Application Layer
High Level
Sits at the top of the OSI model, the Application Layer provides services for applications via the socket API.
HTTP
HTTP is a stateless protocol which uses TCP (instead of UDP) in the Transport Layer. From the developer’s point of view, an application can use TCP via TCPSocket Interface and UDP via UDPSocket Interface, more about that here.
HTTP is defined in [RFC 1945] and [RFC 2616]. The requests are in plain text. You can try it with telnet.
telnet cis.poly.edu 80
GET /~ross/ HTTP/1.1 Host: cis.poly.edu
~ Computer Networking: A Top-Down Approach (6th Edition)
Note that HTTP works in a pull protocol. This means the machine that receives data initiates TCP connection.
Cookies
Cookies is a feature to help make HTTP stateful. It is described in several RFCs. The latest as of this writing is [RFC 6265]. Cookies works because browsers/agents actually support it. If you try using an agent that does not support it, it just won’t work.
Web Cache
A Web cache—also called a proxy server—is a network entity that satisfies HTTP requests on the behalf of an origin Web server.
~ Computer Networking: A Top-Down Approach (6th Edition)
In a typical settings, HTTP requests from machines in a LAN network are routed to a web cache inside that LAN network. Web cache fetches HTML objects for other machines from remote web servers and stores a copy. Web cache will utilize this copy for future request for the same object instead of having to fetch it from the remote web server.
Question What if the cached object is older than what is served on the web server?
The answer is that whenever a web cache fetches an object from web servers, it stores a copy with the fetching time. Web cache typically sends a Conditional GET to web server with Header if-modified-since:<time>, asking if the object has been modified since the last time it was fetched. Response code 304 means the object has not been modified since.
FTP
File Transfer Protocol is stateful and also uses TCP.
It uses two TCP connections over:
- port 21 on server: for sending control commands
- arbitrary port on client: for sending actual data
The protocol is defined in [RFC 959]
Electronic mail
A typical setting of how electronic mail works:
Sender's mail agent --> mail server A -> mail server B -> Receiver's mail agent
Mail agent could be a mail application (iOS Gmail app).
Connection between mail servers:
The protocol between mail servers B is SMTP, defined in [RFC 5321]. The sending server establishes TCP connection to port 25 of receiving server.
As compared to HTTP, SMTP is a push protocol in the sense that the machine sending data initiates TCP connection.
Just like HTTP, the protocol is composed of 7-bit ASCII-encoded commands. You can try sending these commands by initiating a TCP connection to port 25 of an email server with telnet.
Connection between agent and its respective mail server:
- Sender’s agent -> Sender’s mail server: SMTP
- Reciever’s mail server -> Receiver’s agent: POP3 ([RFC 1939]), IMAP ([RFC 3501] or HTTP
Domain Name System
See more here Domain Name System (DNS) Part 1
BitTorrent
See more here BitTorrent Protocol