There is really no RFC violation here.  The value written in the RFC is 
a mere recommendation.  Whether to lower or make that higher is a matter 
of design.  As Dale pointed out,  the default RFC timeout of 64 * T1 (32 
seconds!) is not realistic.  Most people would abandon that call prior 
to sipx trying an alternate route.   One fast work around for your issue 
is to use an SBC that honors the RFC timeout.  You can point your 
registrations to the SBC.  There's a free one out there called Karoo 
Bridge.  As to why T1 is not configurable IMHO is decision made to not 
make users configure a stone to hit their heads with.


On 11/12/2011 01:20 AM, Ari Sonesh wrote:
> Content-Type: text/plain;
>    charset="utf-8"
> Content-Transfer-Encoding: 8bit
> Organization: SipXecs Forum
> In-Reply-To:<[email protected]>
> X-FUDforum: 08063afcdd00a6e76393c5b9527381e8<64454>
> Message-ID:<[email protected]>
>
>
>
> We have two issues here:
>
> 1. Invite is being sent only 4 times, while according to the
> RFC we need to send 7 for total timeout of 64 x T1.  We
> should really follow the RFP.
>
> 2. Indeed T1 should be set as an estimate of network
> latency, but an estimate of 100msec for a RT on WAN in
> general (which may include international locations) and
> mobile network in particular is too low, and should be at
> least doubled (providing that we fix item 1, otherwise we my
> need a higher number for T1, because it takes mobile
> networks several seconds to establish a data channel
> following issuance of the first invite).
>
> Ideally T1 should be a parameter, which would be adjusted to
> strike a balance between  or reliably reaching users, and
> need for acceptable timeout for redundant configurations.
> This balance  depends  on application, topology, and other
> specifics.
> _______________________________________________
> 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