I'm but a measly c++ guy, all the lisp examples make no sense to me, so obviously I fail too see any kind of fundamental problem here with copyleft, p2p, and the "code = data" thingie.
Yes, if the photograph of a birthmark on your butt that you embedded in the software is a central and irreplaceable part of an algorithm that objectively changes/enhances the software as observed by the rest of the p2p network, then I'm afraid it is only ethical to share that piece of personal information. However otherwise, if it is not so essential, in the version that you share you can replace it with any other image. Nobody sensible will have a problem with that, and nobody malicious will be able to prove that you did that, without braking the law or a Kafkaesque court siding with their demands. That said I'm not a lawyer and I would not claim that the wording in the license is perfect in this regard.
Definitely a practical consideration if such a replacement is difficult in your language or paradigm of choice, but not a fundamental issue I think.
Taking this to extreme you arrive at neural networks, as in "I've trained a neural network that filters emails for me but it also spews out my home address if you ask it nicely". I think it is obvious that neural networks are not source code as far as software freedom is concerned, only the training tools are. The difficulty of training is a separate issue, that is solved/mitigated by decentralization.
@namark @cwebber @kragen @phoe @ivan
I updated the post last night at around 3am EST (if that gives you a sense of how well I'm sleeping these days)
I mentioned AI explicitly,. and the challenge there is that if reproducible builds are hard to do generally, with AI they're impractical or sometimes impossible.
Sure it is impractical if you try to do it on every node of the p2p network, but if you have any kind of trust mechanism in place, the network can collaboratively build it for everyone, and distribute it in that built/trained state(along with all the tools, which I guess they would have anyway as they took part in the process). Software freedom does not imply isolation, and this is true in much more general sense. As an individual I might not care if you provide me with source code to build something myself as long as you provide it to my trusted dev team(be that local independent organization, government organization, international organization or my homies on darknet) that can do it for me personally if I can afford it, or for the community they maintain/support which I'm a part of. A collective will always have this kind of advantages over an individual.
Not being physically able to build it on your own is not a reason to consider what is essentially machine code - source code.
Another similar case would be renting a supercomputer that's running free software. Same exact software with same exact settings might take several lifetimes to yield any results on my own machine, it is for all practical purposes impossible to use, but that doesn't mean that it violates freedom 0.
I would be surprised if AGPL does not accomodate things like that.
