[Freeswitch-users] NAT traversal - the final say..!

[Cal Leeming [Simplicity Media Ltd]]

*Any and all feedback on this thread would be much welcomed.*

Hello,
*
*
There seems to be a large number of discussions surrounding NAT traversal,
as well as lots of documentation, but with no concrete answers.

The NAT related wiki documentation is tedious, and depending on the outcome
of this thread, I'd like to spend some time cleaning it up.

The most common problem (the same as ours) was having a router with broken
ALG and a softphone that does not seem to work with STUN.

The following REGISTER is sent from a phone.

REGISTER sip:1.2.3.4:5060 SIP/2.0
Via: SIP/2.0/UDP 192.168.1.102:57787
;branch=z9hG4bK-d8754z-b31b18401713de75-1---d8754z-;rport
Max-Forwards: 70
Contact: <sip:2000 at 192.168.1.102:57787;rinstance=0c7190b115a36513>
To: "foxx"<sip:2000 at 1.2.3.4:5060>
From: "foxx"<sip:2000 at 1.2.3.4:5060>;tag=83311448
Call-ID: NGQyMjJkODlhMzQ1ZWY4ZDk4ZjZmZWRhODU0NWE5YWI.
CSeq: 7 REGISTER
Expires: 120
Allow: INVITE, ACK, CANCEL, OPTIONS, BYE, REGISTER, SUBSCRIBE, NOTIFY,
REFER, INFO, MESSAGE
Supported: replaces
User-Agent: 3CXPhone 6.0.25732.0
Content-Length: 0

As you can see, the client's public IP is not specified
anywhere. FreeSWITCH offers several ways around this, the main ones being;

* NDLB-connectile-dysfunction
* NDLB-force-rport
* apply-nat-acl
* sip-force-contact

The one that has worked in our case was "NDLB-connectile-dysfunction"
(otherwise known as NAT HACK), however there seems to be a lot of negative
comments about using this.

>From what I can tell, the general argument is that NAT HACK is considered a
non RFC compliant hack, and the SIP phones should be doing a better job of
keeping to the RFCs.

In principle, this is a fair argument - but in practise, it's not a
reasonable assumption that all phones are RFC compliant, and (imho) not a
reasonable argument to have this functionality disabled by default.

So, I'd like to present the following arguments;

* Are there any other negative aspects about
using NDLB-connectile-dysfunction, other than it is a non compliant RFC
hack?

* Why is NDLB-connectile-dysfunction not enabled by default when certain
conditions are met? In the event that FreeSWITCH receives a REGISTER from a
phone specifying a Contact/Via as 192.168.0.0/16, but received on a public
IP, then it should be obvious that NAT is broken and automatically try to
circumvent it.

* People seem to get confused between server side and client side NAT
problems, and that they both need to be resolved in a different way. The
documentation doesn't seem to reflect this clearly.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freeswitch.org/pipermail/freeswitch-users/attachments/20121216/f4a8a968/attachment.html