| The User Datagram Protocol (UDP) is described in RFC 768 [Post80] and represents a minimal transport protocol. It runs on top of the Internet Protocol (IP) and essentially offers the same functionality as IP itself: an unreliable, connectionless datagram service. In this case, unreliable means that there are no mechanisms to detect and handle lost or duplicate packets. However, packets can be optionally protected against bit errors by using a checksum, which covers both the packet header and the payload of each packet, in contrast to IP. Otherwise, there is only one additional option, compared to IP: Port numbers can be used to address different applications in a specific end system. On the one hand, UDP is used for transaction-oriented applications, such as the Domain Name System (DNS), where only one request and the relevant reply have to be transmitted, so that it is not worthwhile establishing a connection context, which would mean, for example in TCP, that three additional messages for the establishment and four messages for the tear-down were required. On the other hand, UDP is also used where the reliability of a transmission plays a secondary role, because one is primarily interested in transmitting data easily and quickly. For example, it is normally not a problem if some packets are lost when audio streams are transmitted in small packets. On the contrary, an automatic flow and error control with retransmission of lost packets would often be disturbing to the smooth playback of the stream. 25.1.1 Packet Format Figure 25-1 shows the format of UDP packets. The header fields are briefly described below. Figure 25-1. UPD packet format. 
Source port: The source port is the port number used by the sending process, in the range from 1 to 65535; normally, the receiver of a request sent over UDP will direct its reply to this port. RFC 768 specifies that giving the source port number is optional, and the field can have the value zero, if it is not used. However, for UDP over the socket programming interface in Linux (see Chapters 26 and 27), this is not possible, because a port number different from zero is automatically assigned to each socket, if the user does not state one. Destination port: The destination port is used to address the application in the destination system that is to receive a UDP packet. Length: The length is specified in octets and refers to the entire UDP packet, consisting of packet header and payload. The smallest possible length is therefore eight octets, and the largest possible UDP packet can transport 65535 ?8 = 65527 payload octets. Checksum: As in TCP, the calculation of the checksum includes a pseudo header, in addition to the packet header and the payload. The format of this pseudo header is shown in Figure 25-2. It includes the IP source and destination addresses, the UDP protocol identifier (17), and the length of the UDP packet. The checksum is computed as a 16-bit ones complement of the ones-complement sum over the data mentioned above, where a zero octet is appended if the octet number is uneven. This method can be implemented efficiently for all processor types (as described in RFCs 1071, 1141, and 1624 [BrBP88, MaKu90, Rijs94]). If the computation results in the checksum zero, the all-1-bit value is transmitted instead, which is equivalent in ones-complement arithmetic. A zero in the checksum field means that the sender has not computed a checksum. Figure 25-2. Pseudo header format for checksum calculation. 
|
