Hi Greg, Ludo',

On 24/6/26 18:25, Greg Hogan wrote:

And
the documentation is clear that simple contributions which separately
are non-copyrightable can as a compilation become copyrightable. But
Guix does not aggregate copyright, instead simply requiring licensing
of the derivative work under the GPLv3. Since "legally significant"
cannot have the same meaning within the Guix project, how are we to
understand the use of this term in GCD 008? The text says "to ensure
the contributor has a valid copyright claim on the code" but the rule
of thumb seems to attempt to list non-copyrightable contributions.

The other day I realized the GCD is actually worse than this: it disallows the use of genAI for uncopyrightable things like (simple) package descriptions, but it allows using genAI in a way that affects our rights of the compilation of packages.

The GCD allows me to setup an agent to ask it something like this: "I'd like to upgrade NumPy to 3.0, what would that entail?" It would work overnight and then reply with something like:

- Simply refresh packages a, b, and c to their latest version.
- Apply patch from P.R. 1234 to package d.
- Deprecate package e, as it has not been updated in three years.
- Write a substitution to allow NumPy 3 for package f, it works fine.

As long as I write all the resulting code myself, it will be acceptable according to the GCD.

But the only creative work here is in the selection of packages and patches, in creating the collection. In Guix as a distribution. The code itself is factual and requires less creativity/originality then cooking recipes (which are famously uncopyrightable).

So the GCD disallows genAI to perform work that is not copyrightable to begin with, but allows using genAI for the copyrightable database part.

(IANAL etc. I think not all countries have "databankenrecht", and if they do it can be copyright, or an independent right.)

Hugo

P.S. in this hypothetical example I'm assuming I would have judged it ethically, environmentally and socially fine to run an agent in this way, a judgement I have not made in actuality.


Reply via email to