By 2000, there will be about 25 billion embedded chips.
No one knows what percentage of these are date-sensitive.
The systems the chips are in must all be checked for noncompliant chips.
No one knows what percentage of date-sensitive chips will be noncompliant.
When discovered, noncompliant chips must be replaced by hand.
This assumes that there is a compliant new model chip.
Chips older than three years are almost impossible to replace.
Then the testing must begin. The system the defective chip is in must be tested up to 30 times.
* * * * * * * * *
Date: Thu, 11 Dec 1997 11:07:05 -0600
From: "David C. Hall"
Subject: Re: Embedded Systems: Diatribe and Rant: example
Not to cause any more consternation, but this is a fine example of why we need to cycle test embedded systems, not just depend upon one test. A couple of people have posted to the SIM Web site (www.year2000.unt.edu, topic 11 page) that they have found a need to do approximately 30 or more tests on EACH system to ensure all logic paths have been checked out. This should not come as a big surprise.
Embedded systems are the equivalent of mainframe systems. How any regression tests do you have to do on a mainframe program to ensure all logic paths have been checked out? Same thing with embedded systems.