From - Thu Dec 31 02:06:15 1998 From: Arnold Trembley Newsgroups: comp.software.year-2000 Subject: Re: STL Time Machine Report #36 Date: 31 Dec 1998 06:25:14 GMT Bradley K. Sherman wrote: > > Arnold, if you have the time and the inclination, would you > comment on how scheduling of your Y2k project conformed (or > not) to reality? > > Thanks, > --bks Sure, I'll take a stab at it. First, the background. I volunteered to work on Y2k, and I went to the first meeting we had in April of 1996. We really fumbled around back then. We had no idea how big the project would be, so we started with paper surveys with simple fill-in-the-blank questions like, How many programs does your department support, how many lines of code, in what language, how are dates used in your applications (this survey was mostly oriented to the mainframe development areas). I filled out the survey for my area. I worked hard on it, ran Endevor reports to get exact line counts and program counts. Of course I had to guess at the date impacts because it was early in the game and I didn't have time to scan every line of code. Still, I had a good working knowledge of my primary application and gave it my best shot. About a year later I got to see some of the survey responses from other groups. One thoughtful person include nice round guesses for all the answers and include comments like, "I don't know, Use any numbers you like". At the start of the project I went to all the meetings and knew everyone on the project since there were fewer than 20 people originally. I don't go to those meetings anymore, and since then several hundred people have worked the project in one capacity or another. I was always a technical resource, not a project leader. And my knowledge is very limited on the non-mainframe portions of the project. Not a lot got done in 1996. I went to presentations by tool vendors. They told us programmer rates would go up and we wouldn't be able to find enough people. They told us they would be turning down Y2k work in 1998. They told us we would need to use all of 1999 for testing. In November, 1996, I worked on a Y2k pilot project with two other employees and three contractors. We performed a detailed inventory of all mainframe application software using the vendor tool. Then we repaired an application of about 50 programs and tested it (not in a time machine!). We tested it two ways, because we were also evaluating an SVC-level file bridging tool. It worked, but we ended up not buying it because it wouldn't support BDAM. Funnily enough, I think we have converted the BDAM files to VSAM anyway. We considered the problem of being backleveled on our O/S and COBOL compiler. I think the big advantage to doing a detailed inventory with a fancy tool was that it allowed the managers to estimate the mainframe effort with a reasonable level of accuracy. The pilot project ended in early February, 1997. We began upgrading my application shortly thereafter and installed in around September, 1997. Then we began trying to prepare for Time Machine testing. As you know, we ran into some problems with CICS, a third-party database, and upgrading the Operating System to MVS/ESA 5.2.2, the minimum Y2k compliant release. Most of the problems we found in internal compliance testing were in building the Time Machine and test setups, but we did find and fix application problems. In general, the application problems we found did not require much time to fix, at least not when in testing mode. -- Arnold Trembley http://home.att.net/~arnold.trembley/tmr.htm "Y2K? Because Centuries Happen!"