Live data from Hacker News

No right to relicense this project

github.com

41–50 of 388 posts

Re: No right to relicense this project

#41
post #12

Earlier quoted context omitted.

That whole clean room argument makes no sense. Project changed governance and was significantly refactored or reimplemented... I think the maintainers deserve to call it their own. Original-pre MIT release can stay LGPL. I don't think this is a precedent either, plenty of projects changed licenses lol. I keep kind mixing them up but the GPL licenses keep popping up as occasionally horror stories. Maybe the license is…

No. Because they couldnt have done any of that refactoring without a licence to do so, and that licence forbids them from relicencing it.

Hmm are we in a ship of Theseus/speciation area? Each individual step of refactoring would not cross the threshold but would a rewrite? Even if the end result was the same?

Re: No right to relicense this project

#43

Earlier quoted context omitted.

>the ai copy pasted the existing project. The change log says the implementation is completely different, not a copy paste. Is that wrong? >Internal architecture is completely different (probers replaced by pipeline stages). Only the public API is preserved.

It's up to them to prove that a) the original implementation was not part of whatever data set said AI used and b) that the engineers in question did not use the original as a basis.

It's up to the accuser to prove that they copied it and did not actually write it from scratch as they claimed.

Re: No right to relicense this project

#44
post #21
post #12

Earlier quoted context omitted.

That whole clean room argument makes no sense. Project changed governance and was significantly refactored or reimplemented... I think the maintainers deserve to call it their own. Original-pre MIT release can stay LGPL. I don't think this is a precedent either, plenty of projects changed licenses lol. I keep kind mixing them up but the GPL licenses keep popping up as occasionally horror stories. Maybe the license is…

Governance change or refactoring don’t give you a right to relicense someone else’s work. It needs to be a whole new work, which you own the copyright to.

Which is what happened here? The maintainers did a rewrite, apparently, but it's not enough!

Re: No right to relicense this project

#45
post #29

Wow that's hot. I was not aware that you need to be "untainted" by the original LGPL code. This could mean that... All AI generated code is tainted with GPL/LGPL because the LLMs might have been taught with it

Being completely untainted is the standard many reimplementations set for themselves to completely rule out legal trouble. For example ReactOS won't let you contribute if you have ever seen Windows code. Because if you have never seen it, there can be no allegation that you copied it.

That is however stricter than what's actually legally necessary. It's just that the actual legal standard would require a court ruling to determine if you passed it, and everyone wants to avoid that. As a consequence there also aren't a lot of court cases to draw similarities to

Re: No right to relicense this project

#46
post #22

> 12-stage detection pipeline What is this recent (clanker-fueled?) obsession to give everything fancy computer-y names with high numbers? It's not a '12 stage pipeline', it's just an algorithm.

Isn’t it? I mean 12 stage pipeline has a very specific meaning to me in this area, and is not a new way of describing something. The release notes description sounds like a multi stage pipeline.

Do you know this kind of area and are commenting on the code?

Re: No right to relicense this project

#47
post #22

> 12-stage detection pipeline What is this recent (clanker-fueled?) obsession to give everything fancy computer-y names with high numbers? It's not a '12 stage pipeline', it's just an algorithm.

"ok chatgpt, what name do i give to this algorithm, so it sounds fancy and advanced?"

Re: No right to relicense this project

#48

Earlier quoted context omitted.

It will hold up in court. The line of argument of “well I went into a dark room with only the first Harry Potter book and a type writer and reproduced the entire work, so now I own the rewrite” doesn’t hold up in court, it doesn’t either when when you put AI in the mix. It doesn’t matter if the result is slightly different, a judge will rule based on the fact that this even is literally what the law is intended to pr…

> “well I went into a dark room with only the first Harry Potter book and a type writer and reproduced the entire work, so now I own the rewrite” This is not a good analogy. A "rewrite" in context here is not a reproduction of the original work but a different work that is functionally equivalent, or at least that is the claim.

Possibly important is that it’s largely api compatible but it’s not functionally equivalent in that its performance (as accuracy not just speed) is different.

Re: No right to relicense this project

#49
post #24

Earlier quoted context omitted.

This seems right to me. If you ask a LLM to derive a spec that has no expressive element of the original code (a clean-room human team can carefully verify this), and then ask another instance of the LLM (with fresh context) to write out code from the spec, how is that different from a "clean room" rewrite? The agent that writes the new code only ever sees the spec, and by assumption (the assumption that's made in al…

It requires the original project to not be in the training data for the model for it to be a clean room rewrite

That only matters if expression of the original project really does end up in the rewrite, doesn't it? This can be checked for (by the team with access to the code) and it's also quite unlikely at least. It's not trivial at all to have an LLM replicate their training verbatim: even when feasible (the Harry Potter case, a work that's going to be massively overweighted in training due to its popularity) it takes very specific prompting and hinting.

Re: No right to relicense this project

#50
post #11

FastAPI's underlying library, Starlette, has been going through licensing shenanigans too lately: https://github.com/Kludex/starlette/issues/3042 Be really careful who you give your projects keys to, folks!

That doesn't seem related at all, this is just adding attribution, not changing the license through LLM-washing
Post reply on HN