tnx! it will be useful not only from a nethserver perspective but also to decide what to do with the other servers… i will just add
springdale as RH clone (http://springdale.math.ias.edu/ also if it seems a too small project)
and opensuse leap /sles
2 Likes
Yeah, as long as kernel modules aren’t required, you can run a lot of Linux binaries under FreeBSD. But it still is a limit there.
They may “suggest” reinstallation, but I’ve upgraded in place between major versions of all of them–only CentOS 6-7 [edit: actually 5-6] required reinstallation.
I know I am chiming in rather late on this, I have been following this pretty closely as it has a high degree of impact in my own job and what the company I work for will be doing for customers.
Like to give a shout out to @danb35 for the comprehensive comparison matrix, its definitely helpful.
I do think that the features that NethServer presents should most definitely be OS/Distro independent, so it should not matter if its a rpm or deb or apk based distro.
The RHEL Licensing costs is just too prohibitive and therefore isn’t much of a viable option.
CentOS Streams is potentially not viable for production workloads and that is by their own admission, so in my books, that puts it out of contention.
It is great to hear about other community driven distributions appearing since the announcement, but until they have proven themselves in production environments for at least 1-2 years, I don’t view them as viable options.
So it really comes down to Oracle Linux, Debian, or FreeBSD.
For Oracle Linux, for some reason this does not sit well with me - not entirely sure why as yet.
For Debian, its good, stable and rock solid. If you were going to choose another Distro like Mint or Ubuntu, you might as well have chosen Debian and I suspect that Debian would offer better stability on multiple fronts as opposed to the other Debian-based Distros.
For FreeBSD, I know that its good and fast and other Distros like MacOS is based off it. I have tried it out once or twice and to wrap my brain, but didn’t put enough concentration and effort into sticking through it and learning it. For me I find it a little difficult and will definitely put new-comers off. But then, if you gave it a super-brilliant installer and a smashing web interface so that the CLI was really not needed for 99% of the time, then this would not be an issue.
This looks like its going to be a very interesting year indead.
2 Likes
Well, they can be, but there’s a lot of friction in switching. I’m not aware of any inherent reason that Neth couldn’t run on any Unix-y OS, though the firewall would probably be the biggest sticking point. But there are a number of lesser, but still very significant, issues involved:
- Packages, and versions of packages, are very different from one distro to the next. It’s unlikely we’d need to make our own Apache package regardless of distro, but the available version would likely be different–which could require significant changes to our config templates. And the package name could very well be different, meaning some research would be needed to find the equivalent. Multiply this by a couple of thousand.
- File locations are different–unlike FreeBSD, Linux doesn’t enforce any particular filesystem hierarchy, so different distros put things in different places. Again, this can be overcome, but it takes some work.
- The build/packaging/distribution process is different. On the one hand, once it’s figured out, this can be automated to a large degree; on the other hand, it’s probably the hardest to figure out.
2 Likes
Awesome comparison post @danb35. Spelled it out nicely.
I have said this before, but I can not imagine the Netherserver devs will ever again have a better reason, opportunity and just as importantly the time to make a relatively transparent, to the enduser, distribution switch than now. Just about everything, though almost all unintended, has lined up. My preference would be for them to choose to do so now and not base the next nethserver version on the hope the choice will not be made for you later. When they will almost certainly have less time to do it as slowly, without the benefit of multiple years forethought, planning and execution.
3 Likes
One of the issues eg on FreeBSD would be FreePBX… It would not be easy to get all those DAHDI stuff on the Kernel.
I did use the Linux compatibility to run the Linux WLan stuff, using drivers written exclusively for Windows. That’s fairly close to Kernel and - it worked!
My 2 cents
Andy
2 Likes
An important “feature” that we need is hardware compatibility and certification.
Unless virtualization and containers will rule the world. 
I think we can rule out FreeBSD, HW support is not as good as Linux.
Moving to deb packages would probably involve more work, but it can be evaluated.
At the moment, Stream is not very different from CentOS, we will keep an eye on it.
Maybe we can make NethServer become distro-independent.
6 Likes
Agree on freebsd.
Distro-independent would be awesome, but almost certainly harder longterm to support than a switch or sticking with a RHEL clone. Would give you a lot of flexibility.
Hi Filippo
FreeBSD and hardware support can’t compare with Linux, if you look at the full bandwidth of hardware out there. But if you look at “server” canditates, even using better PCs, the comparison gets more and more even. FreeBSD can run on almost all server capable systems, and I’m including better PCs here.
But as I mentionned, FreePBX and Asterisk can easily present difficulties! And there are a lot of other stumblestones, although the latest WLan drivers are not a critical issue, even if FreeBSD does not have them. Even Linux can’t beat a real WLan-AP. (Not even wrt).
Debian is my favorite candidate, now that Centos/RH based isn’t really feasible anymore. It did serve me good the last years…
But then, anything I couldn’t easily install on Centos7/Neth, I used Debian, and, like Centos, it has never let me down!
Stream is a developers playground, not a server platform. That’s what “they” say…
In 2-3 days Stream may still be close, at the end of the year, it’ll already be light years away!
My 2 cents
Andy
4 Likes
That’s a great constructive approach thanks for that.
Agree, we can work on that a find a good solution. I guess that @giacomo have a few ideas about that.
Such a great comparison, thanks for that 
We have time, CentOS 7 is going to be supported for a few years.
1 Like
something else which I guess it should be read, also because as already said by others, we have some time…
https://blog.centos.org/2020/12/centos-stream-is-continuous-delivery/
1 Like
I don’t think this is really much of an issue for any hardware that would be remotely suitable for a server. But with that said, I really don’t see FreeBSD as a live option.
Yes sir, you have plenty of time to put this together, which is why I hope you go in the debian direction and take advantage of this long window while it is your choice and can be done slowly and well.
My comment on time was more in regards to how much more time you will spend supporting Nethserver8 after it is released if is is distro-independent relative to one target and one set of patches options and abilities upstream that are not released at multiple different times and with different capabilities.
1 Like
@danb35 thanks for your more than extensive insights. And I can agree with your points. As a possible alternative base I would like to add Springdale
This is also a rebuild of RHEL (just as CentOS Linux).
IMO it is important to choose a base that is trusted. (imo oracle does not fit this demand)
You can’t have 100% certainty, but part of the choice is investigating how the base distro is governed. What failsaves are built in the structure of the entity providing the distro.
For now it is way too soon to decide, especialy if we want initiatives like Rocky Linux a fair chance to see how they develop. I have spent quite some time the past days in what is happening in the Rocky Linux project and I must say: they are on a fasttrack.
As for non EL options:
Debian could be an option. Ubuntu absolutely not (as mentioned by Dan, because of the decisions made by Canonical the past few years)
Then there is FreeBSD. In name: super stable and robust. I really can’t say much about it since I never used it. (well, I used pfSense, but that was more as a ‘user’ so I don’t call that 'using FreeBSD)
I repeat myself saying to keep options open. Although IBM/RH is very low on my list of options. Even if they revert this decision (what I doubt they will do) they have lost a lot of trust from the CentOS community and a lot of CentOS users. And I agree that swapping an EL OS for a rolling release playgarden is NOT what I would want and IMO NethServer should not want to consider this.
7 Likes
Why not to make a multidistro control panel like webmin?
This would perfectly fit in the NS timeframe: if they need 6 months to become production ready, there is another 1,5years or so to prove they are there to stay. Then we are end 2022, 2 years before CentOS7 EOL. Then we have 2 years for NS8 based on a community driven CentOS Linux alternative.
IMO that should not be written off as an option.
2 Likes
That probably will be the case.
I am too noob to bring useful hints to this discussion, but just curious to ask two questions:
- Why I see so low enthusiasm on rebasing on Ubuntu LTS? I always dreamed of making Nethbuntu if only I had the capabilities
- When I read distribution independent, do you refer to something like Bitnami LAMP stack and family?
Hi
I for one dislike the snap system. I just don’t trust it, nor the shop. Also, it brings insecurities to the system. That’s not all, but that’s the major itch!
Seconds, We aren’t the only group… Mint Linux, with a fairly large user base, has for years built on Ubuntu. Since a couple of years now, just to be on the safe side, they are now providing two distros, almost identical. One still based on Ubuntu, the second on Debian. Just in case, you never know, nor can corporations always be trusted…
Hope that answers at least one of your two questions…
My 2 cents
Andy
2 Likes
I saw your previous posts and never thought of this double channel as a backup solution, but more like an “exercise”, but I am not a dev. Also because from what I read LMDE is a kind of rolling distro so I felt it like a kind of experimental. From a non-dev perspective snaps are a dream: I mean, I work with a software which requires some particular 32bit libs and it is a bit tricky to install in modern distro (it was a nightmare in the previous Debian, did not tried on Buster). In facts this is why I originally knew Nethserver. From what I understand I could pack those libraries in a snap and avoid problems with my software dependencies and also these lines would not interfere with dependencies of other software, like a kind of sandbox.