Manuel Landesfeind. August 31th 2011

Size: px

Start display at page:

Download "Manuel Landesfeind. August 31th 2011"

Claire Wheeler
7 years ago
Views:

1 Manuel Landesfeind Source Talk Tage 2011 August 31th 2011

2 HTTP basics GET /index.html HTTP/1.1 Host: /path/to/index.html HTTP/ OK Content-Type: text/html <html>...</html>

3 HTTP header Using HTTP 1.1 a client may fetch data from a webserver using different methods. request GET / HTTP/1.1\ r User Agent : c u r l / ( x86 64 pc l i n u x gnu )... \ r Host : www. l a n d e s f e i n d. de\ r Accept : / \ r response HTTP/ OK\ r \n Date : Tue, 30 Aug : 38: 21 GMT\ r S e r v e r : Apache / ( Debian )... \ r Last Modified : Wed, 29 Jun : 50: 19 GMT\ r Content Length : 1778\ r Content Type : a p p l i c a t i o n / xhtml+xml\ r <?xml v e r s i o n= 1. 0 e n c o d i n g= UTF 8? >...

4 HTTP header More data may be send using POST method: POST / index. html HTTP/1.1\ r Host : www. example. com\ r Content Type : a p p l i c a t i o n /x www form u r l e n c o d e d \ r Content Length : 27\ r \ r param1=v a l u e 1&param2=v a l u e 2 POST / HTTP/1.1\ r Host : www. example. com\ r User Agent : c u r l / ( x86 64 pc l i n u x gnu )... \ r Content Length : 150\ r Content Type : a p p l i c a t i o n / base64\ r \ r cgrmbgf0zxg9cgrmbgf0zxgglwhhbhqtb24tzxjyb3iglwzpbgutbgluzs1lcnjvcgokywxsogojjhtw ZGZsYXRleH0gUHJhZXNlbnRhdGlvbgoJJHtwZGZsYXRleH0gUHJhZXNlbnRhdGlvbgoK

5 Extending the transport protocol WebDAV RFC 2518 extends HTTP for well known filesystem methods and allow the manipulation of data online, e.g. files. Methods include COPY, MOVE, (UN-)LOCK. REST-API Uses few well defined methods executable on URI ressources. Initially described in the context of HTTP using its huge vocabular of methods and status codes. webdav.org other DAVs CalDAV (RFC 4791) CardDAV GroupDAV... and others

6 Scheme of a cgi-communication Instead of just returning a file, software is executed: GET /cgi-bin/foo.cgi HTTP/1.1 Host: system("/path/to/foo.cgi"); HTTP/ OK Content-Type: text/html <html>...</html>

7 Data transmission HTTP header information is stored using environment variables. HTTP body data is written to the standard input of the executed program. Programm standard output is returned to the client. $ENV = { SCRIPT NAME => / c g i b i n / t e s t. c g i, SERVER NAME => l a n d e s f e i n d. de, SERVER ADMIN => webmaster@landesfeind. de, HTTP CONNECTION => Keep Alive, REQUEST METHOD => GET, HTTP ACCEPT => /, SCRIPT FILENAME => / v a r /www/ de. l a n d e s f e i n d / c g i b i n / t e s t. c g i, SERVER SOFTWARE => Apache / ( Debian )..., QUERY STRING => param=value, REMOTE PORT => , HTTP USER AGENT => Wget /1.12 ( linux gnu ), SERVER SIGNATURE => < address>apache / </adress >, SERVER PORT => 8 0, REMOTE ADDR => , SERVER PROTOCOL => HTTP/ 1. 0, PATH => / u s r / l o c a l / b i n : / u s r / b i n : / bin, REQUEST URI => / c g i b i n / t e s t. c g i? param=v a l u e, GATEWAY INTERFACE => CGI / 1. 1, SERVER ADDR => , DOCUMENT ROOT => / v a r /www/ de. l a n d e s f e i n d / htdocs, HTTP HOST => l a n d e s f e i n d. de }; Modules/Packages to parse are available in all languages: $ GET and $ POST arrays in PHP, CGI.pm for Perl,...

8 Pros and cons pros The simple unified interface allows the development of software in every available programing language. Create dynamic websites and process data online. cons A new process has to be created for every request which is time and CPU consuming. Executed software may harm the system when executed with wrong permissions ( suexec). a first approach: modules Load the interpreter at the start of the webserver (e.g. mod perl, mod phyton, mod php within Apache) and process the request without executing the external programm. malicious code may infiltrate or crash the webserver software applications with high initialization costs still need a lot of time

9 Scheme of a fcgi-communication Skip the loading time of interpreter and initialization by starting the script once and use it for several requests: Webserver communicates with an already running process via Unix-Domain-Sockets or TCP. GET /cgi-bin/foo.cgi HTTP/1.1 Host: HTTP/ OK Content-Type: text/html <html>...</html> fastcgi.com

10 Using FastCGI Modules for many programming languages are available 1 for transparent handling of the requests: $ENV = { SCRIPT NAME => / c g i b i n / t e s t. f c g i, SERVER NAME => l a n d e s f e i n d. de, SERVER ADMIN => webmaster@landesfeind. de, HTTP CONNECTION => Keep Alive, REQUEST METHOD => GET, HTTP ACCEPT => /, SCRIPT FILENAME => / v a r /www/ de. l a n d e s f e i n d / c g i b i n / t e s t. f c g i, SERVER SOFTWARE => Apache / , QUERY STRING => param=value, REMOTE PORT => , HTTP USER AGENT => Wget /1.12 ( linux gnu ), SERVER SIGNATURE => < address>apache / </address >, SERVER PORT => 8 0, REMOTE ADDR => , SERVER PROTOCOL => HTTP/ 1. 0, PATH => / u s r / l o c a l / b i n : / u s r / b i n : / bin, REQUEST URI => / c g i b i n / t e s t. f c g i? param=v a l u e, GATEWAY INTERFACE => CGI / 1. 1, SERVER ADDR => , DOCUMENT ROOT => / v a r /www/ de. l a n d e s f e i n d / htdocs, HTTP HOST => l a n d e s f e i n d. de FCGI ROLE => RESPONDER, }; 1

11 Using FastCGI A simple application dumping all environment variables and a counter variable: #! / u s r / b i n / p e r l use Data : : Dumper ; my $ c o u n t e r = 0 ; p r i n t ( Content t y p e : t e x t / p l a i n \ r \n\ r \n ) ; p r i n t ( Request $ c o u n t e r \n\n ) ; p r i n t ( Dumper(\%ENV) ) ; $ c o u n t e r ++; The same application using FastCGI: #! / u s r / b i n / p e r l use CGI : : F a s t ; use Data : : Dumper ; my $ c o u n t e r = 0 ; w h i l e ( CGI : : Fast >new ( ) ) { p r i n t ( Content t y p e : t e x t / p l a i n \ r \n\ r \n ) ; p r i n t ( Request $ c o u n t e r \n\n ) ; p r i n t ( Dumper(\%ENV) ) ; $ c o u n t e r ++; } The CGI script will allways print 0 as counter variable while the FastCGI script is stateful and will increment the counter.

12 Possibilities using FastCGI one time initialization Not only the process is executed once, also the initialization (e.g. loading a framework) has to be done once. asynchronous communication FastCGI data header contains a request ID which may be used for asynchronous communication between webserver software and CGI software. Note: Need a asynchronous working webserver (not Apache). load balancing Webserver may send CPU intensive request to more than one compute server processing the request.

13 Pitfalls webserver configuration The configuration of the webserver has to be changed and maybe additional modules have to be loaded. Note: Apache2 is very simple to prepare for fastcgi from script to application The process will not terminate and needs carbage collection. Filehandles have to be closed carefully. Database handles Connections to databases have to be connected and disconnect or need to follow the Singleton design pattern. Only few database modules support connection sharing after a fork().

FastCGI allows implemention of initialization intensive processes. implementing FastCGI applications There is no overhead implementing FastCGI applications.

14 FastCGI allows implemention of initialization intensive processes. implementing FastCGI applications There is no overhead implementing FastCGI applications. Existing cgi scripts may easily be upgraded using provided modules. community While CGI is widely used FastCGI has a small but very active community:

1. When will an IP process drop a datagram? 2. When will an IP process fragment a datagram? 3. When will a TCP process drop a segment?

Questions 1. When will an IP process drop a datagram? 2. When will an IP process fragment a datagram? 3. When will a TCP process drop a segment? 4. When will a TCP process resend a segment? CP476 Internet