In some ways EAP performs the role of an actor's agent. When an actor is looking for work, the agent takes her to a movie director and introduces them. The agent sits back while the actor and director talk about the job, but jumps in again at the end to close the deal.
EAP has a set of messages that it uses to make the introductions and to close the deal. These are used with all upper-layer authentication methods. EAP also allows two parties to exchange information that is specific to the authentication method they want to use. The content of these authentication-specific methods is not defined in EAP. In fact, they can be completely proprietary authentication methods or newly invented ones. EAP's ability to handle part of the communication in a standardized way and part in a specific way is the key to its extensibility. We refer to these authentication-specific messages as "middle messages" because they occur after the introduction and before the closing.
 Upper-layer authentication methods are discussed in Chapter 9 and include methods such as SSL, TLS, and Kerberos V5.
Quite a lot of these middle messages can be exchanged before the authentication is completed. The reason why EAP is extensible is that the details of these special messages are left to other RFCs to fill in. For example, there is an RFC saying how to use Transport Layer Security (TLS) over EAP; another (draft) says how to use Tunneled TLS (TTLS) over EAP, and so on. It also means that if you invent a new method later on, you can write a new draft called "mymethod over EAP"; and if it becomes popular, other people can implement it on existing systems.
RFC2284 (EAP) is a very short document as these things go. In fact, not counting references, acknowledgments, definitions, and so on, it is only nine pages long. RFC2284 (EAP) specifies that four types of message can be sent:
Request: Used to send messages from the authenticator to the supplicant
Response: Used to send messages from the supplicant to the authenticator
Success: Sent by the authenticator to indicate access is granted
Failure: Sent by the authenticator to indicate access is refused
Note that these messages are described here in terms of the authenticator. However, in the IEEE 802.1X scenario, the authenticator forwards the messages on to the authentication server, most likely using RADIUS. In this case it is the authentication server that generates request, success, and/or failure messages and the authenticator just relays them to the supplicant.
Request and response messages are further subdivided using the EAP Type field. The Type field indicates what information is being carried in the EAP message. The first six message types are defined in the standard; all the others are reserved for specific authentication methods. The most important predefined type is Identity (type value 1). Typically, this is used as part of the EAP introduction phase: the message EAP-Request/Identity is sent by the authenticator to a new supplicant. The supplicant replies with the message EAP-Response/Identity containing its user name or some other identifier that will be understood by the authentication server.
Type numbers higher than 6 are not defined by RFC2284 (EAP), but they are issued (uniquely) by IANA for each new authentication method that is introduced. Some are even issued for vendor-proprietary methods. The type number for TLS, for example, is 13, which means that all EAP-Request and EAP-Response messages with this type field contain information that is specific to the TLS upper-layer authentication method.
The use of the Type field is a bit inconsistent. For the most part, it indicates the authentication method. But in a few cases, it defines a special-purpose message. For example, a message with a type value of 2 is called a notification message and is used to send some user-displayable text. This could be anything from "Please enter your password" to "Prepare to meet thy maker"?it really doesn't matter. The message is intended to appear on the screen of the user's system (although few systems actually support this). A message with a type value of 3 is called a NAK and is used when a request is made for an authentication method that is not supported. If an EAP request with type TLS is sent to a peer that doesn't support TLS, it can respond with a Type field of NAK.
Type value 1 Identity could be considered a special-purpose message or it could be considered a very simple authentication method. Under IEEE 802.1X, this request is often the first thing sent and the supplicant will reply with a response message giving its identity information. Originally this was treated as a special type to be used prior to the main authentication phase. However, this has been subtly changed in the revised EAP draft (while remaining compatible with the previous version). The simplest authentication exchange could go:
EAP-Identity request (from authenticator)
EAP-Identity response (from supplicant)
EAP-Success (from authenticator)
Here the device has been "authenticated" on pure trust: "I choose to believe that you are who you say with no proof." Or perhaps proof is available by some other means. For example, the identity might be generated by a smart card that changes every second, synchronized to the authentication server. This type of null authentication can be used with simple wireless LAN networks that have preloaded secret keys (called preshared keys) and then rely on the encryption to prevent unwanted communications.
 This is often referred to as a one-time password.
Because the EAP-Identity exchange can be considered a complete authentication method by itself, when you do the identity exchange followed by another method such as TLS, you are really running two authentication methods in sequence. This concept of serial authentication has been generalized in the new EAP draft, which simply lists the EAP-Identity message as a basic authentication method and then says that you are allowed to run as many authentication methods in sequence as you wish prior to the final EAP-Success or EAP-Failure message.
This ability to run multiple authentication methods in sequence can be exploited in new approaches that allow the client to authenticate the network before revealing its identity. One approach, PEAP (Protected EAP), is discussed in more detail in Chapter 9.
All EAP messages have a similar basic format, as shown in Figure 8.7. Code is one byte indicating the type of message:
Identifier is a value in the range 0?255 and IEEE 802.1X indicates that it should be incremented for each message sent. When a response is sent, the identifier is set equal to that in the request. This helps for checking which response goes with which request. Length is the total number of bytes in the EAP message (including Code and so on). It is a 16-bit value. Finally, Data is the actual request or response data being sent.
We have already discussed the Success and Failure packets. These messages are short and contain no data. One of these messages is used at the end of the authentication process to signal the result. Because the Success and Failure are common across all authentication protocols, intermediate devices (such as the access point) can detect when an authentication completes without understanding all the details of the authentication method. The access point should wait for the RADIUS Accept message before making any decision about access rights.
The details of the authentication method are sent in the request and response messages. These have an extra field called Type. The format of an EAP-Request or EAP-Response message is shown in Figure 8.8.
You can see the Type field, which is used to identify the request or response. The Type field is essential to separate all the different authentication methods. In fact, it is the key to the extensibility of EAP. Each new authentication method is assigned a unique value so the system knows whether the request contains information relevant to, for example, TLS or PEAP.
 These authentication methods are described in Chapter 9.