History of CentOS: How a biochemist's Linux hobby project became the enterprise world's default operating system INTERVIEW Gregory Kurtzer, CentOS's founder, tells the story of how the Red Hat Enterprise Linux clone was born of a small group of rebuild hackers and Linux fans who were angry that Red Hat Enterprise Linux had replaced Red Hat Linux and convinced they could do better. Back in 2003, Linux fans were ticked off at Red Hat because …
I used RH linux back in the day and even became RHEL certified (I needed it for one job) but at home, it was always CentOS.
Then Red Hat FSCK'd up and clones like Rocky and Alma were born thanks to the power of Open Source Software.
These days, I use a mixture of Rocky and Alma Linux on servers and desktops with either XFCE or Cinnamon. I suspect that I'm not alone in that sort of choice.
Dodged that bullet I guess. ;)
Pre·RHEL Red Hat was also being used in scientific/research contexts.
I had a 20 node Pentium 4 Beowulf cluster built using RH7.2 which very soon ticked a few unfortunate boxes.
64 bit hardware (and the large virtual address spaces) were needed for most applications which, until the Opterons arrived and were affordable, meant 64 bit Unix boxes were still desirable.
DEC had a version of RH7.2 that ran on some of their Alpha workstations which worked but I found slower the Digital Unix 4 (OSF/1.)
I imagine once IBM decides that RHEL is just a platform to run their proprietary software and services they might decide it is cheaper to outsource the development and maintenance of the boring FOSS components to the RHELatives.
I definitely wouldn't say there is nothing cheap about IBM but rather in common with most US enterprises, there is everything cheap about IBM.
You need a free version to attract the curious, the tinkerer, and the innovator.
But who wants to maintain a free version? Why not hive it off to a dedicated community so they can do all the heavy lifting, while you get on with selling to your locked-in and $$$ loaded support junkies? The freebie brings in new junkies clients - marketing bait for free.
RH once tried to kill off CentOS, until the penny dropped on the market presence it brought them.
Marooned open source projects should choose whichever distros that suit their needs. The fraud that is enterprise Linux is a necessary evil in modern corporate landscape. You just can't stop the subversion by multiple mega billion dollars players.
For what's it's worth I see Devuan and Artix with the non-systemd OS and going back to Sys V Init and X11Libre now instead of Xorg has been a breath of fresh air in this stale corpo stink space.
We should FORK more of the "enterprise" projects and challenge the status quo especially when they do not want the open source design philosophy but all the freee beer benefits.
They underestimate IBM - they 100% knew what they were doing with CentOS - they deliberately derailed the project because somewhere high up in IBM someone did the calculation that there are X number of CentOS users and if they fuck the project Redhat will gain X number of subscribers.
WE all know that was never going to happen - half the CentOS user base were people spinning up cheap virtual servers who were never going to pay an OS subscription fee and the other half were all techies who chose CentOS because they wouldn't touch (let alone pay) IBM with a 10 foot barge pole. But Redhat believed it, so they fucked with the project just enough knowing the user base would jump.
What they probably didn't expect was for it to fork into two new projects (they'd maybe factored in one fork). The great thing is that by splitting into two projects, both very successful, it stops future IBM meddling. They'll never again throw a ton of money at a project to deliberately fuck it (what's the point when two new ones will just spring up in its wake?).
RIP CentOS and thanks for all the fish. For a while you beat all the odds (and the established giants in the industry) and were the Linux server of choice for the wider tech community.
Not to defend IBM by any means, quite the opposite. But I think Red Hat management for some time was/is pretty capable of messing up things on their own.
I give Red Hat full marks for bringing Linux more into mainstream corporate use, I had several good jobs for years around Red Hat Linux 6,7,8,9 as Solaris and other Unixes were fading away. But when they dropped Red Hat Linux for RHEL, you could see what was coming.
Well said. In a very interesting historical article, the Big Blue factor was surprisingly absent. In any account of Centos, fingers need to be widely pointed to avoid any false interpretation of what really happened and the terrible result. Release 6.4 was the pinnacle for me, I'm far too old for any corrupt, greedy, fake agenda corporate nowadays, imho ;->
When I first got interested in Linux, it was a Red Hat CD that made sense. I tried Debian, but back then it seemed too alien to what I was used to (OS/2 - never been a Windows fan) and so it got filed. Then RHEL turned up and I switched to Fedora, then got fed up with the need to re-install every 18 months and so looked around, eventually switching to the Debian-based Mint and learned about how they did everything and where the default config files lived. That's what I use for user machines now, although Ubuntu and Debian are in use on servers where I just interact over SSH and don't need the overhead of the GUI. I did look at Centos at one point, but decided I liked Mint better.
> Kurtzer thinks that Red Hat's messaging "was just a complete cluster… nobody, including the people at Red Hat, really knew what they were saying."
I think that describes it pretty well. As upset as many people were about the more technical and release-related changes, Red Hat's marketing of the change probably made things worse.
People were already shell-shocked from Red Hat cutting short the announced 10-year support window, my guess is it hit hardest at companies which were already started down the road of CentOS 8.
Places were scrambling to figure out what to do, in some cases affecting longer-term budget and support plans. They'd read the announcements and blogs from Red Hat marketing and product directors and community mangers(?), trying to claim this would actually be better for them, somehow. Maybe someday, down the road a few years, it might turn out that way but I doubt it felt that way to CentOS users at the time.
We were just migrating to CentOS 8 (we never started deploying before a .2 minor release) and luckily both Rocky and Alma , apart from very quickly building a complete enterprise distro rebuild , also published pretty solid docs on migrating. So we migrated around 20-30 servers (over half production machines , about 5 physical and the rest VMware) in the course of 2-3 weeks from CentOS to Rocky Linux (our choice, largely decided by an upstream vender switching to Rocky)
All of the migrations worked just fine and a few of those machines are still up and running today, showing what a great job community rebuilds did and still do !
There are a lot of stories similar to that. At $WORK we had started checking out CentOS 8 but production was still running 7, so we weren't committed at the time and could change course without too much pain.
I think when Red Hat announced Stream, if they'd simply kept the existing support durations for 8 intact, giving people (especially those who'd already upgraded) the long-term safety net they expected and planned for, instead of cutting them off early, there might have been less backlash.
That felt like Red Hat betraying trust, pushing customers into an RHEL situation they didn't expect and hadn't asked for, and no amount of marketing or spin is really going to fully repair that.
Quite. It was a moot point for us at $WORK, but I expect we'd have moved production infrastructure to something in the debian family before running Oracle's linux.
We'd deployed Ubuntu for specific projects at times, and there was already some interest in having Ubuntu available in the fleet anyway. It would've taken some work but I imagine it would've turned out well enough.
... not that we had more faith in Canonical than Red Hat/IBM, mind you, merely more than Oracle.
It is a little weird to write a column like this presenting Alma and Rocky as the "children" of CentOS, and to never once mention CentOS's primary successor, CentOS Stream, as a successor.
It is a little weird to quote Kurtzer talking about the "nasty" responses to Red Hat without acknowledging that a great deal of that came from people associated with the Rocky Linux project, or directly from the project itself, by way of its own merchandise.
It's a little weird to suggest that CentOS release latency improved, making the Stream change unexpected, when the opposite is true. Release latency was getting worse. Users of the GA users were running unpatched systems for 2-3 months out of the year. It's actually hard to imagine continuing that process without major changes.
>"It's a little weird to suggest that CentOS release latency improved, making the Stream change unexpected, when the opposite is true. Release latency was getting worse. Users of the GA users were running unpatched systems for 2-3 months out of the year."
And who was corporately responsible for the entire CentOS project at the time and were themselves the cause of any release latency and unpatched systems? Wouldn't that be IBM/RedHat??
It's absolutely true.
Wikipedia is a quick reference for this information: https://en.wikipedia.org/wiki/CentOS#Older_version_information
If you "show" the content of any of the tables of releases, you will see a "delay" column for each release. That is the amount of time between a minor release of RHEL and the corresponding minor release of CentOS Linux. During that time, no security patches were pushed to CentOS's GA repositories. So any time a security patch for RHEL was published during that delay, users had to wait for the CentOS Linux minor release to get the security update.
During the CentOS Linux 7 release, delays were typically 30-45 days. Twice a year, users were waiting 30-45 days without any security updates.
You are using numbers based on the release of ISO images and those were only intended for use for new installs. CentOS would build the updates as a priority and get a working tree and then publish that in the CR repo for anyone to update a running system from. Once that was done then work was started on building the ISO images but since the updates were already out there, this was done at a more leisurely pace.
I checked 7.8 which was released by Red Hat on 2020-03-30 and the updates for that were built and in the CentOS CR repo and available for anyone to use by the following Monday, 2020-04-06. That's 8 days.
It's a similar story for 7.9 which RH pushed 2020-09-29 and was available to CentOS users on 2020-10-14. That would be 16 days so not as good but still less than half the numbers you are using.
The article sez He came out of biochemistry and genomics, where compute‑hungry (Basic Local Alignment Search Tool) (BLAST) jobs were chewing through early SGI systems.
Perhaps I'm misunderstanding the article but if I am undertanding it, the time frame is 2003-ish.
At this point you might call them "late SGI systems."
Silicon Graphics had been around for a couple of decades at that point, formed in late 1981,1
I probably first encountered them at SIGGRAPH a year or so after that.2
Really, SGI was about to become the late SGI, since it went bust in 2009.
______________
2 I remember being on a plane flying out of San Jose, California, headed to the conference and was seated across from a bunch of SGI salespeople who were loudly lamenting that SIGGRAPH was too "academically oriented" to sell much. Probably true since this was before computer graphics really broke through into the gaming, advertising, and Hollywood markets, so a majority of the attendees were broke graduate students and adjunct professors with zero purchasing power.
I started working at a research facility at about this time. The computational chemists were very unhappily moving off the SGI boxen to Linux workstations[0] and the bioinformatics folks were moving off Solaris boxen to Linux servers.
[0] They seemed to resent moving off colourful SGI workstations that cost more than many people's salary onto dull HP workstations
(Spelling intentional).
Kurtzner also promoted a cluster environment based on network boot of multiple nodes and NFS serving standard /usr/bin executables. Called Warewulf.
It was a clever notion: nodes won't fail due to disk drive issues. I think he built a few 1000+ node clusters at LANL and elsewhere.
I ran it in my home lab (couldn't get $WORK approval to build it).
Verrrry interesting.
Bit off topic as not CentOS but still OA mentions other clones of EL.
I recollect reading about the events that led up to Scientific Linux / Fermi Linux after the event. Something along the line that Red Hat changed the pricing model for RHEL in high energy physics installations from an annual site licence fee to a per core subscription, without much warning, and part way through a budget cycle. These installations have rather a lot of cores, and high energy physicists are good at arithmetic, so it was obvious that it would viable to employ some people (among them Connie Sieh) to produce and support a clone from the srpms that Red Hat made available. SciLi was OK until the CentOS 8 truncation of support, then most moved over to Rocky/Alma whatever. Some are actually using Stream I gather.
Some here might know more, I would be grateful for corrections.
PS RIP Springdale Linux aka PUIAS Linux. I enjoyed versions 5 and 6.
To be honest I don't think IBM/RedHat has good Marketing or sales team. What they could have done is keep CentOS old cycle and change it's name it "CentOS LTS" and Current CentOS stream cycle named as "CentOS" they could have gradually let community decide which one is better in terms of ecosystem and gradually kill the cycle which doesn't work for community.
Why? Because Alma is basically an LTS version of CentOS Stream, not a bit-for-bit clone of REHL. Rocky plays cloak-n-dagger games to get the code to build their distro and I won't be surprised if IBM/RedHat eventually drop the legal hammer on them. That would probably be a bad move by IBM but then we all know the amount of stupidity there.
Give credit to Red Hat/IBM they finance a lot of necessary technical work and they do not close access to Fedora to things that have no obvious corporate value. Pity corporate greed is unable to do it without being forced by an IBM/Red Hat middleman. Pity all the people that benefit from this technical work love to pretend it is their due and they are building something autonomously to “save” the world from Red Hat/IBM greed.
Centos and the other clones, while necessary to provide access to the general public, were never that interesting from a technical point of view. Mostly because they all restricted themselves to “serious” “useful” established stuff, so tech innovation happened elsewhere. IBM was in total agreement with them when it tried to steer its parters towards Centos stream and its reduced set of serious packages. However techs know better, so Amazon Linux, Azure Linux, and IBM’s own Hummingbird, all build upon the unloved Fedora bypassing Centos Stream.
