UDP question

[Dar Scott]

On Nov 10, 2006, at 8:29 PM, Mark Wieder wrote:

> The server
> could send a broadcast packet which was received by the client. The
> client would then respond to the packet but the response was
> apparently never picked up by the server. The same thing works fine
> using TCP. There seems to be a bug with either a) responding to UDP
> broadcast packets or b) receiving responses to UDP broadcast packets.
> We had to code around this by using known IP addresses.

I would guess the problem is b.  The socket might be willing to  
receive something from, say, 255,255,255,255, but that will never  
happen because that cannot be a sender's IP address.  I guess Rev  
sockets are more picky (and moralizing) then TCP/IP requires, often  
more than Unix sockets.

Back when a team was looking at communication among Rev tasks, I made  
something that got around this by sending responses to all of a batch  
of ports on the requester's IP, a sort of poor man's Rendezvous.   
(The multiple ports allowed multiple tasks per computer.)

Several years ago I suggested some improvements that would allow  
better UDP use.  I don't know if I'd agree with those suggestions now  
or not.

Dar