FWIW one reason the service mentioned in the blog was so fast was because of laziness. Because it didn't need most of the request, the work required to parse it was mostly avoided.