This was posted on Peter de Jager's discussion forum. The man who posted it is part of the Taskforce 2000 project, a private nonprofit organization set up at the request of the British government to foster awareness of y2k.
* * * * * * * *
Date: Fri, 15 Aug 1997 16:54:58 +0100 To: firstname.lastname@example.org From: Ian Hugo
Subject: Sighting: IBM legacy
At the meeting with UK IT Minister Barbara Roche which I referenced in another posting, some interesting stats on IBM software emerged which sheds light on the real legacy (not mainframe per se) problem. It didn't come from IBM so IBMers reading this should feel free to comment.
The stats are as follows. Since IBM started selling software, it has produced around 250,000 (numbered) program products. Of these, some 19,000 are currently marketed and supported (across all platforms); however, around 90,000 are still in use. That's ~71,000 program products (times the number of users) potentially to be replaced. Even allowing for some that won't have date dependencies, the implications are interesting and won't be cheap.
There's a corollary I'd like to add. There's no reason I know of to suppose that the user base of any other supplier of significant longevity (say 20 years?) is in any better shape.