I have on more than one instance come across statements which suggest that zero-knowledge proofs are not practical. One example is: "Our theoretical constructions use zero-knowledge proofs, and therefore they do not suggest a practical way of implementing a pseudonym system." from the paper Pseudonym Systems.
What is the basis for this? Is it a caution which was common in the past which is now obsolete? I can think of recent systems, e.g. Z-cash, which uses zero-knowledge proofs in practice.