On Sat, Oct 18, 2025 at 12:17:53AM +0200, Antoine Le Gonidec wrote: > Le Fri, Oct 17, 2025 at 09:49:20AM -0700, Russ Allbery a écrit : > > Greg McPherran <gm@mcpherranweb.com> writes: > > > I am considering using unstable, and would be happy to provide feedback > > > on any findings. If someone could point me to the best contact for such > > > feedback, so that I may be helpful, I would be grateful. > > (…) > > I would recommend testing over unstable unless you're very familiar with > > Linux and comfortable with your ability to roll back to previous kernels, > > downgrade and pin packages, and fix weird problems. Unstable doesn't break > > that often, but it is prone to more low-level breakage than testing is. > > And I would recommend unstable over testing, for security reasons, less > risks of packages disappearing, and bugs actually being fixed in a > timely manner. > > In my 15 years of user support, a huge majority of reported problems > were with testing. But the sample might be biased by testing being the > one chosen by less experienced users, wrongly thinking it would be some > kind of middle ground between stable and unstable. In a way, it is: when packages land in testing they have passed through a rough first test. So in a way, you get less surprises in testing. In some other way... it's the other way around: packages "disappear" from testing when some bigger issue is found, but they stay on stable. Plus, since both don't have a security policy, security issues get (arguably) fixed faster on testing. If you (try to) install packages from a future suite (FrankenDebian), then testing is more often the better bet. If you run a future suite wholesale, the situation is not so clear. The difference to a rolling release is that in the latter case, more effort is put to avoid breakage. Cheers -- tomás
Attachment:
signature.asc
Description: PGP signature