OpenIndiana: Community-Driven Illumos Distribution
openindiana.org
openindiana.org
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.
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
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
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.
> 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.
It is odd to boot it and see sendmail running from my native ksh93 root login.
using our package manager pkgsrc [https://pkgsrc.smartos.org/], you can install the MTA of your choice too!
So OpenIndiana has MATE and a couple of other options; Tribblix has about 30 desktop options; OmniOS you can install a desktop stack from pkgsrc.
Not sure if characterizing this as a fork is right. OpenSolaris (aka project Indiana) was the development version for Solaris 11.
Between late 2010 and 2012 I was playing with OI in a VM, and I can read a lot of excitement about it in my old emails :)
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.
Thanks for the PDF suggestion, I didn't think about looking for a book to get more context!
I learned a lot about Plan 9 by following a boot camp ran by SDF, I wish there was something like that to promote Illumos distributions
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.
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'll keep the small "i" in mind as well. :)
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.
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.
So from that standpoint it makes sense that they acquired it. They probably just didn't care about any non-Oracle users.
I wouldn't bet on that. Their Linux distro is a RHEL clone, but they have total control over Solaris.
https://www.zdnet.com/article/sun-set-oracle-closes-down-las...
I guess my question - I honestly have no idea and it's been years since I paid attention to Oracle or Sun's inner workings - is whether they're still developing SPARC hardware? The Google seems to say no, they've abandoned that... so I'm inclined to agree with you guys now.
I don't have a ton of nostalgia left for Sun stuff, but still... what a waste.
Their core reference platform moved from SPARC to Linux + X86 almost a decade ago. They pushed Exadata down everyone's throat at the time with their "it'd be too bad if we have to audit you unless you buy this new hardware" approach.
Sadly it is still years away on ARM and x86, Linux and BSD systems.
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.
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...
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).
[citation needed]
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.
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.
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.
To be short, it is opensource implementation of SUN Solaris OS. I don't know if it is already developed much.
I did a sort of review 3Y ago. It may throw some light:
https://www.theregister.com/2022/12/07/new_version_of_openin...
Oracle bought Sun and killed OpenSolaris. There is a continuing community fork called illumos. OpenIndiana is a distro of illumos.
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)
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.
The repo has a list of packages
Times I've heard it's time to migrate from with no one willing too and so the swamp just becomes swampier.
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.
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.
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.
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:
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".
This is really good for them, and I do hope they do succeed, but if it's a matter of an individual developer deciding on their optimal area of systems expertise, this is still putting too many eggs in a single basket.
The reality is that even Oxide.Computer clients would still be running Linux on their machines.
So, as an engineer, you're literally shooting yourself in the foot by NOT moving to Linux, when everyone else has.
I think it's a standard sunk cost fallacy here. Some of us continue using BSD, but others switch to Linux and go work at Google, Meta, Amazon and even Oracle and Netflix, at 2x to 4x+ the income we get as BSD and Solaris devs.
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!
Linux is too bloated.
I mean, even if you get a job at Oxide Computer, what do you do if you decide to move on because of your manager or whatnot?
Literally all the sectors are Linux now, you'd think BSD would be huge in embedded because of licensing, yet it's still Linux. OpenWrt is Linux, Android is Linux, OpenIPC is Linux, not to mention the entire data-centre stacks being all Linux now.