Silicon Graphics files for Chapter 11

Silicon Graphics, Inc. has filed for Chapter 11 bankruptcy protection. It’s a sad day for all SGI fans; many have said that SGI started going downhill when they decided to start selling Windows workstations.

I have fond memories of SGI workstations. It was on an Indy that I first surfed the World Wide Web at the Ontario Science Centre back in about 1992 or 1993; I had no idea that one of the buttons on the front was the system shutdown button, so I pressed it out of curiosity. Nothing much seemed to happen except that the front panel LED started blinking, so I happily kept using the web browser (some version of Mosaic) only to have the system shut down on me about five minutes later.

Much later in life I worked for cbc.ca and in the storage room I came across another Indy, which I was told was originally used to host the entire cbc.ca website before it was moved to a Sun E450! My colleague Blake booted up the Indy and found that it still ran, happily starting up Apache 1.3.9 and a bunch of other software with lots of remote security holes, I’m sure.

I also used SGI servers while at the University of Toronto; the Engineering Computing Facility‘s main computing server was an 8-way Origin 3200 with 2 gigabytes of memory, an unfathomable amount of RAM in 1996.

So it’s very sad to see one of the pioneers of high-powered computing be brought to its knees like this. At least if they go down in history they will be remembered for bringing superior technologies like XFS to the world.

1986 was not a good year

Yesterday was the twentieth anniversary of the Chernobyl nuclear power plant disaster. I have to wonder: to what extent did the usability, or lack thereof, of the control room instruments contribute to the disaster? I mean, just imagine trying to make sense of a reactor’s status from this panel indicating fuel rod positions. (The complete gallery has more interesting pictures)

Wikipedia’s article on the disaster has this telling quote:

The unstable state of the reactor was not reflected in any way on the control panel, and it does not appear that anyone in the reactor crew was fully aware of danger.

Still — even if the “unstable state” was reflected somewhere on that massive control console, would anyone have been able to find it in the haystack of instruments?

The Three Mile Island accident is another example of how control console complexity can contribute to exacerbating emergency situations. Again, in that accident, there was no reliable instrument to indicate the statuses of various critical components that had failed. Another telling quote:

There is general consensus that the accident was exacerbated by incorrect decisions made because the operators were overwhelmed with information, much of it irrelevant, misleading, or incorrect.

I’m not convinced that control room “habitability” has improved much since these accidents occurred. Just have a look at this section of the newly-refurbished Pickering A Unit 1’s control room and you’ll see that the instrument panels are just as complex as ever. I worry — if another serious incident were to occur in a nuclear reactor today, would the operators be able to correctly interpret their instrument panels in time in order to prevent an accident?

I leave you with this sobering view of reactor #4’s control room as it sits today, courtesy of Robert Polidori.

Control Room of Unit 4 reactor after the meltdown.

Unbuttoned

This appeared in the The Globe and Mail a little while ago:

It’s a ritual. You walk up to an intersection, press the little pedestrian button that is supposed to change the traffic lights for you, and nothing happens for a long, long time. In this computer-controlled age, you are merely letting the traffic light know you are there. It won’t do anything until its next natural cycle.

And sometimes it won’t do anything at all. New York City’s Department of Transportation acknowledge the other day that 80 per cent of the 3,250 buttons in New York were deactivated long ago, even though the official signs telling pedestrians to push the button to cross the street remain in place. They encourage pressing engagement even as they cause people to miss pressing engagements.

Still, it’s not worth hitting the panic button. That’s probably a sham, too.

/dev/kitty: Inappropriate ioctl() for device

… and now for something light-hearted: How to tell when your cat has been on your keyboard.


Jul 15 22:17:10 jupiter kernel: atkbd.c: Keyboard on isa0060/serio0 reports too many keys pressed.
Jul 15 22:18:00 jupiter last message repeated 33 times

Haha!