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

[Mikael N. <mik...@fa...>]

Answering myself with this FSFE article about source code, copyrights and some considerations about AI:

Legal Corner: The threshold of originality for copyrightable source code:

https://fsfe.org/news/2025/news-20250515-01.en.html

This topic is a rabbit hole...

73,
Mikael OH3BHX

On Sat, Feb 28, 2026, at 01:23, Mikael Nousiainen wrote:
> Thanks Nate, enjoy the fine weather and time outside! Here in Finland 
> spring is still 1-2 months away ...
>
> I spent some more time reading the Debian mailing list, which has some 
> good discussion on the genAI topic and found this message, which also 
> discusses potential accidental plagiarism and code cloning, please have 
> a look:
>
> https://lists.debian.org/debian-vote/2026/02/msg00032.html
>
> More specifically, there are links in that message to the Linux kernel 
> code generation and assisted coding guidelines:
>
> - Kernel Guidelines for Tool-Generated Content:
>   
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/generated-content.rst
>
> - AI Coding Assistants:
>   
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/coding-assistants.rst
>
> Especially the latter document is rather clearly worded and simple - 
> and still gives the option for developers to use coding assistants and, 
> most importantly, emphasizes that contributors must use their best 
> knowledge and skills to assess the output for potential suspicious code 
> (regarding cloning/plagiarism/correctness/security/...).
>
> My personal thoughts about Hamlib and the risk of code cloning and plagiarism:
>
> Regarding potential license incompatibility and plagiarism caused by AI 
> code, I think the _vast_ majority of code that should ever end up in 
> Hamlib is likely to be rather simple in terms of what it does - and 
> very specific to the "niche" of controlling (ham) radio devices: we're 
> talking about (mostly) trivial abstraction layers for devices, Hamlib 
> command parsing, device-specific command generation and response 
> parsing. Of course, the documentation used for controlling devices must 
> originate from a freely available/compatible source, but that's what we 
> trust contributors to do anyway - even if they write all the code 
> without genAI.
>
> Anything more advanced - let's say we'd add a custom implementation of 
> an audio codec for a network rig or maybe a custom, non-trivial 
> algorithm for doing some data transformations - deserves more attention 
> and research on the origin of the code and I'd argue it is rather easy 
> to spot this kind of changes. Here is where the responsibility of the 
> contributor(s), reviewer(s) and maintainer(s) matters - just like with 
> any code contribution.
>
> Additionally, for many cases that need more complex/advanced features, 
> there are freely available libraries available (ones that have a 
> compatible license). As an example, FlexRadio rigs seem to support Opus 
> audio codec in their RX audio streams (in addition to PCM). Here, 
> instead of trying to "blindly prompt the LLM and generate an Opus 
> decoder", a good choice would be to add libopus (with MIT-style 
> license) as an (optional) dependency. The code generated as the result 
> would only involve calls to libopus, which make the contribution 
> significantly easier to review.
>
> What I'm trying to say is that even with code generated by coding 
> assistants, it's still possible to see whether a piece of code (e.g. a 
> pull request) consists of what I'd call "trivial" changes and that it 
> follows Hamlib code conventions/patterns (arguably indicating 
> negligible risk of license violations or code plagiarism) - or whether 
> it's a larger or a significantly more complex contribution.
>
> -> There are ways to use genAI where it's practically impossible to end 
> up with code output that would cause issues regarding license 
> compatibility.
>
> It's also worth thinking for a moment: what can even be interpreted as 
> plagiarism or a license violation? It has to be something non-trivial 
> and/or a larger piece of code. I have to admit my knowledge is limited 
> here. Does anyone here happen to know if there are any reliable sources 
> for guidelines on what can be considered a "work of art", something 
> that can be licensed and something that's clearly copyrightable? After 
> all, source code often contains lots of patterns that are common to 
> thousands of applications - so simply looking at individual snippets of 
> code that look similar or even identical doesn't always count.
>
> Seems I ended up writing here a lot more than I originally intended - 
> oh well - I'll stop writing/posting now for some days and wait for 
> everyone to consider their opinions and chime in.
>
> 73,
> Mikael OH3BHX
>
> On Fri, Feb 27, 2026, at 21:22, Nate Bargmann wrote:
>> Thanks, Greg.
>>
>> I'm not ignoring the discussions!  The weather has been too nice and
>> with the lengthening of the days, I'm spending more time outside as we
>> turn the corner into spring farm work.  As an example I need to get back
>> outside and put a wheel bearing assembly back together with new bearings
>> and seal.
>>
>> Next week the weather is supposed to be wet so I'll likely have more
>> time to sit and collect more thoughts and merge PRs.
>>
>> 73, Nate
>>
>> -- 
>> "The optimist proclaims that we live in the best of all
>> possible worlds.  The pessimist fears this is true."
>> Web: https://www.n0nb.us
>> Projects: https://github.com/N0NB
>> GPG fingerprint: 82D6 4F6B 0E67 CD41 F689 BBA6 FB2C 5130 D55A 8819
>>
>>
>> _______________________________________________
>> Hamlib-developer mailing list
>> Ham...@li...
>> https://lists.sourceforge.net/lists/listinfo/hamlib-developer
>>
>> Attachments:
>> * signature.asc
>
>
> _______________________________________________
> Hamlib-developer mailing list
> Ham...@li...
> https://lists.sourceforge.net/lists/listinfo/hamlib-developer