I would disagree that the situation is the same. The way to remedy the BSD situation is simple attribution. The way to remedy the GPL situation is release of the entire source code.
>> Worse, the source code of all versions from _before_ that checkin automatically become GPL-licensed (because they contain code that, transitively, becomes GPL because it later was linked with GPL-licensed software)
>This is simply false.
Is it? If you are forced to release all your source under GPL any new release of that software will contain software released under GPL, even if the GPL specific changes are rolled back.
My interpretation is this: If you have two functions, A() and B() where A() is released under GPL. If you then release software where B() is defined as :
B():
...
A()
You now have a product which is a work derived from GPL code so you must release the source code for B(). Now you release B() under GPL. So you go back to the commit right before including A() and make a new branch with a new commit:
B():
...
C()
BUT! B() is now source code released under GPL so any changes to it must also be released under GPL.
Years of conversation about this kind of stuff, I don't remember specifically. Looking it up wikipedia says $150,000 is the upper limit for wilful infringement, accidental infringement is lower, but that doesn't affect the argument.
> The way to remedy the BSD situation is simple attribution. The way to remedy the GPL situation is release of the entire source code.
The damages are statutory plus actual damages (almost certainly $0 for someone who doesn't make money off their software directly), not "what you would have done". You have the same options in either case: reach agreement with the person whose copyright you infringed, or go to court and pay the damages. So your worst-case liability is the same either way.
> BUT! B() is now source code released under GPL so any changes to it must also be released under GPL.
As you acknowledged in your reply, you still hold all the copyrights on B and can still release B under any terms you please.
(Aside: it doesn’t matter whether B calls A; linking with it is sufficient. In your example, B’s source became GPL licensed because it was linked with a publicly released A)
My logic was incorrect. I assumed that the exact same source code for now GPL licensed B from earlier revision revisions would have to be GPL licensed, too (and because of it, anything linked with that, linked with code linked with that, etc.).
However, licensing (unlike copyright) doesn’t work that way. Copyright owners can even release the exact same set of source code 7nder multiple licenses.
My conclusion still stands, though. Companies are right to worry about using GPL-licensed software.
I would disagree that the situation is the same. The way to remedy the BSD situation is simple attribution. The way to remedy the GPL situation is release of the entire source code.
>> Worse, the source code of all versions from _before_ that checkin automatically become GPL-licensed (because they contain code that, transitively, becomes GPL because it later was linked with GPL-licensed software) >This is simply false.
Is it? If you are forced to release all your source under GPL any new release of that software will contain software released under GPL, even if the GPL specific changes are rolled back.
My interpretation is this: If you have two functions, A() and B() where A() is released under GPL. If you then release software where B() is defined as :
B(): ... A()
You now have a product which is a work derived from GPL code so you must release the source code for B(). Now you release B() under GPL. So you go back to the commit right before including A() and make a new branch with a new commit:
B(): ... C()
BUT! B() is now source code released under GPL so any changes to it must also be released under GPL.
How have I misunderstood?