What is the difference between "tail -f" and "tail -F"?
You describe the GNU tail utility. The difference between these two flags is that if I open a file, a log file for example, like this:
$ tail -f /var/log/messages
... and if the log rotation facility on my machine decides to rotate that log file while I'm watching messages being written to it ("rotate" means delete or move to another location etc.), the output that I see will just stop.
If I open the file with tail like this:
$ tail -F /var/log/messages
... and again, the file is rotated, the output would continue to flow in my console because tail would reopen the file as soon as it became available again, i.e. when the program(s) writing to the log started writing to the new /var/log/messages.
On the free BSD systems, there is no -F option, but tail -f will behave like tail -F does on GNU systems, with the difference that you get the message
tail: file has been replaced, reopening.
in the output when the file you're monitoring disappears and reappears.
YOU CAN TEST THIS
In one shell session, do
$ cat >myfile
That will now wait for you to type stuff. Just go ahead and type some gibberish, a few lines. It will all be saved into the file myfile.
In another shell session (maybe in another terminal, without interrupting the cat):
$ tail -f myfile
This will show the (end of the) contents of myfile in the console. If you go back to the first shell session and type something more, that output will immediately be shown by tail in the second shell session.
Now quit cat by pressing Ctrl+D, and remove the myfile file:
$ rm myfile
Then run the cat again:
$ cat >myfile
... and type something, a few lines.
With GNU tail, these lines will not show up in the second shell session (where tail -f is still running).
Repeat the exercise with tail -F and observe the difference.
Simplified, when you open a file, you will get the inode that contains some metadata of where exactly the file is located on your disk. Tail will then listen for changes to that file.
If you remove the file, and create a new one with the same name the filename will be the same but it's a different inode (and probably stored on a different place on your disk). tail -f fill not retry and load the new inode, tail -F will detect this.
The same effect will happen if you rename/move a file. If you for example follows /var/log/messages and logrotate rotates the log to /var/log/messages.1. tail with -f will still listen to the old inode that points to messages.1. tail with -F will realize this and read the new inode.