Are there any REAL examples ?greenspun.com : LUSENET : TimeBomb 2000 (Y2000) : One Thread
I know of course that there are micro-processors everywhere since many years and that when there are micro-processors there are programs. And since programs are written by programmers anything can of course happen. I am not convinced however that embedded processors should be a major case of failure, since I believe that if many of them have to keep track of time, they dont usually track the date. If a microprocessor tracks a date, then there must be a facility to restore or change the date. One should probably only be suspiscious of anything where the date can be changed one way or the other. This is not the case of "real" embedded systems. There is a difference to be made between embedded systems (for me this means hidden, black box type systems), and computers controlling something. Is there anyone who could contradict me by providing a REAL example. By REAL I mean not something that could theoratically happen (in fact embedded systems can fail anytime, not just on 1 jan 2000 and the worst could happen), but some tangible facts.
-- SCHEIDT Robert (firstname.lastname@example.org), June 23, 1998
The reason so many microprocessors are installed and working is because they are relatively cheap. The reason they are cheap is because semiconductor manufacturers made them into a commodity. They way they did that is by marketing the same chip for different applications. Therefore the same functionality and feature set is available across a given manufacturers line regardless of whether or not a given application accesses that functionality. Virtually all MP sold as ECU rely on date/time functionality at some level. Whether or not this will shut down the chip remains to be seem. You can go to some manufacturer's web sites to see which ones (www.intel.com)they sold that will shut down. (Of course that won't tell you which application is going to fail..)
As for the logic that if mp tracks a date then there must be facility to change it has 2 problems. 1) That's not necessarily true because of what I stated above. The application may not even use the time/date at all, but because it is there the chip will fail. 2) Let's say all MP with this problem COULD be changed. (although we know they can't - some must be replaced)Then sombody magically identifies 1/10 of 1 percent (most experts say the number is actually MUCH higher) of all 50 Billion Embedded Controllers need to be fixed. That's 250 Million Microprocessors that need to be reprogrammed.
Your job, should you decide to accept it, is to fix all of them June of '99 so you have at least 6 short months to make sure the fixes actually work.
Good luck, Jim.
Maybe someone else can point you to examples. There are many. Try www.euy2k.com.
-- Sam Loy (email@example.com), June 23, 1998.
Here is some information contained in a report e-mailed to me today which may be helpful, since it names names and quotes numbers:
The Year 2000 in Industrial and Process Control Systems Conference was held in Houston, May 18-20, 1998. The conference was hosted by the International Quality and Productivity Center.
A session was put on by Evan Hand, a Kraft engineer who talked about what Kraft was doing about its embedded systems problems. Kraft Foods has more than fifty manufacturing and distribution centers and more than 50 co-manufacturers with over 55,000 potential problems. They also have to assure the compliance of 5000 vendor products. At the beginning of his talk, Mr. Hand told the audience that he would give them an overall view of the remediation effort at Kraft, then he would get into specifics and give them the actual numbers. Kraft has already had some year 2000 problems surface. Several million dollars worth of food was destroyed because the food expiration date was after the millennium. Because of two digit date encoding, the warehouse had processed the shipments as expired and destroyed them.
Kraft has tested 832 Programmable Logic Controllers so far and had found 10% of them to have a date capability, a higher number then they had initially estimated. He also talked about what kind of problem the tested machines had such as a subtraction problem or date roll over. At Kraft, the PLCs control safety and food production, so any glitch will shut down the whole line.
-- Dan Hunt (firstname.lastname@example.org), June 23, 1998.
Thank you all for your feed back.
Sam pointed me to a quite good site giving some good examples. See in particular : http://www.euy2k.com/reallife.htm http://www.euy2k.com/plcemb1.htm which explain quite well how those things happen.
However, they all refer to systems which receive a date from an external systems, and as I said in my question, one should be suspicious on systems where the date can be changed, since obviously this indicates that the date can be used and misused. I also saw an example : http://www.euy2k.com/embedded.htm where in fact it is the PC giving the date which is non year 2000 compliant and which has the problem most PC's sold before 1996 have. I do not fully agree though with Sam when he says that virtually all MP rely on a date and time function. I would say that nearly all rely on a time function to track elapsed time between two events (expressed by a value in some counter, which can be translated in millisecons, seconds or more), but they dont necessarily need a date. In fact micro processors dont have a true date functionality : it is microcode or programs who translate the values of a counter into dates.
I think this discussion is interesting and I appreciate the feed back. Note that I work on computers since 1961, and have quite an experience both on mainframes and on micro processors from fixing them and programing them. My experience is however more in the commercial and financial world and not at all in the industrial part. I am involved in some projects on Y2OOO essentially in banks and have practical knowledge of what could happen there.
-- Robert SCHEIDT (email@example.com), June 24, 1998.
For what it's worth, Kraft Corporation has been very forthcoming in it's findings...even sharing them with competitors. They've found an average of 10% of their embedded systems date problematic, which is about double what they thought it would be (5%).
-- Pastor Chris (firstname.lastname@example.org), June 24, 1998.
:) Didn't read the post previous to the one I just sent...apparently my embedded system needs a little work also. :)
-- Pastor Chris (email@example.com), June 24, 1998.
Yes, there are REAL examples. To *begin* your research look at the following:
http://web.lexis-nexis.com/more/cahners/11388/3430638/20 http://www.nando.net/newsroom/ntn/info/061698/info19_9529_noframes.html http://www.pathfinder.com/fortune/1998/980427/imt.html http://www.senate.gov/~bennett/pr609b98.html http://www.cbn.org/news/stories/980609.asp http://www.orlandosentinel.com/news/0525bug.htm
-- Rocky Knolls (firstname.lastname@example.org), June 24, 1998.
You misread what I stated. I said virtually all MP HAVE date/time functionality. I never said that most MP DEPEND on date/time. (In fact I thought I did a pretty good job of explaining it) That was the whole point: Most MPs contain date/time functionality regardless of whether the MP itself or its embedded application even use that functionality. And yes, some chips actually do depend on it even though the application may not. Does that help?
-- Sam Loy (email@example.com), June 24, 1998.
A related post that you might want to look at is:
-- De Lewis (firstname.lastname@example.org), June 24, 1998.
Since Rocky was so good as to point you to my newspaper, I'd like to upgrade his URL to our full coverage (such as it is):
Mark Zieg Technical Producer Orlando Sentinel Interactive
-- Mark Zieg (email@example.com), June 25, 1998.