>>Also, Dmitry's theory is that the memory is corrupt. Yeah, maybe that's possible, but in my lifetime I've never encountered that too many times if ever.
>
>Back in the Ethernet days, we apparently had gotten a batch of bad NIC. How do I know this? Because suddenly an application that hasn't had a change done to it in over a year is getting 'Unable to access' errors and "line noise" being saved to the dbf's. The only person who wouldn't be getting these errors was on the machine with the NIC going south. They would have no problems. Have them log out of the app, and *poofda!* no one else had a problem. The first NIC that went bad took some time (and massive arguments with the hardware people) to find. The 2nd, much easier. The 3rd? Just had to tell CAST (the hardware peoples) which machine.

Yes, I remember back in the day of Token Ring a NIC was bad on a machine and flooding the network and the trick was finding the machine in a multi-story office building.