This is misleading at best. "verified" only means that Flathub knows who did the upload. It doesn't tell you anything about the safety of the app, and it doesn't mean the app is affiliated with who you think it is.
There is no such thing as a “package”. It is a repository of binary data with references to data in it (ala git). The whole repo and all data is gpg signed.
> ostree show flathub:runtime/org.kde.Platform/x86_64/6.6
commit a7443e846cf67d007fcecda5c9dc27844001cfb8929064395cfc25c6d71d9474
Parent: 23107550082daf3b2892a4a0db2543838578ca882340a756b988bc5c1614540c
ContentChecksum: 607ba9475d32a24c51509bc7919f5a93d401f8f7198c30ad93ad74051d966c41
Date: 2024-01-30 13:55:08 +0000
build of org.kde.Sdk, Tue Jan 30 11:23:00 UTC 2024 (5998d2f3ef21414d14f066ab91fa44e5aef65b90)
Name: org.kde.Platform
Arch: x86_64
Branch: 6.6
Built with: Flatpak 1.14.4
Found 1 signature:
Signature made Tue 30 Jan 2024 12:21:18 PM CST using RSA key ID 562702E9E3ED7EE8
Good signature from "Flathub Repo Signing Key <flathub@flathub.org>"
Primary key ID 4184DD4D907A7CAE
Key expires Mon 14 Jun 2027 08:19:40 AM CDT
Primary key expires Mon 14 Jun 2027 08:18:56 AM CDT
I still don't understand why a central repository for AppImages exist. The moment you are using a repository (and possibly version management), the format looses its reason to exist.
I don't see how that's true. The main point of AppImage is it 'just works' on any distro. If you have one primary place to distribute them to any distro - it's still meeting AppImage's vision.
No. Appimages are selfcontained and thus useful for archiving software or carrying it around in random ways. Flatpak could do this too but not as easy.
I personally use a few AppImages, but want replace them with Flatpaks. Flatpaks have their own issues, and because I did not want to troubleshoot in case I encounter another issue, just carry on using AppImages for these selected applications. Also I was not able to archive Flatpak easily, its very complicated with keys and not. Compared to it, I just have the AppImages included in my regular backup process with regular files.
My point was not if AppImages are useful (they clearly are and I use them), but was talking bout repositories. However after some other replies I thought about it and indeed such a repository makes sense even for AppImages. I personally just don't have to use them.
Not really. AppImages are as much secure as any other executable you run on your system. If you download it from a trusted source, like you download trusted Flatpaks or your systems repository, then they are not worse. If you say AppImages are highly insecure, because you run executable code, then you have to take that logic to any other executable format. The problem is not the format itself that makes it insecure, it's the source.
I read that page and there is nonsense included too. Just because I read that page does not make it correct. If you think that AppImages insecure, then you did not understand my point that its not the format thats insecure, but the source where you get the files. Every packaging system is insecure if you get it from bad source.
That's not even a question. AppImages are fine and not insecure if you download it from a secure place you trust (like your system packages, you trust your distro maintainer fully). Would you trust every distribution maintainer on every distribution? Let's say a Chinese Linux distribution, that maintains Flatpaks and native packages. Let's say they are flaky. See? It's the source you don't trust, not the file format or packaging system.
Read my replies (just like you said I should read the linked post). And understand the issues.
Appimage is not a neutral packaging format. Of course "an app packaged as .zip is as secure as packages as .tar.gz". But the format causes all the things mentioned in the post.
libraries are often the oldest non-EOL possible to support old kernels
no transparency about used libraries and possible vulnerabilities
no upgrades of libraries, always just the wanted app and then passively also the libraries
no sandboxing without firejail (which is a root binary and thus can lead to privilege escalation of rootless processes if it has a vulnerability which it had in the past)
no GUI sandboxing
even with a repo no cryptographic signature verification like on Android (not sure about Flatpak which uses OSTree)
requires users to execute code in random locations
So it is way less secure than Flatpak, thats a fact. It may not be worse than tarballs, but if those dont include the libraries even less secure than them.
I partly agree. But your tone changed a lot from "highly insecure" to "less secure", which is a complete different statement. An application does not need to be in a sandbox to be secure, so I don't accept that as an argument against trusted applications. I only accept your arguments if we talk about random downloads from random places.
Also the argumentation that AppImages are usually run from random locations doesn't make them unsecure, it's a feature. BTW I have a dedicated folder where I put them, but that's my personal organization. Did you know that you can unpack an AppImage back to its original folder (like an archive)? You need appimagetool for that.
The only thing I fully agree with you and is a weak spot about the AppImage format is, that it can or will include outdated and not updated libraries (or executables). Which is the point of the format on the one hand, but a curse on the other hand. Normally it is recommended to build on the oldest supported LTS Ubuntu, because of older libc. libc is the root of many problems in Linux (for compatibility).
This is a good step but I still feel like it's pretty obscure where a package is actually coming from. "by Google" or for the Steam package "by Valve" is really confusing and makes it sounds like it's coming directly from the company. Unverified tells the user to pay attention but there is no hover over to say what it actually means.
Wait… so the author displayed in “by <author>” is the supposed author of the software, not the one that put it on the store? That’s insane! Also sounds like you’d be open to massive liability since the reputation of the software author will be damaged if somebody publishes malware under their name.
Also maaany packages direct to issuetrackers of projects not supporting that flatpak.
If someone knows where that flathub metadata is stored I would love to know, as the manifest is not it. I would like to fix those to link to their own bugtrackers
Ain't that the truth. I don't know why KDE Discover is so sluggish when it comes to Flatpak, it takes me like 10+ seconds to load the landing page and see the popular apps.
First time I've heard someone call Gnome Software fast. In my experience that app feels like it's on it's last legs; the Flatpak CLI is far better than any desktop GUI.
Gnome Software has received numerous updates over the last few years which make it considerable faster. Searching and viewing apps is now fast enough to be usable, compared to it taking many seconds to minutes for basic tasks.
I've stopped removing Software on every system, altough I'm not usually using it. I've not tested it, but I feel like Discover is now slower than Software.
My main issue with Gnome Software is if I queue something to install, and go back to browse for more apps, once something is done installing it "refreshed" and I lose the spot I was at. Makes me feel I can only install one thing at a time.
It's definitely faster than it used to be. But yeah, searching for app updates is still more sluggish than through the terminal, at least on Fedora Workstation.
Verification doesnt help at all if the source is not trusted. All this says is "upstream developers maintain this package". Unofficial packages can be safe too, like VLC.
It does help prevent actual malware from being downloaded, though, since upstream developers probably won't publish malware on Flathub.
But this is still a half-measure. I don't understand why Red Hat and Canonical don't treat this issue seriously; people on Linux are used to assuming software installed from the repos are safe, and yet Snap and Flatpak are being pushed more and more despite their main repositories being potentially unsafe.
I can't find it now, but I read that the verification process also includes human review (for the initial verification, not every update), so it should actually prevent "verified" malware (though it does nothing against unverified malware).
This unverified badge does not prevent from malware being downloaded. This is a false statement! An upstream developer can have malicious intention and be verified as the upstream developer. This unverified badge only helps identifying its not a modified version by someone else and is guaranteed to be from the original developer. It does not prevent anyone from downloading and installing unverified apps. If that was the goal, then why having unverified apps in the first place on the store? Yes, because its useful. Therefore people will download unverified apps or just blindly trust verified apps.
At the moment his is enough. But if the Flathub store grows, this can be an issue. Look at the Android and ios app stores; there are plenty of apps from original developers with malicious intentions.
It is reasonable to assume that a verified Flatpak will have a lower chance of containing malware, since initial verification includes manual review (by a Flathub maintainer), and certain changes (like default permissions) also require manual review.
So the way I see it, it does help, but not in a meaningful way.
Because both Red Hat and Canonical are of the "pay us to care" mindset. If you aren't paying for support, you're a freeloader and need to do your own research.
That's not entirely true with Red Hat. There's a lot of work that they've done in the open source community that they haven't shared back. And canonical seems to think this is a good idea.
Next step, display the "potential unsafe"-badge next to verified or unverified, that can be found on the same page. In example https://flathub.org/apps/io.github.shiiion.primehack is marked as verified, but if you scroll down you can see the application has full system and data access and is marked as potential unsafe.
It makes it obvious to people whether they are downloading Google Chrome as packaged by Google or as by someone else. That being said, Google Chrome is malware. That being said there is a lot more that needs to be done to truly prevent malware, which will be costly but will hopefully take effect when they've got the budget for it
Because if you search Firefox and see a badge that says verified, you can be confident that it was Mozilla that packaged it and added it to FlatHub as opposed to some random scammer.
