CSCI.4220 Network Programming
Class 6,Monday Feb 7, 2005
Some Application Layer Protocols

Here is Graham Doig's slide of the submtting for the Biotech Center

The HTTP protocol is ridiculously simple. A request consists of the request line, followed by zero or more headers, followed by a blank line (that means two conscutive newline characters), followed by a body, which is only present in POST requests.

In HTTP/1.0 there are only three types of requests.
GET - requests a document
HEAD- requests only the header info for a document
POST - sends some information

The newer version of HTTP, version 1.1 added two more request types, PUT and DELETE, that are hardly ever used.

Here is a link to an overview of the HTTP headers

The response consists of the response line, zero or more headers, a blank line, and the body.

There are only two important response codes, 200 OK and 404 document not found.

Do this right now. Log onto a unix machine and at the prompt, type 

telnet www.rpi.edu 80
then type 
GET / HTTP/1.0
and hit the Enter key twice

You should get the reply from the RPI web server, which consists of the headers followed the html code for the RPI home page.

If you just want to look at at the headers, type 

HEAD / HTTP/1.0
Here is a request sent from firefox. I ran a server on monica at port 8888, then typed http://monica.cs.rpi.edu:8888 in the browser. Here is what the brower sent. 
GET / HTTP/1.1
Host: monica.cs.rpi.edu:8888
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7.5) Gecko/20041107 Firefox/1.0
Accept: text/xml,application/xml,application/xhtml+xml,text/html;q=0.9,text/plain;q=0.8,image/png,*/*;q=0.5
Accept-Language: en-us,en;q=0.5
Accept-Encoding: gzip,deflate
Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7
Keep-Alive: 300
Connection: keep-alive
A Proxy Server is a network entity that satisfies HTTP requests on behalf of an original web server.

It caches recent documents. If the document is cached, it returns it immediately. Otherwise it passes the request on to the server. The server sends the document back to the proxy. The proxy stores a copy of the document and passes it on to the requesting client.

Note that it is both a client and a server at the same time.

In practice hit rates range from .2 to .7, thus dramatically reducing network traffic.

P2P File Sharing

P2P accounts for more Internet traffic than any other application. In P2P (Peer to Peer) everyone is both a client and a server.

To download a file, the client has to locate someone who happens to have that file, and then sends an http GET request. The search is the hard part. There have been three models.

  1. A centralized directory (Napster) which maintains a database of who is offering which files. The centralized directory collects information from peers as they connect, and periodically checks to see if they are still up.

    Three problems

    1. The centralized directory is a single point of failure
    2. It can be a bottleneck
    3. It is easy for the authorities to shut it down if they think it is illegal
  2. A fully distributed system (Gnutella) in which the peers which are connected at any given time form a network

    Each peer knows about its immediate neighbors and establishes a persistent connection to them. When someone wants to make a request, they send it to their immediate neighbors. If one of them can satisfy it, it sends a reply back and the requester then sends a GET request to download the file. Otherwise, each neighbor sends the request on to each of its neighbors, and the process continues.

    This process is called query flooding. One result is an exponential growth of packets. If each peer is connected to ten neighbors, the first round results in 10 requests, the second round results in 100 requests, the third round results in 1000 requests, etc.

    To squelch this, each request has a TTL (Time to Live) field, set to, say, 7. Each cycle decrements this, and when it gets to zero, it is not propagated any more.

  3. Kazaa (formerly KaZaA) used a hybrid system in which not all peers are equal. Some peers are group leaders. These are more powerful peers with greater responsibilities. Ordinary peers connect to group leaders, and each group leader has connections to many ordinary peers.

    Each group leader has connections to other group leaders and they are constantly sharing information.

    Kazaa also uses some performance enhancements. (it can limit the number of simultaneous uploads) It gives incentives to those peers who upload more files. It can support parallel downloads (If a file is on more than one peer, one can upload the first half and the other can upload the second half.)