A Quick Discussion of Active and Passive FTP Connections

I get a lot of questions about Active versus Passive FTP Connections, specifically when people are configuring their FTP firewall settings as described in my articles like Configuring FTP Firewall Settings in IIS and FTP Firewall Support, and I get related questions when people are trying to figure out why they can't use the command-line FTP.EXE utility that ships with Windows over the Internet. With all of this in mind, I thought that I would put together a quick blog that explains Active and Passive FTP connections and why those matter when you are connecting to an FTP server over the Internet.

Here is the briefest description that I can think of to describe the different between Active and Passive FTP:

• When you are using Active FTP, your FTP client determines how your data connections will work.
• When you are using Passive FTP, your FTP server determines how your data connections will work.

That sums up the whole concept into two bullet points. Unfortunately, neither of those bullet points are the least bit significant to you if you don't understand what they actually mean, so I will describe them in detail below.

How Active FTP Works

The following output shows the communication between an FTP client and FTP server using Active FTP to retrieve a simple directory listing, and I manually color-coded the output so that the client and server's responses would be a little easier to see:

OPEN ftp.contoso.com
Resolving ftp.contoso.com...
Connect socket #1920 to 169.254.10.1, port 21...
220 Microsoft FTP Service
HOST ftp.contoso.com
220 Host accepted.
USER robert
331 Password required for robert.
PASS **********
230 User logged in.
PWD
257 "/" is current directory.
PORT 169,254,10,2,72,50
200 PORT command successful.
LIST
125 Data connection already open; Transfer starting.
drwxrwxrwx 1 owner group 0 Feb 15 19:26 aspnet_client
-rwxrwxrwx 1 owner group 689 Jan 31 22:27 default.htm
226 Transfer complete.
Transferred 106 bytes in 0.008 seconds
QUIT
221 Goodbye.

In the beginning of this exchange, the client connects to the server from one of its ephemeral data ports to the server's port for the FTP command channel. After negotiating the FTP host, username, and password, the client retrieves the name of the current directory. So far all of the client/server communication has taken place over the FTP command channel, and up to now the conversation is identical to Passive FTP, but that is about to change.

The client's next task is to request a directory listing, which is denoted by the LIST command. The server will return its response to the client over the data channel, so before FTP client can send the LIST command, the client has to specify whether to use Active or Passive. In this example, the client has specified Active FTP by sending a PORT command. The syntax for this command is PORT A1,A2,A3,A4,P1,P2, where A1 through A4 are octets of the client's IPv4 address, and P1/P2 are two bytes that make up a 16-bit (0-65535) port address on the client. (Note: if you are using IPv6, there is a similar EPRT command that works with IPv6 addresses.)

Here's what the information in the PORT command means: the FTP client is essentially telling the FTP server, "For the upcoming data transfer, you need to talk to me at this IP address on this port." This means that the FTP client is actively in control of how the subsequent data communication is going to take place.

If we analyze this information, you can easily see why Active FTP will often fail to work over the Internet. As a relevant example, if you were to use the FTP.EXE client that ships with Windows, it can only use Active FTP. So when a client computer requests something from the server that needs to use the data channel, the client computer sends its IP address via a PORT command. If the FTP client is behind a firewall or NAT server, then the client is going to send its internal, LAN-based address, to which the FTP server will more than likely fail to connect. For example, if you are on a LAN that uses a NAT server and you have a 192.168.0.nnn IPv4 address, that IP address is invalid over the Internet, so the server will never be able to establish a data connection to your client to send the data. (Note: This is the reason why many customers contact me with the following problem description: "I can use FTP.EXE to connect to my server, and everything works until I try to retrieve a directory listing, then it hangs until the connection times out." What is actually happening is the FTP server is trying to connect to the FTP client's IP address and port that were specified by the PORT command, but the connection does not succeed because the server cannot connect to the private IP address of the client.)

How Passive FTP Works

The following output shows the communication between an FTP client and FTP server using Passive FTP to retrieve the same directory listing as my previous example, and once again I manually color-coded the output so that the client and server's responses would be a little easier to see:

OPEN ftp.contoso.com
Resolving ftp.contoso.com...
Connect socket #2076 to 169.254.10.1, port 21...
220 Microsoft FTP Service
HOST ftp.contoso.com
220 Host accepted.
USER robert
331 Password required for robert.
PASS **********
230 User logged in.
PWD 
257 "/" is current directory.
PASV
227 Entering Passive Mode (169,254,10,1,197,19).
LIST
Connect socket #2104 to 169.254.10.1, port 50451...
150 Opening ASCII mode data connection.
drwxrwxrwx 1 owner group 0 Feb 15 19:26 aspnet_client
-rwxrwxrwx 1 owner group 689 Jan 31 22:27 default.htm
226 Transfer complete.
Transferred 106 bytes in 0.008 seconds
QUIT
221 Goodbye.

As I mentioned in my earlier example, the beginning of this conversation is identical to Active FTP: the client connects from one of its ephemeral data ports to the server's port for the FTP command channel. After negotiating the FTP host, username, and password, the client retrieves the name of the current directory as in my earlier example - and here's where the difference begins.

Once again the client's next task is to request the directory listing, which is still denoted by the LIST command. But in this second example, the client has specified Passive FTP by sending a PASV command. The server responds to this command with a reply that is in the format of 227 Entering Passive Mode (A1,A2,A3,A4,P1,P2), where A1 through A4 are octets of the server's IPv4 address, and P1/P2 are two bytes that make up a 16-bit (0-65535) port address on the server. (Note: if you are using IPv6, there is a similar EPSV command that works with IPv6 addresses.)

Here's what the information in the response to the PASV command means: the FTP client is essentially telling the FTP server, "For the upcoming data transfer, you need to tell me which IP address and port I should use to talk to you." This means that the FTP client is passively allowing the server to control how the subsequent data communication is going to take place.

If we analyze this information, you can easily see why Passive FTP often works over the Internet; when the FTP server is in control of the communication parameters, it doesn't matter whether the FTP client is behind a NAT server, because the server is telling the client how it should communicate with the server.

All of this leads to an obvious question: what happens when both the server and the client are behind NAT servers or firewalls? This is where a little bit of configuration comes into play. If you read my Configuring FTP Firewall Settings in IIS and FTP Firewall Support articles, you would notice that you can configure the IIS FTP service to tell the FTP client which IP address to use; when your FTP server is located behind a firewall, you would configure your FTP server to send the external IP address of your firewall, and then you would configure your firewall to route FTP requests on that IP address to your FTP server. Since the FTP server is sending the external IP address of your firewall, the client knows how to communicate to the FTP server even though it is behind a firewall, even if your server is using a LAN-based internal IP address.

In Closing...

Having explained everything in my preceding examples, you should now understand what I meant earlier when I described the difference between Active and Passive FTP with these two simple points:

• When you are using Active FTP, your FTP client determines how your data connections will work.
• When you are using Passive FTP, your FTP server determines how your data connections will work.

I hope this clears up some questions you might have about Active versus Passive FTP, why you sometimes need to configure your firewall settings for your FTP service, and why the built-in FTP client for Windows seldom works over the Internet.

That wraps it up for today's blog post. ;-]