Red Hat makes things worse for paying customers to sell more subscriptions which will fail in the end.
One example for this is their UBI (universal base image) which should be used to build RHEL containers. The image is "free to use" and therefore they removed lots of packages from the included repos. Many packages cannot be installed - with the current version 9 they removed even more packages. This is really frustrating and time consuming.
My recommendation: use Alma or Rocky Linuxes as an alternative
Because the thing that centos has morphed into isn't terribly useful. It gets most updates early, and as a result has many bugs, which often take a good while to get fixed.
However for security issues it's the other way around - RH won't release security fixes into centos until they've rolled out across all their sold products, even long-term maintenance releases of old OSes. This can take weeks.
End result: many bugs, often broken, and can be missing security fixes for far longer than any sensible user would be comfortable with. It's a product to make facebook (who are the primary user of it) happy essentially, and very little else.
My source comes from watching the IRC channels every single day, and seeing what users complain about and how long it takes to get fixed, compared to before the switch to stream. And for important security fixes there's a huge change - I've seen upwards of two weeks for kernel vulns in 8-stream that would have been O(hours) previously.
> Practically how is that damaging your use case by using CentOS stream? Do you see many things being unstable?
I don't know what the actual rate of bugs ended up being; I burned CentOS from my infrastructure when the news dropped.
> It doesn't refer CentOS as a beta release of RH.
Yes, RH is quite attached to their lie that the thing that gets changes before the next stable version of their commercial product isn't a beta. It's unclear to me whether they somehow actually believe this (and if so, why they're giving free users a supposedly better product than paying customers), or if not why they expect anyone else to believe them, but that is indeed their official claim.
UBI images work fine for many use cases, and importantly, they're free to redistribute, not just use. If you're selling a containerized software solution to people, it might make sense to use the UBI as it will get real security updates, etc.
"A single platform, where all operational data sits powering real-time AI driven feedback loops for continuous workflow improvements.
In the platform, automation enablement scales to the size of the wanted business rather than being limited by the operational headcount."
I feel like an entire graduating class of MBAs is kept employed just maintaining this page.
If you are a VP of a mobile operator who has KPI's of "integrate AI", "drive automation", and "stable headcount" then that page pretty much says take-my-money.
Rakuten's Cloud business (aka Symphony) is made for Telecom Companies wanting to switch away from traditional RAN model to OpenRAN. It's not really a direct competitor to AWS or GCP (although AWS, GCP and Azure seems to be making big progress in getting Telecos to switch their 5G infra to public cloud).
Symphony is nothing but a wrapper on top of vendor software from the likes of Mavenir, Nokia, NEC, Altiostar etc. There is not much real engineering done by Rakuten themselves, they just get different vendors under same umbrella and ask them to play nice with each other.
> "A single platform, where all operational data sits powering real-time AI driven feedback loops for continuous workflow improvements. In the platform, automation enablement scales to the size of the wanted business rather than being limited by the operational headcount."
Reads as someone shuffled some buzz words and took a random sample.
But OK, I guess you can't do anything wrong with this approach in MBA land…
I'd not heard of Rocky Linux or CIQ before so I thought I'd check them out.
Sadly looking at their corporate About Page[1] tells me nothing about who they are, when they were founded or who their key people are.
Instead it's all "Why People Choose CIQ" and "Empowering People" type meaningless marketing fluff and I'm none the wiser. That doesn't fill me with confidence.
As a user, I struggle because of one main reason. The seeming lack of LTS version either currently or planned.
I currently am predominantly Debian. Which is inherently LTS, and generally works Just Fine (TM) except when you fall into something where they are a just a bit too cautious (e.g lack of in-house PHP 8 option in the current Debian Bullseye release - nothing in main repo, nothing in backports, you have to go to an external repo).
I refuse to use Ubuntu because of their increasingly commercial smell. I tried SUSE but always felt a bit uncomfortable with some of its more exotic sysadmin tendencies.
Which kind of leaves me in the market for another LTS platform.
Yeah, that. One of the driving points of RHEL (and all the downstreams as a result, like Rocky) is that you get a minimum of 10 years of support for every release.
I haven't willing used Redhat since they dropped desktop support, and spun out Fedora as a community supported replacement. I had just purchased a boxed copy with a book from best buy, and they let me return it even though it violates their policy against returning open software. I knew the manager that was working and explained open source to her, but I think she just let me slide cause she liked me. I used the money to buy a physical copy of Suse, which came with a bigger book. Suse was my main driver for a few years after that, but I switched to debian when Novell bought them and started causing trouble. I really liked Suse though, Yast was way ahead of it's time.
Sorry, that took a bit long to get to the point that these days I only use Redhat because it's the only linux distro available at my job. I know I'm not the only one who feels this way. Maybe IBM is just realizing no one gets joy from using RedHat anymore, so they might as well squeeze every bit of money out of it before it dies.
You're right, I should have said home user support. This was before they had RHEL or Fedora. Maybe 2002 or 2003? It was very annoying at the time. I haven't trusted them since.
First, you won't use arch on production servers. Second what you said is incorrect. Arch Wiki has info on a lot of topics, but mostly for desktop or casual server usecase and certain aspects are completely ignored. Arch's Wiki is excellent but its also about the quality of documentation and how its structured.
RHEL's docs are simply unparalleled in the Linux world. I only wish search was better, but if required I can download them and use my own search anyways.
> First, you won't use arch on production servers.
I don’t know, for a pet server this might not be as dumb an idea as it sounds (though without automatic updates, obviously). Perhaps with aconfmgr[1]?
I haven’t done it in anger, but over seven years of running Arch on a personal server I’ve had a total of three instances of breakage: upstream strongSwan systemd integration change[2] (when using a config format they’ve considered deprecated for years), upstream nftables parsing bug[3] (I knew in advance that nftables is for adventurous people), and upstream Kea config format change[4] (and, well, Kea really bloody sucks—suggestions for anything else that can do DHCPv6 PD welcome). All of those resulted in the relevant service failing to work at all and would’ve been caught by a reasonable staging environment. My (Arch) GNOME desktop broke much more frequently during that time.
I guess nobody is seriously using Arch as a server OS in production. So this is kind of comparing apples to oranges, sure.
But Arch's documentation is indeed one of the best around!
I'm a die hard Debian user since forever but for the last years I don't even bother to try to find anything in the abysmally bad Debian docs. I go straight to the Arch wiki (or to the Gentoo one for even more in depth docu on internal aspects of Linux software).
Contrary to that I've almost never found anything of value in the RedHat docs (that wasn't RedHat specific).
If the Arch wiki would expand on some things that are more related to running servers in production it would be the Linux documentation.
The overall quality and general applicability of Arch's docu is already unparalleled, imho.
Reminder, to stay true to the trope: I'm not using Arch Linux (as such) for anything.
A not insignificant portion of resources at access.redhat.com are locked down behind the RH subscription paywall. It’s fairly trivial however to just obtain a free RH subscription via the RH Developer program:
It's a bit of split as to how Red Hat considers the term "documentation". The actual product docs are free to access while the Knowledge Base, which is considered a value-add for customers, does have a lot of its content (usually troubleshooting or specific instructions/guides) behind either a Red Hat login or needing an active subscription. For which, as you mentioned, the Developer Subscription for Individuals qualifies.
Ever since the systemd thing, I can’t get it to work. :-(
Devuan is OK, though apt-get upgrade broke grub on one of my machines.
I settled on manjaro. I kind of hate it, but at least it sort of works. In particular, I have it on a shared desktop, and “switch user” works.
It took 3-4 hours to get xscreensaver to work (poorly) with multiple sessions. There’s no way to save games in micropolis (open source simcity) because it is a flatpack. I frequently reboot because PulseAudio is still terrible.
Proton supposedly runs steam games on freebsd these days. Maybe that’s worth a spin.
If you're willing to run Manjaro (which uses systemd), any opposition to running e.g. Fedora with GNOME? I'd guess multi-session would function fine with GNOME on GDM, and pulseaudio has been replaced by pipewire. Although you could also use pipewire in Manjaro...
> Proton supposedly runs steam games on freebsd these days. Maybe that’s worth a spin.
Via running a Linux userland with syscall emulation. Since the GUI and user session stuff in FBSD is mostly the same as Linux (DRM/KMS for graphics, libinput and the xf86 DDX drivers, logind/ConsoleKit for session tracking, D-Bus for IPC) so I don't think this will address many grievances...
> I want Debian circa 2002 back.
Did display managers back then even support running multiple concurrent X servers on the same seat?
I never made the switch from Gnome 2 to 3; I switched to KDE instead. I use a redhat derivative at work, and it's a non-starter for me. RPM + yum is still a hot mess vs. dpkg.
I thought they were recompiling proton for freebsd (it is definitely open source), but I haven't looked carefully.
The "switch user" button is one feature. Whether or not Debian had it in 2002 isn't the point. The point is that what was supported was stable. That hasn't been my experience with it since the pulse audio disaster, and then the init wars.
I should look into switching to pipewire, but the only reason I'm using manjaro is that it mostly works out of the box. I'm not really interesting using a distro that makes me tinker with it to make my sound card (which has perfectly good drivers!) stable.
You know what's better than free? Something that's free and also being used by the majority of industry players.
Personally I think this news is a huge one for Linux and RHEL based users community in general. This is the main reason why FSF's GPL is a great license for open source software.
According to the article not only the Rocky Linux becomes the natural successor for CentOS, it also added the real-time capability to RHEL based clones, and all it's done with the collaboration with the Linux community. This looks to me like a RHEL distro on steroid and before long there will be tens if not hundreds of books using the Rocky Linux in its title descriptions.
I kind of feel Rakuten is just moving their problem down the line, or up the stack. Key focus areas in Open RAN is containers now. The commodity hardware at the edge needs to be able to run many of the network functions in containers, and to do that, a service provider is going to need very good k8s orchestration. There are not a ton of "free" options any mobile service provider will choose to manage this at scale.
Regarding Rocky Linux, it is "...an open-source enterprise operating system designed to be 100% bug-for-bug compatible with Red Hat Enterprise Linux®.": https://rockylinux.org
Does Rakuten ever give back to to the community ? I doubt it very much, just look at the struggles many open-source communities have with these large companies sucking blood from them over the years (openssh,openssl, ntp...) It is Halloween today, had to get a some kind of spooky over the top reference in :)
Red Hat made a gamble and it may turn out to be a miscalculation. By buying CentOS and leaving it be, they had tighter control over it's release cycle (how fast patches and newer versions were made available etc) as well as metrics on who was using it. Additionally, all of the CentOS users were still "in the Red Hat family" and at least kept the door open for future conversions/sales. They threw that all away now, they don't have any control over these forks and have wasted a lot of goodwill on short term profit.
Red Hat should've gone with the Ubuntu model instead and maybe even change to usage based pricing instead of a yearly subscription too, IMO.
> Additionally, all of the CentOS users were still "in the Red Hat family" and at least kept the door open for future conversions/sales.
How? When Ubuntu tried advertising Ubuntu Pro via apt-get, everyone was outraged, and rightly so. How else could Red Hat talk to a CentOS user to convince them to pay up?
By providing value beyond just support. Ubuntu does this via Landscape, Red Hat is still trying to figure this out with Insights and the obtuse licensing for Satellite IMO. Packaging software has a high level of effort but really low value in customers eyes due to how many Linux variants are out there (again, IMO), so putting that behind a paywall these days is dumb. Containers exasperate this problem as most folks don't care if RHEL ships with PostgreSQL, they'll just run the Debian-based container.
For support, folks need enterprise support contracts. Maybe not everyone, maybe not every system, but folks that need it will pay whatever for it, it's the cost of doing business. Charge accordingly.
Out of curiosity, which aspect of the subscription terms do you find obtuse? Simply put, any system managed by Satellite needs to have a Smart Management subscription. You can get these individually, or bundled with a RHEL purchase as the add-on. For consumers of the public cloud marketplace RHEL images, they include the client-side aspect of Smart Management (no Satellite infra, but you can request this) so you don't need extra subscriptions to cover those.
Ah, if only they’d thought of that. The longtime successful and solidly profitable (~$500 million a year) Red Hat corporation should copy the business model of Canonical, which makes 1% as much profit. Do you charge by the hour for this kind of advice?
Another mistake Red Hat has made over a long is having a "talk to the hand" attitude about doing anything differently.
As an example of the pathology, there is a long standing bug in GTK that causes the positions of things to be off by few pixels if you are running an X server like xming in multiwindow mode. It doesn't sound like much but it can make it impossible to click on menu items for menus at the top of a window.
The people involved at Red Hat would say in no uncertain words that they couldn't care less if it worked when you were running a multiwindow X server on Windows -- out of nothing but pure hate for Microsoft and people who use a Microsoft OS.
Expecting devs who work primarily on Linux to install a completely different OS and spent a few hours configuring it in a particular way, to work on an issue which is neither relevant to their own interests as FOSS devs, nor relevant to their employer, doesn't seem entirely reasonable.
If someone came along and fixed the bug, submitted a PR, did all the legwork and then it was still left to languish anyway then you would have a reasonable complaint. Could you provide a bit more context?
Fortunately for Red Hat, all of the makers of "cross platform UI" platforms are in a race to the bottom and that is why we are stuck with Electron apps that are all (30+x) MB in size and depend entirely on engineering work from Google.
I don't think that's the case. The underlying issue is that while Btrfs is growing in popularity across the Linux space, Red Hat just hasn't attempted to staff for supporting Btrfs in RHEL. No staff to support it to the level they wish to offer technologies in RHEL means no Btrfs in RHEL.
I would hazard to say that is probably because they haven't experienced any requests for the filesystem from their customer base. If customers started specifically asking for it, I think that would change.
I've used a Linux desktop for years, back in the KDE 3.x days (2005 something) where everything worked well. Then KDE 4 came, GTK 3 and Gnome and that's when I left because it was just unusable and went to Windows.
ChromeOS and Android are nice but are not desktop user interfaces.
I do have a Macbook Pro but only use it for testing purposes. It works well but I only use it for running Safari so can't really say how it works for a power user.
XFCE has been working well since inception, and Mate was working well during that time too. On windows, when win 8 came out, you either got left in the dust or had to endure it. On linux, when groups fuck up their thing like that you have the liberty to switch at very little cost.
The desktop bit is indeed a little unstable on Linux for some features. My experience as well. Especially with monitors, plugging/unplugging and window placement.
The underlying OS and command line world is sooo much more stable than Windows.
I agree KDE3 was the most stable Linux experience (I'm with a few small adventures still a KDE user).
I want corporate open source projects to production-harden their stuff on my OS, so I can force them and their communities to buy licenses from me. They said no. Boo hoo. :-)
> Amin teamed up with one of CentOS’ founders, Greg Kurtzer. Apparently the two have been working together for 13 months to build a real-time kernel to run complex workloads and complies with the Linux open source community.
One example for this is their UBI (universal base image) which should be used to build RHEL containers. The image is "free to use" and therefore they removed lots of packages from the included repos. Many packages cannot be installed - with the current version 9 they removed even more packages. This is really frustrating and time consuming.
My recommendation: use Alma or Rocky Linuxes as an alternative