Re: [Hamlib-developer] The use of LLM generated code in Hamlib (long)

[Daniele F. <iu...@gm...>]

TL;DR I'm not against LLM but I'm worried that they could produce code
identical to projects with incompatible licenses or projects with
compatible licences but that require attribution and it is missing. In
one word: plagiarism.

I think that the thing to address is the licensing part because we
aren't writing software only for ourselves, but we are also giving it
to other people saying that all the authors have licensed their
contributions under GPL/LGPL. This affects distros and possibly
commercial users.

If children contributed code to Hamlib I believe that we would need to
ask their parents to license the code, the same for
employees/employers. Is this even possible with LLM's? I believe that
Github had an AI assistant which allowed users to choose the license
of its data.

For some Hamlib is a hobby project and for some is a commodity. One
goal for Nate IMHO is to make maintenance easier for everybody.

To make it easier for downstreams that want to update the 4.x releases
already in use, I hope that we can clearly show that all of the
contributions are properly licensed whatever was their source.

For the future should we ask contributors to add a note to their
commit messages like the Signed-off-by tag required to contribute to
the Linux kernel?
It just tells that the person has the right to submit that code:
https://www.kernel.org/doc/html/latest/process/submitting-patches.html#sign-your-work-the-developer-s-certificate-of-origin
it's an annoyance but it works for them that have tens or hundreds of
new contributors each year that have to learn this additional step,
will it work for Hamlib too? Would it make any difference? Probably
for some downstream users, yes.

BTW, Gmail has suggested several corrections while I was writing this
email, I accepted some and refused others and some it did
automatically, was it AI or just a spelling and grammar checker?
-- 
73 de IU5HKX Daniele