Alberto Potterini, Wizard

Way back in the 1990's computer engineers were exasperated by random errors that occurred once every few thousand times. I was exposed to this at work by people trying to handle data that came back from scanner devices. We were getting data errors every few million bytes, or every few thousand scanned pages. These errors were probably caused by quality issues in metals used in conductors, ground wires, connectors and internal components on the scanner and computer. The scanner errors were different depending on which brand of computer was used.

But things have changed and nowadays the peskiest errors are ones that occur every few trillion times. Computers are not only much faster but they are also more reliable if you look at it this way. The other day I was talking to a friend who works with switches meant to handle fiber optic data traffic. These things are pretty fast - I don't really know how fast - except I do know that the frequency of light is the speed of light divided by the wavelength - so the speed of these switches are probably no faster than a few thousand gigahertzes. Nothing new here.

My friend said that the errors occurred about one time in 10 trillion. This seemed like a lot for a device that was completely electrical and not at all mechanical, and he told me the switches were made out of silicon, some capacitors and some wire. The signal they were trying to switch was made out of lased light, running on optical fiber, and was a lot faster than silicon and capacitors and wire. In fact, the light moves through the fiber at close to the speed of light in a vacuum - relativistic speeds.

One way to notice this is to look at light flowing through a prism. The light moves through the thicker part of the prism at a different speed, causing a phase translation that people can visually perceive as a rainbow. The lased light is so fast that it seems likely that the theory of relativity applies differently to the fiber, silicon, metal capacitors and metal wire. This was the first time I had to admit that a device with no moving parts had to be viewed as a mechanical thing at some fundamental level.

My friend was telling me how they worked on the problem - trying to analyze the errors using the same method we had used five years earlier - catching the error in software, comparing the bad data to the correct answer, and then trying to see what the errors were made of. He was told he would have plenty of time to catch the error with software and handle it.

"How much time?" he asked the switch engineer.

"Oh, let's see" and the engineer put his fingers in the manual and measured between the timing diagrams on chart A: "About this much..."

Then chart B.. "Plus this much..."

Then chart C.. "Plus about his much."

Then he waved his hands in the air, adding the three segments of timings together and said "You should have ~at least~ 150 nanoseconds to handle the error!"

My friend said he laughed and told the engineer "I'm not going to be there. Not in 150 nanoseconds! Not with software running on silicon."

A recent review of semiconductors made of pure silicon-28 show speed improvements of approximately 30%, energy savings of 30% due to better insulation, and open the door for greater miniaturization due to less current leakage. Also, P N junctions made out of gallium arsenide are a bit faster than silicon, but gallium arsenide is brittle. Unless we develop lightweight small stable superconductors the problem seems unsolvable: light in a vacuum travels 450,000 meters in 150 nanoseconds, electricity only about one tenth as far.

***
Costs and fees

"Jurisdiction is retained by this Court over this action and the parties thereto for the purpose of enabling the parties to this action to apply to this Court at any time for further orders and directions as may be necessary or appropriate to carry out or construe this Final Judgment, to modify or terminate any of its provisions, to enforce compliance, and to punish violations of its provisions.

"In accordance with the imposition and affirmance of liability, the Plaintiff States shall submit a motion for the award of costs and fees - United States of America

***
"Dammit Jim, I'm a doctor not a lawyer" - De Forrest Kelley as Gene Roddenberry's character Dr. McCoy