The OSI and TCP/IP network models

When TCP/IP was invented, the designers sought to make the network scheme as flexible as possible. So the network functionality was divided into layers, and different functions were allocated to the different layers. Years later, some of the wrinkles and problems with the function allocation were addressed in a new networking scheme, called Open Systems Interconnect (OSI). While OSI never became popular (neither has Ipv6)1, the OSI model represents a convenient way for thinking about networks.

The following table describes how functionality is allocated in the OSI and TCP/IP systems.





OSI layer

TCP/IP layer

Functionality

Devices

7

Application

Application

The applications

Clients, Servers, Proxies

6

Presentation

Data translation, encryption

5

Session

TCP


Creating and destroying connections

Virtual Servers (VS), or Virtual IPs (VIPs), firewalls

4

Transport

UDP

Reliable delivery of data

3

Network

IP

Routing packets

Routers, SNATs, DNATs

2

Data link

Physical

Control of the wire

Switches, bridges, hubs

1

Physical

Moving data on and off the wires

Wires

Table 1: Comparison of the TCP/IP and OSI stacks

The different devices modify the traffic that flows through them.

1DECnet, SNA, Novell, OSI are all examples of networking technologies which tried to compete with IPv4 and failed. IPv6 has not caught on yet. Why? In the case of DECnet, SNA, and Novell, they were proprietary technologies. Although there are implementations of these systems in the open source world, it is understood that the implementations are for compatibility reasons. In the case of OSI and IPv6, although they are open and they solve certain problems with Ipv4, they are not so much better as to provide a compelling reason to abandon the existing, ubiquitous IPv4 infrastructure.