BIND 10 trac1948, updated. 55a817bff19c1b61e2ce32d22a743d4a3073036c [1948] Suggest listing all addresses in listen_on

[BIND 10 source code commits]

The branch, trac1948 has been updated
       via  55a817bff19c1b61e2ce32d22a743d4a3073036c (commit)
      from  2801bd9baadb96c2965b2cddc729e9ac2ec6dbb0 (commit)

Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.

- Log -----------------------------------------------------------------
commit 55a817bff19c1b61e2ce32d22a743d4a3073036c
Author: Michal 'vorner' Vaner <michal.vaner at nic.cz>
Date:   Thu May 24 12:55:24 2012 +0200

    [1948] Suggest listing all addresses in listen_on
    
    The guide is modified to suggest to list all the addresses explicitly
    for a multi-homed server, since the wildcard addresses don't work well
    yet.

-----------------------------------------------------------------------

Summary of changes:
 doc/guide/bind10-guide.xml |   18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

-----------------------------------------------------------------------
diff --git a/doc/guide/bind10-guide.xml b/doc/guide/bind10-guide.xml
index e2b3961..c868a55 100644
--- a/doc/guide/bind10-guide.xml
+++ b/doc/guide/bind10-guide.xml
@@ -1352,6 +1352,24 @@ This may be a temporary setting until then.
       and <varname>port</varname> number.
       By default, <command>b10-auth</command> listens on port 53
       on the IPv6 (::) and IPv4 (0.0.0.0) wildcard addresses.
+      <note>
+        <simpara>
+          The default configuration is not appropriate for a multi-homed host.
+          In case you have multiple public IP addresses, it is possible the
+          query UDP packet comes through one interface and the answer goes out
+          through another. This will probably be dropped by the client, as it
+          has a different source address than the one it sent the query to. The
+          client would fall back on TCP after several attempts, which works
+          well in this situation, but it is clearly not ideal.
+        </simpara>
+        <simpara>
+          There are plans to solve the problem in a way the server will handle
+          it itself. But until it is actually implemented, it is recommended to
+          alter the configuration — remove the wildcard addresses and list all
+          addresses explicitly. Then the server will answer on the same
+          interface as the request came, preserving the correct address.
+        </simpara>
+      </note>
               </simpara>
             </listitem>
           </varlistentry>