Rust Is Eating JavaScript (2023)
leerob.com
leerob.com
I have worked at 2 of those named companies where people made conscious decisions of moving away from TS backend because JS is an awful language to be used for system critical things.
I've been writing backend TypeScript for 8 years and it's more than fine.
It's not my favorite language but it's a great compromise, and I would never choose a Lisp, Haskell, Java or C over it (the other languages I know).
We built some stuff with Go and then went with full .Net. Its like fresh air after years of Js.
- Proper multithreading and performance story. Performance itself is not the most important thing when your app mostly waits on IO. However you still get to use thread pooled async in .net or parallel goroutines in go etc and they sometimes make a difference.
- ASP.NET Core is fully feature packed and easy to work with. .net has a more cohesive ecosystem where tools, libraries, and frameworks are designed to work together seamlessly. In js runtimes, you often need to piece together different approaches and deal with compatibility issues.
- Proper standard library. Both node.js stdlib and the js globals are really weak compared to what Go or .NET or Java etc provides you out of the box. I hate dealing with npm dependencies with a passion at this point and you can have minimal dependencies when the standard library is decent. There is also the fact that upgrading an npm package is a dice roll as you need to trust the semver or check update logs etc. At least with a static typed language you get to catch public api changes on your dependencies during compile time, not in runtime.
- Linq is good, Linq expressions are great. I don't really love ORMs but I tend to use Entity Framework Core on top of low level SQL access in .NET. Makes life easier.
I don't hate TypeScript though. I still use it on the frontend obviously and I'm fine with it but given the choices on backend, it does not make much sense for me anymore.
It was originally designed for writing throwaway frontend code, but people liked it so much that they started using it to build their system architecture—only to realize it doesn’t work well for anything beyond glorified RPC backends.
The type system is wishy-washy, and TypeScript needs a massive type space to compensate for it. Python is also dynamically typed, but it has a strongly typed system that saves you from runtime blowups. JavaScript doesn’t even properly blow up at runtime—you just get [object Object] or some random undefined error. TypeScript is a fantastic piece of engineering from the C# guy, but even they couldn’t fix all of JavaScript’s language-level blunders.
Few people want to build mission-critical backends on a weakly typed language.
The ecosystem is a mess, and things randomly break after a few days. My Python and Go apps from 2018 still work exactly as they did on day one. Go’s gorm and Python’s SQLAlchemy are the default ORMs that pretty much everyone uses. And how many ORMs does the JS ecosystem have?
And let’s not even start with frontend frameworks—no one loves them two days later.
The Next.js project I built yesterday is already showing 69 vulnerabilities. This sorry excuse of a language, coupled with terrible design and an indecisively childish community, makes it difficult to take seriously.
By the way you can be extremely strict and safe in typescript, it's really up to the team using it, but you can encode virtually everything and have it fully type safe.
Also, there's great tools like effect-ts if you're more functionally leaning.
It's fine to not know about the goods in the TypeScript ecosystem.
Again, I recommend you checking libraries like effect-ts or effect/schema or even better trying them.
Sure, JS has some quirks that language design gurus complain about, but nothing that actually matters on the job. Like, == operator is weird, oh well.
Python is rightfully king for certain things like data science, but parallelism and package management are two big messes in it. ||ism is trying to be fixed with asyncio, but it's kinda too late. You know packaging is broken because every Python repo has a Dockerfile, not something you see in JS where npm is solid.
Golang is more for different use cases. Looks good but really should have done error handling like Rust, or used exceptions.
sure it is
My NodeJS apps from 2014 still work exactly as they did on day one as well, what's your fucking point?
Alternatively you could go the openAPI route and declare your API as a spec with types for a client to consume and the endpoint to return. This works cross-language and gives you docs first. (Start from the docs, convert that to types, then implement).
I also thing that the tooling hasn’t kept the same pace as it has with graphql, where it’s easy to document your schema and the inspector makes it super easy to play with.
I have found that the tooling tends to have the problem of not knowing how to catch edge cases like webhooks, websockets, complex auth and additional logic beyond just an HTTP request. Many tools add layers of configuration and extension to work around these, but i feel it misses the mark.
I'm building Borea.dev to solve this pain by keeping the source of truth in your OpenAPI doc and allowing for custom code implementations. We support Python rn, are building out generators for multiple other languages and we're open source :P hope it helps
I like to say that the main value of GraphQL was never really in the dynamic queries (in many ways an anti-feature) but rather in having a nice typed spec that lets you generate typed clients.
https://oxide-and-friends.transistor.fm/episodes/the-fronten...
Can I tell you about live debugging of an app server and hot reloads? Pretty wonderful stuff really.
I can't believe how much knowledge, tooling and experience was thrown in the dumpster when Java was abandoned en masse by the younger generation of devs due to the rise of node.js. And if it weren't for Typescript the tooling would still likely suck.
I haven't used it for a long time, but it's still alive: https://www.gwtproject.org/
I feel like there's an adjacent universe where the JVM lives in every browser and JS was a weird failed experiment. But it turned out that the browser was the universal UI everyone had waited for, and Java didn't connect to it while JS entrenched itself there.
GWT was a glimpse of trying to send us to that universe, via a twisty back door. But it was never gonna happen.
but yeah js is a strange creature
NB: I was very distantly part of the Swing team. I did early reviews of its design documents.
I wish us devs would have access to that intermediate representation.
And that's the exact reason I'm done with it.
Typescript is somehow becoming a laguage that supports writing extremely performant high throughput and very scalable server code? That seems unlikely. So no worries, it can't become Java.
Java, as a language is fine I guess, but the culture around it sucks.
Culture is what you want it to be.
There are indeed some ridiculous subcultures within the world of developers who use Java. Like J2EE-anything or any of the baroque frameworks.
It is important to realize none of that is part of Java and none of that is required unless you really want to.
You can use Java simply as a memory safe language with reasonably good type safety that can create wicked fast code and has huge library and tooling support. On all those checkboxes, there are few-to-none options that are so good across all those aspects at once.
Unfortunately so many people just complain about writing TaskStubConsumerWatcherVisitorServiceContainerListIdentifier classes, without realizing nobody is forcing you to do that. So don't.
do you mean rmi/javabeans or something else ?
beside that, I left java long ago, and believed in simpler REST stuff but now I realize that it's indeed rediscovering stuff people dealt with long ago
As for rmi, when Java was in its heyday the most popular protocol for remote api handling was SOAP. But when RESTful era arrived Java tooling adapted very quickly and it made little difference what you were using to call outside the app server.
@Validated
public interface PetOperations {
@Post
@SingleResult
Publisher<Pet> save(@NotBlank String name, @Min(1L) int age);
}
@Client("https://example.com/pets") @Inject PetOperations pets;
@Controller("/pets")
public class PetController implements PetOperations {
@Override
@SingleResult
public Publisher<Pet> save(String name, int age) {
Pet pet = new Pet();
pet.setName(name);
pet.setAge(age);
// save to database or something
return Mono.just(pet);
}
}
But yeah, I built business apps with just Tomcat and no J2EE components and it was fine.
It’s easy to say “because they’re hipsters following trends” but I dipped my toe in the Java world years ago and found it immensely frustrating. Eclipse was awful compared to the lightweight editors I was using at the time. The language got in my way more often than it helped me.
I’m sure that’s not the case these days (indeed I’ve done a little Android-adjacent stuff and found it to be fine) but the idea that there were no downsides to Java at the time strikes me as unlikely.
Since Java was developed the world has shifted to accommodate a lot more rapid development and code you throw away rather than maintain. But experienced developers also know how often "prototypes" are still around decades later.
Java is also not perfect, and a lot of new languages are conscious of its faults. But the reason it's still around is that it's core principles are oriented to code that sticks around.
You could equally say this of COBOL.
> They haven't yet looked back at code they wrote and said, "ugh, what the hell was I thinking?"
Ironically, "what they hell were they thinking" is exactly the reaction I have to even Greenfield code written using Spring Boot or any of the other annotation driven nonsense.
The idea that programs written in Java are more maintainable than programs written in (say) Go, where you can read what the program does end-to-end instead of having to guess at what garbage has been generated by Lombok or overridden by some random JAR on the class path is bizarre. Let's not even talk about Gradle.
A lot of external Java framework is about inversion of control. It looks good on PowerPoint slides, but I think it's horrific to maintain (as well has hard to write). Inversion of control means you have no control, and the most common error case is "why didn't my code run?" But you can't use a debugger to tell you what didn't happen. Any breakpoints have to be set inside the framework code, which is open source but is not intended for your eyes.
Yep no magic...
Have you ever read Kubernetes source code?
Kubernetes is not a good example of idiomatic Go in any sense. Perhaps Spring Boot and friends are also not good examples of idiomatic Java, but they are absolutely the common case of what is found in the wild.
Kubernetes is a single (bad) Go project. Every piece of Java I ever see is AbstractFactoryProxyBean nonsense with annotations all over the show.
I don’t know of a Go project where I can’t start at main (for a given program) and work out how it works without a bunch of ecosystem context. I can’t think of a single Java program where I can.
My problems with Java are startup time and memory consumption (and NPEs).
If that's not your problem - fine. But writing web apps in the Java ecosystem has been terrible for many of the most important years (let's say 2005 to 2015) and that's where all the "young people" forsake it for... PHP, Rails, node.js.
(You can already guess that I'm not even counting myself as part of the younger generation here. I'd used used Java and the JVM intermittently between 2000 and 2020 but it has never been my favourite stack).
Yet it blames "hipsters" for throwing "it all in the garbage," implying that there was no rational reason for people to switch away from Java. Naturally, people respond by sharing their reasons.
And talk not to me about lightweight. I'm currently on Day 4 of setting up a Python environment in VsCode that doesn't suck. I'm nearly there though it's still a bit inferior to a 20 year old JBuilder install for Java but I guess it needs a few more years to get there. Oh and as I said, I'm on a long quest here to pave the road for my devs but it looks like this hodgepodge of extensions still won't deliver the experience I had decades ago.
I'm also not saying there's a better option for you and your team, either, my friend.
As a python programmer using bare-bones vim (on purpose), having abandoned all fancy IDEs, I really wouldn't know where the state of the art is, anyway, but I doubt the MS leopard has changed its spots.
Embrace & extend complete ;-)
This time it looks like 'integrate' will be the third step.
Thanks for the great info.
Python is pretty close to a 'Microsoft language' these days. They've hired several Python core developers to make sure windows is first class citizen for python and the developing python on windows and with VSCode is as good an experience as possible. They also develop the official VS Code Python extension and language server as well as many other python tools and extensions.
I'd say the current state-of-the-art is either VSCode with the Microsoft python extensions or Jetbrain's PyCharm. Personally I slightly prefer PyCharm Pro, but it's pretty much a toss up.
As a newcommer to javascript, I was able to learn faster and produce same quality faster then non-IDE developers. Not because I would be smarter or more talented, but because I had no emotional need to prove myself by not using ide.
It's probably unavoidable that people's memories will differ, but personally, as someone who worked with Java 10–15 years ago and then ran away as fast as he could, these are the things I remember most: the IDEs being awfully slow and unstable (though quite powerful for their time) and the language/ecosystem being painfully verbose and full of boilerplate and layers of indirection. I remember how, after switching away from Java, I felt elated by how easy and maintainable things could be when I was allowed to use simple, fast tools and write straightforward code that just did what it was supposed to do.
> there were not real licensing issues
Oh yes there were. Even in 2024 the “which JDK are you running on” debate still goes on. There are notable differences between corretto and openjdk, and openjdk and Oracle JDK. Here [0] is a stack overflow question that notes it was asked 10 years ago that talks about the various JDKs and the license change where oracle started charging for JDK access.
As for your point about IDE’s vs non-IDE’s that’s a straw man. I said that you were forced to used an IDE for Java (which is of course not technically true but compared to rails which took of at that time it is), but also that the IDE du jour (eclipse) was pretty monstrous.
I’ve said this on here before but the minute maven gets introduced into a project build times are measured in minutes. My current work project is c# and I can run my entire test suite faster than the jvm got itself up and running on my last kotlin project. Even IntelliJ and co suffer from these limitations - manually bumping the memory limits for IntelliJ and friends in 2024 is practically a requirement, and the startup times are very Java inspired.
[0] https://stackoverflow.com/questions/22358071/differences-bet...
Java and Eclipse are waaay better then C# and visual studio. Kotlin is a pure pleasure against C#.
Also, I see people using different JDKa on the same project. That there is competition is a good thing, it is not like you was locked in one.
Except, unlike Kotlin, C# has a rich selection of what to use for development across a variety of platforms and is not constrained to JVM's type system limitations, allowing the syntax differences to have meaningful impact on the code execution.
It also has excellent, comparable to Go and Rust, CLI tooling to further streamline the process of quickly managing, building, testing, running and deploying projects, something JVM ecosystem is yet to provide an alternative for.
I disagree. dotnet has it's quirks, but so does every mature ecosystem. I wouldn't use "piece of crap" to describe it (or java) and I get that it's an opinion, but I think your opinion is more misguided than what yo're accusing other people of believing.
> Kotlin is a pure pleasure against C#.
My last project was a kotlin backend powering a C+ desktop app. The kotlin backend took longer to compile than the C++ did. Kotlin is the kitchen sink bolted onto Java. It's definitely a nice environment to work in, but it's certainly not a more streamlined java.
> That there is competition is a good thing, it is not like you was locked in one.
If by competition you mean OpenJDK forks with minor differences then yes there's competition. To me that's not competition, that's fragmentation.
3-10 crashes per day were common.
Lots or coffee breaks, or at least switching back to emacs to just finish editing whatever classes you were touching.
The second thing however is that fronted is better off in non-java languages, you can afford to do small backend in typescript and what not. And larger enterprise projects were done in java the whole time.
I wasn't fond of JavaScript (In particular, I never liked dynamic typing), but when I had the opportunity to work in node.js instead of Java I took it immediately, just to avoid the pain and suffering that the Java toolchain inflicted on me.
I started my career in Java, and worked in it for 20 years before I really started using Javascript in anger. I didn't (and don't) like Javascript, but the freedom for prototyping was a revelation. When Typescript appeared, I fell in love.
I'm of the opinion that the world was a better place with Ant and when we had third party dependencies manually, deliberately adn deliberatively managed by hand in /jars/thirdparty
As someone that was writing Ant scripts when CruiseControl was modern, and there was no Ivy in sight, not sure if it was such a better place.
I hated Maven at the beggining, exactly because we got some hipster company pushing for it during Maven early days, it was still hot out of the oven.
20 years later, I don't see a reason to use anything else other than Maven.
I've looked at it also. The problem is incentives. Will developers pay for a better build system? Probably not. Then, what people use is whatever happened to be given away for free first.
Another problem with shifting Gradle is that lots of libraries now come with build system plugins too, so you'd need to replicate all that logic into whatever new build system you create. And it's actually not easy to get everyone to agree on what they don't like about Gradle. Say "imperative not declarative" and half of devs will agree and the other half will say, no, you need programmatic configuration because build systems are genuinely complex. Etc.
https://gradle.github.io/declarative-gradle/
Fingers crossed.
The biggest problem with Gradle is not Kotlin or even Groovy syntax. Kotlin isn't a bad language to use even for writing configuration. It's that the entire thing is designed in extremely confusing ways from back to front, and has a lot of awkward design choices and bugs that result in a hyper-complex user model.
What the JVM world needs is a totally new tool that has the same feature set as Gradle but with a drastically simpler mental model behind it. The only real contender right now is Bazel, I think.
Android started being devleoped mostly by C++ refugees, that is why originally all framework code is full of m_ prefixes.
Then the whole standard library fragmentation that persists to this day.
Despite the whole excuse why Dalvik was created, Nokia and Sony-Ericson had quite good JVM implentations for Symbian.
When Kotlin was introduced they stiffled Java on purpose, left it on Java 8, using Java 8 examples to promote Kotlin, and have since Android 12, started to finally update ART to more modern versions, because as it turns out, even with Kotlin, loosing compatibility with the Maven Central ecossytem isn't that great.
Still, it is mostly Java 17, and all the ongoing improvement in Java land across all JVM vendors, will most likely never land on Android.
Android is not Java, even though it relies heavily on the ecosystem for its tooling.
Android is basically Google's .NET, and Kotlin plays the same role as C# in relation to Microsoft's J++, that lead to Sun's lawsuit.
Java today looks like a fine platform, though.
I've grown up a bit since then, and so has Java. These days I work in Java for my day job, and it's perfectly fine. I don't even mind all the boilerplate, I can look at a page of code and just see blonde, brunette, redhead.
But I still love Common Lisp and Scheme much more.
Got in the way how? By insisting that things have types and one cannot blindly add an object to an array [] + {} and expect it to behave? Never mind that in Javascript adding an array to an object {} + [] gives a different result.
ETA: for bonus points, what does {} + {} return?
I hear Java has gotten a lot better. But then again there is C#, which is pretty much just better Java. Still too enterprisy for the tastes of most HN types though
I thought JS was bad at the time, but node.js was a breath of fresh air for API development, and client side JS has been fine for like 10 years (assuming you care about UI development).
I'd never go back to Java except for maybe some backend service away from the UI.
.NET has also had all these features for just as long, and is less prone (though definitely not immune) to AbstractFactoryProxyBean nonsense.
Last I checked If I change a signature or something in Java, I am not guaranteed the whole thing "checks". If I add a new exception to a function, how do I know what else is effected?
I give it five years until Typescript and peers adds checked exceptions.
[0] https://www.artima.com/articles/the-trouble-with-checked-exc...
We tried. Java was fine, JBuilder was fine. When they instead started pushing J2EE, EJB and all the other acronyms I've forgotten, this 'youngster' realised he'd never be smart enough to be a Java developer and bailed back to C, Perl and Python.
Had they focused on Java and making Java a great language in its own right, it no doubt would have gone a lot better. But all the 'hype' and marketing around Java was focused entirely on these huge frameworks and enterprise 'solution' architectures which most 'youngsters' just didn't understand and didn't see the need for.
If someone makes it clear their product isn't for you and that they don't care about solving your problems, is it strange they abandon you?
While I haven't gone too deep down this particular rabbit hole, if you build a Rust backend you can use function signatures and annotations to automatically build an OpenAPI specification for it. You can then consume that specification to automatically build a typescript client library, giving you an automatically updating Typescript interface to your Rust backend.
Similarly, there are multiple projects to automatically provide python type hints for your Rust python extensions. None of them really production-ready yet, but this is something people are actively working on for quite a while
Only extra step is running a tool to download the swagger and produce auto-generated typescript definitions, otherwise the workflow is the same as shared code (Some projects are typescript on both ends but C# gives a ton of other benefits for backends).
Like mentioned, our regular pipe is C# producer (server) and TS consumer (client). Consuming services authored in Java,etc is usually not a problem either but the TS system definetly has a bunch of more cases that aren't easily mapped.
Have you ever seen Scala.js? Same language on the frontend and backend, in a way that actually feels first-class and real. Type system that's as powerful as Typescript but also sound. Full IDE functionality if you need it.
There are tons of generators that can create a client with types from an OpenAPI spec. Not only to/from TS but also tons of other languages.
It seems like in almost all cases, you want to evolve in a forward compatible manner, with a very slow deprecation process.
Except for a nod to WASM, all the examples are about the JavaScript toolchain.
Besides Bun is not Rust even though Deno and some tools are. The JS crowd generally does not care about coding Rust, though would welcome faster tools which are sometimes rust based. Also Go based and even Zig based.
For example, you’ll still write your React components, but the React framework itself might be written in Rust and compiled down to WASM to make everything snappier (e.g. DOM diffing).
I personally really love this pattern because it strikes a balance between speed and DX. Not everything has to maximize performance, and not everything has to maximize productivity.
UV is a year and a day old. It's not surprising that there isn't huge industry adoption yet.
I sort of have the impression that Ruff does have a reasonable degree of industry adoption. Certainly linting does.
There mypy alternative doesn't yet exist in a usable state
I'm not seeing anything worrying here?
To add to the issue, a huge amount of Python in public repositories is untyped and unversioned- which means that LLMs are pretty bad at getting typing correcting in Python, so the debt is growing much faster than the solutions.
I honestly think that the most viable solution at this point for typing is a typescript-adjacent solution for Python. I’m kind of surprised Microsoft hasn’t put one out yet
Eg. My webpack builds in 35 seconds. The i try esbuild and it does the same job in 35ms. Insane! Ok now lets try rust, now i get 20ms. Wow!
Do i care about 35ms -> 20ms? Nope, not one bit. I just want it to compile under 1second.
It talks about projects like Rome as if they're current when Rome was archived a year and a half ago: https://github.com/rome/tools
I think it used to have one but it was lost in a site redesign a few months ago: https://github.com/leerob/site/pull/727/files#diff-888ddb306...
C/C++ developed thanks to the effort of hardware, kernel, and os makers who invested billions into its development. Development of the JVM once occupied the effort of 10k engineers at SUN. JavaScript has a complex history, but has received significant investment from Netscape, Mozilla, Facebook, Google, and Microsoft.
Rust needs one or two singular players with 5+ billion to spend supporting and improving devs lives. The nascent efforts I’ve seen so far are just that - nascent.
Why must we rehash this type of post every few months?
was between (js) and (go/C/rust)
See the problem to me is that this comparison isn't nice. Javascript comes with a lot of its problems , so of course people would want go / C / rust
but Now the author thus concluding that people want "the fastest" out of go/C/rust is just wrong.
I love golang.
People like fast software but hate abandoned software even more . Yes there are rustaceans who would love the project but now hand anybody genuinely interested to contribute a rust project and ask them to learn rust and they would come hating at you.
And now ask them to learn go. They can actually much easily do that.
I stand with this opinion that go code can be read / written by anybody. There is very rare bad go code but rust code feels "magic" to people.
Admire simplicity. That being said , I use archlinux with paru instead of yay , I use a lot of rust tools as well.
To be honest , to me it doesn't matter that much , what matters is if its being maintained stable / has active community. It could be written in any language just make it easy to embed / provide bindings for major languages like python node golang etc. Or though this might come overkill but I wish if there was some sort of glue code which can bridge languages , most of these glue code libraries end up having so many use cases. so I like the idea of wasm but its just not that performant. If we can have insanely performant wasm ...
it's of course very subjective, but for me golang feels more magic, especially with importing GitHub repositories like local files.
The easiest language is of course the one you already know.
i would also argue that it's easier to write bad go code (I personally wrote some of it while learning) in rust usually if it compile and your are not using unsafe, it's at least ok quality and clippy is great with guiding you how to improve your code.
to be fair JavaScript also is weird and different than OOP languages but from some time it tries to feel more traditional with classes and this handling.
wasm is already fairly performant in comparison to native code, it might lack in come cases like simd (there are extensions already, but I'm not sure about adoption) but generally it's only about 20% slower than native binary.
I'm not sure how much wasi is practical for cli tools, but using was for plugins to allow any language someone might be comfortable with seems like no-brainer to me (also isolation will allow you to control potential memory leaks and crashes so your application will not get corrupted or anonymously destabilised)
That would be nice. Instead we’re seeing all these JS/Python tools written in Rust - a complex language whose complexity comes providing memory safety without garbage collection.
Their authors are deliberately choosing to take on that complexity for programs where it’s unnecessary - programs which would work equally well with a GC.
Here on HN there is plenty of announces of new things “like X but in Rust” - even when X would be fine in Python or Bash.
The reason Rust and Go tooling is getting popular is because they are appropriate for the job.
C and C++ are riddled with footguns and non-ideal for this kind of tooling, Ruby and Python provide only disadvantages compared to current JS tools, Haskell and OCaml are good but too niche. And so on.
And the reason Rust and Go are getting usage in WASM is because they have a good WASM story. They are static, compiled and born at the right time. Other than C and C++, they're the only ones that put the effort.
The reason people are making projects in Rust and Go is because they are accessible to those people. For me both hit a sweet spot that none of the languages above do.
JVM is like the Cologne Cathedral of the software world.
And supposing you took start time out of it, how long would an equivalent Java program take to run?
Memory isn't as much of an issue, but all other things being equal users do prefer the program that consumes less memory.
That wouldn't be true for gccgo implementation, but then again, it is now abandoned in pre-generics Go, stuck at version 1.17, no one seems to care to update it.
Java doesn't even have null safety or ergonomic algebraic data types.
Besides people here sometimes have a hard time understanding English, given that I acknowledged Rust being faster.
And the success of ripgrep, mcfly/atuin, zoxide, hyperfine and starship, just reinforces the feeling that yeah, CLIs written in Rust can succeed.
So yeah, strongly disagree with you about Rust’s place.
But from your previous comments I also know you’re very knowledgeable about Rust so I’m surprised you came to this conclusion.
Depending on how many years into the past of HN headlines we would search for, it would be Go, Scala, Haskell, Ruby,...
Maybe all those Rust CLIs will be all being rewriting into Zig, as soon as version 1.0 gets out.
Roc language is already one of the first.
One of the reasons Rust succeeded, especially in CLI was
1. Low fixed memory overhead (compared to most popular VM based language - Python, Ruby, Java).
2. Startup time is nearly instantaneous, compared to the same languages. Don't have to wait 50-100ms or so just to start executing code.
3. Run time is faster as well, especially so for programs like ripgrep that use SIMD.
4. Single static binary with no dependencies. Didn't need a VM or toolchain to be installed on the user's computer.
Having a great CLI library like clap.rs brought it all together.
If you think this is simply a cycle and Zig will replace Rust, I'm slightly skeptical. Rust had an easier time competing with the status quo because of the 4 advantages it had.
Zig however, will have to compete with Rust. Sure, you can start ziggrep, but it'll be about equal in all these 4 factors. So why would anyone adopt something that has fewer features than ripgrep and about the same performance, memory consumption and ease of use? Just hype? Some people will, but not enough to get the critical mass ripgrep has.
That's why I'm bullish about Rust - the performance of existing codebases means that in many cases, it's hard to migrate away from it, or replace Rust programs with reimplementations. You'd have to be getting something really unique that Rust doesn't offer. You said Roc, and Roc is actually a great example of benefiting from something unique. Roc plans to reuse the Zig compiler's library that emits LLVM bitcode. There's no Rust equivalent.
But I'd argue that that's pretty niche. How many other examples are there of Zig libraries that have no Rust equivalent? Something that would make Zig the no-brainer choice over Rust, even if you had to reimplement in Zig.
Again, I'm sure people will try to rewrite things in Zig. But I'll confidently assert -> the most popular toolchain for Python in 5 years will be Rust based, not Zig or any other language. $10 on that.
And for CLI tools, you need those to be tiny. You're going to many of them, and you'll want to build a whole distribution, which you'll want to be small (for small images, small containers, etc).
Another benefit (that Zig won't offer) is good shared library support, which distribution maintainers will always prefer. I don't think there's a good language for that, but if one will appear, it will definitely be favored.
AFAIK, ripgrep's biggest deployment is through VS Code, by far. Nobody is going to notice that some part of VS Code is now using 8MB instead of 4MB (or whatever it is, I think it's around that). It's largely a non-issue from my perspective.
Now, compile times... That's a different story. I'll happily complain about that one.
For 1 really good utility like ripgrep, yeah. You'll convince many people to use it based on its merits. You might even have 5 or 10 of these utilities, if they're really good.
But for Linux system builders, for the things that go into standard distributions, where you need hundreds of such utilities, Rust CLI binaries are simply not sustainable. Not for size, not for security, not for simplicity of bootstrapping, nor, as you said, for compiling an OS from sources.
You will be able to get ripgrep into many distributions, but it will be more unlikely to become one of the tools installed by default for that reason.
Oh I don't agree with that all. It might be a factor, but it isn't going to be installed by default because grep will already be there. Unseating a POSIX user land in a Linux distro is not going to be blocked just on binary size IMO.
I also think what you're saying is not inconsistent with what I'm saying. The default tools installed on a Linux distro is a niche use case. ripgrep doesn't need to be installed by default in any Linux distro in order to be successful.
If you're setting out to build tools to replace, say, a GNU user-land to fulfill POSIX (or provide an entirely different sort of user experience), then binary size is probably the least interesting thing to overcome. And if you "try" hard enough, Rust binary sizes can be reduced quite a bit. ripgrep being 4MB is the result of me not caring at all about binary size (beyond stripping debug info). But this is a niche goal IMO, and IMO does not support your initial claim:
> And for CLI tools, you need those to be tiny.
Notice how if you instead said:
> And for CLI tools that are installed by default on a Linux distro, you need those to be tiny.
then it takes almost all the air out of your comment.
Pretty sure JVM is also a lottt more overhead.
Also since Java Loom project, it has virtual threads.
Additionally, since at least a decade you could your own thread scheduling if feeling so inclined to do some low level coding, as the scheduling architecture is customisable.
There is probably no language in existence that "ticks all the boxes". And this is good.
Fashion and hype has always played a big role in software and our industry has at times unwisely abandoned good technology in favor of the 'New hotness'. We gyrate between local apps and cloud apps, thin and thick clients, static and dynamic typing, micro-services and monoliths, relational databases and keyval stores. Our industry is not very rational and the best tools don't always win. Marketing and market effects (winners keep winning) play a big role.
Recognizing the role hype plays in our industry is not myopia.
The things people are shipping in Rust are stuff Rust is good for. As far as JS goes: mostly tooling.
Notice how virtually nobody is shipping websites in Rust. Sure there's frameworks and async, but we're not there yet.
The fact people are building libraries for servers, websites, gamedev, etc, in it is something that is required for experimentation.
People don't deserve the disrespect shown here for exercising their curiosity and trying to advance the field.
We all know that rust has a lot of merits, in that it is trying to establish a different, ambitious new trade off wrt. safety ganranties vs expressive power, and as a result is one a the few serious contender to replace C++ in the very areas where C++ is king, trading just a bit of performance for a vastly more sane language and better static garanties.
But it is also fair to acknowledge that more often than not, in the many "old_thing but in rust" that are being posted on HN, the old_thing at hand does not actually fall in that very important but also very small subset of programs that can not tolerate automatic memory management.
In other words, rust was not the good choice from a technical stand point.
Maybe it's particularly obvious to me because I have tried to promote rust before it became cool? At that time my coleagues could only tolerate Python or Go. These days the same people probably program in rust (or are at least considering it), and that's not because rust improved that much, or because they have sudenly learned to use the right tool for the job.
A lot of the things being posted in HN are not products but rather small-to-mid-size programs written by (quite often) experienced developers trying out a new language and somehow managing to scratch theirs or someone else's itch. On top of that, they are learning and we as a community are also learning from their hits and their misses.
A lot of things, such as WASM frameworks, are heavily experimental, and in this very thread there is an example of an author of those saying they're back to working JS. This is fine.
I would be totally fine with comments like "Rust is not good for this" or "Go is not good for that", and I have given some examples myself above. But this is clearly not what's happening here. It's ok to not want to participate on something, but the vitriol and flamewar-bait on those topics is not that.
To me the actual target of those critical comments are not the projects at hand but rather that trend in our industry to let popularity lead our decisions, and how easily we form small chapels that fight each others instead of encouraging learning and practicing several competing techs. Not that it will change anything, though.
And the top level comment about "cool kids" I answered to definitely doesn't have the nuance you suggest it has, nor does the flagged comment comparing it to a religion, or the comments calling people "hipsters". And this thread here is actually better than average.
It is no different than those folks, now 20 years later, that adopt Typescript, but really the only thing they do is to rename the file extension, and keep writing plain old JavaScript as always.
It has been the Most Loved/Most Admired language on Stack Overflow’s developer survey for 9 years in a row. I know that’s a relatively short amount of time for a programming language, but to me it seems like it has some serious staying power.
I guess 'automatic memory management' is the less controversial term which covers both traditional GC and RAII, the downsides are very similar though.
The article does mentions GC, but that's just demonstrating differences between JS and Rust.
The other popular JS-tooling alternative is esbuild, which is built in Golang, which has a GC.
If anything, the two most common arguments in favor of Rust for this specific use case are null-pointer safety and speed.
At this point, garbage just means whatever does not follow RAII. And RAII is not the best solution either. For throughput reasons, GC languages can be siginificantly faster exactly because it doesn't have to clean those "garbage" at every step.
Which GC languages are significantly faster than Rust and C++?
Which language? Does one exist?
The DOM API is defined as a JavaScript API. The browser, JavaScript, and the DOM API have been evolving in parallel for over 20 years. If you want to interact with the DOM from WASM, what you can do is write a function in JavaScript that interacts with the DOM, and then import that function into your WASM module.
At least for now.
The existing workarounds for this glue the JS DOM API to the guest language through glue on both sides, but this means that you have to pay the penalty of marshalling through JS for every single call you make.
There's a solution in the works to make direct interaction possible, but it's proven to be quite challenging to finalise: essentially, you need a high-level ABI that can provide a safe and fast interface between the host and whatever guest language, and defining that in a way that makes everyone happy is non-trivial.
For more information, see the Component Model [0] and WIT [1]; the idea is that you will eventually be able to compile your WASM to WASM Components, which can then interface with the host (or other WASM components) through the WIT IDL, including interacting with the browser's DOM.
[0]: https://github.com/WebAssembly/component-model [1]: https://github.com/WebAssembly/component-model/blob/main/des...
No, not even that (the second thing, passing pointers)! WASM has a different address space (starting from 0), effectively indexing into the WASM module's linear memory. A pointer/reference from the outside env can not be dereferenced from within the WASM bytecode.
The multiple linear memory proposal can circumvent this (allowing you to build something somewhat similar to a simple MMU/address translation engine), but language support for that feature is sparse (AFAIK, Rust does not expose any way to access a different than the default linear memory for example).
There is various hacks around it, but sharing memory block-wise between WASM and outside env is involved already.
It's defined via WebIDL (https://webidl.spec.whatwg.org/) which is theoretically language agnostic, but practically speaking the primary implementation target is JavaScript.
Definitely leading the way in segfaults.
That doesn’t mean I’m under the delusion that JS is going away anytime soon; at least from the FE. But its monopoly on the web took a big hit—and rightfully so.
JS is a terrible language, and people who only know that one language will fight tooth and nail with anyone who opposes them.
This gave birth to the awful trend of using JS in the backend. I understand that sometimes JS is the only option for frontend work, but keep that junk out of my sysarch. There are better languages your backend can benefit from—Go, Rust, Zig, Kotlin, Python—anything but JS/TS.
I’m glad the industry has learned from its mistakes. I’m seeing far fewer Node.js backend jobs and a lot more Python, Go, Rust, and C# roles.
I’m also seeing a ton of Java, C#, and Go jobs in London. At my workplace, Wolt/DoorDash, all new services are being spun up in Go, where Node.js used to be the default.
Plus, LLMs have gotten really good at Node.js and Python, driving backend salaries down quite a bit. In this age, being a one-trick pony just won’t cut it anymore.
As for LLM's, from what I see the languages that LLM's are most often used for are those with the lower barrier to entry / and those that UNI's and bootcamps teach, which would very much also include Java. However, what I see is actually my salary going up as the average candidate quality has been also decreasing, almost in correlation with the usage of LLM's. So while maybe the junior/mid-level salary goes down, the senior salary, in my observations, has not. And I keep hearing from companies how finding actually capable developers has become harder and harder, so I expect my salary to only go up.
Now, of course, being a one trick pony has never been a good thing, I agree on that (did you assume that I'm a just JS dev?), but I was just answering to your original comment.
Right, famous developer Ryan Dahl who knows only one language.
Come everyone. C is dead, Javascript is dying.
Also it is used by the bigdiks, like facebook who can't make a proper comment system, yet they market react as another shiny thing. And Apple who always turn your bluetooth back on after update. And google, who act like inbred mutants thanks to the incest that is going on there. A W S, the three letter magic thing also uses it. Plus cloud!
And the list goes on! Plus it has fugly syntax, but who cares, it is memory safe!
Jump onboard, this is the future!