Re: problems in TCP RTO calculation

[Erik Fair <fair%netbsd.org@localhost>]

It strikes me that we should just use a 32 bit integer and represent 
microseconds; that would give us a maximum measurement of an hour and 11.5 
minutes, which would allow for TCP connections to devices beyond the moon's 
orbit, and should be fine enough for 10 gigabit Ethernet latencies, no?

Or would nanosecond units be better?

        Erik <fair%netbsd.org@localhost>