Flatpaks aren’t huge at all. This is a debunked myth. I can’t recommend reading this article enough.
So you only need to use two technologies that add complexity and cost performance (filesystem compression and deduplication) to get to the point where you are still 10+% higher in disk space use? I am not sure your post supports the argument it is trying to make.
Author here. The distro comes with the filesystem compression and deduplication already set up and I don’t need to manage it, so of course I’m going to use it.
Given the cost of storage I have no problems spending a barely noticeable amount of space to use flatpaks given all the problems they solve.
given all the problems they solve.
?
End of text?
What’s the use case where storage is at enough of a premium to matter? None of this is targeting a server where you’re getting silly with optimizing storage, and even the smallest storage on most consumer facing hardware is filled by media one way or another. It straight up doesn’t matter to a reasonable end user. Storage is less than dirt cheap.
Ah yes. The mindset of: I have lots of money to spend on storage, so we shouldn’t care about optimisation for less fortunate users.
No, the mindset that the storage is less than pennies worth and this usage would have to explode massively to even approach negligible.
A device that is affected in any way by a GB of storage space is going to choke on 50 other things way before you get to that.
I have a cheap laptop with a small SSD dual booting Windows. To me, a couple of GB does matter.
Not when the manufacturers solder the storage and mark it up 1000+%. For many devices, 1GB is still worth over $1.
Deduplications comes with flatpak for free. Both systems had filesystem compression, so this one doesn’t count. 10% higher disk space is neglectible on most systems and the containerisation makes it worth it.
Compression often improves performance as it means reading less data from storage. Deduplication, as flatpak uses it, is free.
If I had to suffer only having 600GB of free disk space instead of 640GB of free disk space I’d shoot myself
This is the exact same mentality that’s resulted in the overconsumption and waste that’s currently killing the planet. “Bandwidth is cheap! Diskspace is cheap! May as well be sloppy and wasteful, because resources are cheap.” Sound familiar? It has an impact on real world resource usage; the computer industry alone is driving strip-mining as we try to satisfy demands for more rare elements needed to make computers-
Bandwidth and storage are cheap… if you live in a first-world country. Increasing storage demands drive up real-world crass consumerism to upgrade, upgrade; it allows developers to be lazy and write unoptimized, crap software and distribute web applications packaged up and thinly disguised as desktop apps that consume significants percentages of CPU, memory, and disk at (apparent) idle, as they waste bandwidth polling the network - I’m looking at you, almost every Electron app.
If you think sloppy and wasteful software (flatpack as an example isn’t sloppy, but it is wasteful) isn’t responsible for real world wasteful consumerism, ask yourself why you upgraded your last computer. Was it too slow? Not enough memory? Did you buy a bigger disk because it was pretty?
People bitch about proof-of-work cryptocurrency wasting electricity, and rightly so. But they do it while installing shit 1GB Electron chat programs on their computers, and 70MB calculators on their phones. Which they then upgrade because it’s “too slow,” or because they need “the bigger GBs.” Flatpack and Snap aren’t as bad as Node, but they’re part of the “waste” trend, make no mistake.
I don’t think the article was defending bloated applications. Instead, it was defending Flatpak’s use of storage.
You are right, and I understood that, but the methodology he uses - and therefore the conclusions - is wrong. He tests two virgin installs, adds some applications, and reaches a conclusion. It’s like saying that I watched a baby be born and live until she was five, and so I’ve proven humans live forever. I also want him to confirn that no Flatpack was used for any packages on the Workstation 36 machine; I can’t speak for Fedora, but on Arch AUR there are some packages that depend on Flatpack and will install it because that’s the only way upstream releases it. So you can easily unintentionally end up with Flatpack on your Arch box if you’re not careful.
Let’s see a real-world, used desktop comparison with multiple package upgrade cycles after a year.
I didn’t use any flatpaks on the workstation install. I’m about three years with this setup on 4 computers through multiple OS updates, works great.
I’m sure it works fine; the question is about how disk space usage compares.
@sxan @beta_tester EXACTLY, I am glad SOMEBODY gets it.
But it occupies a freaking crazy amount of space. People do really be on drugs when going with these religious strong stances.
Yeah last time I tried Flatpak it took like 3 apps to completely fill up my laptop’s root partition and use nearly as much space as my Arch install on its own. For some reason they all used a different platform/runtime/whatever they call it. Oh this one uses the latest Gnome 3, this other one the version before, and that other one Gnome 4. Same with KDE apps, they’d also pull different versions of KDE frameworks and Qt versions. How many versions of Gnome and KDE do I need, just run it on whatever’s the latest.
Granted, my fault for not having quite a big enough root partition. But I’m skeptical about the methodology of the article because it doesn’t match real world experience at all, at least for me.
deleted by creator
I still don’t like that flatpak never knows how much it has to download
100MB/80MB
The issue with flat packs is the more you use it, the higher the chance that you get less shared runtimes and the higher the chance of the duplication. And at some points it really does get to awfully ridiculous levels.
A while back, I had run everything I possibly could with Flatpak to the point I’d even make my own Flatpak to try and see how well it would work. Instead of using the AUR. And it worked great for the first little while. I’d installed all of my apps and it was fine, but as I kept using the system, kept installing new apps and not uninstalled the old ones, it really started to build up awfully quick, especially with older apps.
I feel like the usefulness of flatpaks is the inverse parabola, where it’s extremely useful in the center use, but when you go to either side of it, it becomes less and less useful.
Apologies for any incoherentcy this was written with a speech 2 text.
Maybe I’m in the wrong here but I would think focusing on management time for Flatpak vs whatever would be the important part, not disk space usage.
deleted by creator
My premise is that sysadmin/user time is more expensive than drive space. Seeing some real world examples of how Flatpak could save time over the long run would probably be beneficial for increasing usage.
Keep in mind I have no dog in this fight, I don’t have a preference of one over the other. I only made that comment because everytime I see a Flatpak reference on the web it’s always in the context of disk usage.
deleted by creator
If I was running business workstations on Linux, I’d probably prefer Flatpak over distro specific package managers most of the time.
I’m the author of the blog post and a former sysadmin, there’s really no maintenance to do with flatpaks, not having to deal with traditional package manager issues have removed that problem completely from my life.
Distros may or may not provide this functionality, but on my systems they’re set up for zero maintenance of the OS base image and the flatpaks via service units and then I don’t have to do anything.
My take: native packages for the core OS, flatpaks for desktop applications. Works for me.
I only resort to using container formats if native packages aren’t available.
Fair enough, but I like the fact that I can keep Firefox or Steam from accessing my bank records and holiday photos.
I don’t really care but I have a 512GB drive, a few extra GB of NVidia packages or whatever means nothing. I just enjoy the containerization and not having to give it my root password to install things. I’m not on an immutable distro and not having an app invade my core system (in whatever way the packager felt necessary) feels really good.
I’m watching the immutable space though, once it matures a bit more might try it. openSuse has an elegant and simple take on it with BTRFS snapshots.
I’m curious on what would happen if you installed lots more applications. If it started at a 3.8GB disadvantage but narrowed to 1.2GB after installing a bunch of apps. Most serious systems will have lots of apps installed and a decent amount of storage, at least 100GB.
What it means is that you’re getting the libs the program uses with the program instead of using the system libs, this defeats the whole point of shared memory and wastes RAM, it is inefficient but saves them from having to compile for each distro, still, the system loader has to resolve and load these making loading slower, if they had to include the libs, a better way to do it is to simply compile the binary as a static binary with all the libs compiled in, at least that way it saves the loader overhead.
I still hate flatpaks a tiny bit.
Why?
Due to the big size of all the packages and package dependencies (which also need regular updates) and yet another package manager. I already have APT on Linux Mint, I do not want Flatpak (and also not snap).
i honestly havent noticed a diffrence since i switched to a flatpack heavy distro (normal fedora to silverblue)
I am confused a little. The space taken by silver blue increased by around 3 gigs but the space taken by workstation increased by 4 gigs. So flatpak actually resulted in less space being used the apps. Is this some sort of faulty space usage reporting¿?
deleted by creator
Are runtimes from the fedora flatpaks remote used by apps that are downloaded from flathub¿? Because the gnome apps that are installed by default in silver blue are from fedora flatpak remote so I am guessing the runtimes are also from the same remote
They sure are huge on my system and spread their shit over half the file systems. Firefux is a complete disaster now that it is flatpack.
Well yes but I am not sure that this is the main problem with flatpak containers.
I’d rather point out that this approach creates a bigger attack surface since the containers tend to ship with outdated versions of libraries, frameworks and tools that the actual application relies on because it is now that specific app developer’s problem to update them inside of the container. So with this, even an up to date system is not really up to date and might suffer from severe vulnerabilities. I’d say it depends on your application, use case and threat scenario; containerization can make sense but is not the holy grail.