It has been indicated that the actual division is made according to the size and shape of the packets. The division corresponds approximately with variation in weight. Few packets weighing more than 3/4 ounce would come within the class of short letters, that is, of letters which can be dealt with at the ordinary sorting frames; but as there is no analysis of the number of packets of less than 1 ounce weight, there is no alternative to the adoption of 1 ounce as the limit of this class.[621] The effect of this is slightly adverse to the short letters and favourable to the heavier packets. As between long letters and the heavier packets the limit is less definite. In many cases packets of the same weight fall into one or other class according to their size or shape, but the mean weight of such indeterminate packets is roughly 4 ounces, and that weight is adopted as giving the mean upper line of division for long letters. In Table H are shown the relative rates of stamping and sorting for each of the various classes of packets. The proportionate cost per packet of stamping and of sorting, based on the rates shown in Table H, is given in Table J.[622] The actual cost is shown in Table L. As regards the cost of delivery some difficulty presents itself. Letter packets, postcards, halfpenny packets, and newspaper packets are, in all cases, taken out for delivery by the same postman, and it is not possible, therefore, as with stamping and sorting, to ascertain the rates of work for the various classes of packets. But the features in the different packets which lead to differences in the rate of sorting, viz. weight and irregularity of shape and size, lead also to differences in the time taken for delivery. In practice the postman makes a division of the packets; and the time occupied in the delivery of the bulky and irregular packets is greater proportionately, as compared with the time occupied in the delivery of ordinary letters, than is the time occupied in sorting. It is not possible, however, to estimate with any degree of exactness the relative amount of time actually occupied in delivering packets of the various classes, and for the division of the cost of delivery (Table J) the rates adopted for the division of the cost of sorting are taken. This method favours the bulky and irregular-shaped packets. CONVEYANCE