On 2021-09-15 at 06:45, Brian wrote: > On Tue 14 Sep 2021 at 22:42:12 -0400, The Wanderer wrote: > >> On 2021-09-14 at 16:33, Tanstaafl wrote: >>> Hmmm... ok, so, I could run sid 'forever', as long as I keep it >>> updated regularly? >> >> In theory you could, but in practice it would break well before >> that. >> The guiding principle of running a system that tracks sid is "if >> it breaks, you get to keep all the pieces". >> >> It is NEVER advisable to track sid on a computer you're not willing >> to blow away and reinstall on demand if necessary. (As distinct >> from installing specific selected packages from sid on a >> case-by-case basis - but be careful even about that, as the >> dependencies of those packages might pull in enough other things to >> lead to a hybrid Debian system and potentially break things.) >> >> I would advise against tracking sid on any computer other than one >> you're running specifically to contribute to the process of testing >> the contents of sid before they migrate into testing. >> >>> Anyone do this for important (maybe not 'mission critical') >>> servers? >> >> I certainly hope not. (And am mildly horrified that someone who >> posts as much good advice here as I believe I've seen from Brian >> has said that he does.) > > I am in agreement with what you say as regards stable vs unstable. > For the avoidance of doubt, I would always advise stable for a user. > It has been thoroughly tested, gets timely security upgrades and is > supported by the images team with point releases. What is there to > dislike about it? There are a few possible reasons; the primary one is that new packages, and new package versions with new features, don't appear for a long (potentially *very* long) time. For myself, I recommend stable for anything you aren't willing to do active maintenance on (the way Tomas referenced doing with an unstable machine), and testing - with, optionally, fallback to stable - for anything else that's even vaguely a production machine. My daily driver is the latter configuration; a server-ish machine I run at work is the former. > My response was simply to indicate that some users do run unstable, > hopefully knowing what they are doing. I see it as a way of > contributing to a future stable and would not use it on an important > machine. Someone has to watch out for Debian and upstream bugs in > packages of interest to the user. Agreed, and appropriate. I sometimes regret that I'm not in a position to be able to afford the workflow that would make "eh, just reinstall" a viable model for any of my computers, because that would make tracking sid and helping test it viable. (Helping test sid was, IIRC, the primary reason why I decided to track it at one point in the past.) > I was also rather hoping Tanstaafl would contribute a few words on > how the unstable model contrasts with Gentoo's rolling release > model. That could be interesting, too. My understanding is that the closest Debian equivalent to rolling release is the sometimes-discussed but never-really-implemented possible goal of "continuously usable testing", i.e. a model in which there's never a release freeze of testing (and releases are done through some other avenue) and - I think - library transitions etc. are handled in a sufficiently atomic fashion that you don't wind up with some packages temporarily unable to satisfy their listed dependencies. I think there's still an interest in principle in achieving that goal, but if there's any movement towards getting closer to a point where it actually happens I'm not aware of that. -- The Wanderer The reasonable man adapts himself to the world; the unreasonable one persists in trying to adapt the world to himself. Therefore all progress depends on the unreasonable man. -- George Bernard Shaw
Attachment:
signature.asc
Description: OpenPGP digital signature