Week 2 Notes - Internet Infrastructure

For a splendid article describing the fundamentals of the Internet's infrastructure, including comments on it's history, evolution, and current state, I wholeheartedly reccommend that you take a look at this article.

If you need to make an informed choice about which provider you should go with or which connections are the "good" connections, have a look at this page.


The following are terms which we will cover in this lecture. You should know what these terms mean.

Brief History

As mentioned in the first week's notes, the Internet originally began as a military project known as ARPANET. It then spread to research organizations and universities.

Then, the NSF (National Science Foundation) decided to fund the development of a number of supercomputer facilities across the United States. The research/academic community wanted to have access to these computing resources, so they got connected. This early setup was the "seed" of the future Internet.

After awhile, the NSF network became popular for purposes other than pure research: email, file transfers, and newsgroups i.e. The network grew and grew, expanding to other research orgs and universities. Eventually, IBM and MCI began to fund the expansion and helped upgrade the lines to T1 and eventually T3 speeds.

The nature of the traffic on the NSFNET was becoming increasingly commercial, and much debate was had over the future privatization or commercialization of NSFNET. The NSF no longer wanted to be in the backbone business, so they sold off their network and established the concept of Network Access Points (aka NAPs). Now, rather than independant connections being made along the whole backbone, people could connect at single designated points. Four of these were originally established accross the U.S. in San Fransisco, Chicogo, Pennsauken (New Jersey) and Washington D.C.

This concept of a NAP is important to understanding the modern layout of the Internet. Building on the idea of NAPs, large network providers can resell their bandwidth to smaller providers, and establish "mini-NAPs" of their own to which other people can connect. These smaller providers can then resell their bandwidth and so forth.

Alternatively, network operators can decline to access the Internet at a NAP and instead just "peer" with each other by laying a line between them.

Normally, it is not in a company's best interest to "share" its customers with other companies. The following describe why they do anyway:

  1. The ARPANET and NSFNET were not developed with Commercial interests in mind.
  2. Transmitting packets throgh long and tortuous routes is INEFFICIENT. It is better to "share" traffic (read: customers) with another provider if it can get the packets to their destination more quickly so you can offload that trafic from your network.
  3. The demand by customers to be connected to "The Internet" has greatly outstripped any company's "best interests". (Case in point: AOL)

Present Day

A picture is worth a thousand words, so have a look at the following illustration. Something to keep in mind as you look at this illustration: There are currently about 4,500+ bandwidth vendors.

Tracing a session from end to end

The following is a painfully oversimplified synopsis of what happens when a request is made from a client (say a browser or a newsreader) to a remote server on the Internet.

Keep in mind that at any point along the way, a request can be satisfied if the destination is reached before the request ascends to the higher levels. Also keep in mind that at any point along the way, the request can be passed LATERALLY (sideways) to somebody at the same level. This is known as *peering*. This is an important concept to know about.

Some Words on Routing

A router is like an intersection of streets. Cars come in to the intersection, cars go out of the intersection. They might turn at the light, they might go straight. It all depends on what the ultimate destination of the driver is. Note that it is impossible to tell what the ultimate destination of any given car is just by seeing which direction it turns (or doesn't).

In like manner, routers do not send packets to their destination, rather they try to get packets closer to their destination. They do this by examining their routing table and determine where to send a packet based on its IP address.

Some Words on DNS

Domain Name Services (or DNS) maps host names to IP addresses. Can be done with a flat text file. DNS software (either bind or named) will communicate with each other to update their lists. There is a "root server" for all DNS entries (InterNIC) which is The Authoritative Source. Some talk is being made about decentralizing the DNS system to make it more robust and fail-safe.

DNS entries are given as domains. Some top-level domains (or TLDs) have been created to make indexing entries more efficient. The second-to-last entry is your organization's domain name. Everything below that (and by that I mean any other prefixes--yes, I realize this seems backwards) are sub-domains within your own domain.

Top Level Domains

Much hubub is being made about increasing the number of TLDs because some of the entries are getting a little full (notably, the .com category).

Problems with the Internet

HTML Stuff

Before embarking on today's excursion into HTML, there are two large divisions of tags that you should be aware of:

The following covers basic text formatting. I expect you to look up these tags in the HTML docs from the W3C site (see the resources page for details). I'm not giving long descriptions here, but we will go over them in the lab and do some experimenting.

*Whew* That's about enough for one week, I think.


1/20/98 - Initial revision.
1/21/98 - Added HTML stuff.
2/2/98 - Added .int Domain Name