NON-AUTOMATIC WATER-TO-CARBIDE GENERATORS.--The only principle underlying the satisfactory design of a non-automatic water-to-carbide generator is to ensure the presence of water in excess at the spot where decomposition is taking place. This may be effected by employing what is known as the "flooded-compartment" system of construction, _i.e._, by subdividing the total carbide charge into numerous compartments arranged either vertically or horizontally, and admitting the water in interrupted quantities, each more than sufficient thoroughly to decompose and saturate the contents of one compartment, rather than in a slow, steady stream. It would be quite easy to manage this without adopting any mechanism of a moving kind, for the water might be stored in a tank kept full by means of a ball-valve, and admitted to an intermediate reservoir in a slow, continuous current, the reservoir being fitted with an inverted syphon, on the "Tantalus-cup" principle, so that it should first fill itself up, and then suddenly empty into the pipe leading to the carbide container. Without this refinement, however, a water-to-carbide generator, with subdivided charge, behaves satisfactorily as long as each separate charge of carbide is so small that the heat evolved on its decomposition can be conducted away from the solid through the water- jacketed walls of the vessel, or as the latent heat of steam, with sufficient rapidity. Still it must be remembered that a water-to-carbide generator, with subdivided charge, does not belong to the flooded- compartment type if the water runs in slowly and continuously: it is then simply a "contact" apparatus, and may or may not exhibit overheating, as well as the inevitable after-generation. All generators of the water-to- carbide type, too, must yield a gas containing some air in the earlier portions of their make, because the carbide containers can only be filled one-third or one-half full of solid. Although the proportion of air so passed into the holder may be, and usually is, far too small in amount to render the gas explosive or dangerous in the least degree, it may well be sufficient to reduce the illuminating power appreciably until it is swept out of the service by the purer gas subsequently generated. Moreover, all water-to-carbide generators are liable, as just mentioned, to produce sufficient overheating to lower the illuminating power of the gas whenever they are wilfully driven too fast, or when they are reputed by their makers to be of a higher productive capacity than they actually should be; and all water-to-carbide generators, excepting those where the carbide is thoroughly soaked in water at some period of their operation, are liable to waste gas by imperfect decomposition.