Earlier quoted context omitted.
If you add a listener for 'data', or call stream.resume() then it'll start flowing data as it receives it, just like v0.8 and before. The only caveat is that listening for 'end' will not automatically start the flow of data, so you do sometimes have to either call resume() or add a data listener. In practice, it's rare that you care about the end of the stream, but not about the data coming through it, except in test…
Thanks. Is there any significant performance difference? (I'm wondering if it makes sense to roll my own "proxy" converting read()s to emit()s.)
No, it does not make sense to roll your own proxy turning read() calls into emits. You probably won't get it to go faster than it does now. Whether you use on('data') or read(), you get equivalent performance.