|
Compared to the connectionless network layer service, the main advantage of the UDP service is that it allows several applications running on a host to exchange SDUs with several other applications running on remote hosts. Let us consider two hosts, e.g. a client and a server. The network layer service allows the client to send information to the server, but if an application running on the client wants to contact a particular application running on the server, then an additional addressing mechanism is required other than the IP address that identifies a host, in order to differentiate the application running on a host. This additional addressing is provided by `port numbers`. When a server using UDP is enabled on a host, this server registers a `port number`. This `port number` will be used by the clients to contact the server process via UDP.
|
|
As the port numbers are encoded as a 16 bits field, there can be up to only 65535 different server processes that are bound to a different UDP port at the same time on a given server. In practice, this limit is never reached. However, it is worth noticing that most implementations divide the range of allowed UDP port numbers into three different ranges :
|
|
a pseudo header :rfc:`2460` containing the source address, the destination address, the packet length encoded as a 32 bits number and a 32 bits bit field containing the three most significant bytes set to 0 and the low order byte set to 17
|
|
A discussion of the ephemeral port ranges used by different TCP/UDP implementations may be found in http://www.ncftp.com/ncftpd/doc/misc/ephemeral_ports.html
|
|
a 16 bits source port
|
|
a 16 bits length field
|
|
a 16 bits destination port
|
|
a 16 bits checksum
|
