[Live-devel] doubt trying to implement a rtsp client

[Ross Finlayson]

>I'm trying to implement a small receiver with live555. What I basically
>do is contacting a RTSP server, browse the subsessions, receive the
>packets of the desired subsession and pass them to a routine that
>process them. I tried following playCommon.cpp , but I'm stuck at a
>point and I don't understand how to finish the program.
>
>What I do is:
>
>* call RTSPClient::createNew to create a client
>* call RTSPClient::describeURL url get a description
>* create a MediaSsesion with MediaSession::createNew and iterate through
>its subsessions looking for the one that I want

At this point, you should call "initiate()" on the subsession object. 
This will create "RTPSource" and "RTCPInstance" objects for the 
stream.  (The "RTPSource" object is the one that actually receives 
data from the network.)

>* if I want to setup a session I call RTSPClient::setupMediaSubsession

Yes.

>but at this point I'm stuck. I think that before calling
>RTSPClient::playMediaSession I need to do something so that each time
>data is received my processing function is executed.

Yes, after calling "setupMediaSubsession()", and before calling 
"playMediaSubsession()" (note that those two functions only send RTSP 
commands; they don't actually do anything with the incoming data), 
you should do the following:

1/ Create a 'sink' object which will actually receive the incoming 
network data (from the "RTPSource" object).  This must be a subclass 
of "MediaSink", and could - for example - be a "FileSink", if you 
want to write the incoming data into a file.
2/ Call "startPlaying()" on your 'sink' object, asking it to receive 
data from the "subsession->readSource()".  See line 712 of 
"playCommon.cpp".  (Note that you should call 
"subsession->readSource()" rather than "subsession->rtpSource()", 
because for some types of stream, they are different.)

Then, after doing 1/ and 2/, you will call "playMediaSubsession()", 
to tell the server to start streaming data.

I'm sorry that the "openRTSP" code is so complicated (because the 
"openRTSP" application has so many 'bells and whistles').  At some 
point, I plan to also release a much simpler "testRTSPClient" 
application that will make it easier for developers to figure out how 
to write their own RTSP client application.  (This will also use the 
new asynchronous "RTSPClient" interface, which is what all developers 
should eventually be using.)
-- 

Ross Finlayson
Live Networks, Inc.
http://www.live555.com/