Make TCP connections
This exercise guides you through establishing a raw TCP connection between two
servers and manually sending an HTTP request to a web server, illustrating how
network protocols work in practice.
Legend
Parts of this exercise are annotated with the following icons:
-
A task you MUST perform to complete the exercise
-
Optional step that you may perform to make sure that
everything is working correctly, or to set up additional tools that are not required but can
help you, or optional extra exercises to explore the topic further
-
Advanced tips on how to go further (or
challenges!)
The end of the exercise
-
The architecture of the software you ran or
deployed during this exercise
-
Troubleshooting tips: how to fix common problems you might
encounter
Establish a bi-directional TCP connection
You need to find a partner for this part of the exercise, since the goal is to
establish a TCP connection between two of the servers you have set up for the
course.
Let’s call you and your partner Bob and Alice.
Alice will need to know the public IP address of Bob’s server. We will refer to
it as W.X.Y.Z.
Bob: listen for TCP clients
Bob should run the nc command on his server to listen for TCP connections on
port 3000:
$> nc -l 3000
More informationAt this point, nc will start listening on port 3000 and take over the
console, waiting for a TCP client to connect.
Alice: connect to Bob’s server
Alice should run the nc command on her server to connect to TCP port 3000 on
Bob’s server:
$> nc W.X.Y.Z 3000
More informationHere, nc is acting as a client, connecting to the listening nc process on
Bob’s server.
Bob and Alice: communicate!
Bob should type “Hello” and press Enter to send this text:
$> nc -l 3000
Hello
It should be immediately displayed in Alice’s terminal:
$> nc W.X.Y.Z 3000
Hello
Similarly, if Alice types “World” after that and presses Enter, it should
immediately appear in Bob’s terminal:
$> nc W.X.Y.Z 3000
Hello
World
$> nc -l 3000
Hello
World
You have a two-way raw TCP connection running.
Once you’re done, you can close the connection with Ctrl-C.
Talk dirty HTTP to Google
You can do the rest of this exercise individually.
Let’s do something that your browser does every day: an HTTP request.
Find out Google’s IP address (O.P.Q.R in this example) using the ping
command:
$> ping -c 1 google.com
PING google.com (O.P.Q.R) 56(84) bytes of data.
64 bytes from google.com (O.P.Q.R): icmp_seq=1 ttl=53 time=0.890 ms
...
Open a TCP connection to the Google IP address you found:
$> nc O.P.Q.R 80
Now, talk to this Google server, but not in English or French, in the HTTP
protocol. Type exactly the following text:
GET / HTTP/1.1
Host: www.google.com
Then press Enter twice and you should receive the HTML for Google’s home page.
TipBe sure to type exactly the text above. Your request must be a valid HTTP
request or Google’s server will not be able to interpret it correctly. If you
have made a mistake, exit with Ctrl-C, re-run the same nc command and try
again.
Once you’re done, you can close the connection with Ctrl-C.
If you open your browser, visit http://www.google.com and display the source
code of the page, you should see the same result.
More informationBy sending this text over the TCP connection, you are communicating in the HTTP
protocol, a text protocol: you are sending an HTTP request to retrieve (GET)
the resource at path / of host www.google.com (the landing page of the
Google website), using version 1.1 of the HTTP protocol.
What have I done?
Like the pioneers of the 1970s who developed TCP/IP, you connected two machines
across the Internet and exchanged a few (hopefully nice) words. Then you did by
hand what your browser does every time you open a page: you asked a web server
for it.
A TCP connection always starts the same way: one process listens on a port
and waits, and another connects to an address and that port. The one that
connects is the client, the one that listens is the server. Bob’s nc was the
server and Alice’s the client, yet both of you could type: once the connection
is open, it carries data in both directions.
To reach Bob’s nc, three things had to be right. Alice needed the public
address of Bob’s server, the one you connect to with SSH, and not the private
one that ip address prints on the server itself. Azure’s firewall had to let
port 3000 through, which is why you opened it when you created your server. And
a process had to be listening on that port: an open port with nothing behind it
is refused.
Talking to Google showed the layers working together, each relying on the one
below it:
- IP (layer 3) brought your packets to an address. That is why you first
turned the name
google.com into an address with ping.
- TCP (layer 4) gave you a connection to port 80 at that address, exactly
like the one between Bob and Alice.
nc does not know what a web server is.
- HTTP (layer 7) is what the browser and the web server say to each other
over that connection. Google answered because what you typed was a valid
request.
Your request had two lines. The first asked to GET the page at path /. The
second, Host, named the site you wanted, although you had connected to an
address: one address can serve many sites, so the request has to say which. The
empty line after them, your second Enter, told the server that the request
was complete. Its response had the same shape: a status line, headers, an empty
line, then the page itself.
You could type the request because the version you used, HTTP/1.1, is text over
TCP. Newer versions keep the same requests and responses but send them
differently: HTTP/2 encodes them in binary, and HTTP/3 does not use TCP at all.
Your browser uses them when the server supports them. The course uses HTTP/1.1
because you can read it, and what a request means has not changed.
Later in the course, your own application will listen on a port of your server
and your browser will connect to it, as Alice did to Bob. And when a single
server hosts several sites, the Host line is what decides which one answers.
Architecture
This is a simplified architecture of the main running processes and
communication flow during this exercise:
Troubleshooting
Here’s a few tips about some problems you may encounter during this exercise.
400 Bad Request
If you get a 400 Bad Request response, it means that your HTTP request is
invalid. You probably did not type exactly the text above.
I don’t get a response
If you don’t get a response, it may be because you took too long to type the
text, and the request has timed out. Try again a little faster.