With classless addressing, the network address and IP prefix are often expressed in the form: xxx.xxx.xxx.xxx/n where xxx.xxx.xxx.xxx is the lowest IP address in the network and n is the number of bits in the network-address portion of the IP address. For example, with a network address and IP prefix of 192.0.2.0/24, the network address is 192.0.2 (three bytes, or 24 bits), and the final eight bits in the IP address are the host address. In routing datagrams for addresses that use classless addressing, routers use Classless Inter-domain Routing (CIDR) protocols defined in RFC 1519.

IP Addresses Reserved for Special Uses

Some IP addresses are reserved for special uses. A network address or host address can never be all zeros or all ones. So, for example, in a network with an IP address and IP prefix of 192.0.2.0/24, the hosts can have a host address of any value from 1 to 254, but not 0 or 255. There is no network at 255.255.255 or 0.0.0.0

The Local Host

The address 0.0.0.0 refers to the local host or network, also called “this” host or network. In a network with a DHCP server, a host sends a datagram with a source address of 0.0.0.0 to request the server to assign an IP address.

Broadcast Addresses

A destination address of all ones is a broadcast to all hosts in a network or subnet. A destination of 255.255.255.255 would appear to be a broadcast to the entire Internet, but in fact, Internet routers and most other routers ignore broadcasts, so the datagram only goes to the hosts in the local network or subnet. Individual hosts may also be configured to accept or ignore broadcasts. A broadcast can also specify a network or subnet, with the host address and subnet ID, if any, set to all ones. For example, a network with this network address and IP prefix: 192.168.100.0/28 can have up to 14 hosts (192.168.100.241 through 192.168.100.254) And a broadcast to: 192.168.100.255 is directed to all hosts in the network. As Chapter 1 explained, an Ethernet frame with a destination address of all ones is another way to do a broadcast.

Loopback Addresses

Addresses with the most significant byte equal to 127 are loopback addresses reserved for loopback tests. On receiving data to transmit to a loopback address, the IP layer passes the data back up to the source instead of passing the datagram down for transmitting on the network. Transmitting to the loopback address can be a useful test of the local networking software. Multicasting Another option for sending datagrams to multiple hosts is multicasting, where a source addresses a datagram to a specific group of hosts that may reside in different networks and subnets. Uses for multicasting include sending audio and video to subscribers. Classful addressing reserves the Class D addresses for multicasting. In practice, multicasting on the Internet has been uncommon because all routers between the source and destination must support multicasting, and many routers don’t. Multicasting is feasible within local networks, however. As explained in Chapter 1, in Ethernet networks, destination addresses can also identify multicast groups.

Local Addresses

In a local network that doesn’t connect to the Internet, the IP addresses only have to be unique within the local network. An address range in each class is reserved for local networks that don’t communicate with outside networks:

Class A: 10.0.0.0 to 10.255.255.255

Class B: 172.16.0.0 to 172.31.255.255

Class C: 192.168.0.0 to 192.168.255.255

These ranges are preserved with classless addressing as well. The addresses are for use within networks where the network administrator can ensure that no two hosts have the same address. A network that uses addresses in these ranges should not connect directly to the Internet or to another local network that might use the same addresses. However, as explained earlier in this networking tutorial, it’s possible to connect computers with local addresses to the Internet by using a router that performs Network Address Translation (NAT).

Other Reserved Addresses

RFC 3330: Special-Use IPv4 Addresses lists other reserved ranges of IP addresses.

Manual Allocation

In manual allocation, the network administrator specifies an address for each host, but instead of configuring the addresses at each host, the administrator configures all of the addresses at the DHCP server. On receiving a DHCPDISCOVER message, the DHCP server returns the address assigned to the requesting host. For example, in an Ethernet network, the network administrator can provide the server with a table that matches an IP address to the Ethernet hardware address of each Ethernet controller in the network. The DHCP server reads the source’s Ethernet address from the Ethernet frame, finds the corresponding IP address in the table, and returns the address to the requesting host’s Ethernet address. Manual allocation is more convenient than configuring an address at each host, but the allocation still requires the administrator to know each host’s hardware address and to assign an address every time the network gains a new host.

Automatic Allocation

In automatic allocation, instead of maintaining a table of values matched to hardware addresses, the DHCP server begins with a list of available IP addresses. On receiving a DHCPDISCOVER message, the server selects any unassigned address to return to the requesting host and marks the address in the table as assigned to that host.