Unix Network Programming Manual
Addresses
In the example program and in the description of functions, you have seen
references to a C strucure named sockaddr. This structure is defined
in socket.h as:
struct sockaddr
{
unsigned char sa_len;
sa_family_t sa_family;
char sa_data [14];
}
So this is a 16 byte structure (sa_family_t is an unsigned char) and it
represents to the operating system what a host address looks like. It
is general in the sense that the family designation allows for lots of
different address types, as long they fit in the 14 byte sa_data field.
After you choose a particular family, the sa_data field can be interpreted
a number of ways. For example, Internet addresses are are 4 byte integer
and the Xerox Network Software network uses a two-byte address.
To make things easier, the more common address types are given their own
structure type. For example, the Internet address type is defined in
/usr/include/netinet/in.h as:
struct sockaddr_in
{
unsigned char sin_len;
sa_family_t sin_family;
in_port_t sin_port;
struct in_addr sin_addr;
unsigned char sin_zero[8];
}
where in_port_t is an unsigned short (two-byte integer) and
in_addr is an unsigned int (4 byte integer). Note that the last
8 bytes are set to zero and unused. When using this structure, the family
should always be AF_INET. Since everything about the address is known from
the family, it is not necessary to set the length.
There are two parts to this address, a port and a host address.
The address part is defined by the Internet structure to be a 32-bit
integer, which is normally displayed as four separate bytes. For example,
153.90.192.3 indicates that the address is 00110101010110101100000000000011.
The most significant 8 bits are the integer 153, the second byte is 90, the
third is 192 and the least significant byte is 3. The dotted-decimal
notation is widely used and recognized for Internet addresses.
The Internet address represents the "name" of a host on the network.
The port is also defined by the Internet to be the "name" of a process on
a particular host. A host, such as 153.90.192.3, might have hundreds of
processes running that want to communicate over the network. In order for the
operating system to distinguish between them, each process has to
specify a port that they want to use.
When a process sends a message to a process on another machine, it sends
its Internet host address and its port. When a response arrives, the
operating system can tell by the port what process should receive it.
You might wonder how a process could ever send a message to another process
without receiving a message telling it what port the process is using.
For such things, well-known ports are used. These are port numbers
that are assigned to specific types of servers. If you look in the file
"/etc/services" on your machine, you will see records of the following type:
ftp 21/tcp
telnet 23/tcp
who 513/udp whod
pop2 109/tcp
These are assignments of specific processes, such as the ftp server (ftpd),
to specific ports (21). If everyone agrees that such an assignment will be
made on every host that runs ftp, a client will always be able to connect
to it. There are other servers that are normally assigned
ports that are not in the table, such as the web server that normally uses
port 80. The first 1024 ports (0-1023) are reserved and can't be used by
normal processes, but those above 1023 are available to any process.
The code in Example1
is going to connect to a
server on the host "191.31.215.210" which is connected to port 32351. The
calls to htons and inet_network will be discussed later, but
the former guarantees that the port is in the correct internal format, and
the latter converts dotted-decimal notation to an 4-byte integer.
