So far, we’ve discussed how to use the OpenSSL programmatic interface for securing arbitrary TCP/IP connections using SSL. While SSL is a great general-purpose protocol, there are situations in which it is not appropriate. For example, SSL can’t be used to store encrypted data, such as on a disk or in a cookie, nor can it encrypt UDP traffic. In these cases, you should use the OpenSSL API for symmetric cryptography.
As you have probably noticed, we’ve been careful to recommend using SSL instead of raw cryptographic primitives for securing your applications if at all appropriate. We do this because it is incredibly easy to apply cryptographic primitives in a way that is insecure. Even professional cryptographic protocol designers have a hard time writing “secure” cryptographic protocols built on these primitives, which is one reason peer-review is so important in the world of cryptography.
If you’re planning to use this chapter to do real work, then we assume that you have some sort of need that SSL cannot fill, such as long-term data storage. We recognize that many people will want to design their own network protocols despite our recommendations. If you are considering such an option, we strongly urge you to prefer well-respected protocols, and even pre-existing implementations of those protocols, if possible. Nonetheless, this chapter is a reference for the basic API, and it is your responsibility to use that API in a secure manner.