@linux Sharing a ‘small’ inconvenience I had to fix with #opensuse #slowroll (I suspect #tumbleweed is the same) - I couldn’t launch snaps (spotify, bitwarden) after update - error was: cannot determine seccomp compiler version in generateSystemKey fork/exec /usr/lib/snapd/snap-seccomp: no such file or directory
The fix (I first tried re-installing, didn’t work) was to:
a. locate snap-seccomp - was in /usr/libexec/snapd
b. symlink: ln -s /usr/libexec/snapd /usr/lib/snapd
Not being able to launch snaps seems like a good feature
Why not flatpak?
@pastermil @linux I use both. There are packages where the website officially lists snap packages, no flatpaks.
Unless the project website has a link/install instruction recommending flatpak, I prefer either the distribution package where available, or snap otherwise - this is more from a supply-chain perspective - since snap requires the original developers of the package to package snaps.
If the developers have officially listed flatpak on their site, that however, is good enough for me.
I would take this with a grain of salt. For me, as long as the package is available and functional for my prefered installation method, I’d go with that.
Take cerbot for example. For some reason, the cerbot developers uses snap in their installation guide. I’ve been using apt on all my projects that requires https, both personal and professional (yes, I get paid to do this, among others). Never had any issue with it.
@pastermil @linux the attack surface for something that isn’t officially maintained by the developers, and that doesn’t have more vetting (e.g. distribution packages) opens up room for malicious actors.
e.g. #arch / #aur recommends verifying scripts manually before installing, and malicious scripts have been found and removed.
There are actors like #jiatan out there. An unofficial #flatpak needs manual verification before install - that’s why I just go with #snap if the flatpak isn’t official
the attack surface for something that isn’t officially maintained by the developers, and that doesn’t have more vetting (e.g. distribution packages) opens up room for malicious actors.
There are actors like #jiatan out there.
Funny that Jia Tan was an official maintainer of xz until he was found to be problematic.
Speaking of verifying, you know you can’t really verify anything on the snap server since they’re proprietary, right? On the contrary everything on flathub is laid to bare for anyone to look at.
In the end, you’re free to choose. Since you’ve kindly provided your argument, I’ve provided mine in hope you’d reconsider.
Are you sure snapcraft requires the original developer publish snaps? This seems unlikely, but they may have updated their policies.
Edit: they aren’t, Signal for example is an unofficial snap not published by the Signal developers but rather “snapcrafters” - https://snapcraft.io/signal-desktop
Snapcraft has hosted multiple malicious applications, so I wouldn’t exactly call it a safe place either.
@Bitrot @linux interesting, thank you for that information: I had been under the impression they did do manual verification of authors.
I did some checking: the closest I found to verification was this (so you’re right- no need to be the original author, but a bit of vetting does seem involved).
https://forum.snapcraft.io/t/manual-review-of-all-new-snap-name-registrations/39440
My takeaway here is to use whatever the authors recommend (assuming trusted authors)!
@thegreybeardofthetree @pastermil @linux
FYI FlatHub uses GitHub Actions, you can check how they build their apps and some of them support reproducible builds, just in case you want to verify GitHub isn’t acting maliciously.
FlatHub and AUR can’t really be compared in terms of security. Flatpak apps also don’t modify the host OS, while AUR packages can.
Personally, I only trust distro packages and FlatHub.
This is why I prefer using Distrobox on my personal computer. No package for Signal-Desktop? No problem, run it through a Debian container using Distrobox.
But snap is cringe
