NInindalf·Dec 17, 2024
> we're not going to even attempt to replicate the rigorous, arduous, and closed-source testing
Almost everything you've said is wrong, but this one is trivially easy to prove wrong. You say "not even going to attempt" but they plan to take testing seriously from the get go. They say:
> Since this is a reimplementation, doesn’t that mean that testing is now even harder? The reality is that it is the other way around. Since we are reimplementing it from scratch, we are doing it with Deterministic Simulation Testing (DST) built-in from the get-go. We have both added DST facilities to the core of the database, and partnered with Antithesis to achieve a level of reliability in the database that lives up to SQLite’s reputation.
> Deterministic Simulation Testing is a paradigm made famous by the folks at TigerBeetle, that we at Turso already di
Read the rest
pped our toes into with our server-side code. With DST, we believe we can achieve an even higher degree of robustness than SQLite, since it is easier to simulate unlikely scenarios in a simulator, test years of execution with different event orderings, and upon finding issues, reproduce them 100% reliably.
Did you make your false claim in ignorance of their announcement post (https://turso.tech/blog/introducing-limbo-a-complete-rewrite...), or did you make it knowing full well that they're trying hard to make a well tested library?
---
But fuck it, I got time. I'll tackle the other dubious claims you've made as well.
> Also, we replaced that pesky, unsavory-to-our-sensibilities “Code of Ethics”
They haven't replaced it because Limbo is a new project. LibSQL is a fork of an existing project but again, they haven't replaced the "Code of Ethics" in the repo (https://sqlite.org/src/dir?ci=trunk) because it never existed in the repo, just on the sqlite.org website.
So they "replaced" nothing. They simply added their own code of conduct, which (I agree with you) is bog standard for a project on Github.
---
> Announcing before it's ready
Yes? That's normal? You're objecting to a project even existing on Github if it's not 100% done? They specifically explain why they've done it. They're building in the open, they're seeing excitement from outside contributors who want to contribute. So they're announcing to make more potential contributors aware this project exists.
Only someone with a warped mind could find something objectionable about starting small and iterating with the community.
---
> Also we're going to extend SQLite to be “cloud-capable” because that's our business use-case, even though it's completely at odds with SQLite's intended use-case
Nothing you've said makes me think you understand the social side of building software, but this makes me think you don't understand the technical side either. Perhaps you missed the link on this thread where they explain why SQLite works well in the cloud.
I'll explain for you though, in case you're unable to find the link at the top of this page. SQLite relies on random access from disk, which made it a poor fit for cloud workloads when the cloud used HDDs exclusively. HDDs work much better for sequential access, which is why LSM trees were invented. So rather than reading from a random part of the local disk (10-20ms) you'd rather pay the network cost (0.5-2ms) and read from a database on a different machine that has optimised it's data storage for sequential read.
But SSDs change the game. A random read of 4k from a local SSD is 150 microseconds (https://gist.github.com/jboner/2841832), always less than the network round trip. Local SQLite running on an SSD is suddenly the lowest latency database.
For infrequently changing data that needs to be served with low latency, SQLite is the best choice in the cloud.
If you don't follow this, it's alright. I didn't expect you to.
---
> SQLite is public domain so there's no problem with us rewriting it and We're going to MIT license our rewrite because there's various reasons why people would rather use MIT-licensed code than public domain code
Yes, this is true. I don't think this was an objection from you.
---
> We're going to rewrite it in Rust because that's going to make it inherently better (and don't question this)
What should they be writing it in? I notice you don't say. You reckon they should start a new C code base in 2024?
---
> Redot comparison
Redot is a fork created solely because the creators didn't like a tweet about Godot being welcoming to LGBT game developers. LibSQL was created because they wanted to add features to the SQLite codebase and weren't allowed.
In any case the comparison is wrong. You can see the two commit histories of redot (https://github.com/Redot-Engine/redot-engine/commits/master/... libsql (https://github.com/tursodatabase/libsql/commits/main/) and see which of these is actually a community creating something new and which is a dead project simply merging in commits from upstream.
---
No one disagrees that SQLite is highly useful, rock-solid reliable software. But I disagree with you that it is perfect in every way and no one should ever second guess any decision made by the maintainers. Turso wants to try something different. Nothing wrong with trying.