The problem is that there’s ambiguity if you add new code to those files because the file header says it is MIT licensed but the project licence says it is GPL3. It’s contradictory.
I believe they could have resolved the issue by deleting the file headers and including a copy of the MIT license in LICENSE with a note saying something like “Code at this commit is fully licensed under the MIT license which is reproduced here; subsequent code is licensed only under the GPL3. Both licenses must be respected.”
But they haven’t done that, and it doesn’t seem like they’ve even thought about it.
I’d opine that the MIT license has no requirements about avoiding ambiguity. That’s kind of its thing, it’s as unobtrusive as possible and minimizes the amount of having to think about license compliance minefields.
And incidentally they have done quite a bit to avoid ambiguity, in readme.md:
License
Forgejo is distributed under the terms of the GPL version 3.0 or any later version.
The agreement for this license was documented in June 2023 and implemented during the development of Forgejo v9.0. All Forgejo versions before v9.0 are distributed under the MIT license.
though they also distribute binary-only copies… the main website even recommends downloading the binary. not even a tarball, just the plain binary. which even in old versions don’t contain an MIT license at all. Even a hexdump of the binary does not contain any representation of the MIT text. I think that’s actually an MIT license violation?
I think that’s probably fine actually since the place they are distributing the binary from (Codeberg releases) has a copy of the licence easily available.
the MIT license has no requirements about avoiding ambiguity
Err yeah of course not. The issue with creating ambiguous or conflicting legal requirements is that they might not get applied how you’d like if it went to court. For example Amazon might fork Forgejo and keep it closed source, saying “we copied the individual source files and those are MIT licensed” and they might win. The license text doesn’t have to say anything about that for it to be true.
The problem is that there’s ambiguity if you add new code to those files because the file header says it is MIT licensed but the project licence says it is GPL3. It’s contradictory.
I believe they could have resolved the issue by deleting the file headers and including a copy of the MIT license in
LICENSEwith a note saying something like “Code at this commit is fully licensed under the MIT license which is reproduced here; subsequent code is licensed only under the GPL3. Both licenses must be respected.”But they haven’t done that, and it doesn’t seem like they’ve even thought about it.
I’d opine that the MIT license has no requirements about avoiding ambiguity. That’s kind of its thing, it’s as unobtrusive as possible and minimizes the amount of having to think about license compliance minefields.
And incidentally they have done quite a bit to avoid ambiguity, in readme.md:
though they also distribute binary-only copies… the main website even recommends downloading the binary. not even a tarball, just the plain binary. which even in old versions don’t contain an MIT license at all. Even a hexdump of the binary does not contain any representation of the MIT text. I think that’s actually an MIT license violation?
I think that’s probably fine actually since the place they are distributing the binary from (Codeberg releases) has a copy of the licence easily available.
Err yeah of course not. The issue with creating ambiguous or conflicting legal requirements is that they might not get applied how you’d like if it went to court. For example Amazon might fork Forgejo and keep it closed source, saying “we copied the individual source files and those are MIT licensed” and they might win. The license text doesn’t have to say anything about that for it to be true.