Local Area Network Technology: Part 5: TCP/IP Protocol

by Jeff Carrell, Electronic Communications Chairman [reprinted from IEEE - Fort Worth Section Signals September 1997]

Computers "talk" together using a set of languages called protocols. Network protocols are as different as spoken languages around the world. Unfortunately, there is not a single protocol standard that all computers use, but there are a few common ones that most computers can use.


The primary protocol used on the Internet and on UNIX based networks is TCP/IP. The actual address component is IP. There has been one protocol version in use for the last 18 years, called IPv4. The maximum limit of addresses available on any one network is over 16 million, but only 127 of these network numbers exist. In 1991 the IETF formed a committee to design a new numbering system that would allow for many more numbers, while still being backward compatible to IPv4. This new IP numbering scheme is called IPv6 and is in the final stages of testing and should become a standard soon.

Any network requiring connection to the Internet needs a valid assigned IP address. The actual controlling body that assigns the addresses is called the InterNIC. Although it looks like a lot of addresses could be used, the actual limit of assigned addresses is about to be reached. Factors contributing to the shortage: many companies with assigned numbers are not directly connected to the Internet and recently, the InterNIC started assigning contiguous blocks of addresses to Internet Service Providers (ISPs), which in turn assigns addresses to its subscribers.

First OctetMax NetworksMax Hosts per networkFormatSubnet Mask (defaultAddress


IPv4 uses a 32 bit number, represented by four 8-bit bytes (or octets) in dotted decimal notation. The address is broken down into two parts: the first is called the "network" address and the second is called the "host" address. There are also certain "class" numbers used and denoted in the first octet. The class designation defines how many actual host addresses are available in that specific network number. Network address 127 and all Class D addresses are reserved for specific use. The chart below indicates the numbering layout.

Intranets use an additional addressing scheme, subnetting, in order to have more host addresses than might normally be available with the IP address assigned.

With subnets, the host portion of an IP address is partitioned into a subnet number and a 'real' host number. The IP address is now defined as net.subnet.host. Routers outside of the Intranet do not interpret separately the subnets and host partitions of the overall IP address. Routers inside the Intranet and connected to the subnets use a 32 bit subnet mask that identifies the extension bits. The network administrator allocates bits within a subnet address and assigns values to the subnet and host addresses. The chart below indicates some subnet mask numbering schemes.

# BitsSubnet Mask# Subnets# Hosts
Cls B
Cls C


IPv6 (formerly IPng) uses a 128 bit number, represented by eight 16-bit sections. An example might be: 5F03:2500:CFB5:D700:0000: 0260:8C6C:9A9C. With leading zero's truncated, you would have: 5F03:2500:CFB5: D700::260:8C6C:9A9C.

Many new features and enhancements are designed into IPv6. Some of these are: enhanced addressing capabilities - from 32 bits (IPv4) to 128 bits; authentication and privacy capabili- ties and extensions to support these security features; and flow labeling capability - providing for special handling of data. There are many tests still in process on Ipv6 and many products will require extensive software rewrites in order to allow IPv6 to interoperate with IPv4. Router software, DNS applications, client protocol stacks, TCP/IP applications, will all be required to update to IPv6 and network administrators will need a general understanding of the IPv6 protocol addressing scheme.

There is a test network available on the Internet, called the 6bone, on which vendors and ISPs are currently conducting stress tests. It may be a year or two before most products are commercially available. Before a company tries to implement IPv6, it is highly recommended that they acquire the IPv6 products and connect to the 6bone and participate in the IPv6 tests to ensure proper operation before introducing the products in their Intranets.


Copyright © 1997 Jeffrey L. Carrell All Rights Reserved