[zeromq-dev] IPC on Windows (again)

[Martin Lucina]

marcelo.cantos at gmail.com said:
>  2. MartinL: The zpmd solution doesn't require a service. Erlang nodes start
>     epmd on-the-fly when they detect that it isn't running, and it runs forever
>     from that point on. This does, of course, require a zpmd.exe to be sitting
>     around in easy reach, and I don't deny that this is a hack, but it's a hack
>     that has proven itself to be effective and reliable.

As Ivan points out, it's not quite so easy to keep a daemon hanging around
"forever" on Windows.

>  3. MartinL: Your proposal for using named pipes to negotiate the port
>     certainly sounds more appealing than my proposal, but I am puzzled about
>     point 5. The non-existence of an endpoint doesn't imply a connection
>     failure in ZeroMQ. Does your approach preserve the ability of connecting
>     peers to wait in a quiescent mode until the other peer binds?

Yes, that would be handled transparently to the user by the existing core
auto-reconnect mechanism. In fact, this is what happens internally with
e.g. TCP transported endpoints if the listening end doesn't exist at the
time of the connect().

Cheers,

Martin