Thanks for your Henning, at least I will sleep better knowing I am not completly idiot :-))<br><br><div>I haven&#39;t noticed related to INVITE timers (such as the 408) because I only make my client behind NAT to register and re-register and verify it is always reachable in time thanks to this TCP keep-alive. Tomorrow I will try to make a couple a calls to see if I get this 408 you mentionned.</div>

<div><br></div><div>Is there a way (either with a parameter or changing source code) a to control the frequency of TCP k.a. ?</div><div><br></div><div>Cheers</div><div>Pascal</div><div><br></div><div><br></div><div><br><div class="gmail_quote">

2010/5/7 Henning Westerholt <span dir="ltr">&lt;<a href="mailto:henning.westerholt@1und1.de">henning.westerholt@1und1.de</a>&gt;</span><br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">

<div class="im">On Friday 07 May 2010, Pascal Maugeri wrote:<br>
&gt; tcp_keepalive=yes<br>
&gt; tcp_keepidle=10<br>
&gt; tcp_connection_lifetime=3600<br>
&gt;<br>
&gt;<br>
&gt; The TCP connection is kept alive and TCP Keep-Alive packets are sent when<br>
&gt; the TCP connection is idle.<br>
&gt;<br>
&gt; But I don&#39;t understand why the time between KA is not linear:<br>
</div>&gt; [..]<br>
<div class="im">&gt; Could you explain me where do come from these values of 75 seconds, 23, 54<br>
&gt;  ?<br>
&gt;<br>
&gt; Everything works well: my client behind NAT is kept connected but I&#39;d like<br>
&gt; to understand why it does work :-)<br>
<br>
</div>Hi Pascal,<br>
<br>
sounds indeed interesting. :-) Perhaps a problem with the internal timers.. Do<br>
noticed also other timer related problems, like for internal INVITE 408<br>
replies or similar?<br>
<br>
Cheers,<br>
<font color="#888888"><br>
Henning<br>
</font></blockquote></div><br></div>