Every node should be able to do the whole job
Quote
each one of the piping servers needs to be able to fully service a request from start to finish all the way through. And that way, there's no dependence on one on one single point of failure.
Episode 35: Episode 35 01:10:41 (play audio from this timestamp, opens in a new tab)
Voice-matched
Summary
Podping's design puts one DNS name in front of a group of servers, run by different parties and reached round robin. Each server must service a request from start to finish on its own. With six or ten of them, any one can run everything, so there's no single point of failure, and publishers only need to know one address.