As someone who worked for Sun in the 1990s, Solaris will always hold a special place in my memories. It ruled the enterprise UNIX world back then and I never would have imagined back then that it would ever fade into obscurity (as it is doing now).
Will I install it again? No. Like most everyone else, I've moved on to Linux and BSD.
But I'll spin up Solaris 2.5.1 on my SPARCbook or Solaris 10 on my Blade 1500 whenever I need a nostalgia kick.
I worked at Intel from 1988-1998 and when I started we had either AIX or SunOS terminals (a few folks had giant HPUX boxes). By 1996, everything, and I mean EVERYTHING, switched to Linux and VNC. It happened quickly. First we started hearing about Linux, then a quarter later all the remote machines were Linux, then the quarter after that they took away our workstations. There were a few HPUX boxes in the larger pools for extremely compute-heavy layout tasks, but everything else was RedHat.
Point is: I remember writing scripts for then main environment and every script had to work on all terminals. The three variations meant constantly armor-plating things with HPUX/AIX/SUNOS variable checks. I hated that. I was glad when RedHat came along and we stopped doing that.
(Also, back then it was a sign of being cool to have an 8514/A monitor, and not the B&W sun terminals.)
I work in enterprise software and around 2000 the combination of Solaris and Oracle database was the most rock solid out of all the various OS/DB combinations. I vaguely remember the threading model was weird and I think limited stack size per thread but overall it was my favorite combo because of how reliable it was.
The speed with which Linux killed Solaris however was astonishing.
I was at university in the late 90s. In exchange for buying us all (usually cheap) pizza, companies could come in and give an hour long pitch for why we should come work for them after graduation. A few big companies wanted more time so they'd cater higher end pizza or more elaborate catering as a bribe to get students to come to their presentation. Microsoft bough a large amount of Mellow Mushroom (the best pizza in our little college town) and did a two hour presentation on what it is like to work at Microsoft and why we should want to work there. Later in the semester Sun came to campus and catered their two hour presentation with chicken nuggets. But real issue was that the people and videos that were part of Sun's presentation didn't tell us much about Sun or why we should work there. It was a two hour long ranting whinefest about why no one should work at Microsoft. This was on the East Coast so these people were doing a tour of all of the major engineering schools and wasting it on their angst over Microsoft. I was around the time of the JVM lawsuit so I understand why Sun was mad at Microsoft but it seemed to have broken something in their corporate DNA and turned them insane.
Perhaps that was just an isolated incident confined to the university recruiters but Sun at that time seemed overly focused on fighting Microsoft instead of improving their product offerings. Perhaps some of anger ended up turning potential users and employees away from Microsoft but I suspect lots of those ended up going with Linux or some other alternative rather than going with Sun.
You've hit the nail on the head, that's a perfect analysis, and it wasn't an isolated incident!
But they'd been like that for a long time, since before I started there in 1990, long before Java. They DEFINED themselves in terms of Microsoft, to the extreme extent that when Sun Microsystems fell apart into separate divisions, they actually named one of them "SunSoft" to directly position it against Microsoft. As if.
The management at Sun didn't consider Java to be a programming language or software platform, they considered it to be first and foremost their primary weapon of mass destruction in their apocalyptic war against Microsoft, and they didn't consider Java developers to be loyal cherished customers, they considered them to be disposable brainwashed mercenaries in their World Wide War against Microsoft.
It was funny when Sun proudly and unilaterally proclaimed that Sun put the "dot" into "dot com", leaving it wide open for Microsoft to slyly counter that oh yeah, well Microsoft put the "COM" into "dot com" -- i.e. ActiveX, IE, MSJVM, IIS, OLE, Visual Basic, Excel, Word, etc!
And then IBM mocked "When they put the dot into dot-com, they forgot how they were going to connect the dots," after sassily rolling out Eclipse just to cast a dark shadow on Java. Badoom psssh!
Sun totally dropped the ball fighting their true original enemy AT&T, and they should have put all that effort and energy into improving SunOS and railing against AT&T after SunOS finally beat System V in the Unix market, instead of capitulating to AT&T AFTER SunOS won the Unix war against System V, and then rolling over, giving up, selling out to their mortal enemy, and becoming Solaris.
To port my favorite cross platform Apple/IBM joke:
The SunOS -> Solaris transition was before I joined the industry and in spite of multiple attempts at reading around the topic I've never found an explanation of what happened that actually made sense to me.
Sun completely forgot they were a hardware company. They made great hardware, but while Microsoft made it easy to develop software for Windows, Sun didn't make it easy to do the same for Solaris (maybe out of fear it'd benefit their remaining hardware enemies IBM and HP). They didn't even try very hard to build low-end machines that could do PC tasks at competitive prices.
I was at a university career fair in, I can't remember the exact year, but I think it was like 2008 or 2009. It may even have been while Sun was in the process of being bought by Oracle? I'm not sure. All I remember was the Sun recruiter handing out OpenSolaris CDs (I think) with a vibe of "well this is all essentially over but it might be fun for you to play with a non-Linux UNIX-like." I'm not sure if "this" was the company or the distro, though, it was a while ago. Mostly I just recall the interesting ambiance of... nostalgia for what could have been, I think...
I was also at university in 98/99 going to job fairs and I was totally sold on the Sun/Solaris/Java future, so I was very excited to see the Sun booth and try to get an interview. But, of all the companies there, they were the least enthusiastic and the least informative. I went to the event sure I wanted to work at Sun and left the event evaluating my other options. Seeing how it turned out in the years that followed, I'm glad I changed my mind!
(Still used a Sun workstation for the next five years at my first job, though.)
Mellow Mushroom had expanded by then to lots of the southeast, including most of the SEC and ACC college towns. I made a joke to a friend who had moved to Denver that he should start a franchise out there because it fit the culture. Turned out someone has already done it, making a little island of Mellow Mushroom in the Rockies.
IMHO the relative cost of Linux hardware created a brain drain that they didn't see coming and ultimately couldn't resolve. The amount of work a single well-equipped E4500 could do vs a cabinet of x86 gear was remarkable, but that E4500 might run $600K and the x86 gear 0-10% of that.
I'm still staggered by it, and I watched it happen, too.
Several times in my computing career -- say, from 1990 on -- I've seen something start to happen that was clearly going to be huge, only to be told over and over that the obviously-inevitable development was Impossible.
The first example was the ascendancy of the microcomputer. "OH, there will ALWAYS be a space for mainframes! Corporations aren't going to do their accounting on some crappy PC! You're just young and don't see how these things really work."
The second was the way Linux completely destroyed the commercial Unix market (and, with it, the "fancy Unix workstation" market). The power of FOSS is, in my view, very very often overstated by idealists, but it's hard not to see the wholesale corporate/academic/research shift from Solaris, et. al., to Linux on commodity hardware as anything but a HUGE win in terms of computing power.
I'm drinking coffee out of an Oracle Linux mug from SCaLE 9x, which would have been 2011... note how Oracle Linux has absolutely taken over the world since then :)
Perhaps free as in money but not in source. All the big boys who deployed it absolutely wanted the source to customize in house. Also too little too late.
It's not only that. I deployed commercial projects when the question "Solaris or Linux" didn't sound ridiculous, and there were times where Solaris had been a reasonable choice. But that was a long time ago. Since then Linux (and the ecosystem) matured a lot, and Solaris didn't have much to counter that, so its relative worth declined rapidly.
Yes! People today may not realize how huge Sun was in the 90's. Almost all startups ran on Sun gear. Almost all ISPs had a Sun system for their "shell account" servers. Also, I had a SparcStation 10 at home in 1996 or so (which was kind of unheard of!)
Totally honest question: WHY does everyone think it was cool? I can still remember running it in Uni on some of the prized workstations with everyone raving about it but never got the point about it.
Back then I was on Windows and everything I saw on Sun systems looked weird, laggy and just super obscure to me. The whole UI felt like a mix of enterprise and neckbeard charme despite being non-open. It did not seem to run none any of the things I cared about back then. Later on I remembered Sun mainly in context of their stewardship of Java which also didn't help connecting any good feelings to it.
So maybe in 2022 someone could enlighten me about what exactly was so cool about Sun?
In the 90s, commercial Unixes really still ruled the roost. Linux was getting traction, but slowly, so for big-biz use you were looking at AIX from IBM, Solaris from Sun, HP/UX from HP, or something like it.
The Sun hardware was materially better, or so I recall. It was also more attractive. I actually still have an Ultra5 on a shelf in my office that, for a time, ran Oakley's online store (which my old company built). It's still a good looking machine.
Solaris, too, was widely viewed as better/more workable/easier to use than its competitors. It had more polish.
If you're looking at Solaris with the latest build of Ubuntu or whatever in mind, yeah, it's gonna look neckbeardy and hokey. But that's not the world Solaris was the king of. Back then, using a commercial Unix meant you got a LOT of shit that Just Worked out of the box, and generally an easier time getting your shit running.
The GUI apps were generally pretty bad. I didn't use the default Sun window manager (CDE), which was kind of slow and bloated. I installed either WindowMaker or Afterstep. I forget which, but it made it look like NeXT. The only GUI app I used on my Sun was Netscape. Everything else was done in a terminal (xterm window.)
I probably spent most of a week building stuff from source and customizing my system. If you wanted a "Unix workstation", Sun was the best. SGI looked cooler but often getting open source stuff built on it was a PITA.
This was during the Windows 95 / NT 4.0 era, so it was definitely a step up on Windows.
Ironically, Sun invested heavily in user studies and development for Gnome 2, which is why is was so user friendly. Sun didn't really benefit from it commercially, then thanks to thr CADT development model, Gnome 3 was a massive regression for normal users (and still was a few years ago).
> Sun systems looked weird, laggy and just super obscure to me.
Workstations typically came with fast SCSI disks. Once I wanted to find a file after a failed install and used `find`. The search came back immediately, empty. I concluded I was doing something wrong, did `man find` to confirm there wasn't any Sun-specific oddity. Then I tried with `find / -name passwd` because I knew that file existed. Result came back immediately. It was probably caching the whole directory structure in RAM, but the instantaneous response really impressed me back then.
Mouse was laggy, because, IIRC, it communicated at 9600 bps or something like that. Macs and SGIs were much smoother.
> So maybe in 2022 someone could enlighten me about what exactly was so cool about Sun?
The one thing I loved was X and its network transparency. I often run NetScape on a beefier box and it was just like having it running on my workstation. Same with software development and engineering tools. I could reach out to multiple computers (and not only Solaris ones, but any Unix) from my desktop easily.
The same could be said about any other Unix-based workstation with X (NeXT didn't come with it).
I had the same experience. The Sun workstations that I was able to use were very slow compared to the same era PCs running Windows. The keyboard was cool as hell, though. The control key was in the correct position (where Caps Lock is in the standard layout). I do remember that the font rendering was horrible. There was no anti aliasing. Windows had just came up with Cleartype.
At the hight of the workstation era, where Sun generally ruled the roost.
You've got to compare what you got with a workstation vs. a PC class of the era.
The PCs of now-a-days are light-years ahead of what PCs were available in the Workstation era, so much beyond that time, that nobody can think back to that time. This comparision kind of smears some time frames together..
So, in the early 90's , Sun was starting to ship 4-way MP, and 8-way MP. In 95, they introduced 64-bit CPUs just as PCs were really starting to use 32-bit computing. Typical speeds were from 50MHz - 400MHz in this time frame. At the high end you could get up to 256MB of RAM.
Sun Frame buffers at the time were doing 1920x1080 resolutions. (or slightly higher, but granted, a lot of these high-res were black & white in the earlier time frames).
They came with multiple fast/wide SCSI drives, tape drives, etc.
They all came with the vendor's unix. System uptime was usually quite good, months to over a year.
You did have some vendor lock with what software was available, but generally Sun was top of the heap, every vendor had something available for Sun.
Contrast this with PCs at the time.
State of the art PCs were 486 running 50MHz-75MHz, with 8MB of RAM. Maybe 32MB if you wanted to get Windows 95 going.
Video was VGA at 640x480, again not much need for more until after Win95 launched.
Hard drives were IDE, a whopping two if you were lucky. While SCSI existed on PC, it wasn't very common.
Linux and *BSD didn't come out until 1992-93
Even then, uptime of DOS systems or PC unix at the earliest time frame could be measured in days. You had to get _just the right_ hardware to make things work together.
By the end of the '90s. Workstations hadn't progressed very far. PC's grew by leaps and bounds, catching up and surpassing workstations handily.
PCs were actually fairly stable and the PC unix could give you really decent uptime.
But in the before the explosion of PCs, if you wanted power or uptime, or graphics, or disk speed, etc. etc. compared to what else was there, Workstations were the way to go.
Sun boxes were way more performant than anything Intel for some time. I used to run a big CORBA app, and windows hosts would only be able to handle half of the workload form a thread perspective. In that era, you usually scaled up, and Sun gear did that, even giving you mainframe-like segmentation on the E10k boxes.
The other thing is that OSS databases mostly weren’t a thing, and commercial database platforms like Oracle or Informix were released for Solaris/SPARC first. I worked for a .gov client that was all about RS/6000 and AIX, and it was painful - you’d wait 3-6 months for some patches.
Linux really broke the workstation market circa 1998, and the cracks in the server platform were sort of hidden after the dotcom implosion put lots of gently used gear on the market.
My Windows 95 and 98 couldn’t run more than 45 mins or so without bluescreening. NT was good but had less software and PCs didn’t have enough RAM. A stable multi proc/user machine was a luxury in those days. Being pretty was not a concern for most, and SGI was the choice when it was.
I only really cared about GNU/Linux in production around 2006, until then it was a means to have a cheap UNIX clone at home, while we were deploying Solaris, HP-UX and Aix into production.
Most of the deployments were naturally Solaris based.
About 2002 here. Some companies were more conservative than others. One was okay with Linux for dev systems, but not production. Another had a "low end" product ported from AIX, which ran on Linux.
It wasn't just startups, of course; we used Sun gear at good ol' Cray Research in that era and it was our first upgrade from dumb terminals on mainframes (& vaxen). Good times.
(I still remember when Sun introduced color systems, and they were such status symbols that they went mostly to managers, including one who was color-blind. A warning sign in retrospect, I suppose.)
It was good, but a lot of people didn't get to find that out because it cost money to play.
Good enough and free will beat out better and expensive almost every time (the key part being "good enough", which, granted, makes that rule someone circular ;) ).
It makes the snobs look a bit ridiculous: "Oh, it only took you five years to get to where we were... but we were on expensive hardware you couldn't buy without a major contract, running fewer applications, and invested in a system that would be eaten by the low end so comprehensively now only snobs know how good it was."
I do know some folks that are still using it (or Joyent's SDC Hypervisor offering, which I guess also has numbered days since the Samsung aqui).
Sadly, I don't think it's got much time left. I do miss it, it was rock solid and the tooling was much better than linux at the time.
Containers were implemented properly from the beginning, SMF is (still) much superior to SystemD, things like resource limits and so on could be applied in a much nicer manner.
And it was fast too, when you had the sun compilers. Oh well.
They don't seem to be a huge player like Sun/Oracle, Joyent/Samsung though.
I really wish them well, I loved that stack -- but I wouldn't bet on that stack anymore. I moved storage to FreeBSD, (internet) networking to OpenBSD and VM's live on proxmox when I am running actual hardware.
I saw the MNX announcement the other day and am still confused by it. What did Samsung acquire Joyent for if not the software, or did things simply not pan out and Samsung is now offloading bit?
>or did things simply not pan out and Samsung is now offloading bit?
Samsung wanted to enter the Public Cloud Space, ( The kind of business they would be interested in ), but their NAND and DRAM division would suffer order reduction from hyper scalers. In the end they sort of back off this segment of business.
It was a rock solid OS, sitting on rock solid hardware. I enjoyed working with Sun kit back in the day, and it's such a shame that Oracle killed yet another great product.
It was already dead when they bought it. What Oracle wanted when they bought Sun was Java.
Solaris and the hardware just came along with the ride. There are a few people that they needed to support for contractual reasons and a legacy install here and there that they could continue to milk, but for everybody else Solaris was dead before 2009.
This was just as obvious then as it is now. It is just that Solaris fans refused to see it.
It’s impossible to reply to this because by virtue of thinking that the systems were good you can label anyone a fan.
The fact remains: it was widely deployed, and we are still catching up to certain components of the operating system.
Dtrace, ZFS and SMF are examples.
Calling it dead only tells me that you weren’t in an industry that used it. It would be like claiming windows is dead because most things these days are cloud hosted and windows is very expensive there. It’s true in some sense but extremely blind to the rest of the industry.
Dtrace, zfs and smf were nice but came too late but they are things that appeal the engineer, not the manager that sign the contracts. And Sun was already on life support when Oracle bought them.
Sure there was still a lot of companies still using mostly solaris back then - I worked for 2 of them as a Solaris "expert" somewhere during last decade - but all of them already had projects running about replacing Solaris for more cost effective redhat, even those that bought exadata racks.
"expert" meaning I was still one of those survivors still young enough to be still willing to work in an operation team and having more than 5 years of experience administrating Solaris and AIX shit.
I must say I loved the integration of zfs with gnome2's nautilus back then on opensolaris. I used it for a while as my daily driver at home.
> This was just as obvious then as it is now. It is just that Solaris fans refused to see it.
Solaris fans (yours truly included) saw it just fine. But Oracle has been a poor steward, and if it had any chance of being rescued, well, they gave it none. And their fame for suing and squeezing their own customers didn't help... every company ran away from it once it was an Oracle property, and suddenly there was no way to run most commercial software on it (My employer was impacted for example).
Sun was killing Solaris way before Oracle acquired it. Because Sun was at the beginning of that era, people look at it with rosy colored glasses.
It had terrible leadership and didn't understand how to fully embrace open source (speaking strictly about OpenSolaris, they never really wanted share control of anything, form a community, etc).
What is Solaris without SPARC? A somewhat slower and not-very-compatible OS for apps that are perfectly content with an x86?
Oracle killed it by stopping development of new SPARC systems. Linux is a much better fit for x86 boxes than Solaris ever was. Solaris for x86 was not a revenue driver or should never be considered that, but a gateway drug to SPARC workstations and servers. Java was the same - a programming environment where you could develop on Windows and deploy on Solaris. Unfortunately, for them, you could deploy to Linux as easily.
> This was just as obvious then as it is now. It is just that Solaris fans refused to see it.
Sun was having trouble funding their chip development operation because they tried to be an everything company.
I ran OpenSolaris on my laptop back in the day; I thought it worked very well. Note I'm not some old-style Unix greybeard who used Solaris back in the 80s: I was just playing around with various systems and ended up sticking with OpenSolaris.
To be fair, from my casual observations it didn't seem there was that much inertia around the system, and Oracle killed OpenSolaris before I had a chance to put it on any servers, but I definitely think there was some potential there.
I think the general consensus is that Oracle bought Sun for Java.
Databases at Oracle have certainly done better than Solaris though. OracleDB is still a cash cow and MySQL I believe MySQL is still the most popular Open Source relational database ( well, other than SQLite probably ). Monty forked MySQL to MariaDB literally the day after Oracle bought Sun but MySQL remains far more popular.
Yeah it's a "why not both" situation. They wanted Java for that sweet Android lawsuit, and they wanted MySQL for the support contracts and to ensure it wouldn't encroach on many of the same use cases Oracle database is still needed for.
Solaris has been a zombie for years. There's a ton of enterprises that still need it. If you suggested using it for a new project anywhere, you'd be laughed out of the room (or the zoom call, at least.)
Never heard of it so I had to look this up, OmniOS is a distribution of illumos.
And for those who dont follow closely, illumos is forked from OpenSolaris. And OpenSolaris is a discontinued open-source computer operating system based on Solaris and created by Sun Microsystems.
> Containers were implemented properly from the beginning, SMF is (still) much superior to SystemD, things like resource limits and so on could be applied in a much nicer manner.
This! Shame they didn’t make it free sooner—it might have lived longer.
I'd never heard of SMF [0] before, and I'm not very familiarized with this problem domain. Maybe this is a dumb question, but if it was so great why hasn't anyone tried to replicate it for Linux?
It always baffles me that superior tools can be developed while still failing to gain traction in alternative ecosystems.
They did, it one of the larger parts of systemd that motivated its adoption and ubiquity.
A bunch of commercial UNIXes sprung init replacements/extensions with process lifetime management systems in the 2000s, systemd is visibly influenced by many of their design choices.
OS X natively uses launchd (uses plists for config, has a weak dependency system) , Solaris uses SMF (uses XML for config). There were also the daemontools/runit family, s6, and a couple others that sprung up from open source quarters.
I remember the Oracle acquisition and the subsequent squelch of chatter in #opensolaris and #nexenta on irc. I worked on porting a ton of Debian packages to Nexenta and used it in a university setting along side Linux and macOS hosts. It was hard to see all of that work become somewhat worthless with Oracle's acquisition of Sun. It was an extra slap on the face that Sun hardware that I personally owned outright had a firmware bug that Sun released for free that I then had to pay for an Oracle subscription to download.
ZFS, Dtrace, Zones, iSCSI, SMF ... all rock solid and extremely usable. I never liked the antiquated package management in Solaris thus my involvement in Nexenta.
Solaris being "free" 8 years later means nothing to revive any of this. If anything, it signals that Oracle is taking steps to stop considering this as a product.
I remember that well too. I ran OpenSolaris om my laptop for years because at the time it had the best hardware support for what I had. I had damn near turned into a fanboy when the Oracle acquisition happened and they yanked the rug out. I was disappointed at the time and remained so for years, especially when I found an old laptop with opensolaris on it a couple years back and got super nostalgic.
In hindsight, it was probably the best thing for Linux though. When I was hunting for open source OSes there were plenty of people like me recommending OpenSolaris to new users. Once that happened all the oxygen got sucked out of the room.
Yeah the ship has long since sailed. This isn't going to make any difference to reviving Solaris. It's too far behind now and even if it weren't, it's a "fool me once" kind of thing.
When I was in college in the mid 1990's Sun was the dominate Unix workstation company. We had a couple of thousand on campus. Then I started working in the semiconductor industry in 1997 and every had a Sun SPARCstation or Ultra 1/2 on their desk. We had some big 16 CPU E4000/5000 class machines in the server room.
All of the EDA / CAD chip design software ran on Unix and some companies started porting to Linux/x86. Some of our jobs used >4GB RAM so we required a 64-bit machine. AMD released their Opteron 64-bit in 2003 and we got a couple right after our tools were ported to 64-bit Linux. We never bought another Sun after that.
Sun ultimately suffered the same fate of all of the Unix workstation companies. By selling only expensive "Serious Business" hardware to corporations they were able to keep the per-unit margins high, but were completely crushed by commodity PCs in volume and ultimately total profit. Ultimately those huge commodity sales volumes were turned into big R&D budgets that allowed them to produce faster and more efficient hardware than the big workstations providers. And then the one advantage they had: the OS, was undercut by BSD and Linux.
So a company has the option of spending 5 or 6 figures on a big Unix box with a 4 figure yearly support contract and pain in the ass license management, or they could spend a low 4 figures on a PC running Linux that is just as good or better in all the ways that matter and doesn't have annoying salesguys to deal with or recurring costs.
Scott McNealy, CEO of Sun, spoke at my biz school ~2000. Someone asked about the competitive threat of Dell + Linux. He ranted for a long time about Dell being junk and SunOS vastly better than Linux. Everyone after felt Sun was doomed. They crashed a few years later. It was a textbook Innovator’s Dilemma case.
> By selling only expensive "Serious Business" hardware to corporations
This is a mistake IBM is making as well. There is no cheap on-ramp for their POWER and mainframe platforms. I doubt those machines are running many new workloads. Almost all new corporate software is developed for Linux-based commodity servers.
It's a shame, really, because they are magnificent machines.
It doesn't even run on qemu /kvm due to very poor compatibility so not sure what Oracle is trying to achieve here - maybe get me to buy their x86 servers if they are still selling one? Or find specific old h/w in my basememt anf try it on there?
If you needed dtrace or zfs luckily FreeBSD runs fine on most hypervisors and real world x86 hardware.
So does Illumos and OmniOS - which are direct forks from the last opensource versions of Solaris. I played around with illumos a few years ago before getting into FreeBSD and it’s a lovely little operating system.
It actually does run on KVM! I spent yesterday trying to get it to work and found the trick was setting the chipset to i440FX, and putting every drive on the IDE bus.
Probably not a reasonable choice for any real usage though.
The specific mention of open source developers makes me think "more open source software that works on Solaris" might be part of it. Doubt it'll help much, but it might move the needle from "they still make Solaris?" to a "guess I'll make sure it at least builds" sort of state.
Perhaps they aren't trying to achieve anything at all, and this opening is just the small visible part of some internal conflict where one participant used to advocate for a grand open source Solaris strategy. Resolved by giving that participant this partial opening as a token concession.
I had the pleasure of programming on Sun machines back in the late 90's. I loved them. They had humongous screens. I wouldn't want one of those falling on me. They had a package management system. I occasionally had to update the system, and I found the update process a really pleasant one. The guy who did my performance appraisals noted that most programmers preferred the Unix environment over Windows.
That was the late 90's, though. This is now. Linux won. It was great for its time, but I feel that Solaris is too little, too late.
I'd say it was a lot, very early. SunOS and later Solaris were awesome before Linux existed and all through the growing-up period of Linux. Even by the time of the acquisition (2010) when Linux was fairly mature, Solaris was still so much better technically (zfs, dtrace, tooling) but just had become niche because everyone was running what was free, Linux.
Yeah if anything what was too late was solaris' open sourcing. It could have had Linux' lunch.
It could have done wish some Linux dev love too, Linux was much further ahead in terms of updates.
I remember updating one of our Solaris servers back in the 9 or 10 days and it was basically just a list of things I had to change manually. Almost nothing automated.
> Yeah if anything what was too late was solaris' open sourcing. It could have had Linux' lunch.
I know some disagree but I maintain that the key factor to Sun decline was the unbundling of the compiler when Solaris came out (vs. SunOS where it was included).
Stepping back to early 90s, open source was generally never packaged and prebuilt. You had to download the tarball and compile it locally. Installing a new machine meant spending a good while building all the tooling to make it useful.
Approximately all of academia and research labs were SunOS shops. All open source code built out of the box (tarball) with ease on SunOS since its developers were almost certainly on SunOS themselves, or at least used it as a primary testing platform given the popularity.
In the early days of Linux it was a lot of hard work to get open source to compile, since the code mostly assumed SunOS. As late as the late 90s, it was still easier to compile many things on SunOS than it was on Linux.
But with Solaris in the second half of the 90s[1] the compiler was unbundled. Suddenly it was impossible to build all the code you needed to make the system useful, without jumping through a buch of hoops. Either buy an expensive SunCC or go get and install the GNU compiler but then the makefiles tended to assume that if the OS was sunos the compiler was SunCC so you had to go resolve differences between Sun and GNU compilers. Meanwhile, compiling packages on Linux was steadily becoming easier.
[1] Well, Solaris came out in the early 90s but was pretty bad compared to SunOS initially. So most SunOS shops, like the research lab where I was, resisted migrating for many years. Solaris 2.5.1 in 1996 was generally considered the first version worth upgrading to from SunOS.
I don’t share the same memories. Working around commercial compilers was horrible. I actually preferred NT for a bit around then. Plus I could afford to buy a machine to run it at home.
Early days of win32 were quite fun. I can see why it won on the desktop and perhaps unfortunately is still doing quite well.
I worked at a computer shop during school. And I remember some time in 1994 a guy coming in very excited about a new operating system. "It's like the school operating system, but you can run from home". He needed a specific network card, foreshadowing the next 20 years of my life, which we did not have in stock at the time.
This is nice for SPARC hobbyists like me, because now I can play with newer packages on my SPARC T4 server. Without this the updated packages would be stuck at versions from 2018.
But in 2022, there is not very big need for commercial operating systems outside of large enterprises. But still I'm happy that Oracle will be supporting Solaris until 2034. I always get sad when something is deprecated.
I understand that everyone wants to use Linux and x86 nowdays, but it still makes me sad that the diversity in operating systems and hardware is decreasing. It feels like everyone is only interested in Linux. Many FOSS software doesn't even work or compile well on other free operating systems such as BSDs and Illumos distros, because the developers use Linux themselves.
I'm mostly a Linux user myself, but I would like to live in a world with a lot of different operating systems and CPU architectures in use.
What's the point? Linux has surpassed Solaris in every way. The only possible explanation for this I can think of is: There may be some companies stuck with some software that only runs on solaris, and they need to keep onboarding new engineers to deal with it, and they want those engineers to go through solaris training, and this makes the training go more smoothly, since the trainees don't have to sync with a license server. But, it's a stretch.
ZFS, NFS, iSCSI, and CIFS services are quite superior on Solaris compared to Linux. On Linux each of these services are a bad joke, but on Solaris they are rock solid.
I think anyone would be a fool to use software licensed from Oracle, but if that part doesn't bother you, on the technical side there may be benefits.
I remember emailing Scott McNealy in 1995 telling him urgently that he should opensource Solaris. If I had persuaded him it would be a different world today. We would not have had to live through two decades of NT on servers.
Was there a meaningful possibility he might have done that? Was your relationship close enough that he'd have regarded you as a competent authority on the subject?
I know people who have written missives like that to both Bill Gates and Steve Jobs. Nothing came of it, obviously, because nothing could have come of it.
Why? There is already an active open source fork of Solaris with better licensing terms than what Oracle offers - Illumos / OmniOS - https://illumos.org/ .
From what I have heard, the core of the OS engineers had left when Oracle acquired Sun back in 2010. And those developers have continued to be the largest contributors to Illumos.
So, in essence, Illumos is the real continuation of Sun Solaris.
This is factually true - The exodus from Sun Microsystems started when the takeover news started doing the rounds. The OS team was a shell of what was before. Then Oracle further squeezed that division ~2015-17. All those remain are the switchovers or people who have been promised a lot of TC to keep that brand just about alive.
> I have a feeling that Solaris is a bit of a late bloomer.
Compared to Linux and FreeBSD? Not a chance, Solaris was running circles around them in the 1990s when Linux was still a toy operating system and FreeBSD was figuring out their infrastructure.
These days, FreeBSD has adopted nearly all of the cool technical advantages Solaris has, and Linux seems to be actively hostile to adopting such things (ZFS, DTrace, ACLs...).
I had luck with Linux when I couldn't get Solaris to install on my personal PC in the early 2000's due to compatibility issues - me and a colleague tried our best. So about 22 years too late.
There is a cautionary tale here for anyone who wants to write great software. Solaris is great software, and yet I wouldn't want to run it because I wouldn't want to hire for it. OSes are complex enough that you want specialists who "dwell" in the space (as opposed to casual things, like libraries), and the community is just too small. (That said, sometimes teams pick the outlier and win - WhatsApp and Erlang come to mind. But this is really rare, I think.)
Solaris supports OracleZFS rather than OpenZFS (forked from old OracleZFS). It's interesting and maybe there are some good feature only available on OracleZFS, but I don't want to build a new server with Solaris that would not good for recent hardware support.
Some of the core offerings in Solaris (ZFS, Zones, dTrace, mdb) were ahead of their time, arguably. And the implementation is far more balanced than what you see in e.g., Linux, with good central documentation.
So if you have a system which was built on Solaris, it might be smart to keep it operating on Solaris because you happen to know the tools and so on.
It's likely any greenfield tech is better built on other systems. I'd use Illumos over Solaris, or FreeBSD.
Solaris used to have some interesting features that were later implemented in Linux and FreeBSD. It took years and not all of these work perfectly well, but at this point there is no reason whatsoever to use Solaris instead of Linux. Unless you want to test if your software builds for the dozen or so of existing Solaris users.
For one, it is one of the few UNIXes with proper hardware memory tagging, so if one cares about UNIX and security critical apps, that is a possible benefit.
Unless you are part of their existing installed base, don't waste your time. It's an attempt to generate interest in a dead product. Oracle long ago killed any chance of it remaining relevant in the larger software world.
In 2010 Oracle -a company BUILT on MINDSHARE- failed to understand that OpenSolaris was an attempt at building MINDSHARE, so they killed it over a minor "embarrassment" caused by a minor "competitor" using OpenSolaris. Now they must be wondering how to re-build MINDSHARE and recoup some of their costs, and they evidently think "free for personal use" is going to do it. That's NUTS.
Between November 2016 and August 2017 pretty much the entire Solaris development team was shown the door. I was on the Solaris Zones team in August 2017. A small number were allowed to stay and they are keeping the lights on. I have no idea what happened on the sustaining engineering (support) side.
If you download a copy of this while at work, be prepared for your company to receive a nastygram from Oracle's lawyers about them now owing Oracle money, just like if you download Virtualbox's extension pack.
That's cool for a bit of nostalgia but sad because...in a world with Linux and freeBSD I can't understand why anyone would do this for any real reason.
If they had done this 10 years ago people would be thrilled, but doing it in 2022...well it's kind of like saying DOS 6.22 is free for personal use.
A not so subtle reminder that at oracle for every engineer or developer there are two lawyers or managers waiting in the wings.
I was just thinking that I used to really like the CDE Motif theme, but after looking up screenshots I realized what I was actually thinking of was Java's cross-platform "Metal" look and feel. I liked how MDI windows looked like they had knurled tool handles for the title bars. I liked the little notches on tab panel selectors. At the time, I liked it because it was novel compared to what I was used to (Win9X), but now I like it because everything looks like it does something.
Now that I think about it, Motif is the least favorite of any DE theme I've ever used.
A lot of good comments about the Unix market back then. The honest truth is that cheap no-name PC's with RedHat and a little later, nVidia graphics cards are what killed the Unix vendors. The price disparity was stark and a no-brainer for companies looking to do big things but with increasingly commodity hardware. It was sad to see Unix vendors die off.
Unfortunately, no. I guess it is because Solaris 11.4 was originally going to be Solaris 12. It has a lot of new stuff and modifications compared to 11.3. I guess they wanted to optimize it for the newer SPARC processors (or get enterprises to update their old systems).
I don’t think Solaris gets much attention from either the blue or red side these days. Patched or not, I would keep a Solaris machine segmented off with my other retro crap and not use it for anything public-facing where it might be exposed to attack.
Huh! I thought x86 Solaris was always free for personal use. I remember installing it on my desktop in 2003? I guess something might have changed in the past 20 years :V
It has remained free for test and dev use during the entire time Oracle owned it, yes. The big deal about this announcement is that it is still free for test and dev, but you get updates! Previously, you could get Solaris 11.4 from 2018, with no updates. Now you can get 11.4 with updates through December 2021, and it sounds like future updates will be made available.
I’m too young and missed the Solaris era. Are there still valid usecases for Solaris today? Or ist this just to support some legacy systems that can’t be upgraded?
I can imagine that, but for me personally it was because I could run Linux on the existing machine in our dining room, and didn't need Big Iron™ to try it out. I'm sure related to that is why the Asus router in my house also runs on Linux, and likely my fridge, and _arguably_ my phone
there really hasn't been a compelling reason to run solaris since linux fixed its resource scheduling issues in the early-mid 2000s. (or freebsd fixed its smp issues around the same time)
Solaris 11 does not work without ZFS. It is the only supported root file system.
Keep in mind that Solaris ZFS is largely incompatible with anything that exists in an open source OS. ZFS started out by incrementing pool and dataset versions as features were added. The pool version that will be created by default in Solaris 11.4 cannot be used by ZFS in illumos, FreeBSD, or Linux.
The open source versions of ZFS (found in illumos, FreeBSD, OpenZFS, ZFS on Linux, OpenZFS on Windows) use feature flags. By default, pools created with these will likely use a pool version that says "I have feature flags", which Solaris ZFS will no understand and will refuse to import.
All implementations of ZFS should be able to create a pool with an explicitly old pool version and be able to be imported everywhere. That doesn't mean that pool will be useful for everything. For instance, open source ZFS does not have dataset aliasing, a feature I added in Solaris ZFS after the SoLongAndThanksForAllTheFish commit but before Solaris 11 was released. This feature is required for Solaris Zones.
I guess my question was more along the lines of "does this at all impact the potential legalities of including ZFS in the kernel"? Very interesting details, though!
Meh. FreeBSD with jails and Linux with containers will send that Sun Solaris wannabe to the trash.
Propietary Unixen are dead at the server space. Nowadays people doesn't care, they just keep spawning vm's and jails/vm's as nothing in order to scale up things.
The hippies from the MIT/Berkeley won.
We have free as in freedom hacker OSes and corporation-level setups with ease. Even FSF distros work relatively well (fck Radeon blobs).
Today tech support and configuring services it's where the money lies on.
When you say "people don't care", I think you have a specific, but not explicitly defined, set of "people" and "organizations" you are presuming.
Proprietary Unixes are still common enough in large traditional organizations to stay profitable and developed. They get purchased or bundled by non-hands-on people likely using different criteria than average hacker news hacker might consider.
(Source or disclaimer - probably half the projects I've been on in ERP space are on proprietary Unix. My problems are very different [not implying better or worse] than those typically discussed here)
organizations with enough cash to use solaris for production environment probably have enough cash to pay for licenses for dev/staging environments too.
this move is likely a (desperate?) try to keep solaris relevant among open source developers.
tbh, if I were developing something and somebody asked me to support a proprietary operating system I can't run in a VM... the least I'd ask for is for them to provide me a licensed environment (like at least a virtual machine or something, with the license paid by the requestor) for me to run tests and stuff.
For sure. Not as many as I would've thought though.
A perspective in large number of orgs I've seen (not claiming right or wrong) is that Linux is better at horizontal scaling while Unix is better at vertical scaling. Vertical scaling is out of fashion but still powerful for many scenarios. So I've been in many rooms where prevalent thought is "Linux is great we love Linux we want more Linux, but we want this database to run on 128 cores and we don't feel Linux is the right choice". So I've seen clean slate ERP implementations on Unix, not just legacy upgrades.
And I mean, what some of these things do in firmware is impressive. They do on firmware and hardware level what we're these days doing in software and software defined whatever.
If you had to pick an inflection point where Sun's dominance in the Unix world began to decline it would be this.
Not in terms of total installations, momentum and exponential expansion of the entire industry distort that, but in terms of it's influence. This is also the exact same time that 386BSD/NetBSD/FreeBSD and Linux started releasing.
As others have pointed out BSD-based SunOS was the king of Unix-like operating systems in the late 80's to early 90's. Sys-V based Solaris (1992) was an immediate flop and quickly earned the nickname "Slowlaris". Only by version 2.4 was it even usable in production. By 2.6 it was a decent operating system but anyone who remembered using SunOS wished they had never switched.
>Michael Tiemann is vice president of open source affairs at Red Hat, Inc., and former President of the Open Source Initiative. [...] He co-founded Cygnus Solutions in 1989. [...] Opensource.com profiled him in 2014, calling him one of "open source's great explainers."
Early versions of Solaris were terrible. It was a pain to build any open source stuff on them. Solaris 2.5 was decent, but I think I kept my personal SparcStation on SunOS 4.1 (retroactively renamed Solaris 1.1) until Solaris 2.6 (aka SunOS 5.6) was out. Sun marketing was pretty confusing.
You know who's a bonafide hard core mega-hippie is Sun's "Science Officer" John Gage, who earned his way onto Richard Nixon's enemy list! We've done bong hits together, and I was incapable of smoking him under the rug, so he's the Real McCoy (and also as smart as Spock).
>Following the Free Speech Movement, Gage became active in opposing the war in Vietnam. He worked on the Robert Scheer for Congress campaign in 1966, almost defeating a Democratic congressman who supported the war. He co-chaired the Robert F. Kennedy for President campaign in 1968 in Alameda County, and was a Robert Kennedy delegate to the 1968 Chicago Convention, representing Berkeley and Oakland.
>At Harvard, he helped form the Vietnam Moratorium Committee with Sam Brown, David Hawk, Marge Sklenkar, and David Mixner, and co-chaired the New England division. He organized the 125,000 person Boston Common Vietnam Moratorium demonstration in October. and coordinated the 400,000 person Vietnam Moratorium demonstration on the Washington Monument grounds in Washington, D.C. Nationwide, the Vietnam Moratorium was the largest mass demonstration in US history, with over two million people involved.
>He organized a number of major antiwar demonstrations in Boston, in New York, in Washington, D.C., and in Philadelphia. In 1972, he was named by White House attorney John Dean to Nixon's Enemies List.
>He was the first field organizer for students for the McGovern Presidential campaign in California, then joined the national campaign as assistant press secretary and trip director, working for Frank Mankiewicz. In that role, he coordinated the day-to-day movement of the press and staff from event to event nationwide with John Podesta and others. In the 1976 Jimmy Carter campaign, he helped train staff, and organized the final rallies in California. In the 1980 Ted Kennedy campaign, he was assistant national press secretary and trip director, and member of the traveling party.
>In 2008, Gage spent several weeks in Ankeny, Iowa, organizing for the 2008 Barack Obama campaign.
>Concert activity
>In 1969, Gage was asked by Bill Hanley, the owner of Hanley Sound, the staging and sound system used at Woodstock, and the system Gage used in Washington, D.C. for the Vietnam Moratorium, to come to Palm Beach, Florida, to take over producing the International Palm Beach Music and Art Festival. B.B. King, Janis Joplin, the Rolling Stones, Sly and the Family Stone, Jefferson Airplane, the Byrds, Sha-Na-Na, Country Joe and the Fish, Steppenwolf, Johnny Winter, Sweetwater and twenty other groups performed; there had been strong opposition from the Governor of Florida, local law enforcement, and some local churches. Heavy rain and unusual cold did not stop some career-best performances, and crowds up to 100,000 people. This Rolling Stone concert just preceded Altamont, six days later.
>Subsequently, Gage was called to rescue the Louisiana Celebration of Life Festival after two people had drowned; produced the New York Shea Stadium Festival for Peace Concert with Peter Yarrow, Janis Joplin, Creedence Clearwater Revival, Herbie Hancock, Miles Davis, Dionne Warwick, Paul Simon, Sha-Na-Na, Johnny Winter, and fifteen more, the Philadelphia Peace Concert, and several other events involving over 100,000 people.
John Gage clip from Berkeley in the Sixties, on peacefully seizing a police car:
California inspired: from flower power to Silicon Valley;
By Chris Edwards.
How 1960s Bay Area radicalism helped shape the technological powerhouse of northern California.
The operating system is basically irrelevant in Oxide's case, it's relegated to a hypervisor with a few web APIs. But they have a very narrow set of hardware they need to support and a lot of desire to hack on the kernel so it makes some sense that they would pick something they're familiar with.
Will I install it again? No. Like most everyone else, I've moved on to Linux and BSD.
But I'll spin up Solaris 2.5.1 on my SPARCbook or Solaris 10 on my Blade 1500 whenever I need a nostalgia kick.