Just a sample of the Echomail archive
[ << oldest | < older | list | newer > | newest >> ]
|  Message 8527  |
|  Michiel van der Vlist to Binkd team  |
|  Bindaddr problems in dual stack  |
|  07 Nov 23 17:20:40  |
 
TID: FMail-W32 2.2.0.0
TZUTC: 0100
CHRS: CP850 2
PID: GED+W32 1.1.5-b20170303
MSGID: 2:280/5555 654a63da
Hello Binkd team,
The bindaddr keyword is used to select the IP address for making outgoing
calls in case there is more than one IP address that the system can use. This
is usefull in case one has more binkd clients running on one and the same
system or when one wants to override the system default in order to present a
2001:DB8::f1d0:zone:net:node address instead of a SLAAC or random privacy
address.
The problem.
------------
bindaddr works fine in a single stack envionment (IPv4 or IPv6 only) but it
encounters problems in a dual stack environment.
bindaddr accepts a literal address (IPv4 or IPv6) or a hostname. Specifying
multiple bindaddr lines in the config does not result in an error, but only
the last takes effect.
specifying:
bindaddr fido.vlist.eu
or
bindaddr [2001:1c02:1103:2a00:f1d0:2:280:5555]
Results in binkd using that address for outgoing IPv6 calls.
In the windows version it results in a warning when making outgoing IPv4
calls: "bind: {W32 API error 10049} The requested address is not valid in its
context" or ."bind -- getaddrinfo: Der angegebene Host ist unbekannt. (11001)"
In the Linux version, the literal address must be specified without the square
brackets and attempts to make outgoing IPv4 calls fail. So one can not use
bindaddr to specify an outgoing IPv6 address without breaking outgoing IPv4
capability.
Proposed solution
-----------------
To remedie the situation I suggest to split up the bindaddr in an IPv6 and an
IPv4 part.
bindaddr6
|
[ << oldest | < older | list | newer > | newest >> ]