They can take longer to start up, which can suck on older hardware. It’s not as bad as it used to be though. Once they’re running there shouldn’t really be any difference. The main drawback is actually that Flatpaks use more storage space.
I am glad that the startup times have improved, that bodes well for future startup times. Using up more storage really is what makes it suck for everyone. I thought that it was more efficient, since I see a lot of .platform, and I assumed those are libraries shared across flatpak apps that use those dependencies.
I am almost sure AppImage has the same problem? I don’t know, people do rated that better though.
Storage space mostly isn’t as bad as it is with AppImages. Each AppImage stores all the libraries it needs, even if they are shared with another one. They can’t even know if they have shared libraries. A single AppImage will probably actually use less storage than a single Flatpak if you only have one, just because the AppImage only uses exactly the libraries it needs, while Flatpaks use shared sets of them. That being said, Flatpaks generally get less bad the more of them you use, because of the shared libraries. They’re still a whole extra set of libraries on top of your system ones though, plus they put out a new set every year. Apps that are still under active development generally get updated to the latest version, but older apps that are basically finished often require older libraries, so that’s more space used. Overall for a one off program when you’re not using universal packaging systems regularly AppImages are mostly better, but if you’re going to be using them regularly Flatpak quickly becomes far better. It still uses more storage space than just using native apps though.
Another difference between Flatpak and AppImage is that it can be kind of a pain to theme Flatpaks to match the rest of your system, and I don’t know of any good way to do it with qt6 apps yet, but it’s just straight up impossible to theme AppImages. They can technically have themes built into them, but unless you’re using Adwaita, or maybe Breeze if you’re lucky, they just don’t, and having to rebuild your own custom AppImage completely defeats the main benefit of using AppImages.
That is what I thought, and I was confused about people complaining about the redundancy. Also, every new program I install manually seems to pull a crap-tonne of new dependencies, so nobody is saving space.
There might be an increase in startup time and RAM usage because it loads it’s own dependencies instead of using system libraries, but the difference is probably very little.
I imagine that is the case. I also feel that is a trifle. Unless one is constantly closing and opening an application they use often, the extra seconds starting should not break a workflow.
Sure. I personally have not noticed a difference. Then again, I recently got a new computer, and all my other computers are over a decade old, so everything feels luxurious.
All I remember about my first time is being tricked into using Slackware. They told me it was the easiest distro. And this was in like 94 or 95; just a year or two after the damn thing came out.
Slackware was mine too - all it took was a box of floppy disks and tens of hours of downloading and installing! It was great though, something so different. But it was just a toy, and I went back to DOS/Windows on PC - mainly for the games and hardware support (Voodoo!)
A year or so later I spent a lot of time playing with Solaris and VAX/VMS at University and really developed a love for the command-line and UNIX environment. It was that which led me to my first job (with HP-UX) and my second (Debian/Yellow Dog). From then on I used it at home a lot more. Now I use Windows for games/gamedev, and Ubuntu for everything else (desktop, laptop, servers).
But it’s amazing how far things have come in some respects, but how some things have regressed over those 20 years - window managers/themes never reached the heights I envisioned in the Enlightenment hay day, session management/restoration/remoting seems to have been eroded away, virtual desktops/window management/tiling regressed and became fractured, the wonder of Compiz didn’t really move things in an interesting way, and I felt sure Quicksilver (for MacOS) was the future of launcher, but it’s not really been taken up - though the Expose feature is an excellent essential part of Gnome now (Activities)!
In some ways I think Linux has lost that “wow factor” that we used to have with all those cool features - but it is much more rock-solid and professional now! I use it more now than I ever have.
I’m a /home on separate drive/partition kind of guy. I like it just following my installs. Though seeing some using guix/nixos to create a config for my desktop has got me wanting to spend a weekend trying that out.
The rest, ansible for any sufficiently complex enough setup at the moment. Good for integration work with LDAP, etc if your using that. Again may play around with guix on that front.
I was just thinking yesterday when looking at how NixOS works. The config file seems to be quite reminiscent of an Ansible Playbook. I mean maybe I’m way off the mark, I haven’t really dug into Nix much yet.
Guix/nix seem very powerful. The reproducibility is something ansible just isn’t built to same level robustness for, which makes them seem very promising to me.
I’ve been using chezmoi for dotfile management and have been really happy with it. You can directly import existing files to get started and template out any differences between systems.
I used SW in kvm/qemu VM with GPU passthrough, you can do gpu passthrough on most machines integrated or dedicated gpu, I’ve passthrough part of a integrated intel gpu on a cheap laptop and performance was really good, way better than anything virtualbox can do, if you have and are able to passthrough a dedicated GPU, even better
Flatpaks are great for GUI apps, and have a sandboxing system that allow them to work well on any system that support flatpak. This allows devs to package once run anywhere, saving Dev time! It also has a portals system to allow for better system integration of the granular permissions needed for the app to actually work (nobody wants a truly isolated sandbox for every app).
Snap is less featureful for GUI apps, but work closer to how native packages do. The real issue is the proprietary app store required for it, making non-foss. If you want the same benefits of snap, check out Guix and NixOS both of which have a more cleaner design, and work better IMHO.
@lemminer@fruitywelsh
Appimage is OK but no auto update makes it download and forget type of deal - definitely not for every app
Flatpak - best for me but permissions on some apps make it unusable e.g. gpodder - command for player as flatpak is unable to access MPV installed from repo flatpak etc. - sandboxing (couldn't fix it with flatseal mpv --profile=... not working)
snaps people love to hate them... no love from me :-)
Repo if it works, is available - the best option
I dont understand how redhat is going to police this policy of “we’ll keep source code open to paying customers, but reserve the right to cancel a customer that shares said source”.
Toss in GUID’s or randomly place identity files to anyone that downloads the RHEL source hoping they get accidentally published as an identifying attribute if someone does decide to publish it elsewhere.
And make sure that identifier scheme still works if different people on different subscriptions download the source and compare to filter identifiers like that out…
They could try that but I suspect it would be rather easy to find anomalies like that. These are ultimately patches to an upstream and already open-source project, so one can just diff the RHEL version with the release it’s based on and quickly notice that random GUID in the sources or random spaces/indentation. Or have multiple sources leak the code independently, and then you can diff them all between eachother to verify if you got exactly the same code or if they injected something sneaky to track it, and remove it.
Lots of companies in enterprise also want to host their own mirror because the servers are airgapped, so they can’t even track who downloaded all the sources because many companies will in fact do that. And serving slightly modified but still signed packages sounds like it would be rather computationally expensive to do on the fly, so they can’t exactly add tracking built into the packages of the repos either. And again easy to detect with basic checksumming of the files.
I don’t think that many companies have their shit together well enough to mirror the source code, besides the RHEL repos aren’t small, so that’ll cost.
The companies I’ve helped either had a minimalist mirror to reduce the surface area of what was installable or to save on cost.
It’s possible that a few enterprises do a full mirror of all RHEL sources, but i doubt it’s many
I don’t know, I’ve worked in Debian/Ubuntu companies mostly. Last two had thousands of servers and both had an apt-mirror custom repo including the deb-src ones. Otherwise we just get ourselves banned from the official mirrors when thousands of VMs pull updates from the same NAT IP.
Not sure how that works exactly on the RHEL side, maybe it’s not nearly as easy or common to do that.
This is not about an individual sharing the source. This is about near verbatim copy distributions like Oracle Linux. And they can easily see who contributes code from RHEL into those distributions.
I think Jeff has a point that a Linux distribution is a collective effort, but I honestly don’t see why he can’t just target Fedora which is for all intends and purposes the testing release for RHEL and most of the development work that Red Hat does goes directly into Fedora. RHEL adds little of value to that other than some compliance BS for large companies.
Fedora isn’t the testing distribution for RHEL, CentOS is. Fedora is upstream of CentOS and could be viewed as the bleeding edge in that regard. CentOS used to be downstream of RHEL, but that changed a few years ago when IBM did its first shitty thing at Red Hat. The tree is like:
Fedora (Top of code stream, “unstable” from a business perspective)
|
|
v
CentOS (midstream, much less frequent feature updates)
|
|
v
RHEL (end of stream, stable/predictable/reliable/etc)
And I couldn’t disagree more about RHEL adding little value. You’re not going to run a server on Fedora for something you want/need to rely on, and especially rely on not to change much/cause breaking changes. That’s what RHEL is for and it is the gold standard in that regard.
And that’s not even mentioning the fact that Red Hat support is some of the absolute best in the world. Motherfuckers will write a bespoke kernel module for you if that’s what it takes to fix your issue. Not sure if that’s still true after the IBM takeover though, but that was my experience with them before that.
You can absolutely run important services on Fedora server edition. Most of the stuff in containerized anyways, so having a more up to date version of the base system is actually an advantage.
It is really only those large corps with massive closed source lagacy applications and loads of compliance regulation that need a stale but long term supported distribution like RHEL.
Sounds like you have some investment into hardware and software not really designed with linux in mind. Running windows in a VM could work. There might be issues with graphics though if your VSTs/DAW have a lot of eye candy, as you’ll usually use a virtual GPU with a VM. You could always try a windows VM inside windows whether it works OK, before committing to linux + VM. An alternative could be to have a dual boot setup. You could use linux for day to day things, and reboot into windows to do music production.
This is a great comment. If it’s dabbling then absolutely, have fun! But it’s a real self-administered kick in the nuts to squander serious investment in SW/HW just for bragging rights, cool as they are.
Dual boot is probably the (annoying) answer. Not sure on the efficacy of a windows VM for music production.
I have been watching system76 from afar for a long time and everytime I upgrade I look at their systems but I was never confident of local support. I bought an equivalent to one of their early laptops from a local company once. I think it is great that they are bringing more design in-house as rebadging generic systems limited their documentation and repairability.
While competition is good I can’t look past Framework at the moment. They shipped to me direct from Taiwan as fast as a local delivery and I know I can repair the system so it removes all the concerns I had about dealing with a niche foreign company. I see no value in PopOS or the other user space stuff from system76. Open firmware is an advantage but I think framework will get there eventually. As much as I respect system76s mission I think their business model is dubious. They should have gone in-house open hardware earlier and I think the userspace stuff is a pointless distraction.
I respect that. I do code occasionally and I was only interested in 16:10 or squarer for a laptop. I was very concerned about the high dpi but it has been fine for me.
Ideally I wanted a 14" 16:10 (ideally 1920x1200 so I didn’t need fractional scaling) with a high refresh rate and integrated amd graphics but the expandability and ability to maintain the system myself in a fairly remote area sold me on the compromise and I don’t regret it but it wasn’t my ideal laptop.
Expanding a custom product line is very expensive and will take time compared with slapping a badge on generic machines. The 16" framework with 16:10 aspect and 165hz refresh is going to expand Framwork’s customer base a lot but my ideal is a system that falls in-between the two.
Without an equivalent to the Framework marketplace or a local presence I don’t see myself ever buying a system76 despite looking at them regularly since they started. I bought an ASUS z35fm in 2007 based on what I think was their Darter at the time. They had 16 years to convert me to a sale and it took Framework a year with a better business model.
That’s your opinion and that’s fine. For me, open hardware + hot swappable mechanical keyboard + trackpoint + designed to be repaired + the general opennes are the reasons why Virgo comes to my desk.
Framework is good for many and thanks to LTT will sell more units but I’ve been only with Thinkpads since 2008 so Coreboot + trackpoint are must-haves for me.
I was on windows 10 without an SSD. There is only so much dog slow you can handle before you want to change it. I was amazed by how much faster linux was. Windows 10 was the first windows I noticed that struggled this much. It’s like they gave up on performance and just relied on the hardware. I dual booted from there with linux mint and over time, I started windows less. I haven’t used it in months.
Without saying why you are leaving Fedora, it's not really possible to advise you... whatever we recommend may have the same mystery issue you are trying to escape.
linux
Active
This magazine is from a federated server and may be incomplete. Browse more on the original instance.