Mainframe support pilot fish gets a call from his boss: A customer at a remote site says the mainframe is spewing out dump after dump of what looks like an opcode exception -- something that just isn't supposed to happen.

"So I barreled down the interstate to a customer 80 miles away that I had never met before," says fish. "I was confronted by the hardware guys telling me 'You've got a software problem.'

"Now, I always got along great with my hardware guys, but sometimes they had a bit of a limited field of vision -- if it wasn't smoking, on fire or otherwise physically busted, they didn't think it was their issue."

First fish looks at one of the memory dumps. Sure enough, it looks like opcode errors, meaning that some machine-language instructions have stopped working. How is that possible?

To read this article in full or to leave a comment, please click here