I have been doing some tests with kamailio and IPv6.
My initial setup was IPv6 only and now am I extending it to a
dual-stack environment. Well, and now I am starting to face some
(interesting) challenges.
So, the first step in the dual-stack environment was to install
RTP Proxy and configured kamailio to use it. With this setup, the
UAs locally registered were able to communicate with each-other no
matter with address family (IPv4/IPv6) they were using. So far so
good.
But now I want to extend my tests a bit more ... I want to
communicate with the "outside world" (using ENUM and domain based
SIP URIs). Do I have a way to know if the "destination" is IPv4 or
IPv6? Because I need that information in order to properly bridge
the calls on rtpproxy.
I tried to use the onsend_route but it didn't work. I mean, I am
able to know if the "next-hop" is IPv6 or IPv4 but it seems that
it is "too late" to use rtpproxy. BTW ... I am assuming that if
the "next-hop" is IPv6 then the final user agent will also be IPv6
(the same for IPv4)
Should I use some other approach? For example, failure route
instead?
Does anyone have a similiar setup? How are you solving this issue?
I see few options here:
- use two outbound proxies, one for IPv4 and one for IPv6 -- from
your main proxy you can do parallel forking to the two edge proxies,
one of the branches will fail if no device/server is listening on
that protocol (or it will be dropped in onsend_route if after dns is
not the expected IP version)
- use lua or other embedded language to do dns lookup from config
and decide in branch route what to do. Alternative is to execute
external script or some web service
- export to the config the internal dns lookup function so it will
be done on demand and detect outgoing IP layer version (this
requires c coding, perhaps not much -- preferable solution, because
will remove it from my to-do)