In The Economist, July 25, 1987, pp. 73-74, there is an interesting look at the coming automation of commercial airports. Deregulation has led to explosive growth in demand, with consequences all of us who fly can talk at length about. The chart in the article shows the current ~10^12 revenue passenger miles/year almost doubling by the year 2000.(~6% growth per year). The article discusses the introduction of automation to cope with increasing airport and aircraft size. After a brief look at automated ground traffic control, including arrival gate assignment, it concentrates on evolving mechanisms for baggage sorting and handling. There is also a brief mention of the problems associated with sharing ticket counter facilities by small airlines when each uses a different, incompatable, reservation system. Large emphasis is placed on a new International Civil Aviation Organisation rule effective next year that will require carriers to match every piece of baggage loaded with a passenger verified to be on the aircraft. In addition to a brief survey of the technologies involved, there is a lot of discussion of the potential problems: Simple system breakdown, incom- patible tag systems at different airports, accidental mischief (child with toy magnet scrambling magnetically encoded tickets), and malicious mischief (as simple as disgruntled employees entering false gate data, thereby running aircraft around in circles). My favorite "fun" prospect mentioned is baggage tags where the human and machine readable destinations differ... This one reminds of the old check kiting scheme based on checks with mismatched bank name/numbers. The article is very witty, and I just hope the people designing these systems are experienced. I think we can find examples of just about every kind of risk ever mentioned in this forum lurking here. Oh yeah, the solution to the incompatible reservation system problem is CUTE...Common Use Terminal Equipment. Dan Graifer
In RISKS 5:15, Andy Freeman asked, in regard to commercial airline operations: > What are the pilot's responsibilities and liabilities? What about the > controller's? As one of many private pilots reading RISKS, I'll try to answer. In so doing, I will attempt to simplify the system enough to a) be understood by non-pilots, b) not insult too many pilots, and c) not reveal too much of my own ignorance. The statement of the pilot's responsibilities is, like the U.S. constitution, very succinct and, also like the constitution, full of implications: The pilot-in-command is responsible for all aspects of the flight. Among other things, this means: 1) The pilot can refuse Air Traffic Control (ATC) instructions. 2) Whatever the co-pilot, navigator, flight attendants, airline executives, controllers, and everyone else have to say, it is the pilot who has the final word on whether a flight is go or no-go. 3) No matter who's being paid to de-ice the wings, fill the gas tank, inflate the tires, clean the windshield, etc., it is the pilot's responsibility to determine that the aircraft is ready to fly. What are ATC's responsibilities? To put it very succinctly (probably too succinctly): Advisory, sequencing, and separation. This, of course, means a lot of different jobs. These jobs are handled by four different parts of the ATC system which usually involve different personnel and, somehow, intermesh, communicate, and coordinate. These four pieces (not found at all airports) are: 1) Ground control (take care of traffic on the ground up to the runways), 2) Tower (control use of the runways, specifically through takeoff and landing clearances), 3) Approach and departure control (control air traffic in the vicinity of the airport), and 4) Enroute traffic control (control air traffic everywhere else). Ground control and tower operations take place in a tower with a view, approach/departure control takes place in a darkened room full of radar scopes, and enroute traffic control takes place in a darkened room at a regional center (for example, a center in Seattle watches over most of the Pacific Northwest). During the course of a typical flight under Instrument Flight Rules (IFR), ATC tells the pilot when to taxi, when to take off, when to ascend and descend (and to what altitudes), when to turn, and when to land -- all based on a flight plan filed by the pilot. The pilot's job is to worry about which way the aircraft is pointed and what's going on in its vicinity -- ATC's job is to allow lots of aircraft to use the same airspace (clouds and all), airports, runways, and terminals. That, in a nutshell, is my attempt to answer Andy's question. So what's gone wrong with the system? There's been no shortage of finger-pointing. Some blame general aviation (little guys like myself), some blame the PATCO strike, some blame drugs, some blame deregulation, some blame pilot training. The answer, of course, isn't all that simple -- it involves economics, politics, technology, the Reagan military buildup, personnel, unions, etc. So far, nobody in any position of power has come up with any solution more creative than to increase regulation of the already heavily-regulated airspace. My own pessimistic prognosis is that the air traffic system has a lot worse to get before it's going to get better. Nathan Meyers (hplabs!hp-pcd!nathanm) [We have noted Henry Petroski's evidence that we tend to learn little from our engineering successes, but that the real advances come from trying to understand our failures. There is MUCH to be learned from the ATC situation, especially when confronted with the realization that many of the existing problems will continue to exist even when the long awaited new computer systems arrive. PGN]
In RISKS 5:18, Mark Day writes: > The fact that nuclear power plants have been run in a generally safe way in > the past tells me very little about the future danger from them. Predicting > the future like that is similar to the statistical fallacy that if a fair > coin has come up "heads" 500 times in a row, it is somehow "more likely" to > come up "heads" the next time that I flip it. Sorry, but that's true only if you assume that the coin is balanced to have equal probability of falling heads or tails. A balanced coin would have a probability of 2^-500 of 500 consecutive heads; while not impossible, a run of that length would lead a reasonable person to conclude that the coin was balanced in favor of heads. I suggest that the nuclear power industry is still demonstrating a good track record and will probably continue to do so. This doesn't say that there won't be a disaster, but if you want to dismiss the operating history, you should explain why it represents a statistical anomoly. (Coverup, progressive deterioration, etc.) After 500 consecutive heads, it isn't unreasonable to expect a 501st head as well. On the other hand, it is perfectly reasonable to consider whether the potential damage from an accident outweighs the small probability of its occurrance, or whether it is outweighed by the reduction in consumption of non-renewable fuels. Unfortunately, the analysis may be technical, but the decision will be political.
In RISKS-5.14, Alex Bangs raises two concerns relating to nuclear power plant safety (construction corruption, intelligent control systems) and requests references on nuclear power plant control. Let us assume, for the sake of argument, that a given plant may be constructed without "corruption"; I interpret that to mean that the plant is built exactly to specification. Such a procedure does not guarantee that the specification itself is correct; nor does it guarantee that a properly constructed plant built to a correct specification will continue to adhere to that specification after it is put into operation. I do not mean to de-emphasize the problems that result from shoddy construction practices, but merely to point out that these are not the only problems; perhaps they are not even the major ones. With regard to the use of intelligent control systems, a similar plethora of problems exist. Although [hopefully!] intelligent control systems known as "human beings" are now used to operate plants, the multitude of sensor inputs (some of which may be erroneous) and the large number of possible failure modes pose problems that are exceedingly difficult to solve even when copious time is available--which it sometimes isn't. Consider also what repercussions have already occurred for the nuclear industry as the result of [possible] human error during crisis situations; now imagine the outcry if the public at large discovered that a nuclear accident occurred in part because of a fault in an expert system---a "computer error". Two excellent references on the subject are: IEEE Spectrum Vol 16 #11 Nov 1979 and Vol 21 #4 Apr 1984 Both issues cover TMI extensively, especially the first. They are eminently readable to anyone with a technical background; I believe the first issue won a number of awards for its coverage. Finally, a brief anecdote. On March 28, 1979, I was in St. Louis for an interview with Union Electric Co. for a possible position at their then-under-construction nuclear facility at Callaway, Missouri. The interview went fine, but when I came out of their office in the afternoon and turned on the car radio, I began to hear news bulletins about a place in Pennsylvania called "Three Mile Island". I figured it was a sign. :-) Rich Kulawiec, email@example.com, j.cc.purdue.edu!rsk
> [Concerning SINGLE-FAULT-TOLERANT SYSTEMS, I noted recently that most > of the nuclear power plants are designed to remain safe as long as only > a single pipe ruptures. Two pipes are too many. Earthquakes could make > things quite difficult. PGN] First, I'd like to point out that I am not an expert in nuclear energy or nuclear power. I am not entirely familiar with the workings of the insides of nuclear power plants. Even if I were, the fact that there is no standardization among these plants makes quantitative statements difficult. Judging from "The President's Commission on The Accident at Three Mile Island", there are three methods of cooling the reactor: 1: Main feedwater pumps. 2: Auxiliary feedwater pumps. 3: High-pressure injection pumps. Unfortunately among this foot-thick stack of reports I cannot find a blasted diagram of the reactor cooling system. I presume these systems operate on separate lines. But for all I know they could feed into a common line (besides the reactor vessel itself). Breaks in the upper half of the cooling system are covered however. Parts of the contingency plans indicate that it is possible to pump water that comes out of a break back into the reactor via sump pumps. Therefore, from what I can gather, two double-guillotine breaks that occur in the lower leg of the reactor cooling system may be fatal, but two double-guillotine breaks, where one occurs in the lower leg might not be fatal. Note: from the definition of double-guillotine breaks, isn't this four pipe breaks in 2 pipes? Ref: double-guillotine break: a pipe break where a section of pipe is completely removed from the line (needing 2 breaks) and the outgoing water does not impede the flow of water coming out the other pipe. I read this somewhere in WASH-1400.
In reference to the above subject of Computer Crime, I have a book at home that told of a case of a stolen program for estimating bids on the type of work involved by two competing companies. The only reason the larger company was convicted for the theft of the program from it's smaller competitor was that they printed out a hard copy of the program. U.S. law at that time said that what was contained in electronic memory, wasn't "real?" or something to that effect. There certainly is a need for new or more comprehensive laws due to new technology. I also understand banks are required to make hard copies of their accounting programs determinations at appropriate places for auditing purposes. Again what is in memory is not "real". SEG, Pac Bell, Rohnert Park, Calif.
This is an old pet peeve/idea/complaint of mine that some recent postings on passwords being broken have finally prompted me to set down on iron oxide: Claim: Any frequent computer user, including the most non-technical, can/should be able to remember a 10 to 15 character password consisting of a "random" sequence of digits. To demonstrate this, consider the following 2 questions: (A) What's your office phone number? (B) What's your home phone number? I suspect almost everyone can answer both questions correctly. The two together give 14 fairly patternless digits, or 8 if you don't count exchanges. Now compare the frequency with which you dial/speak/write your *own* phone number with the frequency with which you type your password. (This is why I only make my claim for any "frequent" user). At least around here, the number labels on office phones are often missing, so the lack of visual feedback for passwords shouldn't be a problem. - Jonathan Thornburg firstname.lastname@example.org email@example.com
Although Ted Lee's interpretation of "2-man control" does describe a form of implicit separation of responsibility, I suspect that Dr. Ware was referring to what has become rather common in NATO defense compusec procurements, the explicit "2-man (or more) rule". This a mechanism whereby two (or sometimes more) mutually cooperating, authorized administrators are required to perform some action that affects the state (particularly the security state) of the system. The cooperating, authorized administrators may also be required to belong to separate operational groups, etc., depending upon how the system has been screwed together. All in all, it's not a bad idea and I'm still surprised that the US compusec community has not yet picked up on it. Pat Farrell (Control Data Corporation)
Please report problems with the web pages to the maintainer