After the Sun (Microsystems) Sets, Stories Come Out | Hacker News Reader

After the Sun (Microsystems) Sets, Stories Come Out

spectrum.ieee.org

64 pointsspectruman37 comments

I had Sun Badge 12386... why do I remember that. (Or my CompuServe account 75460,1375, sigh.)

The area between San Antonio and Shoreline I guess is now mostly Google; before that it was bracketed by Sun and SGI. I know lots of people that basically stayed working in the same set of buildings throughout the last 20 years: different companies, same view. That's soil has grown a lot of wealth.

There was a point in my life, having grown up in north-east England, when I thought this Sun-SGI strip sounded like the most wonderful otherworldly place you could possibly be. Sun workstations were the first machines I'd used that had seemed to be really limitless. And while wasting time at work in 1993 or so, I'd read Usenet posts by people like CJ Silverio at SGI that (though often seriously dissatisfied) had a mood that seemed to highlight how far away that world was from the one I'd grown up in.

I pictured Silicon Valley as a real valley, with low orange sunlight picking out the slabs and flat roofs of the research centres, roads winding to the valley floor below, the sea at the end of the valley.

How disappointed I was when I finally visited the place and found an extended grid-plan office park on flat reclaimed land with the odd hill in the distance.

different companies, same view

I'm not sure if that's an original quote of yours, but it certainly captures the ethos poetically well.

I've spent a large part of the previous decade there. It's been, and still is, a special place in the Valley.

For more Sun stories, the "Life in the Boys' Dorm" blog series is a great read. Here's the first post -- there doesn't seem to be a single page aggregating them all:

http://consultingadultblog.blogspot.com/2010/01/life-in-boys...

I'm speechless...

Bob Coe once told me he did not have to interview the candidates for his Administrative Assistant position. I should simply ask each candidate to link her hands behind her head with her elbows pointing forward and walk toward the wall. If her elbows were the first part of her anatomy to touch the wall, she was eliminated from candidacy. All applicants whose breasts touched first, he would interview.

Sexism is sometimes its own punishment - with a hiring system like that, how many good candidates do you think were rejected?

That just makes me picture an office full of trex-shaped assistants.

These are fantastic stories.

Where's the story about the surge of Linux and commodity technology in the late 90's and Sun's big pivot to avoid a slaughter?

Seriously, Sun sold for less than Skype. Internet start-ups with no revenue have sold for more. What happened in Sun's final days is embarrassing.

As I understand it, a lot of the failure there was Sun's total lack of interest in direct sales between what you could put on a credit card on their web site and mainframe levels of $$$ (several million). In theory various distributors or VARs and the like were supposed to handle this gap, but I can't count the number of times I read a blog posting by a startup founder who found that didn't work and bemoaned the fact that they have to buy lower quality servers from Dell simply because Dell was actually willing to sell to them.

And of course once such companies got big enough to interest Sun's direct sales people, they were adept at making do with lower quality and cheaper hardware; I'm sure Sun missed out on massive sales before they got desperate and degraded the quality of their designs and what they were shipping, at which point it was game over.

SGI's story was somewhat similar. Some really bone-headed decisions were made at a high level, which resulted in the company losing contact entirely with the wide bottom of the market, instead retreating to the narrow top.

It's pretty cool to be able to sell a mega-supercomputer for the price of a small castle once in a while, but it turns out it's not a viable business model long-term.

It is, until people can just click a button and buy a bunch of smaller machines with the same aggregate power for less.

Neither SGI nor Sun were boneheaded, per se. They simply missed the signs of the Internet and cheap commodity hardware. Just as MS missed the Internet and Mobile. And I'm sure today's high-flyers are missing something too.

I guess what I'm trying to say is that in an industry moving as fast as ours, the only viable long-term model is to change your model all the time.

I'm pretty sure a lot of Sun were boneheaded. I had an IPX at work in the early 90s and the 486/66 I bought for home ran rings around it.

Their off again on again attitude to Solaris on x86 and their closing down of their free C compiler ceded the market to Linux.

Oh and about Java. When they brought in their Coolthreads machines Java ran abysmally without special settings. You would think the Java guys would talk to the Coolthreads guys to get nice settings out of the box. Sun were like three companies in one who didn't talk to each other.

It's not just the hardware made Java perform like crap. Java made Solaris perform like crap too.

    http://www.brendangregg.com/DTrace/smc.html

Java reading a single file 12million times increased the boot time from 10 minutes to 30 minutes!

There's a story that the Sun486i, which went as far as a small pre-production batch, was vetoed by the internal SPARC types in part because it was faster than what they were delivering at the time, so the company didn't followup on the wildly popular Sun386i....

On the other hand, you're comparing a 1991 or earlier SPARC chip with the late 1992 486/66. Weitek offered their own after market clock doubled SPARC POWER µP, although I'm not sure it gained as much as the clock doubled 486.

The Solaris story on x86 was poor, indeed: in 1999 I set up a x86 machine with a removable hard drive so it could run the various operating systems our customers used, Solaris was the only OS we couldn't get running on it (and I first bare metaled a UNIX(TM) system in 1981 (sic) so I don't it was our fault). One or two BSDs including the commercial one, Linux, forget what else, but not Solaris.

Then again, as I note plenty of people were willing to run another OS on Sun x86 hardware, but couldn't buy it.

And echoing your other Java comments, a 2001 project I was involved in, running on SPARC hardware, dropped the idea of using Java because the SPARC implimentation at the time was too buggy.

As for their closing down their free C complier, I myself never found it difficult to get GCC running on a SPARC machine through the '90s. But, yeah, that added a barrier.

Re the IPX acknowledged, but I think Sun were still selling them at many times the cost of a 486/66.

Re Sun C and GCC - it's not the difficulty of getting GCC on; its the fact that the whole culture of development started to revolve around GCC rather than Sun C.

Errr, I may not have made it clear, but Sun did make commodity hardware, as in AMD and later Intel server boxes that would run Linux: http://en.wikipedia.org/wiki/Sun_Microsystems#x86-based_syst...

Not "cheap", but of at least perceived high quality, and something that many startups found attractive. The latter just couldn't actually buy these boxes in sufficient quantity.

Lots of companies suffer because they screw up their sales, fail in that sort of basic blocking and tackling.

At one point I hid tchotchkes in my office--coffee mugs, an umbrella--from at least three defunct Sun resellers. And that was before Sun itself went under.

I was an intern at Sun in menlo park in 2002 for a year. One of the oddest memories was going to a town hall and listening to Jonathan Schwartz (then CTO) answering the question "but how will this make the company any money?"

His answer was that engineers shouldn't concern themselves with how any of their work would make money - that was the job of other people. Even then, it seemed like a recipe for making cool stuff no one would buy.

I was at Red Hat's then tiny Asia Pacific office with a couple of ex-Sun sales guys in 2003. The recipe was to walk into all their old customers, replace whatever expensive-yet-underpowered SPARC-based box they were were running with a Xeon box running RHL (then later RHEL), and outperform it.

Sun had great software (shitty stuff, like the old IP stack, got replaced), but their on-again-off-again relationship with Intel screwed them.

Isn't that guy on the left in the top photo the guy who doesn't want people on his public beach in NorCal somewhere?

Yeah. After conveniently forgetting everything in court, his lobbyists now claim he owns everything out to the tides. He probably owns you and me as well.

http://blog.sfgate.com/stew/2014/05/28/billionaire-claims-he...

>"We should have charged $1 a seat for every Java license" and that would have generated billions in cash annually, perhaps saving the company. "There's a fine line between doing good for the community and doing too good."

Would Java have been as widely adopted if it cost money to use the language?

No, and James Gosling's experience with the abject market failure of his NeWS because Sun demanded you to pay "$100,000" vs. X's $0 had a lot to do with the Java team insisting on it being free except "for some specialized versions", i.e. mobile.

My fav sun memory from the time hanging out there (I was a student intern also) was playing with the Sun SPOTs. Apart from the "cloud" - SUN also saw the "internet of things", and had devices that look much like arduinos and RPis, much much before their time.

I dont think the community is that active anymore: http://www.sunspotworld.com/

1) What could Sun have done differently to monetize Java? Or was there no opportunity?

2) Sun was one of the true pioneers of the cloud, but they couldn't capitalize and lead this next generation of computing. (They should have created AWS, for instance.) Was it due to Christensen's innovator's dilemma, or was there another reason?

I have a few theories--I'd love to be corrected.

1. "Cloud" basically didn't matter until we had widespread broadband internet access. Around the beginning of the 00's, that still wasn't the case for many businesses.

2. Consumers hadn't really accepted that the 'net was going to be a big deal, because there really wasn't a lot of stuff there yet that put them in contact with it every day--maybe, for younger people, Xanga or LiveJournal, or whatever, but the massive publishing empires you see today hadn't really formed yet.

3. Sun was very invested in workstation and server hardware, big iron stuff. That core business supported a whole lot of growth, but once Linux and comparatively cheap commodity server hardware became available, that core business got eroded hardcore.

4. Computers were still pretty weak, as were phones and wireless networking. Technology just wasn't primed for mass adoption yet.

Sun never decided whether it was in the hardware or software business. They made outstanding hardware but made software that ran on good-enough x86 commodity boxes. Niagara was way ahead of its time and Solaris was probably the best Unix you could get. The worst error, IMHO, was to kill the workstation business. If they wanted people to develop for SPARC, the only way was to put a SPARC machine on developer's desks. We write software on the machines we have on our desks. If my machine is a x86, I probably won't be very interested in exploring an architecture I don't have access to.

To me as a teenager in the late 90s, Sun workstations still had enough prestige that I wanted to get my hands on one. In hindsight it was irrational, but I felt that these "workstations" were real computers, as opposed to mere garden-variety PCs running Windows or even Linux. Then, at my first job in 2000 as a code monkey in a cubicle, I got to use a Sun workstation for some of my work (by then I was the second or third employee to have that particular box). I discovered that a PC running Linux was better in every way... except that I couldn't run Linux at work (Windows NT 4.0 on the PC). So much for the vaunted Sun workstation, at least by that time.

If you did extensive floating point or needed lots of fast graphics, or needed lots of RAM or I/O, the Sun workstation was more powerful than a PC running Linux, at least up until 1998. By 2000, it was probably less clear. As I recall, it was typical to pay 15 or 20K for the workstation plus display.

RIP my Ultrasparc workstation from back in the day (early 00's) running RF and Microwave design software. Computer labs back in EE school in the 90's were 5s and IPC/IPX, with a few 20 with 64MB of RAM for serious electromagnetic computations. I thought it was the cat's meow when I had my own IPX at home I bought for only $200. Now everything is win blows.

Wasn't the stealth bomber's radar cross section optimized on a Sparc 1?

A lot of Late 80's and 90's CAD was done on SPARC's. In fact you can still find some shops running old Sun CAD machines since its the only CAD software the Production Manager knows or some strange reason like that.

Those old sparcstations ran, and ran, and ran. My IPX outlasted it's battery a couple times, hence programming the mac address COFFEE.

In the 90s Martin county FL the govt workers all had Sparc5 (and later Ultra 5) workstations on their desks. It was very advanced for govt workers at the time, and so were the management practices, Jumpstarting end-user workstations in place while the worker took a lunch break for example.

Eventually they switched everything to Windows due to widespread demand from users, and from a sysadmin perspective, went back to the dark ages.

I've had several Suns over the years (Sun 3/80, Sparc 10, Ultra 10) though I haven't turned one on in ages. In the 90's, Sun workstations were the shit! Until Linux ate their lunch.

At my first real admin job I had root on an E6500 machine and really thought I was the shit.

At my first real job I got to use a Sparc 20 as my workstation. Fantastic machine.

Reply on news.ycombinator.com