If SipX would follow the RFC then it would send out 7 INVITE's totalling 
to
0+0.5+1+2+4+8+16=31.5 seconds of INVITE's.
I don't think the client will wait that long, but the current SipX could 
be improved.
If the client would terminate the call before the 31.5 seconds expire then 
I assume the 
Bye will be forwarded by SipX.

Paul 

Ari Sonesh <[email protected]> wrote on 08-11-2011 14:21:05:

> 
> Tony, consider a scenario where most phones are connected to
> wireless modems, and servers are on a public IP.  There are
> no gateways or trunk providers, just extensions are calling
> extensions.  There are many possible applications, but think
> about application where care is provided to old people at
> their homes that do not have connection to the internet
> (most really don't), so the connection is via mobile data
> network.
> 
> The case of inherent (by design to save energy)latency in
> opening a connection on a mobile data network is well
> described in the following AT&T labs paper
> http://www.research.att.com/articles/featured_stories/2011_03/
> 201102_Energy_efficient?fbid=0lDQwAs5gdi
> (see the graphical depiction of the phases of data
> connection setup).
> 
> The conclusion is that it takes typically 2.4 seconds,
> before packets start flowing, and including RT for DNS and
> RT for first invite and ack  it's 2.6 sec.   I believe that
> these numbers are  "typical" sometimes less (and this is why
> sometimes calls get through), and sometimes more. If this is
> typical, i.e. a "mean" than probably doubling this number
> would be "safe" timeout, i.e. 5.2sec. 
> 
> Also, in any case, if and when the SIPX gives up after a
> timeout, it needs to send a disconnect message not leaving
> the called extension in limbo (currently on timeout the
> called phone doesn't stop ringing). 
> 
> _______________________________________________
> sipx-users mailing list
> [email protected]
> List Archive: http://list.sipfoundry.org/archive/sipx-users/
_______________________________________________
sipx-users mailing list
[email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-users/

Reply via email to