MDI-QKD: Continuous- versus discrete-variables at metropolitan distances

In a comment, Xu, Curty, Qi, Qian, and Lo claimed that discrete-variable (DV) measurement device independent (MDI) quantum key distribution (QKD) would compete with its continuous-variable (CV) counterpart at metropolitan distances. Actually, Xu et al.'s analysis supports exactly the opposite by showing that the experimental rate of our CV protocol (achieved with practical room-temperature devices) remains one order of magnitude higher than their purely-numerical and over-optimistic extrapolation for qubits, based on nearly-ideal parameters and cryogenic detectors (unsuitable solutions for a realistic metropolitan network, which is expected to run on cheap room-temperature devices, potentially even mobile). The experimental rate of our protocol (expressed as bits per relay use) is confirmed to be two-three orders of magnitude higher than the rate of any realistic simulation of practical DV-MDI-QKD over short-medium distances. Of course this does not mean that DV-MDI-QKD networks should not be investigated or built, but increasing their rate is a non-trivial practical problem clearly beyond the analysis of Xu et al. Finally, in order to clarify the facts, we also refute a series of incorrect arguments against CV-MDI-QKD and, more generally, CV-QKD, which were made by Xu et al. with the goal of supporting their thesis.
Updated reply to Xu, Curty, Qi, Qian and Lo (arXiv:1506.04819), including a point-to-point rebuttal of their new "Appendix E: Addendum"