Apologies,

"I don’t want to contribute to problems that sees generated code..." should read "I don’t want to contribute to projects that see generated code..."

 -- Ian

Ian Eure <[email protected]> writes:

Hi Florian,

"pelzflorian (Florian Pelz)" <[email protected]> writes:

Hello Ian.

What else can we fix? Or is it only a policy issue? Or unrelated
to
LLM?

It is a combination of multiple things.

The LLM policy issue is the main one. I don’t want to contribute to projects which treats recombinations of infringing code as having value equal to original work; I don’t want to contribute to problems that sees generated code as more important than the planet’s climate (or ones where those seem likely outcomes). At least to me, this is not a difficult question, so even seeing the discussion around this
subject has been a disheartening look into the mismatched values
between myself and others in the project. The end result of GCD 008
was much weaker than what I’d like to see, and still failed.

The asymmetry of expectations within the GCD process. We are expected to accept the objection, while the objector is allowed to reject the
expectations of the project, ex. they seem to have blown up the
process and vanished rather than engaging to solve the single, minor, material issue they raised. While I do not agree with invalidating
the objection, I also believe that a process tolerant of this
asymmetry is unlikely to result in any but uncontroversial changes.

Lack of visible effort to address either the direct failure of GCD 008 or the underlying issues leading to it. Maybe they’re happening, but if so, they don’t seem to be happening in public. And we (guix-devel) were explicitly told to stop talking about these issues. How are we
supposed to find a way forward when discussion is prohibited?

LLM commits hitting master. While it’s only a couple at this time, there is no policy against them, and there seem to be many people who wish to add them. Call them the straw(s) that broke the camel’s back, but my heart absolutely sank when I saw. Absent a policy prohibiting LLM commits, I expect more. Absent fixing the dysfunction in the concensus process, I expect no prohibition. The more LLM commits that
get added, the harder it will be to unwind; the window to change
direction is rapidly closing.


Reply via email to