Re: GR -- Allow AI-Assisted Contributions

On 04/03/26 at 10:37 -0700, Bdale Garbee wrote:
> I'd love to see this conversation focus less on concerns about who is
> or is not using whatever tooling to help them write code, and more on
> the longer term support implications.  Things that worry me, like what
> is the preferred form of modification for code written by issuing chat
> prompts?

Assuming we are making the hypothesis of an LLM that is packaged in
Debian and used as part of the build process of the package (so it's a
build-dependency, and does not require internet access during build),
how is that different from the 'bc' source package's use of flex/bison
to generate C source files[0], or the 'swiglpk' source package's use of
swig?

In both cases, the preferred form for modification would be the input to
the tool, not the generated source code.

> And what does a reproducible build look like if we someday
> decide prompts are source code and there's a significant
> non-deterministic element operating between the human expression of
> intent or need and a compiled object code? 

I agree that's a difficult problem. I wonder if we already have packages
that are unreproducible in some cases due to floating-point arithmetic?

Lucas

[0] actually the 'bc' source package does not call flex/bison to
regenerate generated source files like bc/bc.c, but it probably should.