Hal Murray wrote:
> 
> Sending 1 out of N doesn't seem quite right.  I think you want to
> send at most N per second.  Am I missing something?
> 
> I'm thinking of something like this:  Don't send if a counter is
> too high.  Bump the counter up by 1 when you send an error packet.
> Bump it down by X % every Y milliseconds.  If you scale by 1000 or
> 1024 you can do it all with integer arithmetic.

IIRC, Solaris limits the amount of packets per second 
and not 1 out of N.

  Guenther