On Mon, 27 Jul 2026 16:03:28 +0200 Gabriel Wicki <[email protected]> wrote:
> That being said, the amount of social pressure to disapprove of such a
> long and hard discussion is almost unbearable. I can imagine others
> preferring not to expose themselves against what could be perceived as
> an overwhelming consensus of a vast majority of GNU Guix. I don't
> know if I am the only one opposing, but the amount of exposure with
> the potential attack-surface ("we had such a great consent except for
> this one turd in the punchbowl") is highly problematic for
> consent-finding processes. Voicing disapproval should be encouraged,
> otherwise we will never be able to truly find the best solutions for
> all of us, and slowly lose valuable participants, that prefer to
> remain silent than exposing themselves, one discussion at a time.
While I did vote in favor of the proposal at the end, I have exactly
the same concerns as you in your shorter explanation.
And I guess I can speak now as the deliberation date is now passed:
> Monday, July 27th, 2026, 23:59 AoE (anywhere on earth)
Personally I went extremely close to have a total mental breakdown
during these discussions.
For me, on top of that:
- The stakes were extremely high for me: I co-maintain GNU Boot
(https://www.gnu.org/software/gnuboot/) and did *a lot* of sacrifices
for that project, and I'm not the only one, and we need to have low
legal risks for many different reasons.
A wrong decision could have shut down GNU Boot and also the
Chipflasher project as well (http://www.zerocat.org/chipflasher.html)
which GNU Boot depends on, and I only knew what to expect after the
deliberations started.
And also note that for people that only value work done, many of my
work in Guix is not very visible because it's done outside
(distribution packages, advocacy, etc).
- The heat waves. I slept ~3 Hours per night for a week or more. And I
usually don't sleep very well.
- A few years ago, I had an extremely bad experiences during charter
chang discussions in a hoster collective
(https://www.chatons.org/en) where I represented a non-profit free
software user group that was a member.
During these discussions, the worst tactics (pushing people to fight
each other, ad hominem attacks, etc) were common practice to push
people to vote for the charter change, and at the end their tactics
worked. This was not subjective: several people confessed that they
voted in favor just to make the discussion end, and some people
pushing for confrontation also openly admitted that they were doing
that to get things done once for all. For the record I didn't use any
of these tactics there and I was a victim of the ad-hominem attacks.
To avoid similar issues we need to find ways to avoid escalation
otherwise each side would be pushed to use these tactics as they do
impact votes.
Having good moderation might help but I don't think that it is magic,
as moderation is subjective. In Guix the people that were in charge
of doing the moderation also had stakes in the discussions.
And I know one case where where a reply on the mailing list may have
been misunderstood as a moderation action and I also disagreed with
the "moderation" comment, though I had other concerns with the mail in
question and took it off list for discussions about it, both about
the content (with the author) and the way it was posed (with the
author and the moderation team which didn't reply if I recall well).
So now that the deliberation is over, we probably need to start new
discussion(s) with ideas about what could be improved in the GCD
process, with a special focus on trying to make decision as smooth as
possible, limit conflicts, etc.
In my opinion the GCD process as it is now, was not ready to deal with
issues as complicated as LLMs.
So let's take the occasion to learn from what went wrong here and adapt
the GCD process to better deal with highly controversial topics[a].
Maybe we need to open new threads about the process, but I also think we
badly need in-person discussions for all that. This could take time (~1
year, maybe a bit more).
This is something that really needs time and research. This means that
we also need to spend time looking into how other projects take
decisions.
Debian would be something interesting to look at. And I also think that
we also need to recognize what worked well as for instance there are
really interesting things in the GCD 008 as well that isn't present in
any of the Debian proposal about LLMs (
https://www.debian.org/vote/2026/vote_002).
> ~~~~~~~~~~~~~~~~~~~~~~~~~~ longer explanations ~~~~~~~~~~~~~~~~~~
============================ Longer replies =======================
> 1. Discussions never really calmed down. The proposal was altered
> until the very last moment. Multiple people asked to withdraw to
> have more time. While I don't think any one of these are absolute
> thresholds to disapprove of a GCD, I think they are good indications
> that more time is needed.
Agreed.
> 2. I just don't buy the "if we don't regulate it now it will be too
> late and Guix is doomed forever" argument. This feels like
> fear-mongering and should not find its way into consent-finding
> processes. Please excuse the probably historically inaccurate and by
> terminology xenophobic comparison (damn my western/european
> upbringing!), but if the Barbarians were to invade our realms and we
> fear they set our village(s) on fire, I can't imagine us sitting down
> for a multi-month process to figure out with what solutions most
> people are most comfortable with.
I have a way better comparison: the process to pass Villainous laws, but
here it's only a partial comparison.
Unlike the discussions in the hoster collective I mentioned above, not
everything apply, so the idea is more to look at how close we are from
that, and try to see the issues with the process.
So let's see. According to a book about the topic ("Enemis d'États"
from Raphael Kempf, in French, I'm unsure if it's translated or not):
1) A villanous law if often adopted in the urgency in the heat of
the emotion of an event. Here we do have the perceived
urgency so I think it applies.
2) It brings back previous attempts that did not make it or stayed in
drawers. It doesn't apply because LLMs and their discussion in Guix
are relatively recent.
3) The discourse around the law does the opposite than the law does.
At the start, it did: the title of the GCD was against LLMs, but the
GCD did open the door to it. And using LLMs were already forbidden by
GNU but not mentioned.
But it stopped applying when the title fixed, which I think
is good.
But as you point it it probably needs to be combined with more time
to deal with the changes in one way or another.
4) During the debates there is often a compromise (between freedom and
some interests) with regard to a future villanous law or more
extreme laws. Here I'm unsure if it applies or not. I didn't follow
enough of the debates for that.
But this is also part of what made me vote in favor: I was afraid
that we could have something worse if it didn't pass. And I found
some part neat and some remaining issues as well.
The rest of the points of the book seems to only apply to state
laws, so I would disregard that for the comparison[2].
The concerns Simon Tournier had are also a direct byproduct of the
process: because of the urgency people can't usually manage to make
villanous laws logical and consistent, but I would also temper that as
here the GCD isn't even close to contradict itself as much as a real
villanous law would.
In any case the process is not good enough in my opinion, but I find the
constructive approach is really interesting.
> If the fear were real, I guess the adequate, temporary solution would
> have been to enact a moratorium until legal questions are
> satisfactorily answered. We do have the tools (consent in each and
> every PR) to black any change that could potentially hurt the project.
We already have that moratorium as decided per GNU. We could push GNU
to make it official or something like that as this would be easier for
everybody because Guix votes and GNU works differently (it has
committees etc), and we need 100% public info, with exact text for
voting, not something that only maintainers can look at, with very few
context obtained though discussions in a GNU maintainers-only mailing
list.
> 3. I find it highly questionable that the first time we should
> introduce contribution guidelines is to tell people not to use the
> wrong tools. We are a project by people, for people, so contribution
> guidelines should be rooted on that fact, and not upon (potentially
> questionable) tooling choices.
I disagree here as the usage of tools can have consequences. Examples
before LLMs are Microsoft Word + its docx file format, or a nonfree
text editor that use a standard text format. Both have different
consequences. And the docx format can also have legal implications as
per the GFDL, depending on the details.
> 4. The term genAI is not (properly) defined. Are we talking about
> proprietary SaaS (which AFAICT is clearly regulated FSDG-wise: we, as
> a project, can and will neither use nor advocate)? Do we include
> everything in the future that can be labeled genAI? Is this "just"
> about large-language models or are we including other "AI"
> technologies that can output code? Are "coding assistants" (like the
> ones currently baked into VS Code and the like by default) meant as
> well, or are we just talking about `prompt-to-code' tools?
I also agree. Personally I disagree that "AI" is meaningless as it
really fits well if we consider it an ideology. And the case for that
is easy to make by following who puts resources into "AI" and why,
under a long term historical perspective.
So to me it would be tools that follow this ideology and generates
something, which is way too vague for Guix. And if people mean LLMs,
let's just use LLMs. For instance:
> LLMs (sometimes mistakenly mentionned as "GenAI", which is not
> clearly defined), [...].
or have some definition.
> 5. The "Luddite" angle reminds me of how people have probably argued
> back in the 80ies, when general purpose C compilers took away the jobs
> of the crafty assembly programmers.
Note that there are 2 "Luddites".
The historical ones are really close to the free software movement, as
they tended to make their own tools, do technical innovations and so
on. We also do that, and Guix is a good proof.
The technique also has a long history, and it is not linear. Different
people pushed different techniques, so you had huge confrontations
between different groups pushing different techniques. We also confront
with nonfree software, SAASS, etc.
And then there is the common perception of Luddites that makes them
become against (advancement in) techniques. This has nothing to do with
the real historical movement.
The question for us and the historical Luddites is more how to
preserve our freedoms when competing techniques push against freedom,
and it is not easy. Luddites literally went to war against people
pushing anti-freedom techniques (after exhausting pacifist means). There
are other examples like community supported agriculture where people are
still there, still exist, still have some freedom left, etc.
Self-subsistance farming also still exist. In many fights to keep doing
self-subsistance farming, freedom is mentioned as the reason of the
fight.
> Without adequately defining *what exactly* we are banning or
> restricting, thus by accepting this GCD, we burden ourselves to
> another one of these absolutely nerve-wrecking processes.
I also agree here.
But we might also need documentation that goes with all that and we
went faster than GNU here, so we lack that.
An example[3]:
> If you use JetBrains to disable LLMs, [...] hoover over "Disable AI
> Assistant" and click "Disable for this project".
References:
-----------
[1] For comparison I managed to follow from start to end the GCD about
the name to use for the main/master branch and I found the process
really adapted for simple cases like that.
[2] The law is not made only for judges but also for the police, It is
made against a specific group of people and then applied more
widely, It aims at the intent more than the action), especially
because my point is to compare the process of passing these
villainous laws and not the laws themselves (Guix isn't a state, and
it's supposed to be inclusive as well).
[3] https://jetbrains.com/help/ai-assistant/disable-ai-assistant.html
note that this page uses JavaScript.
Denis.
pgpvrG1KzgKHW.pgp
Description: OpenPGP digital signature
