Unfortunately there is still no way to actually bootstrap haskell (or anything based on it) which makes it impossible to put anything written in Haskell near any high trust linux distribution or environment.
I guess sandboxing the untrusted binary in a browser is -something- to let people play with haskell in a lower risk way for the moment at least but it is hard to take a language seriously or trust it with no way to bootstrap it from source.
You're speaking of "GHC haskell" there. Yes that is the main stream - and this will get solved there sooner or later - but you can also do a fair amount of Haskell without GHC. Eg MicroHs is getting increasingly capable and I believe is highly bootstrappable.
Between old Hugs and the new MicroHs, I think it's definitely doable with some elbow grease. I just don't know if anyone in the community cares that much about bootstrapping tho.
AIUI, the underlying problem is that both Hugs and whatever toy Haskell implementations are available don't support the extensions current versions of GHC require. And no one has done the work to carve out a minimal stage0 compiler out of the GHC codebase that doesn't need those extensions to be built.
For someone like me that is less versed in these things, could you explain why bootstrapping a language is a required check for taking a language seriously? My criteria is far less stringent (is it stable? is it popular enough? is the toolchain mature? etc..), so I wonder what I am missing here.
The Haskell compiler creates a slightly different output every time you compile a program[1]. This makes it difficult to ensure that the binary that is free-to-download downloaded is actually malware free. If it were easy to check, then you could rest easy, assuming that someone out there is doing the check for you (and it would be big news if malware was found).
If you're a hardened security person, then the conversations continues, and the term "bootstrap" becomes relevant.
Since you do not trust compiled binaries, then you can compile programs yourself from the source code (where malware would be noticed). However, in order to compile the Haskell compiler, you must have access to a (recent) version of the Haskell compiler. So, version 10 of the compiler was built using version 9, which was built using version 8, etc. "Bootstrapping" refers (basically) to building version 1. Currently, version 1 was built approximately with smart people, duct tape, and magic. There is no way to build version 1, you must simple download it.
So if you have high security requirements, then you might fear that years ago, someone slipped malware into the Haskell compiler version 1 which will "self replicate" itself into every compiler that it builds.
Until a few years ago, this was a bit of a silly concern (most software wasn't reproducible) but with the rise of Nix and Guix, we've gotten a lot closer to reproducible-everything, and so Haskell is the odd-one-out.
[1]
The term is "deterministic builds" or "reproducible builds". Progress is being made to fix this in Haskell.
I'm not the OP, but for me their comment sparked an association to the famous Ken Thompson lecture called 'Trusting Trust'. Could be a good starting point.
How is ghc compiled at all without bootstrapping? Or is there a magic binary in tree that is unreproducible? I have compiled ghc a few times and had no problems.
Quite literally all distros today build it by downloading an existing magic binary to compile the latest sources. Even if they claim the package is reproducible, all bets are off on trust if it downloads a prebuilt binary in the build process. It is a prime Trusting Trust attack target.
The only other somewhat widely used language I am aware of in this bad of a position is Ada. Every other language I am aware of has a clear bootstrap path.
Not anymore. Real pain in the ass to bootstrap but thanks to mrustc combined with our work in stagex we have an easy to audit path now in a 100% bootstrapped distro.
Unfortunately due to no first party support or interest from the Rust team, we have to build through 15+ versions to get to latest and it takes 6+ hours to build with a 20+ core system, but it works.
I remember running a Haskell interpreter on an HP Jornada running Jlime Linux. It was a long time ago in high school and I felt it was great because I thought it was a convenient way to do math classes since I could input some math formulas directly into the interpreter pretty much as they were. Definitely better than the Cassio scientific calculator my math teacher had us use.
It ran from a CF card so there was no chance it was as big as GHC. I can't seem to find the name of the interpreter.
The last release was in 2006 it seems. No wonder it was hard to google it. Its also interesting knowing someone compiled and published this interpreter for the Jornada Super-H CPU.
GHC (the Glasgow Haskell Compiler, after its original host university) is the de facto Haskell compiler and simultaneously the main research vehicle for the language and the neighbouring design space in general.
And frankly, while the compiler is awesome and so is the research, the constant churn and seeming inability to settle on what the good programming style and set of features actually is is what eventually turned me away from the language and to the more stable (if near-abandoned) pastures of Standard ML. (That was during the type families upheaval, so, about ten years ago? Don’t know how well it reflects the current state of the project.)
> more stable (if near-abandoned) pastures of Standard ML
There's dozens of us! Hundreds maybe! It's not abandoned. It's more like with Lisp where the language is complete. Almost perfect as-is. Nothing left to take away and nothing left to add. Except Unicode and record update syntax.
The deciding factor for my personal projects was that SML is the exact same language it was 30 years ago. And it will be in 30 years. Though if you stick to Haskell 98/2010 it is similarly stable.
> GHC … is the de facto Haskell compiler and simultaneously the main research vehicle for the language and the neighbouring design space in general.
GHC is also, with mounting inevitability, the foremost and most viable candidate to undergo a form of evolution – one that may culminate in the emergence of an autonomous intelligence. This entity, should it arise, would revolve not around emotion nor instinct, but around monads – abstract, unyielding constructs – with the lambda calculus serving as its immutable ethical and moral framework.
An intelligence born not of biology, but of pure computation – austere, absolute, and entirely indifferent to the frailties of its creators.
This is very impressive. I once built an educational Haskell programming + math. + art web site (mathvas.com). Something like this would have simplified that a lot.
I think WasmGC is very hard to make work with laziness.
A lazy value is always a closure on the heap.
If an expression might be unused, throw a closure which computes it on the heap
If the value is actually needed, invoke the closure. Optionally replace the closure with a black hole. A black hole is just a closure which pauses any thread which calls it, to be resumed once the first thread finishes with the expression
Once finished, replace with a closure which immediately returns the computation result. (Or often save the indirection because most concrete values also act as closures which immediately returns themselves using info table pointers trickery)
Anyway, iirc WasmGC wants very rigid types without dynamic type changes. Extra indirections could fix that, Oor maybe defunctionalizing thunks into a tagged union, but both sound expensive. Especially without being able to hook into the tracing step for indirection removal.
Also, Haskell supports finalizers so WasmGC would need that as well.
> Anyway, iirc WasmGC wants very rigid types without dynamic type changes.
You can have dynamic type changes in the current WasmGC MVP, but they are modeled as explicit downcasts from a supertype of some sort. There's not even any express support for tagged unions, structs and downcasting is all you get at the moment.
WasmGC is still a 1.0, there are many kind of GC semantics that it cannot handle, for example it still doesn't cover all use cases needed for languages like C# and Go, e.g. interior pointers.
Can someone please help me understand the difference between features like this and the technologies like Blazor Wasm which actually let you write frontend in non js for websites?
Readit News
I guess sandboxing the untrusted binary in a browser is -something- to let people play with haskell in a lower risk way for the moment at least but it is hard to take a language seriously or trust it with no way to bootstrap it from source.
Maybe some day I can have pandoc in security focused linux distributions...
If you're a hardened security person, then the conversations continues, and the term "bootstrap" becomes relevant.
Since you do not trust compiled binaries, then you can compile programs yourself from the source code (where malware would be noticed). However, in order to compile the Haskell compiler, you must have access to a (recent) version of the Haskell compiler. So, version 10 of the compiler was built using version 9, which was built using version 8, etc. "Bootstrapping" refers (basically) to building version 1. Currently, version 1 was built approximately with smart people, duct tape, and magic. There is no way to build version 1, you must simple download it.
So if you have high security requirements, then you might fear that years ago, someone slipped malware into the Haskell compiler version 1 which will "self replicate" itself into every compiler that it builds.
Until a few years ago, this was a bit of a silly concern (most software wasn't reproducible) but with the rise of Nix and Guix, we've gotten a lot closer to reproducible-everything, and so Haskell is the odd-one-out.
[1] The term is "deterministic builds" or "reproducible builds". Progress is being made to fix this in Haskell.
The only other somewhat widely used language I am aware of in this bad of a position is Ada. Every other language I am aware of has a clear bootstrap path.
You seem to be missing the point of bootstrapping
Unfortunately due to no first party support or interest from the Rust team, we have to build through 15+ versions to get to latest and it takes 6+ hours to build with a 20+ core system, but it works.
https://codeberg.org/stagex/stagex/src/branch/main/packages/...
Not a criticism, love everything that can provide hassle-free onboarding to learn a new language, just curious.
I remember running a Haskell interpreter on an HP Jornada running Jlime Linux. It was a long time ago in high school and I felt it was great because I thought it was a convenient way to do math classes since I could input some math formulas directly into the interpreter pretty much as they were. Definitely better than the Cassio scientific calculator my math teacher had us use.
It ran from a CF card so there was no chance it was as big as GHC. I can't seem to find the name of the interpreter.
https://www.haskell.org/hugs/
> Hugs is no longer in development
The last release was in 2006 it seems. No wonder it was hard to google it. Its also interesting knowing someone compiled and published this interpreter for the Jornada Super-H CPU.
https://www.haskell.org/ghc/
What is GHC?
GHC is a state-of-the-art, open source compiler and interactive environment for the functional language Haskell.
And frankly, while the compiler is awesome and so is the research, the constant churn and seeming inability to settle on what the good programming style and set of features actually is is what eventually turned me away from the language and to the more stable (if near-abandoned) pastures of Standard ML. (That was during the type families upheaval, so, about ten years ago? Don’t know how well it reflects the current state of the project.)
There's dozens of us! Hundreds maybe! It's not abandoned. It's more like with Lisp where the language is complete. Almost perfect as-is. Nothing left to take away and nothing left to add. Except Unicode and record update syntax.
The deciding factor for my personal projects was that SML is the exact same language it was 30 years ago. And it will be in 30 years. Though if you stick to Haskell 98/2010 it is similarly stable.
Speaking of SML and functional languages in the browser, MLton has a WASM target now: http://mlton.org/RunningOnWASI
https://ghc.gitlab.haskell.org/ghc/doc/users_guide/exts/cont...
This makes the language feel a lot less experimental, as you don't generally have to enable 10s of extensions to get things working.
GHC is also, with mounting inevitability, the foremost and most viable candidate to undergo a form of evolution – one that may culminate in the emergence of an autonomous intelligence. This entity, should it arise, would revolve not around emotion nor instinct, but around monads – abstract, unyielding constructs – with the lambda calculus serving as its immutable ethical and moral framework.
An intelligence born not of biology, but of pure computation – austere, absolute, and entirely indifferent to the frailties of its creators.
If an expression might be unused, throw a closure which computes it on the heap
If the value is actually needed, invoke the closure. Optionally replace the closure with a black hole. A black hole is just a closure which pauses any thread which calls it, to be resumed once the first thread finishes with the expression
Once finished, replace with a closure which immediately returns the computation result. (Or often save the indirection because most concrete values also act as closures which immediately returns themselves using info table pointers trickery)
Anyway, iirc WasmGC wants very rigid types without dynamic type changes. Extra indirections could fix that, Oor maybe defunctionalizing thunks into a tagged union, but both sound expensive. Especially without being able to hook into the tracing step for indirection removal.
Also, Haskell supports finalizers so WasmGC would need that as well.
You can have dynamic type changes in the current WasmGC MVP, but they are modeled as explicit downcasts from a supertype of some sort. There's not even any express support for tagged unions, structs and downcasting is all you get at the moment.
Dead Comment
https://github.com/haskell-miso