In cryptography, there are three goals: security, simplicity, and efficiency. You may choose two.
One-time pad is secure as long as the key is kept secure. Key generation is difficult. Transferring the key is difficult when the plaintext is large. If the key is stolen in transit, the security is all for nothing. Thus, one-time pad has some problems with both efficiency and security.
Then, in the 1970s came the ingenious invention of public key cryptography. This made it feasible to send encrypted text over public networks like the Internet. It is not simple, but it is both efficient and secure. Once it's understood, public key cryptography works fast. While the encryption isn't as secure as one-time pad, when one considers the hypothetical amount of computing time needed to attack the encryption, it is good enough.
The big downside of public key cryptography is that the sender and the recipient must keep their "secret keys" private and secure. They can generate them in isolation, but cannot allow them to be accessed by outsiders. Importantly, to avoid this pitfall the user of public key cryptography must understand the basics of public key cryptography. Once understood, key management is efficient. Yet key management is not simple. In the area of key management, the ordinary user must take security very seriously, a job that the ordinary user of computers does not ordinarily do or understand. This is the crux of the problem.
SSL uses public key cryptography. The secret keys are kept by a trusted third-party. As long as the third-party is worthy of trust, it works. The experience of most Internet users with trusted third-parties has been seeing a corporate name, and trusting them on that basis. Then, after they use SSL a few times, they learn from experience that trusted third-parties exist. In some situations, though, there are no trusted third parties.
PGP was brilliant. It put public key crypto tools in the hands of ordinary folks. The use of PGP has literally saved lives. It has a few drawbacks, however. The most notable is that the interface was initially limited to the command line, and even the later GUI interface was slightly clunky. Given time, money, talent, and will, the interface can be made better. Nevertheless, the basic problems of understanding and execution at the level of the ordinary user will remain.
There are various kludges that avoid the need to educate and train users. These involve a competent third person, a "system administrator" who handles key management behind the scenes. For example, a webmail host may use public key cryptography to provide secure e-mail to its users. Obviously, though, this scheme relies on trusting a third person, just like SSL, and sometimes you just can't trust any third party.
You can simplify the PGP interface and things will get slightly better. The main problem continues, however: How to educate and train ordinary users on the concept of public key crypto, and also on the software product. The paper linked to by K5 talks about the need for usability.
There is a difference between simplicity and simplification. Simplicity is fine. Simplication though would blind the user to the internal mechanics of the cryptosystem. Any decrease in complexity = an increase in the system's vulnerability. Designers should find out what threats the user can best take care of themselves. Then they should design the software to take care of all other (known) threats, exposing only certain threats to the user. Finally, the user needs to be educated to handle these vulnerabilities himself.
The user must learn the basics of the technology, and apply that knowledge daily. Some people want a cryptosystem where they don't have to think. Such a system won't be secure. The learning curve for public key crypto is steep, but there is no way around it. The user has to climb it or abandon public key crypto altogether. The user must be patient enough to manage keys competently every time, or a vulnerability will be introduced.
Even if a good design is found, though, one big problem remains. If there is not a threat that people can conceptualize, they will not act to minimize that threat. People put private letters in sealed envelopes because everyone can visualize that an unwanted person, be he a government agent or a random acquaintance, could otherwise read their mail (easily). People don't even consider that their e-mail is being intercepted every day. People think that Internet privacy issues concern only cookies and credit card numbers.
Therefore, the problem is not only educating and training ordinary users, it's also convincing them that a problem exists.
It is conceivable that we will someday dispense with all this and invent a crypto system that is very secure, very simple, and very efficient. Such a system will rely necessarily on a technology distinct from public key cryptography, however. No such technoloogy is known to exist.
Finally, where I can download this PGP 10? :-)