Defaults matter a lot, and snaps are the default in Ubuntu.
The topic is not whether snaps are avoidable or not, but the Ubuntu is going downhill. And snaps are purported to be part of that downhill, which would be Ubuntu's NIH syndrome. As far as I know, Ubuntu's only successful development is Ubuntu itself - the other projects have all failed over the years, and snap, while ongoing, is not winning any popularity contests either.
Snaps per se are no better or worse than flatpak. Canonical's mistake, IMO, was to make their store the only place snaps can be hosted. That is the "proprietary" bit everyone keeps talking about.
But in practice even for flatpak the only realistic place you can publish your flatpak if you want any traction at all would be flathub, so both formats have only one store right now. But flatpak allows a custom store while for some strange reason Canonical decided not to allow snap that freedom.
Another problem is, Canonical promised to release server components and enable alternative stores, and just forgot that they made that pledge.
Also, rugpulling users and migrating things to snaps without asking their users in order to "create a positive pressure on snap team to keep their quality high" didn't sit well with the users.
> But in practice even for flatpak the only realistic place you can publish your flatpak if you want any traction at all would be flathub
But, for any size of fleet from homelab to an enterprise client farm, I can host my local flathub and install my personal special-purpose flatpaks without paying anyone and thinking whether my packages will be there next morning.
Freedom matters, esp. it that's the norm in that ecosystem.
I was neutral-ish about Ubuntu, but I flat out avoid them now, and migrate any remaining Ubuntu server to Debian in shortest way possible.
I'm using Debian for the last 20 years or so, BTW.
Yes, same. I started with Ubuntu back in the day, because the server I inherited ran Ubuntu, and it was just natural after that for me to run it on the desktop as well. I grew to dislike their NIH over the years, tried distro hopping, and settled on Debian.
Yes, I agree. Snaps or Flatpak, not much of a practical, technological difference. What sets them apart is the way the distribution is handled, including the open source availability of the backend, which enabled for example Red Hat and Elementary to run their own stores.
If you are making your own distro, creating your own flatpak store is trivial, that's all what matters. Linux Mint doesn't use snap exactly because Canonical forces everyone to use their snap store.
Canonical doesn't force anyone to use anything. Snap is open source, just modify it to use a different store if you want. Mint literally forked a zombie DE, but changing a few lines of code in snap is an issue...
Defaults matter a lot, snap is not open source (client is, backend isn't), you cannot "just modify it (Ubuntu)" to use a different store, because Ubuntu installs snaps even with apt. Mint is not part of the discussion.
Red Hat do the same. They reinvented the wheel on multiple occasions (systemd and it's whole ecosystem like systemd-resolved and timed and the whole kitchen sink; podman, buildah, dnf, etc etc.)
They just have more success on getting their NIH babies accepted as the standard by everyone else. Canonical just fail at that (often for good reasons, Unity was downright crap for some time) and abandon stuff, which doesn't help their future causes.
Canonical did their own NIH init daemon called Upstart which failed due to the fundamental design and the implementation being plain bad. Redhat builds better software which is why their NIH gets more adoption.
They're not forced on anybody, they're not required by systemd, and many distributions use more feature-rich alternatives (including, afaik, RHEL — last time I looked at it, they used dnsmasq and chrony). They're also often shipped as separate optional packages:
> Still not anywhere near as popular as Docker. Although technically they're far better than Docker, and if anyone is using them, it's for that reason.
NIH packages are generally expected to be less popular, yes. They have some technical merit, though in my opinion that's mostly trade-offs rather than one being strictly better than the other. I would be surprised if everybody using them is using them because of technical merit as opposed to it being pushed by the distro.
Red Hat builds really good stuff. NIH is sometimes right because nobody invented the stuff at all. Standard Unix tools are great but they don't solve everything, so we've ended up with most distros having "the Debian way" or "the Red Hat way", the main difference of course being deb/apt/dpkg vs rpm/yum/dnf. When building an embedded system with Yocto, the basic choices are also Debian or Red Hat style, though you can of course do anything.
Special mention goes to NetworkManager, which has become the de facto standard way to configure networking because it's good. And with nmcli I can even remember how to connect to wifi from single user mode.
>They just have more success on getting their NIH babies accepted as the standard by everyone else.
This depends on the phrasing. We could also say that Red Hat produces actually useful software, in contrast with Canonical, whose developments don't seem to provide value over existing solutions.
We could also say that Canonical tries really hard to do exactly what Red Hat does, but in a slightly different space, and not very successfully.
A major difference is that Canonical projects have copyright assignment policies, while Red Hat projects don't - this probably explains a lot of the difference in adoption dynamics.
The topic is not whether snaps are avoidable or not, but the Ubuntu is going downhill. And snaps are purported to be part of that downhill, which would be Ubuntu's NIH syndrome. As far as I know, Ubuntu's only successful development is Ubuntu itself - the other projects have all failed over the years, and snap, while ongoing, is not winning any popularity contests either.