4 ms·

The rust reimplementation of the node modules is interesting to read. I took some ideas for the llrt runtime modules. As a comparison Bun Zig implementation is

by Sytten 1y ago

The rust reimplementation of the node modules is interesting to read. I took some ideas for the llrt runtime modules. As a comparison Bun Zig implementation is scarily ignoring a lot of edge cases.

bsder 1y ago
So, basically, people don't seem to value security (Deno) very much but value speed (Bun) quite a lot. That is pretty much the standard problem across programming.
- hu3 1y ago
  I value the path of least resistance, and Bun wins hands down because that's one of its priorities. In contrast, Deno gambled that Node compatibility wasn't critical and lost. Now they're backpedaling.
- pjmlp 1y ago
  Mostly because liability is only a thing in some areas, when programming becomes as scrutinised as other industries, and basic stuff like returns due to failed functionality are normal instead of "try to reboot it", security and quality will have other priority.
hardwaresofton 1y ago
Everyone complains about Rust until they need high quality software that doesn't cut corners or isn’t impossible to write correctly. I’m biased but learning difficulty aside, Rust is very optimal as a language.
- pjmlp 1y ago
  Except other than the affine types, we can get the same high quality in ML derived languages, nothing special about Rust there. In general the lack of quality is a side effect of the lack of liability in many software fields, not the programming language that gets used.
  - unchar1 1y ago
    Library and ecosystem support is a big thing. Rust being ML-ish, and having a large ecosystem is probably it's greatest strength
    - pjmlp 1y ago
      There is plenty of choice in Maven Central, NuGet, Opam, and Hackage for most tasks anyone can think of.
  - hardwaresofton 1y ago
    Fully agreed, other than the fact that Rust is absolutely beating Haskell in being the usable-in-production choice (and Haskell is WAY ahead of any other ML languages). Rust has just enough ML type goodness to make a big improvement AND it has the affine types which are legitimately novel. Haskell is still struggling with what exactly to do about Linear types (and get the whole ecosystem on board), and strictness. While lazy-by-default is really amazing for certain things, Rust's approach just wins out for production code, I think. > In general the lack of quality is a side effect of the lack of liability in many software fields, not the programming language that gets used. This is true, until you have languages with truly novel features, like an implementation of affine types, and strong compilers. Even just the difference between errors-as-values versus exceptions is a huge difference in my book -- Rust code gives you a clear chance/requirement to handle every possible error.
    - pjmlp 1y ago
      There are other options besides Haskell.
      - hardwaresofton 1y ago
        Thinking production readiness, not many? Ocaml maybe but seems a bit unused outside certain firms
        1y ago

by Sytten 1y ago

The rust reimplementation of the node modules is interesting to read. I took some ideas for the llrt runtime modules. As a comparison Bun Zig implementation is scarily ignoring a lot of edge cases.

bsder 1y ago

So, basically, people don't seem to value security (Deno) very much but value speed (Bun) quite a lot. That is pretty much the standard problem across programming.

hu3 1y ago
I value the path of least resistance, and Bun wins hands down because that's one of its priorities. In contrast, Deno gambled that Node compatibility wasn't critical and lost. Now they're backpedaling.
pjmlp 1y ago
Mostly because liability is only a thing in some areas, when programming becomes as scrutinised as other industries, and basic stuff like returns due to failed functionality are normal instead of "try to reboot it", security and quality will have other priority.

hardwaresofton 1y ago

Everyone complains about Rust until they need high quality software that doesn't cut corners or isn’t impossible to write correctly. I’m biased but learning difficulty aside, Rust is very optimal as a language.

pjmlp 1y ago
Except other than the affine types, we can get the same high quality in ML derived languages, nothing special about Rust there. In general the lack of quality is a side effect of the lack of liability in many software fields, not the programming language that gets used.
- unchar1 1y ago
  Library and ecosystem support is a big thing. Rust being ML-ish, and having a large ecosystem is probably it's greatest strength
  - pjmlp 1y ago
    There is plenty of choice in Maven Central, NuGet, Opam, and Hackage for most tasks anyone can think of.
- hardwaresofton 1y ago
  Fully agreed, other than the fact that Rust is absolutely beating Haskell in being the usable-in-production choice (and Haskell is WAY ahead of any other ML languages). Rust has just enough ML type goodness to make a big improvement AND it has the affine types which are legitimately novel. Haskell is still struggling with what exactly to do about Linear types (and get the whole ecosystem on board), and strictness. While lazy-by-default is really amazing for certain things, Rust's approach just wins out for production code, I think. > In general the lack of quality is a side effect of the lack of liability in many software fields, not the programming language that gets used. This is true, until you have languages with truly novel features, like an implementation of affine types, and strong compilers. Even just the difference between errors-as-values versus exceptions is a huge difference in my book -- Rust code gives you a clear chance/requirement to handle every possible error.
  - pjmlp 1y ago
    There are other options besides Haskell.
    - hardwaresofton 1y ago
      Thinking production readiness, not many? Ocaml maybe but seems a bit unused outside certain firms
      - 1y ago