Alice and Bob each know their own private key and the other's public key, they can communicate securely, through any number of public key based protocols such as IPSec, PGP, S/MIME, or SSL. However, how do they know each other's public keys? The goal of a public key infrastructure (PKI) is to enable secure, convenient, and efficient discovery of public keys. It should be applicable within as well as between organizations, and scalable to support the Internet. There are various types of PKI that are widely deployed or have been proposed. They differ in the configuration information required, trust rules, and flexibility. There are standards such as X.509 and PKIX, but these are sufficiently flexible so that almost any model of PKI can be supported. We describe several types of PKI and discuss the advantages and disadvantages of each. We argue against several popular and widely deployed models as being insecure, unscalable, or overly inconvenient. We also recommend a particular model.