People will continue to use ubuntu if it meets their needs, mostly in terms of packages. I'm a seasoned sysadmin and sometimes it's difficult to make things work "the right way" with little effort in centos/RHEL.. but with ubuntu it's very easy to make reproducible machines even with no proper ops experience.
Honestly, I don't like ubuntu server, but this is my choice.
I also dislike systemd, and I laud any distro that doesn't use it, even if they will begin using it soon.
systemd has its place, and that is "Linux on the desktop." Say what you will about it being a pain in the ass to debug or maintain or how it eats every Unix service in order to grow like a Katamari, it works beautifully for package maintainers to keep everything in order. If you like `apt-get install`s or `pacman -S`es, this really helps those who need to keep that in order.
However, I would never, ever, recommend it on a server. Unless you've got some sort of really complex deployment, get a barebones kernel and SysVinit and put it on a read-only FS. Do you need any more than that?
Deploying custom services on systemd requires a 5-line service file, only 1 of which is not boilerplate. If it works as a foreground service, then it'll work as a start-up service and your logging will go into syslog correctly.
The "it's too heavy" argument doesn't hold much water either. I've not yet heard a realistic description of a lightweight server which can actually be a practical server.
This is terrible clickbait, it reminds me of that 4chan HN spoof thread.
The author disregards many of his points through his article, yet only acknowledges this with a small footnote at the end of each section. Those sections should have a note at the start, or be removed altogether.
He praises Red Hat's use of upstream LT kernels, yet ignores the fact that RHEL's lifecycle is WAY longer than upstream LT, essentially negating any benefit after some time.
Perhaps his comments about Upstart and logind are valid. However I doubt Ubuntu devs are going to break their entire init daemon, and surely as long as logind works well this isn't an issue either?
Only the AppArmor point is very valid. Ubuntu has never been a good choice for a virtualization host. Red Hat, KVM, and SELinux are very closely knit and very good.
I don't think the author meant that you should never ever use Ubuntu, I think the author meant that this version of Ubuntu's guarantee of LTS is not very good. The support for all of the components between Upstart and Mir and Compiz is derived solely from Canonical & co. Further, the patching that will need to be done to the kernel is non-negligible.
It's going to be a pain to hold together, and thus support may end up being worse. I think that's the point. Just trying to be charitable.
I actually laughed out loud when right after the section "No-one else is using compiz", he has a section saying "So what if ubuntu is widely used, 'it is popular' isn't an argument".
Also, the author isn't very familiar with ubuntu - because there are official backported kernels. I'm using one on my servers stuck on 12.04LTS. Those backported kernels wouldn't have been there on release day, but were definitely there by the time of a single Fedora release cycle.
The primary point the author makes is that this version of Ubuntu LTS has a kernel and init system that only it is going to be supporting for most of the lifetime of the release. Bringing up RHEL's use of kernel versions no longer supported upstream isn't really relevant - RHEL has a proven track record of supporting a kernel with patches and backports single-handedly.
I'm not very fond of Ubuntu, actually I hate it.
But this is an incredibly weak arcticle. Most of the points
involve Ubuntu not adhering to what all the other distros that
the author considers supperior are using. So? Isn't the point
of a distro to do things differently?
Upstart & Compiz are unsupported but will have to be backported to for five years. logind has been ported to Upstart for... some reason, and that needs support too. No one else uses Mir, but maybe it's "the future" or something, so I guess that one's okay. But oh, don't forget the kernel. That needs backported fixes too, because no one else will do it.
Basically, they've made a mess for themselves, and by extension, they've made a mess for "you," the user.
Making distros that investigate different ways to use Linux is a fantastic thing. Without it, we wouldn't have Alpine Linux or Arch Linux. But this seems to lack any sort of actual logical basis. Alpine is about being super-light-weight, Arch is about always staying fresh so devs don't need to recompile their tools manually. So what's Ubuntu's excuse?
Debian voted for systemd in Feb 2014. Ubuntu 14.04LTS was released in April 2014, and was intended to be a stable release with no major changes. This was bad timing for Canonical, but the way it's been phrased in the article is that Canonical are idiots for keeping on with their tried-and-true Upstart instead of the not-quite-ready-yet Debian systemd.
Note that it's April 2015, 12 months later. Debian's 'systemd release' is still not here yet.
I read the "no gratuitously negative comments" rule, and I will have you know that I spent a full 10 seconds thinking about different ways to phrase my comments on this (old) article.
Article summary: "I like RH/CentOS, and so you should use it" - as suprjami points out in his comment, he disregards most of his own points.
We use ubuntu for general purpose stuff, and alpine linux for more task-specific stuff - especially for container images. I have a bitter resentment towards rpm and yum hasn't done anything to make that go away. We don't like systemd, and are evaluating alternatives to ubuntu for that reason.
Debian if you're OK with the packaging system??? The packaging system is probably the biggest win, hands down! I was using RedHat back when RPMs installers still couldn't do dependency management. Even though things have gotten better, I still hate RPMs. This article is flat out terrible.
I had the same response as well. It's more annoying to get a package into dpkg format, but once it's in, apt is so much better than yum. More flexible, more informative, less chaff in the messages...
Too late for that. I've installed 14.04LTS, then upgraded to 14.10 when it came out. None of the author's arguments really made much difference to me since I use Ubuntu mainly as a desktop and dev machine. I like to move with the cutting edge, as much as Ubuntu moves forward that is. If I wanted a longterm server, I'd probably use RHEL or CentOS.
OP was specific that you shouldn't use 14.04 LTS, and all his reasons were because they were writing checks they couldn't cash as an LTS. He's even careful to point out that he's not anti-Ubuntu, just that you should avoid that particular release if you're looking for long-term support.
As a contributor to Ubuntu ( on the desktop bits ), I find this article quite offensive. There are a ton of smart people working on Ubuntu Server ( both working at Canonical and otherwise ) and this kind of click bait just disparage's their efforts to provide a amazing product :/
20 comments
[ 5.6 ms ] story [ 61.2 ms ] threadHonestly, I don't like ubuntu server, but this is my choice.
I also dislike systemd, and I laud any distro that doesn't use it, even if they will begin using it soon.
Is it really worth swimming against the tide?
That depends where the tide is taking you....
However, I would never, ever, recommend it on a server. Unless you've got some sort of really complex deployment, get a barebones kernel and SysVinit and put it on a read-only FS. Do you need any more than that?
The "it's too heavy" argument doesn't hold much water either. I've not yet heard a realistic description of a lightweight server which can actually be a practical server.
The author disregards many of his points through his article, yet only acknowledges this with a small footnote at the end of each section. Those sections should have a note at the start, or be removed altogether.
He praises Red Hat's use of upstream LT kernels, yet ignores the fact that RHEL's lifecycle is WAY longer than upstream LT, essentially negating any benefit after some time.
Perhaps his comments about Upstart and logind are valid. However I doubt Ubuntu devs are going to break their entire init daemon, and surely as long as logind works well this isn't an issue either?
Only the AppArmor point is very valid. Ubuntu has never been a good choice for a virtualization host. Red Hat, KVM, and SELinux are very closely knit and very good.
It's going to be a pain to hold together, and thus support may end up being worse. I think that's the point. Just trying to be charitable.
Also, the author isn't very familiar with ubuntu - because there are official backported kernels. I'm using one on my servers stuck on 12.04LTS. Those backported kernels wouldn't have been there on release day, but were definitely there by the time of a single Fedora release cycle.
Basically, they've made a mess for themselves, and by extension, they've made a mess for "you," the user.
Making distros that investigate different ways to use Linux is a fantastic thing. Without it, we wouldn't have Alpine Linux or Arch Linux. But this seems to lack any sort of actual logical basis. Alpine is about being super-light-weight, Arch is about always staying fresh so devs don't need to recompile their tools manually. So what's Ubuntu's excuse?
Note that it's April 2015, 12 months later. Debian's 'systemd release' is still not here yet.
Article summary: "I like RH/CentOS, and so you should use it" - as suprjami points out in his comment, he disregards most of his own points.
We use ubuntu for general purpose stuff, and alpine linux for more task-specific stuff - especially for container images. I have a bitter resentment towards rpm and yum hasn't done anything to make that go away. We don't like systemd, and are evaluating alternatives to ubuntu for that reason.