334 HISTORICAL NOTICES OF LOP, SHAN-SHAN, AND LOU-LAN [Chap. IX
leads from Tun-huang to the Lop tract along the northernmost range of the Altin-tagh. It is still used at the present time as an alternative to the desert track, and during the summer months is the only practicable route for caravans. But the result would differ very little if the reckoning were assumed to refer to the desert route, since the distance from Miran to the point on the ancient Limes where my explorations have shown the Yü-mên gate-station to have been situated, also amounts by the map to about 295 miles. To Nan-hu or Yang kuan it would be about 3o miles further.4
The distances given to other localities are somewhat more difficult to check, because the routes to which they refer are not determined with equal clearness by the natural configuration of the intervening ground. In the case of the Governor-General's seat at Wu-lei, which may be looked for about the present Chadir on the high road west of Korla,5 we can scarcely go wrong in assuming that the route mainly followed the line of the terminal Tarim and of the Konche-darya, by which all traffic still proceeds from the Lop tract to Korla. The bearing to the north-west is certainly correct, and the distance of about 320 miles, as shown by Dr. Hassenstein's map, does not seriously differ from the 1,785 li of the text, seventeen daily marches being the present road reckoning.
The ` kingdom of Shan ' ill can only be roughly located in the Western Kuruk-tagh. If its identification with Singer, the only place of actual cultivation and that of the smallest size, is accepted,6 we can account for the estimated distance of 1,35o li by assuming that the route, as at present, led up the Tarim to the vicinity of Turfan-karaul and thence past the ruined station of Ying-p`an north-eastwards.' The most direct route from Lop to Turfan, as still used in recent times, lies through Singer, and as the latter place is shown by our Map No. 55 to be separated from the old Turfan capital by about 110 miles, it is possible to assume that the distance of 1,890 li given between Shan-shan and Chü-shih was reckoned along this line of route. But it must be remembered that in ancient times, when desiccation had not progressed as far as now, other routes through the Kuruk-tagh may also have been practicable, and in any case the erroneous bearing indicated, which makes Chü-shih lie to the north-west of Shan-shan, must serve as a warning that we are here on less secure ground.
Distance reckoned to Wu-lei.
4 Dr. Herrmann, Seidenstrassen, p. 106, assumes that the distance from the Yang barrier to Yii-ni was calculated by a route which led first to the Lou-Ian' Site north of Lop-nôr and then turned to the south-west to Charkhlik, where he assumes YU-ni to have been situated, Reference to the map shows that such a route would have implied a very considerable détour, not warranted by physical conditions, which, as far as transit between the Charkhlik tract and Tun-huang is concerned, must always have favoured the use of the shorter lines either straight through the desert or along the A1tin-tâgh. It deserves to be noted that Fa-hsien, who certainly could not have travelled by the then abandoned route passing the Lou-lan settlement, took seventeen days to travel from Tun-huang to Shan-shan (see above, p. 324). This is exactly the number of the rather long daily marches which brought me from Abdal to Tun-huang by the direct desert route.
5 Cf. above, p. 296 ; Herrmann, Seidenstrassen, p. 38, note 4 ; Wylie in J. Anthrop. Inst., 188r, p. 23.
6 Cf. Chavannes in T'oung-pao, 1905, p. 552, note 7, where other references to this petty territory of Shan (called
Mo-shan W by the commentator of the Shui thing, ibis.
p. 570) are discussed. Its identification with ' Kizil-sangir' (i. e. Singer) was first suggested by Grenard, Mission D. de Rhins, ii. p. 61. [My surveys of 1915 showed traces of former cultivation on patches of ground at Po-ch'êng-tzü and Shindi to the north-west and south-west respectively of Singer. This name is now in use for the tiny colony that alone remains.]
7 This well-known route from Tikkenlik on the Tarim to Singer via Ying-p`an was duly surveyed by Col. Kozloff in 1893 and is shown in the map illustrating Roborovsky's expedition of 1893-95. Dr. Herrmann also accepts the location of Shan at' Kizil-sangir ' and the use of the Ying-p`an route (Seidenstrassen, p. 112), but believes, in accordance with his system of interpreting the reckonings of Ch'ien Hanshu's road, that the distance of 1,35o li was obtained by first going north-east from Yti-ni to the ' Lou-lan ' site, then due west to Ying-p'an, and thence to Singer. The map shows the huge détours such a route would have implied. From Col. Kozloff's surveys it is seen that Singer can even now be reached from the Lou-lan ' Site by comparatively short routes provided with springs of some sort. [Our surveys of r91415 have fully confirmed this.]
Routes to Singer and Turfan.