page data like crypto initialization vectors or txid epochs.
A significant chunk of time went to triaging long-standing commitfest patches. I picked up several for review in the coming weeks. The day closed with more discussion, still circling back to collation problems and LLVM compatibility.
PGDay and the Main Conference
Friday's PGDay featured Joe Conway presenting a compatibility library AWS extracted from glibc, which makes older collation versions available on newer RedHat releases. The approach works, but I believe the real goal should be making upgrades easier rather than postponing the problem. I spent most of the day in hallway discussions with fellow hackers, particularly with Jeff Davis on what built-in collations could look like.
Saturday was the main FOSDEM conference. The hallways felt as crowded as ever, though others mentioned it seemed quieter than previous years. The PostgreSQL booth stayed busy with questions and merchandise pickup. On the main track, I attended a talk on the EU Cyber Resilience Act from one of the law's authors, who confirmed that concerns raised by Open Source projects were incorporated into the text. Bradley M. Kuhn's GPL talk didn't offer much new for me.
Sunday's PostgreSQL devroom — FOSDEM's term for conference tracks — skewed toward a general audience. I caught Charly Batista on PostgreSQL data types and Umair Shahid on clustering, then moved to other tracks. One notable session covered the Linux kernel keystore for secure password storage, which we plan to use for CYBERTEC's Transparent Data Encryption in PostgreSQL Enterprise Edition.
The long weekend was well worth it. I'm bringing back concrete review tasks and plenty of ideas — and I'll be back in Brussels next year.
