nullswan / webserv Goto Github PK

Handle Location blocks and their attributes.

Restrict support to HTTP/1.1 or HTTP/1.0 ONLY.
In any other cases, we must return a 505 error.
https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/505

/!\ In HTTP 1.0, the connection is closed by default.
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Connection

Enhance Client timeout checker.
In case our server is under high load, we must still trigger garbage collection.
This will be also used to clear any CGI process that are running for too long.

View: https://developer.mozilla.org/en-US/docs/Web/HTTP/Methods/POST

check for cgi,
check for upload,
otherwise, send a 403 (Forbidden) https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/403

#7, #15 Dependent

View: https://developer.mozilla.org/en-US/docs/Web/HTTP/Methods/DELETE

resolve_uri => using location

if ( stat(file) )
    if (delete())
        return 204
   if (not allowed())
      return 403
    if (error())
       return 500
if (not found)
   return 404

Depend of #7

Setup default settings when no argument (ie. no config file) is passed.

Default config eventually expose a /ping GET endpoint, otherwise anything will be 404.
(Default autoindex off)

• Before using non-blocking file descriptors, recv() function was waiting for the end of the request message.
• Now, fds are set to non-blocking operations: the goal is to be able to read a full request with multiple call to recv().
• To solve the issue, when the server receive a new input on an active socket, it has to create a new Request object if client input is a request-line (i.e. GET / HTTP/1.1), but after that, it should parse headers and optional body (if post request) to the existing request object.
• Tricky point: handling both application/x-www-form-urlencoded and multipart/form-data;boundary="boundary"

Handle chuncked_request using special body reader.

Send a 408 error page to timed out clients.
https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/408

https://tryhexadecimal.com/guides/http/431-request-header-fields-too-large

Handle timed out client by destroying their Webserv::Http::Client object.

https://developer.mozilla.org/fr/docs/Web/HTTP/Status/414
https://stackoverflow.com/questions/1067334/how-to-set-the-allowed-url-length-for-a-nginx-request-error-code-414-uri-too

Process GET CGI

• ISO C++ doesn't allow to have a type name following an enum declaration
  

Setup following rules:

server {
	server_name (IServer._name<std::string>)
	listen		(IServer._host<std::string>):(IServer._port<int>);

	// or

	listen		:(IServer._port<int>);

	// or

	listen		(IServer._port<int>);

	// or

	listen		(IServer._host<std::string>);	// port 80 is used

	/* 
		If listen is not defined, *:80 is used 
		if webserv runs with the superuser privileges, 
		or *:8000 otherwise. 
	*/
}

Handle CGI system.

This must be asynchronous.
In case a CGI is too long, it must be killed by our garbage collector, triggered when epoll_wait contain no events, view #14 .

Define how we await CGI.

Return error if file is empty

Make logger uri placeholder longer to prevent overlap.

[WEBSERV] 2021/12/01 - 13:54:49 | GET     / | 404 |    63 μs | 127.0.0.1 -> 0.0.0.0
[WEBSERV] 2021/12/01 - 13:54:49 | GET     /favicon.ico | 404 |    34 μs | 127.0.0.1 -> 0.0.0.0

It should be aligned, ie:

[WEBSERV] 2021/12/01 - 13:54:49 | GET     /                             | 404 |    63 μs | 127.0.0.1 -> 0.0.0.0
[WEBSERV] 2021/12/01 - 13:54:49 | GET     /favicon.ico            | 404 |    34 μs | 127.0.0.1 -> 0.0.0.0

View: https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Host

The host header is not optional and must appear in any HTTP/1.1 request. If it is missing or contains multiple Header fields, we should return 400 (Bad Request).

• Managing getline() eof;
• exit and quit command correction.

https://developer.mozilla.org/en-US/docs/Web/HTTP/Methods/GET

resolve_uri => use locations

( is_cgi() )
   handle_cgi()

// direct file serv
( stat() == -1 )
404 not found,  500 server error

if (path_is_directory)
try config.index resolve, catch 500 (Server Error)
try autoindex, catch 500 (Server Error)

=> 404

Handle server block with their attributes.

• server_name
• host
• port
• index
• error_pages

In case we have a duplicate host:port, only the first one will be used.

Raise error when attempting to configure a nested location.

Call a specific method invoker depending on the HTTP method provided within the request.
Otherwise, return a 501 error if none match.
https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/501

if (GET)
=> GET()
if (POST)
=> POST()
if (DELETE)
=> DELETE()
return (501, NOT IMPLEMENTED)

• In the request-line (first line in request), apart from the method which is case sensitive and has to be in uppercase, the URI and http version are case insensitive: they are converted to lowercase.
• Headers are case insensitive too, and we should convert them to lowercase when we parse them.
• Case variations should be kept in the payload!

Resolve $query, replacing them with real value and calling the correct server by returning 301 (Moved Permanently: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/301)

Pre-generate any error pages, load from file if supplied in configuration file.

https://github.com/c3b5aw/webserv/blob/ab947a3907eb766511ee6a48876eb24096726ea2/srcs/models/IServer.hpp#L132

Solver must split at ':' if any, meaning there is a port and attempt to match the extract port with a matching vhost if any has been found.

Response must be sent with some header any time:

• Server: https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Server
• Date: https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Date
• Content-Length: https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Content-Type
• Content-Type: https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Content-Type
• (Eventually) Connection: https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Connection

Setup a upload pass member that allowed file upload and root them to this specific directory.

Handle virtuel hosts when merging servers.
If one vHost exist, we must try to match before using default server host configuration.

• Handling different invalid requests: bad request, wrong http version, not implemented, invalid multipart request, payload too large, etc..

Setup Webserv::Http::Request object lifetime.

This object should be instanced while reading request in Webserv::Http::Client.
He must contains several members referring to her buffer parsing result.
Along with several methods, is this a closed, chuncked request ?

Handle IBLock and their attributes.

• autoindex
• root
• redirect
• body_limit
• allowed_methods
• cgi_pass

Build an autoindex generator.

Used by: #13

Handle body size being too large.

Client usually send Content-Length header withing their payload. (https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Content-Length)
We can use this header to prevent full reading of a payload.

In case the payload size is too large, we can return a 413 error. https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/413

Define and handle behavior when there is a duplication for a specific host:port key.