This letter came from a mainframe programmer. I shall respect his desire for anonymity, although there is nothing mysterious about his information. His message: forget about replacing bad chips that operate systems unless they are relatively new chips.
I keep thinking about oil pipelines, natural gas pipelines, and electrical power systems. Are they all run by new, compliant chips? Nobody who is in a position to know is public about this information -- not on the Web, anyway. I won't bet my life on old chips. But hundreds of millions of people are unknowingly betting their lives on old chips.
Not everyone can get out at the top of a market. A few can. Maybe.
* * * * * * * * * * *
As someone who has programmed chips, and replaced chips in embedded devices, I can tell you from actual experience, that reprogramming and replacing chips older than 3 years will be almost an impossibility.
Here's the problem . . .
First, the chip design business is highly competitive, and chips have a techical lifespan of just few years. When these chips are superseded by better, faster chips, the old ones are no longer manufactured. In the case of EPROMs (these are programmable chips), if you can't find a pin compatible chip for the chip that needs to be replaced, you most likely have to discard the entire device. Older chips have little market value, and are almost impossible to come by. The only ones I know of are the ones pulled from junk boards (and these aren't too reliable.)
Second, even if you can find the chip, you still have to write the new code for the chip (and that code has to be compatible with all the other code in all the other chips in the same device), and once that code has been written, it must be compiled before it can be programmed into the chip. Compiling the code means you must find a compiler for the chip, the compiler must be compatible with the version of the chip you are using, and the compiler must not have it's own inherent y2k flaws.
Unforunately, finding these older chip compilers is getting difficult, especially in the case of chip specific compilers (which most are). If the chip is rare, the compiler is ever rarer - and you usually have to find a version that is compatible with the chip release.
If the compiler is very old, it is likely that it has inherent y2k flaws in the machine language (i.e. the binary date field will most likely by mm/dd/yy instead of mm/dd/yyyy.)
And, finally, even if you can find the chips and locate the correct compilers, you've still got to find someone who can still program in the language of the compiler. In the computer business, finding someone willing to code chips in an ancient chip-specific language is going to be exceedingly difficult. Most people who used to do this have gone on to bigger and better things. Few would be interested in going back to the highly technical coding requirements of chip programming. (I know I wouldn't.)
Most chips are not socketed onto circuit boards. This means the chips must be desoldered, and new one soldered back in (without damaging the board in the process.) The manufacturers did not intend most of these chips to be removed, and in some cases included devices that would disable the entire board if specific eprom chips were removed. (They didn't want competitors removing the chips, and downloading the programs so they could create their own products.)
Anyone who suggests the problem is a matter of just removing the old chips and putting new ones in, obviously has never tried to do it himself. It would be easier to remove spots from a cow and replace them with new ones. (Of course it would probably kill the cow in the process.)