We got into Rust the hard way

Everyone is rewriting everything in Rust these days. Performance is great. Agents have turned this into a near-zero-effort task. Fair. Not us though.

We did something else. We contributed to Turso, an incredible database. It can be simplistically described as a Rust rewrite of SQLite. A couple of bug fixes are in; Clippy for Linux arm64 and case sensitivity in computed columns. Most importantly, an exciting new feature is already cooking.

These approaches differ a lot, because the outcomes differ. We were not trying to build something for self-consumption. We want some hardworking strangers on the internet to trust and accept our work as part of a codebase they are maintaining. At a time when anyone can one-shot open a pull request, that requires meticulous work.

We had to work meticulously with our agents to make the answers to two questions trivial. First, is the claimed issue, to be addressed by each pull request, actually valid? For this part, we had to prepare a laconic, crystal clear description of the issue with the fewest, yet sufficient, references proving it. Second, does the patch actually do what it should with nothing less and nothing more? Here, we had to go the extra mile and ensure that every single line is both absolutely needed to solve the issue and in its clearest possible version. But talk is cheap; you can see these in practice in Turso's GitHub.

Finally, why did we do this? Well, we are building something new. Without many spoilers, we are building Fakos, an extremely efficient OpenTelemetry server. It can handle thousands of OTEL requests with tens of records each, with very few resources. The backbone of its storage engine is the impressively well-performing Turso.

If you are a bit more curious and want to learn more about Fakos, you can check my presentation at Django Day Copenhagen just a month ago.