Maybe true for proprietary software, but for FLOSS it is a non issue once your software has traction and gets picked up by distributions. I wrote a piece of open source in ~2002 and it is still found in pretty much all distros AND I never spend a minute in pain over packaging difficulties.
IMHO, the basic package managers work REALLY well for FLOSS. The gave me a peak into the AppStore/PlayStore(tm) experience (but then without all the spyware), waaaaay before they even existed.
I'm not sure though "almost identical packaging work" is accurate, maybe that applies to Debian and Fedora (and their derivatives), but Gentoo and Nix are examples of doing something quite different (not to mention the different embedded distros), and so the "duplicate work" is anything but that.
> it would solve one of Linux's biggest pain points as a software developer.
Not the pain of the distro-people. That may be mitigated with a universal standard of snap or flatpak.
At the same time snap packages have serious downsides for end users.
Then 99% went with Flatpak, except one, Ubuntu with their own proprietary solution called snap.
It's not the same thing. We have a decent (yeah not perfect, I'll spare commenters the effort of posting that same old tired flatpak rebuttal) cross distro packaging solution, but Canonical always wants to play different and not collaborate with the other standards. So here we are.
In 10 years the "Ubuntu Store" will be a pretty frontend on top of Flatpak just like Gnome Software, and just like every NIH Canonical product made before it.
Could you briefly summarize those for us?
AUR is explicitly a community-driven effort. The others are packing standards, some of which (like DEB and RPM) are also used in community-driven efforts. Debian and RedHat derivatives have repositories that use DEB and RPM, respectively. There's not a lot of bike-shedding, here. Availability of functional software is a pretty big feature of Linux distributions.
And there is a big difference between those community efforts and the likes of Snap, AppImage, and FlatPack, which are developer-centric. User preferences on software distribution aren't always or even usually based on irrelvent arguments, but fundamental ones. The community efforts centralize dealing with compatibility and, to an extent, security with community standards, while the developer-centric model largely leaves this work to software developers alone. This is a meaningful distinction.
DEB vs RPM is an ancient debate, each comes with its pros and cons. Flatpak and Snap try to solve a different problem (sandboxing + dependency hell vs stable system-wide dependencies).
MacOS has PKG, but most software I've seen actually comes through DMG images or ZIP files. Windows has MSI, which, if you use it correctly, actually elegantly solves most problems; the features that were added to allow developers to bypass the declarative installation process are what usually breaks MSI installers.
But then modern Windows now features has packages installed through .appx, .appxbundle, .msix, .msixbundle, and plain old .exe. Windows Mobile (based on Windows CE) had .cab; Windows Phone 8/8.1 (based on Windows 8) had .xap; Windows 10 Mobile had .appx. Now Windows 11 also supports .apk files from Android!
In the end, the package method doesn't matter. Users don't care where the package comes from, they just want to install the application and run it. Deb, Flatpak, RPM and Snap all work fine from tools like Discover or GNOME Software; it really doesn't matter to normal users what kind of packaging system they use.
AppImage and Flatpak are fine, and AppImage is the first choice amongst this "everything included" bundle formats.
However, Ubuntu is forcing snaps as a power move. Both the software, and the treatment from Canonical is off-putting a lot of enthusiasts right now.
I know what I’m saying can be regarded as “+1 different standards”, but Nix is the only truly novel approach to package management that can reliably install a huge-ass behemoth package group like Gnome, install Plasma next to it and remove both without a single garbage file laying around afterwards.
I think we really should embrace this uniquely good solution.
Just ship an ELF binary in an archive. Include any files and libraries you need. Compile against the oldest glibc found in any system you care about supporting. That's all you need.
[0] https://download-installer.cdn.mozilla.net/pub/firefox/relea...
The sooner we all begin treating Ubuntu and Red Hat Enterprise Linux as separate Operating Systems the sooner we can all move on from the "single standard" and "Linux on Desktop" debates. As an application vendor I always find it amusing when customers ask if we support Linux. My response is no, we don't support Linux but we do support Ubuntu 18+ and RHEL 8. Linux distros are so wildly different these days, not just in software but in ideologies, that this distinction is important, and there's nothing wrong with that.
Yeah, the year of the Linux desktop isn't coming, but it's not 2001 either. We've standardised onto a few technologies pretty much everyone adopt except the ones that want to be different: systemd, Flatpak being the most used cross-distro system, pulseaudio (and pipewire offers a 100%-compatible shim for it), dbus, NetworkManager and now finally with GNOME 42/libadwaita we can have even standardised UI design.
And like many other things frontend, things move at a pretty rapid pace. Ubuntu 18 is ancient, might be good for a server, but it's not a good idea to use that as a metric for where the Linux desktop is.
This is why my daily driver isn't Ubuntu (or Debain, because I also hate apt - it also sucks).
MSI on Windows is completely fucked, PKG on Mac has a few footguns, but at least they're universal, decades-old standards supported by mature tooling. You release an MSI/PKG, and you're done. Works on every Windows/Mac system, no issues.
On Linux, an OS with 3% market share, there are more competing standards to count: Deb, RPM, snap, FlatPak, AUR, AppImage and probably a dozen other semi-popular ones. Every individual user has their opinion (and they'll voice it!) as to which standard is the best. At best, this leads to God knows how many man-hours of duplicated work packaging and QAing. At worst, it leads to the dev abandoning the thought of Linux altogether.
Linux on Desktop simply can't move forward by continued bike-shedding over frankly irrelevant details. Even if the only rationale for a standard is "Because Mike Shuttleworth said so, and he got a phone call from Mandela in space", that's a massive improvement over the current status quo.