Hi Daniel, Hi Klaus,

On Thu, Jan 22, 2015 at 10:48 PM, Daniel-Constantin Mierla <miconda@gmail.com> wrote:

1) I guess the initial author know there is no nat in ipv6, so he didn't
bother with. I just pushed a patch for it in master (814c08f3), if
tested and reported to work ok, it can be backported

Thanks for the patch, I just patched our 4.1.3 with it, and now the Contact IP is surrounded by square brackets. So I guess it can be backported.

2) the contact uri example in the first email is perhaps not properly
reflecting the contact uri that was generated, because it should have
been with the ip address and the port. It seems to be only the ip
address. If there was an omission, that's ok, because I expect the uri
parsing error is due to hostpart having the port following the ipv6
address -- that requires the ipv6 between []. If the port was missing,
that can be another issue, but the code shows the port is always added
and it wouldn't worked at all so far without it.

Sorry, that was my mistake. The request came in from an odd port, so I removed it while adjusting the line for the mailing list. Of course, after fix_nated_contact the port is appended.

3) use set_contact_alias() if use modules that need the new contact
(like dialog, presence, ...) for later usage. The *contact_alias()
function don't change the host/port part, they just add a new parameter,
so it would have been safe with or without []. Anyhow, the code adds []
if the address is ipv6

I have had a look into these functions, and they seem a lot more appropriate (and do work with IPv6). Guess they haven't been there, when we originally built that part of our configuration back in 2007. I think we will use those functions now.

Thanks for the quick help.

Best Regards,
Sebastian