m********5
发帖数: 17667 | 1 Please describe the principle and assumptions of its algorithm.
How accurate it can be?
Does it assumes the network latency is a stable constant + noise?
Thanks in advance. |
m***x
发帖数: 20 | |
m********5
发帖数: 17667 | 3 多谢, 不过还是希望实际研究过这个东西的朋友来说说, 到底这个基本假设在现在的互
联网上有多大程度 valid.
【在 m***x 的大作中提到】
: http://en.wikipedia.org/wiki/Network_Time_Protocol
|
a****l
发帖数: 8211 | 4 It looks like the basic idea is to combine multiple measurements of time
difference to calculated a theoretical offset which might be very close to
the true time.
Given that nowadays internet is way faster than when NTP was designed, I
think its accuracy is getting better. As long as you are not using a Chinese
server from NA (which is probably taken care of by the protocol) the delay
seems to be generally very small, so it's claim of ms accuracy seems to be
reasonable.
【在 m********5 的大作中提到】
: 多谢, 不过还是希望实际研究过这个东西的朋友来说说, 到底这个基本假设在现在的互
: 联网上有多大程度 valid.
|
m********5
发帖数: 17667 | 5 非常感谢!
Chinese
delay
【在 a****l 的大作中提到】
: It looks like the basic idea is to combine multiple measurements of time
: difference to calculated a theoretical offset which might be very close to
: the true time.
: Given that nowadays internet is way faster than when NTP was designed, I
: think its accuracy is getting better. As long as you are not using a Chinese
: server from NA (which is probably taken care of by the protocol) the delay
: seems to be generally very small, so it's claim of ms accuracy seems to be
: reasonable.
|
