What I'd wanted was for this

http://snapon.lab.bufferbloat.net/~d/codel_sim.py

to show

* slow start until we hit the first drop (exponential increase in delay)
* go additive (congestion avoidance)
* have the first drop start to take effect, reducing the delay somewhat
* second drop take more effect
* third drop almost succeeding
* fourth drop dropping it below target for a while, drops stopping
* going above target again, starting a drop at 1/sqrt(3)


At least it gets the invsqrt right...

(toke tossed this off before leaving for vacation, and I only got to
hacking on it at 2am this morning. Wanted to use it to illustrate codel...)


-- 
Dave Täht

Fixing bufferbloat with cerowrt:
http://www.teklibre.com/cerowrt/subscribe.html