innfeed, bindaddress, force-ipv4, ipv6

["Miquel van Smoorenburg"]

In article <64D9EAF56C934931BCA4BD1107D7EB7D at Iulius>,
Julien ÉLIE  <julien at trigofacile.com> wrote:
>Hi Miquel,
>
>>>* Marco d'Itri added a force-ipv4 peer configuration option for innfeed that,
>>>if set, tells innfeed to never attempt an IPv6 connection to that host.
>>
>> FYI, the "innfeed bindaddress per peer" patch I sent also allows
>> you to do that.. if you set "bindaddress: any" on a peer then
>> innfeed will only use IPv4 connections for that peer, and if you
>> set "bindaddress6: any" then innfeed will only use IPv6 for that peer.
>
>Sure.  However, force-ipv4 was added in August 2006 to STABLE (but never yet
>officially released).  Is it wise to remove this parameter from STABLE
>and to put instead this new bindaddress per peer parameter?

Ah, OK.

>> If none of the IPv6 addresses work, do fall
>> back to IPv4. Next, make innfeed reckognize "bindaddress: none"
>> and "bindaddress6: none" config settings instead of yet another
>> config item like force-ipv4 / force-ipv6.
>
>It is true that it is better to use bindaddress instead of force-ipvX.
>It is used only at one place in innfeed/host.c:
>
>    hints.ai_family = host->params->forceIPv4 ? PF_INET : NETWORK_AF_HINT;
>
>But I do not know what innfeed's behaviour with "bindaddress: none"
>and "bindaddress6: none" at the same time would give.  It should
>perhaps be forbidden to do that.

I've updated my "innfeed bindaddress per peer" patch to do exactly
that. It now behaves as I described. And it doesn't let you
set bindaddress and bindaddress6 to 'none' simultaneously.

It compiles, but I haven't tested it yet - will do that tomorrow or so.

BTW, my old patch was wrong. With current inn you can set bindaddress6
and ipv4 will still work, my patch broke that.

Mike.
-- 
The From: and Reply-To: addresses are internal news2mail gateway addresses.
Reply to the list or to "Miquel van Smoorenburg" <miquels at cistron.nl>