[dbcurtis]

> but probably just happens by simple association "blue screen of death = Windows problem"

This certainly happens. Before driver signing, an extremely common cause of BSODs was a page fault in the kernel caused by a driver bug that failed to lock down a page during I/O. Only if you had the hex codes of the various exceptions memorized would you be in a position to tell a driver-caused BSOD from some other cause. So.... "it must be Windows again". This was a powerful motivation for MSFT to start a driver validation lab that they forced vendors through.

And then... you have OS/2 -- where they actually used more than two security rings. Kernel in ring 0, user space in ring 3, and drivers in ring 1. Now the kernel can properly blame the driver. But of course, that can't be ported to CPU's with only 2 security levels.