World Wide Web

Major problems with the net (circa 1994)

Cryptic command line interface
Complete lack of organization (also a strength)
Incredible growth, change

WWW is the best answer to these major problems so far

GUI, hypertext, searchable, de-centralized

WWW - mixes content with menus. Gopher - separate content and menus.

hypertext on the Net
Swiss army knife of Net tools (subsumes functionality of ftp, Gopher, archie, news readers)

Publishing model

From few-to-many, with high costs and long delays, to many-to-many with low costs and no delay. Business model

Targeting, interactivity, customization URL - Uniform Resource Locator

http, ftp, file (local disk), news, gopher, mailto, telnet

Client - Web browser

Browser history

Browser features

current problems: HTML isn't very rich

Role of MIME

The web uses MIME (Multimedia Internet Mail Enhancement) to encode data.

Hyper Text Transport Protocol - HTTP

References

Webmonkey: backend: HTTP Transactions and You

http://www.hotwired.com/webmonkey/backend/

Simple page transfer scenario <figure 7-59 Tanenbaum>

Browser displays default page
User enters URL (http://www.site.domain/~person)
Browser library call resolves hostname to IP address via DNS server
Browser makes TCP connection to port 80 of host
Browser sends a GET command for page ~person/Welcome.html
Server www.site.domain sends desired HTML page
TCP socket connection is closed
Browser interprets HTML, displays page
Browser establishes separate connections for images, may use threads

The server side

The httpd daemon is waiting for incoming connection requests. When it hears one it:

allocates a new socket for this client request
forks a kid process (or creates a thread)
hands the client off to the kid
goes back to listening.

HTTP is built on TCP. Version 1.1 came out fall of 1997. This brings up some questions:

Is HTTP connection-less or connection-oriented?

Is TCP a good match to the idea of web browsing?

What are the advantages and disadvantages of using TCP for HTTP?

How might efficiency be improved?

Their are seven built-in methods to HTTP, and a means of extending the protocol. Each request is followed by a status line, optionally followed by data (e.g. the URL content from a GET). The built-in methods are

GET - request to send a web page

An If-Modified-Since header allows for the server to reply without data to this request so a cached copy is used instead, saving transfer time

HEAD - request to read a web page's header

Handy if you want to check a URL, get its modification time, or collect info for indexing.

PUT - request to store a web page

Web server must be configured to allow for this. Options to allow for authentication header.

POST - request to append to a named resource

Like PUT, but does an append. Used for adding a file to bulletin board.

DELETE - request to delete a web page

Authentication header may be used. File system ultimately determines success or failure of this request.

LINK/UNLINK -

The request would look like this

GET /mattmarg/ HTTP/1.0

User-Agent: Mozilla/2.0 (Macintosh; I; PPC)

Accept: text/html; */*

Cookie: name = value

Referer: http://www.webmonkey.com/html/96/47/index2a.html

Host: www.grippy.org

The status replies are things like

A redirect response can be sent by a server when a URL is requested that the server knows isn't right (for example, when you ask for someone's home page without including the ending /) or when a URL has a redirect in it (e.g. someone moved the page and didn't want everyone to get 404s on the old URL). The redirect is put in the .htaccess file in the URLs directory, or in the server's configuration file.

After the response code comes the response header

HTTP/1.0 200 Found

Date: Mon, 10 Feb 1997 23:48:22 GMT

Server: Apache/1.1.1 HotWired/1.0

Content-type: text/html

Last-Modified: Tues, 11 Feb 1997 22:45:55 GMT

Performance

The first version of HTTP was really inefficient. One request per socket meant multiple socket setups per page. Various things have been done to improve performance and waste less bandwidth.

Keep-Alive

To improve efficiency and reduce latency of downloading complex web pages a server allows for more than a single request to be sent down a TCP socket connection. These multiple requests may be pipelined (i.e. more than one can be sent before a response to the first is received). There is a configured maximum number of requests per socket, plus a timer value the server uses to tear down idle sockets. Caching

All browsers keep a disk cache for URLs they've fetched before. The cache is checked before the URL is downloaded. HTTP 1.1 gives some fine grained control over what is cached, so a page can be written which specifies that some pieces of it may come from a cache, but that others must be reloaded every time. Proxy servers

Note that the cacheing done by a proxy server is common to all users/client browsers, and not just limited to one single user/browser, so the performance boost is potentially greater. One large campus proxy server sees hit rates of between 50 and 59%.

Configuration of web server

Directives

Logging

Accesses
pm087-07.dialip.mich.net - - [31/Aug/1996:05:00:25 -0400] "GET /~wolfd/now8.gif HTTP/1.0" 200 1135
muskie.csis.gvsu.edu - - [04/Sep/1996:12:20:23 -0400] "GET /~erickson/CS380/Content.html HTTP/1.0" 200 1404
yandr-bh.yr.com - - [04/Sep/1996:12:06:07 -0400] "GET /~vanoflej/HenryMiller/images/cubeneg.jpg HTTP/1.0" 200 1784
148.61.25.48 - - [04/Sep/1996:12:02:15 -0400] "GET /~levinm/237downl.html HTTP/1.0" 200 1606

Errors

[Fri Aug 30 15:30:08 1996] httpd: access to /home/erickson/public_html/NetworkingArchives/WindowsSockets/KenCoviak/descript.html failed for barpm3_p9.caribsurf.com, reason: file does not exist from http://www.altavista.digital.com/cgi-bin/query?pg=q&what=web&stq=20&fmt=.&q=wsacleanup