I've only commented the parts that are different from the directed_client.
// page04.html,v 1.11 1999/09/22 03:13:43 jcej Exp #include "ace/SOCK_Dgram_Bcast.h" #include "ace/INET_Addr.h" static const u_short PORT = ACE_DEFAULT_SERVER_PORT; int main (int argc,char *argv[]) { ACE_INET_Addr local ((u_short) 0); /* Instead of creating the ACE_SOCK_Dgram we created last time, we'll create an ACE_SOCK_Dgram_Bcast. "Bcast" means, of course, "Broadcast". This ACE object is clever enough to go out to the OS and find all of the network interfaces. When you send() on a Dgram_Bcast, it will send the datagram out on all of those interfaces. This is quiet handy if you do it on a multi-homed host that plays router... */ ACE_SOCK_Dgram_Bcast dgram; if (dgram.open (local) == -1) ACE_ERROR_RETURN ((LM_ERROR, "%p\n", "datagram open"), -1); char buf[BUFSIZ]; sprintf (buf, "Hello World!"); /* The only other difference between us and the directed client is that we don't specify a host to receive the datagram. Instead, we use the magic value "INADDR_BROADCAST". All hosts are obliged to respond to datagrams directed to this address the same as they would to datagrams sent to their hostname. Remember, the Dgram_Bcast will send a datagram to all interfaces on the host. That's true even if the address is for a specific host (and the host address makes sense for the interface). The real power is in using an INADDR_BROADCAST addressed datagram against all interfaces. */ ACE_INET_Addr remote (PORT, INADDR_BROADCAST); ACE_DEBUG ((LM_DEBUG, "(%P|%t) Sending (%s) to the server.\n", buf)); if (dgram.send (buf, ACE_OS::strlen (buf) + 1, remote) == -1) ACE_ERROR_RETURN ((LM_ERROR, "%p\n", "send"), -1); if (dgram.recv (buf, sizeof (buf), remote) == -1) ACE_ERROR_RETURN ((LM_ERROR, "%p\n", "recv"), -1); ACE_DEBUG ((LM_DEBUG, "(%P|%t) The server said: %s\n", buf)); /* Using the "remote" object instance, find out where the server lives. We could then save this address and use directed datagrams to chat with the server for a while. */ ACE_DEBUG ((LM_DEBUG, "(%P|%t) The server can be found at: (%s:%d)\n", remote.get_host_name(), PORT)); return 0; }
About that subnet thing:
If you run this client on a host that has multiple network interfaces, the broadcast will go to all of those (sub)networks. What do you do, though, if you need to get past a router? My advice is to write a server that will run on hosts on both sides of your router. When a server on one side of the router receives a broadcast, it would send a directed datagram to it's counterpart on the other side of the router. The counterpart would then re-broadcast the original datagram on that sub-net. Cheap, simple and effective.One final word of warning:
When creating your broadcast datagrams you may see something like this: ACE_SOCK_Dgram_Bcast::mk_broadcast: Broadcast is not enable for this interface.: Unknown error. There are some interfaces (ppp, slip) that don't support broadcast datagrams. That's what you're seeing here.Ok, one more warning:
If you happen to have multiple servers running on your network when you invoke this client, the response could come from any one of them.