For 26.12: stay on Debian sid, or move to trixie + backports?

[MTSistemi]

The most reasonable criticism this project has had is that it is built on Debian sid:
newest software, and it can break. On a machine that is also a family games console, that
matters more than it would on a workstation.

The plan for 26.12 is to move to trixie (stable) plus backports. Before we commit,
we want to know what people running it actually prefer.

	Option	What you get
👍	Move to trixie + backports	A stable base, with a recent kernel and graphics stack where it matters. Fewer surprises, slightly older userland.
❤️	Stay on sid	Newest everything, including Mesa and Wine improvements that reach gaming quickly. You keep the risk of a bad day.
🚀	Both — sid as a separate edition	Honest about the cost: it is a third image to build, capture, verify and support, on a project with one board.
👀	No preference / I follow whatever you pick	Also useful to know.

Btrfs snapshots exist precisely because sid can break, and they stay either way. The
question is whether you would rather not need them so often.

---

Voting is by reaction: GitHub's API cannot create native polls, only the web interface can, and we would rather have a working poll today than a perfect one later. One reaction per option, react as many times as you like if more than one applies.

[Wendigogo]

Why not moving to Forky instead? Wouldn't it be easier to move to the testing version as you're based on Sid? And as it is planned to be the stable one in mid-2027, you don't have to move to Forky only 6 months after the Trixie release...
Moreover, Debian 14 “Forky” will also be the first Debian release to ship with native rollback, undo, redo, and history features for its default APT package manager. It sounds like your snapshots feature but from the OS point of view.

My 0.02 cents

[MTSistemi]

Thank you @Wendigogo, that is a fair point and it was not in the poll.

What Forky would give us: no second migration six months after the first one, since trixie hands over to Forky in mid-2027 and we would already be there; and APT's new history with undo and redo, which overlaps with what our snapshots do today.

What weighs against it: testing gets security fixes later than stable, because they normally reach it by migrating from sid after the usual delay, and it can still move under us during transitions. The transitions are exactly what bit us this year: the LLVM 22 move broke our Mesa, and WirePlumber 0.5.18 replaced the version our audio package was pinned to. In sid we see those the day they land; in testing a few days later; in trixie not at all.

So it is really three options, not two, and Forky goes on the list. Nothing is decided for 26.12 yet. If others have a view on sid, Forky or trixie + backports, this is the place for it.

[Wendigogo]

Thanks for detailed explanation.
Then, stable Trixie + backports are the best option IMO.
I can't understand how "no tinkering" is able while basement is unstable. 😅

And Debian will probably provide a migration guide at some point.