Rant /
gonna be honest here, tying software versions to the OS in LTS releases is the dumbest idea ever.
Just because I want a stable OS doesn’t mean I want to be stuck with outdated apps for years. It’s not the 70s anymore, having duplicate dependencies isn’t gonna make my computer die.
Their solution? Flatpak! Downloading gigs of useless bullshit to run a 30mb app. Make it make sense!
/
If they’re gonna go ahead and call a release LTS, shouldn’t they actually support it? It’s a stable API target. Just keep building new versions against that target and releasing them.
That would simply push the burden to software devs, as they’d need to make sure their app still works using several years old dependencies.
The solution is the separation of The OS space and user installed software. Microsoft solution of letting each app bring it’s own dependencies, while only providing some basic shared libraries was IMHO the best solution. The disk space wasn’t a problem anymore so a few gigs of duplicate files across hundred of apps isn’t the end of the world.
Yes, the Microsoft style solution is actually what I had in mind. A set of extremely common libraries that follow a standard LTS-style release schedule (so that breaking changes only occur once in a while rather than constantly). On top of that you’d have all the countless other long tail libraries that individual projects need. The standard libraries would be dynamically linked and the long tail libraries would either be statically linked or bundled with applications similar to how Windows applications bundle DLLs.
Developers should also be free to override a standard library with their own bundled version if they need one that is incompatible with the system’s included version.
This would make Linux a hundred times more usable and less prone to autodestruction ( steam nuking the OS, freaking installing python…).
Appimages are the closest thing we have of working .exe files, but it saw almost no adoption and it’s dying because “not secure enough” or some other dumb reason. How hard is it to convert appimages into msi style installers that would unzip the app on some directory, notify the OS it got a new app, add a shortcut and uninstaller program ? I don’t care if it’s not secure, that’s what antivirus softwares are for. Even then, I basically only install cracked software and I didn’t have had a virus in years.
How hard is it to convert appimages into msi style installers that would unzip the app on some directory, notify the OS it got a new app, add a shortcut and uninstaller program ?
I think we can still have the security, integrity, and convenience advantages we have gained from package managers while still having a better approach to LTS releases, shared libraries, and up-to-date applications.
The problem isn’t that it’s possible in some distros with some tinkering, albeit that warning would scare the shit of someone that is used to computers, but not a Linux expert yet (ex : me), let alone your average normie user that can’t do anything more complex than change their wallpaper. The problem lies in the existence of such a system in itself in the modern day single-user centred computing. I simply can’t fathom a reason for it to be this way other than a holdover from the past, back when Linux was only for companies and servers.
Rant / gonna be honest here, tying software versions to the OS in LTS releases is the dumbest idea ever. Just because I want a stable OS doesn’t mean I want to be stuck with outdated apps for years. It’s not the 70s anymore, having duplicate dependencies isn’t gonna make my computer die.
Their solution? Flatpak! Downloading gigs of useless bullshit to run a 30mb app. Make it make sense! /
If they’re gonna go ahead and call a release LTS, shouldn’t they actually support it? It’s a stable API target. Just keep building new versions against that target and releasing them.
That would simply push the burden to software devs, as they’d need to make sure their app still works using several years old dependencies. The solution is the separation of The OS space and user installed software. Microsoft solution of letting each app bring it’s own dependencies, while only providing some basic shared libraries was IMHO the best solution. The disk space wasn’t a problem anymore so a few gigs of duplicate files across hundred of apps isn’t the end of the world.
Yes, the Microsoft style solution is actually what I had in mind. A set of extremely common libraries that follow a standard LTS-style release schedule (so that breaking changes only occur once in a while rather than constantly). On top of that you’d have all the countless other long tail libraries that individual projects need. The standard libraries would be dynamically linked and the long tail libraries would either be statically linked or bundled with applications similar to how Windows applications bundle DLLs.
Developers should also be free to override a standard library with their own bundled version if they need one that is incompatible with the system’s included version.
This would make Linux a hundred times more usable and less prone to autodestruction ( steam nuking the OS, freaking installing python…).
Appimages are the closest thing we have of working .exe files, but it saw almost no adoption and it’s dying because “not secure enough” or some other dumb reason. How hard is it to convert appimages into msi style installers that would unzip the app on some directory, notify the OS it got a new app, add a shortcut and uninstaller program ? I don’t care if it’s not secure, that’s what antivirus softwares are for. Even then, I basically only install cracked software and I didn’t have had a virus in years.
You described this almost verbatim: https://github.com/ivan-hc/AM
I think we can still have the security, integrity, and convenience advantages we have gained from package managers while still having a better approach to LTS releases, shared libraries, and up-to-date applications.
Alpine Linux allows choosing a repository manually for each package, so it’s possible to have an LTS release but install some packages from the edge repository. https://wiki.alpinelinux.org/wiki/Repositories#Repository_pinning
The problem isn’t that it’s possible in some distros with some tinkering, albeit that warning would scare the shit of someone that is used to computers, but not a Linux expert yet (ex : me), let alone your average normie user that can’t do anything more complex than change their wallpaper. The problem lies in the existence of such a system in itself in the modern day single-user centred computing. I simply can’t fathom a reason for it to be this way other than a holdover from the past, back when Linux was only for companies and servers.