Y2K Bug Found (AREV Specific)
At 26 SEP 2006 06:20:27PM Victor Engel wrote:
I just found a Y2K bug in our software, after all these years. Part of the software involves running batches. A key to the batch transaction is in the format XXXX_mm/dd/yy_hh:mm:ssYM, where Y is A or P and XXXX identifies the type of batch job.
These summary keys are stored in index records styled XXXX.INDEX, oldest first.
There is a purge routine that loops through, concatenating 19 or 20 before the yy, I guess in an attempt to be Y2K compliant. Unfortunately, this effort makes it noncompliant. The code looks at the first y digit, and if it's a 9 concatenates 19, otherwise concatenates 20. The result is then used, along with an assumption that the rows are in chronological order, to identify records that can be deleted, based upon a 5 day retention period.
Another assumption, a bad one, is that dates are after 1989. We have some sites that have been around since the 1980's and have records that old extant in the logs. This causes the index record not to be purged, and it grows until a linebreak occurs.
I don't need help with this. The fix is obvious. I just thought it might be fun to see how late Y2K bugs can show up.
At 26 SEP 2006 06:42PM dbakke@srpcs.com's Don Bakke wrote:
Victor,
I think it is important to underscore that this wasn't an AREV Y2K bug but rather it was inherent to the way the application was designed. Interestingly enough, as you pointed out, some attempt was made to deal with the perceived Y2K issue, but it was obviously not well tested.
Future Y2K bugs will eventually appear as solutions based on pivot dates will begin to fail. Of course, most pivot dates are set so far in the future that no one expects the original code to still be in use. Guess they didn't cound on AREV32 being developed…
So…exactly how long is the half-life of plutonium anyway?
dbakke@srpcs.com
At 27 SEP 2006 04:20PM Steve Smith wrote:
Don, I never heard the term "pivot date" before. But I knew this girl once…
At 27 SEP 2006 07:32PM dbakke@srpcs.com's Don Bakke wrote:
Can't say I know the same girl, but she sounds very intersting…
Although most of the Revelation world was immune to Y2K issues (except for poor application design, as in this case) I was involved in a lot of Y2K auditing for some other organizations. Pivot dating was allowed for to be Y2K compliant…although technically it just pushes out the problem another 99 years (or less).
dbakke@srpcs.com
At 28 SEP 2006 05:13PM Steve Smith wrote:
Could you describe what a "pivot date" actually is?
Is it where you define the DDMMYY for any YY ] 25 as 1900's and any year lower than 25 as 2000's ???
At 28 SEP 2006 05:35PM dbakke@srpcs.com's Don Bakke wrote:
Steve,
Yep, that's exactly what a pivot date is.
dbakke@srpcs.com