McMinnville, OR (AP, 17 Nov 1992) -- The Yamhill County clerk discovered a computer programming error that reverses the election results of the county's district attorney's race. Incumbent District Attorney John Mercer didn't lose in the November election -- he won by a landslide. Clerk Charles Stern said the computer error occurred because the program failed to list the candidates in alphabetical order, as they were on the ballot. Mercer had supposedly lost to Bernt "Owl" Hansen, 16,539 votes to 8,519 votes. On Monday, the clerk's office told him Hansen's votes were actually his votes. Mercer said he was astounded at the turn of events. "The feedback I was getting everywhere during the campaign was very positive. And that's why it was such an emotional extreme to see that I'd lost," Mercer said. "But this is really just as shocking the other way." Nathan Meyers firstname.lastname@example.org [Stern Warning: Once Bernt, Twice Mercerized. (in the proper 2:1 ratio) PGN]
In high school I campaigned for a Democrat in a town near Buffalo with a Republican "machine". He said the the single most important thing to do on election day is to get someone to EVERY voting machine at the very hour the polls open, to cast their own votes but also to *test* the machines. And if ANY problem is found, you demand the machine's closure, and telephone the Board of Elections just to make sure. In our case every voting machine in town was set up to allow only straight party-line voting. Hurried calls to the county Board of Elections [run by the Democrats] got the machines closed until they were set right, later the same morning. To keep this relevant, in the example RAY cited it was quite evident that the voting system was not working properly, but in general can *-electronic-* voting and tabulating systems be checked by users for correct operation ? If so, RISKS readers can offer their services for the morning of polling day, to the party of their choosing. If not, don't be surprised when "accidents" occur. fred :: email@example.com [Better make it the WHOLE DAY, not just the morning. And keep your eyes open for those curiosities about which you should by now be aware, as well as others as yet unexposed. PGN]
Time-out costs BA dear Its computer system may have cost the airline millions in lost earnings, missing spare parts and legal expenses. Chris Blackhurst, Independent on Sunday, 15th November, 1992 That was the headline over an article relating to a computer system called `Total Inventory Management Engineering' (Time) which British Airways (BA) introduced in July 1987, at a cost of 10M UK pounds. Time, designed in-house, governs BA's aircraft parts and stock control operation, handling 250,000 parts worth 400M pounds. It is suggested that problems have arisen because when Time was installed it was initialised with inaccurate current stock levels taken from the original manual stock system (known to be as much as 45% out), and these have apparently never been corrected. The article claims that this has affected BA in a number of ways: a) General Electric took over servicing aircraft engines for BA in 1991. BA initially claimed the transfer of 53M pounds worth of spare parts. General Electric have nearly finished counting them, and have found only 30-35M pounds worth. b) In October 1991, BA submitted an insurance claim for a fire at its Gatwick (London) warehouse. The claim included 50M pounds worth of spare parts. The loss adjuster's report hints that BA's figures are not entirely reliable, and valued the lost spares at 28M pounds. c) The prosecution of 12 people on theft of aircraft parts and conspiracy was based substantially on evidence from Time. 9 were acquitted, some of whom are bringing legal proceedings against BA for wrongful arrest. During cross-examination, the person managing Time admitted that when it was installed in 1987 40,000 items, including 94 complete aircraft engines each valued at 250,000 pounds, were found to be missing. (The article is not too clear on this point, but I presume it means to imply that concern over accuracy of the data produced by Time contributed to the collapse of cases against some individuals.) d) Lack of confidence in the reliability of Time has lead to it being ignored in some instances. In one particular case, an engineer did not consult it when refitting a cockpit windscreen. As a result, he used the wrong bolts, and the windscreen blew out in flight, almost sucking the pilot to his death (June 1990). BA dispute the interpretation of events referred to in this article, suggesting that there is no disagreement with General Electric, and that in the case of the fire an initial `guesstimate' had later been revised. John Jones, Department of Computer Science, University of Hull, UK.
The British magazine _New Scientist_ has in its issue of November 14, 1992, two articles of interest. The first relates to the recent discussion of the London Ambulance Service. The article states that the review began last week and a report is due in February. The article begins: "An overcomplicated system and incomplete training for control staff and ambulance crews are the likely causes of the collapse of London's computerised ambulance dispatch service two weeks ago. One software company says that the London Ambulance Service (LAS) underestimated the pressure placed on staff at the control center, and that it makes working there `like a wartime action room.'" The article continues with general observations about system complexity and a description of the process of ambulance dispatching that the system was intending to automate. The computer consultant working on the review panel, Paul Williams ("from the City firm Binder Hamlyn"), is described as having 13 years experience but he has never reviewed a safety-critical system. He intends to compensate with "expert help from his firm and the computer industry." [The Tied Typer of Hamlyn? PGN] The second article is a four-page discussion entitled "Battling on with veteran computers." The major theme is the problems that are created by trying to keep aging software and hardware going. Examples discussed include the Patriot missile system, IBM's Customer Information Control System package, the recent upgrade to the Space Shuttle on-board computer system (we're up to a whole megabyte of memory now!), porting the software for power distribution in Britain from archaic Ferranti Argus 500 machines to modern equipment -- (I interject here a wonderful vignette: "The software for the initial system was written in a language called April, which disappeared long ago. But the problem was not the rarity or age of the language, it was the lack of documentation. Three years after the system was delivered , the CEGB [Central Electricity Generating Board] decided to develop its own software. Today the system is maintained by a lone programmer who has been working on the system in assembler for 20 years. Ask Derek Roberts, the group head of control facilities at the national centre of the National Grid Company what would happen if that person fell under a bus, and he pauses. Then he replies: `we don't like to think about that.'" We now return to our regularly scheduled programming.) and the early flight control system for the Boeing 747-400. According to the article, so long as there are three copies of any aircraft type still flying in the US, the avionics manufacturer is required by law to continue support -- so Honeywell (which bought Sperry Flight Systems some time ago) is still cranking out gauges and regulators for DC-3s. Something new to add to everyone's burgeoning files.
According to Marketplace on American Public Radio, a stock on the New York Stock Exchange (I don't remember the company) closed above 10000 on 16 Nov. This is the first time any stock has been above $10000, and as you might expect, the stock exchange's computers couldn't handle the 5 digit price. The price rise wasn't incredibly fast, as the stock was up 400 for the day, so one hopes they saw this problem coming and dealt with it, but the report I heard had no further details. There's nothing particularly surprising about this report, as we've seen lots of similar examples. After a while it's more depressing than surprising to see the same mistake over and over again. --David Wittenberg
The NewYorkTimes Magazine had an article on October 18 about CHIPS, the financial clearinghouse for major American banks, which handles one trillion dollars electronically every day. Although 85 percent of all transactions are still made in cash, and only 2% electronically, the electronic payments make up 85% by value. The article examines some of the possible risks in this system. The hardware is run off of storage batteries, in a room with a Halon fire extinguishing system. But on Oct 1981, "a hardware breakdown took out both New York computers" and "processing was interrupted for five minutes" until backup systems (on an "independent communications grid") in New Jersey were brought up. Users "would never have known" if they hadn't been told. Messages are verified/encrypted in such a way that someone intercepting a message couldn't just change a dollar amount. Once, in 1989, some criminals (with inside help at a Swiss bank) used CHIPS to help steal $20M(illion). They wired money from the Swiss bank (entering a fake deposit on the books) to Australia, and quickly spread it around. Though they have been caught, only $8M has been recovered. The electronic system merely helped them disperse the large amount quickly. The bigger worry is a loss of confidence. Unlike in the similar European system, all debts are netted at the end of the day. Each bank either owes some amount to the center, or is owed money. If one bank fails to meet its obligations, all transactions involving it that day are supposed to be "unwound". This could, of course, lead some other bank to no longer be able to meet its own obligations for the day, causing a cascade. CHIPS does allow each bank to set a limit on how much it is willing to be owed by all other banks; this limit is monitored continuously, and so a cautious bank could avoid problems. The Federal Reserve runs a similar system, and once had to make an overnight loan of $24 billion to the Bank of New York "in order to settle the day's accounts on transfers of Government securities that got fouled up in a software snafu." Of course, these days such securities are really just electronic entities stored with the Fed, so the overnight loan was well collateralized, and evidently the situation was fixed the next day. The article says this could not happen on CHIPS, because each transfer must be originated by the payer. [I don't know what this implies about the Fed system.] The article concludes that "what all the experts fear is what they do not know." -John Sullivan@geom.umn.edu
BANK SYSTEM IN CHAOS AS MICROCODE BUG STRIKES By Randall Jackson November 1, 3pm: a date and time users of Tandem's CLX systems around the world won't forget in a hurry. That's when a microcode bug struck, sending system timers incoherent and causing chaos in applications such as EFTPOS and automatic telling machines. The bug was discovered first in New Zealand, which is the first country to greet the new day. "Literally, a bit seemed to fall off the field and the timers went incoherent and began talking to themselves," says Ken Hennessy, chief manager at Electronic Transfer Services (ETSL), which manages EFTPOS in New Zealand. "They took the date back to December 1983." There are five CLX installations in New Zealand, including Westpac, whose ATM system crashed at the same time as EFTPOS. Hennessy says Australia was the next affected, then Asia. "I believe Japan was a hell of a mess. "We had been in touch with Australia because ETSL operates contracts there, and they started to notice the problem. They contacted Tandem and the Americans became involved. "By midnight, Tandem had worked out a way of getting around the problem." That was important, because Tandem was able to advise all its users in America and Europe and prevent systems crashing there. Hennessy says EFTPOS in Wellington was up and running again by 6.30pm. "We turned the clocks back two years to give us a clearance into 1990 at least. Then we had to raise each host and hope it didn't cause problems of irreconcilability. It didn't, because it was day-to-day, month-to-month. "Our Auckland node came up at 9:40pm and in the early hours of Monday morning we got back to 1992." Hennessy says that there were two fixes: rolling the clocks forward past 3pm then shifting them back so 3pm wasn't hit, or waiting until 3pm rolled around, and doing a cold start. Typically, New Zealand businesses affected on a Sunday were supermarkets and petrol stations. Foodstuffs Wellington retail systems manager Alistair Garvie syas the loss of EFTPOS was a major inconvenience. "One of out largest stores does 25% of its business through EFTPOS, and customers were complaining about having to pay cheque fees instead," he says. BP spokesperson Beppie Holmes says there was some inconvenience but the company was able to revert to paper based transactions. "Where it did affect us was in our ability to provide cash to customers, which has an effect on residual business," she says. Tandem New Zealand manager John Simms says it took about four hours to work out an answer to the problem, then communicate it to customers. "There was a microcode defect that caused the internal clock to be read incorrectly. It affected different applications in different ways," he says. "It was a field where at rollover the bug caused the data to be interpreted wrongly. "We got our customers to cold load and then reset correctly." Simms says Tandem acted quickly to provide a fix. "It would happen again in 2001 if we hadn't fixed it," he says. From Computerworld New Zealand, November 9, 1992: Don Stokes, Network Manager, Computing Services Centre, Victoria University of Wellington, New Zealand +64-4-495-5052 firstname.lastname@example.org (wk) email@example.com
The following material is from "Space Digest" v15 #425, distributed as "Space@ubvm.cc.buffalo.edu" The article deals with the newly available, from the RUSSIANS, satellite photo imagery with resolutions of 1.5 meters. This is good enough, to pick out individual cars in parking lots (although not to read the apocryphal license plates). They expect a bit more sharpness after some technical problems get resolved. This is a curious "RISK." On the one hand, it makes all sorts of overhead photographic info available. On the other hand, it also makes it (almost) available to the general public. Is it a "RISK" to find out how many Japanese fishing trawlers are out there? What about which cars are parked overnight at the take-a-buck hot sheets motel? article follows: 4- RUSSIAN MILITARY SPACE OBSERVATION DATA ON THE MARKET [Ran across a couple of interesting notes, with interesting ramifications.] Central Trading Systems in Arlington, Texas has a new product. Digitized, very high resolution Russian "Earth Observations" data. This data showed up about a month ago when some demonstration data was circulated within the industry to see if there was some interest in buying it. Folks who've analyzed the data say it's in the 1.5-2 meter resolution range. At that resolution, you can pick out the Christmas tree in front of the White House, or pick out individual cars in the Pentagon parking lot on the demo tapes data. Some rumors circulating in the industry claim the data could have even a higher resolution quality, but the data has been poorly digitized from photos. This data is obvious from a former "strategic asset" of the Soviet Union. Central Trading systems, can't identify what satellite generated the photo data, but that the Russians call it a "DD5" system, for Digital Data 5. As a representative of the data seller Central Trading Systems is offering global coverage with an extensive data archive of digital images. If the scenes are in the archive, customers can have the images on data tapes within 2 weeks, delivered by Federal Express. If new scenes are required, they can be delivered with 45 days, weather permitting. Central Trading Systems thinks the data is delivered digitally in Russian, transferred to photos, and then re-digitized. His offers the possibility that resolution can improve as more advanced digitizing and image processing systems are applied. Cost for the data is $3180 (including shipping and handling) for a 13 x 13 Km, 8-bit scene, of 40 mps at 1600 bpi. Demand is reportedly high. As a side note, on 2 October, a top Russian space commander stated the Russian military space program will only survive by sharing its expertise and hardware. Col General Vladimir Ivanov was quoted in a Krasnaya Zvezda interview as recommending Russian military space systems be used for commercial and civilian purposes. In particular, he was reported to have stated "Reconnaissance satellites can be successfully used for long-distance probing of the Earth's surface and for ecological monitoring without impairing their main task." [Commentary: New competition in the Earth Resources market area. There are reportedly warehouses of high-resolution Earth observation data on both sides of the ex-Iron curtain. Different organizations have been selling ex-Soviet observation data in the 10-meter resolution class, but the data availability and market response has been poor, partially because the data was only available sporadically or only in photographic form. (For obvious reasons, the preference is for data in digital format.) But if true, a marketable archive of global 2 meter or better data could be a market gold mine. And the Krasnaya Zveda quote could indicate regular availability to high-resolution data from Russian military systems could become official policy and routine. SPOT and Landsat data is about an order of magnitude more coarse, with some gaps in the digital data coverage available. The Russian data prices are also very competitive. I expect if the initial expectations are proven for this Russian data, then it will capture a large share of the market within a few years. Again, there can be a substantial commercial market pact from an ex-Soviet system. Due to policy considerations, the US government has been reticent to release high-resolution Earth Observation data, and has encouraged the use of 100-meter resolution Landsat Data for commercial or non-critical government needs. It was only last month the US Department of Defense even officially revealed the existence of the office which controlled such space assets. Similarly, SPOT, which has a very large ownership share by the French government, has not striven to achieve the maximum resolution in its system. A higher resolution has been expected in the French military HELIOS observation system under development. Perhaps the sale of high-resolution Russian data will encourage the release of high resolution data by Western governments. But this will also decimate the existing SPOT or Landsat/EOSAT data markets, when they still have not reached a critical mass for full commercial viability. The best result would be the encouragement of the construction of commercial Western systems with equivalent capability, which is well within the capability of the industry. As it stands now, there are still significant unknowns in the future of commercial Earth observations data. This new source of data, if it is proven as reliable and accurate, could substantially change some of the market assumptions for Earth resources data.]
>From the December issue of Popular Mechanics, Tech Update column (paraphrased) Workers are installing on all 2400 Greyhound buses an on-board radar system made by VORAD Safety Systems. One radar beam will scan ahead for obstacles while a second will probe the driver's blind spot. Steering, braking, speed and obstacle closing rates will be recorded by a 'black box'. VORAD is already testing a system on passenger cars that links the radar and cruise control, enabling the car to maintain a constant distance away from the vehicle ahead. (no longer paraphrasing the magazine) "The next step, says VORAD, is to connect the radar directly with the brakes, to decelerate the car before the driver has time to react to an obstacle." The RISKS seem obvious enough to me... Steve Mestad, Physics Research Division, Superconducting Super Collider Lab 2550 Beckleymeade Ave., MS 2003 Dallas TX 75237 firstname.lastname@example.org
>From the Dallas Morning News Friday Nov 13 issue, in the Line One column (an advocate column of sorts): Person's problem: (paraphrasing salient points) Person went to renew their driver's license during lunch; paid; was photographed and taken to the back. There they were informed of an outstanding warrant and told to either pay the fine or be arrested. Person admitted to old speeding ticket which was allegedly paid. Previous queries of driving record and traffic stops did not reveal anything about the warrant nor was any notification received by mail. Response from Texas Dept of Public Safety: (again paraphrased) Signature on citation is promise to contact/appear in court by date on citation. Failure results in issuing the warrant. Issuing trooper enters warrant into the Warrant Data Bank (WDB). Warrants are placed in WDB are for traffic citations issued only by the Dept. Anytime license record is checked, outstanding warrants will be indicated. Some police depts. do not serve warrants on license checks so a person may not be notified at a stop. Warrant information is not provided on driver's record checks. With the start of the WDB, the Dept. no longer sends mail to advise of issuing a warrant. Steve Mestad, Physics Research Division, Superconducting Super Collider Lab 2550 Beckleymeade Ave., MS 2003 Dallas TX 75237 email@example.com
I recently heard from someone who *works* on British Airway's flight booking system that it is only lack of access that keeps hackers out - the system it runs is completely unprotected - a multitasking system where every task can access the memory of other tasks. And they're scared to make major changes to it in case it falls over. So he told me. Season with salt as desired.
The coin collecting column in the Books section of the Chicago Tribune of Sunday, Nov 8th is about counterfeiting paper money. Among other things it says: Meanwhile, Canon USA has reported that it soon will add either one or two counterfeit deterrents to its new color copiers in an attempt to thwart would-be forgers. One technology places an invisible code on every copy made so that police could trace the machine that duplicated a dollar bill or other documents. The company also might produce machines that print black copies of greenbacks and other bank notes because of information programmed into the machine's computer memory. I see a risk that these "invisible codes" will be used not only to track counterfeiters but also whistleblowers, government critics, and those who only want to be able to communicate privately. The risk increases if (when?) the authorities require that each color copier/printer's "invisible code" be registered. I'm also unhappy with the idea that my printer will try to enforce laws about what I can and cannot put on paper. How accurate will it be? Also, the scheme creates the risk that more color copies of money will be produced. Who could resist trying to fool the censor-in-the-machine? Carl Kadie -- firstname.lastname@example.org -- University of Illinois at Urbana-Champaign
Just discovered a feature that will probably amuse other readers of RISKS. A certain very-popular-workstation-tape-storage-device will reload its firmware upon finding a firmware-reconfiguration tape within its maw upon power-cycling. Presumably it reads whatevers loaded upon start up and upon finding the right code, interprets the data as destined for its EEPROMS. Totally convenient but amusing to a reader of RISKS. David Honig
The November 92 issue of Scientific American has an interesting article on the risks of computers and proposes three different mechanisms to limit them. Nothing there that regular readers of this forum won't have seen before, but spelled out in clean language that anyone can understand. Note that this is a case of circular reference, since the article ends by recommending this forum as a good place to learn more about risks. Greg Captain W. Greg Phillips, Royal Military College of Canada 613-541-6367
Please report problems with the web pages to the maintainer