>
>You are probably aware that different sets of data might result in the same checksum. (It seems to be a 16-bit checksum - see www.protocols.com ) On the other hand, I would expect single-bit errors to be the most likely, and I am quite sure that those are caught.

Yes I am aware of that, but to be honest I had ruled that option out because the odds of that are less than winning the lottery (well maybe not that low). That could be a reason though now that I think about it. On the other hand why is the distortion limited to one of two bytes (byte #3 or byte #4) and why is the value received always (as I have seen it) 0x10 less than what the PC sent? I find it hard to believe it to be random. (maybe it is the ghost in the machine like in the movie IRobot with Will Smith lol)

>
>But perhaps there are really many more distortions, and only the ones that don't change the checksum pass on to the other end.

That is true. I might have to pick the network admins brain about this one.

>
>Still, 2% would seem rather high under these circumstances - forget my previous remarks! - since they most likely hide a much higher incidence of errors that are don't reach the recipient (after TCP processing).

I agree it is much too high.

Thanks for the link I will review that information tomorrow after getting some much needed sleep.

Thanks for discussing this with me.

Einar