["Pete Holzmann" <pete@xxxxxxxx>]

I realize that Ethereal cannot normally track actual bad packets, 
since device drivers and OS' do not pass them on.

However, at least on Linux, the OS keeps driver-level counts of 
collisions etc., e.g. in /proc/net/dev 

Wouldn't it be a big win to monitor these bad-packet counts, and 
record "virtual bad packets" in the data stream when bad data is 
detected?

We wouldn't get to see the actual bad data, but that's far less 
valuable (since the data is bad anyway) than seeing the timing and 
existence of bad data.

With that, wouldn't it be MUCH easier to detect and debug network 
problems?

Has anyone ever attempted or accomplished this?

Thanks,
Pete