[slgeorge]

  It certainly is positive for the ability to create new free software. It is negative for the ability to control other people's actions through copyleft licenses, since the same ability to create new free software also allows creating new permissive licensed or commercial software rather than adhering to the copyleft terms.

Agreed. As a point of practicality I can't think of anyone whose created 'commercial' open source who depends on copyright protection 'of the API' for their commercial elements though, can you?

It feels like the OSS ecosystem has always concluded that creating drop-in replacements using API's are 'fair use'.

[schoen]

Hurst's argument would be somewhat convincing if there were people trying to apply the GPL to prohibit, say, the Windows Subsystem for Linux (or other efforts to make Linux binaries run natively under Windows). I don't think it would be easy to find a consensus that that is appropriate or supported by the law, and indeed, there's been tons of concern that proprietary software copyright holders will try to sue people for writing new interoperable implementations that use the same protocol or API.

Maybe there are some narrower cases involving proprietary Linux kernel modules (which many people have suggested could be improper derived works that constitute copyright infringement of the kernel); there this result might make the infringement argument more difficult.

But as people here have pointed out, almost nobody ever attempts to limit implementation or use of APIs as a "work based on the program" for GPL purposes, almost nobody assumes that would be successful, and almost all uses of the GPL are based on limiting the literal copying of code from one program into another.