WEBSERV

What is a Web Server?

A web server is software that understands URLs and the HTTP protocol. It stores content and delivers it depending on the user's request as an HTTP response. For example, the content could be anything from images and videos to HTML pages and JSON files.

Why did we build a Web Server?

Building a Web Server is the foundation for understanding web technologies. It's a great tool to dive deeper into how networking in browsers works and see what is happening behind the scenes. It broadens our perspectives and skills and enhances our educational values through a demystifying complex system process. In the end, we achieve a better knowledge of HTTP protocol, sockets, request and response cycles, software architecture and so on.

Installation Process

In your terminal run:

cd ~/Desktop
git clone https://github.com/Sepulven/webserv.git
cd webserv
make
./webserv config.yml

Warning

We do not advise you to run it in any virtual terminal. For example, VS Code's terminal may create a zombie child process when you finish executing, which blocks the port. In case that happens, you need to find the process that uses the PORT and kills it through its PID.

In your terminal run:

lsof -i :[PORT]
kill -9 [PID]

Caution

When testing it, use Firefox as a browser. Otherwise, you will encounter unexpected behaviors.

In your terminal run:

firefox 127.0.0.1:8080/

Development Environment

Operating System:

**Unix/Linux**
Distributor ID:	Ubuntu
Description:	Ubuntu 22.04.4 LTS
Release:	22.04
Codename:	jammy

Compiler:

> c++ (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0

Browser:

> Mozilla Firefox 127.0.2

How to use it

Create the executable file as explained before and create a config file following the rules of the config file. It will server locally all of the files requested and execute all of the defined scripts (nodejs, bash, php, python, ..., etc.).

Understanding the config file

The config file is defined in a YAML-like file. Its structure is very similar to YAML except for the fact we can have multiple definitions of the same variable, in this case, the variables are going to be stored as an array.

For example:

server:
	[...]
server:
	[...]

Thus we can have multiple definitions of servers listening to different ports and multiple definitions of routes. We must have at least the / route definition which is in-build in the main scope, it cannot be defined as any regular route.

Definitions

• listen:
  • host : string | number The host can only be either localhost or 127.0.0.1
  • port: number Any avaible port, the server with throw an error in case the port is busy.
• server_name: string Names the server block.
• maxcbsize: string Megabyte or bytes. Sets the max. size of the content-body(http protocol).
• max_conn: number 1 is enough as the max. connections. C++ is fast to deal with thousands of incoming request flawlessly.
• error_pages: Sets up the path to the http error pages.
  • number : path Error code followed by the path of the file that is going to be served.
• cgi: Defines all of the cgis.
  • string : path Extension followed by the path to the interpreter.
• root: path The root to the / route.
• index: path The index for the / route.
• http_methods: list of strings Allowed methods for the / route.
• route path: Defines a route except the / route.
  • root: path Defines root.
  • redirect: path Defines the redirect route for when the http method completes.
  • index: path Defines the index file.
  • dir_listing: on | off Directory listing.
  • http_methods: list of strings Defines the allowed methods.

If any rule is broken or a logic gap is presetended, the server won't run.

Errors such as: - Bad indention; - Wrong attr. definition, like attr. name and value; - Non-existing file path; In case no error pages is provided for an error code, a standart html page with the error code is created.

Tip

The config file accepts comments. They must begin with # followed by the comment. # This is comment!

server:
    listen:                     # mandatory
        host: localhost   
        port: 8080
    server_name: main.server.com
    max_cbsize: 2000m		#mandatory
    max_conn: 1			#mandatory

    # / route definition, mandatory
    root: /www   # * Root definition 
    index: pages/index.html
    http_methods: GET POST

    error_pages:   # * Error pages map;
        403: error/403.html
        400: error/400.html
        404: error/404.asf    # works with every type of file
        404: error/404.html

    cgi:   # If not defined, no script can be interpreted
        .py: /usr/bin/python3
        .php: /usr/bin/php

    # * Server routing;
    route /main: # AN EXAMPLE OF A REGULAR ROUTE
        root: /www/pages
        redirect: /up
        index: index.html
        dir_listing: off
        http_methods: GET POST 
    route /up: # ROUTES THAT DEALS WITH FILES;
        root: /www/uploads
        dir_listing: on
        http_methods: GET POST DELETE

File structure

The whole project is structured into pieces for each piece you have the declaration and the implementation, namely a .h and .cpp respectively. Each feature is a folder defined, and all of its extra dependencies go into the __utils folder.

Every utils file starts with ___ and ends with utils followed by its extension for better identification.

Note

If you find any issues, or have question, or have suggestions for improvements feel free to open an issue or send an email.

If I want to build a webserv in C++98 how should I begin?

First, understanding the basis of a client and a server is a good start. Next, how the HTTP request and response cycle works. After that, how file descriptors can be used as web sockets. Then, how to syncronize fd's events with epoll or any of its variants. Finally, use all of the tools C++98 has to give to manage the requests and responses, you're going to see that it is pretty much managing I/O operation with file descriptors.

Test performance and availability with Siege

Siege is an http load tester and benchmarking utility.

Install siege with:

sudo apt install siege

Siege Manual

Main Tests

Evaluation sheet

How it works?

Overview

Single threaded
CGI
Parsing
Event handling with epoll

[... here comes the whole explanation]

References

• Epoll Performance
• How NodeJs works
• Internal Architecture of NodeJs

Made by:

@ratavare - Rafael Tavares
@terezamr - Tereza Ribeiro
@sepulven - Arthur Aguiar