The most impressive thing about visiting the old CRAY HQ back in Minneapolis was this huge building, located in a sub-zero environment, comfortably heated entirely by computers. I'd say that they'd successfully achieved HAC on the inside and CAC on the outside.

Our neighbors have frequently speculated on why we have window air conditioners in our closets. I tell them that we grow mushrooms. Fortunately, they don't yet know that half of our home's 200A service is routed out to the garage.

My experiences: I've been pretty lucky overall. My first "real" machine was a CDC 3300 running MASTER and Compass. Among others, I experimented with writing hardware-level I/O and turned off all of the interrupts during one run (card batches). It never occurred to me that no one would reset such things between runs, and the entire payroll system went down in flames the following weekend. I was not a popular camper after that.

I liked CP-CMS. MULTICS was massively unreliable and it was way too easy to pillage the password file. DEC machines ran RSX, RT-11 and then VMS. I don't remember them failing much (except for those dogs, the microVaxen), but then again they were sort of like personal appliances and restarts were pretty simple and didn't leave psychological scars. Also, no one thought twice about going after one with an extender board and a scope. (Both circuitry and code was pretty much open source.) Applying system patches was a bear, as were system installations.

You could usually debug your way out of a Symbolics crash.

I only used Macs regularly through OS7.5, but I remember being generally satisfied, except when a bad install created extension conflicts: bad stuff. Windows 3.1 and 95 were islands of OK-ness in a world of bad OS versions from MS. Not enough Unix experience to assess it, but I liked BSD.

Our Corporate NT workstations used to blue screen about once a month, mostly due to buggy applications. (One could argue that a buggy user application should not be able to kill off an OS...) But I thank the folks from DEC who created it anyway; it brought the PC into the 20th century.

My Win2K was babied along and ran nearly 24x7 for 5 years, on its original install, before a toxic software download killed it off. I finally switched to XP at that point (2007) because it was apparent that this type of event was going to be occurring more and more often. (My disk drives also started hitting their 5 year MTBFs.) XP SP2 has been rock stable (except for IE). I have one Vista laptop and it is OK as a "media machine", but I otherwise hate its UI-for-dummies. I'm holding out great hopes for Windows 7; hopefully there will be lessons learned.

In all of this, I have been pretty careful about what I downloaded or installed, generally having learned to keep away from the dark corners of the software and hardware worlds. These days, I am mostly disappointed that certain web browsers can bring a machine down. It just seems wrong, sort of like allowing a non-privileged user to take over all of the hardware interrupts.

It has always seemed to me that a deterministic universe, as envisioned by Descartes and Newton, was completely specified by its past history. If all particles in the universe could be exactly tracked through all of their elastic and inelastic interactions, from the distant past on into the distant future (assume that vast computer, Deep Thought, playing celestial pool), then any given human action should be completely determined by the momentary convergence of the past histories of all his constituent particles as predicted by their prior interactions. Thus sounds to me like complete determinism, with no room for free will. The escape clause from this state of one's distant past being their fate seems to appear in the form of uncertainty and other quantum indeterminacy, preventing the possibility of a completely deterministic past history.