Hi Enrico,
Am Tue, Nov 03, 2026 at 06:33:22PM +0100 schrieb Enrico Zini:
> Now, what are you doing these days to bring Debian somewhere
> interesting?
My first approach is to fix RC bugs in my team that nobody had the
capacity to address so far.
As one example, I have been working on upgrading a rather complex
package (vg). It had two open RC bugs for nearly two years, which could
actually be fixed with relatively straightforward patches to its CMake
files. However, we were also lagging about 20 upstream releases.
The new upstream release required packaging around 20 additional
dependencies, as well as upgrading and patching some existing libraries.
Analysing this dependency network with an LLM and getting a sorted todo
list was an enormous help. Working through the issues and
interdependencies and preparing patches that are useful to upstream
became feasible within the time I can realistically invest in this kind
of work. Without inference.debian.net, I probably would not even have
started this effort.
There are quite a few similar challenges waiting in my team. Of course,
removing complex packages from Debian is an option. But I think we
should be aware that refusing tools that help us deal with complexity
may gradually shape Debian into a distribution that can no longer afford
to provide some of the more demanding software our users need.
Once this backlog is under control, I would like to spend more time
bringing complex workflows into Debian: packaging groups of tools that
interact with each other, making their integration work properly, and
doing the necessary communication with upstream projects. This is
particularly relevant in Debian Med and the other scientific Blends.
This has been a long-term goal of mine and is at the heart of the Blends
concept. Making progress towards it has been difficult because day-to-day
maintenance work consumes all available time of the team.
Am Sun, 1 Nov 2026 10:36:45 +0100 schrieb Enrico Zini:
> Can we design for where we could go, rather than adopt an LLMs to make
> it supposedly easier to stay where we are?
I like this question, and I agree with Joey's underlying point[1]. But
I don't think these approaches are mutually exclusive. Debian is not a
black-and-white world. I'd be happy if we could acknowledge that there
are areas where LLMs enable us to make progress that would otherwise be
out of reach, while continuing to think ambitiously about where we want
Debian to go.
For me, the interesting question is not whether we should use LLMs to
preserve the status quo, but what additional things we can achieve with
the capacity they give us.
Kind regards
Andreas.
[1] https://joeyh.name/blog/entry/Debian_and_the_sirens/
--
https://fam-tille.de
