Pure fork/join concurrency isn't a novel idea. I was doing some work on it in grad school. Without blocking channels, fork/join has some really nice properties,

Pure fork/join concurrency isn't a novel idea. I was doing some work on it in grad school. Without blocking channels, fork/join has some really nice properties, such as guaranteed deadlock freedom. However, I agree with you that there's no one-size-fits-all approach to concurrency. Sometimes you want long-lived tasks that communicate—the actor model, in other words—in which case fork/join doesn't buy you much. I do think that fork/join is frequently what you want, though.