This year, like many years before, I attended FOSDEM PGDay. Below are my semi-structured thoughts - reflections on sessions attended, and conversations had.
Daniel Gustafsson (Microsoft) talked about a patch he's been working on since 2017, when Magnus first brought up the possibility of data checksums in PostgreSQL. The suggestion gathered 100+ replies on the mailing list, after which Daniel and Magnus hacked something together at the Stockholm PUG (PostgreSQL User Group).
The patch ended up so complicated that senior devs were like "looks correct, but I'm not touching/committing it". "Restartability" as a feature was desirable, but ultimately made it uncommittable. Removing all traces of restartability, thus removing complexity, was necessary even if manual restarts are not "sexy". Now the basic functionality has been vetted through review and there's the conviction it can be finetuned to the finish line.
It might of course still not make it (in PG 19 or at all), does that make it a bad idea to spend (some part of) ~10 years on? Daniel doesn't think so: "attempting to solve hard problems is foremost an investment in yourself". Of course Daniel, like the rest of us mere mortals, needs to report to a manager on what he's worked on, and he has "nothing" to show for this multi-year investment. That's where he highlights he works on other "side-quests" as well. Many of those prompted by bugs found in PostgreSQL while working on this patch.
Robert Treat (AWS) slightly cheekily asked Daniel if he ever stopped to think whether users still need this feature. Daniel's response: while the feature itself is not terribly important, the capability / possibilities it unlocks is the real value.
Daniel helps his colleague Claire Giordano (one of those side-quests he mentioned) with her talks covering "what goes into a PostgreSQL release" (latest slides). Beyond code contributions, Claire also looks at the volunteering folks do for the different working groups (security team, infrastructure team, etc), on committees, or by adding value through blog posts, talks, free workshops. Attributing code contributions and mailing list activity, Daniel developed wrapper Ekorre, which in turn builds on my colleague Robert Haas' work.
My colleague Bruce Momjian in "What's missing in Postgres"... did just that. Instead of talking about existing features, or a patch idea, Bruce covered why certain functionality never made it into core. This was the second iteration of the talk, having presented it already at P2D2 (Prague PostgreSQL Developer Day), earlier this year. The feedback and comments from that event and FOSDEM PGDay combined will certainly make for a super well-rounded talk when Bruce ultimately presents at PGConf DEV in May this year.
The day before Bruce met with other Core members for their developer meeting scheduled just ahead of FOSDEM. And other Core members were at Bruce's talk, with Alvaro even providing extra context here and there.
Bruce highlighted that all requests he listed are performance optimizations, with the exception of encryption (or: a community TDE solution). Some requests come from people coming from other ecosystems, but Bruce found that once they're using PostgreSQL in production they're not asking for any of those anymore.
It's a pity FOSDEM PGDay sessions aren't recorded, several times sessions I wanted to see were scheduled in parallel. But it's hard to make a schedule that works for everyone... I did very much enjoy the lightning talks (and delivered one, with Jonathan Gonzalez), meeting with my fellow PGDay Lowlands organizers, EDB colleagues, and Christoph Berg and Joe Conway, who I'm on the Contributors Committee with. A great start into the new year!
Top comments (0)