Ars Technica heeft uitleg over de wazige 49.7 dagen bug in Win9x:
A few people e-mailed me regarding the "why" of the 49.7 day Windows 95 bug; for those interested, here's the story. Windows 95/98 is of course a 32 bit operating system with integers as large as 32 bits. To keep track of the order of messaging events in the operating system, the OS uses one of these 32 bit integers that starts at zero when the operating system boots and increments by 1 every millisecond. So let's whip out a calculator, shall we?
The variable that holds this value is a 32 bit unsigned integer. Expressed in decimal terms, this has a range of 0-4294967295. Division time! 4294967295 milliseconds divided by 1000 equals a 4294967 second range of values. Divide that by 60 and you get a 71582 minute range. Divide that by 60 and you get an 1193 hour range. Divide that by 24 and you get.... a 49.71-day range. Coincidence? I think not.
This means the "how many are affected" question Microsoft neglected to answer is pretty much everybody who makes it to 49.7 days without crashing. (What--two, three people, tops?) The system queue is used all the time in Windows, and it doesn't take kindly to its counter being reset to zero right in the middle of things. That's pretty much a guarantee for a system crash or, at best, severe system instability. Thanks to Mark Pritchard for a lot of this info. He also informed me that NT is immune to this bug because it handles messages in a time-independent fashion.