The cold-air flues leading from either side of the furnace must be conducted to their respective inlets. If possible, at least two inlets should be provided, facing different ways: this with regard to the possibility of certain winds drawing the air out where it is wanted to enter. The openings should be vertical, like windows, and, in cities, furnished with a solid frame and casement, fitted with louvres of plate glass with polished edges. Between the rebate and the casement it is a good plan to leave a space of an inch and a half for a movable stretcher-frame holding several layers of "cheese-cloth" to filter the air. The construction of such an air filter is shown at Fig. 7. The glass louvres keep out the wet, and throw off coarse particles of falling soot; and the provision of a movable stretcher permits the cloths to be frequently changed for clean ones--a very important point, though little heeded, if not, perhaps, wholly ignored. [Illustration: FIG. 8. Plans and Section of a Furnace Chamber, &c., for a Bath on the ordinary Hot-air Principle.] The position of air intake is a matter of great importance, especially in large towns. It evidently is bad to draw a supply of air from the bottom of an area. Even the position shown in Fig. 8 is not good: the shaft should be carried higher. The best places for the intakes are where there is always a current of pure air blowing, and away from smoky chimneys. Theoretically, it would seem that the higher the level of intake the better; but in cities, by going high we get among the belching chimney-tops, even if we escape the stagnation below. Moreover, a high inlet with a strong wind tending to exhaust the air in the shaft might find the architect with the cold air sweeping through his bath, and all the heated air rushing up the supply-shaft. A large "lobster-back" automatically turning _towards_ the wind, would in many cases prevent such a disastrous result. Even in low-level intakes, as I have said, trouble will sometimes arise from the same cause. This may be remedied by providing more than one inlet, so that only the one facing the current of air will be employed, the other being closed, which could be effected by fixing the glass louvres, spoken of above, on pivots, and connecting them with a rod and adjustable rack. It would be a very simple matter to make the wind itself automatically open and shut the louvres.