Paolo Bonzini <[email protected]> writes:

> On 10/10/25 14:38, Markus Armbruster wrote:
>> The boundary between legal and illegal is a superposition of fuzzy,
>> squiggly lines, one per jurisdiction.
>> 
>> We can only try to approximate it from the legal side.
>> The tighter we try to approximate, the more risk we take on.
>> 
>> In addition, tighter approximations can be difficult to understand and
>> apply.
>
> I agree.
>
>> [Strong argument why type annotations are low risk snipped...]
>
> Note that type annotations are pretty much the upper bound of what I 
> would consider a mechanical change.  I would expect, for most cases, 
> that "include the prompt in the commit message" and the boringness of 
> the change are together already a satisfactory explanation.
>
> At the same time, I decided to try with a more complex change to 1) 
> avoid a slippery slope; it's easier to do so if you look at the hard 
> cases from the beginning, and Daniel did that very, very well; 2) probe 
> the limitations of the tool and ascertain if it's even worthwhile having 
> an exception.

It was a useful experiment.

>>> There's a definite "slippery slope" situation. The incentive for
>>> contributors will be to search out reasons to justify why a work
>>> matches the AI exception,
>> 
>> „Libenter homines id quod volunt credunt.“
>
> Yes, and that's why we should strive for simplicity if we are to have 
> exceptions.  If you cannot convince me with the prompt that your change 
> is mechanical/non-creative, don't even bother making complicated and 
> probably wrong legal arguments.
>
> Paying a certain price upfront (i.e., now) is fine, but in the long term 
> the maintainer's job wrt AI should be and remain easy.  There has to be 
> a cost, but then the same would be true with any policy other than 
> "don't ask, don't tell"---including zero tolerance.

Yes.

Cost is fine when the benefits are worth it.

Maintainer bandwidth is precious.  But it's not infiniyely precious :)


Reply via email to