[zeromq-dev] OpenPGM and IPv6 interfaces

Martin Lucina mato at kotelna.sk
Mon Feb 1 17:24:08 CET 2010


Hi Steven,

we've found some odd behaviour with OpenPGM and IPv6 interfaces.

When trying to bind a socket to a 0MQ endpoint such as
udp://eth0;224.0.0.1:5555 the call was failing intermittently.

Martin Sustrik tracked down the error from OpenPGM and it was
PGM_IF_ERROR_NOTUNIQ. It turns out this was happening because his box has
IPv6 enabled by default, so OpenPGM (correctly) says that "eth0" is not
unique. Not sure why it fails sometimes and sometimes it works though.

Anyway, we discussed further how IPv6 support should resolve names and I
think we're getting a consensus that 0MQ should have separate IPv4 and IPv6
transports, otherwise things get messy for the caller. So something like:

tcp://, tcp6://, udp://, udp6://, pgm://, pgm6://

Where the IPv4 transport would not accept or resolve IPv6 addresses and
vice versa.

What do you think, is this a good idea?

Further, is there a way to tell OpenPGM to please resolve only IPv4/v6,
which is something we'd need to implement this?

-mato



More information about the zeromq-dev mailing list