The current upstream maintainer of python-chardet recently used the Claude Code LLM to completely rewrite the library for version 7.0.0[1]. The source code now has no obvious direct relationship with previous versions.
The maintainer used the LLM-based rewrite as justification to unilaterally relicense the project from LGPL-2.1-or-later to MIT. This has prompted a debate about whether or not the maintainer truly has the right to do so, given their deep familiarity with the LGPL code and other considerations. The original author of the library feels that the relicensing was improper[2].
There is no great rush to make a decision about how to handle this particular situation. I can keep python-chardet at 6.0.0 in Fedora for quite a while. However, we in Fedora may end up having to make a decision sooner or later on what to do about cases like this. Even if the question can be somehow sidestepped for chardet, we can expect to be faced with an increasing number of similar situations.
On Thu, Mar 5, 2026 at 5:26 AM Ben Beasley via legal legal@lists.fedoraproject.org wrote:
The current upstream maintainer of python-chardet recently used the Claude Code LLM to completely rewrite the library for version 7.0.0[1]. The source code now has no obvious direct relationship with previous versions.
The maintainer used the LLM-based rewrite as justification to unilaterally relicense the project from LGPL-2.1-or-later to MIT. This has prompted a debate about whether or not the maintainer truly has the right to do so, given their deep familiarity with the LGPL code and other considerations. The original author of the library feels that the relicensing was improper[2].
There is no great rush to make a decision about how to handle this particular situation. I can keep python-chardet at 6.0.0 in Fedora for quite a while. However, we in Fedora may end up having to make a decision sooner or later on what to do about cases like this. Even if the question can be somehow sidestepped for chardet, we can expect to be faced with an increasing number of similar situations.
I've been following this. Typically Fedora does not second-guess upstream relicensing acts (or forks that essentially involve project relicensing), but there have been some notable exceptions.
Richard
On Thu, Mar 5, 2026 at 10:45 AM Richard Fontana rfontana@redhat.com wrote:
On Thu, Mar 5, 2026 at 5:26 AM Ben Beasley via legal legal@lists.fedoraproject.org wrote:
The current upstream maintainer of python-chardet recently used the Claude Code LLM to completely rewrite the library for version 7.0.0[1]. The source code now has no obvious direct relationship with previous versions.
The maintainer used the LLM-based rewrite as justification to unilaterally relicense the project from LGPL-2.1-or-later to MIT. This has prompted a debate about whether or not the maintainer truly has the right to do so, given their deep familiarity with the LGPL code and other considerations. The original author of the library feels that the relicensing was improper[2].
There is no great rush to make a decision about how to handle this particular situation. I can keep python-chardet at 6.0.0 in Fedora for quite a while. However, we in Fedora may end up having to make a decision sooner or later on what to do about cases like this. Even if the question can be somehow sidestepped for chardet, we can expect to be faced with an increasing number of similar situations.
I've been following this. Typically Fedora does not second-guess upstream relicensing acts (or forks that essentially involve project relicensing), but there have been some notable exceptions.
That parenthetical about forks was confusing. I think I was thinking of some of the *exceptions* to the general rule. We probably *would* see an upstream fork that was coupled with a license change as a non-green flag.
Anyway, Red Hat is facing this same issue in at least one non-Fedora-derived context. I think (seeing this from Red Hat's point of view) we probably would want to adopt a consistent approach, whatever that ends up being.
On Fri, Mar 6, 2026 at 9:30 AM Richard Fontana rfontana@redhat.com wrote:
On Thu, Mar 5, 2026 at 10:45 AM Richard Fontana rfontana@redhat.com wrote:
On Thu, Mar 5, 2026 at 5:26 AM Ben Beasley via legal legal@lists.fedoraproject.org wrote:
The current upstream maintainer of python-chardet recently used the Claude Code LLM to completely rewrite the library for version 7.0.0[1]. The source code now has no obvious direct relationship with previous versions.
The maintainer used the LLM-based rewrite as justification to unilaterally relicense the project from LGPL-2.1-or-later to MIT. This has prompted a debate about whether or not the maintainer truly has the right to do so, given their deep familiarity with the LGPL code and other considerations. The original author of the library feels that the relicensing was improper[2].
There is no great rush to make a decision about how to handle this particular situation. I can keep python-chardet at 6.0.0 in Fedora for quite a while. However, we in Fedora may end up having to make a decision sooner or later on what to do about cases like this. Even if the question can be somehow sidestepped for chardet, we can expect to be faced with an increasing number of similar situations.
I've been following this. Typically Fedora does not second-guess upstream relicensing acts (or forks that essentially involve project relicensing), but there have been some notable exceptions.
That parenthetical about forks was confusing. I think I was thinking of some of the *exceptions* to the general rule. We probably *would* see an upstream fork that was coupled with a license change as a non-green flag.
Having looked at this a bit more closely, I haven't seen anything persuasive that the non-use of LGPL on chardet 7.0.0 was improper in a legal sense, i.e. I'm not persuaded it should be considered a copy or derivative work of LGPL-licensed versions of chardet. The possibly questionable thing was the maintainer's use of the MIT license on what is apparently more or less entirely AI-generated code (assuming such code is not copyrightable). [1] However, I have recently thought that projects using AI in a similar manner ought to have an easy way to signal a preference or intention that subsequent human creative contributions be placed under a particular FOSS license.
On Thu, 2026-03-19 at 19:12 -0400, Richard Fontana via legal wrote:
On Fri, Mar 6, 2026 at 9:30 AM Richard Fontana rfontana@redhat.com wrote:
On Thu, Mar 5, 2026 at 10:45 AM Richard Fontana rfontana@redhat.com wrote:
On Thu, Mar 5, 2026 at 5:26 AM Ben Beasley via legal legal@lists.fedoraproject.org wrote:
The current upstream maintainer of python-chardet recently used the Claude Code LLM to completely rewrite the library for version 7.0.0[1]. The source code now has no obvious direct relationship with previous versions.
The maintainer used the LLM-based rewrite as justification to unilaterally relicense the project from LGPL-2.1-or-later to MIT. This has prompted a debate about whether or not the maintainer truly has the right to do so, given their deep familiarity with the LGPL code and other considerations. The original author of the library feels that the relicensing was improper[2].
There is no great rush to make a decision about how to handle this particular situation. I can keep python-chardet at 6.0.0 in Fedora for quite a while. However, we in Fedora may end up having to make a decision sooner or later on what to do about cases like this. Even if the question can be somehow sidestepped for chardet, we can expect to be faced with an increasing number of similar situations.
I've been following this. Typically Fedora does not second-guess upstream relicensing acts (or forks that essentially involve project relicensing), but there have been some notable exceptions.
That parenthetical about forks was confusing. I think I was thinking of some of the *exceptions* to the general rule. We probably *would* see an upstream fork that was coupled with a license change as a non-green flag.
Having looked at this a bit more closely, I haven't seen anything persuasive that the non-use of LGPL on chardet 7.0.0 was improper in a legal sense, i.e. I'm not persuaded it should be considered a copy or derivative work of LGPL-licensed versions of chardet. The possibly questionable thing was the maintainer's use of the MIT license on what is apparently more or less entirely AI-generated code (assuming such code is not copyrightable). [1] However, I have recently thought that projects using AI in a similar manner ought to have an easy way to signal a preference or intention that subsequent human creative contributions be placed under a particular FOSS license.
If AI-generated output is later deemed to not be copyrightable, what happens though? Would they be considered public domain in the US (which still does not help in the EU), or would it be something else?
Looks like faced with a similar issue, OpenBSD's position is that given the doubt about what license will eventually apply, such code should not be accepted for redistribution
https://lwn.net/Articles/1064541/
Best regards,
On Thu, Mar 19, 2026 at 7:12 PM Richard Fontana rfontana@redhat.com wrote:
On Fri, Mar 6, 2026 at 9:30 AM Richard Fontana rfontana@redhat.com wrote:
On Thu, Mar 5, 2026 at 10:45 AM Richard Fontana rfontana@redhat.com wrote:
On Thu, Mar 5, 2026 at 5:26 AM Ben Beasley via legal legal@lists.fedoraproject.org wrote:
The current upstream maintainer of python-chardet recently used the Claude Code LLM to completely rewrite the library for version 7.0.0[1]. The source code now has no obvious direct relationship with previous versions.
The maintainer used the LLM-based rewrite as justification to unilaterally relicense the project from LGPL-2.1-or-later to MIT. This has prompted a debate about whether or not the maintainer truly has the right to do so, given their deep familiarity with the LGPL code and other considerations. The original author of the library feels that the relicensing was improper[2].
There is no great rush to make a decision about how to handle this particular situation. I can keep python-chardet at 6.0.0 in Fedora for quite a while. However, we in Fedora may end up having to make a decision sooner or later on what to do about cases like this. Even if the question can be somehow sidestepped for chardet, we can expect to be faced with an increasing number of similar situations.
I've been following this. Typically Fedora does not second-guess upstream relicensing acts (or forks that essentially involve project relicensing), but there have been some notable exceptions.
That parenthetical about forks was confusing. I think I was thinking of some of the *exceptions* to the general rule. We probably *would* see an upstream fork that was coupled with a license change as a non-green flag.
Having looked at this a bit more closely, I haven't seen anything persuasive that the non-use of LGPL on chardet 7.0.0 was improper in a legal sense, i.e. I'm not persuaded it should be considered a copy or derivative work of LGPL-licensed versions of chardet. The possibly questionable thing was the maintainer's use of the MIT license on what is apparently more or less entirely AI-generated code (assuming such code is not copyrightable). [1] However, I have recently thought that projects using AI in a similar manner ought to have an easy way to signal a preference or intention that subsequent human creative contributions be placed under a particular FOSS license.
The chardet maintainer responded to the above issue by relicensing from MIT to 0BSD.
Regarding the question posed in this thread, I want to point to the clarification I made here: https://github.com/chardet/chardet/issues/355 and I'd note also the comment made by Bradley Kuhn of the Software Freedom Conservancy https://github.com/chardet/chardet/issues/355#issuecomment-4145369025
Richard
-
Ben Beasley -
Michel Lind -
Richard Fontana