This is a hard one to answer definitively... here are a few shots at it...
Conversation
The original architecture and design from Stonebreaker deserves a nod, I mean you don't win a Turing award on accident.
1
2
From there a small group of stewards for many years, many thanks to likes of Bruce Momjian, Marc Fournier, and over a number of years (past 20 or so Tom Lane), in addition to MANY MANY others
1
2
as a database it took a counterintuitive approach to most. It was never the "easy to use" database, tbh it still isn't when compared to some others. But it focused on correctness.
It has since become more easy to use, but that dedication for 20 years on correctness matters
1
2
I'd also say there is true software engineering within Postgres. A lot of software isn't engineering. Not saying some SaaS app that helps order lunch is bad, but calling it engineering in the dictionary sense is a stretch.
1
4
I've heard it described "I don't like writing C" but I don't mind writing Postgres C.
It's a quality code base that has focused on quality for a long time.
Mailing lists may not always be friendly (unfortunate) but it has kept a quality bar.
2
4
Here's a great read from that was a tribute to Stonebreaker that dives into a lot of this. may also have further ideas on how Postgres was able to do what it did/does - arxiv.org/pdf/1901.01973
1
4
What a fantastic read. Who was the hacker that desperately wanted MVCC and won the argument against Mike Olson about adopting Berkeley DB transaction support because he was willing to do the work and build MVCC using no-overwrite storage?
1
2
Don't recall. Discussion would have been somewhere in the 1998 - 2000 time frame, maybe a scooch later. Pretty sure the work didn't actually get done then -- continued to use no-overwrite for a longish while.
2
3
Vadim B. Mikheev added MVCC in late 1998, early 1999:
