COS 461, Princeton University Spring 2015
Contents |
In this assignment, you will implement a web proxy that passes requests and data between multiple web clients and web servers, concurrently. This will give you a chance to get to know one of the most popular application protocols on the Internet -- the Hypertext Transfer Protocol (HTTP) -- and give you an introduction to the Berkeley sockets API. When you're done with the assignment, you should be able to configure Firefox to use your personal proxy server as a web proxy.
The Hypertext Transfer Protocol (HTTP) is the protocol used for communication on this web: it defines how your web browser requests resources from a web server and how the server responds. For simplicity, in this assignment, we will be dealing only with version 1.0 of the HTTP protocol, defined in detail in RFC 1945. You may refer to that RFC while completing this assignment, but our instructions should be self-contained.
HTTP communications happen in the form of transactions; a transaction consists of a client sending a request to a server and then reading the response. Request and response messages share a common basic format:
The initial line and header lines are each followed by a "carriage-return line-feed" (\r\n) signifying the end-of-line.
For most common HTTP transactions, the protocol boils down to a relatively simple series of steps (important sections of RFC 1945 are in parenthesis):
It's fairly easy to see this process in action without using a web browser. From a Unix prompt, type:
telnet www.yahoo.com 80
This opens a TCP connection to the server at www.yahoo.com listening on port 80 (the default HTTP port). You should see something like this:
Trying 69.147.125.65... Connected to any-fp.wa1.b.yahoo.com. Escape character is '^]'.
type the following:
GET http://www.yahoo.com/ HTTP/1.0
and hit enter twice. You should see something like the following:
HTTP/1.0 200 OK Date: Tue, 16 Feb 2010 19:21:24 GMT (More HTTP headers...) Content-Type: text/html; charset=utf-8 <html><head> <title>Yahoo!</title> (More HTML follows)
There may be some additional pieces of header information as well- setting cookies, instructions to the browser or proxy on caching behavior, etc. What you are seeing is exactly what your web browser sees when it goes to the Yahoo home page: the HTTP status line, the header fields, and finally the HTTP message body- consisting of the HTML that your browser interprets to create a web page. You may notice here that the server responds with HTTP 1.1 even though you requested 1.0. Some web servers refuse to serve HTTP 1.0 content.
Ordinarily, HTTP is a client-server protocol. The client (usually your web browser) communicates directly with the server (the web server software). However, in some circumstances it may be useful to introduce an intermediate entity called a proxy. Conceptually, the proxy sits between the client and the server. In the simplest case, instead of sending requests directly to the server the client sends all its requests to the proxy. The proxy then opens a connection to the server, and passes on the client's request. The proxy receives the reply from the server, and then sends that reply back to the client. Notice that the proxy is essentially acting like both a HTTP client (to the remote server) and a HTTP server (to the initial client).
Why use a proxy? There are a few possible reasons:
References:
Your task is to build a web proxy capable of accepting HTTP requests, forwarding requests to remote (origin) servers, and returning response data to a client. The proxy MUST handle concurrent requests by forking a process for each new client request using the fork() system call. You will only be responsible for implementing the GET method. All other request methods received by the proxy should elicit a "Not Implemented" (501) error (see RFC 1945 section 9.5 - Server Error).
This assignment can be completed in either C or C++. It should compile and run (using g++) without errors or warnings from the FC 010 cluster, producing a binary called proxy that takes as its first argument a port to listen from. Don't use a hard-coded port number.
You shouldn't assume that your server will be running on a particular IP address, or that clients will be coming from a pre-determined IP.
When your proxy starts, the first thing that it will need to do is establish a socket connection that it can use to listen for incoming connections. Your proxy should listen on the port specified from the command line and wait for incoming client connections. Each new client request is accepted, and a new process is spawned using fork() to handle the request. To avoid overwhelming your server, you should not create more than a reasonable number of child processes (for this experiment, use at most 20), in which case your server should wait until one of its ongoing child processes exits before forking a new one to handle the new request.
Once a client has connected, the proxy should read data from the client and then check for a properly-formatted HTTP request -- but don't worry, we have provided you with libraries that parse the HTTP request lines and headers. Specifically, you will use our libraries to ensure that the proxy receives a request that contains a valid request line:
<METHOD> <URL> <HTTP VERSION>All other headers just need to be properly formatted:
<HEADER NAME>: <HEADER VALUE>In this assignment, client requests to the proxy must be in their absolute URI form (see RFC 1945, Section 5.1.2), e.g.,
GET http://www.cs.princeton.edu/index.html HTTP/1.0Your browser will send absolute URI if properly configured to explicitly use a proxy (as opposed to a transparent on-path proxies that some ISPs deploy, unbeknownst to their users). On the other form, your proxy should issue requests to the webserver properly specifying relative URLs, e.g.,
GET /index.html HTTP/1.0 Host: www.cs.princeton.eduAn invalid request from the client should be answered with an appropriate error code, i.e. "Bad Request" (400) or "Not Implemented" (501) for valid HTTP methods other than GET. Similarly, if headers are not properly formatted for parsing, your proxy should also generate a type-400 message.
We have provided a parsing library to do string parsing on the header of the request. This library is in proxy_parse.[c|h] in the skeleton code. The library can parse the request into a structure called ParsedRequest which has fields for things like the host name (domain name) and the port. It also parses the custom headers into a set of ParsedHeader structs which each contain a key for the header field name and value corresponding to the value to which the header is set. You can search for headers by the key or header field name and modify them. The library can also recompile the headers into a string given the information in the structs.
More details as well as an example of how to use the library is included in the library file, proxy_parse.h. This library can also be used to verify that the headers are in the correct format since the parsing functions return error codes if this is not the case.
Once the proxy receives a valid HTTP request, it will need to parse the requested URL. The proxy needs at least three pieces of information: the requested host, port, and path. See the URL (7) manual page for more info. You will need to parse the absolute URL specified in the given request line. You can use the parsing library to help you. If the hostname indicated in the absolute URL does not have a port specified, you should use the default HTTP port 80.
Once the proxy has parsed the URL, it can make a connection to the requested host (using the appropriate remote port, or the default of 80 if none is specified) and send the HTTP request for the appropriate resource. The proxy should always send the request in the relative URL + Host header format regardless of how the request was received from the client:
Accept from client:
GET http://www.princeton.edu/ HTTP/1.0Send to remote server:
GET / HTTP/1.0 Host: www.princeton.edu Connection: close (Additional client specified headers, if any...)Note that we always send HTTP/1.0 flags and a Connection: close header to the server, so that it will close the connection after its response is fully transmitted, as opposed to keeping open a persistent connection (as we learned in Recitation 2). So while you should pass the client headers you receive on to the server, you should make sure you replace any Connection header received from the client with one specifying close, as shown. To add new headers or modify existing ones, use the HTTP Request Parsing Library we provide.
After the response from the remote server is received, the proxy should send the response message as-is to the client via the appropriate socket. To be strict, the proxy would be required to ensure a Connection: close is present in the server's response to let the client decide if it should close it's end of the connection after receiving the response. However, checking this is not required in this assignment for the following reasons. First, a well-behaving server would respond with a Connection: close anyway given that we ensure that we sent the server a close token. Second, we configure Firefox to always send a Connection: close by setting keepalive to false. Finally, we wanted to simplify the assignment so you wouldn't have to parse the server response.
The following summarizes how status replies should be sent from the proxy to the client:
Run your client with the following command:
./proxy <port>, where port is the port number that the proxy should listen on. As a basic test of functionality, try requesting a page using telnet:
telnet localhost <port> Trying 127.0.0.1... Connected to localhost.localdomain (127.0.0.1). Escape character is '^]'. GET http://www.google.com/ HTTP/1.0
If your proxy is working correctly, the headers and HTML of the Google homepage should be displayed on your terminal screen. Notice here that we request the absolute URL (http://www.google.com/) instead of just the relative URL (/). A good sanity check of proxy behavior would be to compare the HTTP response (headers and body) obtained via your proxy with the response from a direct telnet connection to the remote server. Additionally, try requesting a page using telnet concurrently from two different shells.
For a slightly more complex test, you can configure Firefox to use your proxy server as its web proxy as follows:
You can find details for the Berkeley sockets library in the Unix man pages (most of them are in section 2) and in the Stevens Unix Network Programming book, particularly chapters 3 and 4. Other sections you may want to browse include the client-server example system in chapter 5 (you will need to write both client and server code for this assignment) and the name and address conversion functions in chapter 9. Please refer to the first precept
slides in order to review the
Socket Programming tutorial and the reference
solution given for Assignment 0 during precept 2 and on Piazza.
In addition to the Berkeley sockets library, there are some functions you will need to use for creating and managing multiple processes: fork and waitpid. These will be reviewed in precept 3.
References:
You should submit your completed proxy by the date posted on the course website to CS Dropbox. You will need to submit a tarball file containing the following:
Your tarball should be named ass1.tgz. The sample Makefile in the skeleton zip file we provide will make this tarball for you with the make tar command.
Your proxy will be graded out of twenty points, with the following criteria:
make on your assignment, it should compile without errors or warnings on the course VM and produce a binary named proxy. The first command line argument should be the port that the proxy will listen from.