Immutability and the Future Direction of Fedora- (Will Fedora eventually become immutable?)

Greetings Fedora Linux Developer Team and Community-

Hello there, and I hope everyone is doing well. I wanted to post in this forum with the intention of inquiring about the current state and future roadmap of the Fedora project and the Red Hat Enterprise Linux ecosystem. To provide some initial context, I am currently considering the transition to using Linux and Fedora as my main desktop operating system, due to its inherent advantages of increased flexibility, modularity, and the project’s ongoing commitment towards leading adoption and innovation of new technologies in the Linux space. I have recently been looking around in the Linux ecosystem and have been interacting with the general community. Recently, I became aware of the introduction and increasing adoption of immutable Linux distributions within the general desktop PC space. Being aware that the Fedora project has been one of the earliest adopters of this new paradigm with established offerings of Fedora Silverblue and Kinoite, I had some questions that I wanted to ask:

  1. Since the Fedora project has been offering immutable spins of the traditional Fedora releases for some time now, are there any current or future plans to gradually and fully replace the traditional, classic spins of Fedora with only immutable versions in the project’s roadmap? TL;DR- Are there any plans for an eventual and complete discontinuation of the classic versions of Fedora Workstation and Spins in favor of the immutable versions in the long-term?
  2. Are there any compromises in control over the system and/or flexibility for the end user of any of the current immutable Fedora spins using RPM-OSTree? (i.e. the ability to use native RPMs and access to the underlying DNF package manager, ability to interact with and modify certain components of the underlying core system, end user customization, etc.)
  3. What are the general benefits, in summary, of the approach to immutability and the atomic update model as seen and implmented by the Fedora project?
  4. What are the future plans regarding the Fedora project as a whole in terms of the general direction that Red Hat and IBM would like to go?

I understand that there are many benefits to immutability for general desktops such as providing a basis for better stability, simplification of updates, and the prevention of accidental breakage by the end user.

I mainly wanted to ask these aforementioned questions because I have been hearing a lot of rumors and mixed opinions on the topic of immutable distributions and the future direction of both the general desktop Linux and enterprise Linux ecosystems, and I wanted to seek some more clearer answers regarding the future plans and roadmap of Fedora Linux. I welcome any input from anyone that is a part of either the main Fedora development team or general community, or any input from any members of Red Hat or IBM, if any of them are hanging around here as well.
Also, I would like to thank all of the Fedora developers, contributors, and community for all the work that has been put into this excellent Linux operating system and I hope that the project will stay around for years to come.

Thank you so much in advance and I hope this post finds anyone who reads this well.

-DTLegit

In regards to your #1 question, I’m not sure about “total replacement”, but Immutable being the majority of Fedora Systems is an Objective of Fedora Strategy 2028.

1 Like
  1. No: Fedora is upstream of RHEL and unless RHEL plans to go all immutable, then Fedora will continue with the RPM based versions. RHs primary target is businesses so they have to address the needs of businesses and as someone who does enterprise Linux support, I don’t see that happening.
  2. Yes: I was using Kinoite for a while on my work laptop. I really liked it, but the biggest issue for me was libvirt. My normal setup for work has the “Virtualization” package group installed and then I use Virtual Machine Manager (VMM) to interact with it and spin up VMs as needed. Installing libvirt in Toolbox or layering the packages via rpm-ostree both have positives and negatives, but in the end it was always getting in my way and I always had issues with it. I’m sure other people have had similar experiences with other tools as well.
  3. The benefits I see:
  • Easy to upgrade to a new release
  • Just as easy to roll back
  • Easy to try a different desktop, like switching between Silberblue, Kinoite, etc.
  • Having all the apps as Flatpacks means all their data is in a uniform location
  • A user could customize an image, push it to a repo, and have all their devices using the same image.
  1. This is really all up to IBM, and from my experience they don’t really take risks or try to push the envelope. With all the new AI stuff Apple and Microsoft are pushing to be deeply integrated into their desktop operating systems, I could see businesses looking for a suitable alternative and then maybe a Redhat Atomic Desktop will be rise to the occasion.
1 Like

You’re on the Fedora forums so you’ll get answers from the community here, not from Red Hat or IBM. If you want their opinions, you should ask them directly. I’m a Red Hat employee but I can’t speak for either Red Hat or IBM. I can only speak as a Fedora community member.

Fedora image based variants (Atomic Desktops, CoreOS, IoT, bootc, etc.) are all built from Fedora packages. So by definition, they can not replace other Fedora variants, because those will likely keep existing as long as Fedora packages exists. Image based variants also don’t need to replace the other ones to be successful (and I would argue that we already are kind of successful).

5 Likes

Red Hat has Image Mode for RHEL in Technology Preview: Image mode for Red Hat Enterprise Linux Overview | Red Hat Developer

The upstream is the bootable containers initiative in Fedora: Getting Started with Bootable Containers :: Fedora Docs

6 Likes

Mind sharing your issues? I use this combo nearly daily, layering the packages, and everything works fine.

I also know siosm has a custom image just for this purpose, Kinoite with libvirt.

I think it would be great if we could have libvirt images, as they would reduce the burden a lot.


Btw should this be moved to Ask Fedora ?

Sort of. I had long expected that the Silverblue codename would eventually be retired and that it would become the new Fedora Workstation. However, as of a few years ago, I’m no longer very confident of that. Red Hat has mostly stopped working on Silverblue. I’m not sure what it would take for Silverblue to become the next Fedora Workstation.

Well you certainly cannot use dnf.

I work for Red Hat. I think Red Hat doesn’t seem very interested in Silverblue anymore. Maybe I’m wrong, or maybe this could change. I don’t know.

I doubt that IBM even knows what Silverblue is.

It’s probably up to the Silverblue contributors to decide where to go from here.

2 Likes

I’d like to add that I’m not an IBM or RedHat employee. I used to work for another company till they sold my department to IBM. I recognized IBM for what it is; a chop-shop, and after a year or so I left for a new job. A few months after I left they closed down my old office and let everyone go. If you work in enterprise Linux support you will eventually hear the phrase “IBM: It’s where good software goes to die”. I don’t like to sound mean or cynical, but I don’t see IBM as the company taking the charge in marketing an immutable Linux desktop to businesses. For now it’ll be kept as a “fun experiment”.

4 Likes

For what it’s worth, they’ve been surprisingly hands-off with Red Hat, even this far into the acquisition. I’ve heard many horror stories like yours, but so far this has been something different. I think there might be some pressures over in the more business-function side of the house (particularly with finances and compliance stuff), but I don’t really know much about that. From the engineering side, basically the pressures I see are exactly the same as Red Hat had as a publicly-traded company. With the possible exception that IBM’s big bet on AI-as-the-future definitely comes down to us — but that could have happened anyway, and it isn’t really dictating specifics: it’s basically “Red Hat should come up with something awesome in this space!”

Really, all of the “IBM is calling the shots for RH’s decisions!” messages anywhere, whether random forums, youtubers, news articles (I’m lookin’ at you, The Register) are uninformed name-calling.

And then, one step closer: as Michael and Timothée say, Red Hat doesn’t call the shots for Fedora either. (Even those of us who are paid to work on Fedora have it in our employment agreements that we can put the community’s best interests first.) But, of course, since Red Hat sponsors a lot of work, where they put resources makes a difference. I’m excited about the bootable containers / image mode thing, because that is major work in the direction we need to go to make Silverblue and the other Atomic Desktops primary.

13 Likes

I’m excited about the bootable containers / image mode thing, because that is major work in the direction we need to go to make Silverblue and the other Atomic Desktops primary.

Hello Matthew! It’s really nice to meet you. From looking at the previous replies to the post and at yours, what I have gleaned is that the Fedora Project’s ultimate direction seems to be eventually making the immutable editions of Fedora the primary (or flagship) offering. However, you guys still plan on still offering the traditional/classic versions alongside due to the immutable versions still relying on the traditional packages and technology of the classic editions of Fedora in order to build. Is this observation correct?

It’s abstraction.

If something breaks and causes Fedora not to boot, you can roll it back on a whim without any insight to what the issue was (and optionally figure it out later). Without immutable, you’re forced to figure out what went wrong :stuck_out_tongue:

Some scenarios favor that, and I guess that isn’t compromising or preventing anything. I’m not for easy-mode and haven’t felt the need to try an immutable distro :stuck_out_tongue: but mainly I don’t want to change my workflow to take immutable-specific stuff into account as long as there’s no need to (I like being able to distro-hop without distro-specific stuff). I’m confident in Linux troubleshooting enough to not feel like I need immutable to solve unforeseeable problems.

I expect we’ll keep offering options like that for a long time, but we’ll de-emphasize them. Alternately, we may have some option to “flatten” an image-based installation into the “traditional” kind.

3 Likes