In "Computer Networks:A Top Down Approach" - can multiple applications using the same protocol use its designated port?

Recall also that when a client or server program implements a protocol defined by an RFC, it should use the well-known port number associated with the protocol

In modern practice, I would amend this statement to: well-known ports are used by servers of such standardized protocols to receive connections from clients.

Clients typically use ephemeral ports, not well-known ports, for the local ends of outgoing connections, meaning that the number of outgoing connections you can have is nearly unlimited: the only restriction is that multiple connections to the same remote IP and port must have different local port numbers, but there are many thousands of those available.

And clients that need to be able to accept incoming connections from servers or from peers typically listen on a random high port and then communicate that port number to others using some other mechanism (for instance, a connection made to the server on a well-known port, or STUN). This is both because of the conflict issue that you raise, and because of the prevalence of NAT in modern times.

There are exceptions to both of those rules, but exceptions to the first rule tend to be very old (DNS queries originating from port 53 and NTP queries originating from port 123 come to mind), and exceptions to the second rule, while more common, are decreasing with time. So that's why I say "modern practice".

But to my understanding a port number cannot be in use more than once. If I had two processes using the same protocol they would use the same designated port right?

Most protocols have a designated default port.

For instance, if you intend to do HTTP, you will use TCP port 80 by default. If you intend to do HTTPS, you will use port 443 by default. And so on.

But most protocols can use alternate ports.

For instance, you can include the port in HTTP or HTTPS URIs (http://www.example.org:1234/this or https://www.example.org:5678/that). This would tell the underlying protocol to use ports 1234 or 5678 instead of the defaults, respectively.

So you could have multiple HTTP servers listening on different ports. This is quite common in some environments (e.g. node processes behind a reverse proxy).

Likewise, a listening port can only have one attached process for a given IP address and IP protocol (an IP protocol is UDP or TCP, through there are others):

• You could have one application using UDP port 1234 and another using TCP port 1234.
• You could have one application listening on port 80 for IP 192.168.0.100 and another one listening on the same port 80 for IP 192.168.0.200.

What the book is probably trying to say, though, is that if you design your own protocol (the Poly-Incremental Neural-Knowledge protocol, known as PINK), you probably shouldn't use a port that is used by other existing protocols. For instance, don't use TCP ports 80, 443, 21, 22, 25, etc, because if you do, when some user wants to install your server on their box, and they already run an HTTP, HTTPS, FTP, SSH or STMP server on the same box, they will have a conflict: the server started last will refuse to run (unless the user switches the default port of one of the two, or uses different IP addresses for each).

If I had two processes using the same protocol they would use the same designated port right?

They might want to, but you can't do that. The server/listener side needs to be on a known transport-layer port and that cannot be shared usually.[*1] The client/initiator side very commonly uses an ephemeral port, so there's no problem there.

In the corner case where you have concurrent server processes that want to use the same port number for listening, you could bind multiple IP addresses to the network interface and bind the processes specifically to an IP:port combination (instead of bind to all IP addresses using the unspecified address 0.0.0.0).

Alternatively, you could use secondary/exotic port numbers and configure the clients to use those for destination.

[*1] Some operating systems do support port sharing, but that's rather random, so you'd only use it with multiple processes for the very same service, a kind of load balancing. Another option is to configure a reverse proxy specific to the application-layer protocol - off topic here.

For a given protocol, think of a port number as the number of a table entry that points to the process using that port for that protocol. There can only be one process associated with the protocol port.

For example, if you run a web server that listens on TCP port 80 (well-known HTTP port), you cannot then have a different web server also use the same TCP port 80.