[e2e] 150ms - tolerable latency for quakeIII

[Christian Huitema]

Grenville,

I really wonder whether the delays that you report characterize the
player's tolerance, or are merely a sampling of the delays observed by
well connected actors on the Internet. The delays you are mentioning are
not terribly different from the delays reported by Telcordia's Netsizer:
http://www.netsizer.com/daily/quality_today.html. Telcordia observed
median roundtrips of 60 ms between their probe and well connected sites,
90 ms for connection to random sites; the 95% RTT would be about 200 ms
for well connected site, 400 ms for random sites. This is not very
different from the typical one way latency of 30 ms for most players, or
even the "break point" at 150 ms. It would be consistent with players
investing on high speed connections, and thus getting reasonable
service; there is actually no need to hypothesize that they are voting
with their feet. 

If you remove the "voting" hypothesis, then the hypothesis that 150 ms
is some kind of magic number for games also goes. In fact, we could draw
another conclusion: that the tolerable delay is a function of the design
of the game, and that the designers of the game take into account the
prevalent conditions on the Internet to make sure that players get a
reasonable experience. In short, the designers program for the network
that most of their customers can get.

-- Christian Huitema