[prutschman]

> Which I believe is OK with everyone

What's the FSF's position on this, I wonder. The whole premise of the GPL as opposed to the LGPL is that even dynamic linking creates a derived work.

If someone creates a completely drop-in replacement for readline (I don't know if libedit and kin are drop-in or not) and releases it under a BSD license, there's no way to distinguish my app linking against one vs the other, I'm simply making use of the API in the abstract.

This isn't exactly the same as the Java case, but the copyright status of APIs per-se sure seems to have implication for the GPL's teeth.

[belorn]

Lets imagine that we create a third way to link software. There is static linking, dynamic linking, and now we got this imagined third thing called adaptive linking. In this made up linker, it magically combines the different parts and uses neural networks (ie, more magic) to figure out how to make the two parts communicate without a API.

Okey, in this fantasy world, do we still have the concept of derivative work? I would say Yes. Would thus GPL still apply? Yes. Is there an API? No.

GPL is not based on the technology of linking. As a software license, it used the concept of derivative work, distribution and other concept in copyright law to define the scope. LGPL in contrast do specify linking technology, but that is only an additional permission which would need to be modify if it were to function with new linking technologies.

If you created a drop-in replacement for readline, the argument that the complete work is readline + some code get weaker. However, if the police finds some emails on your drive that states your intention to combine readline + some code, and then go to copy readline by reimplemented the software in a identical (copy'ish) way, then you might very well be infringing on the right of the readline authors. It depend what a judge/jury think your intentions was.

[prutschman]

It's an interesting thought experiment.

I agree that GPL doesn't get it's "teeth" unless linking creates a derivative work. The FSF (and Linus) for that matter have Opinions about what does,though, so I'm curious how and whether this ruling changes those.

I don't know whether to be thankful or not that there's still a circuit split in the US over what constitutes a derivative work in the software realm.

[esarbe]

RMS and the FSF seem to be okay with copying header files.

see http://lkml.indiana.edu/hypermail/linux/kernel/0301.1/0362.h...