|
Posted by Wolfgang S. Rupprecht on August 7, 2006, 12:59 pm
If you were Registered and logged in, you could reply and use other advanced thread options
> "Wolfgang S. Rupprecht"
>
>>> Does RTP packets flow between endpoints once the sip call is setup
>>> in an Ip
>>> Pbx. and hosted Ip Pbx.
>>
>> If the PBX and endpoints are setup correctly
>
> ...and the endpoints are not both behind NAT, which unfortunately
> happens very often...
True. Putting SIP phones behind NAT isn't set up correctly. They
really should have a real IP.
(My feeling is that NAT is evil and breaks way to many symmetrical
protocols.)
>> the RTP data stream is
>> redirected to go between the endpoints. You really don't want the
>> sip/rtp proxy to stay in the data path. It just chews up bandwidth on
>> the server and adds delay and jitter.
>
> Well, there are cases when have to. If DTMF signalling is passed over
> RTP, as per RFC2833, the removal of the PBX from the data path makes
> it unable to respond to requats such as "call transfer"... Same story
> if the PBX performs other useful functions like transcoding between
> endpoints that do not share a common codec. You may argue that this is
> not a case of endpoints correctly set up, but it's a reality
> especially in a peer-to-peer, providerless environment without a
> guaranteed interoperability baseline.
True. This is exactly why one wants to route the DTMF via SIP-Info
messages. One needs to keep the SIP proxy "in the loop".
I've been running with reinvites active for 2 years now. It can be
made to work well and does in fact help quite a bit with
jitter/drop-outs when running a soft-pbx on a general purpose
computer. Having reinvites working is very important when running
off-site extensions. It would be very wasteful to haul the RTP/ulaw
stream onto one's net and then send it right out the same connection.
RTP isn't that efficient and a 64kbit/sec ulaw stream ends up taking
96kbits/sec on the wire (if I'm remembering the numbers correctly).
That comes out to 192kbits/sec just for looping the voice stream
through the proxy.
As an aside, the first issue with each phone needing its own IP (or
suffering higher delays and jitter) is what I think will drive ipv6
adoption. ISP's, at leas in the US, are in general very stingy with
how many IP's they hand out to customers. With ipv6 it is hard for
them to not hand out a full set of 2^64 IP's to each customer.
-wolfgang
--
Wolfgang S. Rupprecht http://www.wsrcc.com/wolfgang/
|