* The part played by Sharduris in the events of the years which followed, passing mention of which was made by Winckler (Gesch.  Bab. und Ass,, pp. 224, 225), have been fully dealt with by Belck and Lehmann (Chaldische Forschungen, in Veriiand. der Berliner anthropol.  Gesellschaft, 1895, pp. 325-336).

Matilu of Agusi, unable to carry the day against the Assyrians in the open field, distributed his men among his towns, and resisted all attacks with extraordinary persistence, confident that Sharduris would at length come to help him, and with this hope he held out for three years in his town of Arpad.  This protracted resistance need no longer astonish us, now that we know, from observations made on the spot, the marvellous skill displayed in the fortification of these Asiatic towns.  The ruins of Arpad have yet to be explored, but those of Samalla have been excavated, and show us the methods adopted for the defence of a royal residence about the middle of the century with which we are now concerned.  The practice of building citadels on a square or rectangular plan, which prevailed so largely under the Egyptian rule, had gradually gone out of fashion as the knowledge of engineering advanced, and the use of mines and military engines had been more fully developed among the nations of Western Asia.  It was found that the heavily fortified angles of the enclosing wall merely presented so many weak points, easy to attack but difficult to defend, no matter how carefully they might be protected by an accumulation of obstacles.  In the case of fortresses built on a plain, where the plan was not modified by the nature of the site, the enclosing wall was generally round or oval in shape, and free from useless angles which might detract from its strength.  The walls were surmounted by battlements, and flanked at short intervals by round or square towers, the tops of which rose but little, if indeed at all, above the level of the curtain.  In front of this main wall was a second lower one, also furnished with towers and battlements, which followed the outline of the first all the way round at an interval of some yards, thus acting as a sort of continuous screen to it.  The gates were little less than miniature citadels built into each line of ramparts; the gate of the outer wall was often surrounded by lower outworks, two square bastions and walls enclosing an outer quadrangle which had to be crossed before the real gate was reached.

[Illustration:  226a.jpg PLAN OF THE ANCIENT CITY OF ZINJIRLI.]

     A reproduction by Faucher-Gudin of the first plan published
     by Luschan.

When a breach had been made in this double enclosure, though the town itself might be taken, the labours of the attacking force were not yet over.  In the very centre of the place, on a sort of artificial mound or knoll, stood the royal castle, and resistance on the part of its garrison would make it necessary for the enemy to undertake a second siege no less deadly and protracted than the first.  The keep of Zinjirli had only a single gate approached by a narrow causeway.