Am 16.02.2010 21:36, schrieb Andrei Pelinescu-Onciul:
On Feb 16, 2010 at 18:30, Klaus
Darilion<klaus.mailinglists(a)pernau.at> wrote:
just as info: I have "dns_try_ipv6=no"
Try dns_try_ipv6=yes (or change it at runtime with
sercmd cfg.set_now_int core dns_try_ipv6 1
but as this destination is already an IP address this setting should not
matter.
I think we better fix the docs. If dns_try_ipv6 is off, no hostname
resolving that would result in an ipv6 address would succeed (it doesn't
matter if an actual dns lookup is to be performed or the host is already
an ip address).
OK.
I don't think that the case where you don't
want AAAA lookups, but you
want IPv6 addresses in hostnames has much practical use, except maybe
some extra optimization (at least to me it sounds strange that you would
support IPv6 addresses, but not dns names).
If it's about the order of AAAA and A lookups, this can be controlled
(setting dns_cache_flags to 4 will try AAAA first and then A,
by default it's the reverse).
So do you really want to optimize away the AAAA lookup made if an A
record is no found?
In my setup I want to enable IPv6 support for local users, but use only
IPv4 interconnect with other SIP domains.
The reason is, that when I call force_rtpproxy() I need to know if the
target is v4 or v6 because I have to set the proper rtpproxy flags to
enable v4/v6 RTP interworking. When the call gets routed to another
domain I do not yet know which target IP address (v4 or v6) will be
chosen when calling rtpproxy.
Of course I assume that RTP will use same IP version than SIP, but I
think this is usually the case and if callee again uses other IP version
than the callee's proxy should take care of transcoding.
regards
klaus