servers.txt
+--

Web servers vs. applications, system and server administration

  Examples so far: web server and web application are the SAME program
  Typical organization SEPARATES web server from web application
  Web server handles generic services needed by any web site
   Serves many applications (programs), in different programming languages
   Manages sockets, multiple connections (like chat example, but many more)
   Access, permissions, authorization
   Redirection, proxies
   etc. ... See docs on how to configure Apache etc.
  Small servers used for development: ws30, SimpleHTTPServer, Django etc. ..
  Industrial strength servers for big sites: Apache, Nginx, Twisted, IIS (MS)
   configuring and running these is a job, a career path ...
  Servers seen by Python devs usually run on Linux
   our VMs provide Apache and Nginx







+--

CGI - Common Gateway Interface 

 A way to put almost ANY program on the web (not just Python)
  "almost any", no socket wrangling needed, all handled by server
 The ORIGINAL way to put programs on the web, influence is still pervasive 
   simple, supported everywhere, always works as a last resort

  1. Put myprogram.py in cgi-bin directory on server, (say) www.acme.com
  2. Point client at http://www.acme.com/cgi-bin/myprogram.py
  3. Server collects data from HTTP request
      passes data to myprogram.py and runs it
     myprogram.py writes output, usually HTML
     Server collects myprogram.py output, serves it back to client

 Much like our ws30.py exercise.  
  BUT - how to you get INPUTS to myprogram.py?









+--

CGI - sources of input

  All data must come from HTTP request: URL, headers, POST body
  HTML forms are the typical way for user to provide data
  HTML forms contents can be encoded in URL or POST body  
  
 
















+--

CGI program can get its inputs from ENVIRONMENT VARIABLES

   CGI program does not (can not) read HTTP request itself
    Only server can see HTTP request - must then communicate to CGI program

   Environment variables are similar to program variables but system-wide
    A way for programs to communicate with each other
    Can be read, set at command line:
     Unix (Linux, Mac):  echo $VAR       export VAR=value   printenv
     Windows:            echo %VAR%      set VAR value      set

    In Python, environment variables are in a dictionary
 
    import os
     os.environ['VAR']           # to read, look up in dictionary
     os.environ['VAR'] = ....    # to set, assign in dictionary












+--

CGI environment variables

  Server reads HTTP request, assigns environment variables
  CGI program can then read the environment variables 
  CGI standard defines which environment variables must be set

  http://www.citycat.ru/doc/CGI/overview/env.html
  http://tools.ietf.org/html/draft-robinson-www-interface-00

  Particularly important, used to store forms data: 
   REQUEST_METHOD (GET, POST etc.) CONTENT_LENGTH, QUERY_STRING
  
  Not just Python - environment variables can be read by any language










+--

CGI Python library
  
  Higher-level alternative to getting forms data
  Deals with GET and POST, treats form items as a dictionary
  Deals with error reporting

  (demo echo_cgi.html, .py)















+--

 WSGI - Web Server Gateway Interface

  CGI Program starts up and runs to completion EACH TIME client connects
   Poor performance
   Like HTTP itself - no memory beyond single request/response

  WSGI can support long-running programs
   Each HTTP request invokes a function call
   Much much better performance
   WSGI program can remember data from all previous requests
 
  WSGI is Python-only (unlike language-independent CGI)
    analogous to Servlet for Java, Rack for Ruby


+--

WSGI - How it works

   A WSGI application must provide a callable (function, class, ...)
    that the WSGI-enabled server can call.  
   Server provides args, application body uses them
   
     application(environ, start_response)
  
      environ - Python dictionary of CGI-style environment variables
      start-response - function that sends headers

   WSGI application (not server) does URL routing

    (demo echo_wsgi.py)


+--

 WSGI applications - layering

  A WSGI application body can itself call another WSGI application
   environ is a passed parameter, NOT global, each layer can customize   

    Server <--> WSGI <--> Application

    Server <--> WSGI <--> WSGI <--> ... <--> Application

 WSGI toolkits: Werkzeug, Paste, ...
  for error handling, URL routing, sessions and authentication

 Supports layering in traditional Internet fashion
  interposed layers are sometimes called "shims"