Class 40: HTTP

Held Monday, May 8, 2000

Overview

Today we conclude our consideration of application-layer protocols by visiting one of the most popular Internet application protocols, HTTP, the hypertext transfer protocol.

Notes

• Grading scheme:
  • Drop lowest homework grade.
  • Homeworks: 40%
  • Final: 40%
  • Best of final and homeworks: 10%
  • Participation: 10%
  • Anyone who showed up for at least 80% of the classes, turned in reasonable work for every assignment, and completes the final will get a B- or better.
• Are there questions on the final examination?
  • I'll also reserve time on Wednesday and Friday for questions.
• Schedule for this week:
  • Today: Discussion of HTTP
  • Wednesday: Course evaluation. Attendance required.
  • Friday: Wrapup

Contents

• Side note on DNS
• Requirements
• HTTP Basics
• HTTP Problems

Summary

Side note on DNS

• The HTTP RFC includes a line of the form

  This requirement [that clients not cache DNS info] also improves the load-balancing behavior of clients for replicated servers using the same DNS name and reduces the likelihood of a user's experiencing failure in accessing sites which use that strategy.

• We might want to think about some implications of this.

Requirements

• Basic requirements:
  • Textual/human-readable protocol (so we can test using telnet :-)
  • Supports document transmission (of multiple kinds of documents)
  • Some user verification
• Questions to elicit design goals:
  • Why do we need something other than ftp?
  • What makes it hypertext?

HTTP Basics

• Request/response paradigm.
• Uses TCP.
• Primarily for getting documents. (Using GET request.)
• Caching/proxies built in to protocol.
• But also support for uploading information.
  • Using parameters to GET
  • Using POST
• Lots of additional fields available for sending and receiving.
• Some security, but with some (noted) flaws.

HTTP Problems

• Too much complexity
  • 109+ proportially-psaced pages
  • Many ways to accomplish the same tasks (five+ different ways to delimit messages; HEAD and conditional GET for checking updates)
• Too many layers rolled into one (not modular)
  • Message transport
  • Remote method invocation
  • Some document processing stuff
• Abused: Too many things built on top of it.
• Abused: Some clients use multiple http connections to improve bandwidth.