The main idea of the paper is that by using a fingerprint scanner, the unique fingerprint of each finger can be identified, and commands or objects can then be associated with each finger as though each finger was carrying things with it.
The authors specifically compared their FUI (Fingerprint User Interface) with that of the Pick-and-Drop, a pen-based direct manipulation technique that allowed a user to virtually hold a data object in a pen. The advantages of FUI are that there is no special tools required (the fingers themselves would suffice), and there is no ID management since each finger already has a name. Also because a normal person has 10 fingers, it is possible to hold 10 objects/command, which is only achievable by 10 pens in the Pick-and-Drop system.
When associating fingers with commands, FUI can be great for operations where the user cannot look at the operation panel, for example, operating machines in darkness or manipulate them in a bag/pocket, because the user would not need to visually identify different buttons for different functions. FUI is also good when the user wants to conceal commands from others, such as opening the cash register when being robbed: one finger would simply open it while another finger would open it and also send a signal to the police.
Fingers can also be used as containers to store information. For example, text strings can be stored with multiple fingers, and when scanning the finger on another machine or in another application, the strings can be retrieved and pasted into the new machine/app. To the user, it felt just like that each finger is carrying very specific information.
It is important to note the drawbacks of such an interface. Firstly, scanning and identifying fingerprints can be very slow (about 2 seconds). The paper was written in 1998. At that time, it took 1.7 seconds to identify one fingerprint. More than 10 years later, technology today is not much better. The Dell XPS m1530 laptops have fingerprint scanner built in, which allows the user to associate different commands with different fingers. However, it still takes about 1.5 second for the application to identify the fingerprint. This delay can be very annoying to the end user.
Secondly, many times, the end user would forget or get confused about which finger carries what information/command. Thirdly, FUI would not work well for one hand operations when that hand has to hold the device (such as a mobile phone for a user who is driving). Fourthly, the fingerprint identification algorithm can make mistakes. And when that happens, it is important that the user can invoke undo operations to recover. And for things that cannot be undone (such as the lunch of a nuclear missile), this technology better not be used.
Some of the example applications presented in the paper are: operating a CD player where a finger could be "play" while another finger be "fast forward"; using fingers to store bookmarks for web browsing. An interesting application suggested by the authors is that a phone number can be stored with a finger at home, and then at public phones, the number can be retrieved from the finger to automatically dial that number. While this idea is cool, I personally wouldn't think it's a good idea because that would have to allow the public phone to have access to people's fingerprint data together with access to user's data repository (whether the home computer or an online storage), which exposes the user's private information to possible unauthorized accesses.
The main advantages of fingerprints are: they are unique to each user, and the user normally don't lose them (as opposed to keys). However, they are also unchangeable, so once others get hold such information, they can impersonate the user for the rest of his life.
If you are sleepy but want to stay awake, do push-ups. At least you'll be stronger.