[Live-devel] how to define that client is died?

[Ashutosh Dutta]

Ross, thanks for your reply. I think my scenario is little different. We 
are assuming a scenario where the server stops sending traffic for 
certain time, either controlled by a third party, or by its own, and 
then it is asked to send the traffic. Thus for a while, client does not 
get any RTP traffic from the server and after a while it starts getting 
traffic. I think it could very similar to network congestion in some sense.

Thanks
Ashutosh


Ross Finlayson wrote:
>> I have a similar question in the case of server inactivity. What happens
>> to the client if it does not receive any RTP stream for sometime from
>> the server? Can RTCP info be used to signal the client to behave
>> accordingly?
> 
> Yes.  Note that RTSP server implementation (e.g, as used in the 
> "LIVE555 Media Server") sends send RTCP "BYE" packets when the stream 
> ends, but only for media types for which 'seeking' is not supported. 
> (For those media types, the server keeps the session open after the 
> stream ends, in case the client wants to replay it from an earlier 
> time.)
> 
> RTSP clients already interpret incoming RTCP "BYE" packets as 
> signalling the end of a stream, and handle this accordingly.
> 
> However, if you want to watch for a server just 'dying', before 
> signalling the end of a stream, then you could also watch for RTCP 
> "SR" packets coming from the server (using the 
> "RTCPInstance::setSRHandler()" function).  In most cases, though, 
> this shouldn't be necessary, because if a server dies, then the RTSP 
> TCP connection will also (eventually) get closed, and the client will 
> detect this.