How to Unslop a Commit?

Philosophical question: You want to implement a feature for a FOSS project. During web research you stumble over virtually the same thing that

* was apparently produced with a LLM / gen AI thing - according to the commit message (co-authored by...). * was not upstreamed (yet) for reasons not clearly stated. might be due to project policy, neglect or something else. * is close to trivial in nature - it's not apparent how the same outcome could be achieved with a significantly different approach at the core of it.

How would one get themselves in a position to - in all conscience - submit the patch in a LLM-free form?

▲

vincent-uden 6 hours ago | parent | next [-]

What perceived problem do you have with reading the LLM-generated commit, understanding its purpose and submitting your own PR solving the same problem in practically the same manner?

Without knowing more about which of the many reasons you might find LLM commits unethical or otherwise bad it is impossible to give any reponse.

▲

blackbit 6 hours ago | parent [-]

This is a question of principle. The objective is to gain the ability to rightfully claim "I have written this". There are a number of imaginable reasons to want this. Copyright, ethics, upstream project policy, you name it.

To put it to the extreme: Let's say I take the commit as is, typing what I learned verbatim off the screen as is but removing the "Co-authored by" note of the commit message. Then the LLM provider sues me, claiming I have no right to remote the note. Who would win in a court of law? I am not looking for binding legal advise here of course, I am just interested in what people here think.

	▲	vincent-uden 5 hours ago \| parent [-]
		To start off, law and ethics are not the same. Law is definitely not the extreme of ethics, I'd almost say it's the other way around. Law is just a practical (and imperfect) implementation of an ethical system. At least in the U.S. (and the european countries I know of) only a human can be considered an author for copyright. Thus the code belongs to the person who prompted the LLM. As far as I know the AI-providers make no claim on the copyright of AI-assissted code generation. Regardless, that is not what stops you. If you find this code online, in some repository that hasn't been upstreamed yet it is extremely likely for the code to be under some license. Many of these licenses have no requirement for attribution at all, thus you can just copy the code no matter who the rights holder is. Some require that the code must be re-distributed under the same license, but if it was a change intended to possible be upstreamed that is already a non-issue. This is my laymans understanding of the legal case, which is almost entirely separate for any ethical analysis (or principles).

▲

pestatije 4 hours ago | parent | prev [-]

well ur stealing from burglars, so the way to go is to announce what ur doing with full disclosure

	▲	vincent-uden 4 hours ago \| parent [-]
		It is almost definitely not stealing if the code is written in a fork as would be the case in the hypothetical scenario of finding non-upstreamed code. It would depend entirely on the license of that code which will almost certainly be the same as the license for the base repo in the outlined case. Most common licenses (MIT, BSD, GPL-2, GPL-3) would allow copying the code without any issues