Jérôme Gravel-Niquet

Rustconf when it was in Montreal (local to me) had special priced tickets for locals ($50 IIRC). There aren’t for Seattle?

Nice. Do you happen to know why it’s comparing to s3fs instead of something like juicefs, geesefs or tigrisfs? I assumed s3fs was antiquated and not used very much.

parking_lot: ffffffffffffffff... Gnarliest Rust proxy bug yet.

JFC this bug. fly.io/blog/parking...

My first reaction is that it’s not really embedded if it uses Postgres :) but maybe a good Postgres setup is what’s expected. If it used SQLite or fjall or another embeddable DB, you could get rid of that dependency.

I’ve only discovered its existence this week and it’s solving a lot of problems for me. Previously tried axum+utoipa and dropshot. Both solutions had their own issues where poem seems to just work and be flexible enough for my needs!

Stop Building AI Tools Backwards | Hazel Weakly I’ve been reading this week about how humans learn, and effective ways of transferring knowledge. In addition, I’ve also had AI in the back of my mind, and...

I realized something recently, which is that we're building AI tools completely backwards.

While I was aware that there's some de-skilling happening already, I finally clicked a bunch of pieces together and decided to write this to talk about how we can do *better*

hazelweakly.me/blog/stop-bu...

crates.io: Rust Package Registry

unsynn v0.1.0 is out!

crates.io/crates/unsynn

What the…

If I understand correctly, their software requires generating Rust code from SQL. Instead of making it all one big crate, they splitted it into a ton of crates. They are not publishing the crates, they’re only generated at build-time for specific uses of Feldera

Taming A Voracious Rust Proxy The Integral Principles of the Structural Dynamics of Async Rust

Peter Cai, on our proxy team, working with Pavel to track down an annoying async Rust bug. fly.io/blog/taming-...

That part is very interesting to me! I’ll definitely take a look.

Hmmm, I like that! What names do you usually use for labels?

Err… `#[serde(skip_serializing)]`

Hmm, maybe `#[serde(skip_serialization)]` on the field would give you what you want?

That said, I’m not opposed to jumping through the hoops and learning new ways, but I was hoping for something that worked out of the box :)

Very interested by this.

Up to now I’ve been able to evaluate Rhai scripts inside a hyperlight VM, but nothing more complex.

Wasmtime requires jumping through hoops to get working no_std. Most things with C dependencies are hard to compile for the x86_64-unknown-none.

Was very interested in that feature and ended up implementing it for fun. Didn’t turn out too bad!

Without foreign key constraints, then you’re implementing it yourself, yeah :( might be good to take that into account when designing the keys you’re using so you can iterate on them. E.g. `device_id:message_id` or the reverse.

Let me know if you do! Happy to chat about it.

If your transactions don't overlap, then this is definitely faster than single writer tx. If they do cause conflicts due to the aforementioned reason, you can decide what to do depending on what you're trying to do. Definitely not a silver bullet heh.

That's correct! Using the `single_writer_tx` means each writer transaction happen one at a time. Using `ssi_tx` (introduced by this PR), you can have any number at the same time, but they can fail (retryable) if a transaction inserts data that was read by another committed previously.

Serializable Snapshot Isolation (SSI) by jeromegn · Pull Request #79 · fjall-rs/fjall This allows multiple concurrent write transactions. Checklist Change MSRV back to 1.74 Uncomment unwrap_used lint Remove #[inline] attributes Keep the option to use the single-writer transacti...

If you’re still using fjall and need transactions, this might interest you: github.com/fjall-rs/fja...

A tweet of the OP followed by a picture of a bodega cat on an ATM machine

ATM surveillance

Tasted great for 10 seconds!

If you see this, post 🐈