When is IPPROTO_UDP required?

When is IPPROTO_UDP required?

Is there ever a case where UDP is not the default protocol for SOCK_DGRAM? (real cases, not hypothetical "it might be", please")

i.e., what are the situations where the fol开发者_如何学运维lowing two lines would not produce identical behavior?

if ((s=socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP))==-1)
if ((s=socket(AF_INET, SOCK_DGRAM, 0))==-1)

Some operating systems (eg. Linux kernel after 2.6.20) support a second protocol for SOCK_DGRAM, called UDP-Lite. If supported by your system, it would be enabled by providing IPPROTO_UDPLITE as the third argument to the socket() call.

It is differentiated from normal UDP by allowing checksumming to be applied to only a portion of the datagram. (Normally, UDP checksumming is an all-or-nothing effort.) That way, the protocol can be more resistant to checksum failures due to fragmented transmission, in the event that some fragments outside the checksummed area may have been lost in transit. As long as the fragments covering the checksummed portion were successfully received, as much of the datagram as possible will still be delivered to the application.

For backwards compatibility with existing code, I suspect (but I cannot guarantee) that the call socket(AF_INET,SOCK_DGRAM,0) will continue to default to normal UDP, even in systems that additionally support UDP-Lite.

Given these declarations:

tcp_socket = socket(AF_INET, SOCK_STREAM, 0);
udp_socket = socket(AF_INET, SOCK_DGRAM, 0);
raw_socket = socket(AF_INET, SOCK_RAW, protocol);

the ip(7) manual page in linux says:

The only valid values for protocol are 0 and IPPROTO_TCP for TCP sockets, and 0 and IPPROTO_UDP for UDP sockets. For SOCK_RAW you may specify a valid IANA IP protocol defined in RFC 1700 assigned numbers.

Those two lines in your questions will always produce the same result.