5 ms·

The Linux Kernel Will Soon Be MIT-Licensed and Copyleft Will Be Dead

26 points by bananamogul 8mo ago

gpcz 8mo ago
Good luck finding every contributor to the Linux kernel and getting them to agree to that.
- OneDeuxTriSeiGo 8mo ago
  lol that's not even their claim. their claim is that AI will rewrite the entire linux kernel from scratch with an MIT license. which. lmao
  - number6 8mo ago
    AI can't hold copyright (as for now) how will this affect licencing?
    - votepaunchy 7mo ago
      The MIT License exists because in the US you cannot make a work public domain. AI works being public domain is effectively equivalent.
OneDeuxTriSeiGo 8mo ago
lmao good luck with using an AI to try to rewrite the entire linux kernel. Doubly so good luck doing so without accidentally recycling a bunch of GPL code in the process.
- 7e 8mo ago
  The Linux kernel isn’t that complex. It’s mainly device drivers. Linus wrote the first version himself as a hobby, for crying out loud. AIs will be creating custom kernels within five years, from scratch, as customized as you wish. Probably based on BSD, or other liberally licensed code, for legal reasons.
  - OneDeuxTriSeiGo 8mo ago
    v0.01 took Linus months of work and even then it barely ran anything. It wasn't POSIX compliant. It couldn't run GNU userspace. It only ran on one arch and had no concept of multi-arch support and limited to no real driver compartmentalization. It wouldn't be until 2 years later that almost a hundred devs joined the Linux project and they'd gradually have achieved compatibility with GNU userspace. Full POSIX compliance wouldn't come until later. And it wouldn't be until another year or two later that they'd finally reach a "production ready" 1.0 by which point hundreds of devs were contributing to the project. You are vastly underestimating the insane amount of work that went into the Linux kernel in the early days to get it to where it is today that it can be developed at such scale and with support for so many intermixed, overlapping, or mutually incompatible features, devices, and platforms. To call that effort (just for core linux, not the drivers) anything less than herculean is frankly an insult to all the people who have dedicated so much time and energy towards that project.
tptacek 8mo ago
This piece was more interesting than I went in expecting it to be. But it hinges I think on a disputable claim, that the GPL's value is based on a network effect, so that the GPL gets dramatically less useful (and thus attractive to newer projects) as the base of mainstream GPL software shrinks. I'm not sure I understand why that would be the case. The AI rewrite of the Linux kernel also seems farfetched. I don't think it really belongs in the title of this post.
- ignoramous 8mo ago
  chardet is a Python module ... with 170 million downloads ... licensed under the GPL. A different developer ... to include it in the Python standard distribution ... [reimplemented] it. With [Claude] ... in 5 days ... the reimplementation ... [has] better performance. May be LLMs learning from all of the FOSS code out there, and out-performing a team of domain experts, is helping us realise the promise of free as in beer.
  - vips7L 8mo ago
    I guess the question is: was Claude trained on chardet? Is that really clean room?
    - spwa4 8mo ago
      What do you mean? It's pretty damn clear how copyright and courts work, but let's review: 1) when a large copyright conglomerate sues a teenager: 30.000 USD per violation + enforcement of the license by an armed police officer coming to their house. 2) when a teenager has copyright violated by a large copyright conglomerate, or any large company: change the rules and force a compromise, resulting in $0 in damages. In fact it has just been made clear that copyright DOES NOT protect against AI training. That's "fair use". But let's pretend the courts are fair. Let's say I take all disney movies and tell AI to make a new one. Will courts suddenly decide Disney owns the copyright on both the model AND the movie? 100%.
    - ignoramous 8mo ago
      > was Claude trained on chardet? Is that really clean room? Not a true clean room; the TFA notes: ... Every file is different. Automated plagiarism checks report only 1.3% in common code ... the reimplementation looks very clean. Different code, different algorithms, and better performance.
iamnothere 8mo ago
AI hopium/copium. Why would you even rewrite it rather than finishing out Redox or Genode OS, which are built on better principles? This is silly.
WhereIsTheTruth 8mo ago
Linux was compromised by Big Tech a loooong time ago, we are lucky to have solid alternatives, but for how long?
notepad0x90 8mo ago
I've always wondered how many proponents of open source software also support the concept of copyrights in general. On one hand I see lot of people in these communities support the right to do whatever you want with the things you own, but on the other hand copyright licenses. I've always liked the MIT license because it is closer to copyleft than GPL variants. I get having to use a license so that people who use the software are legally protected, and attribution can be nice too, so while I don't agree with the legitimacy of software licensing as a whole, in practical terms the MIT is a good and safe license that doesn't impose lots of restrictions on its users. WTFPL and unlicense are better in my opinion, but lawyers might not like them. if you don't like the idea of lawyers running the world though, they're great. Public domain is the way. Even then, I despise the idea of even acknowledging "Public Domain" as a concept. But back to my original question, are most people using these licenses because they actually believe in their legitimacy? I always assumed it was to facilitate nonsensical copyright laws. From wikipedia: """ Copyleft is the legal technique of granting certain freedoms over copies of copyrighted works with the requirement that the same rights be preserved in derivative works. In this sense, freedoms refers to the use of the work for any purpose, and the ability to modify, copy, share, and redistribute the work, with or without a fee. Licenses which implement copyleft can be used to maintain copyright conditions for works ranging from computer software, to documents, art, and scientific discoveries. Similar approaches have been applied to certain patents. """ Ugh..yeah. Then I don't get it OP, I hope copyleft does indeed die. Either you have a commercial agreement or a contract with a person or you don't. The idea of publishing some work (software, book,etc..) and then by default and without any contractual agreement, dictating what a person does with that work that you published publicly is ridiculous. I know it is the law, but that doesn't make it right. What you put out there to the public, you have no right to control. Either close source your software and require a proper contractual agreement to use it, or make it actually free and actually open. It isn't free if you're going to tell me what to do with it later on, that's a deceptively hidden cost, it isn't really free. Wouldn't it be in the true spirit of open source and the Linux kernel if absolutley anyone can do whatever they want with the kernel's source code? The thing that irritates me is that if they want money, let them charge for it. No problem with that. You can charge for your work. But to make it "free" and then demand that the law enforces control over the software after someone has adapted it feels like such a crooked way of doing things. I would prefer to pay for it, so that I too can get some guarantees of the software being viable in return as well. It sounds to me like 1) You don't truly own copyleft software 2) You don't get any warranty, or expectation of viability 3) certainly, there is no expectation of support of any kind 4) It is marketed as 'free' (deceptively) if people make modifications to the software and keep that private later on, let them. it's free, it's theirs now, they can do what they want. If they value it though, it is in their interest to keep the upstream project viable, so sensible users will contribute back. Especially considering how globalized software dev has been, you can't even practically enforce something as weak as copyleft outside of the EU, Canada, Japan, and maybe (long shot) in the US.
friedtofu 8mo ago
Gross. Not sure about y'all but seeing an obviously AI-generated image at the top of the article is an instant nope/close tab for me. Not going to flag but these articles should be DOA.
ctoth 8mo ago
There is no test suite that captures kernel correctness. This is simply wrong. and will not happen.
- 7e 8mo ago
  Well the UNIX certification and POSIX comes close. But man pages might be all an AI needs.
up2isomorphism 8mo ago
Even GPL never seems to be great idea to me. But at least it considers a core problem of any knowledge sharing seriously. Can you just steal other people’s work? I would rather trust GPL folks than this guy.
7e 8mo ago
GPL licenses are already only a small minority of the total OSS world. But the GPL will die because AI can write equivalent software easily, and it’s more customized to boot. And for that reason OSS will also wither away; there’s little reason to upstream AI changes. Code is cheap and the FSF is a dinosaur.
bloppe 8mo ago
This article ignores the other side of the coin. You can argue that AI makes "clean room" re-implementation so cheap that copyleft is doomed. But then it follows that re-implementing proprietary software as free software would also become "free". So mb AI will kill copyleft, but then it would also kill the need for copyleft in the first place.

26 points by bananamogul 8mo ago

gpcz 8mo ago

Good luck finding every contributor to the Linux kernel and getting them to agree to that.

OneDeuxTriSeiGo 8mo ago
lol that's not even their claim. their claim is that AI will rewrite the entire linux kernel from scratch with an MIT license. which. lmao
- number6 8mo ago
  AI can't hold copyright (as for now) how will this affect licencing?
  - votepaunchy 7mo ago
    The MIT License exists because in the US you cannot make a work public domain. AI works being public domain is effectively equivalent.

OneDeuxTriSeiGo 8mo ago

lmao good luck with using an AI to try to rewrite the entire linux kernel. Doubly so good luck doing so without accidentally recycling a bunch of GPL code in the process.

7e 8mo ago
The Linux kernel isn’t that complex. It’s mainly device drivers. Linus wrote the first version himself as a hobby, for crying out loud. AIs will be creating custom kernels within five years, from scratch, as customized as you wish. Probably based on BSD, or other liberally licensed code, for legal reasons.
- OneDeuxTriSeiGo 8mo ago
  v0.01 took Linus months of work and even then it barely ran anything. It wasn't POSIX compliant. It couldn't run GNU userspace. It only ran on one arch and had no concept of multi-arch support and limited to no real driver compartmentalization. It wouldn't be until 2 years later that almost a hundred devs joined the Linux project and they'd gradually have achieved compatibility with GNU userspace. Full POSIX compliance wouldn't come until later. And it wouldn't be until another year or two later that they'd finally reach a "production ready" 1.0 by which point hundreds of devs were contributing to the project. You are vastly underestimating the insane amount of work that went into the Linux kernel in the early days to get it to where it is today that it can be developed at such scale and with support for so many intermixed, overlapping, or mutually incompatible features, devices, and platforms. To call that effort (just for core linux, not the drivers) anything less than herculean is frankly an insult to all the people who have dedicated so much time and energy towards that project.

tptacek 8mo ago

This piece was more interesting than I went in expecting it to be. But it hinges I think on a disputable claim, that the GPL's value is based on a network effect, so that the GPL gets dramatically less useful (and thus attractive to newer projects) as the base of mainstream GPL software shrinks. I'm not sure I understand why that would be the case. The AI rewrite of the Linux kernel also seems farfetched. I don't think it really belongs in the title of this post.

ignoramous 8mo ago
chardet is a Python module ... with 170 million downloads ... licensed under the GPL. A different developer ... to include it in the Python standard distribution ... [reimplemented] it. With [Claude] ... in 5 days ... the reimplementation ... [has] better performance. May be LLMs learning from all of the FOSS code out there, and out-performing a team of domain experts, is helping us realise the promise of free as in beer.
- vips7L 8mo ago
  I guess the question is: was Claude trained on chardet? Is that really clean room?
  - spwa4 8mo ago
    What do you mean? It's pretty damn clear how copyright and courts work, but let's review: 1) when a large copyright conglomerate sues a teenager: 30.000 USD per violation + enforcement of the license by an armed police officer coming to their house. 2) when a teenager has copyright violated by a large copyright conglomerate, or any large company: change the rules and force a compromise, resulting in $0 in damages. In fact it has just been made clear that copyright DOES NOT protect against AI training. That's "fair use". But let's pretend the courts are fair. Let's say I take all disney movies and tell AI to make a new one. Will courts suddenly decide Disney owns the copyright on both the model AND the movie? 100%.
  - ignoramous 8mo ago
    > was Claude trained on chardet? Is that really clean room? Not a true clean room; the TFA notes: ... Every file is different. Automated plagiarism checks report only 1.3% in common code ... the reimplementation looks very clean. Different code, different algorithms, and better performance.

iamnothere 8mo ago

AI hopium/copium. Why would you even rewrite it rather than finishing out Redox or Genode OS, which are built on better principles? This is silly.

WhereIsTheTruth 8mo ago

Linux was compromised by Big Tech a loooong time ago, we are lucky to have solid alternatives, but for how long?

notepad0x90 8mo ago

I've always wondered how many proponents of open source software also support the concept of copyrights in general. On one hand I see lot of people in these communities support the right to do whatever you want with the things you own, but on the other hand copyright licenses. I've always liked the MIT license because it is closer to copyleft than GPL variants. I get having to use a license so that people who use the software are legally protected, and attribution can be nice too, so while I don't agree with the legitimacy of software licensing as a whole, in practical terms the MIT is a good and safe license that doesn't impose lots of restrictions on its users. WTFPL and unlicense are better in my opinion, but lawyers might not like them. if you don't like the idea of lawyers running the world though, they're great. Public domain is the way. Even then, I despise the idea of even acknowledging "Public Domain" as a concept. But back to my original question, are most people using these licenses because they actually believe in their legitimacy? I always assumed it was to facilitate nonsensical copyright laws. From wikipedia: """ Copyleft is the legal technique of granting certain freedoms over copies of copyrighted works with the requirement that the same rights be preserved in derivative works. In this sense, freedoms refers to the use of the work for any purpose, and the ability to modify, copy, share, and redistribute the work, with or without a fee. Licenses which implement copyleft can be used to maintain copyright conditions for works ranging from computer software, to documents, art, and scientific discoveries. Similar approaches have been applied to certain patents. """ Ugh..yeah. Then I don't get it OP, I hope copyleft does indeed die. Either you have a commercial agreement or a contract with a person or you don't. The idea of publishing some work (software, book,etc..) and then by default and without any contractual agreement, dictating what a person does with that work that you published publicly is ridiculous. I know it is the law, but that doesn't make it right. What you put out there to the public, you have no right to control. Either close source your software and require a proper contractual agreement to use it, or make it actually free and actually open. It isn't free if you're going to tell me what to do with it later on, that's a deceptively hidden cost, it isn't really free. Wouldn't it be in the true spirit of open source and the Linux kernel if absolutley anyone can do whatever they want with the kernel's source code? The thing that irritates me is that if they want money, let them charge for it. No problem with that. You can charge for your work. But to make it "free" and then demand that the law enforces control over the software after someone has adapted it feels like such a crooked way of doing things. I would prefer to pay for it, so that I too can get some guarantees of the software being viable in return as well. It sounds to me like 1) You don't truly own copyleft software 2) You don't get any warranty, or expectation of viability 3) certainly, there is no expectation of support of any kind 4) It is marketed as 'free' (deceptively) if people make modifications to the software and keep that private later on, let them. it's free, it's theirs now, they can do what they want. If they value it though, it is in their interest to keep the upstream project viable, so sensible users will contribute back. Especially considering how globalized software dev has been, you can't even practically enforce something as weak as copyleft outside of the EU, Canada, Japan, and maybe (long shot) in the US.

friedtofu 8mo ago

Gross. Not sure about y'all but seeing an obviously AI-generated image at the top of the article is an instant nope/close tab for me. Not going to flag but these articles should be DOA.

ctoth 8mo ago

There is no test suite that captures kernel correctness. This is simply wrong. and will not happen.

7e 8mo ago
Well the UNIX certification and POSIX comes close. But man pages might be all an AI needs.

up2isomorphism 8mo ago

Even GPL never seems to be great idea to me. But at least it considers a core problem of any knowledge sharing seriously. Can you just steal other people’s work? I would rather trust GPL folks than this guy.

7e 8mo ago

GPL licenses are already only a small minority of the total OSS world. But the GPL will die because AI can write equivalent software easily, and it’s more customized to boot. And for that reason OSS will also wither away; there’s little reason to upstream AI changes. Code is cheap and the FSF is a dinosaur.

bloppe 8mo ago

This article ignores the other side of the coin. You can argue that AI makes "clean room" re-implementation so cheap that copyleft is doomed. But then it follows that re-implementing proprietary software as free software would also become "free". So mb AI will kill copyleft, but then it would also kill the need for copyleft in the first place.