Given sources, fibres and detectors for use at 4 um, what are the engineering implications? The most striking, clearly, is the greatly increased stage length that appears to be possible. Some caution is needed here since even if fibre could, as suggested, be prepared with 10 kB/km transmission loss, there are other factors that would limit stage length in a practical communication system. For example, cabling losses will undoubtedly be encountered, as will be losses incurred at joints between lengths of cable and in coupling cable to source and detector. Even though the fourfold increase in waveguide size (compared with the situation at 1 um) will lead to reduced jointing and coupling losses, it seems unlikely that a repeaterless transatlantic cable system would be feasible.