He was a bit of a curmudgeon, going on about how his business partner screwed him out of a "seven-digit payout" when his domain eventually got bought by some Japanese company. But a minivan rental and some elbow grease later I had a whole pile of hardware that he was all to happy to be done with: A Sun SPARCstation 20, a Sun ULTRA 1 Creator, an Axil Ultima 1 (a third-party Sun clone), an gorgeous amber Wyse CRT terminal, and a few other odds and ends.
I wrote a more detailed list on my blog [1], but so far all I've managed to do with them is get the drives out and cloned and the ULTRA 1 running (an involved process as the internal BIOS memory lost power long ago, wiping such transient properties as "what is my MAC address".
[1]: https://sidneys1.com/retrocomputing/2022/06/03/retro-roundup...
I remember that the programmable MAC address feature occasionally came in handy when dealing with recalcitrant / braindead software ‘entitlement’ schemes vendors would occasionally require.
Some time around 2003 my boss had a Sparc machine as his desktop, though this was widely viewed as unix nerd nostalgia even then. As a junior I got a PC. It turned out my PC built the project 5x faster than the Sparc managed. After this I don’t think we bought much from Sun except one of those big tape drives, and lots of Dell servers appeared instead.
It must've cushioned the blow for orgs that were still investing in new Suns but were able to plug in regular commodity peripherals, at long last.
The big vendors used to be differentiated on how they communicated on a network, shared files, type of bus architecture, and workloads, and so each company or unit would choose a vendor and rely on them for solutions, and the staff had specialized expertise in those systems.
But all those vendors found themselves adopting TCP/IP and Ethernet. And Sun's NFS was widely adopted. X11 also became standard. The BSDs already had a wide hardware compatibility list, so that spare machine with no OS license could be returned to service. A few generations of college grads had direct experience with Unix and building PCs. Once Linux on PC-compatibles began showing up in the server room, Windows NT was mature, and software vendors were porting to Linux, it was a fait accompli.
It all comes back to SPARC. Price to performance. If Sun had amazing high quality x86 servers and workstation by 1998, their software would have been perfectly fine. Its not like there was anything missing from Solaris at that time.
Many people were not unhappy, they just realized that they could get a faster cheaper machine, and run the same software on a linux. Even if linux was less optimized, still end up cheaper overall for the majority of workloads.
Its not like running linux on unsupported PCs in the early 2000 was that much fun. But Sun by 2002 they were still trying to kill x86 based stuff. They bought x64 server line in 2004.
I got blank look when I asked "why?". Sure they were snappy, and you could run StarOffice on them, but really there wasn't a lot that they were useful for in our day to day work. Nice machines to be sure, but completely extraneous. I already had a fleet of Sparc build servers running everything from Solaris 2.5.1 through 2.9 which I used to build and package open source stuff for our corp servers. Turned out there were just some leftover funds at the end of the financial year in our departments budget and he had to spend it somewhere.
Ah, the beloved "use it or lose it" end-of-year crap. So much needless waste just to keep the beancounters happy.
Back then with just dual cores running heavily loaded VMs always was annoying, and slowed your main system down noticeably - so I preferred not doing that on my x86 box. My Sun on the other hand had three SunPCI cards I used for Windows development and testing - which had pretty decent performance, still allowed me to have the disk images on my UNIX filesystem, but didn't ruin my native performance.
The contemporary PC running MS-DOS or early Windows was just a toy by comparison.
And under high load.
I've had Linux live lock on more than one occasion when load averages hit >100. I've never a Solaris (or FreeBSD) system live lock even under ridiculous loads (200+), and was always been able to login and kill whatever process(es) went sideways.
We used a Sparc Ultra 10 for a Authentication server in 2000, it supports concurrent 100K users without any issue, obviously you need to write your own software, but the server is super stable. And yes, we use cheap x86 + Linux for all sorts of thing from 1996 and it was quite faster but you can not trust it the same way as a Sparc.
We even went through a phase of email on OpenBSD before being bought by a company that insisted on Exchange.
Linux didn’t seem to pull decisively ahead of the BSDs until multicore x86 became mainstream. Up until then it always seemed flaky.
For a brief moment before epoll came along it looked like Linux would get /dev/poll.[1] In fact, IIRC, epoll started as a /dev/poll implementation. I don't think Sun's /dev/poll ever saw much uptake because, aside from the limitations mentioned the kqueue paper, the pace of software development was much more rapid and dynamic in the FOSS and web worlds, and the center of gravity had already shifted to BSD and Linux.
For better and worse, Linux developers seemed more inclined toward adopting extensions from SysV and SunOS/Solaris than from the contemporary BSDs.
[1] See, e.g., https://lwn.net/2001/0712/a/devpoll.php3
It certainly was from 2011 to 2019: https://en.wikipedia.org/wiki/K_computer
> obviously you need to write your own software
Kerberos ran fine for me on cast-off SPARCstation 10(?) which was well obsolete at the time.
I have plenty of fond memories about the stability of Spark Stations but none as fun as that.
Thanks for sharing.
I did once have a Bullfrog TSS that survived a literal explosion. They certainly don’t build hardware like they used to.
HP PA-RISC, RS/6000, MIPS on SGI and others, DEC Alpha itself. Every workstation and server vendor had a proprietary architecture and a Unix variant to match.
In the 90’s, after all, aliens were coming to Earth to steal Intel’s chips.
Granted, the Intel CPU at the high end is pulling 250W+ (or was it 300W+?).
There are places for both architectures. I don't see x86 going anywhere unless Intel folds and ceases to design chips. Not sure AMD could power on alone given their current market share, though I certainly hope they could as a user of their chips in the desktop (and conversely, an M2 Air for my laptop).
A minor quibble—high end desktop is a niche almost seemingly intentionally(?) ignored by ARM. Ampere puts out interesting server chips, and Apple puts out interesting laptop chips. What a high end desktop ARM chip? Apple’s max and ultra chips maybe, but they are pretty clearly a compromise with the fact that that market is pretty niche.
The A18 launched with the iPhone 16 was also supposedly beating Intel single core as well.
The servers didn't seem much better. They'd handle a ton of users, but each would get the same slow experience.
So if you just wanted a good Unix environment, SparcOS was it.
Java/ZFS were both Sun products, and we're still using them today. Just not SparcOS. Sun tried with Project Indiana, but they were getting outpaced by Linux and the open source movement.
OpenSolaris was an interesting experiment.
I was specifically talking about the corporate experiment of deciding to go that hard for open sourcing your crown jewels, and Oracle has notably discontinued their participation in that experiment.
Solaris had (and illumos has) truly unmatched tooling around a number of things.
Solaris 2 and up were derived from System V release 4, which had actually merged the best of System V with both Xenix and BSD, so rather than being purely AT&T Unix, SVR4 was promised as the best of all worlds, with some ability to pick and choose which variety was in play, based somewhat on provision of both types of utilities in separate directories, and appropriate libraries and APIs.
SVR4, IMHO, was the best and most stable Unix, and the right choice for vendors to adopt in those days.
sun's pivot from bsd to at&t was a very nice and clean change (I was the one who ended up upgrading our sunos servers to solaris when the time came in the 90's), sequent's switch was a nightmare.
I still miss my e4500, though, but not the noise or electric bill.
Still true:
$ uname -sr
SunOS 5.11
Not entirely unlike Windows 11, a.k.a. Microsoft Windows [Version 10.0.22631.4169]
[1] Windows 95 also tells 16 bit apps that it's v3.95, but 32 bit apps get the correct 4.0.
(The Sun 386i didn't get SunOS 4.1 nor Solaris 2, at least not at our site, where we had a few sitting around in empty cubicles, and occasionally used for random things.)
oh gosh, I always forget about those, thanks! around here, it was mostly the 68k suns, followed by sparc. the 386 unix variants were mostly sequent.
I remember one fun job interview in the 90's, where before we went to grab beer the interviewer (my future boss) stopped in the office and said "and this will be your sparc" - to be quite honest, that was such a huge perk!
Incidentally, we were a Sun ISV and customer out in the Silicon Forest, where Sequent was located (and Intel, Tektronix...). I initially learned C++ and Smalltalk from an adjunct professor from Sequent. Also where Cray Research Superservers (nee FPS) was, who developed a multiprocessing SPARC system before Sun. Which is how, as a teen, I got sent by a marketing guy to onsite at Cray, to "port" some of our software to the Cray S-MP. It was a nice time and place, with a little like a mini version of being in Silicon Valley, but with more rain.
I also spent a lot of time in the old sequent campus, at the OSDL, and a bunch of time at OGI before it got subsumed.
I still remember the day when Microsoft visited the office, saw I had a Microsoft keyboard connected to my sparc, and asked if I ran internet explorer (they had a solaris version). I laughed and said, "no, I only have 96mb of ram, you can't run internet explorer in only 96mb of ram".
Also in the OGI science park was Verdix (makers of Ada development tools, and some kind of multilevel-secure workstation software).
I don't know what all the other companies were. But amenities included private showers for biking to work, a small forest jogging trail, and a restaurant that made nice turkey sandwiches and huge blueberry muffins to replenish those calories.
the ISP was founded by a mixture of ex-intel and ex-ogi telecom, so there was always a bit of closeness there.
eventually, at the OSDL, we had a Fred Meyer across the street, and not much else. but damnit, we had a great raised floor datacenter!
Oral History of Grant Saviers, part 2 of 2 https://www.youtube.com/watch?v=Od830KDrLUU
Oral History of Grant Saviers part 1: http://archive.computerhistory.org/resources/access/text/201...
Oral History of Grant Saviers part 2: https://archive.computerhistory.org/resources/access/text/20...
'As DEC’s Corporate Vice President of PC Systems and Peripherals from 1990 to 1992 Grant successfully restarted DEC’s PC business from a dormant state and grew revenues to $350M and break-even profitability in 18 months.'
@18 minute timestamp - they copied DELL strategy and did pretty good, business was growing and then DEC founder and CEO Ken Olsen decided to kill it. Grant got recruited to lead Adaptec.
The SunOS -> Solaris transition is an important piece of Sun history.
It really is an incredibly important part of Unix history, not just Sun. It basically really started the outright Unix wars. Had Sun just gone with BSD and tried to create that as a standard, they could have taken most of the world with them without creating a massive blowback counter-reaction that their alliance with AT&T provoked.
And AT&T would have been dead in the water, they might have tried with somebody else, but that would have just cast Sun as the good guys and AT&T and whoever as the bad guys and with Sun already being the market leader, that standard would have been pretty dominant I would think.
I think it was mostly that he liked the bsd tooling over the at&t tooling, but it instilled so much in me that I still have a hard time remembering the gnu command line options: I tend to default to svr4, then to bsd, and finally to gnu. probably one of the reasons I still feel at home on macOS.
Before that, opposite coast, the company had a larger and older 4/390. Which historically performed all "server" tasks (NFS, email, even quietly a UUCP node). Eventually, 1GB IPI hard drive for it, arriving in a package the size of a person, wouldn't scale fast enough to storage needs, so we started getting multi-drive SCSI chassis, and hanging them off random Sun workstations in cubicles. Unlike the mainframe-like big fridge in the locked machine room. And some flaky Exabyte drives scattered around, each handling multiple workstation-turned-server nightly network backups, which backups would fail more often than they worked.
Most folks were either using the new OS X, or Windows, with a custom Linux distribution on the servers, eventually replaced by Scientic Linux distribution.
There were still some Sun stations kind of serving the X Windows sessions on the restaurant area, and even those didn´t last much longer.
In the UK Sun had done some deals of the type that in North America are dominated by IBM (banking infra etc.), so they were probably absolutely milking those clients and giving the rest of us more slack than they wanted to let on.
Still a very cool form factor, though nowadays beware that the PSU is a cap goop timebomb:
https://users.glitchwrks.com/~glitch/2017/07/24/ipc-recap
I greatly enjoy the Sun3 and Sun4c/Sun4m era, always under SunOS 4. Probably because that's what I started on with my first UNIX account (Grex). Probably why I still prefer the BSD-style distros nowadays.
https://www.glennklockwood.com/benchmarks/performance.html#s...
My favorite video of the time is still a guy screaming into a rack of spinning rust and watching read latencies spike on the drives nearest his mouth.
There's a reason Citrix ended up worth $16.5B when they went private a couple of years ago, they were highly successful propagating the thin client vision that Sun championed but fumbled.
And, the future marches on, such as the DECstation 3000 emulator that runs on an RP2040. [0] Seeing a cheap microcontroller doing something like that makes me laugh out loud.
https://users.glitchwrks.com/~glitch/2022/09/19/sun-3-50
...but yeah, the slow factor means it doesn't get bench time, the 3/160 almost always gets picked over the 3/50 for hacking. I'd still like to do a dataless install for the 3/50, I recently refurbished a Sun SCSI shoebox with an 80 MB MFM drive on an Adaptec ACB-4000 and that'd probably be the "best" use for it.
In one of my groups we had a Sun V480 and we ran all kinds of stuff on it and it never had the slightest hiccup. It was rock solid.
Fun times!
You can just cheat: https://virtuallyfun.com/2024/08/12/vncfox-better-way-of-bro...
https://cdn.netbsd.org/pub/pkgsrc/packages/NetBSD/sparc64/10...
A lot of the "wow, this actually Just Works on OpenBSD!" support is because one or more OpenBSD developer actually runs the given system. Supposedly that's why there's really good support for a large number of laptop peculiarities that NetBSD lacks -- OpenBSD devs are actually dailying OpenBSD on those machines.
[0] https://www.forbes.com/2002/05/06/0506sun.html#703713c16a5e