Gossamer: a Rust-flavoured language with real goroutines and pause-free memory

by mwheeler • 27 days ago

92 90 27 days ago

quotemstr6• 27 days ago

Gossamer has a cycle collector and eager reference counting. Good luck dropping the last reference to a 10,000-node graph, especially if cyclic. That means it doesn't have "pause free" memory. If you want pause freedom, go use ZGC or another modern GC on a modern VM.

I just can't take seriously this spate of languages that ignore the past 30 years of research into automatic memory management. We have multiple open-source pauseless miracles GCs right there before our eyes, yet it's the trendy thing in language design to foist memory management on users.

You don't even have to use a big VM if you want good GC. Go use MPS. Lots of options out there, even if you want to implement your own VM.

platinumrad3• 27 days ago

We have multiple open-source pauseleses miracles right there before our eyes

Is this meaningfully true in a practical sense? I've been writing code with soft real-time requirements and I don't think your notion of "pauseless" suffices. And if these miracles are open-source and right before our eyes, why do languages like Crystal and D still use Boehm?

quotemstr6• 27 days ago

The charitable explanation is the authors lack the time to rebase onto something more modern.

gomoboo• 27 days ago

D uses a homegrown GC not Boehm: - https://dlang.org/spec/garbage.html - https://dlang.org/blog/2017/03/20/dont-fear-the-reaper/

nu11ptr2• 27 days ago

why do languages like Crystal and D still use Boehm?

Languages use Boehm for exactly one reason: it is easy to shim into an otherwise manual memory system (it was designed for use in C/C++). I mean no respect to its authors, but using Boehm in production is the worst of all worlds: slow allocations (free list allocator), poor cache locality, and not precise (so you can expect memory leaks). If you are going to do a GC language you want: 1) precise 2) bump allocator 3) compacting collector 4) generations. Essentially you want to allocate fast, only touch live objects (most objects die young), compact them for locality, and only process objects each cycle of similar age. There is a huge amount of engineering that goes into a state of the art collector, but those are the basics.

iyn• 27 days ago

We have multiple open-source pauseless miracles GCs right there in front of us

Can you share some links/references?

quotemstr6• 27 days ago

ZGC is extremely good work.

https://wiki.openjdk.org/spaces/zgc/pages/34668579/Main

ZGC performs all expensive work concurrently, without stopping the execution of application threads for more than a millisecond. It is suitable for applications which require low latency. Pause times are independent of the heap size that is being used. ZGC works well with heap sizes from a few hundred megabytes to 16TB.

Go's GC is also very good: https://go.dev/blog/greenteagc.

V8's Orinoco is also pretty good now. It's improved a lot over the past decade and is now mostly-parallel. (A decade is about how long one of these things takes: high-performance GC is hard.)

I'm also a fan of MPS: it's a big of dark horse because it's more a GC construction kit than a ready-to-go GC, but it's fast and flexible, and I'd start with it any day over Boehm if I were making a VM from scratch.

elitepleb• 27 days ago

https://github.com/pizlonator/fil-c/blob/deluge/libpas/src/l...

smj-edison• 27 days ago

The one plus I'll give reference counting is it still takes the cake for interoperability with C. Which is only important if you need good interoperability, but when you do, tracing GCs don't play nice.

poulpy123• 27 days ago

Sorry but the name means bitterkid in french, I can't get over it :D

ambicapter• 27 days ago

I'm french and had to think about it for a little bit.

mkl2• 27 days ago

https://en.wiktionary.org/wiki/gossamer#Noun

throwrioawfo• 27 days ago

There was once a time when I'd see a page like this and think "wow, must be a great project with such a polished website".

Now, it's just a neutral or perhaps even very slightly negative signal (especially the em-dash in the very first line of the page).

Anyone able to tell me if this is a project actually worth paying attention to, or just another raindrop in the current monsoon of slop?

klardotsh• 27 days ago

Somewhat with you on this. I got slightly excited for a brief moment, but then the site starts to scream "an LLM threw this together super quickly" which doesn't spark joy at all.

I then started digging into the code examples and quickly determined that nothing about this project is for me, even as a fan of Rust and some of its influences it has on recent languages. That web routing example is absolutely gross to my eye, for example.

Different strokes for different folks - my own thoughts on language design (I'm hacking on one in private over the past several years, maybe one day it'll be shareable) would probably make some folks have a similar reaction, despite taking a wildly different approach than here. But it does suck to see Yet Another Vibe Looking Site hosting a language that feels like Yet Another Flavor Of Similar Stuff. Really looking forward to a language that wildly shakes things up in a usable way, and has a lot of care put into the DX... this one did not check that box for me.

charlieflowers2• 27 days ago

The language choices demonstrate excellent taste though.

mattkrause• 27 days ago

The web-routing thing is especially gross because a different example shows off pattern-matching.

bscphil• 27 days ago

It's clearly vibecoded if you look at the commit history.

Zak2• 27 days ago

A card grid with rounded corners as the second section and a dark blue or purple theme just scream "designed by Claude".

It's decent design, but not a useful quality signal.

felixgallo• 27 days ago

you know what's even worse than slop? Zero-effort 'is this maybe slop can someone tell me' posting.

cure_422• 27 days ago

No.

A site for an entirely new programming language being slop is atrocious.

An entirely unimportant one minute comment on social media being low effort? Normal and expected and appropriate.

What on earth are you on?

sesteel• 27 days ago

Looks like we may be going to go back to the geocities days of primitive websites. I use still use hugo to generate my personal site automatically from markdown upon PR approval.

samuell• 27 days ago

Glad to see more languages adopt true goroutines [edit: lightweight threads or fibers] with M:N scheduling. Surprised more haven't. Among compiled language I'm only aware of Go and Crystal off the top of my mind.

kccqzy2• 27 days ago

Haskell does too. And it predates Go by a large margin, such that calling it goroutine is weird. And within Google, the C++ implementation fiber also predated goroutines. It really shows that this is more of a library feature rather than a language feature.

Balinares• 27 days ago

In fairness, goroutine is far catchier than >>=<%>.

throwaway17_175• 27 days ago

I agree with your objection to treating goroutines as the baseline implementation of fibers. However, I would disagree with categorizing the feature as a library feature vs a language level feature. In “low-level” languages (C, Rust, Zig, C++, etc) the ability and option exists for having ‘green threads’ with most of their commonly assumed characteristics be library based constructs (although see [1] for why that’s not true for threads in general ). However, almost all managed/runtime-required/scripting/“high level” languages lack the facilities to implement almost any realistic types of fibers and any FFI based implementations are going to suffer severe syntax integration issues (I am assuming somewhat on this point).

TL;DR Green threads are on a library feature for low level languages.

1 - Threads Cannot Be Implemented as Libraries (Boehm 2005), https://dl.acm.org/doi/10.1145/1064978.1065042

gf0002• 27 days ago

Haskell, Java

jeremyjh8• 27 days ago

Java and Crystal are not equivalent - neither have preemptible threads. Lots of other languages have co-routines or other cooperative systems - async/await are really just variants of that.

I didn't check if Gossamer is actually preemptible.

I believe the short list of production languages with preemptible M:N schedulers are limited to:

Erlang/Elixir

Haskell

Twey• 27 days ago

Green threads predate modern async/await by quite a long way. Since async/await was developed, recent languages tend to prefer the ‘principle of least surprise’ by making the yield points explicit, since they interact in important ways with the code around them, with the notable exception of Go (which is very weird considering how explicit they decided to make their error handling).

lstodd4• 27 days ago

IDK what's the fuss. m:n scheduled by io is .. 2000s-era. The implementation was so obvious in like 2005, that even I patched then Python 2.6 in, so we at my then company could get rid of Twisted.

Also let you remind that M:N scheduling was the FreeBSD's pthread implementation for quite a bit too long. No, it didn't play well with MySQL at the time.

rsyring• 27 days ago

If you are interested in Gossamer, you may also be interested in Lis, which is Rust flavored and compiles to go: https://github.com/ivov/lisette

From their readme:

  Safe and expressive:

    - Hindley-Milner type system
    - Algebraic data types, pattern matching
    - Expression-oriented, immutable by default
    - Rust-like syntax plus |> operator and try blocks
    - Go-style interfaces, channels, goroutines

  Quietly practical:

    - Interop with Go ecosystem (WIP)
    - Linter, formatter, 250+ diagnostics
    - Fast incremental compiler, readable Go
    - LSP for VSCode, Neovim, Zed, Helix, GoLand

etaioinshrdlu• 27 days ago

I would also like to toss my project into the ring, which also allows mixing most real-world Rust & Go packages together, on runtime & syntax compatible level. https://github.com/deepai-org/omnivm

Pardon the sloppy readme - it actually does work :)

jeremyjh8• 26 days ago

You were so preoccupied with whether or not you could, you didn't stop to think if you should.

pantsforbirds• 27 days ago

I'm fairly sure the code snippets aren't equal in the last python example:

```python names = sorted({name.lower() for name in users if name}) ```

vs.

Python is sorting a set (unique values only), but I'm not seeing a unique or set approach for gossamer.

mkl2• 27 days ago

Also the Python sorts by string content and the Gossamer sorts by string length.

IshKebab• 27 days ago

It's only 2 months old. Clearly vibe coded. Still, kind of crazy what you can vibe code now.

Also the actual language design seems quite nice. I'd love something like this that was embeddable (and not vibe coded). There are basically no good easily embeddable languages. Everyone used Lua but it sucks.

cure_422• 27 days ago

Why not embedded rust?

mkj• 27 days ago

Interesting idea, but wouldn't an embedded language normally be interpreted/runtime editable? I guess you could do it shipping a rust compiler, but it'd be big. A bit like Numba?

throwaway17_175• 27 days ago

If this is a joke, I actually chuckled.

But, maybe this is a terminology issue conflating an embeddable language with a language used for embedded programming. I think this is one of those CS issues prevalent for non-native English speakers (especially when those devs are relying on translations of text without LLM style semantic analysis of the source).

The first refers to the shipping of a language interpreter within the executable of a program, where that interpreter then processes “scripts”, data files, high level non-programmer editable object attributes or behaviors, etc. The most prevalent example of this is shipping a Lua interpreter inside a video game for interpreting hot reloadable asset scripts. The second usage of embedded is the implication that code is being run on a microcontroller or SoC ‘embedded’ within some other product or device.

jeremyjh8• 27 days ago

This is exactly the language I've been yearning for - the exact motivations and intersection of features that would be the sweet spot for me.

Kotlin without the Java baggage. Rust but with automated memory management and without async bifurcation. Go with a modern type system. Swift but with green threads and a linux community. Haskell without the hair shirt. Elixir with a full type system and native performance.

But yes there are some flags others have mentioned, I won't repeat them. "goroutines" ?

Even on the off chance this project is not entirely vibe coded how could it ever build an ecosystem? How can any new language? All the libraries would be suspect for the same reason this repo is. The agents won't know anything about it, no one will believe it can get momentum so it won't.

LunicLynx• 27 days ago

Just a comment: Rust does not have a garbage collector

jeremyjh8• 27 days ago

Thanks - edited - I meant to say "but with a" garbage collector. I like Rust, but I don't like thinking about ownership all the time. Although when I just make Codex do it its really not that bad.

Jtsummers• 27 days ago

I can't figure out the point of having both go and spawn. `spawn` seems to be an ordinary thread spawning mechanism, generating a handle that you can join with (but not cancel? can't find that for certain, it's not actually in the tour but is in the SPEC.md). `go` inherits all the problems of go routines. You can't join with it, you can't cancel it, you have to build other infrastructure on top of it to achieve those effects. As neat as golang was 17 years ago when it came out, this was and still is a major weakness of the language which has resulted in the community developing conventions (now standardized in the go standard library and particular usage patterns) around how to deal with it. I don't get why you'd half fix it (by introducing spawn) but then leave it in anyways. Just let the handle from spawn be ignored and you remove that particular footgun (it becomes a choice to ignore it, which still causes problems but at least the programmer chose to shoot towards their own foot).

reocha3• 27 days ago

Three things stick out to me on https://gossamer-lang.org/docs/migration/rust/

* No user macros at all. Six fixed format! / println!-family macros expand at parse time. - Meta programming is incredibly important in rust.

* (unsafe is) Forbidden at the language level. No unsafe keyword in Gossamer source. std is safe-Rust too. - No low level programming then.

* No move semantics. Non-trivial values are heap-allocated, reference-counted, and shared by reference; primitives are copied the same as Rust. - Again, no low level programming.

Calling this rust flavored (or even a systems programming language) seems a bit bold.

bfung• 27 days ago

Totally agree, misleading.

The syntax looks like rust, but esp w/the memory management model (reference counting), it’s going to have more overhead than rust when it’s running, more like Swift or at worse, Python.

charlieflowers2• 27 days ago

Originally I replied here that I thought you were both missing the point (but I was wrong).

I wrote: It has a garbage collector and goroutines, so clearly it is not trying to be a systems programming language.

Then ... I saw that it does indeed pitch itself as a systems programming language. So, I guess you both are right.

If Gossamer were to drop that claim, then I'd say it looks impressive to me. I have often wanted this particular mix of language features.

jeremyjh8• 27 days ago

Google called Go a "systems language". I agree its not the right term, but people do use it to describe a language used in infrastructure software like Docker or Kubernetes.

To me it makes a lot of sense as a starting point. The world does not need another Rust since we have a perfectly good one. People who like Rust's type system but don't care to think about ownership when writing application code, and who appreciate the language asyncing all the things invisibly have plenty to like here.

I do agree though any serious language MUST have an FFI capability, and macros or at least some form metaprogramming of seem like tablestakes. Green threads are not impressive if they are not preemptive. But its not describing itself as finished, its 0.19. Rust looked completely different at that stage in its development - I think it still had green threads itself at that point.

librasteve• 26 days ago

does rust have FFI, then? (aka stable ABI)

norman784• 26 days ago

This language reminds me of Borgo https://borgo-lang.github.io

SkiFire13• 26 days ago

It has a Rust-like syntax, enums, matching, traits, etc etc. Yes, it also loses a lot of special characteristics of Rust, but it has to be different somewhere. Moreover a lot of people like Rust as a high level language, i.e. ignoring the lower level capabilities and lifetimes, and this seems to be a direct response to that feeling.

Recurecur3• 26 days ago

All three points are erased by its ability to call Rust functions…

It’s not a bad approach, in that for most general programming one doesn’t need those things. Granted, macros can be convenient.

reocha3• 25 days ago

Any language with FFI support can call rust functions.

sdicker• 27 days ago

This post reminded me of Swift– a modern high-performance language that uses automatic reference counting. Its got a REPL and compiles via LLVM. It does trade goroutines for something with a little more safety built in.

Zak2• 27 days ago

I'm seeing a big red flag here for what purports to be a systems programming language: it isn't used for its own compiler. The compiler is written in Rust.

A systems programming language should be able to self-host its compiler. Writing compilers is one of the canonical systems programming tasks. Making that happen may not even be hard in the LLM and LLVM era as it's a fairly mechanical task for an LLM to execute, and you can output textual LLVM IR to bootstrap on any architecture LLVM supports.

throw10920• 27 days ago

Does it have a REPL? That's been one of the greatest sins of Rust - designing a compiled language after 1990 that doesn't have an interactive REPL.

Gossamer: a Rust-flavoured language with real goroutines and pause-free memory

Related Stories