Building Your Very Own Web Server

1
Fun Fun Project One

• Building Your Very Own Web Server

2
What is a Web Server?

• Program that understands the HTTP protocol and
  generates appropriate responses
• Clients connect to the machine
• Clients send a request
• Server reads request, generates response
• Client interprets response appropriately

3
A Simplified Web Server

• Client asks for file
• Server finds appropriate file
• Server sends back a response header followed by
  the files data
• Server closes connection

4
What Does Connect Mean?

• For all practical purposes, it looks like theres
  data available via a file descriptor
• Stream of bytes
• Can be treated like any other file descriptor
• Not a FILE (like stdio, stderr)
• Must use read() and write() system calls

5
How Do You Identify Machines

• Names or numbers and ports
• http//www.domain.com implies a machine named
  www.domain.com and a default port of 80
• http//127.0.0.131415/index.html
• Refers to current box (127.0.0.1 is me)
• Port is 31415 (used for this project)
• File is named index.html

6
How Do You Identify Files?

• File name is specified in request
• Server maps that name to a real file
• Mapping can be whatever server wants
• For example, /vivek/index.html is really
  /n/fs/fac/vivek/public_html/index.html

7
Whats In A Request?

• GET /index.html HTTP/1.0\r\n
• Connection Keep-Alive\r\n
• User-Agent Mozilla/4.72 en (X11..)\r\n
• Host 127.0.0.131415\r\n
• Accept image/gif, image/jpeg, /\r\n
• Accept-Encoding gzip\r\n
• Accept-Language en\r\n
• Accept-Charset iso-8859-1,,utf-8\r\n
• \r\n

8
What Do You Care About?

• GET /index.html HTTP/1.0
• In particular, just index.html
• Assume / means /index.html

9
What Could They Want?

• An honest-to-goodness file (me.jpg)
• An indirect request for such a file (such as /
  meaning index.html)
• An implied directory with index (/home/vivek
  instead of /home/vivek/)
• Just a directory listing
• A query (we dont care about these)
• An invalid/nonexistent file

10
Whats In A Response?

• HTTP/1.0 200 OK\r\n
• Date blah-blah-blah\r\n
• Server blah-blah-blah\r\n
• Content-Type important\r\n
• Last-Modified blah-blah-blah\r\n
• \r\n
• Raw data

11
Whats a Minimal Response?

• HTTP/1.0 200 OK\r\n
• Content-Type stuff\r\n
• \r\n
• Data
• HTTP/1.0 302 Moved\r\n
• Location newurl\r\n
• \r\n

• HTTP/1.0 404 Not Found\r\n
• \r\n
• But also
• Connection close\r\n
• Content-Length yyy\r\n

12
How Do You Decide?

• File exists? Send it
• Directory without / suffix? Redirect
• Directory with index.html? Send it
• Directory with no index.html? List it
• For each list entry, add / if needed
• Failure? Send 404

13
How Do You Test It?

• Use a browser
• Use wget or create hierarchy
• Include some big images

14
What is Content-Type?

• text/html
• image/gif
• image/jpeg
• Others not needed

15
Where Can I Find More?

• Google HTTP 1.1 Specification - painful
• man pages
• man man
• man k blah
• read( ), write( ), open( ), close( )

16
Why open instead of fopen?

• Compare fopen, fread, etc., with open, read, etc
• Were dealing with functions closer to the OS
  easier to use in some cases
• Practice

17
Whats a File Descriptor?

• Sort of like a FILE
• Its an integer provided by OS
• Used to represent a stream of bytes
• Can represent file or network connection
• Behavior is slightly different
• Especially when reading/writing network

18
General Steps

• Setup, and then
• Get next connection
• If file, read from disk
• If directory, generate listing
• Send all to client
• Close connection, wait for next one

19
What Am I Given?

• Setup function
• Makes server available for connections
• Accept function
• Gets a connection from a client
• File type function
• Tells you what kind of file, if it exists
• Tells you how many bytes if a regular file
• Directory listing functions
• Gives you the file names one at a time

20
Help! Im Lost!

• Dont know HTML?
• Use Netscape composer to see what to do
• View page source for various pages
• Do telnet www.domain.com 80 and issue the GET
  manually (need to add Host www.domain.com
  header)
• Ask

21
Why Are We Doing This?

• Infrastructure for future projects
• Some OS/Networking interaction
• Its fun, and not too bad