If you've ever used Rust, you've felt jyn's positive influence — even if you never realized it.

I hope my token of recognition is just the start of jyn's invaluable work getting the funding it deserves.

Folks, can we get jyn some more funding? RT 🙏
github.com/sponsors/jyn...

It wouldn't! I was trying to imagine that it would be possible to compile the module twice with different names/paths, so that a1 depends on b, and b depends on a2. But in order for that to work, you would need to be able to do it at the crate/workspace level, not the module level.

Could that be used to prevent cyclic dependencies? That's the use case that comes to mind.

The repo is archived now :(

I also prefer A for the unique filename reason. I have a SvelteKit repo with a thousand `+page.svelte` and `+layout.svelte` files and it drives me bonkers.

Just a thought: What if we prepended the name to `mod.rs` when it was inside the directory:

name/name.mod.rs

I'm so accustomed to having different semantic highlighting for Trait, Struct, and Enum, that when I open a foreign repo, I feel like I can't even start reading the code until RA has run :) I do feel like RA has gotten much better at handling large repos - startup times especially.

Are there people that can write Rust with RA disabled? I would be like a deer in headlights :)

I think people who are happy about the correctness of poisoning have always been relatively quiet, so mostly what the team has heard over the years is that tokio and parking lot don't poison, everyone just unwraps anyway, etc. But now that a change is imminent we're hearing from the other side.

You're trading a dep that requires a lot of setup/config for one that requires almost none and has much better perf. What do we gain by having the node team duplicate that effort?

Are you not happy with the out of the box lints provided by oxlint? You can have it in Neovim with 2 lines of setup:

github.com/dustypomerle...

I'm not overly concerned about the perf aspect, but people naturally have a lot of inertia to overcome. I think a small collection of very concrete examples would provide a lot of motivation to dip a toe in. Right now there are a lot of spectators staying on the sidelines :)

In its simplest form, my question is just "What motivates an end user to add runtime cost like vtable lookups to something like serialization?" Perhaps the cost is not much, but since we already have derives for this, there must be a reason you're working so hard to serialize this way.

the dns is probably off by one :)

Does Weak count as a back pointer? Or are you saying that back pointers can be avoided because we have things like Weak?