Actually, short polling tends to be more scalable, because long polling holds a socket open longer on the server, and fds are a more scarce resource than bandwidth.
Not that it's likely to matter that much on a LAN scale.
Not that it's likely to matter that much on a LAN scale.
As sproingie said, considering that you want to implement this in a LAN environment it doesn't matter much which option you choose.
What you might think about though is that with the first method (polling every x seconds) you have a chance of the client having the wrong number for quite some time (bit less than x seconds in the worst case). Using the second method you have the new number on the client always only short time after the server has the new number (since it's the server that can then complete request whenever he wants - e.g. when he has a new number to send)





*facepalm*. Sorry for that. Force of habit.




