Stump's three criteria for a distributed system has a partial-truth in it that might confuse people (and perhaps he himself is confused.) A distributed system can work just fine if there is a master, as long as the system does not rely upon that node always remaining the master and/or has no way to elect a new master if the old master can no longer connect to a quorum of the nodes in the system. Whenever you see nodes described as the "master" this is usually a strong indicator that the system is not really a distributed system, but you have to dig a bit deeper to make sure.

"Distributed systems" is not a new term. Please don't use it as though it only applies to systems designed to distribute data (ala NoSQL). Distributed computing has classically been focused on the use of network nodes for both data storage and computation. See: MPI, Linda, etc.