Are neural nets (the weights file) copyrightable?

[syzygy]

The “Corresponding Source” for a work in object code form means all the source code needed to generate, install, and (for an executable work) run the object code and to modify the work, including scripts to control those activities. However, it does not include the work's System Libraries, or general-purpose tools or generally available free programs which are used unmodified in performing those activities but which are not part of the work. For example, Corresponding Source includes interface definition files associated with source files for the work, and the source code for shared libraries and dynamically linked subprograms that the work is specifically designed to require, such as by intimate data communication or control flow between those subprograms and other parts of the work.

So distributing a Stockfish executable compiled for Windows almost requires you to provide GPL'd sources for the Microsoft Windows dlls it links to, except that those are "System Libraries". So it's OK to link to a dll as long as it one shipped by Microsoft... Weird.

[gonzochess75]

...
So indeed there is an attempt to force you to release the source code of a dynamically linked (shared) library when distributing a program. Interestingly, this still does not seem to cover the case of a program linking to a GPL'd library. The program's copyright is still untouched by the dynamically linked library.
...

It looks like we're getting closer and closer to agreeing actually! However, now that we see that "derived work" was specifically taken out go back to your earlier analysis where you said the language about mere "aggregation" was superfluous given that it seemed to be trying to do the work of "derived work." Now, do you change your mind and agree that section is not superfluous?

[gonzochess75]

So distributing a Stockfish executable compiled for Windows would require you to provide GPL'd sources for the Microsoft Windows dlls it links to. That seems tricky...

That would be tricky! Hence the system library exception

[cpeters]

MySQL is a GPL program.
Your medical records (if stored using MySQL) are therefore modified code.
This code therefore belongs to anyone who asks.

You could link Linus' kernel against the beelzebub (running at home/a server in a hospital) or store him (beelzebub) in a MySQL-database on Hartmann's ext4 filesystem and no mortal (even not the demon-commanding Richard: https://marc.info/?l=openbsd-misc&m=119775767819426)
could force you to release him.


It's all about distribution.

[syzygy]

...
So indeed there is an attempt to force you to release the source code of a dynamically linked (shared) library when distributing a program. Interestingly, this still does not seem to cover the case of a program linking to a GPL'd library. The program's copyright is still untouched by the dynamically linked library.
...

It looks like we're getting closer and closer to agreeing actually! However, now that we see that "derived work" was specifically taken out go back to your earlier analysis where you said the language about mere "aggregation" was superfluous given that it seemed to be trying to do the work of "derived work." Now, do you change your mind and agree that section is not superfluous?

No, there is no need to change my position.

What is your position exactly if we assume that the FF2 NNUE net would be copyrighted (which we agree it is not)?

My position:
- when distributing an SF executable that includes the FF2 NNUE net, the whole executable must be released under the GPLv3.
- when distributing an SF executable that is modified to load the FF2 NNUE net (and many other nets with the same format), only the executable must be released under the GPLv3. It is fine to include the FF2 NNUE net in the same zip file.

[syzygy]

So distributing a Stockfish executable compiled for Windows would require you to provide GPL'd sources for the Microsoft Windows dlls it links to. That seems tricky...

That would be tricky! Hence the system library exception

Yes, I already corrected myself.

But this exception is quite arbitrary. As long as I create a machine that includes the necessary code that I choose to refer to as "system libraries", I can distribute modified GPL'd programs that link to that code.

[gonzochess75]

What is your position exactly if we assume that the FF2 NNUE net would be copyrighted (which we agree it is not)?

My position:
- when distributing an SF executable that includes the FF2 NNUE net, the whole executable must be released under the GPLv3.
- when distributing an SF executable that is modified to load the FF2 NNUE net (and many other nets with the same format), only the executable must be released under the GPLv3. It is fine to include the FF2 NNUE net in the same zip file.

Very close. The only difference is the last point. I believe:

- when distributing an SF executable that is modified to load the FF2 NNUE net (and many other nets with the same format), both the executable and the net file specifically intended to act as an extension of and be combined with SF must be released under the GPLv3. It is fine to include the FF2 NNUE net in the same zip file provided it is also licensed in a GPL-compatible manner.

[Dann Corbit]

Of course it does.
You said that data is code.

This is a well known theoretical computer science question. A lot of people said that Lisp proved that data is code. However, they don't know their history. The lambda calculus proves it. Of course Turing aficionados would say that Turing proved it.

Therefore, anyone who asks for the data must be given that data.

You are misunderstanding me.

Whether data is copyrightable or not is also irrelevant to the embedding argument.
If I embed a copyright movie into a GPL program does that void the copyright?
A book can be written as data and it may be copyright or it may be public domain.
I can have a database that stores moves, songs and books.
I can read these items into memory.
I can even make the program dependent on them.

I think you want the data to be code when you want it to be code and you want the data to be data when you do not want it to be.
But data is either code or it isn't.

I think the meaning of the license is clear. The FSF wants to guarantee availability of the program's code. Not the program's data.
I think that you do not like what Albert did, and so you are trying to twist the meaning of what they wrote into "source code + data"

Not only do I think that is not what was intended, but if (indeed) that is what was intended, nobody would use GPL software.

[gonzochess75]

So distributing a Stockfish executable compiled for Windows would require you to provide GPL'd sources for the Microsoft Windows dlls it links to. That seems tricky...

That would be tricky! Hence the system library exception

Yes, I already corrected myself.

But this exception is quite arbitrary. As long as I create a machine that includes the necessary code that I choose to refer to as "system libraries", I can distribute modified GPL'd programs that link to that code.

And then the plaintiffs call in an expert to ask if your libFoo is really a system library and what does "system library" really mean and the reasonably intelligent judge laughs at you at best and cites you with contempt at worst

[syzygy]

- when distributing an SF executable that is modified to load the FF2 NNUE net (and many other nets with the same format), both the executable and the net file specifically intended to act as an extension of and be combined with SF must be released under the GPLv3. It is fine to include the FF2 NNUE net in the same zip file provided it is also licensed in a GPL-compatible manner.

And the FF2 NNUE net and not any other NNUE net because we know that that was the plan all along, I suppose?

Obviously it would be fine for anyone else to distribute a non-GPL NNUE net that is compatible with SF13, or with FF2-SF for that matter. (Still assuming that NNUE nets are copyrightable.)