Single Page Web Server

Sounds useless, doesn't it? This is a C implementation of a very simple HTTP/1.0 web server. Just for the fun of it.

On a side note, it's not finished yet. See TODO chapter.

Features

• Configuration is done by command-line parameters.
• Listens to a socket on a given port. Both IPv4 and IPv6 are supported.
• Only the method of the incoming HTTP request is parsed. Only GET is supported.
• Returns only a single HTML page as a response. Maximum file size is 65536 characters.

Requirements

• Linux (tested with kernel 3.10.0)
• GCC (tested with version 4.2.8)
• GNU make (tested with version 3.82)

Potential use cases

Some potential use cases:

• It's very fast to setup.
• It doesn't require any extra libraries.
• It doesn't require Ruby or Python.
• Could work as a sorry-server for a load-balancer. Remember, HTTP/1.0 could not care less about Host headers. When multiple sites are pointed to this web servers, all HTTP responses are the same, no matter what is the Host header.

Building the application

Source code is located in the src directory. Run make to build.

$ make
gcc -lpthread -Os -s -c -o obj/util.o src/util.c
gcc -lpthread -Os -s -c -o obj/server.o src/server.c
gcc -lpthread -Os -s -c -o obj/sockets.o src/sockets.c
gcc -lpthread -Os -s -c -o obj/http.o src/http.c
gcc -lpthread -Os -s -o bin/single-page-web-server obj/util.o obj/server.o obj/sockets.o obj/http.o

This creates two directories: obj for objects and bin for the resulting binary file.

If you want to clean the object files, run make clean.

$ make clean
rm -f obj/util.o obj/server.o obj/sockets.o obj/http.o

Running the application

Print out the help message.

$ bin/single-page-web-server -h
ingle Page Web Server
Options:
   -p <port>: TCP port number, which will be listened to. Defaults to port 80.
   -f <path>: Path to file, which will be served. Required.
   -6:        Use only IPv6 sockets. By default both IPv4 and IPv6 sockets are used.
   -d:        Enable debug logging.
   -h:        Display help.

An example run:

$ bin/single-page-web-server -p 8080 -f index.html

TODO

• An option run this as a daemon instead of as a foreground process.
• Instead of spawning a new thread when a HTTP request arrives, a thread pool of some sort could be used.
• It's probably not compliant with the HTTP/1.0 specification, even though it only accepts GET requests. I should read the RFC once again.
• Bug hunt. I'm pretty sure the protocol handling isn't perfect.
• Timeouts should be used. If a client opens up a TCP connection but doesn't send any requests, we shouldn't waste our threads on that.
• Add possibility to specify a listen address.

Development ideas

• The server could serve any kind of file instead of a HTML page.

• SSL support. Again, with the same simplicity. Perhaps the web server could generate a SSL key and a bogus certificate on start-up, if an -ssl command-line parameter is given. A new key and a new certificate would be generated when the application is restarted. Just a thought.