[Live-devel] RTCP round-trip time estimation problem is now fixed

[Ross Finlayson]

>By the way, what is the status of RTT negative value issue?

FYI, I have fixed this in the latest software release (2004.04.27).

An explanation:
The round-trip time estimate is computed as S-R, where:
         S is the difference between the time that the most recent SR was 
sent (by the sender), and the time that the most recent RR was received (by 
the sender), and
         R is the difference between the time that the most recent SR was 
received (by the receiver), and the time that the most recent RR was sent 
(by the receiver).

The trouble is that S and R are computed using two different clocks: S is 
computed using the sender's clock; R is computed using the receiver's 
clock.  If (as is typical) there is some drift between the sender and 
receiver clock rates, and if the actual round-trip time is very small, then 
it's possible for S to be smaller than R.  This would return a 'negative' 
round-trip time.

The code now returns 0 in this case.

(Dixon, thanks to alerting us to this issue.)


	Ross Finlayson
	LIVE.COM
	<http://www.live.com/>