specifically, they are one of the coolest features in postgres. Have you ever wanted to pretend that a CSV (or your filesystem or google) or basically anything else was a SQL database? With a couple hundred lines of code, you can do this. The performance won't be great, but it's ridiculously flexible.
