SGI: Development

Was it because they wanted a fresh start, or did they think that they'd never be able to port IRIX for whatever reason, or did they just think it wouldn't be competitive.

I think they just thought they'd save a bundle by making the switch to Linux. Bye bye IRIX it was nice while it lasted...

SGI was hemorrhaging cash at the time. I was told by people at SGI that there was brief consideration of doing a port, but given SGI's rapidly deteriorating financial position, it made more sense to try to IRIX'ify Linux (by doing things like developing the SGI ProPack and open sourcing XFS and PCP) than to try a full port.

Given the financial circumstances and that SGI's differentiating market was far more interested in the performance you could get out of very large systems than in what might be done at the workstation level (i.e. no one who could write big checks cared about the GUI or desktop stuff), the idea of porting a few key technologies made some sense, especially if you had QuickTransit for running many IRIX binaries under Titanium/Linux and you simultaneously dropped everything else that went into porting, developing, and maintaining what was, by then, a niche OS.

Keep in mind, it would not be long before SGI would start cost-saving measures like dropping licenses for PostScript, etc.

(Note: I am not commenting on the wisdom of moving to Itanium, itself. I am only commenting on SGI's Linux on Itanium vs IRIX on Itanium decision.)

^ That's the clean and pretty version. Here's the down and dirty one :

SGI management was as stupid as a dead fish. They couldn't pour piss out of a boot with instructions on the heel. Every action they took from about 1993 on was dumber than the one preceding it.

Ed McMuffin got starstruck eyes in Washington, bought Cray at Bill C's request, overpaid, gave away the good parts to Sun and moved into retirement.

Tricky Ricky Belluzzo pissed away millions ... hundreds of millions ? ... on his quest to enter the Wintel commodity market. When that failed miserably he took a job as vice-president of toilet bowl cleaning at Mickeysoft, for a few dozen million a year. Hmm.

Bumble Bob, not sure what he did except run the SGI corp-speak program to create the occasional oracular pronouncement for the gullible tech press (an arm of the Ziff-Davis Boot-Licking Toady Marketing Department). He may have been responsible for buying Alias high, then selling low. Two years later the buyer for Alias | Wavefront sold it to Autocrap for a couple hundred mil more. Mostly Bumble just filled his bank account and kept a low profile until they went bankrupt.

They got Dennis McKenna after going tits-up. The first tits-up, I think. He seemed to be doing a decent job of pumping out the flooded compartments and getting the engine running again, so as soon as the ship was stabilized they tossed him.

Last but not least they hired Bozo Ewald, who spent most of his time grinning for the cameras and hiring his friends and buying their companies. And oh yeah, scrapping millionsss of dollarsss worth of SGI product so that ... ummm ... well, I'm sure there was a good reason. Fiduciary responsibility, maybe.

In short, by 1995 they were the west coast's largest group of clowns outside of a circus. Everything they touched turned to shit : Alias, MIPS, whatever. Every time they re-sold a company it turned back into gold. They were convinced they were the coolest thing since Hawaiian shirts and coke spoons but cool doesn't pay the bills - except for the managers, who had several million dollars in cash to piss their way through. Blow dispensers in the bathrooms doesn't create good hardware.

And let's not forget nVidia ...

By the time they got to embracing Linux and Itanic, everyone with a nickel knew their products were way overpriced, the quality was in the toilet, and the SGI sales guys were the most arrogant assholes since Cornwallis' Hessians. When Bozo said, "Okay guys, we want you to migrate" all the customers* said, "YES !!" and proceeded to migrate as far away from SGI as they could get.

*footnote : a long time ago in a galaxy far away, I had some goofy beginner problem with my Indigo2. We had newsgroups then and some guy posting to alt.sys.sgi.* seemed to know his stuff. I sent him a note. His answer burned off my ears. SGI hardware was crap, their software (xfs) was crap, their service was crap, their sales force was crap and evil too, he hated them with a passion. The minute he could talk his superiors into putting up with the hassle, SGI was gone .

He ran the data storage system for the Hubble telescope. ***

I immediately thought of Irv Something, the IBM guy in charge of service for OS/2. Irv posted in the newsgroups all the time, answered questions. So did Scott Garfinkle, who did the OS/2 kernel. Someone once asked if they could do something the poster wanted and Irv answered, "Well, no, probably not for two people." Poster complained "You'd do it for Deutschebank !" and Irv replied "For Deutschebank we drop in a planeload of developers by parachute."

SGI wouldn't send the cleaning lady by Greyhound.

** Oh yeah, the "Inventor's Dilemna", I almost forgot. Does anyone remember Richard Pryor ? And his eight year old who broke the lamp ? "It wasn't me, Dad ! Really ! no, well, uhh, see, like, I was doing my homework in the living room when these aliens landed on the roof, see, and they were chasing me and one of them knocked it over, not me ! Honest !"

What did we hire you for, you stupid dipshits ? to make third-grader excuses for why the company pissed away hundreds of millions of dollars then went broke ? While you walked away with a pretty penny ? How about we set Palantir onto your bank account ? Fiduciary responsibility my rosy red ass.

*** An underlying problem with the "abandon the desktop to concentrate on Big Iron" scenario is that by 2000, the desktops were still decent - a Fuel was quite fast in those days and 4DWm was way nicer than any other Unix desktop. But their Big Iron was crap. So what they did was put their money into the product that had cancer while starving the healthy Ajax. This was to be expected because, as I mentioned at the top, the people running that place had the brains of a mentally-retarded salamander. Everything they knew about computing they learned reading a magazine on the grunter while having a cocktail over at Biff and Muffy's house.

And dat's da ugly truth of why they didn't, Black Fox.

I'd say there was also such an in-built mentality for not seeing the writing on the walls and planning properly. OK they'd planned to move for Itanium for years (and were WAY too vocal about it - hence putting IRIX/MIPS into the same boat as Nokia's Symbian handsets a few years back), but they never really seemed to think about the software side until it was too late.

In comparison, for many years Apple had an internal port of Mac OS X running on x86 hardware "just in case". They knew that PowerPC had its weaknesses and wanted to be prepared. Hence when the time came (ie, PowerPC was too hot for mobile and Intel was finally in the process of dropping the awful Pentium 4), the transition was fairly straight forward and painless because a lot of the work had already been done.

Of course, these companies were in different markets, however one of the main unique things that sets a Mac apart from any other x86 box out there is Mac OS X. Bung Linux on instead and it's not the same.

For SGI, by the time they really started thinking properly about Itanium, they'd done no work in porting, it looked like a huge and expensive job to do, and Linux just looked cheaper and easier. Of course, the downside is why buy an Itanium desktop unit with an Nvidia graphics card and Linux on it from SGI, when you can buy one for a lot less from plenty of other companies? With no IRIX, no internal graphics division and customer CPUs from Intel, there was nothing unique left.

SGI focused for a while on high-end systems as has been written above, but they really threw away too many markets without a fight.

Then there's this surprise:
IRIX for x86 and the 700MHz O2

so sad to see that often for whatever reason only the crap survives

Kumba wrote: Then there's this surprise:
IRIX for x86 and the 700MHz O2

Yeah, an interesting tidbit, for sure. I had heard about an x86 port before they announced the Itanium move, but no one I knew at SGI seemed to take x86 seriously (aside from the NT-oriented folk that Rick Belluzzo brought in). To the degree they thought of x86, it was either as a desktop or a cluster solution, and SGI was all about the large, single system image (SSI) at the high end, and they didn't really care much about the low end or clusters at the time. "Clusters are for kids."

I'm not sure if SGI would have been successful if they threw a lot of resources into a serious commercialization of an x86 flavor of IRIX, but it sure would've been interesting to see how it might have turned out.

At this point I just hate to see the source of IRIX remain unreleased and bitrotting in Rackable's hands. IRIX under BSD or CDDL would be awesome.

TeamBlackFox wrote: At this point I just hate to see the source of IRIX remain unreleased and bitrotting in Rackable's hands. IRIX under BSD or CDDL would be awesome.

Not likely, since the source for much of what we think of as IRIX doesn't belong to SGI (or Rackable).

The issue has been numerous times over the last few years. Here's one: viewtopic.php?f=6&t=8536&p=69371&sid=0acd2b06815dd857d5eb4e1d538d9c79#p69371 (or just try a neko-search of "open source irix").

TeamBlackFox wrote: At this point I just hate to see the source of IRIX remain unreleased and bitrotting in Rackable's hands. IRIX under BSD or CDDL would be awesome.

Reportedly, some amount of source code from an early IRIX 6.5 version got dumped on the Internet somehow. Anyone with a brain will stay away from it, though.

That said, SGI probably faces similar problems as Sun did when open-sourcing Solaris many years ago. With Solaris, you basically got the core kernel and basic system elements, but a lot of auxiliary stuff was kept closed-source due to MOA's, NDA's, copyright and other agreements. Not to mention the whole "Who owns UNIX" ordeal that SCO put the world through 10 years ago. As much as I'd love SGI to put IRIX source out under some OSI-approved license, this would likely require a lot of internal effort and man-hours, something that wouldn't have much of a return on investment.

What I'm hoping for instead is that they become more receptive to requests for internal hardware documentation, such as programming info for various ASICs like HEART, BRIDGE, XBOW, XBRIDGE, BUZZ, etc. That'll be of better use to both Linux and *BSD efforts to port code to these platforms by implementing functionality without having to worry about code contamination.

TeamBlackFox wrote: At this point I just hate to see the source of IRIX remain unreleased and bitrotting in Rackable's hands. IRIX under BSD or CDDL would be awesome.

I don't understand why people have this fixation with "release the source code" ... Is this some kind of nostalgia for Free Willy ?

There is nothing special about Irix beyond 4Dwm, the applications that ran on it, and its integration with SGI hardware.

The applications now run on something else.

The hardware is "old and deprecated".

In twenty years the open source people could easily have created their own version of 4Dwm. If Rackable freed their willy, the only thing the open source people would do to 4Dwm would be to fuck it up.

Freeing the source is about as relevant as releasing the prints to a 1948 Humber. Why ?

hamei wrote: In twenty years the open source people could easily have created their own version of 4Dwm. If Rackable freed their willy, the only thing the open source people would do to 4Dwm would be to fuck it up.

Technically, someone in open-source did: MaxxDesktop. Seems dead, though. I sent the only e-mail address I found a note asking if the project really is dead or not, but no reply thus far. If another open-source project tried to re-implement it, I worry they'll do silly things like tie it to systemd or such.

Kumba wrote: Technically, someone in open-source did: MaxxDesktop. Seems dead, though.

Exactly. Just like every other wonderful "project", it opens to great fanfare and a lot of promise. They get one or two versions down the road and it starts to look useful.

Then it's time to change toolkits, refactor the code, change repositories, change from cvs to subversion to git to boogle, deprecate the old cruft that supports the dead operating systems, change the build system, implement new technologies, write their own buggy streaming editor, change compilers, rename all the icons in the new toolkit, invent a new language to use, remove the icons from the toolkit and make them a separate download, deprecate the new icons and cancel the site where they used to live, change build systems, oh look, it's time to change toolkits again !

Let's not waste any of our precious energy making the program work.

Worthless crap. Discouraging.

josehill wrote: "Clusters are for kids."

well they are but also cheap so ...

Maxxdesktop/5DWM is useless for non GNU/Linux users. I for one don't want an awesome interface over a CLI that is anti user by design and pisses all over what makes UNIX great

Beyond that I'd really like to have their X11R6 implementation opened if nothing else. There are still some shortcomings of X.Org, which could be supported by the code of XSGI.

TeamBlackFox wrote: Maxxdesktop/5DWM is useless for non GNU/Linux users. I for one don't want an awesome interface over a CLI that is anti user by design and pisses all over what makes UNIX great

CLI that is anti-user by design? Do you mean bash?

Zing!

(tcsh user)

Kumba wrote: CLI that is anti-user by design? Do you mean bash?

All inclusive of the GNU commandline tools included with a GNU/Linux distro. Examples:

Sed: BSD and Solaris sed don't have issues with syntax, which makes commands portable between them. However I ALWAYS have to make adjustments to the -i syntax when testing on GNU/Linux.

Bash: Slow and adds useless features to a shell ( pdksh and zsh user here )

Awk: GNU Awk, again incompatible with BSD awk and nawk, meaning I have to add checks for each, or else pollute the other systems with GPL code.

GCC/binutils: Memory hogging, slow to compile, breaks binaries run as -O3 ( Clang doesn't )

TeamBlackFox wrote: ...or else pollute the other systems with GPL code.

I agree that the GPL is a pollutant.

TeamBlackFox wrote: GCC/binutils: Memory hogging, slow to compile, breaks binaries run as -O3 ( Clang doesn't )

I use the shell color syntax code from GNU binutils on my Octane, it works great...