308 THE SEARCH FOR THE ANCIENT CHINESE HIGH ROAD [Chap. VIII
ing passage the distinct statement that the region in question ` constitutes the northern limit of the lake '. This will be fully understood if we keep in view the fact that now, as formerly, those very limited portions of the ancient sea-bed which still hold patches of salt marsh, with soft boggy ground or with open streaks of salt water, are to be found along the southern shore of the otherwise dried-up basin.12 It must be remembered also that the terminal marshes of the Tarim represented by the present Kara-koshun can be shown to have already occupied in early historical times the southwestern extremity of the basin, and that those once formed by the Kuruk-darya delta, the ` Yu marshes ' of Li Tao-yuan, similarly extended into the south-western portion of the basin. Thus Li Tao-ytian's authority was fully justified in describing the region of the ` Town of the Dragon ' and the great waste of hard salt crust eastwards as ` the northern limit of the lake '.
From Li Tao-yüan's interesting account we may turn next to the notice which is nearest in time and, though brief, has the merit of clear topographical sequence. It is contained in the record which the Wei lio, composed between A. D. 239-65, supplies of the ` route of the centre ', leading from Tun-huang past ancient Lou-lan ' to the great northern highway of the Tarim basin. This important record, which M. Chavannes' translation first rendered accessible, has been fully analysed and discussed in Serindia.13 It will therefore suffice to quote the passage concerning this route, while restricting my comments, except those on the locality which directly interests us here, to a brief indication of the identifications there established or proposed. The passage runs thus : ` The route of the centre is the one which starting from Yii-mên kuan, sets out on the west, leaves
the well of the Protector General VS 4, turns back at the northern extremity of the San-lung
(` Three Ridges ') Sands w, passes the Chii-lu granary g. 1. A ; then, on leaving
the Sha-hsi well #, turns to the north-west, passes through the Lung-tui f (` Dragon
Mounds '), arrives at the ancient Lou-lan i j .'
Even though this itinerary lacks those indications of distances between the several stages which would have been very helpful, the certainty acquired in the course of my explorations as regards the starting and terminal points and the guidance afforded by the recorded bearings made it possible for me, when discussing the passage in Serindia, to locate the intermediate stages mentioned with great probability. Starting from Yii-mên kuan, the famous ` Jade Gate ', the position of which in Han times near the ruined fort T. xiv of the Limes west of Tun-huang is established beyond doubt,14 the ` route of the centre ' followed the Limes line westwards, just as the present caravan track does, to its extreme end near the watch-towers T. iv. a, b. There I place the ` well of the Protector General '.15 As regards the ` Three Ridges Sands ', the evidence furnished by the actual configuration of the ground, by the reference to the route which there ` turns back ', and by the very name, makes it practically certain that we have to place them at the northern extremity of a belt of high dunes crossed by the present caravan route to the east of Bésh-toghrak.16 It is at or near the last-named important halting-place that I consider that the
Chü-lu granary ' was probably situated.
For the location of the Sha-hsi well we are afforded valuable help by the statement that the route on leaving it turned to the north-west. Reference to the map clearly shows that the route coming from 13ésh-toghrak must have kept to the northern edge of the Bésh-toghrak valley, in order to avoid the troublesome and needless crossings of the large eastern inlet of the dried-up sea. It had then, in order to reach Lou-lan, necessarily to turn to the north-west at the western