Does Identity-Based Encryption actually solve any problem?

Identity based encryption schemes [*] seem to have great potential in high-latency Delay-Tolerant and mobile, ad-hoc networks since they apparently seem to avoid the need for key negotiation and exchanges. Used soley for key exchange, they allow any pair of members in the same group (however you choose to define it) to establish a unique pairwise secret that can be computed by each member up-front and as a function of their respective identities.

However, since each pairing based scheme - that I'm aware of - is predicated on a pre-distribution of some shared secret (e.g., the hashed ID of the agent raised to some secret integer), do any of these schemes offer any significant advantages over, let's say, a simple authenticated diffie-hellman (at least for purposes of symmetric key establishment)?

[*] Including pairing-based symmetric-key establishment schemes.

-

 But where is the difference between Alice needing Bob's DH key or Bob's certificate versus needing (in an authenticated way) Bob's ID? I only see that the ID might be shorter and thus is easier written down on paper (and typed from this). – Paŭlo Ebermann♦ Jun 18 '12 at 19:53 The difference is that Alice already has Bob's ID (in this case, "Bob"). $\:$ – Ricky Demer Jun 18 '12 at 20:01 @PaŭloEbermann: when Alice sends the message to Bob, she needs identify how the message gets to Bob. For example, if she sends it via email, she needs Bob's email address (bob@foo.com). Since she has that already, that can be used as the ID. – poncho Jun 18 '12 at 20:25 @RickyDemer: How does Alice know that the person known to her as "Bob" is the same person as the one known to the IBE central (or whoever provides the private keys) as "Bob"? – Paŭlo Ebermann♦ Jun 18 '12 at 20:28 @poncho: So this only helps when the central can identify Bob by his address? – Paŭlo Ebermann♦ Jun 18 '12 at 20:30