The relicensing of chardet
LWN.net needs you!Chardet is a Python module that attempts to determine which character set was used to encode a text string. It was originally written by Mark Pilgrim, who is also the author of a number of Python books; the 1.0 release happened in 2006. For many years, this module has been under the maintainership of Dan Blanchard. Chardet has always been licensed under the LGPL, but, with the 7.0.0 release, Blanchard changed the terms to the permissive MIT license. That has led to an extensive (and ongoing) discussion on when code can be relicensed against the wishes of its original author, and whether using a large language model to rewrite code is a legitimate way to strip copyleft requirements from code.Without subscribers, LWN would simply not exist. Please consider signing up for a subscription and helping to keep LWN publishing.
The fact that chardet is LGPL-licensed has indeed caused some unhappiness in the past. That license is incompatible with the requirements for the Python standard library, frustrating those who would like to see chardet become one of the "batteries" that are included with Python; that licensing has also blocked the inclusion of some other modules that use chardet. Blanchard bemoaned his inability to relicense the code back in 2021:
Unfortunately, because the code that chardet was originally based on was LGPL, we don't really have a way to relicense it. Believe me, if we could, I would. There was talk of chardet being added to the standard library, and that was deemed impossible because of being unable to change the license.
In 2026, though, that inability has, according to Blanchard, been overcome by virtue of a complete rewrite — done using Anthropic's Claude LLM — of the source. Pilgrim did not see it that way:
However, it has been brought to my attention that, in the release 7.0.0, the maintainers claim to have the right to "relicense" the project. They have no such right; doing so is an explicit violation of the LGPL. Licensed code, when modified, must be released under the same LGPL license. Their claim that it is a "complete rewrite" is irrelevant, since they had ample exposure to the originally licensed code (i.e. this is not a "clean room" implementation). Adding a fancy code generator into the mix does not somehow grant them any additional rights.
Blanchard, unsurprisingly, disagreed.
A clean-room reimplementation, he said, "is a means to an end, not the
end itself
", and that there are other ways to reach that end, including
an LLM rewrite. He pointed to results from a code-comparison tool showing
that there was almost no similarity between version 7.0 and the
previous versions, and concluded:
I then started in an empty repository with no access to the old source tree, and explicitly instructed Claude not to base anything on LGPL/GPL-licensed code. I then reviewed, tested, and iterated on every piece of the result using Claude. You can see the history of all the design and implementation plans that were used to create 7.0.0 here. I did not write the code by hand, but I was deeply involved in designing, reviewing, and iterating on every aspect of it.I understand this is a new and uncomfortable area, and that using AI tools in the rewrite of a long-standing open source project raises legitimate questions. But the evidence here is clear: 7.0 is an independent work, not a derivative of the LGPL-licensed codebase. The MIT license applies to it legitimately.
Simon Willison has observed, though, that the LLM did indeed access the LGPL-licensed source at one point. Beyond that, as others have pointed out, it is easy to ask an LLM to reimplement a body of code in a style different from the original, with the result that similarity checkers will see something entirely new. That does not necessarily break the derived-work link, though. Had an LLM been employed to translate chardet to, say, Lisp, the level of similarity would be quite low, but most would agree that the new code was derived from the original. The fact that the training corpus for Claude surely included all previous versions of chardet also muddies the picture.
A lot of people who are not lawyers have offered opinions on whether chardet 7.0 is derived from previous versions. I, too, am not a lawyer, and will not add to that pile. But it is worth saying that, if instructing an LLM to rewrite an existing body of code is sufficient to strip copyleft requirements from that code, then the future of copyleft looks even dimmer than it did before. But, then, the future of any sort of software licensing scheme could be threatened. The death of copyleft could, ironically, be part of its real goal: the end of copyright.
Meanwhile, of course, had Blanchard simply shown up with a new Python module, let's call it "detectchar", that implemented the same API as chardet, the overall level of eyebrow elevation would have been considerably lower. Replacing the existing code, under the same name but with a different license, drew a lot more attention to this move than it would have otherwise attracted.
Nobody involved in the current discussion is showing any sign of backing
down. That means the license change seems likely to stand, unless Pilgrim
decides to bring in real lawyers, which would be an expensive and uncertain
prospect at best. But if the change stands, it would not be surprising to
see a lot more people engaging in this sort of license-stripping exercise.
That may eventually lead to a court decision (or, more likely, a series of
conflicting decisions) on whether an LLM can be used to launder source code
in this way. The old Chinese curse — may you live in interesting times —
would certainly appear to be upon us.
Copyleft is a means, not an end
Posted Mar 5, 2026 21:13 UTC (Thu)
by geofft (subscriber, #59789)
[Link] (54 responses)
> But it is worth saying that, if instructing an LLM to rewrite an existing body of code is sufficient to strip copyleft requirements from that code, then the future of copyleft looks even dimmer than it did before. But, then, the future of any sort of software licensing scheme could be threatened. The death of copyleft could, ironically, be part of its real goal: the end of copyright.
Posted Mar 5, 2026 21:13 UTC (Thu) by geofft (subscriber, #59789) [Link] (54 responses)
This is a very important point! I think we as a community have gotten so wrapped up in (justifiably) protecting FOSS from people who want to subvert it to proprietary ends that we've become reliant on copyright as a tool, and forgotten that the goal of copyleft was to work around copyright, not to embrace it.
Suppose it is true (as it seems to be) that within a couple of years, you can point an LLM at any piece of software and say, give me a clean-room permissive-licensed version of this. Wouldn't that mean we've won?
The FSF's four freedoms are:
- The freedom to run the program as you wish, for any purpose (freedom 0).
- The freedom to study how the program works, and change it so it does your computing as you wish (freedom 1). Access to the source code is a precondition for this.
- The freedom to redistribute copies so you can help others (freedom 2).
- The freedom to distribute copies of your modified versions to others (freedom 3). By doing this you can give the whole community a chance to benefit from your changes. Access to the source code is a precondition for this.
I would argue that we've been significantly helped by the fact that every major LLM provider has steamrolled over copyright in their training process and we now have a significantly expanded view of fair use, i.e., that the output of LLMs is not considered a derivative work of every work in the training set, and that they owe at most financial compensation to authors for the initial copies but not for ongoing inference on the trained model. If megacorporations had all lobbied to, say, limit copyright to two weeks after the death of the author instead of 70 years, we probably would cheer that on, even though it would weaken the GPL compliance case for some software and even though it would transparently be in the service of the megacorporations being more profitable. I think we should see the current situation similarly.
It is weird to admit that the GPL has run its course. But that was already true for the megacorporations anyway. Richard Stallman convinced Steve Jobs that he had to release the source for his patches to GCC, which worked for a while, but eventually Apple invested heavily in the development of LLVM and was free from any requirements of GCC's license. Now the ability to do that is in the hands of everyone, which certainly feels worse, but I would argue it's strictly better from only being in the hands of Apple-sized companies. And as it happens, Apple still posts their patches to LLVM publicly, not because they're obligated to, but because they get better results from opening up their code and sending it upstream. (There's also an interesting argument about that link being about CLISP and Readline, and how Readline is certainly no longer leverage with the existence of libedit and several other options.)
I also wonder how proprietary software companies are feeling about this. The example in the comment thread of Windows and ReactOS is a good one. Or how do companies making proprietary, out-of-tree Linux drivers feel? Is there a world where they are soon convinced that there is no competitive advantage in shipping a closed-source Linux driver, because their competitors can learn everything that's in that driver anyway, and so they decide to open-source things because this brave new LLM world means that source code isn't as much of an asset? That also seems like it would be a significant victory.
Copyleft is a means, not an end
Posted Mar 5, 2026 22:47 UTC (Thu)
by bluca (subscriber, #118303)
[Link] (52 responses)
Posted Mar 5, 2026 22:47 UTC (Thu) by bluca (subscriber, #118303) [Link] (52 responses)
We should, except we are now finding out that, as it is extremely often the case, it was really never about free software in itself, for many people. It is, and always was, a matter of identity. A very large number of people have built their whole personal identity around the idea of free software as an end, not as a means, and so they unsurprisingly lash out in violent, angry and cringeworthy rants against a mantainer who has spent ~15 years working alone to maintain a free software project alive, as we can see in some of the comments in the GH issue linked in the article. Not because they really care about some random python module that they never knew existed until last week, and that they'll have forgotten about by next week, but because they don't know how to answer the question: "what happens to me after the thing I've built my whole identity upon collapses into dust?" - and that must be truly terrifying
Copyleft is a means, not an end
Posted Mar 6, 2026 0:40 UTC (Fri)
by josh (subscriber, #17465)
[Link] (51 responses)
Posted Mar 6, 2026 0:40 UTC (Fri) by josh (subscriber, #17465) [Link] (51 responses)
The maintainer violated the license on the work of prior contributors, falsely claimed a "clean room" reimplementation, and uploaded the resulting slop as a "new version" of an existing package.
As with attempts to bait-and-switch Open Source projects with proprietary alternatives, this needs a fork of the last LGPLed version, and a backport of the useful optimizations in the new version.
> Not because they really care about some random python module that they never knew existed until last week, and that they'll have forgotten about by next week
I've known about chardet since long before LLMs existed. If its maintainer was seeking help, I would happily have amplified that call for help and sent some sponsorship its way. But its maintainer wasn't seeking help; its maintainer was looking to replace copyleft code with permissive code.
AI reimplementations aren't some magical realization of the FOSS dream. They're laundering and producing fallout that's drowning out and alienating human collaboration.
Copyleft is a means, not an end
Posted Mar 6, 2026 0:43 UTC (Fri)
by josh (subscriber, #17465)
[Link] (44 responses)
Posted Mar 6, 2026 0:43 UTC (Fri) by josh (subscriber, #17465) [Link] (44 responses)
Copyleft is a means, not an end
Posted Mar 6, 2026 13:33 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (43 responses)
Posted Mar 6, 2026 13:33 UTC (Fri) by dskoll (subscriber, #1630) [Link] (43 responses)
if copyright were abolished tomorrow, I'd celebrate
I wouldn't. I happen to think that composers, authors, photographers, film-makers and other artists should be compensated for their work, and copyright is pretty much the only means they have to guarantee such compensation.
I think a world without any copyright protections would be much worse and it would make artists and creators lives even more precarious than they are now.
Copyleft is a means, not an end
Posted Mar 6, 2026 13:48 UTC (Fri)
by bluca (subscriber, #118303)
[Link] (35 responses)
Posted Mar 6, 2026 13:48 UTC (Fri) by bluca (subscriber, #118303) [Link] (35 responses)
They live precarious lives because of capitalism and large corporations syphoning off all the wealth. Copyright helps those corporations to do so. Aside from a tiny minority of ultra-rich super-stars, everyone else would be better off.
Copyleft is a means, not an end
Posted Mar 6, 2026 15:18 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (34 responses)
Posted Mar 6, 2026 15:18 UTC (Fri) by dskoll (subscriber, #1630) [Link] (34 responses)
No, I disagree. Even small artists benefit from copyright. I get royalties still from comedy tracks I recorded a few years ago, and without copyright, I wouldn't be earning that money.
Copyleft is a means, not an end
Posted Mar 6, 2026 15:27 UTC (Fri)
by bluca (subscriber, #118303)
[Link] (33 responses)
Posted Mar 6, 2026 15:27 UTC (Fri) by bluca (subscriber, #118303) [Link] (33 responses)
Copyleft is a means, not an end
Posted Mar 6, 2026 15:29 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (32 responses)
Posted Mar 6, 2026 15:29 UTC (Fri) by dskoll (subscriber, #1630) [Link] (32 responses)
You are missing the point. First of all, I'm pretty happy with the royalties I receive relative to the amount of work I did to earn them. Secondly, without copyright protection, I would receive nothing instead of something.
So sure, let's stick it to those greedy corporations! And if independent artists lose their sources of income, well hey... collateral damage. ¯\_(ツ)_/¯
Copyleft is a means, not an end
Posted Mar 6, 2026 15:38 UTC (Fri)
by bluca (subscriber, #118303)
[Link] (31 responses)
Posted Mar 6, 2026 15:38 UTC (Fri) by bluca (subscriber, #118303) [Link] (31 responses)
Copyleft is a means, not an end
Posted Mar 6, 2026 15:57 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (29 responses)
Posted Mar 6, 2026 15:57 UTC (Fri) by dskoll (subscriber, #1630) [Link] (29 responses)
Copyright law needs reform, not elimination. At least, not before you have a proposal for something better and hard evidence that it will actually work.
But just imagining that getting rid of copyright will improve things, without having a viable alternative in place, is naive at best and disingenuous at worst.
Copyleft is a means, not an end
Posted Mar 6, 2026 16:18 UTC (Fri)
by bluca (subscriber, #118303)
[Link] (28 responses)
Posted Mar 6, 2026 16:18 UTC (Fri) by bluca (subscriber, #118303) [Link] (28 responses)
As for alternatives, there is plenty of evidence already that UBI works, and works well, for example.
Copyleft is a means, not an end
Posted Mar 6, 2026 19:35 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (27 responses)
Posted Mar 6, 2026 19:35 UTC (Fri) by dskoll (subscriber, #1630) [Link] (27 responses)
Sorry, but BS.
Copyright is necessary to protect and incentivize creators. In an earlier comment, you wrote: You get the crumbs, while a handful of corporations get literal truckloads of money.
Here's the reality: For streams of my comedy tracks, I get 90% and the record label gets 10%. For broadcasts on Sirius XM, I get 50% and the record label gets 50%. Streaming revenue is next to nothing because streaming companies are terrible, but Sirius XM royalties are decent.
If you think I'm getting "crumbs" while the record company is getting "truckloads", then it just goes to show you know nothing. And I'm fine with that split; the record label took the risk of producing the show, taping and producing the album, and marketing it.
UBI is great. But someone has to pay for it, and it provides no incentive for artists... if you get UBI whether your music is heard by 10 people or 100,000 people, where's the incentive?
The anti-copyright zealots are full of horsefeathers and live in some idealistic world that does not resemble reality in the slightest. Copyright reform could fix most of the problems and not cause additional harms.
Copyleft is a means, not an end
Posted Mar 6, 2026 19:57 UTC (Fri)
by bluca (subscriber, #118303)
[Link] (12 responses)
Posted Mar 6, 2026 19:57 UTC (Fri) by bluca (subscriber, #118303) [Link] (12 responses)
> UBI is great. But someone has to pay for it, and it provides no incentive for artists... if you get UBI whether your music is heard by 10 people or 100,000 people, where's the incentive?
You know, there are people out there who do those kind of activities because they actually _enjoy_ it. If you refuse to do "art" unless it gets you filthy rich, then actually I am pretty sure the world is better off without your "art". Everyone deserves to make a good living out of what they love doing, whether it's arts, crafts or whatever else. Nobody deserves to be filthy rich.
Copyleft is a means, not an end
Posted Mar 6, 2026 20:11 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (11 responses)
Posted Mar 6, 2026 20:11 UTC (Fri) by dskoll (subscriber, #1630) [Link] (11 responses)
Are you an artist who has income that would be lost if copyright were abolished? If not, then I don't think your opinion should count for much.
It's not just about me. I know many artists, comedians and musicians who make a reasonable living off of their work. They are not superstars by any means, and their income isn't enormous, but it's significant and not something they'd enjoy losing.
Again: Copyright reform to get rid of the leeches is a great idea.
there are people out there who do those kind of activities because they actually _enjoy_ it
Ah yes. Straight out of the mouths of exploitative producers who pay musicians and comedians in "exposure". I've certainly dealt with my share of them.
And thanks for saying that I said I need to be "filthy rich" from my creative endeavours; nothing like putting words into someone else's mouth to bolster your argument.
Copyleft is a means, not an end
Posted Mar 6, 2026 20:36 UTC (Fri)
by pizza (subscriber, #46)
[Link]
Posted Mar 6, 2026 20:36 UTC (Fri) by pizza (subscriber, #46) [Link]
...Not only does "exposure" not pay your bills, enough of it will quite literally kill you.
(and s/musicians/all creative endeavors, including most foss authors)
Copyleft is a means, not an end
Posted Mar 6, 2026 20:37 UTC (Fri)
by bluca (subscriber, #118303)
[Link] (9 responses)
Posted Mar 6, 2026 20:37 UTC (Fri) by bluca (subscriber, #118303) [Link] (9 responses)
What part of "Everyone deserves to make a good living out of what they love doing, whether it's arts, crafts or whatever else." is not registering exactly?
Regardless of whether it earns _you_ a satisfying amount of money, copyright has been perveted into an exploitative system that benefits the few at the expense of the many. Like many other aspects of capitalism, there's no "reforming" it, it needs to die.
Copyleft is a means, not an end
Posted Mar 6, 2026 20:43 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (8 responses)
Posted Mar 6, 2026 20:43 UTC (Fri) by dskoll (subscriber, #1630) [Link] (8 responses)
> Are you an artist who has income that would be lost if copyright were abolished?
What part of "Everyone deserves to make a good living out of what they love doing, whether it's arts, crafts or whatever else." is not registering exactly?
A simple "No" would have sufficed. Plenty of people make good livings doing things that don't require copyright protection. It's a little rich of them to say that people who do rely on copyright protection for their incomes should just abandon the protection and do without their incomes.
copyright has been perverted into an exploitative system
If it has been perverted, it can be de-perverted. But please don't promote a course that will harm millions of creators just to satisfy your personal vendetta against aspects of capitalism.
Copyleft is a means, not an end
Posted Mar 6, 2026 23:09 UTC (Fri)
by bluca (subscriber, #118303)
[Link] (6 responses)
Posted Mar 6, 2026 23:09 UTC (Fri) by bluca (subscriber, #118303) [Link] (6 responses)
I'm really not sure what is so hard to understand about:
"Everyone deserves to make a good living out of what they love doing, whether it's arts, crafts or whatever else."
It's really, really not that hard. Try reading it again a couple of times.
Copyleft is a means, not an end
Posted Mar 6, 2026 23:39 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (5 responses)
Posted Mar 6, 2026 23:39 UTC (Fri) by dskoll (subscriber, #1630) [Link] (5 responses)
You're being disingenuous. It is fantastically easy to trumpet:
Everyone deserves to make a good living out of what they love doing, whether it's arts, crafts or whatever else.
while not offering any feasible, proven way to do that. Heck, if it were up to me, everyone would have kittens and rainbows and all the chocolate they could eat while never gaining an ounce.
Go on. Come up with a feasible alternative to copyright that protects creators' livelihoods and that actually has a hope in hell of happening, and then maybe I'll take you seriously. Until then, you're just wasting electrons with meaningless slogans. (And don't say UBI. Much as I support UBI, it's not going to do the trick.)
Copyleft is a means, not an end
Posted Mar 6, 2026 23:49 UTC (Fri)
by bluca (subscriber, #118303)
[Link] (4 responses)
Posted Mar 6, 2026 23:49 UTC (Fri) by bluca (subscriber, #118303) [Link] (4 responses)
"give me an alternative"
"ok, here's the alternative"
"NO NOT LIKE THAT"
Copyleft is a means, not an end
Posted Mar 6, 2026 23:59 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (3 responses)
Posted Mar 6, 2026 23:59 UTC (Fri) by dskoll (subscriber, #1630) [Link] (3 responses)
I agree with you about UBI!
If you're claiming UBI will provide sufficient income for people to live on whatever they enjoy doing, you're dreaming in technicolor.
I am not aware of a single UBI program that's generous enough, by itself, to even make people's income meet the poverty level. If you know of one, post a link.
If you know of a way to give UBI to every single person in a country without causing enormous deficits, I'm all ears. As an example, I live in Canada and there are over 30 million of us over the age of 18. If we gave every one of those people $1750/month, which is much, much lower than the lowest minimum wage in Canada and certainly not enough to live on, we'd have to find $630B somewhere... about a quarter of our entire GDP. How do we do that? Magic?
Copyleft is a means, not an end
Posted Mar 7, 2026 0:08 UTC (Sat)
by bluca (subscriber, #118303)
[Link] (2 responses)
Posted Mar 7, 2026 0:08 UTC (Sat) by bluca (subscriber, #118303) [Link] (2 responses)
The most recent trial:
https://www.theguardian.com/world/2026/feb/10/ireland-bas...
> If you know of a way to give UBI to every single person in a country without causing enormous deficits
Yes. Tax the fucking rich and their corporations.
https://www.oxfamamerica.org/explore/issues/economic-just...
"12 people own more wealth than half the world"
That's where the money is - hoarded by literal dragons
Copyleft is a means, not an end
Posted Mar 7, 2026 0:26 UTC (Sat)
by dskoll (subscriber, #1630)
[Link] (1 responses)
Posted Mar 7, 2026 0:26 UTC (Sat) by dskoll (subscriber, #1630) [Link] (1 responses)
I'm sympathetic to taxing the rich. But even if we did that, it would not be enough for UBI for everyone on Earth. As I mentioned, this would cost one-quarter of the entire GDP of Canada were it implemented here, and would still not be enough to live on. We don't have enough rich people or corporations to tax to afford this. [We should tax them anyway, but...]
The Irish study gave a small number (2000) of artists €325/week. The poverty level in Ireland is for a single adult is €346/week. Great. 0.04% of Irish people were given income below the poverty level. Woo.
Now sure, the Irish artists in the study probably had other income... income that was protected because of copyright law.
UBI simply cannot be scaled up to be truly universal, even if we taxed all the rich and all the corporations on Earth. The numbers just don't work.
Copyleft is a means, not an end
Posted Mar 7, 2026 0:57 UTC (Sat)
by bluca (subscriber, #118303)
[Link]
Posted Mar 7, 2026 0:57 UTC (Sat) by bluca (subscriber, #118303) [Link]
Billionaires and corporations hoarding uninaginable amounts of wealth in tax havens don't.
Copyleft is a means, not an end
Posted Mar 8, 2026 21:52 UTC (Sun)
by da4089 (subscriber, #1195)
[Link]
Posted Mar 8, 2026 21:52 UTC (Sun) by da4089 (subscriber, #1195) [Link]
It might be that a world where creatives are justly rewarded for their work while unjust exploitation is avoided is possible, but eliminating copyright as a first step to that means a lot of collateral damage.
And without broad agreement on what the end goal should be, that damage will continue for some time until a new stable state is established.
So sure, right now there’s a heap of unjust exploitation enabled by copyright laws. But there’s also a lot of people who justly make a living off it too. Getting there from here needs more than just smashing the system.
Copyleft is a means, not an end
Posted Mar 6, 2026 20:16 UTC (Fri)
by marcH (subscriber, #57642)
[Link]
Posted Mar 6, 2026 20:16 UTC (Fri) by marcH (subscriber, #57642) [Link]
It looks like ACPS has already fixed this: https://support.deezer.com/hc/en-gb/articles/360002471277...
Now it's just the small matter of generalizing it. It's basic maths, so there's probably as little hope to get it fix as getting rid of "winner takes all" voting systems. Basic maths almost never makes a good TikTok story.
Copyleft is a means, not an end
Posted Mar 6, 2026 21:05 UTC (Fri)
by joib (subscriber, #8541)
[Link] (12 responses)
Posted Mar 6, 2026 21:05 UTC (Fri) by joib (subscriber, #8541) [Link] (12 responses)
Copyleft is a means, not an end
Posted Mar 6, 2026 21:23 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (11 responses)
Posted Mar 6, 2026 21:23 UTC (Fri) by dskoll (subscriber, #1630) [Link] (11 responses)
Not really. Participating in open-source is optional. If copyright is abolished, artists and creators will be forced into a new regime without any choice.
Copyleft is a means, not an end
Posted Mar 6, 2026 21:53 UTC (Fri)
by josh (subscriber, #17465)
[Link] (10 responses)
Posted Mar 6, 2026 21:53 UTC (Fri) by josh (subscriber, #17465) [Link] (10 responses)
There is no inherent right to a business model that depends on preventing copying. If that puts people out of business, so be it; copyright does more harm than good, now that far more people have the means to benefit from the ability to copy.
All that said, the worst of all worlds would be the one in which *large companies* can misappropriate licensed FOSS works and proprietary work alike as AI training data, while smaller entities are still punished for copying. I'd like to see copyright abolished *across the board*, not just waived for AI training.
Copyleft is a means, not an end
Posted Mar 6, 2026 22:42 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (9 responses)
Posted Mar 6, 2026 22:42 UTC (Fri) by dskoll (subscriber, #1630) [Link] (9 responses)
If that puts people out of business, so be it;
Very easy to say when you're not the one being put out of business.
now that far more people have the means to benefit from the ability to copy.
This is why copyright is needed. Back before modern technology, if you were a painter or a musician or an actor, you made your money directly by performing live or from patrons, and it was infeasible to simply copy your work to deprive you of income. Now that copying is so easy, why would anyone spend months or years working on a creative work, only to be completely unable to make money from it?
I'd like to see copyright abolished *across the board*
OK. How are you proposing to compensate musicians, performers, composers, authors, photographers, painters, and film-makers? Or or they part of your callous "so be it" attitude?
Copyleft is a means, not an end
Posted Mar 6, 2026 23:11 UTC (Fri)
by mb (subscriber, #50428)
[Link] (6 responses)
Posted Mar 6, 2026 23:11 UTC (Fri) by mb (subscriber, #50428) [Link] (6 responses)
This is completely different from what has been done with chardet, though.
Copyleft is a means, not an end
Posted Mar 6, 2026 23:40 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (5 responses)
Posted Mar 6, 2026 23:40 UTC (Fri) by dskoll (subscriber, #1630) [Link] (5 responses)
I was responding to bluca's comments, not the original article.
Copyleft is a means, not an end
Posted Mar 6, 2026 23:43 UTC (Fri)
by mb (subscriber, #50428)
[Link] (3 responses)
Posted Mar 6, 2026 23:43 UTC (Fri) by mb (subscriber, #50428) [Link] (3 responses)
More explicit: Please stop responding.
Copyleft is a means, not an end
Posted Mar 6, 2026 23:47 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (2 responses)
Posted Mar 6, 2026 23:47 UTC (Fri) by dskoll (subscriber, #1630) [Link] (2 responses)
That was uncalled for. If an LWN editor asks me to stop, then I will. But random people should not make such a request.
Copyleft is a means, not an end
Posted Mar 6, 2026 23:52 UTC (Fri)
by mb (subscriber, #50428)
[Link]
Posted Mar 6, 2026 23:52 UTC (Fri) by mb (subscriber, #50428) [Link]
Copyleft is a means, not an end
Posted Mar 7, 2026 0:35 UTC (Sat)
by corbet (editor, #1)
[Link]
I have been hesitant to intervene here; copyright is obviously highly relevant to both our community and this article. I do think that the time is coming, though, for the various folks involved in this discussion to conclude that any possible changing of minds will have occurred by now, and that perhaps it's time to declare victory and enjoy the weekend.
Posted Mar 7, 2026 0:35 UTC (Sat) by corbet (editor, #1) [Link]
Copyleft is a means, not an end
Posted Mar 6, 2026 23:44 UTC (Fri)
by dskoll (subscriber, #1630)
[Link]
Posted Mar 6, 2026 23:44 UTC (Fri) by dskoll (subscriber, #1630) [Link]
Ugh, sorry. I was responding to your "so be it" comment.
Copyleft is a means, not an end
Posted Mar 7, 2026 0:42 UTC (Sat)
by josh (subscriber, #17465)
[Link] (1 responses)
Posted Mar 7, 2026 0:42 UTC (Sat) by josh (subscriber, #17465) [Link] (1 responses)
How are you proposing to allow people to creatively remix and build upon literally everything that gets released the moment it's released? Or are you writing that off, as something you don't envision because you prioritize copyright higher?
As for your point, I pay literally well over a hundred per month to authors whose work I enjoy reading, even though their work will be released for free to everyone a few weeks later, and I'd pay that even if someone was posting all the content elsewhere. I subscribe to LWN, even though their content becomes free a week later. I work professionally as a software developer, and would continue to do so if copyright didn't exist. I would propose a combination of services, patronage, crowdfunding, UBI, and any number of other things that don't depend on copyright.
I'm not being callous about it. I acknowledge that the world will be vastly different without copyright. Some things will stop being as profitable; there will almost certainly be fewer billion-dollar movies, and far more indies. Some things will be more so. And on net the world will be better.
Copyleft is a means, not an end
Posted Mar 7, 2026 0:46 UTC (Sat)
by josh (subscriber, #17465)
[Link]
Posted Mar 7, 2026 0:46 UTC (Sat) by josh (subscriber, #17465) [Link]
Copyleft is a means, not an end
Posted Mar 6, 2026 15:58 UTC (Fri)
by dskoll (subscriber, #1630)
[Link]
Posted Mar 6, 2026 15:58 UTC (Fri) by dskoll (subscriber, #1630) [Link]
Sorry, forgot to add this in my other reply.
Do you work with or know many artists? Because I do, and they're certainly not the 0.01% of superstars. And I can tell you that not a single one of them wants to get rid of copyright and all of them would be fierce in their opposition to your position.
Copyleft is a means, not an end
Posted Mar 6, 2026 14:49 UTC (Fri)
by Wol (subscriber, #4433)
[Link] (6 responses)
Posted Mar 6, 2026 14:49 UTC (Fri) by Wol (subscriber, #4433) [Link] (6 responses)
The other thing, of course, is all those works that are being lost because nobody knows the copyright status, and nobody dares copy it.
Imho copyright should be maybe 25 years, and the ORIGINAL AUTHOR and their descendants *alive at creation* (real people) have the right to renew indefinitely.
For things like Mickey Mouse, or the Marvel Universe, or whatever, we should use trademarks not copyright.
Most works have no real value after ten years or so, LET them fall into the Public domain.
Cheers,
Wol
Copyleft is a means, not an end
Posted Mar 6, 2026 15:20 UTC (Fri)
by dskoll (subscriber, #1630)
[Link] (5 responses)
Posted Mar 6, 2026 15:20 UTC (Fri) by dskoll (subscriber, #1630) [Link] (5 responses)
Yes, I do agree that copyright term should be limited to 25 years after the creation of the work and should not be renewable. IMO, that would balance the interests of creators and of society.
Copyleft is a means, not an end
Posted Mar 6, 2026 17:47 UTC (Fri)
by mathstuf (subscriber, #69389)
[Link]
Posted Mar 6, 2026 17:47 UTC (Fri) by mathstuf (subscriber, #69389) [Link]
Copyleft is a means, not an end
Posted Mar 6, 2026 19:17 UTC (Fri)
by rgmoore (✭ supporter ✭, #75)
[Link]
Posted Mar 6, 2026 19:17 UTC (Fri) by rgmoore (✭ supporter ✭, #75) [Link]
It's not just balancing the needs of creators against the rest of society. It's also about balancing the needs of past and present creators. Copyright is a double-edged sword. On the one hand, giving a creator exclusive right to their work makes it easier for them to earn money from it. On the other hand, all creative work is derivative at some level, so giving past creators the power to interfere with present ones interferes with the creative process. Extending copyright benefits past creators at the expense of present ones. Of course the dichotomy isn't quite so precise; most present creators are also either real or anticipatory past creators, so cutting copyright terms too short would hurt them. There has to be some happy medium, and I'm pretty sure we've gone way overboard in making copyright terms too long.
Copyleft is a means, not an end
Posted Mar 7, 2026 0:09 UTC (Sat)
by Wol (subscriber, #4433)
[Link] (2 responses)
Posted Mar 7, 2026 0:09 UTC (Sat) by Wol (subscriber, #4433) [Link] (2 responses)
So you don't want to be able to retire on the benefits of your creativity?
There's a whole bunch of reasons behind the various extensions to copyright length - I don't know the reason for 50 years, or lifetime, or lifetime+50, but the 70 year extension is interesting - it's all those people who died young in the 1st or 2nd World War.
Imho, my proposal would allow copyright to leave a legacy to the artist's family - which could be important, but restricting it imho to real people who were around "at the start" gives a good compromise between "no legacy" and "protecting my family".
Cheers,
Wol
Copyleft is a means, not an end
Posted Mar 7, 2026 19:50 UTC (Sat)
by nix (subscriber, #2304)
[Link] (1 responses)
Posted Mar 7, 2026 19:50 UTC (Sat) by nix (subscriber, #2304) [Link] (1 responses)
No. No other kind of productive work gets to earn money for work done decades ago: you earn money as you work, and (in the case of things later sold) for a reasonable period afterwards: not indefinitely, and not for decades. Even copyright accepts this principle via the first-sale doctrine: you don't get to claim royalties on secondhand sales.
You retire based on pensions, investments, or other things of that nature, just like everyone else, not based on endless revenue streams for things you did fifty years ago.
Copyleft is a means, not an end
Posted Mar 8, 2026 9:04 UTC (Sun)
by Wol (subscriber, #4433)
[Link]
Posted Mar 8, 2026 9:04 UTC (Sun) by Wol (subscriber, #4433) [Link]
And what is a pension or an investment, but an endless revenue stream for things you did 50 years ago? (Plus, of course, copyrights are investments :-)
Be careful for what you wish for - do you really want to find the value of your work going through the floor when they abolish copyright, just as your need for revenue goes through the roof because they abolished copyright!
"There's always a solution that is simple, easy, and WRONG". I really don't see what's wrong with rewarding the creator of an artistic work, and I don't have any problem with that mechanism being copyright. It's all fiat money anyway.
The problem is the greed and rent-seeking culture that's grown up around it, with perpetual copyrights, and trying to tie everything up, and copyright franchises still spewing money long after the creator is dead - and the amount of work that's rotting because everybody's scared to preserve it because of copyright!
Cheers,
Wol
Copyleft is a means, not an end
Posted Mar 6, 2026 0:58 UTC (Fri)
by bluca (subscriber, #118303)
[Link] (3 responses)
Posted Mar 6, 2026 0:58 UTC (Fri) by bluca (subscriber, #118303) [Link] (3 responses)
Or, how about they can do whatever the heck they want with their project? Or are you paying their salary now?
> The maintainer violated the license on the work of prior contributors, falsely claimed a "clean room" reimplementation, and uploaded the resulting slop as a "new version" of an existing package.
Yeah, well, you know, that's just, like, your opinion, man.
> They're laundering and producing fallout that's drowning out and alienating human collaboration.
If stallman apologists and other such religious fanatics as it can be seen in that GH issue end up being alienated I, for one, won't be missing any of them. Time to find something else to tie their whole persona around. Copyleft is dead, copyright is dying, and not a moment too soon.
Copyleft is a means, not an end
Posted Mar 6, 2026 1:41 UTC (Fri)
by Kluge (subscriber, #2881)
[Link] (2 responses)
Posted Mar 6, 2026 1:41 UTC (Fri) by Kluge (subscriber, #2881) [Link] (2 responses)
Who says it's *their* project? It's only their project if they own the copyright, and that's what's at issue.
I won't bother responding to your other "points".
Copyleft is a means, not an end
Posted Mar 6, 2026 10:19 UTC (Fri)
by bluca (subscriber, #118303)
[Link]
Posted Mar 6, 2026 10:19 UTC (Fri) by bluca (subscriber, #118303) [Link]
Copyleft is a means, not an end
Posted Mar 6, 2026 10:35 UTC (Fri)
by kleptog (subscriber, #1183)
[Link]
Posted Mar 6, 2026 10:35 UTC (Fri) by kleptog (subscriber, #1183) [Link]
Depends on what you consider the "project". As just literal source, you may be right. But when it comes to the community: the ones actually doing the maintenance, managing releases, etc, then ISTM the original author ceded ownership a long time ago.
I've always been of the view that open-source projects are more like mini-cooperatives, with members doing their part. Ownership in such cases becomes a bit fuzzy. But someone who is not present at all is not a candidate.
The (L)GPL is means, not the goal. You have projects like PostgreSQL which maintain a vibrant community with no copyleft licence support.
Copyleft is a means, not an end
Posted Mar 6, 2026 14:44 UTC (Fri)
by Wol (subscriber, #4433)
[Link]
Posted Mar 6, 2026 14:44 UTC (Fri) by Wol (subscriber, #4433) [Link]
Is there any of that work left, to violate the licence of? Serious question, if all that work is at least 15 years old ... ie out-of-copyright as per the US's original copyright laws :-)
Cheers,
Wol
High-quality "slop"?
Posted Mar 9, 2026 4:17 UTC (Mon)
by gmatht (subscriber, #58961)
[Link]
Posted Mar 9, 2026 4:17 UTC (Mon) by gmatht (subscriber, #58961) [Link]
Copyleft is a means, not an end
Posted Mar 6, 2026 1:43 UTC (Fri)
by dvdeug (guest, #10998)
[Link]
Posted Mar 6, 2026 1:43 UTC (Fri) by dvdeug (guest, #10998) [Link]
Each person who has contributed has their own motivations. There's a lot of teamwork and community without getting ripped off motivations, the same sort of attitude that created a lot of non-commercial shared-source software like POV-Ray. A lot of people use GPL as the most protective license generally accepted as open source and thus by Debian, Red Hat and friends.
> If I can take an existing piece of software, perhaps even a binary, and point an LLM at it and get a permissively-licensed reimplementation
Except that's not really true. There's a lot of things that will always be non-trivial to clone; basically, unless you already have an open-source driver, you don't have enough information to write a hardware driver without reverse engineering and playing around with actual hardware.
> I can study how it works and change it, because I now have source code.
To the extent that's true, you clearly have a copyright infringement. You're not studying the original code, you're studying code that an AI wrote.
>If megacorporations had all lobbied to, say, limit copyright to two weeks after the death of the author instead of 70 years, we probably would cheer that on
No. I doubt even RMS would support his work under ND licenses, like much of the GNU webpages, going into the public domain right after he dies. It would have hurt historical uses of copyright, like Grant's writing of his autobiography right before he died (producing a valuable historical artifact) to provide money for his family.
>I also wonder how proprietary software companies are feeling about this.
Microsoft has huge budgets to sue for copyright infringement. Even in a world where you can feed a program into an AI and get out a copy that is by law a non-infringing work, that's only going to come after the big companies have spent years suing everyone.
If LLM can strip LGPL code of copyleft, it can strip copyright out of any copyrighted work
Posted Mar 6, 2026 1:08 UTC (Fri)
by atai (subscriber, #10977)
[Link] (6 responses)
Posted Mar 6, 2026 1:08 UTC (Fri) by atai (subscriber, #10977) [Link] (6 responses)
If LLM can strip LGPL code of copyleft, it can strip copyright out of any copyrighted work
Posted Mar 6, 2026 1:18 UTC (Fri)
by Cyberax (✭ supporter ✭, #52523)
[Link] (5 responses)
Posted Mar 6, 2026 1:18 UTC (Fri) by Cyberax (✭ supporter ✭, #52523) [Link] (5 responses)
Should that author be guilty of copyright infringement?
Harry Potter? What? I'm talking about "Starman Jones", of course.
If LLM can strip LGPL code of copyleft, it can strip copyright out of any copyrighted work
Posted Mar 6, 2026 1:41 UTC (Fri)
by pizza (subscriber, #46)
[Link] (3 responses)
Posted Mar 6, 2026 1:41 UTC (Fri) by pizza (subscriber, #46) [Link] (3 responses)
Well done, sir.. well done.
If LLM can strip LGPL code of copyleft, it can strip copyright out of any copyrighted work
Posted Mar 6, 2026 14:54 UTC (Fri)
by Wol (subscriber, #4433)
[Link] (2 responses)
Posted Mar 6, 2026 14:54 UTC (Fri) by Wol (subscriber, #4433) [Link] (2 responses)
No again I'm not talking about Hogwarts. He *really was* accused of copying Hogwarts in order to create Unseen University :-)
Cheers,
Wol
If LLM can strip LGPL code of copyleft, it can strip copyright out of any copyrighted work
Posted Mar 7, 2026 19:53 UTC (Sat)
by nix (subscriber, #2304)
[Link] (1 responses)
Posted Mar 7, 2026 19:53 UTC (Sat) by nix (subscriber, #2304) [Link] (1 responses)
If LLM can strip LGPL code of copyleft, it can strip copyright out of any copyrighted work
Posted Mar 8, 2026 8:54 UTC (Sun)
by Wol (subscriber, #4433)
[Link]
Posted Mar 8, 2026 8:54 UTC (Sun) by Wol (subscriber, #4433) [Link]
Cheers,
Wol
Starman Jones
Posted Apr 1, 2026 21:26 UTC (Wed)
by sammythesnake (guest, #17693)
[Link]
Posted Apr 1, 2026 21:26 UTC (Wed) by sammythesnake (guest, #17693) [Link]
Joke
Posted Mar 6, 2026 1:11 UTC (Fri)
by pabs (subscriber, #43278)
[Link]
Posted Mar 6, 2026 1:11 UTC (Fri) by pabs (subscriber, #43278) [Link]
https://malus.sh/
https://fosdem.org/2026/schedule/event/SUVS7G-lets_end_op...
Rewrite proprietary software too!
Posted Mar 6, 2026 1:15 UTC (Fri)
by pabs (subscriber, #43278)
[Link] (2 responses)
Posted Mar 6, 2026 1:15 UTC (Fri) by pabs (subscriber, #43278) [Link] (2 responses)
Rewrite proprietary software too!
Posted Mar 6, 2026 7:56 UTC (Fri)
by LtWorf (subscriber, #124958)
[Link] (1 responses)
Posted Mar 6, 2026 7:56 UTC (Fri) by LtWorf (subscriber, #124958) [Link] (1 responses)
Rewrite proprietary software too!
Posted Mar 7, 2026 2:45 UTC (Sat)
by pabs (subscriber, #43278)
[Link]
Posted Mar 7, 2026 2:45 UTC (Sat) by pabs (subscriber, #43278) [Link]
While LLMs can't yet do decompilation satisfactorily, they can clean up the output of NSA Ghidra satisfactorily, and convert between source code languages easily.
https://reorchestrate.com/posts/your-binary-is-no-longer-...
https://reorchestrate.com/posts/your-binary-is-no-longer-...
Another post claimed to be able to recreate a proprietary backend service when given FOSS frontend apps or even just a minified web frontend.
https://news.ycombinator.com/item?id=47259485
I can't find the post now but ISTR a post claiming LLMs were good at reversing WASM already.
win-win?
Posted Mar 6, 2026 1:20 UTC (Fri)
by shironeko (subscriber, #159952)
[Link] (1 responses)
Posted Mar 6, 2026 1:20 UTC (Fri) by shironeko (subscriber, #159952) [Link] (1 responses)
win-win?
Posted Mar 6, 2026 1:59 UTC (Fri)
by atai (subscriber, #10977)
[Link]
Posted Mar 6, 2026 1:59 UTC (Fri) by atai (subscriber, #10977) [Link]
No. Everything gets loose.
Conservancy
Posted Mar 6, 2026 1:26 UTC (Fri)
by pabs (subscriber, #43278)
[Link]
Posted Mar 6, 2026 1:26 UTC (Fri) by pabs (subscriber, #43278) [Link]
https://sfconservancy.org/blog/2026/mar/04/scotus-deny-ce...
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 1:34 UTC (Fri)
by marcH (subscriber, #57642)
[Link] (14 responses)
Posted Mar 6, 2026 1:34 UTC (Fri) by marcH (subscriber, #57642) [Link] (14 responses)
How does that prove anything? Are there several variants of Claude, some trained on (L)GPL code and others not? If a Claude variant had GPL code in its training, then that GPL code will influence its output no matter what the prompt is. This is only an LLM, it's not "intelligent".
I'm not saying that the influence of that GPL code in the training is enough to affect the license of the output (like everyone else, I have absolutely no idea) I'm only saying that the "explicit instruction not to base anything on (L)GPL code" in the prompt seems utterly ignorant and ridiculous. As ridiculous as asking chatbots not to lie.
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 1:39 UTC (Fri)
by bluca (subscriber, #118303)
[Link] (13 responses)
Posted Mar 6, 2026 1:39 UTC (Fri) by bluca (subscriber, #118303) [Link] (13 responses)
It proves something extremely important, as it was explained by that same comment: the old code was not in the local context of the tool.
The local context is extremely different from the training set. Training is performed under explicit exceptions to copyright law, so the license of the dataset is irrelevant. The local context is a whole different story, obviously.
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 9:30 UTC (Fri)
by excors (subscriber, #95769)
[Link] (1 responses)
Posted Mar 6, 2026 9:30 UTC (Fri) by excors (subscriber, #95769) [Link] (1 responses)
It doesn't prove that, because Claude Code is capable of downloading the old LGPL chardet code from GitHub and inserting that into its LLM context. From Simon Willison's post linked just after that quote, Claude Code wrote a plan that explicitly said it was going to do that for one source file. It's evidently quite happy to ignore the instruction to not base anything on LGPL code.
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 10:23 UTC (Fri)
by bluca (subscriber, #118303)
[Link]
Posted Mar 6, 2026 10:23 UTC (Fri) by bluca (subscriber, #118303) [Link]
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 14:03 UTC (Fri)
by marcH (subscriber, #57642)
[Link] (2 responses)
Posted Mar 6, 2026 14:03 UTC (Fri) by marcH (subscriber, #57642) [Link] (2 responses)
So what are all those debates raging about then? Copyright is dead, let's celebrate and move on!
Explicit where? Afraid I (and others) missed that legal milestone.
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 14:57 UTC (Fri)
by Wol (subscriber, #4433)
[Link] (1 responses)
Posted Mar 6, 2026 14:57 UTC (Fri) by Wol (subscriber, #4433) [Link] (1 responses)
EU law explicitly treats training an LLM the same as training a schoolchild. It is not a copyright violation to READ a book.
Cheers,
Wol
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 19:57 UTC (Fri)
by marcH (subscriber, #57642)
[Link]
Posted Mar 6, 2026 19:57 UTC (Fri) by marcH (subscriber, #57642) [Link]
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 18:46 UTC (Fri)
by valderman (subscriber, #56479)
[Link] (7 responses)
Posted Mar 6, 2026 18:46 UTC (Fri) by valderman (subscriber, #56479) [Link] (7 responses)
And no, the training exception does not make the LLM output exempt from violating the copyright of its training data. It only protects the model itself.
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 18:52 UTC (Fri)
by bluca (subscriber, #118303)
[Link] (6 responses)
Posted Mar 6, 2026 18:52 UTC (Fri) by bluca (subscriber, #118303) [Link] (6 responses)
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 20:00 UTC (Fri)
by marcH (subscriber, #57642)
[Link] (3 responses)
Posted Mar 6, 2026 20:00 UTC (Fri) by marcH (subscriber, #57642) [Link] (3 responses)
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 20:07 UTC (Fri)
by bluca (subscriber, #118303)
[Link] (2 responses)
Posted Mar 6, 2026 20:07 UTC (Fri) by bluca (subscriber, #118303) [Link] (2 responses)
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 20:19 UTC (Fri)
by marcH (subscriber, #57642)
[Link] (1 responses)
Posted Mar 6, 2026 20:19 UTC (Fri) by marcH (subscriber, #57642) [Link] (1 responses)
Or, switch to a different topic instead.
Scary misunderstanding of how LLMs work
Posted Mar 6, 2026 20:34 UTC (Fri)
by bluca (subscriber, #118303)
[Link]
Posted Mar 6, 2026 20:34 UTC (Fri) by bluca (subscriber, #118303) [Link]
Scary misunderstanding of how LLMs work
Posted Mar 10, 2026 12:49 UTC (Tue)
by diegor (subscriber, #1967)
[Link] (1 responses)
Posted Mar 10, 2026 12:49 UTC (Tue) by diegor (subscriber, #1967) [Link] (1 responses)
Also who is the author of the new version? The LLM? The mantainer? If it's the mantainer, can he really claim that is a clean room implementation? If it's not the author, he can license the code, not being the copyright holder.
I'm not convinced that "wildly different" is a proof of a clean room implementation.
Scary misunderstanding of how LLMs work
Posted Mar 10, 2026 12:54 UTC (Tue)
by bluca (subscriber, #118303)
[Link]
Posted Mar 10, 2026 12:54 UTC (Tue) by bluca (subscriber, #118303) [Link]
Is a port to another a language a derived work?
Posted Mar 6, 2026 6:30 UTC (Fri)
by rsidd (subscriber, #2582)
[Link] (2 responses)
Posted Mar 6, 2026 6:30 UTC (Fri) by rsidd (subscriber, #2582) [Link] (2 responses)
Had an LLM been employed to translate chardet to, say, Lisp, the level of similarity would be quite low, but most would agree that the new code was derived from the original.
If this argument is correct, chardet should have been under the MPL and not LGPL, since it is apparently a straighforward port of the C++ version of chardet which is under the MPL:
[dan-blanchard] The only reason I am advocating for MPL in this particular case is that a straight literal port like chardet is usually considered a modification/derivative of the original code. Given the terms of the MPL, it seems to me like chardet should have been required to be released under MPL in the first place, and that the LGPL choice was a legal oversight on Mark's part.
Elsewhere Blanchard hints that Pilgrim is not easy to reach on email; but apparently he is able to send email, and it is weird that, after so many years, he comes out of the woodwork demanding that his original licence be respected.
Is a port to another a language a derived work?
Posted Mar 6, 2026 16:15 UTC (Fri)
by Wol (subscriber, #4433)
[Link]
Posted Mar 6, 2026 16:15 UTC (Fri) by Wol (subscriber, #4433) [Link]
MPL I or II? Iirc MPL II explicitly permits relicencing to (L)GPL?
There were complaints that MPL I and (L)GPL weren't compatible so I believe that was the fix that went into MPL II.
Cheers,
Wol
Is a port to another a language a derived work?
Posted Mar 6, 2026 17:20 UTC (Fri)
by Kluge (subscriber, #2881)
[Link]
Posted Mar 6, 2026 17:20 UTC (Fri) by Kluge (subscriber, #2881) [Link]
Why is it odd if there was no indication that his license would not be respected until now?
Just rename it and move along.
Posted Mar 6, 2026 8:01 UTC (Fri)
by edomaur (subscriber, #14520)
[Link] (9 responses)
Posted Mar 6, 2026 8:01 UTC (Fri) by edomaur (subscriber, #14520) [Link] (9 responses)
Just rename it and move along.
Posted Mar 7, 2026 7:47 UTC (Sat)
by josh (subscriber, #17465)
[Link] (8 responses)
Posted Mar 7, 2026 7:47 UTC (Sat) by josh (subscriber, #17465) [Link] (8 responses)
Just rename it and move along.
Posted Mar 8, 2026 5:41 UTC (Sun)
by edomaur (subscriber, #14520)
[Link] (4 responses)
Posted Mar 8, 2026 5:41 UTC (Sun) by edomaur (subscriber, #14520) [Link] (4 responses)
Just rename it and move along.
Posted Mar 8, 2026 5:46 UTC (Sun)
by edomaur (subscriber, #14520)
[Link]
Posted Mar 8, 2026 5:46 UTC (Sun) by edomaur (subscriber, #14520) [Link]
Just rename it and move along.
Posted Mar 8, 2026 8:40 UTC (Sun)
by josh (subscriber, #17465)
[Link] (2 responses)
Posted Mar 8, 2026 8:40 UTC (Sun) by josh (subscriber, #17465) [Link] (2 responses)
Yes, it would be. You can rewrite a piece of software and still be a derivative work of the software you based it on. By way of example, if you spend a while staring at the code of some software, then set it aside, go into an airgapped room, and type in a fresh implementation that follows the same spec, the result is likely still a derivative work, as you've based it on your knowledge of the source code. (The same thing applies if you try to write a copy of Lord of the Rings from memory and publish it, even if you don't end up copying any specific sentences word-for-word.)
The reason people talk about a "clean-room reimplementation" of a piece of software is that doing so typically involves having a careful firewall between those who read the source or reverse engineer the software (who use it to write documentation that explains *behavior* but not *implementation*, where that documentation may need to be reviewed by a lawyer to make sure it includes nothing copyrightable), and those in a "clean room" who read only the documentation and use that to reimplement the software based solely on its documented behavior. There's case law upholding that doing *that* is a defense against copyright infringement and can produce an independent work that isn't derived from the original.
Contrast that with having a person who has worked on the software for years and is deeply familiar with it, taking an LLM trained on most of the Internet *including the software source code in question*, and using that LLM to reimplement the software, with access to the original source that it has been *told* to only to use for its test suite. Consider the vast gulf between those two processes. There is zero legal precedent that the extremely *non-clean-room* approach used here would avoid producing a derivative work. It is my sincere hope that there will soon be legal precedent that doing so *does* produce a derivative work.
Just rename it and move along.
Posted Mar 8, 2026 12:50 UTC (Sun)
by pizza (subscriber, #46)
[Link]
Posted Mar 8, 2026 12:50 UTC (Sun) by pizza (subscriber, #46) [Link]
...Methinks copyleft projects need to start making their test suites private.
(I'd argue the same goes for all F/OSS but Apache/BSD/MIT-licensed stuff already allows for proprietarization)
Just rename it and move along.
Posted Mar 8, 2026 12:59 UTC (Sun)
by bluca (subscriber, #118303)
[Link]
Posted Mar 8, 2026 12:59 UTC (Sun) by bluca (subscriber, #118303) [Link]
You keep saying this, and yet providing zero proof for it. It doesn't work like that. A "clean room" implementation is one way to err on the safe side, but it is not a legal requirement. There are several criterias and tests that are typically taken into consideration in various jurisdictions to decide whether something is a derivative work or isn't, and feelings and vibes and wishful thinking do not come into play.
> Contrast that with having a person who has worked on the software for years and is deeply familiar with it
Doesn't matter, never did, never will. Otherwise it would be impossible for people to do things that are routinely done, like changing jobs and going to work for competitors. You obviously cannot take IP with you, and obviously cannot reimplement the same thing verbatim, but there is absolutely no requirement not to be "deeply familiar" with copyrighted work in order to be able to work on similar but different projects.
That's one of the reasons why large companies, especially in the US, try to impose draconian non-compete private agreements on their employees - because the law is not on their side on this. And thank fuck for that, because otherwise it would provide _yet another_ way for big players in capitalism to screw over ordinary people.
Just rename it and move along.
Posted Mar 8, 2026 9:08 UTC (Sun)
by Wol (subscriber, #4433)
[Link] (2 responses)
Posted Mar 8, 2026 9:08 UTC (Sun) by Wol (subscriber, #4433) [Link] (2 responses)
As has been pointed out repeatedly, *what* other contributors? The empty set?
I wouldn't have done it this way, I would have done it the way they converted the GPL'd Star Office into the MPL'd Libre Office, and the tragedy is it sounds like that would have been an easy thing to do! But what's done is done.
Cheers,
Wol
Just rename it and move along.
Posted Mar 8, 2026 9:24 UTC (Sun)
by josh (subscriber, #17465)
[Link] (1 responses)
Posted Mar 8, 2026 9:24 UTC (Sun) by josh (subscriber, #17465) [Link] (1 responses)
`git shortlog -es` shows several dozen, and that's from the git era, not counting those who contributed prior to the first git commit (which includes the original author Mark Pilgrim).
> But what's done is done.
That's not how license violations work, no.
Just rename it and move along.
Posted Mar 8, 2026 9:59 UTC (Sun)
by Wol (subscriber, #4433)
[Link]
Posted Mar 8, 2026 9:59 UTC (Sun) by Wol (subscriber, #4433) [Link]
Ah! Okay. But you do know, you are the first person in this long thread to actually come up with a list?
How many of these are recent, and how many predate the current maintainer taking over? Sounds like he really should have gone down the LibreOffice type route ...
Cheers,
Wol
I could suggest
Posted Mar 6, 2026 15:35 UTC (Fri)
by frankie (subscriber, #13593)
[Link]
Posted Mar 6, 2026 15:35 UTC (Fri) by frankie (subscriber, #13593) [Link]
Has the horse left the barn?
Posted Mar 12, 2026 12:52 UTC (Thu)
by karim (subscriber, #114)
[Link]
Posted Mar 12, 2026 12:52 UTC (Thu) by karim (subscriber, #114) [Link]
Companies are already shipping code generated by LLMs -- whether proprietary or not. OpenClaw was vibe coded and was sold 1B$. Need we say more? Any PM with half a brain is likely planning their way towards this direction, licensing be damned.
Don't get me wrong. I'm not promoting copyright infringement, etc. But if the cost of generating a given functionality is essentially zero then what "right" is being protected?
That said, would I lay my life down on code generated by Claude? Nope :)
Pulling this off requires some special circumstances ....
Posted Mar 21, 2026 12:46 UTC (Sat)
by ras (subscriber, #33059)
[Link] (1 responses)
Posted Mar 21, 2026 12:46 UTC (Sat) by ras (subscriber, #33059) [Link] (1 responses)
implementation.
- You need a definition of the API. It's in the source that implements the API of course, but you can't look at that and claim it is clean-room. Perhaps you are allowed to look at the documentation? chardet seems to be small and have good documentation. Failing that, you can give it lots of examples of other modules calling the API.
- You need a very good test suite. Without this, you get what would be called crap (if you are being kind). With it, you usually get something that passes 100% of the tests. There is a surprisingly sharply defined moment when LLMs "became useful". Before that point when an LLM tried to fix some code to make failing unit tests pass, it usually introduced more bug than it fixed so it spiralled into a complete mess. In June 2024, Claude 3.5 Sonnet usually converged on code that made all tests pass. Once you have convergence the model just needs to bash on it long enough.
- The lower-end APIs (the ones the program uses to get stuff done) also have to be well defined. chardet looks to be pure python which is not only pretty well defined, it's also something the existing LLM models know very well.
- It can't be too big. The current models advertise 200k tokens, but that isn't near as large as it sounds. Some models advertise 1M tokens, but those models tend to forget random bits of the 1M tokens so they become less reliable, and reliability isn't stellar to begin with. Everything the LLM reads, outputs and thinks to itself in its chain of thought gets appended to that window. If source gets read multiple times, which will happen when its searching for a bug, then it gets appended multiple times. If it has to scan unit test output, or raw data feed to unit tests, that gets appended too. Running out of context is handled by a mechanism that is both less reliable than the OOM killer and more devastating as it gets rid of 90% of what it knows. Tell me again - how do I compile this, I've forgotten.
chardet met all those conditions. I'm guessing it was fed the documentation, told to make the unit tests pass, and it delivered something that passed the unit tests. I am not qualified to say if that means it complies with the LGPL. Others have pointed out in chardet's issues that it can't be MIT licensed as LLM output can't be copyrighted, but I suspect that becomes mushy if Dan hand-modified the source. Besides copyright law tends to change to accommodate new technologies, and we haven't seen the law makers hammer out a change to accommodate LLMs yet.
Any software that meets those conditions is reproducible via an LLM. Not effortlessly, as passing unit tests does not mean it works. A human has to review the code (or at least the result), and go through many rounds of pointing out the errors and getting the model to fix them. But if you do that, human + model in my experience turns out something better than a human could alone. If you already have documented API and unit tests (they are a fair chunk of the work), it's going to happen faster than the original was written.
One final twist - it is possible to get one LLM to read the original work, and produce an English specification. Then get a 2nd LLM to produce code that does the same as the original. That is what I was taught was the "clean-room" process you could use to work around copyright, albeit done by two teams of humans rather than LLMs. I dunno if doing with an LLM changes that calculation. I've done it as a lark to see how well it works. It worked, very, very well. You don't need the documentation or the unit tests in that case, but to get a useful result the LLM has to be able to run the old and new versions of the program and compare outputs. If this process is a way around copyright, any open source projects that meet the last two conditions is reproducible with an LLM.
Pulling this off requires some special circumstances ....
Posted Mar 21, 2026 14:09 UTC (Sat)
by karath (subscriber, #19025)
[Link]
Posted Mar 21, 2026 14:09 UTC (Sat) by karath (subscriber, #19025) [Link]
If my understanding is correct, the case rulings in the US regarding copyright of the output of an LLM are very, very narrow. The relevant ruling does not state that the output of an LLM cannot be copyrighted. It states only that an LLM is not (currently) eligible to be be the copyright holder of the output.
IIRC, in the case, the plaintiff had disclaimed copyright in the prompt, and was using an LLM model where the creator had disclaimed copyright in the output of the LLM. The plaintiff claimed that the LLM was creative enough that the output could be copyrighted by the LLM. The judge declined to rule on the the creative enough part, and ruled only on whether the LLM was eligble to claim copyright.
I'm not a lawyer, I don't live in the USA, and I don't obsessively follow the relevant cases. So it it is entirely possible that there are later rulings that I have not yet heard of.
Protecting the Four Freedoms
Posted Mar 21, 2026 18:52 UTC (Sat)
by jondo (guest, #69852)
[Link] (1 responses)
Posted Mar 21, 2026 18:52 UTC (Sat) by jondo (guest, #69852) [Link] (1 responses)
Protecting the Four Freedoms
Posted Mar 21, 2026 20:04 UTC (Sat)
by Wol (subscriber, #4433)
[Link]
Posted Mar 21, 2026 20:04 UTC (Sat) by Wol (subscriber, #4433) [Link]
Probably not. Certainly legally in many jurisdictions, and indeed morally in the view of many, such a restriction would be seen as making the work non-free, because you are explicitly forbidding the freedom to study and learn from the source.
(Never mind that an LLM is not a person, EU and others explicitly equate training an LLM to teaching a person. And that does make a lot of sense. I'm personally extremely happy with that - treating an LLM exactly the same as a person ie copyright material can go *IN* unhindered, what comes out may or may not be a copyright violation.)
Cheers,
Wol
SFC is analyzing chardet LGPL situation
Posted Mar 27, 2026 21:40 UTC (Fri)
by bkuhn (subscriber, #58642)
[Link]
Posted Mar 27, 2026 21:40 UTC (Fri) by bkuhn (subscriber, #58642) [Link]
Take a look at my comment on chardet's issue tracker.
TL;DR: I'm leading an effort at Software Freedom Conservancy (SFC) to analyze this situation. The results will be published. It will take a long time — for good reason. Meanwhile, anyone using chardet commercially should call their lawyer.
— bkühn, Policy Fellow & Hacker-in-Residence at SFC