Can't wait for LLRT to be marked stable. Was spending time looking into Rust lambdas, but was reluctant about the total build times for my stacks and getting new team members onboarded with lambdas. (Though in fairness, Rust lambdas tend to be short and digestible, which are perfect for Rust novices.)<p>LLRT is basically no learning curve. Same old esbuild most TS and JS devs are used to. I guess we're sticking with one language from front to back again:<p>CDK in TS, lambdas in TS, and frontend logic in TS.
As always, emacs did it first: <a href="https://lwn.net/Articles/707615/" rel="nofollow">https://lwn.net/Articles/707615/</a>
This post is a reminder that you can run anything on Lambda if you're willing to run it as a Docker container! <a href="https://docs.aws.amazon.com/lambda/latest/dg/images-create.html" rel="nofollow">https://docs.aws.amazon.com/lambda/latest/dg/images-create.h...</a><p>I don't mess with AWS's default runtimes anymore!
Also though, this is a gross gross mis-optimization. Every tenant having to carry their own runtime is a huge cost penalty. That it's a much less cost? Fine great wonderful, yeah. But there should be much much much more ability to share runtime & ideally modules too. V8 isolates & wasm sandboxed modules hopefully can deliver on a thicker runtime, but which doesn't require each tenant to host the world themselves. That is never going to be fast or good or to your scale-to-zero advantage. We can should & must explore better.
I noticed that you can use QuickJS as en embedded scripting language within Python.<p>[1] <a href="https://github.com/PetterS/quickjs">https://github.com/PetterS/quickjs</a>
> This also means that its size is tiny, in the order of a few kilobytes in contrast to the 20+MB required by its counterparts.<p>A full, performant ES2023 runtime, in "the order of a few kilobytes"? Even when you remove all JIT and GC, that sounds too fantastic.<p>I'll believe it when I see it.