Further Notes on Chapter Two
Posted 15 Sep 2005 by herman

Some extra information (mostly discussed in class).

Terminology

Chapter Two of the text is heavy on terminology, much of it not explained. I'll try giving at least a superficial idea of what the following mean during the week's lectures.

  • The OSI seven-layer model (7 physical; 6 link; 5 network; 4 transport; 3 session; 2 presentation; 1 application, but layers 2 and 3 are obsolete as far as we're concerned)
  • The Internet TCP/IP protocol stack (often merge link and physical layers into a "LAN" or "MAC" layer; IP is the network layer; TCP or UDP are at the transport layer; other protocols such as SMTP, HTTP, FTP are at application layer)
  • Layered Design Philosophy -- implement Layer k-1 functions by software in Layer k, by invoking only the functionality of Layer k+1.
  • Data Terms -- bit/signal at physical layer, frame at link layer, datagram at network layer, packet (sometimes also segment) at transport layer, message or stream of bytes at application layer. Frames, datagrams, packets, each have their own headers that are intended only for the corresponding layer's processing.
  • IP addressing: all IP datagram forwarding decisions are on the basis of IP addresses, currently 32-bit numbers, which become 128-bit numbers when IP version 6 is deployed.
  • Port Numbers: 16-bit numbers used to distinguish different network applications running on same host; a TCP connection is mathematically identified like a 4-tuple (IP-x,IP-y,Port-x,Port-y). Some port numbers are well known and associated with certain application protocols, such as 80 for HTTP.
  • Network Metrics: [not actually defined in the chapter, but basic to networking]: latency, bandwidth, recovery delay, availability.
  • URL, URI, URN: slightly different meanings, as explained in the text.
  • hourglass waist: a way to view the IP protocol stack.
  • NAT is Network Address Translation, which enables a number of hosts to share one IP address. Often used along with a firewall, which typically can filter datagrams to disallow hacker access or other unauthorized network traffic.
  • Proxy hosts work with web browsers by indirect HTTP requests, that can be forwarded past the firewall.
  • Persistent Connection is a technique where a web client can use one connection for multiple HTTP GET requests, rather than obtaining new connections for each GET (see Fig 2.5 in the text to see the overhead for opening a TCP connection).
  • Checksum and CRC are techniques to detect, with high probability, the presence of a data error in a [frame, packet,file,etc].
  • Stateful vs Stateless Servers -- explained in class.
  • Cookies
  • Access Control, Authorization, SSL, TLS
  • Multicast vs Unicast: the need to go beyond communication between only two endpoints; the possibility of implementing multicast at different places in the network and in the layered architecture.
  • The End-to-End Principle. Explained in the textbook and in class. Further references (optional reading, not required):

Home PageHelp PageView ArticlesView Public DocumentsLogin, do stuff (record diaries)List Group MembersList of Projects