Addressing

Each network interface of each host or gateway must be identified with an IP
address, and these addresses must not be repeated. The IP addresses are structured
in a hierarchical way, as they are made up of numbers of 32 bits, and
one part identifies the network, netid, and the other identifies the host within
the network, hostid.
It was impossible to predict how the networks would be configured
when the Internet was created, and it was necessary to adapt the format of
the IP address to all types of current and future networks. For this purpose,
five address formats were defined corresponding to a series of classes, from
class A to class E.
Classes A, B, and C are currently used, whereas class E is reserved for
future use. Figure 3.3 shows the format for each of these classes. From the
figure, it can be deduced that class A has a great capacity to address hosts
within the subnetwork and so is an appropriate class for large networks,
while class C has a much smaller capacity for addressing hosts within this
same subnetwork and so is suitable for subnetworks with up to 256 hosts.
On the contrary, as only seven bits have been assigned to netid, in class A, the
number of addresses of this class is low, exactly the opposite of what happens
with class C addresses. Finally, class D belongs to what is called multicast address and is used to
transmit datagrams simultaneously to a group of hosts making up a multicast
group.
In practice, the network address makes use of a notation corresponding
to the decimal value of each octet separated by points. Table 3.2 shows the
ranges of valid values for the different classes expressed in decimal notation.
There is a series of addresses that by convention have a special meaning,
which means that they are reserved. They have all the bits at 0 or all the bits
at 1 of the netid and hostid fields. Thus, if the hostid field has all its bits at
zero, it is understood that the address corresponds to the network and not to
a specific host. On the contrary, if this field has all of the bits at 1, it indicates
that the datagram should be broadcast to all of the hosts connected to the
subnetwork. This is denominated broadcast address.
The network addresses are assigned by the Network Information Center
(NIC). Generally, portions of the network are assigned to organizations, and
these organizations are left to handle addresses internally. When working in a network isolated from the exterior, free addressing can be used, but it is
advisable to have a network number set by the NIC in case of connection
with the exterior.
Given that the human handling of the IP addresses is not simple, each
address has an associated name. Thus, for example, the IP address
193.144.186.2 is assigned the name moon.tlmat.unican.es. This assignation
of names is hierarchical; it is based on a scheme of the form system.subdomain...
subdomain.subdomain.domain. There is a distributed application,
Domain Name System (DNS), permitting the resolution of associations
among names and IP addresses.
To finish this section, some of the problems associated with the
addressing scheme are commented:
• If a host transfers from one network to another, it must change its IP
address completely.
• When a class C surpasses 254 hosts, it must change to class B, which
obliges the readdressing of all of the hosts of this network.