Earlier quoted context omitted.
My two cents, you don't have to agree but thought I'd just add a different perspective. I prefer my own package manager (pacman) over snaps for the following reasons: 1) I like to upgrade on my own schedule. I use my computer for work, and I cannot have things break in the middle of the day or in the morning, just as I get started with work. I usually save upgrades for when I less things to do, so that in case stuff…
My position is not that certain users would prefer to update on their own schedule. Of course there are. Or that new updates sometimes break things and that's a hassle. Of course they do and it is. The problem is if you want to distribute an important security update, what do you do? Ask everyone nicely to upgrade? How? Again, what % of users will manually update their software? Not a lot. For #2, that seems like a r…
Even I recognize that this doesn’t make sense in the Linux world, though. Ubuntu is trying to be something it’s not—they’re trying to appeal to a new demographic, and, in doing so, driving away their existing users.
Even with my stance on auto-updating, snaps are a problem for me because I mostly use Linux in the context of servers. Like it or not, that’s where Linux has the largest market share; Android aside, Linux’s consumer market share is negligible.
In that context, snaps have problems:
- I can’t have my servers updating on their own. Security updates rarely break things, but most other updates need testing.
- I use auto-scaling. That means servers need to come up quickly when load increases. If a bunch of new servers come online and all decide to update, that’s worse than no servers coming online.
- I don’t want or need a sandbox. In a cloud environment, the server is the sandbox.
- Environments and server states need to be reproducible for testing and auditing. If I’m doing a post-mortem, I need the software on the relevant image to be in the exact same state as when the problem occurred.
- Performance is critical. I’m already paying AWS for sandboxing in the form of many small EC2 instances; I don’t need the additional overhead of snaps. I’m not working with bare metal.
All of these issues could be resolved, and I wouldn’t object to this experiment running in a non-LTS or desktop-only release. But it is truly an experiment: snaps aren’t ready for prime time. My options are to pay Canonical for extended support for old software, wait it out and hope the issues are sorted before I stop receiving security updates, or switch to something like Alpine or Debian.
Keep in mind, this is coming from someone who loves automatic updates, generally prefers systemd, rather liked Unity, and didn’t see what all the fuss was with Upstart and Mir:
- systemd works pretty damn well and is a big improvement, although it has its hiccups
- Unity was fine. It looked nice out-of-the-box and wasn’t a resource hog.
- Upstart usually worked well enough, though it sometimes had reliability issues.
- Mir never really saw the light of day, so it didn’t matter.
Snaps are where I draw the line. They might be the future, but they’re not ready for the present. And that’s not for lack of trying on my part—I had no trouble embracing Upstart and later systemd.