drop rates

Eric Hoffman (hoffman@cs.washington.edu)
Fri, 26 Jun 1998 09:32:32 -0700 (PDT)

this is excellent supporting data, even better than the delay.
(modulo the usual concerns about the measurement process).

its odd that there is smooth behaviour right up until 5% (clarks
out-of-the-air number for acceptable tcp loss)...and also that it
occurs about halfway through the AB pairs, just as in delay (same
pairs?)

can we work it in?