On Mon, 09 Nov 2026 16:55, Lucas Nussbaum <[email protected]> wrote:
On 09/11/26 at 12:47 +0100, Matthias Geiger wrote:
3) Directly contradicting with the GR outcome
As Ian and Maytham already pointed out, this a direct violation of the GR.
It is a large-scale change, that was neither discussed, nor was consensus
sought, nor announced. It just was there one day.
I would expect at the very least that the DFSG team puts this on hold,
disables the mails from the LLM, and has a discussion on -project.
Otherwise, what good is the GR if it is factually ignored ?
As a data point, the GR option I proposed (option B or 2 in
https://www.debian.org/vote/2026/vote_002.en.html ), which was only
beaten by the winning option, had:
Prior Discussion of Bulk or Automated Changes: Similarly to the
mass-bug filing process (Developers Reference section 7.1.1),
contributors should discuss their intention before submitting bulk or
autonomously generated contributions. Any such automated process
should be overseen by a human who remains accountable for its behavior
and output.
And I think that the process implemented by the DFSG team is OK.
I draw the line on whether the automated process involves sending
information only, or whether it requires an action afterwards. If I
submit many bugs, or open many merge requests, one has to close them. If
I mass-"fix" many packages by pushing to their Git repositories, someone
has to clean up the mess.
But here, there are no consequences if you just ignore the notification
that the review output is available.
The main point isn't whether this has consequences for NEW, but rather
that I get a non-human message I didn't ask for. And this is very
different from mass bug filings: Those need to announced on -devel,
and then after a discussion those are sent. However, those provide
value, such as coordinating a transition or removing outdated
dependencies from the archive. Those reports serve a goal to simplify
and coordinate work across the archive. An LLM "review" of my copyright
does not. It it just a wall of text from a model, that provides no value
for the uploader. I can just run licensecheck locally and can get the
same result.
We have many existing tools to check the copyright, and I use those to
ensure my package is license compliant. Very rarely I miss something,
and I am grateful for the DFSG team pointing it rightfully out then.
However, sending mails to every person uploading to NEW is an incredible
waste of maintainer time and resources.
Also, I think it's obvious that those reviews are generally useful and
correct, and that the prompt used to generate them is fairly detailed,
so those reviews cannot be considered slop.
As I pointed out earlier, regardless what one thinks of the reviews, any
due process in the project was ignored. IMO the reviews hold no value,
as they are wrong at times, as pointed out by others, too. Yes, they
might be 95% correct, but still contain errors (see my first mail to
this thread). And sending an automated mail generated by a statistical
model to anyone uploading to NEW is slop in my book.
It's akin to spam. Again, if people want it, they should be able to
*opt-in*. Pushing this by default without any discussion or announcement
is outright hostile IMO.
best,
--
Matthias Geiger <werdahias>
Debian Developer -- KiCad Contributor
GPG: ECBEDBB607B9B2BE
