The killing of CentOS Linux: 'The CentOS board doesn't get to decide what Red Hat engineering teams do'

Brian Exelbierd, Red Hat Liaison and CentOS board member, gives the company perspective


Interview Brian Exelbierd, responsible for Red Hat liaison with the CentOS project and a board member of that project, has told The Register that CentOS Linux is ending because Red Hat simply refused to invest in it.

Early last month Red Hat shocked users of CentOS, a free community build of the same sources that make up the commercial Red Hat Enterprise Linux (RHEL), by revealing that CentOS Linux would cease and be replaced by CentOS Stream, a build of what is likely to be in the next RHEL update. What happened, and how is it possible that the supposedly independent CentOS project conformed to a change of direction that was not driven by the wishes of its own members?

"Red Hat participates in lots of open-source projects and communities. Red Hat sponsors some open-source projects and communities," said Exelbierd. "CentOS is a sponsored project, we are the funding agent and we happen to also be a heavy contributor. We have learned that open-source communities do well with independence. We let those governing bodies govern."

The CentOS board doesn't get to decide what Red Hat engineering teams do

There is a caveat, though, which is that "the CentOS board doesn't get to decide what Red Hat engineering teams do," Exelbierd told us. In this case, the thing that changed was Red Hat's strategy in terms of what the company was willing to sponsor. "Red Hat said, we're going to make some fundamental changes in how we direct our investment. Then we went to the CentOS project and said, here is a thing Red Hat is going to do. We believe there are consequences of this action... the end result was the decision that got made by the project," he said.

What was the specific change? It seems that Red Hat said it would invest in Stream but not CentOS. "CentOS Stream is critical. It meets very specific needs for us," said Exelbierd. "We laid out our case and we said we're moving our engineering contribution, people time in some cases... we want to call your attention to them because depending on what you decide to do, there are potential liability issues that could result, so we want to make sure you have a plan."

The outcome was the end of CentOS Linux. But was this in part a plan for Red Hat to profit by migrating some users of CentOS Linux to paid-for RHEL?

Exelbierd's response was nuanced. "It would be easy for us to say this is about our business model... if business model includes non-revenue," he said. He insisted that driving CentOS users to become paying customers was not the primary intention, and that the email address for those considering switching to RHEL was not that of the sales department.

"It goes to me and two of my colleagues in the business unit," he said, the goal being to find out more about how CentOS is being used and to work out what new no-cost and low-cost programmes would be required. According to Exelbierd, one of the issues was that CentOS Linux had many unknown users. These are people who, the CentOS team said, "never called, never write, they don't interact with us."

Part of the rationale for getting more people onto RHEL is to collect feedback. "It is the business model. It is absolutely not a mailing list for salespeople," Exelbierd said, adding that Red Hat has no commercial interest in small-scale users. "Nobody wants to go after the person with one server, two servers, 16 servers. There are definitely going to be some folks for whom their CentOS Linux, if it's going to become RHEL, will become paid RHEL, absolutely. But our goal was not to sit down and make every CentOS Linux user a revenue RHEL customer."

Red Hat said recently that using up to 16 RHEL servers will be free, via an amended developer licence, but many CentOS deployments have more servers than that. Why doesn't Red Hat simply say, you can use RHEL in production for free if you do not require support? "That's not the way we think that everything should work out," Exelbierd told us. "We think that there are values to RHEL that go way beyond the bits. We're creating opportunities for people to use RHEL in different ways, but you can't just make a blanket statement... it starts with how you define support. A lot of people would say support is 'can I just call someone', but support is also all of those updates and security fixes," he said.

The company's intention is to cement the idea that the production version of the operating system is RHEL. "One of the things that this change has done is it's made it pretty clear RHEL is for production," Exelbierd said.

Why has Red Hat taken its developer programme and confusingly made it the vehicle for small-scale production use of RHEL, which is not for developers? "Unfortunately, Red Hat is still an IT company and IT in general struggles with names," said Exelbierd. "The core of the programme is still focused on developers, but there is going to be a use case here for hobbyists, system administrators and others who don't self-identify as developers. Remember that a component of the announcement that we made is around development teams in enterprises. So I don't want people to think we bolted this on."

Our intention is not to stop cancer research, that was never a goal here, nor do we wish it to be an accidental side effect

How will Red Hat define a server, in the 16-server programme? "We're working on an FAQ which has all of the specifics. It gets into things like cores and other pieces. A supercomputer sitting in my basement was not intended for this programme." That said, "the intention is not to put significant limitations either on the actual use cases or the hardware," Exelbierd told us. We should know soon: he added that "the terms and conditions should be ready before 1 February."

Red Hat said this was the first of many possible new free and low-cost plans – what else is in the pipeline? "We are looking at everything from something like, say, high-performance computing, where the models keep evolving, academic and research, non-government and small government organisations... our intention is not to stop cancer research, that was never a goal here, nor do we wish it to be an accidental side effect," said Exelbierd.

"A third large group is how we interface with our partners. And maybe ISVs that haven't chosen to partner with us in the past. There's clearly been friction there. We want to use this as the opportunity for examining all of the pieces to help those folks as well grow the ecosystem," he said.

In near-contradiction of his earlier statement about clarifying which Linux editions are suitable for production, Exelbierd said that Stream OS has been misrepresented. "With regard to Stream, I think that people are choosing to use very specific lenses when looking at it, that they're deliberately ignoring pieces of information, in some cases, or in some cases it's truly accident," he said.

The next version will not be like Stream 8. "We have said publicly this is not the way it's going to be. Stream 9 is going to show you that code flows from Fedora to CentOS Stream, and then the periodic pulls from Stream that become the RHEL minor releases. We're working right now on getting some demonstration repositories out to show what we're looking at for the contribution workflow.

"I think people are forgetting CentOS Stream tracks ahead of RHEL, it doesn't track ahead of the universe. RHEL is prided for the fact that it is a stable operating system over that 10-year lifespan. How much change goes in from minor release to minor release? That's the delta you could expect out of CentOS Stream.

"This stuff has been tested. It is not work in progress. If our customers could consume change that fast, it's what we would ship. I talked to a lot of CentOS Linux users who run 'yum update' every single night. Well, then Stream's going to do the same thing for you. You are already leading the Stream life in a lot of ways. I think that's the piece that people are just overlooking." ®


Biting the hand that feeds IT © 1998–2021