Hacker News Reader

Comment by std_throwaway | Hacker News Reader

Isn't SQLite an application calling for a clean implementation in Rust?

Why? SQLite is one of the highest quality codebases out there! For fun, sure. But I think you are underestimating the size of the SQLite project and it's tests. Just because the library is small it doesn't mean it's simple.

It's massive testing means that it would be easier to implement in another language; you'd be pretty sure of conformance with a port.

That said, it's one of the most reliable projects in existence so why rewrite it?

To reboot WebSQL?

IIRC Mozilla's main gripe with it was the fact that there was only one implementation, and no standard beyond "Whatever Sqlite3 does".

How would a rewrite change that? Addressing that objection would require some sort of open standards process and commitment to follow it, which is a rather orthogonal concern and would be somewhat expensive as well.

You'd also need to turn "whatever sqlitev3.x does" into a spec, indeed which is a lot of work that will never happen as there's only one implementation that passes the test suite.

Same reason as always: https://www.cvedetails.com/vulnerability-list/vendor_id-9237... I'd guess around half of those wouldn't happen in rust.

It would be a colossal waste of time. At 125KSLOC code + 91MSLOC tests, the engineering effort is near-prohibitive. There is no single area where SQLite3 would be severely lacking. Just look at SQLite4 and say if you care about any of the differences.

It's been well-tested enough not to matter. Plus that testing would need to be duplicated. Very costly. Uch.

I think you could repurpose the original test suite as is. Rust can interface with C seamlessly.

Reply on news.ycombinator.com