#1741 F27 System Wide Change: Graphical Applications as Flatpaks
Closed: Fixed 2 years ago Opened 2 years ago by jreznik.

I would like to ask FESCo to review this Change proposal:

This change is to enable package maintainers to build Flatpaks of their applications and make those Flatpaks available for installation.


I spoke with @otaylor about this yesterday at a conference, and I'm inclined to vote +1 on this, but I'm a little concerned about the impact on Release Engineering. If they're OK with the proposed changes, I'm wholeheartedly +1 for this change... I'd just like to gauge their reaction first before voting.

Metadata Update from @jforbes:
- Issue tagged with: meeting

2 years ago
  • AGREED: F27 System Wide Change: Graphical Applications as Flatpak
    deferred a week for further discussion (jforbes, 16:17:00)

Please consider the opposition in the mailing list thread and also the long-term implications (see the plan the initiators posted to the thread) before giving this a blanket "+1" vote!

Oh god no.

1) Steam has realised what a mistake this is - and are actively trying to move away from the "Steam Runtime" model.

2) The security implications are massive. Flatpak is the perfect method to have out of date libraries that will never / rarely get changed and open up the user to a much wider vulnerability vector than just using the libraries the OS provides via DNF.

Why the hell do we want to repeat those mistakes (and the ones that every other packaging system has had years to overcome) at an OS level?

Sorry, but this is a seriously silly move.

+1.

Third-party developers will see that Flatpak is a viable way of shipping applications, as they can reach a much wider user base than by packaging only for the major distributions.

I believe this change is essential for shaping a better future of Linux on desktops and Fedora, as a leading-edge distro, should take the first step.

Third-party developers will see that Flatpak is a viable way of shipping applications, as they can reach a much wider user base than by packaging only for the major distributions.

And that helps Fedora users how? It only hurts them, by taking away native packages from them in favor of much less efficient Flatpaks.

And eventually, if the Flatpak approach succeeds, the distro becomes completely interchangeable, so why would you then use Fedora over any random one-man-show distribution on Distrowatch? So it also hurts Fedora as a distro.

Fedora users will benefit from all the Flatpak advantages including sandboxing and shared runtimes. Also, distro upgrades will never break applications. These advantages alone shows that users aren't going to get "hurt" and they are not receiving a less efficient product - it's completely the opposite. Flatpak is developed exactly to solve the efficiency issues with the default software shipping method.

Distros are already interchangeable, with or without Flatpak. There's nothing you run on Fedora that you can't make run on any random one-man-show distribution, because they're all Linux. People aren't choosing Fedora or any other distro because they're "locked in" - we're not talking about proprietary ecosystems here. There's a lot of factors that play a role in the process of choosing a distro (release/support cycle, out-of-the-box experience, distro's main purpose...) and "exclusive software" isn't one of them.

Sorry, but this does hurt users. The whole point of distro provided updates is that one update covers all programs linked to that library.

When you divorce from that process - every flatpak needs to be updated as an individual process. This is doubled-down when you have a runtime base as well. On worst case, this would mean 3 different places the library could exist - all with different versions, and all with different update procedures.

It would be short sighted to assume that this would be an ideal solution - or offer any benefits at all to Fedora users.

Shared runtimes already exist - they're called system libraries - and Fedora uses them to great effect. The sandboxing is the security level of a wet paper bag. Lets not try to shovel BS and call it air freshener.

Indeed, the word "shared" being applied to Flatpak runtimes (in the plural!) is really funny, because having multiple runtimes means libraries are NOT shared between the runtimes, and because the runtime also does NOT share libraries with the underlying distribution.

I also think that the huge software selection available in the repositories of distributions such as Debian or Fedora is a major factor in selecting a distribution. Sure, theoretically, you can compile anything on any distro, but not many users want to compile all their software from source. (Some do, and are happy with Gentoo or AUR, but they are the minority.) Compiling is slow, has a higher risk of errors than installing a prebuilt package, and wastes resources on every computer rather than on a central build system. If it were that easy to compile everything from source, we would not have this discussion about Flatpaks at all. So it really matters how many packages are available for each distribution.

I agree that having only one version of a library is better than two. But things aren't that perfect and seamless to assume that this is the best solution, if they were, there would be no need of developing Flatpak and projects like Fedora Modularity in the first place.

There's a need to solve real world issues with the way software is distributed on Linux. We need to see what's actually happening. We have conservative distros like RHEL and Debian, and more leading-edge distros like Fedora or openSUSE Tumbleweed. With the conservative ones, upgrades are constantly being delayed, because they depend on libraries that haven't been updated to implement a certain feature. With the leading-edge, constant library upgrades may break existing software.

Developers being forced to delay upgrades or to constantly change their implementations in order to keep their software working. Users getting older versions or having headaches because something stopped working correctly. How this is good for anyone? The goal of using Flatpak is to eliminate these problems by allowing developers to choose what dependences works best with their software, and allowing users to use the latest version available without having to make possibly risky system-wide changes in order to get something working.

You're aware that compiling from source isn't easy or eficient and how a big software selection is good, kkofler. And me too!

Flatpak is trying to bring even more software because it allows developers to see Linux as one platform, regardless of the distribution type or version. They will feel motivated because they will know that their software will work the way they intended everywhere, and all users will benefit from timely upgrades, no matter what distro they choose.

When you're against that, you're explicitly saying that you're not concerned about the Linux growth as a desktop platform, instead you're only caring about your distribution(s) of choice and how it can boost software availability, hurting anyone, especially the users that choose something out of your scope.

If you replaced Fedora and Linux with Microsoft and Windows, we would be laughing our heads of at how much of a ridiculous notion this is.

@ichvin Nobody here is explicitly against making it easier for developers to distribute their software to Fedora users. However, developers of a particular project have often a rather limited world view: they usually care only about their project working well. We distribution developers must have a wider view and consider all interdependent software that we distribute as Fedora. So, while I'm willing to allow alternate software distribution formats as long as they're well-integrated with Fedora infrastructure (built from Fedora-distributed sources) and user tools (installable using the same tools as other formats), I'm definitely against making them the only or even primary format any time soon.

@rathann - I agree. One thing that would be really useful is a flatpak of a Fedora devel environment - like a snapshot of mock / rpmbuild system that would allow developers to throw source code in one end, and have RPMs come out of another.

This would be orders of magnitude more useful for external developers that find it too hard to make Fedora packages (but somehow can work out how to make flatpaks) than what is being proposed.

FESCo Meeting 2017-07-21

  * AGREED: - APPROVED: F27 System Wide Change: Graphical Applications
    as Flatpaks (+1:6, +0:0, -1:0) (Includes a +1 in ticket from absent
    FESCo member)  (maxamillion, 16:34:38)

Metadata Update from @maxamillion:
- Issue untagged with: meeting

2 years ago

This is a bad descision!

You give up a perfectly managed package system, that covers all use-cases of a operating system, for what ?

You really want to end up in a mess, where parallel package manager do not recognizes the requirements and dependencies of the other one. You really want to have a situation, we hate on Windows ?

I was working on packaging on unix for many years and i can imaging, the administrative overhead you have in the fedora project. But anyway, your work guaranteed, that the installation on systems are predictable and reliable. The performance overhead on the installations side is really acceptable. There is no reason for to start this risky b-shit.

If Fedora starts to support bshit-bingo design, then thank you for your past professional work. I am not willing to play the guinea pigs of a mad scientist.

Good bye!

oh...

i have overseen to inform you, that i used to document incidents under bugzilla. I tracked back issues from the backend server and GUI related topics we discussed there in order to fix bugs also in the UI side.

How do you think we can go on in future, iff the committers are not able to
discuss the experiences they made with fixes as rpm-updates, that could be
installed, tested an rated.

We are losing a great mean to control software updates, you will loose quality.
I remeber lots of discussion when we had to fix networkmanager issues.
And especially this topic is a case example, why the base-os can not be
separated from the user-interface.

You are doing wrong! Stop the flat-braining!

I have no problems with the tools being available. The part I have a problem is this:

"If an application a user want to install is available as a flatpak on registry.fedoraproject.org, it will be transparently be installed rather than installing the RPM."

That idea needs to die now - before we put any code into it. Flatpak should never be a replacement for Fedora provided software. Sure - for people who want to distro their own software as a flatpak for compatibility with all distros - go ahead! But prefer flatpak format over RPM packages is 'pants on head retarded'.

Yeah,

it reminds me a bit of using npm as a ugly sidecar of rpm, that i prefer.
We had a discussion of the fancy way to develop and lean-install software.
The discussion changed immediately, after i asked for security, and the
discussion became more seriously sick.

Well RPM is wonderful, i get:

  • predictable and reliable installations
  • selinux adjustements
  • via rpm -qf /.... configgile a mean to tell a developer on bugzilla.redhat.com , what went wrong.

I future, i will have shitty unpacked tar-ball. Why do we relinquish a useful mean to make a good
qm tool, to make a realiable distro even better.

How should i document in bugzilla in future a malfunction, if i can not assign a file to a concrete package and version. "heya there is a config error in file, but i cannot tell you, which hillybilly supplier is responsible for that. gimme help now!"

Please refrain from using profanity and disparaging others work. That is not welcome behavior in the Fedora community.

Your technical points have been noted. This ticket should have been closed last week, and I am doing so now.

Metadata Update from @jwboyer:
- Issue close_status updated to: Fixed
- Issue status updated to: Closed (was: Open)

2 years ago

People only start getting profane when you piss them off.

1) It is not my intention to blame somebody, so sorry if you felt p....off.

2) i was surprised of your decision. It reminds me a bit of the "Hitchhikers guide to the galaxy, when an alien civilisation decided to blow up the earth to make place for an interstellar highway. Well mankind had some years time to send an objection to a planet that was lightyear away from earth. When heard from your decision, i was looking for a towel ;-)
Why did'nt you asked the users ?

3) Mr jwboyer, i looks like to me, that you have a misunderstanding of how and who is using your distribution. I can only speak for my motivation, but i am sure that i am not alone when i say, that i preferred Fedora over Ubuntu/snap and other rough managed kiss distros, because i wanted especially the pendantic and accurate work of that package-managers. You are going to kill the unique selling point of Fedora and Redhat.

4) If i would prefer flatpack or any similar technique, i would chose another distro. I would never waste my time i a protestant church, waiting in the wings to see it becomes catholic. It would be easier to change the building ;-)

Login to comment on this ticket.

Metadata