[asterisk-users] Re: NAT: RTP Path Optimization

Brad Templeton brad+aster at templetons.com
Tue Jan 30 18:33:11 MST 2007


On Tue, Jan 30, 2007 at 10:23:09PM +0100, Benny Amorsen wrote:
> >>>>> "PC" == Patrick Cervicek <patrick at cervicek.de> writes:
> 
> PC> But then all RTP Traffic of my internal phones will go over
> PC> Asterisk. I want RTP to go "Peer-to-Peer". ==> "Intern-2-Intern"
> PC> and "Extern-to-Extern" should go P2P and "Intern-2-Extern" should
> PC> go over Asterisk, see picture
> 
> I understand what you want. I am telling you that you cannot get what
> you want, and the best compromise you can achieve. Either your
> internal-to-internal calls go direct, or your external-to-external do.
> Pick one.

You can almost get it, if your NAT will hairpin, by having your
"internal" phones all present their external addresses.   Then
all phones will appear external, and all can talk to one another
(though there can still be port change problems on reinvite if you
don't do explicit ports)  -- but internal to internal (and all other internal
involved calls) will go through your NAT box, not your Asterisk box.

However, the NAT must hairpin audio, and a lot of them don't.

Here is Cullen's latest chart of who hairpins:

http://tools.ietf.org/wg/behave/draft-jennings-behave-test-results-03.txt

Only a few do it, and only a few make his "group A" category at all.

Of course, many of these boxes are not expensive so it may be worth
switching, though many of us love the WRT54G and its clones because
of the ability to use open source firmware.  However, the bastard
does not hairpin, I don't know if any of the new firmwares do.


More information about the asterisk-users mailing list