> There are a small handful of illumos components for which source code is not available. Over time, we have replaced most of the closed source components from the Sun era with new open source versions. This work is ongoing
Anyone run this on top of KVM? I tried a while back but I don't know KVM or Qemu well and all the examples were ancient and didn't work for me. I'd love to play around with Solaris again (last I used it was Solaris 8 on an Ultra-1) but I don't have any decent hardware I'm willing to dedicate to the task.
If we are going to talk about 35 years of illumos linkage, we as well go all the way. Using you notation: Solaris <—- SVR4 [0] <—- AT&T System III [1] <—- Bell Labs Version 7 Unix [2] <—- the birth of Unix at Bell Labs in 1969 [3].
Besides the direct lineage, it’s interesting to see cross pollination between different operating systems over the years. Like BSD’s socket interface spreading everywhere (including Windows), ZFS from OpenSolaris to FreeBSD & Linux, then bhyve from FreeBSD to illumos.
You missed the branch there for Solaris 1.x (aka SunOS) which was a BSD derivative.
One of the nifty things about Solaris was all the compatibility code they had. You could set environment variables and compile against BSD, SysV, and a variety of other "standards" - they tried to make it so that anything that could compile on a UNIX would compile on Solaris.
> OpenIndiana obtains its name from Project Indiana, an open source effort by Sun Microsystems (now Oracle Corporation) to produce OpenSolaris, a community developed Unix-like distribution based on Sun Solaris. Project Indiana was led by Ian Murdock, founder of the Debian Linux Distribution.
(I never understood the naming either)
But here's an ArsTechnica article from 2007 talking more about those origins from back when Sun was still trying to win back marketshare from Linux. It had long since lost that war, but was still trying to stay relevant.
> OpenIndiana takes its name from Project Indiana, the internal codename for OpenSolaris at Sun Microsystems before Oracle’s acquisition of Sun in 2010.
OpenIndiana is more of a downstream distro than a fork of illumos, much like how Ubuntu is a Debian-based distro. Whereas illumos itself is a fork of OpenSolaris.
Open Maryland, Not Indiana :) (So named in this case because the company that released it, OmniTI, is based in Maryland. Source: worked there at the time) It's a good OS, as I imagine the other Illumos derivatives are, but sometimes the relatively small size of the community can be felt, e.g. in breadth/depth of available third party packages and update availability/timelines.
I've been using OpenIndiana for a file server in my homelab since it came out and it's been quietly doing its job ever since without much issues. Coming from Linux it's not easy to find the equivalent commands of what I do on my other servers but it's also what I like about this project, it's another flavor of Unix to learn.
Sun had excellent documentation. But of course Oracle crapped all over it...
If you can find documentation for Solaris 9 I imagine a lot of it would be the same. And there's the Solaris 8 Administration book from O'Reilly - I'm sure you can find PDFs of that floating around.
At Oxide, we have our own illumos (in my understanding, you're supposed to lowercase the i) distribution, discussed on HN a while back https://news.ycombinator.com/item?id=39178521
I think a lot of decisions that eventually lead to the Oracle buyout were all pretty bad and Oracle itself being where good ideas go to die. As bad as MS is at extracting value out of its windows users, Oracle seems to fleece it's enterprise customers far, far worse. I don't think I would ever choose Oracle or IBM for anything.
It would be interesting to see a little more diversity in common operating systems in the wild though. Linux has pretty much taken over the server space, and iOS/Android have split the more common usage outside that, with what's left of desktop still mostly Windows.
I still think there's opportunity for something like Flutter as a cross-platform library that actually works with multiple backing languages.
It still is, but at this point I'm not sure why anyone would pick it over Linux for something new. All the killer Solaris features (ZFS, dtrace, zones, SMF) have good enough Linux equivalents (OpenZFS, eBPF, containers, systemd) and I'm not aware of any usecases where illumos outperforms Linux anymore.
OpenIndiana also has the problem that every commercial illumos user is using it for some niche purpose (networking infrastructure, storage appliance, that sort of thing) so it's basically up to a few unpaid volunteers working in their free time to adapt it for general desktop use. I'm not sure what the state of stuff like audio support or accelerated graphics looks like if you're on modern hardware.
Solaris was the development OS for Oracle for years. I presume that's now shifted to their own linux distro but for many years that was the case, to the point that if you were a serious Oracle customer you ran it on Solaris (and Sun hardware) because all the bug fixes and updates came out first for that platform.
So from that standpoint it makes sense that they acquired it. They probably just didn't care about any non-Oracle users.
And DTrace. And NFS/NIS. And SunRPC. And OpenOffice (sort of - they bought the company that made it and then open sourced it). And... you get the idea.
That's what I loved about Sun, really. They strive for a leadership role in the UNIX world by actually leading, instead of just trying to dominate. No company is perfect, but Sun was better than most.
I was sad to see them go, but with Windows NT taking over corporate and Linux taking over networking, they just didn't have a place. They kept pushing "The network is the computer" at a time when PCs were cheap. If only they'd held out until the cloud craze hit...
Yeah, except it exists since 2015, and while it is great that macOS is getting them, which I am perfectly aware of, it isn't what I meant as "Sadly it is still years away on ARM and x86, Linux and BSD systems.", because it isn't neither a Linux system, nor a regular BSD.
Apple OSes don't count for the BSD group I was talking about, because they only took whatever NeXT felt like back at the time, and there have been very few updates other than those required for UNIX certification.
While on Linux, you might be tempted to point out Android, however NDK alone hardly makes it a GNU/Linux system.
I'm really disappointed that Solaris picked a "let's screw Linux" license, relegating some otherwise interesting technologies to only run on Solaris, on permissive OSes like BSD, and on systems that don't care about license compliance.
There was nothing about screwing anyone involved in the choice of the license. The license had to be something all the copyright and license holders were prepared to accept (and getting them to accept CDDL was hard enough, not everyone did hence the few closed components).
Our belief was that Linux would be unlikely (and unwise as the overall system architecture is sufficiently different that it would be hard to port) to take the code. We expected - and encouraged - the concepts to be taken (as with the slab memory allocator).
It's not hard to find Linux users shit-talking Solaris - or "Slowaris" as it was known at the time. Just browse the Slashdot archives.
Reaching feature parity and taking Solaris' market share was an often cited goal by people in the Linux community, and one they achieved.
That said, I don't remember much actual hate directed at Sun in the way that Microsoft got. As far as companies go, Sun was a better member of the UNIX community than most.
I don't know about Illumos but setting up a Solaris 10 machine to provide block devices through multi path iSCSI was a breeze with all the tooling it had. No config files needed it was all command line.
It was for a university lab and it was the only ever contact I had with Solaris 10 and later versions or forks. I was mildly interested once since it had KVM but support was Intel only so that kept me away.
I remember some hypervisor and storage products based on Illumos like Nexenta and Soylent OS. I'm guessing those projects faded into obscurity.
Which is a shame. A lot of people was optimistic about OpenSolaris when it came out but Oracle gutted it.
AFAIK the interesting thing about these Illumos Forks is ZFS support together with a kernel based SMB server ability using it as file server.
After enabling a share, connect as root and set NTFS alike ACL via Windows.
The SMB server is using Windows SID (not uid/gid) as ACL and Windows-ish SMB groups on Illumos (not Unix groups)
OI uses IPS packaging, which is the same packaging used by Solaris 11 (some find it over engineered). Tribblix, on the other hand, is also an illumos-based distro but is based on Solaris 10 (and prior’s) SVR4 packaging which is managed via a utility called “zap”.
Both are good distributions, but I strongly encourage you to try Tribblix if OI is problematic; for whatever reason the latest OI installers do not seem to include the same amount of driver support as Tribblix, in my experience.
Solaris, basically. Illumos was derived from the old OpenSolaris project. I don't know how much it's diverged, as I don't really follow this much. But that's more or less what it is.
Solaris is still developed and sold by Oracle, and expected to be supported for a very long time. The current version, Solaris 11.4, is planned to be supported until 2037 at the earliest. The previous major version, Solaris 10, was released in 2005 and is supported until 2027. So they are unlikely to sell the trademark any time soon.
Who are still using Solaris though? And I say that as a previous SunOS and Solaris user for a couple of decades. Great technology. That's where I first got to play with dtrace and containers/zones.
Is there really a valid use case these days if you are not a hobbyist/fanboy or have existing Solaris workload compared to FreeBSD (or Linux if licensing isn't an issue)? ZFS seems quite well-supported on those.
Solaris is an actual System V, not a clone like Linux distros often are or a BSD.
It's got excellent support for compiling software written for different UNIX systems.
It does a lot of things with commands that other systems do with config files.
But no, there really aren't any compelling reasons why a non-hobbyist would be interested. If Oracle had continued development on OpenSolaris then yes, but they have little interest in their own version of Solaris.
Not that I'd really use it, but the integration of ZFS with the package manager on the one side and things like NFS on the other side made it really nice to use.
If a package update failed for whatever reason one could simply revert to a previous boot environment and configuring permissions on NFS exports worked directly integrated with ZFS. Really nice.
On other platforms there is a lot more fiddling around required.
Everything is converging on Linux these days, and even the majority of the people who used to promote Solaris, DTrace and ZFS, have seemingly moved on, mostly to Linux, somewhat to FreeBSD, too, per Brendan Gregg:
In the BSD land, things aren't that much better, either.
Netgate, the makers of pfSense, are now using Ubuntu for Netgate TNSR product.
iXsystems, the descendants of BSDi, have moved FreeNAS / TrueNAS from FreeBSD to Linux. They're basically d/b/a TrueNAS now, and it's all Linux now.
You used to need Solaris or illimos or FreeBSD for production-ready ZFS support, but now OpenZFS is provided exclusively for Linux and FreeBSD; note that Linux already comes first in the title; it would seem like it's only a matter of time before FreeBSD support may follow Solaris and illumos.
Joyent, the commercial shepherds of OpenSolaris descendants like SmartOS, were acquired by Samsung, but the entire Solaris part of the equation, including Triton DataCenter orchestration, were subsequently offloaded to mnx.io, a tiny cloud hosting provider based out of a small town in Michigan. (Frankly, without Triton, I don't even understand what remains of Joyent at Samsung? Just the physical servers with the third-party software? It's basically just a name for Samsung's data centre ops and their presumably-Linux-based Private Cloud?)
Apple used to use NetBSD for AirPort WiFi routers, but the whole router line has been discontinued. (I thought Apple actually already dropped NetBSD but couldn't find a source right now.)
Last not least, DJB used to run OpenBSD, then FreeBSD, but then switched to Ubuntu after possibly being annoyed that too many steps were required to make FreeBSD work as a desktop: http://cr.yp.to/unix/feedme.html (my fav is that in FreeBSD the audio doesn't work unless you recompile the kernel). I think he initially may have abandoned OpenBSD because it was crashing too often for undetermined causes: http://cr.yp.to/serverinfo.html.
FreeBSD is still used by Netflix OpenConnect Appliance, which is a huge win, but feels like too many eggs in a single basket:
"There’s a big difference between mostly dead and all dead. ... Now, mostly dead is slightly alive. Now, all dead…well, with all dead, there’s usually only one thing that you can do."
I personally believe that illumos will survive for decades, and it very well could rise again for those users that want/need robust stability.
One particularly notable design win was Oxide Computer Company as their hypervisor. They published their reasoning for choosing illumos here:
Also Oxide just raised a $100 million series B with their illumos based product [1].
While I respect Brendan I think the arguments he makes are kinda weak. For example take OpenZFS. OpenZFS on Linux or FreeBSD isn't that great;
OpenZFS on Linux is only an option if you want to support the compatibility. As long as OpenZFS can't have (parts) inside the main Linux kernel source tree there is going to be breakage on updates. Which means manually testing and maintaining updates. Or you have to confine yourself to Ubuntu because they are of the opinion you can combine the CDDL and GPL. Don't think Ubuntu indemnifies you against an Oracle lawsuit though.
You could use FreeBSD as an alternative. FreeBSD is great, but lacks a lot of functionality. For example a good I/O scheduler is missing in FreeBSD. The FreeBSD I/O scheduler will basically just do what is most advantageous for it. If you have competing I/O workloads which you want to serve "fairly" there is no way to do that on FreeBSD. Basically the load which "pulls the hardest gets the most".
Great episode! The Dell PERC "parity issue" "you're the only customer having this issue" story is hilarious, reminds me of my own experience getting support over the years as a non-Windows user!
Sadly, today, "non-Windows" is simply being replaced with "non-Linux", so, if you run BSD or Illumos, it's still the same issues.
Oxide Computer is great, I really do hope they succeed, because it does seem like an uphill battle!
> illumos-closed package, containing binary blobs still necessary to build illumos-gate, was added
https://docs.openindiana.org/release-notes/2016.10-release-n...
> There are a small handful of illumos components for which source code is not available. Over time, we have replaced most of the closed source components from the Sun era with new open source versions. This work is ongoing
https://illumos.org/docs/developers/build/#getting-the-close...
> From this, however, project founder Garrett D'Amore took the last drop of the gate and announced illumos in mid-2010.
https://illumos.org/docs/about/history/
What does "took the last drop of the gate" mean?
The last code pull from Sun/opensolaris before they started completely diverging.
Anyone run this on top of KVM? I tried a while back but I don't know KVM or Qemu well and all the examples were ancient and didn't work for me. I'd love to play around with Solaris again (last I used it was Solaris 8 on an Ultra-1) but I don't have any decent hardware I'm willing to dedicate to the task.
As a Hoosier... I had to check the relevance... and learned it's basically a fork^3 of Solaris, a Sun operating system.
OpenIndiana[3] <-- Illumos[2] <-- OpenSolaris[1] <-- Solaris[0]
Note: I guessed here at <-- meaning fork of... any other options I should have used instead?
[3] https://en.wikipedia.org/wiki/OpenIndiana
[2] https://en.wikipedia.org/wiki/Illumos
[1] https://en.wikipedia.org/wiki/OpenSolaris
[0] https://en.wikipedia.org/wiki/Oracle_Solaris
If we are going to talk about 35 years of illumos linkage, we as well go all the way. Using you notation: Solaris <—- SVR4 [0] <—- AT&T System III [1] <—- Bell Labs Version 7 Unix [2] <—- the birth of Unix at Bell Labs in 1969 [3].
Besides the direct lineage, it’s interesting to see cross pollination between different operating systems over the years. Like BSD’s socket interface spreading everywhere (including Windows), ZFS from OpenSolaris to FreeBSD & Linux, then bhyve from FreeBSD to illumos.
[0]: https://en.m.wikipedia.org/wiki/UNIX_System_V#SVR4
[1]: https://en.m.wikipedia.org/wiki/UNIX_System_III
[2]: https://en.m.wikipedia.org/wiki/Version_7_Unix
[3]: https://en.m.wikipedia.org/wiki/Unix
You missed the branch there for Solaris 1.x (aka SunOS) which was a BSD derivative.
One of the nifty things about Solaris was all the compatibility code they had. You could set environment variables and compile against BSD, SysV, and a variety of other "standards" - they tried to make it so that anything that could compile on a UNIX would compile on Solaris.
From: https://docs.openindiana.org/misc/openindiana/#what-is-the-o...
> Why is it called OpenIndiana?
> OpenIndiana obtains its name from Project Indiana, an open source effort by Sun Microsystems (now Oracle Corporation) to produce OpenSolaris, a community developed Unix-like distribution based on Sun Solaris. Project Indiana was led by Ian Murdock, founder of the Debian Linux Distribution.
(I never understood the naming either)
But here's an ArsTechnica article from 2007 talking more about those origins from back when Sun was still trying to win back marketshare from Linux. It had long since lost that war, but was still trying to stay relevant.
https://arstechnica.com/uncategorized/2007/07/understanding-...
Illumos based OS's have been kicking around a lot longer than I anticipated.
Not sure if Ian Murdock had a say in the naming but his family was from Indiana.
I think what's important is that the name matches on /ian/.
> OpenIndiana takes its name from Project Indiana, the internal codename for OpenSolaris at Sun Microsystems before Oracle’s acquisition of Sun in 2010.
https://en.wikipedia.org/wiki/OpenIndiana#History
Ian Murdock, who grew up in Indiana, joined Sun Microsystems in early 2007 to lead Project OpenIndiana.
Yeah, just thought the wikipedia kind of cleared that up, Ian Murdock was an interesting fellow.
I have also used SmartOS, which imports KVM from Linux.
It is odd to boot it and see sendmail running from my native ksh93 root login.
https://www.tritondatacenter.com/smartos
https://en.wikipedia.org/wiki/SmartOS
While we do still ship KVM on SmartOS, we've moved development over to bhyve - https://docs.smartos.org/bhyve/
using our package manager pkgsrc [https://pkgsrc.smartos.org/], you can install the MTA of your choice too!
OpenIndiana is more of a downstream distro than a fork of illumos, much like how Ubuntu is a Debian-based distro. Whereas illumos itself is a fork of OpenSolaris.
There is OmniOS[1] as well, fits next to OpenIndiana.
https://omnios.org.
Open Maryland, Not Indiana :) (So named in this case because the company that released it, OmniTI, is based in Maryland. Source: worked there at the time) It's a good OS, as I imagine the other Illumos derivatives are, but sometimes the relatively small size of the community can be felt, e.g. in breadth/depth of available third party packages and update availability/timelines.
> OpenSolaris[1] <-- Solaris[0
Not sure if characterizing this as a fork is right. OpenSolaris (aka project Indiana) was the development version for Solaris 11.
I've been using OpenIndiana for a file server in my homelab since it came out and it's been quietly doing its job ever since without much issues. Coming from Linux it's not easy to find the equivalent commands of what I do on my other servers but it's also what I like about this project, it's another flavor of Unix to learn.
Sun had excellent documentation. But of course Oracle crapped all over it...
If you can find documentation for Solaris 9 I imagine a lot of it would be the same. And there's the Solaris 8 Administration book from O'Reilly - I'm sure you can find PDFs of that floating around.
Wow. Illumos / Solaris is still kicking, eh?
Really too bad Solaris didn’t stick around and was so horribly mismanaged by Sun.
Solaris and Vax/VMS is where I started my career decades ago, and still brings back memories.
It still is!
At Oxide, we have our own illumos (in my understanding, you're supposed to lowercase the i) distribution, discussed on HN a while back https://news.ycombinator.com/item?id=39178521
Very cool! Thanks for sharing that.
I'll keep the small "i" in mind as well. :)
I think a lot of decisions that eventually lead to the Oracle buyout were all pretty bad and Oracle itself being where good ideas go to die. As bad as MS is at extracting value out of its windows users, Oracle seems to fleece it's enterprise customers far, far worse. I don't think I would ever choose Oracle or IBM for anything.
It would be interesting to see a little more diversity in common operating systems in the wild though. Linux has pretty much taken over the server space, and iOS/Android have split the more common usage outside that, with what's left of desktop still mostly Windows.
I still think there's opportunity for something like Flutter as a cross-platform library that actually works with multiple backing languages.
It still is, but at this point I'm not sure why anyone would pick it over Linux for something new. All the killer Solaris features (ZFS, dtrace, zones, SMF) have good enough Linux equivalents (OpenZFS, eBPF, containers, systemd) and I'm not aware of any usecases where illumos outperforms Linux anymore.
OpenIndiana also has the problem that every commercial illumos user is using it for some niche purpose (networking infrastructure, storage appliance, that sort of thing) so it's basically up to a few unpaid volunteers working in their free time to adapt it for general desktop use. I'm not sure what the state of stuff like audio support or accelerated graphics looks like if you're on modern hardware.
I’m of the opinion that the acquisition of Sun by Oracle was the worst possible outcome; it guaranteed that Solaris would decline.
Solaris was the development OS for Oracle for years. I presume that's now shifted to their own linux distro but for many years that was the case, to the point that if you were a serious Oracle customer you ran it on Solaris (and Sun hardware) because all the bug fixes and updates came out first for that platform.
So from that standpoint it makes sense that they acquired it. They probably just didn't care about any non-Oracle users.
> Solaris was the development OS for Oracle for years. I presume that's now shifted to their own linux distro
I wouldn't bet on that. Their Linux distro is a RHEL clone, but they have total control over Solaris.
Given that they killed the roadmap and admitted there will never be a Solaris 12 after laying off 90% of the core developers - I would bet on that.
https://www.zdnet.com/article/sun-set-oracle-closes-down-las...
Still mad at IBM for screwing the pooch on that one. They basically handed the company to Oracle.
Did IBM consider purchasing Sun?
Yes, although IBM only wanted Java and was planning to cancel SPARC and maybe Solaris.
Makes sense given that IBM already had the POWER arch & AIX, which was based on BSD rather than SysV for Solaris.
Kind of like Oracle did.
It (Solaris) was also the origin of ZFS, if I'm not mistaken.
And DTrace. And NFS/NIS. And SunRPC. And OpenOffice (sort of - they bought the company that made it and then open sourced it). And... you get the idea.
That's what I loved about Sun, really. They strive for a leadership role in the UNIX world by actually leading, instead of just trying to dominate. No company is perfect, but Sun was better than most.
I was sad to see them go, but with Windows NT taking over corporate and Linux taking over networking, they just didn't have a place. They kept pushing "The network is the computer" at a time when PCs were cheap. If only they'd held out until the cloud craze hit...
Not only it is still around from Oracle, it is the only production UNIX with hardware memory tagging enabled.
Sadly it is still years away on ARM and x86, Linux and BSD systems.
Like https://www.usenix.org/conference/usenixsecurity23/presentat... ?
Yeah, except it exists since 2015, and while it is great that macOS is getting them, which I am perfectly aware of, it isn't what I meant as "Sadly it is still years away on ARM and x86, Linux and BSD systems.", because it isn't neither a Linux system, nor a regular BSD.
Apple OSes don't count for the BSD group I was talking about, because they only took whatever NeXT felt like back at the time, and there have been very few updates other than those required for UNIX certification.
While on Linux, you might be tempted to point out Android, however NDK alone hardly makes it a GNU/Linux system.
I'm really disappointed that Solaris picked a "let's screw Linux" license, relegating some otherwise interesting technologies to only run on Solaris, on permissive OSes like BSD, and on systems that don't care about license compliance.
There was nothing about screwing anyone involved in the choice of the license. The license had to be something all the copyright and license holders were prepared to accept (and getting them to accept CDDL was hard enough, not everyone did hence the few closed components).
Our belief was that Linux would be unlikely (and unwise as the overall system architecture is sufficiently different that it would be hard to port) to take the code. We expected - and encouraged - the concepts to be taken (as with the slab memory allocator).
Some Linux people were saying "let's screw Solaris" first and Sun people are only human so that's the result.
> Some Linux people were saying "let's screw Solaris" first
[citation needed]
It's not hard to find Linux users shit-talking Solaris - or "Slowaris" as it was known at the time. Just browse the Slashdot archives.
Reaching feature parity and taking Solaris' market share was an often cited goal by people in the Linux community, and one they achieved.
That said, I don't remember much actual hate directed at Sun in the way that Microsoft got. As far as companies go, Sun was a better member of the UNIX community than most.
Ahh man the nostalgia. I miss Sun. Oracle buys and ruins everything.
If only Broadcom would buy Oracle...
I don't know about Illumos but setting up a Solaris 10 machine to provide block devices through multi path iSCSI was a breeze with all the tooling it had. No config files needed it was all command line.
It was for a university lab and it was the only ever contact I had with Solaris 10 and later versions or forks. I was mildly interested once since it had KVM but support was Intel only so that kept me away.
I remember some hypervisor and storage products based on Illumos like Nexenta and Soylent OS. I'm guessing those projects faded into obscurity.
Which is a shame. A lot of people was optimistic about OpenSolaris when it came out but Oracle gutted it.
Oracle ruined Sun Microsystems for sure, incl. OpenSolaris.
I finally got my hands on using a Solaris machine at my last job to upgrading the backup software.
All I could explain the OS as, was as a swamp. It's wants to work but doesn't work as it should and you always were knee-deep in sludge too.
AFAIK the interesting thing about these Illumos Forks is ZFS support together with a kernel based SMB server ability using it as file server.
After enabling a share, connect as root and set NTFS alike ACL via Windows. The SMB server is using Windows SID (not uid/gid) as ACL and Windows-ish SMB groups on Illumos (not Unix groups)
See https://illumos.org/man/8/smbadm
Look to figure out what is it: https://docs.openindiana.org/misc/openindiana/
To be short, it is opensource implementation of SUN Solaris OS. I don't know if it is already developed much.
I've spent two minutes look at this website and I still have no idea what this project is.
https://docs.openindiana.org/misc/openindiana/ looks like what you wanted, was not straightforward to get there from the homepage.
I have to say, that I really loved OpenSolaris. I really have to try OpenIndiana. Can anyone tell me about their package manager?
OI uses IPS packaging, which is the same packaging used by Solaris 11 (some find it over engineered). Tribblix, on the other hand, is also an illumos-based distro but is based on Solaris 10 (and prior’s) SVR4 packaging which is managed via a utility called “zap”.
Both are good distributions, but I strongly encourage you to try Tribblix if OI is problematic; for whatever reason the latest OI installers do not seem to include the same amount of driver support as Tribblix, in my experience.
Thanks! Tribblix seems interesting at a quick glance. Does it have i3? Or is there a list of packages that I can see online?
It has 30 desktop options, including i3
The repo has a list of packages
http://pkgs.tribblix.org/
I found it, thank you. :)
I can’t figure out what this is from anything on the site. I suppose I need to know what Illumos is? My best guess is that it is an OS?
Solaris, basically. Illumos was derived from the old OpenSolaris project. I don't know how much it's diverged, as I don't really follow this much. But that's more or less what it is.
Maybe the Perifractic equivalent of the Sun fanworld can buy the Solaris trademark and rename it!
Solaris is still developed and sold by Oracle, and expected to be supported for a very long time. The current version, Solaris 11.4, is planned to be supported until 2037 at the earliest. The previous major version, Solaris 10, was released in 2005 and is supported until 2027. So they are unlikely to sell the trademark any time soon.
Who are still using Solaris though? And I say that as a previous SunOS and Solaris user for a couple of decades. Great technology. That's where I first got to play with dtrace and containers/zones.
Banks, large institutions and other enterprises all seem to keep the lights on for Solaris.
Times I've heard it's time to migrate from with no one willing too and so the swamp just becomes swampier.
Is there really a valid use case these days if you are not a hobbyist/fanboy or have existing Solaris workload compared to FreeBSD (or Linux if licensing isn't an issue)? ZFS seems quite well-supported on those.
Solaris is an actual System V, not a clone like Linux distros often are or a BSD.
It's got excellent support for compiling software written for different UNIX systems.
It does a lot of things with commands that other systems do with config files.
But no, there really aren't any compelling reasons why a non-hobbyist would be interested. If Oracle had continued development on OpenSolaris then yes, but they have little interest in their own version of Solaris.
Not that I'd really use it, but the integration of ZFS with the package manager on the one side and things like NFS on the other side made it really nice to use.
If a package update failed for whatever reason one could simply revert to a previous boot environment and configuring permissions on NFS exports worked directly integrated with ZFS. Really nice.
On other platforms there is a lot more fiddling around required.
Everything is converging on Linux these days, and even the majority of the people who used to promote Solaris, DTrace and ZFS, have seemingly moved on, mostly to Linux, somewhat to FreeBSD, too, per Brendan Gregg:
Solaris to Linux Migration 2017 - https://news.ycombinator.com/item?id=15177118 - Sept 2017 (129 comments)
In the BSD land, things aren't that much better, either.
Netgate, the makers of pfSense, are now using Ubuntu for Netgate TNSR product.
iXsystems, the descendants of BSDi, have moved FreeNAS / TrueNAS from FreeBSD to Linux. They're basically d/b/a TrueNAS now, and it's all Linux now.
You used to need Solaris or illimos or FreeBSD for production-ready ZFS support, but now OpenZFS is provided exclusively for Linux and FreeBSD; note that Linux already comes first in the title; it would seem like it's only a matter of time before FreeBSD support may follow Solaris and illumos.
Joyent, the commercial shepherds of OpenSolaris descendants like SmartOS, were acquired by Samsung, but the entire Solaris part of the equation, including Triton DataCenter orchestration, were subsequently offloaded to mnx.io, a tiny cloud hosting provider based out of a small town in Michigan. (Frankly, without Triton, I don't even understand what remains of Joyent at Samsung? Just the physical servers with the third-party software? It's basically just a name for Samsung's data centre ops and their presumably-Linux-based Private Cloud?)
Apple used to use NetBSD for AirPort WiFi routers, but the whole router line has been discontinued. (I thought Apple actually already dropped NetBSD but couldn't find a source right now.)
Last not least, DJB used to run OpenBSD, then FreeBSD, but then switched to Ubuntu after possibly being annoyed that too many steps were required to make FreeBSD work as a desktop: http://cr.yp.to/unix/feedme.html (my fav is that in FreeBSD the audio doesn't work unless you recompile the kernel). I think he initially may have abandoned OpenBSD because it was crashing too often for undetermined causes: http://cr.yp.to/serverinfo.html.
FreeBSD is still used by Netflix OpenConnect Appliance, which is a huge win, but feels like too many eggs in a single basket:
Serving Netflix Video at 400Gb/s on FreeBSD [pdf] - https://news.ycombinator.com/item?id=28584738 - Sept 2021 (293 comments)
But that's about it. Linux has won.
"There’s a big difference between mostly dead and all dead. ... Now, mostly dead is slightly alive. Now, all dead…well, with all dead, there’s usually only one thing that you can do."
I personally believe that illumos will survive for decades, and it very well could rise again for those users that want/need robust stability.
One particularly notable design win was Oxide Computer Company as their hypervisor. They published their reasoning for choosing illumos here:
https://rfd.shared.oxide.computer/rfd/0026
https://www.youtube.com/watch?v=5P5Mk_IggE0&t=2556s
Also Oxide just raised a $100 million series B with their illumos based product [1].
While I respect Brendan I think the arguments he makes are kinda weak. For example take OpenZFS. OpenZFS on Linux or FreeBSD isn't that great;
OpenZFS on Linux is only an option if you want to support the compatibility. As long as OpenZFS can't have (parts) inside the main Linux kernel source tree there is going to be breakage on updates. Which means manually testing and maintaining updates. Or you have to confine yourself to Ubuntu because they are of the opinion you can combine the CDDL and GPL. Don't think Ubuntu indemnifies you against an Oracle lawsuit though.
You could use FreeBSD as an alternative. FreeBSD is great, but lacks a lot of functionality. For example a good I/O scheduler is missing in FreeBSD. The FreeBSD I/O scheduler will basically just do what is most advantageous for it. If you have competing I/O workloads which you want to serve "fairly" there is no way to do that on FreeBSD. Basically the load which "pulls the hardest gets the most".
[1] https://oxide.computer/blog/our-100m-series-b
> https://www.youtube.com/watch?v=5P5Mk_IggE0&t=2556s
Great episode! The Dell PERC "parity issue" "you're the only customer having this issue" story is hilarious, reminds me of my own experience getting support over the years as a non-Windows user!
Sadly, today, "non-Windows" is simply being replaced with "non-Linux", so, if you run BSD or Illumos, it's still the same issues.
Oxide Computer is great, I really do hope they succeed, because it does seem like an uphill battle!
Sony still uses FreeBSD for PlayStation.