[e2e] 150ms - tolerable latency for quakeIII

Manuel Oliveira m.oliveira at cs.ucl.ac.uk
Fri May 25 04:35:44 PDT 2001


> if you look at the game source code you would delete the idea that
> game programmers "design" the game for any kind of network.
I can say from personal experience that the game industry does not
understand the network any better than the Virtual Environments community.
All their networking experience revolves around sockets and their perception
of a black box where you send/receive messages. Some of the more network
aware people have the notion of delay, but what causes it is the black box.

I am not even sure if the ping measurement that the server provides is
correct or matches the network semantics of the word. In some of Tristan's
measurements (Half-Life game server), he has players playing more than 10
minutes with "game pings" way beyond the 150ms. We suspect that this might
just mean that the game server ping mechanism might be incorrect or just a
hack that expects a network in a particular state.

I would be careful before setting in stone the measure of 150ms. For
starters, I would be curious to understand how the sample you have is
representative of the remainder population (note that the same player coming
back again and again counts as one unique person). Also missing are a wide
range of social factors regarding the users that need to be considered.
Unlike previous networking multimedia studies, the objective is not task
performance but enjoyment and sometimes users accomodate for less than ideal
network conditions.



More information about the end2end-interest mailing list