Previously undiscovered bugs in a legacy software system and record high trading volumes in one stock are being blamed for crashes that stopped trading on the Toronto Stock Exchange several times in the last few days. The stock is question is Bre-X Minerals Ltd. Its price is in free fall due to alleged misrepresentations concerning assay results for their large gold discovery in Indonesia. Although the Toronto Stock Exchange (TSE) is able to handle large trading volumes (e.g. 23.2 billion shares/year), the frantic buying and selling of Bre-X shares exceeded the normal volume for a single stock by a couple orders of magnitude leading to memory and system congestion problems. In "TSE speak", the active set of buy and sell orders for a particular stock is a "book". Under normal conditions a book contains 200 .. 300 orders. The largest book size encountered before Bre-X was around 1600 orders. The Bre-X book was averaging 2,500 orders with peaks to around 4,500 orders. The TSE trading system is estimated at 3,000,000 LOC (language unspecified). The article describes the code as poorly documented and heavily modified. The software is designed so that when it tries to execute a single order (buy or sell transaction), it reads the entire book for the stock into main memory (probably to match buy and sell orders). Normally this isn't a problem but the size of the Bre-X book caused memory contention/overload problems that apparently crashed the system. This problem was fixed by adding more memory to the system. Subsequent attempts to run the system exposed another bug in the legacy software. There is apparently a memory leak that occurs when an order is canceled. The memory for the order (or possible the entire book) isn't released properly and eventually the system strangles for lack to available memory. The very high trading volume in Bre-X caused a much higher than normal number of cancelled orders. The TSE is working on a solution to this problem and as an interim measure is carefully controlling trading in Bre-X. These problems hadn't surfaced in 20 years of operation because the order books had never been big enough and the trading in one stock volatile enough to trigger the memory problems. [Source: Abstracted from a very well written article by Geoffrey Rowan in the *Toronto Globe and Mail*, 12 Apr 1997]
On Monday, 14 Apr 1997, as a result of greatly increasing volume of e-mail traffic, two Microsoft MSN servers bellied up -- one for users with logon names beginning with C through E, the other with names beginning with T through Z. After two days of persistent reboots, further crashes, and lots of customer complaints, Microsoft finally shut down the entire MSN e-mail service on 16 Apr 1997 for a day or two, to increase storage capacity. This affects all 2.5 million customers. [Source: an article by Julia Angwin, *San Francisco Chronicle*, 17 Apr 1997, D1]
*IEEE Spectrum*, April 1997 (pp.43-47) includes a marvelous article by Nickolai Grudinin and Ilya Roytelman of Siemens Power Systems Control, with the title as in the "Subject:" line above and a subtitle of "The blackouts that swept power systems out West last year could have been prevented by a centralized automatic response system." The lead figure is a map of the Western states showing in sequence the locations of 48 distinct events that occurred during the cascading outage of 10 Aug 1996. (For background, see the discussion in RISKS, beginning in RISKS-18.32.) The title and subtitle give a clear indication of what the authors have in mind -- protecting the power system as a whole, not just protecting the individual transmission lines from overloads.
The April 1997 *Communications of the ACM* (pp. 30-37) has a risks-relevant article by Marc Eisenstadt (title = subject: above), presenting an analysis of some debugging tales reported to him. My favorite among those he cites is this: I once had a program that worked properly only on Wednesdays... The documentation claimed the day of the week was returned in a doubleword, 8 bytes. In fact, `Wednesday' is 9 characters long, and the system routine actually expected 12 bytes of space to put the day of the week. Since I was supplying only 8 bytes, it was writing another 4 bytes on top of the storage area intended for another purpose. As it turned out, that space was where a `y' was supposed to be stored for comparison with the user's answer. Six days a week the system would wipe out the `y' with blanks, but on Wednesdays, a `y' would be stored in its correct place.
The military is filled with people who are constantly trying to justify their existence and, worse yet, think that what they do is the most important concern of everyone else. The computer people are no exception. Recently, there has been an enormous push for computer security. The computer people have gone crazy promoting COMPUSEC (computer security) including safeguarding passwords. As a matter of fact, they have made everybody who touches a computer take an eight block test with 15 questions each on computer security (complete with boring tutorials, fortunately non-mandatory). Since they didn't want to install the test everywhere, the software was placed on central computers. Well, the systems people finally got around to getting my work area attached to the installation LAN. Being the curious type, I nosed around some to find out what resources were available. On my second day of exploring, I found a computer with some files that weren't password protected. One of the files was a database file that I decided to look at. Lo and behold, to my amazement I found dozens of unencrypted passwords. I knew that the program wasn't critical, but I also realized that people reuse passwords. Being the security conscious kind-of-guy that I am, I called the systems office to let them know about this breach of COMPUSEC. I described what I found and tracked down the computer on the LAN to aid the systems office. Wouldn't you know it? I happened upon one of the central computers for the COMPUSEC tests and found the password file for it. Later that day there was an e-mail to all of the users (using a different e-mail system) advising them to change passwords if they used that machine. Systems people: please make sure you follow your own rules. You can be sure that other people aren't.
This was sent to me by a friend who works at a newspaper that I'll leave unnamed: A funny thing happened this weekend. The **ONLINE EDITION** copy is generated by an automated conversion from typesetter copy to online copy that works for the **PAPER**. The program runs on a schedule, every morning at 2:15 a.m. But, of course, there was *no* 2:15 a.m. this Sunday!! The NT servers dutifully jumped from 2 a.m. to 3 a.m. without missing a beat. But there was no copy on **ONLINE EDITION** come Sunday morning. Oops. They had to move as much of it by hand as they could. I'm sure this RISK has been well-discussed in RISKS, but I can't wait to see what happens this fall. Stephen K. Doig, Professor, Cronkite School of Journalism, Box 871305, Arizona State Univ., Tempe, AZ 85287-1305 1-602-965-0798 email@example.com [I presume you might get TWO copies! PGN]
Many readers might be tempted by the availability and reliability of GPS to buy an inexpensive receiver and use it as a highly accurate time source for computer systems. I would caution against relying blindly on GPS to set your clock by. In addition to 1999's 13-bit overflow problem, GPS is currently 11 seconds ahead of UTC. http://tycho.usno.navy.mil/leap.html If one system takes its time from GPS, and another from UTC (like most telecommunications networks do?), then there could be trouble. In some applications, 11 seconds is an eternity. Bernie
The claim was apparently in a forged e-mail asserted to be from me (firstname.lastname@example.org)! I would like to take this opportunity to formally deny that I ever made a 5-line program that cracks PGP available via FTP or Telnet on the all.net server. Fred Cohen can be reached at tel:510-294-2087 fax:510-294-1225
You may have heard of the effort to crack the DES challenge by a group originating from Sweden (http://www.des.sollentuna.se/). This has one very worrying aspect: The organizers don't give out the sources. The reason given on their web site is: > Q5: Will you release the source-code? And why not!? > No, unfortunately we will not release the sourcecode for the client. > This is due to the fact that people may, advertently or inadvertently, > modify the client so that it breaks. This would of course jeopardize > the entire effort, since some clients would not be able to find the > correct key. When the project is finished, we will release all of the > source-code used in the project. There are quite a lot of things a malicious binary expected to soak up cycles of CPU could do: a) The program could do any of the the traditional naughty things (send out password information, install Trojans or back doors, ...) b) The program could look for local passwords, try to crack them, and send them back to the master server. c) The program could also try to crack other codes. The master DES keys of the EuroCheque ATM cards, for example, would be a an attractive target. [There are about 40 million EC ATM cards in use in Germany today; fraud involving EC cards is increasing]. Point c) is especially worrying. I do assume the organizers themselves are honest (mostly because at least two people I know quite well by 'net reputation are involved in this). But even with that assumption, a criminal could still break into the organizer's web site and substitute modified clients. The organizers have take no precautions against this that I can see. There are no PGP signatures of the supplied binaries, not even MD5 checksums (which a criminal could also alter on the web pages). Finally, the organizers also rely on security through obscurity to ensure integrity of their clients: > Both between a client and a server and between a server and the > masterserver, a special authentication method is used to make sure > that it is the correct program in the other end. This is done to avoid > people from disturbing the challenge by reporting in blocks as > finished even if they are not. It's almost unnecessary to say that this is not good enough. Thomas Koenig, Thomas.Koenig@ciw.uni-karlsruhe.de, email@example.com.
Sometimes one's person's privacy is a risk to the rights of others, and privacy should be sacrificed. > "In this instance, people familiar with the new Social Security system > say, there is danger for abuse from many directions: a legal adversary, an > employer seeking to learn about an employee's outside income, an ex-spouse > contemplating adjustments in support." Those sound like usually legitimate uses: 1. Nearly all employers require an affirmation not to work for anybody who might conceivably be a competitor. If a person lied about it, this is the first step to investigating such abuse. 2. An ex-spouse with a support payment of a percentage of income is [otherwise] at the mercy of the person reporting income. Of course, I am an extremist, and I believe that if we want to claim our government is a democracy, then *ALL* government information should be public. Period. That includes all tax forms for all taxing bodies and all taxed individuals, groups and businesses. Only military and court documents could be sealed, and then only for a minimum amount of time. The absolute maximum for court documents, for instance, should be the lifetime of the litigants and occasionally of a witnesses (If a witness admitted to having an affair with one of the litigants, and it was a significant portion of the case, the records could be sealed till the witness' death). Just a side note: If tax returns were made public, that might spur the greatest tax simplification in history, to the point where almost everybody could do their Federal Income taxes in an hour or two.
Oh please, when will the media, and RISKS, stop pampering to the misinformed just because it seems to make a story? ActiveX objects don't attempt to prevent any action, beyond the security pre-existing on the running OS. Authenticode is for identification and integrity, not prevention of malicious apps doing malicious things. Who's said otherwise? Sun says "security loophole" and the media jumps up and down with glee. Fact is there is no "security loophole" since there was never an attempt to prevent the applications functionality in the beginning. Sun says "specially written program containing ActiveX", what they really mean is simply an ActiveX object. What's specially written? Sun says "the program then took over the user's computer", why'd they bother to do that? Besides, its highly unlikely that the ActiveX object actually prevented the user from doing other things. Sun says "personal financial information", how'd they distinguish between personal and business financial information? > > McNealy ... said they see security as a major issue differentiating Java... Oh, so like all the looming issues regarding Java have now disappeared forever? Finally, why we needed to hear from David Kennedy that the ActiveX object was signed is beyond me, and beyond what I thought was the scope of Risks. Who cares if it was signed, was it signed by some reputable business attempting to be legitimate or was it signed by Sun for Sun's internal consumption (or maybe to be included in an up-coming release to their most favoured developer?). The signature is irrelevant, its merely intended to guarantee who wrote the object and that the object is delivered as it was written. It has absolutely no bearing on the content of the object (beyond telling you who wrote the hack), and certainly no bearing on Risks, IMO. Stick to Risks. Accepting ActiveX objects across untrusted boundaries without prior understanding of the importance of the digital certificate is a risk, period. Russ R.C. Consulting, Inc. - NT/Internet Security owner of the NTBugTraq mailing list: http://ntbugtraq.rc.on.ca/index.html
Without fanfare, Pacific Bell has modified the switch software in downtown San Diego to allow 11-digit dialing even of local (7-digit) numbers. By making it simpler to program (especially mobile) computers (or people) to make calls reliably, this reduces some of the risks which RISKS contributors have lamented over the years.
Regarding the use of a database of street names to spew out thousands of amendments to a proposed government bill in the Legislature of Ontario, technology is a sword that cuts both ways. It turns out that, as with many a similar database, the one used by the New Democratic Party (NDP) had duplicate or similarly named items in it (Old Orchard Grove and Old Orchard Grv., for example). The NDP found itself needing to occasionally interrupt the proceedings on points of order to withdraw from consideration those errant amendments. The proceedings were carried on the Legislature's cable TV channel, and I watched the shenanigans at various and found the whole display hugely entertaining (but then, I once participated in a performance of Erik Satie's "Vexations"). I happened to catch this little nugget one night, from Gilles Bisson (Member for Cochrane South): Point of order, Chair: You'll note that the following amendment has the following: Richmond Street and Richmond Court -- r-i-c-h-m-o-n-d street and Richome, spelled differently, r-i-c-h-o-m-e, Court. It looks like the IBM PC broke down just like normal and the PC printed two of them at the same time. So if we used the Mac maybe this wouldn't have happened. It's a PC problem. I'd like to withdraw that amendment. Hansard has transcripts of the whole thing up on the Web, another example of my tax dollars hard at work. The above snippet can be found at: <http://www.ontla.on.ca/hansard/36_parl/session1/house/0497/L176_18.htm> I'd guess that Mr. Bisson was enjoying the fact that the "PC problem" he referred to could be taken to mean a problem with a personal computer or a problem with the governing Progressive Conservative party. Mark Connolly Connolly Design Inc., Waterloo, Ontario, Canada http://www.connollydesign.com
As a ham-radio operator, I use satellite tracking programs. One of the best of these programs requires the user to enter the timezone offset. Here in Melbourne, Australia, the offset is +10 hours. Then it asks to set the Daylight Flag if it is now Eastern Summer Time. When I do, the tracking program shows the time in UTC with Local Time as +11 hours. When we switch back to Standard Time [Spring ahead, Fall behind], I remove the Daylight Flag. Why is such an implementation so difficult, Microsoft? Maggie firstname.lastname@example.org
The AIX time rules can be changed using the TZ environment variable. It is documented under "environment File" (/etc/environment) in Files Reference book. I'm currently using "TZ=GMT0BST,M3.5.0,M10.4.0" which works for most years, depending on our legislators. Andrew Yeomans, Andrew_Yeomans@uk.ibm.com NOSS/VNET: WINVMD(YEOMANA) [Also noted by "Eric Ball" <ericball@VNET.IBM.COM>. PGN]
>It should be pointed out that the United States Naval Observatory ><http://tycho.usno.navy.mil/> distinguishes between UTC and GMT (which is >currently one hour ahead of UTC). No. It quotes "GMT/BST" as currently BST and one hour ahead of UTC. It also confusingly uses GMT as an abbreviation for "Greenwich Mean Time/British Summer Time". The authority defining GMT is the Royal Greenwich Observatory <http://www.ast.cam.ac.uk/RGO/>. In <http://www.ast.cam.ac.uk/pubinfo/leaflets/time/time.html> they state: RGO>In the UK we use Greenwich Mean Time (GMT) which in fact nowadays is RGO>Universal Time (UTC). In the summer we adjust our clocks to British RGO>Summer Time, BST, which is one hour in advance of GMT. > It would seem, then, that Windows 95 is correctly advancing GMT when the > user selects "adjust for daylight savings changes." No. GMT is always UTC. The current UK time is BST = GMT+1. [...]
GMT (Greenwich Mean Time) is the standard time at the Meridian at Greenwich Observatory in London, at zero degrees longitude. It was originally derived from the sun's noon position at Greenwich. For many years it was the "base" standard for the world's timekeeping, local timezones being promulgated at some offset from GMT. UTC (Coordinated Universal Time) is a newer time standard, sanctioned by international standards bodies and kept by scores (hundreds?) of advanced "clocks" around the world. There's also International Atomic Time (TAI) which currently differs from UTC by 30 seconds. UTC differs from GMT by some small number of seconds. These time standards vary a little because of occasional corrections (such as the addition of leap seconds) to account for tiny changes in the Earth's rotation speed. A fuller explanation can be found at: http://www.ast.cam.ac.uk/pubinfo/leaflets/time/time.html For all practical purposes, however, UTC and GMT are the same - and always stay the same. GMT does not change: it neither "springs" forward nor "falls" back. GMT is used as standard (winter) time for Britain and Ireland, but for historical & cultural reasons it is still called GMT and not British <something> Time. I suspect that this is what may cause confusion in the minds of US citizens, programmers included. At 01:00 GMT on the last Sunday in March each year Britain and Ireland advance clocks by one hour. This is British Summer Time (BST), or GMT+1. Clocks are put back an hour at 02:00 (local time, still 01:00GMT) on the last Sunday in October each year. GMT itself stays put. This system of advancing clocks by one hour from spring to autumn was used in Britain from 1916 up to the Second World War. During WWII Britain was run on GMT+1 during the winter and GMT+2 during the summer, partly to allow munitions factories a longer production day. Britain reverted to GMT/GMT+1 in 1948. The changeover dates were adjusted in recent years to harmonise with the rest of Europe. http://www.ast.cam.ac.uk/pubinfo/leaflets/summer/summer.html Interestingly, the UK's Royal Society for the Prevention of Accidents (RoSPA) believes that many lives could be saved in the UK each year if Summer Time was adopted permanently. Using GMT+1 all year round was tried in a 3-year experiment from 1968 to 1970. This was called British Standard Time [unfortunately, also BST! PGN]. RosPA predict more road accidents in the dark mornings, but fewer in the brighter evenings as tired drivers make their way home. They estimate the net benefit for the UK's 50 million or so population at up to 400 fewer fatalities and about 10,000 fewer serious injuries. Political considerations within the UK have prevented this idea from being implemented so far. The Labour party (expected to win the forthcoming election) have said that they will adopt GMT+1, at least experimentally. The idea is that shifting the clock forward for the summer "saves" energy by moving the work day closer to the actual hours of daylight. In our modern, high-speed, global, 24-hour society this concept is perhaps of less benefit than it was. Especially with abundant artificial light and too-cheap energy... but that's an argument for another time and place ;-). However, there are now many more time-dependent systems to change twice a year than there were in the 1940's and perhaps the costs are beginning to outweigh the advantages. The RISKS of something new or unexpected going wrong with computer-controlled systems because clocks are out of sync seems to me to be increasing with the number and interconnection of such systems, and our increasing dependence on them. Perhaps it's time to seriously consider dropping Daylight "Savings" Time (originally an American idea!) altogether. Bernard Lyons (email@example.com) Dublin, Ireland.
[... stuff duplicating Bernard Lyons' note omitted. PGN] BST is sometimes used (totally incorrectly) by the uninformed to mean "GMT or Summer time, as the case may be, depending on the season". There are examples of this on various web pages, some I am ashamed to admit originating in the UK. I note that during WW2 in the UK Double Summer Time (GMT+2hr) was kept - if carried on today (as has been suggested), this would give scope for confusion with DST=Daylight Saving Time? Useful basic info giving an idea of the complications of time zones in Europe is shown at: http:/wsspinfo.cern.ch/file/sunos-europe It is clear that it is impossible to program in future summer time changes more that a year ahead, given the reliance on government decision rather than precise formulae. Very RISKy to rely on your OS getting it right for you. Ian Stephens <Ian.Stephens@b-g-trading.btx400.co.uk> [And there will still be screwups... PGN]
"Network Security", Charlie Kaufman/Radia Perlman/Mike Speciner, 1995 %A Charlie Kaufman firstname.lastname@example.org %A Radia Perlman email@example.com %A Mike Speciner firstname.lastname@example.org %C One Lake St., Upper Saddle River, NJ 07458 %D 1995 %G 0-13-061466-1 %I Prentice Hall %O +1-201-236-7139 fax: +1-201-236-7131 email@example.com %P 505 %T "Network Security: Private Communication in a Public World" For communications security, this is the text. A solid conceptual background covers cryptography and authentication. The number theory basis of much of modern encryption is provided as well. In addition, there is overview coverage of specific security implementations, including Kerberos, PEM (Privacy Enhanced Mail), PGP (Pretty Good Privacy), and a variety of proprietary systems. Where many security texts use only UNIX examples, this one gives tips on Lotus Notes, NetWare, and Windows NT. The explanations are thorough and well written. The organization of the book may be a bit odd at times (the explanation of number theory comes only after the discussion of encryption that it supports), but generally makes sense. The end of chapter "homework" problems are well thought out, and much better than the usual reading completion test. copyright Robert M. Slade, 1996 BKNTWSEC.RVW 961209 firstname.lastname@example.org email@example.com firstname.lastname@example.org
Please report problems with the web pages to the maintainer