17:24 <ncm[m]> The notion of a "data diode" makes no more sense than for a real diode. The downstream can pass information up just by varying the timing of how fast it accepts input.

17:59 <zgrep> The idea that it "accepts" an input would be the communication back. I assume that such a system would be designed such that it's the writer's responsibility to not send data too fast or too slow.

17:59 <zgrep> (And with enough redundancy to recover from errors.)