The post isn't really about the ship of theseus case. Most comments are about the cases discussed in the post, but the SoT case is worth mention as that case seems pretty clear:*
1. If the code started with was *GPL'ed, each edit resulted in a GPL'ed piece of code, thus the final edit was a change to a GPL'd piece of code and the result is GPLed.
This is basically no different from doing the same with a proprietary library you'd licensed from a vendor.
Because you looked at the code to begin with you couldn't make the "clean room implementation" argument even if every line were different.
2. MIT or related licenses: just depends on whether what remains is "substantial".
* modulo the effect of interested lawyers, of course.
> If the code started with was *GPL'ed, each edit resulted in a GPL'ed piece of code, thus the final edit was a change to a GPL'd piece of code and the result is GPLed.
The second clause is not correct. Each edit does not result in GPL'd code. GPL only applies to the code you start with. If you want to redistribute the code with the modifications, then the GPL requires that you make your changes available under the GPL. If you cannot license the work in this way, then you may not redistribute it; the licensing under the GPL is not automatic, and it may not even be possible (e.g., if the modifications are copied from code with an incompatible licence). But this does not apply if none of the code you want to distribute is covered under the GPL. So if you remove all of the original code, and your final work is independent of the original, then the GPL would not apply.
The license still applies even when the redistribution clause is not being exercised (else any edit on your machine would strip the license!). This is an unremarkable part of the language and of contract law.
If you mix in other code that you don’t have the right to distribute into the GPL code this has the same kind of impact as doing that with non-GPL code.
Yes, of course the licence applies even without redistribution. Without a copyright licence, the user does not have the right to use, modify or distribute the code.
But what the licence applies to is the original work. If you replace half of the original work with your own work, then the GPL applies to the remaining half of the original work, while you would own the copyright to your own half of it. GPL requires you to license modifications under the GPL only if you redistribute them. But if what you are distributing is modified to the point of being completely independent of the original, with none of the original code or any dependency on it remaining, then the GPL is no longer relevant to it.
According to that logic, you may just rewrite any GPL code line by line and then apply any license you want to distribute the result. It doesn't sound right.
Copyright also applies to the structure of the code. So rewriting it line by line would not be enough to remove all of the GPL-licensed elements. You would also need to restructure and reorganise it.
I'm unaware of it ever having been tested and upheld in court in a SoT situation. And while it's easy to argue that it should be upheld because that's what the license says, it's also just as easy to argue that it shouldn't. Because it's easy to argue that a chain of derivation loses all meaning under copyright law once there are no recognizable elements of the original work. Quite simply, making a claim of copyright infringement requires elements of the original being used. If no original elements exist, then no license can bind, no matter what path it took to get there. Just because you put something in a license doesn't mean it's enforceable.
Yes, I think it's pretty clear (though yes, a lawyer can bring anything up). The basis is the propriety code case I mentioned which has been well litigated over the decades, its clear definition of "derived work" (in particular for the ToS case, not calling into it) and the implication of derivation vs clean room (discussed by others in this thread).
When people try to attack the GPL in court it's typically over it being a contract of adhesion, enforceability, or applicability of calling into it (none of which are an issue in this discussion). The aspects I mention above are too well trodden.
(I'm talking about the USA only, of course, and I assume you, the author of the post are too).
I think you always retain the right to change the license of your own code. If something is 100% your own code, you can choose whatever license you like.
Yes, but you can't take back old licenses from anyone who got it before you changed it (unless they specifically say you can, which no FOSS license does, by definition).
But anyway, I think this is about code that's a derivative of someone else's and so not entirely your own.
No, the idea here is that after all the original GPL licensed code (not own code) was step by step fully replaced by by own code (Ship of Theseus). Then you can change the license, because it is 100% your own.
There is plenty of legal case law on this ranging from BSD to various proprietary licenses. You would not win on this. This is why Compaq had to do a clean room implementation of the BIOS back in the 80s -- the law of doing it otherwise was already clear back then.
1. If the code started with was *GPL'ed, each edit resulted in a GPL'ed piece of code, thus the final edit was a change to a GPL'd piece of code and the result is GPLed.
This is basically no different from doing the same with a proprietary library you'd licensed from a vendor.
Because you looked at the code to begin with you couldn't make the "clean room implementation" argument even if every line were different.
2. MIT or related licenses: just depends on whether what remains is "substantial".
* modulo the effect of interested lawyers, of course.