Long Live RHEL Clones

Red Hat's decision to stop publishing RHEL source code to git.centos.org really rocked the community boat. Instead, they want the Enterprise Linux community to build downstream of CentOS Stream, which does not get 100% of the RHEL source code added to it. I have spent a lot of time reading opinions from all sides and then some. I have deliberated a lot. I have experimented with running various distributions to replace the RHEL 9 clones I'm running today.

My decision is to stick with RHEL clones for the foreseeable future. I am running Alma Linux and Oracle Linux on Raspberry Pi (RPi) and in the public cloud. I have swapped Oracle Linux with Rocky Linux and keeping Alma as is.

I see merit in Alma's decision to be compatible with RHEL by building on top of CentOS Stream, just like Red Hat wants. Alma wants to remove the barriers of the famous "1:1 bug compatability" by aiming to be "only" binary compatible.

On the other hand Rocky continues to "maintain our commitment to 1-to-1 compatibility with RHEL for all EL8 and EL9 releases." I still don't know how they will continue this way for too long, given Red Hat appears motivated to end all 1:1 clones asap, but this vigorous defense of their founding mission is exactly the commitment I'm looking for.

I don't see merit in the explanations Red Hat has given for their decision. They said that "1:1 bug compatible" was merely a marketing term and not really true. But everyone understood it to mean that you got something really, really close to the current release of RHEL without being RHEL. Red Hat themselves say CentOS Stream is not really, really close to current release of RHEL. I need something exactly like that: really, really close to RHEL without being RHEL. They said downstream rebuilders were not adding value. But does RHEL install on RPi with an easy to use image? No. Does Alma and Rocky and Oracle Linux install on RPi with an easy to use image? Yes. That's value enough for me. I don't need support for my public cloud instances so I don't need to pay for RHEL. Alma and Rocky provide that value to me. They said that downstream rebuilders are dangerous for the open source community because they are freeloading on the massive investment Red Hat makes in building RHEL. Isn't RHEL made from the unpaid work of hundreds if not thousands of non-Red Hat people writing or contributing to open source code? Does Red Hat compensate them fairly for their work and genius? Are Red Hat not the same freeloaders for monetizing those people's massive investment for their own benefit? When people contribute to open source they do so for many reasons but the primary aim is to make a difference in everyone's life. Up to some point Red Hat appeared to be on the same mission but now something has changed.

I've worked in the industry for many years. More recently in enterprises I've seen RHEL and Ubuntu side by side. I consider both to be Enterprise Linuxes. RHEL clone of yesteryears, CentOS, has been popular at many places where I worked. SUSE Linux Enterprise Server (SLES) was used at one place and I liked it a lot. Some places used Debian in production as well. But when it comes to regulated environments it's always RHEL in my experience. This spills over in commercial environments too but there Ubuntu has been equally successful.

I was a FreeBSD user for personal things for years. I still like it. But I decided a while ago to switch back to Linux and I'm not changing that decision. That leaves me in a bind.

I love the idea of using a community distribution like Debian. Is there any other stable server-oriented community distribution? Most are forks of others or heavily sponsored by enterprises (see Ubuntu, openSUSE, and Fedora). The world of open source has changed tremendously and corporations are heavily invested in it. One cannot move away from their influence in the Linux ecosystem either. So while I would love to be a community-only cheerleader I have experienced the massive contributions of enterprises to such an extent that I prefer the experience they provide. I'm at a stage in my life that I prefer a distribution that has one way of doing things; opinionated wins over impartial. While Debian is impartial it also reflects in its user experience. For this reason I prefer a cohesive and "crisp" user experience, such as provided by Ubuntu and Fedora.

The next thing on my list is a dsitribution that supports RPi and is available on tier-1 and tier-2 public cloud providers. Here Ubuntu, Debian, and Fedora win hands down. They are everywhere and they support RPi. More recently Alma and Rocky have also shown up in more public clouds. openSUSE has been the toughest to find, although my current provider, Linode, supports it very well.

I want a stable distribution, once that I don't have to re-learn every couple years. Ubuntu is great and all but they experiment too much and one LTS release can be vastly different from the next. After spending years with FreeBSD, where the basics are stable and new changes are additive, Ubuntu is too much like a toddler. RHEL also has changes between major releases but recently they have a more settled base.

SUSE has introduced their Adaptable Linux Platform (ALP) concept. It will take years to become stable. I'm pretty sure their customers will not switch to ALP from SLES until at least five or more years after it stabilizes. ALP has already affected the plans for openSUSE Leap. While Leap 15.6 is planned which will be supported until end of 2025, a Leap equivalent for ALP also needs to be built. Although I love being part of the community, this is too much churn for me to keep up with in an otherwise busy life. I can revisit the openSUSE option in a couple years while wishing the community the best.

Aren't the recent RHEL changes also disruptive and causing churn? For sure. But as I said, in the workplace RHEL is a staple and I get paid to keep up with it. That means I can keep my RHEL skills current at work. When I bring the same skills home I don't have the context switching that occurs with using a different distribution.

Ubuntu became popular almost immediately after it was released. Even today it's the most popular distribution with developers especially in the workplace. This gave Ubuntu a way to sneak into production workloads on-premises and in the public cloud when it became popular. But Ubuntu has not displaced RHEL which is more time the default that teams reach for. In my case I use RHEL at home because it's popular at work.

I think that sums up my reasons for sticking with the RHEL ecosystem.

If Alma is successful over the next couple years in continuing to be binary compatible while innovating in some areas then that must be celebrated. If Rocky is successful in the same time frame with being one-to-one compatible that will be very surprising. Add to it the recently announced Open Enterprise Linux Association (OpenELA). It looks like it'll be an equivalent to CentOS Stream that SUSE, Oracle, and CIQ can use to build their downstream compatible distributions. But how compatible -- binary or one-to-one -- will that code base be with RHEL 10 and beyond is anyone's guess.

If Rocky fail in the next two years to stay true to their founding mission and have to fall back to Alma's strategy then will/can Oracle Linux keep going with the same mission? Remember, Oracle Linux is not available in tier-2 public clouds today (I wonder why?). So it won't fulfill all my use cases.

If OpenELA is successful in being as close to RHEL as possible then that too works for me if they support my use cases.

In conclusion, I think Alma and Rocky are the least disruptive choice for me. Ubuntu will be a context switch from RHEL at my day job. openSUSE has its own churn to sort out. OpenELA will take time to provide tangible value. These are very interesting times in the Enterprise Linux ecosystem for sure. I'm here for the joyride.