I would have liked to have seen Martin Fowler mentioned the parallelism and state implication of the unix pipeline, somewhat limited (see http://www.quora.com/P

I would have liked to have seen Martin Fowler mentioned the parallelism and state implication of the unix pipeline, somewhat limited (see http://www.quora.com/Parallel-Computing/Do-pipelined-unix-commands-run-faster-on-multicore http://www.quora.com/Parallel-Computing/Do-pipelined-unix-co...) but this is for me a very profond way of looking at this kind of pattern: whether you finish completely your processing on the data structure or you emit the result downstream as soon as you have finished processing one element. Stream is for me the fundamental nature of the unix pipeline, and this is rarely met in programming language (channels to the rescue like in Clojure and Go). The Java Streams seems to work this way with the distinction between stateless and stateful processing element (http://docs.oracle.com/javase/8/docs/api/java/util/stream/package-summary.html http://docs.oracle.com/javase/8/docs/api/java/util/stream/pa...).