This post is about computing systems, not embedded controllers. Most Y2K pundits know that the Gartner Group has projected that a sizeable fraction of Y2K "failures", whatever that word means, will occur in 1999, perhaps 25% of the total burden. In contrast, Ed Yourdon mentioned an anonymous company, no doubt confined to a particular sector, whose internal analysis indicated they would encounter only about 3% of their Y2K problems in 1999. There is a graphic presentation at http://www.senate.gov/~y2k/hearings/030599/gartner.html which, I guess, Gartner Group made at a US Senate hearing in March. What surprises me is the second graph, whose title is "Failures- How Many and When." You have to look carefully to see that the years on the graph are separated by thin vertical lines, and that the numerical names of the years lie in the middle of each respective year. The graph shows that failures now in the first half of 1999 are on an upward slope, and in midyear (probably July 1), the slope will suddenly become steeper, with failures continuing to increase on the same slope through the end of this year. That's when the surprise comes. I say surprise, because it is counter to anything I've heard from geeks or nongeeks (I'm a nongeek) on any part of the Y2K opinion spectrum. Not only does the slope of the graph fail to increase in early January, it immediately decreases to zero (horizontal graph), and soon begins to have a slightly negative value. The number of failures in early January is essentially no different from late December. My wife is a programmer and cannot explain this portion of the graph. She thinks that even if application software doesn't experience a higher frequency of failures in January, problems with operating systems might cause a surge of failures just after the century rollover. If this graph is correct, then I suspect that almost everyone, even those who follow this forum avidly, has an incorrect concept (leaning too far towards pessimism), of what will happen in January relative to December. On the other hand, if it is incorrect, someone needs to tell the Gartner group so they can tell the US Senate. Can any of you IT people shed more light on this subject? Incidentally, the whole Gartner presentation is worth looking at. I suspect its biggest potential flaw is the same problem we've been talking about for a year- the fact that all of the data on status of governments and industries are based on self-reporting.

-- Bill Byars (billbyars@softwaresmith.com), April 28, 1999

Answers

Bill,

Too late at night for me to comment on the overall question about the Gartner graph, etc -- looks like other people are doing that.

One small correction, though: I noted in one of my essays that an "anonymous" company (which happens to be a large financial institution in NYC) did some statistics on the Y2K bugs that they had found during the course of their testing in 1998. What they discovered was that approx 3% of the bugs would have caused trouble on Jan 1, 1999 -- not the entire year of 1999

-- Ed Yourdon (ed@yourdon.com), April 28, 1999.

Link

-- regular (zzz@z.z), April 28, 1999.

Can't contribute much except "It doesn't make sense to me - logically or "intuitiively."

They must have a reason, but it's their job to justify their assumptions and their conclusions, not mine. Maybe the "label" on the graph is for some particular "kind" of system, rather than a toal of all systems?

-- Robert A. Cook, PE (Kennesaw, GA) (cook.r@csaatl.com), April 28, 1999.

Bill,

I think it's a mistake to interpret the slopes of the plotted lines as continuous predictions. All those lines really do is to connect the values for four predictions per year, and their slopes between those quarterly prediction values are basically meaningless for interpolation.

The type of graph and plotting chosen by Gartner are unsuited to representing data with significant discontinuities such as the widely-expected spike in failures right after the 1999->2000 rollover. (A bar graph would be more appropriate for this data.)

In particular, there is no data presented in the graph to support a comparison between an interpolated value for late December with an interpolated value for early January.

Divisions on the horizontal scale could represent either the beginning, the middle, or the entirety of each quarter year. The absence of one marker line for the fourth quarter of 1999 is probably inadvertent, because close examination of a magnified view shows that there is a visible change of slope in the "Computing System Failures" line at that time, indicating a distinct data point plotted there.

-- No Spam Please (No_Spam_Please@anon_ymous.com), April 28, 1999.

No Spam, I follow you- sort of. But I don't see how any shortcomings in their graphing method could take an expected increase in failures at the century rollover, and turn it into the most sharply negative downturn in slope of any point on the entire graph- precisely at the change of century- unless they really mean that's what they expect.

-- Bill Byars (billbyars@softwaresmith.com), April 28, 1999.

Bill.....

i did the same thing, at first glance (it does look like the spike occurs sometime in the middle of 1999)

but if you take a second look you'll see that the 25% spans year 1999.......the 55% spans year 2000.......and the 15% spans yeas 2001..................that puts the spike right were it belongs...at the beginning of year 2000

they could have easily made this chart more readable if they'd just "colored" the year spans, instead of drawing those heavier dotted vertical lines

-- andrea (mebsmebs@hotmail.com), April 28, 1999.

--the most obvious answer is the most obvious answer, regarding these doods, and folks like "boil up the furniture" dicky and others. Who pays 'em? Follow the money. People making big money say in public what the clients want. Who's gonna bite the hand, etc. It's always different in reality. Their predictions are bogus, my opinion-what have they been right on? Maybe some stuff, but not much....tell the king what he wants to hear, and you get to keep your jesters job, again my opinion. Of course, if it's all really slick loking and formal and professional looking and all, well, it's just got to be real, right? Who sees the "real" stuff in a grand scale, the actual real picture all over? Answer is no one, no "predictions" can be accurately made. There are gross patterns is all, and they appear to be pretty gross to me...LOL! "They" are running out of deadlines to miss....5% or 95% of no water or food or electricity still equals zip water or food or electricity. I got over it a long time ago, and following on an earlier thread, I can afford to be wrong, most that will happen is embarrassement--now the minimalists and dgi's, if they are wrong....m-m-m-m-m-m-m-m delicious no gas Lexus, pass me a slice! HAHAHAHAHA Have some cell phone-pie for dessert! Whoops--not only no cushy job, but any jobs available are not heated or airconditioned anymore, and no "power lunches", or flying off for a "conference". Gotta be scary when you really can't do anything but what ya do now...so it just can't happen! more HAHAHAHAHA

-- zog (zog@avana.net), April 29, 1999.

Bill,

The highest point on each of the three lines of the graph _is_ at the beginning of 2000. Isn't that exactly where you expect the maximum failure rate?

>the most sharply negative downturn in slope

I still think you're making a mistake in interpretation. The downturn in slope is not meaningful, except insofar as the curve _has_ to turn down after it reaches its highest point. After you've climbed to the top of a hill, the only way to go from there is down -- that doesn't mean the top of the hill isn't its highest point.

-- No Spam Please (No_Spam_Please@anon_ymous.com), April 29, 1999.

Several good points made. I was looking for an upward spike after 12- 31-99, and maybe I'm just wrong. Thanks everyone.

-- Bill Byars (billbyars@softwaresmith.com), April 29, 1999.

Bill,

I think your post and the Gartner reference are extremely important and should not be ignored. Even if this graph is only an estimate I believe that the important fact is that the y2k errors are occuring now and will continue throughout this year. It is a clear fact that computers are constantly being feed the 2000 date now and this will increase as the year progresses.

You asked if some IT folks could shed more light on this and so far we haven't heard any substantial replies. Somehow, we do need to hear more from the IT folks.

The absence of answers from the IT folks does not lesson the importance of the issue. It appears that IT part of y2k is not a singularity but is spread out over a six-month to one year period. The question now becomes, will we start to see a lot of failures in Nov/Dec of this year?

IT people, where are you?

-- Tomcat (tomcat@cat.com), April 29, 1999.