Why not port the Linux kernel to Common Lisp? (2009)

http://stackoverflow.com/questions/1848029/why-not-port-linux-kernel-to-common-lisp

by lcasela • 13 years ago

29 28 13 years ago

rbanffy• 13 years ago

I can understand why the question was closed as not constructive, but it's still an interesting question. An OS written in Lisp targeted towards current needs (as opposed to the needs we had in the 70's) would be a worthy research goal.

But I wouldn't go for Common Lisp. I'd go for something smaller and cleaner.

perturbation• 13 years ago

I grant you that this would be an interesting research project, but Linus isn't going to rip out the internals of the linux kernel, start from scratch, and re-implement everything in Lisp just for kicks. Too much work for too little return, and Iron Lisp is still not as performant as C. Also(I would confidently wager) more kernel programmers are proficient in C than in Lisp, at least for systems programming.

It's difficult to take the question seriously rather than as an attention-grab.

zhemao5• 13 years ago

Not to mention that one of the main arguments for using Lisp was "Lisp programs are less buggy than C programs". That's assuming that you trust your Lisp implementation. All implementing it in Common Lisp does is move some of the potential bugs from the kernel to the Lisp implementation.

zhemao5• 13 years ago

Yeah, you'd need a more focused Lisp implementation specifically designed for writing low-level code. Among its features should be

1. The ability to manually manage memory instead of relying on a GC.

2. Low-level access to memory. As in, the ability to address an arbitrary location in physical memory

3. Inline assembly and the ability to link with assembly code

There have been operating systems written in Lisp in the past, such as the operating systems for the various Lisp Machines. https://en.wikipedia.org/wiki/Lisp_machine

betterunix3• 13 years ago

All of those features are already in SBCL, in one form or another. I think the issue is a little deeper than that; for example, the difficulty of producing a small executable in SBCL, one which does not come with the entire CL runtime attached to it. There is also the difficulty of debugging low-level code in SBCL.

In other words, the problem is not with Lisp, nor with a lack of support for low-level features, but with tooling. The tool support for writing a kernel in Common Lisp is pretty bad at the moment, and there are not many people trying to improve on that.

porlw2• 13 years ago

If you use a CPU that is specifically designed to run LISP, then you don't need any of that:

[http://dspace.mit.edu/handle/1721.1/5731]

hannibal5• 13 years ago

Common Lisp can be separated into core of 25 or so core operators and rest could be considered as library.

I think CL is in a way more operating system like than most other Lisps and would make good base for operating system. You can load and unload stuff, clean the system, etc.

616c• 13 years ago

For those are interested, I am sure it has been brought up before, the inverse has been brought up before, an in-kernel Scheme for debugging and some more crazy testing, schemix. [0] Unfortunately, the project is long since deprecated.

[0] http://www.abstractnonsense.com/schemix/

coryrc• 13 years ago

I updated it long ago to sort-of work with 2.6, but lost the code in a hard drive crash (right after getting commit access but before check-in, sigh), then lost interest on redoing it.

codex2• 13 years ago

It's turtles all the way down.

mburns2• 13 years ago

That idiom doesn't apply to this situation.

codex2• 13 years ago

Doesn't implementing a kernel in LISP really just move the interesting bits into the LISP GC, compiler, and runtime?

greenyoda• 13 years ago

Why port the Linux kernel to Lisp? What advantages would it have that would justify the huge amount of work that would have to go into it before it could even begin to compete with Linux?

Linux took many years to get to the point where it had the stability and performance to be able to compete with commercial operating systems. Users of Linux will not care what language the kernel is written in. They're only going to care whether it will run all of their existing applications without changes, as fast and reliably as Linux does. If the new system doesn't, it will never have any users. (People were willing to put up with the early versions of Linux because it was free and its competitors were not. Any new OS would have to compete with an existing free OS.)

Who would write and maintain it? Linus and most of his followers would stick with the C version, which they know and love, so where are we going to suddenly find the hundreds or thousands of contributors who are both expert Lisp programmers and expert kernel hackers?

derleth• 13 years ago

Why port the Linux kernel to Lisp?

For many of the same reasons people started writing OS kernels in C rather than assembly: Less scut work, wrong code looks wrong more often, and more opportunity for code reuse.

Other reasons include being able to provide a higher-level API, which only kinda happened with the move to C. A Lisp API would enable things like system-level garbage collection and possibly things like a single-address-space OS where the kernel is responsible for fine-grained memory protection and multiprocessing becomes as cheap as multithreading without becoming as unsafe as multithreading.

But that wouldn't be Linux anymore, I suppose; it would be a Linux-compatible system with extended features beyond what the C kernel provides. On the other hand, is Linux 3.11 the same as Linux 1.0 or is it merely mostly-compatible with Linux 1.0?

DigitalJack• 13 years ago

Less scut work is a good arguement when writing a kernel from scratch. The scut work has already been done in the linux kernel.

lukego2• 13 years ago

This seems like a perfectly reasonable reaction :-)

I would add: if someone really _did_ do all the work to write a whole modern operating system in Lisp, it would be criminally unimaginative to use the Linux ABI as your interface to the world.

You would have to write your own userspace too for the whole enterprise to be worth bothering with. And then you have basically written SqueakNOS (http://squeaknos.blogspot.ch). :-)

malandrew• 13 years ago

Are there any operating systems that run on an x86 machine that are open source, well-written and written in Common Lisp or Scheme? It doesn't need to be a production OS, just a toy one to learn from.

lukego2• 13 years ago

Movitz was neat back in the day (last update 2008): http://common-lisp.net/project/movitz/

That's a one man effort to write a Lisp OS for x86 from scratch. He wrote his own compiler, basically with the chain of macroexpansions bottoming out in WITH-INLINE-ASSEMBLER.

I only played with it once. I was struck by how cool it was that I could play with the Lisp REPL from inside the garbage collector (also written in Lisp). This was surprising but then seemed obvious: the GC is simply a Lisp function that's moving objects from old-space into new-space (and then deleting old-space), and it doesn't mind if you wake up and interleave a few more allocations into new-space.

Movitz was current back in 2004 when Pascal Costanza organized the first real european Lisp workshop in Oslo and a whole bunch of Lisp hackers who only knew each other through the 'net got to meet up in person. fun times.

acqq• 13 years ago

Anybody who claims that this has sense, just do it yourself and let me know when you succeed.

Why not port the Linux kernel to Common Lisp? (2009)

Related Stories