TCP/IP Configuration and Management Manual

NonStop TCP/IP Processes and Protocols
TCP/IP Configuration and Management Manual—427132-004
B-5
The Problem of Resolving Addresses
Encapsulation of Messages
Because TCP/IP uses encapsulation, (placing a message of a higher layer into the
data portion of the lower layer which then adds its own header), we cannot rely solely
upon IP addresses to communicate with other machines on the network. When the
network interface is assembling its own frame to place onto the medium, it adds its
own header information which must contain, not an IP address of the destination, but a
hardware destination address. The IP address of the destination is buried in the data of
the network interface frame.
Static Binding of Addresses
Some networks use tables that include both the hardware and IP addresses. Other
networks encode the hardware addresses into the IP addresses. In either case, this
method of mapping IP addresses to physical addresses is known as static address
binding. It is static because the information is static or when the information does
change, it is done manually. Changes to this information are usually accomplished by
updating a table that may be stored in an edit file.
The NonStop Implementation of TCP/IP over X.25 uses static address binding.
Dynamic Binding of Addresses
The NonStop implementation of TCP/IP over an Ethernet network or an IEEE 802.3
(SNAP) network avoids the limitation of static address binding by using the Address
Resolution Protocol (ARP). The ARP is fully described in RFC #826 of the Internet
standards.
Once a host receives information relating to the IP and hardware addresses of other
hosts on the network, that host stores the information in cache. This is known as
dynamic address binding. Since most Internet traffic consists of communication
between pairs of machines, the cache eliminates the need for most ARP requests.
Consequently, this Internet-to-physical address information is stored in each host
and/or router that is attached to the Internet. But doesn’t this result in a large number of
addresses stored in all machines on the Internet? In fact, it doesn’t because ARP is
used only on the local physical network. Thus, only those hosts and gateways/routers
that are directly connected to the same physical network share one another’s Internet-
to-physical address information.
Subnet Addresses
In the model, each network can be divided into subnetworks, or subnets. Within a
given network, each subnet is treated as a separate network; outside that network, the
subnets appear as part of a single network. Each subnet attached to a host is reached
through a device (or a pseudo-device, such as a loopback driver) that provides an
interface for IP data transmission. In this manual, the term subnet also refers to the
interface through which the subnet is reached.