[VOIPSEC] Session Border Controller use

[Simon Horne]

At 03:43 AM 22/06/2006, Robert Welbourn wrote:
>You'd want to deploy an SBC in either scenario for a number of important 
>reasons:
>
>
>    NAT traversal.  Even if you mandate that your customers use approved 
> terminal adapters with some form of NAT support, you probably don't have 
> complete control over the broadband routers used by your customers, which 
> may negate the NAT support in the terminal devices.  (And I'm sure you 
> don't want your support folks telling customers how to turn on UPnP in 
> their routers, for example.)

Having written a open source Implementation for UPnP for VoIP, support for 
it in SIP it is functionally impractical. The principal reason is that RTP 
in SIP is symmetric so you must know the port forwards to open during the 
call setup. UPnP port opening can be notoriously slow and it may take 
several seconds to finish opening the ports so the wait for audio to flow 
at the start of a call can very annoying. However in other protocols the 
RTP does not have to be symmetric so the media ports can be opened in 
advance and simply allocated to the call thereby giving the user a seemless 
experience. Traversal of NAT is quite trivial actually, it just depends on 
who is the connection initiator, to work the initiator must be the person 
residing behind the NAT so as in the H460.18/19 the VoIP call flow path is 
altered to have the NAT device initiate the connections.

Forcing customers to purchase approved terminal adaptors greatly reduces 
the market potential. With the advent of wireless Internet, people expect 
to be able to send/receive calls no matter where they are which is why 
SKYPE is now so prevalent.

Simon