This is the most elaborate and far-reaching library classification yet put forth, claiming superior clearness, flexibility, brevity of notation, logical cooerdination, etc., while objections have been freely made to it on the score of over-refinement and aiming at the unattainable. What is known as the decimal or the Dewey system of classification was originally suggested by Mr. N. B. Shurtleff's "Decimal system for the arrangement and administration of libraries," published at Boston in 1856. But in its present form it has been developed by Mr. Melvil Dewey into a most ingenious scheme for distributing the whole vast range of human knowledge into ten classes, marked from 0 to 9, each of which sub-divides into exactly ten sub-classes, all divisible in their turn into ten minor divisions, and so on until the material in hand, or the ingenuity of the classifier is exhausted. The notation of the books on the shelves corresponds to these divisions and sub-divisions. The claims of this system, which has been quite extensively followed in the smaller American libraries, and in many European ones, are economy, simplicity, brevity of notation, expansibility, unchanging call-numbers, etc. It has been criticised as too mechanical, as illogical in arrangement of classes, as presenting many incongruities in its divisions, as procrustean, as wholly inadequate in its classification of jurisprudence, etc. It is partially used by librarians who have had to introduce radical changes in portions of the classification, and in fact it is understood that the classification has been very largely made over both in Amherst College library and in that of Columbia University, N. Y., where it was fully established. This only adds to the cumulative proofs that library classification cannot be made an exact science, but is in its nature indefinitely progressive and improvable. Its main object is not to classify knowledge, but books. There being multitudes of books that do not belong absolutely to any one class, all classification of them is necessarily a compromise. Nearly all the classification schemers have made over their schemes--some of them many times. I am not arguing against classification, which is essential to the practical utility of any library. An imperfect classification is much better than none: but the tendency to erect classification into a fetish, and to lay down cast-iron rules for it, should be guarded against. In any library, reasons of convenience must often prevail over logical arrangement; and he who spends time due to prompt library service in worrying over errors in a catalogue, or vexing his soul at a faulty classification, is as mistaken as those fussy individuals who fancy that they are personally responsible for the obliquity of the earth's axis.