Draketo neu
debian and debbugs and fomenting problems
Comment to the video Debian is going through a crisis.
Upgrading systems to systems that new people want to work on can also mean that old people with experience have to change without benefit to them, and that is a main contributor to FLOSS devs burnout. 73% of FLOSS devs already experienced burnout at least once. There’s a report by itsfloss: "Open Source Developers Are Exhausted" that describes this finding along with causes.
It’s possible, not to break old systems, but that takes careful consideration.
About debbugs there’s a small, easily overlooked fact:
It was mainly written by Ian Jackson, the former Debian project leader. Ian Jackson resigned in 2014 when systemd was pushed through as init system with a crazy amount of vitriol. You can get an inkling of the amount of pain its critics went through by reading entry by OpenRC:
OpenRC vs Systemd
I have decided to not write anything in this section, considering the aggressive tone I'm getting in return, which is all but fun. Anyway, the problems with Systemd have been debated a lot already, so it is useless to list them here again.
Well paid devs pushed out the maintainer of debbugs (Ian Jackson, debian project lead till 2014) by pushing through systemd, and the paid devs who caused the damage didn’t care about finding a new maintainer for a decade. That’s a crisis a long time in the making.
But Debian is still an essential project and does a lot of things right, especially not breaking things. If you want to help it to preserve and improve a corner stone of Free Software, there’s a link "get involved" on the debian frontpage. That has a link "New member". There you find the information how to join.
If you check that out: Thank you for wanting to step up!
Debian is useful for corporations because of its principles, but it does not have those principles to please corporations. It has them because they are the right thing to do.
For example “you’d break Debian for the Dreamcast tinkerers” is an argument to preserve compatibility that very likely annoys corporations who want it to move faster and break things (but only for others). Where else than in debian would you -- when you make that argument -- actually have a Dreamcast tinkerer answer in the mailing list. Happened to me november 2025.
LLMs require human input, else their implicit priors become dominant
When I did inverse modelling in my PhD we always made sure that we know what’s prior model info and what’s actual data.
You had to know which evaluation includes prior data in addition to measurement data, because assimilating multiple sources with the same prior could cause the prior to become dominant.
Todays LLMs repeatedly ingest their own outputs -- outputs that contain their own implicit priors.
It’s obvious that preserving the model quality requires pure human input.
Though from inverse modelling I’d guess that improving the quality of the prior -- of the output the model would generate without training -- could improve the quality of the results when trained with less data.
This is not new, but I now connected the dots from what I know to LLMs. For a discussion and refrences see Wikipedia: Model collapse.
Diese Seite nutzt Cookies. Und Bilder. Manchmal auch Text. Eins davon muss ich wohl erwähnen — sagen die meisten anderen, und ich habe grade keine Zeit, Rechtstexte dazu zu lesen…
