In the north it is clear that at an early period a second branch broke away from the Euphrates at a point about half-way between Sippar and the modern town of Falûja, and, after flowing along the present bed of the river as far as Babylon, rejoined the main stream of the Euphrates either at, or more probably below, the city of Kish. It was the extension of these western channels which afterwards drained the earlier bed, and we may conjecture that its waters were diverted back to the Euphrates at this early period by artificial means.[12] The tendency of the river was always to break away westward, and the latest branch of the stream, still further to the west, left the river above Babylon at Musayyib. The fact that Birs, the site of Borsippa, stands upon its upper course, suggests an early date for its origin, but it is quite possible that the first city on this site, in view of its proximity to Babylon, obtained its water-supply by means of a system of canals. However this may be, the present course of this most western branch is marked by the Nahr Hindîya, the Bahr Nejef, and the Shatt 'Ateshân, which rejoins the Euphrates after passing Samâwa. In the Middle Ages the Great Swamps started at Kûfa, and it is possible that even in earlier times, during periods of inundation, some of the surplus water from the river may have emptied itself into swamps or marshy land below Borsippa. The exact course of the Euphrates south of Nippur during the earliest periods is still a matter for conjecture, and it is quite possible that its waters reached the Persian Gulf through two, if not three, mouths. It is certain that the main stream passed the cities of Kisurra, Shuruppak, and Erech, and eventually reached the Gulf below Ur. Whether after leaving Erech it turned eastward to Larsa, and so southward to Ur, or whether it flowed from Erech direct to Ur, and Larsa lay upon another branch, is not yet settled, though the reference in Hammurabi's letter may be cited in favour of the former view. Another point of uncertainty concerns the relation of Adab and Umma to the stream. The mounds of Bismâya and Jôkha, which mark their sites, lie to the east, off the line of the Shatt el-Kâr, and it is quite possible that they were built upon an eastern branch of the river which may have joined the Shatt el-Hai above Lagash, and so have mingled with the waters of the Tigris before reaching the Gulf.[13]