THE EXCAVATIONS 39 



fragmentary, and are so constantly under water, that a 

 reliable plan of that entrance can scarcely be hoped for. 

 Such measurements as have been taken, however, indicate 

 that it was similar to the other double gateways. 



DIMENSIONS OF THE FORT. 



The uncovering of the north and south gateways made 

 it possible for the first time to obtain the exact dimensions 

 of the fort. Turning to the plan, it will be seen that the 

 enclosure is almost a rhombus, with the comers 

 rounded off, as was usual. As is explained elsewhere, ^^ 

 the departure from the rectangular shape is no doubt due 

 to a slight error in setting off the right angle in the centre 

 at the outset. It will be seen that the plan of Gellygaer 

 received a similar twist in the opposite direction. 

 Apparently, the angle was only set off once, after which 

 measurements were made with ten-foot rods (decem-pedoe), 

 along and parallel to the two base lines at right angles. 

 This explains the repetition of the error throughout. 

 Curiously, another error appears in both plans. If the 

 front line of the central building be produced, it will be 

 found in each case to pass out at about the centre of one 

 of the western gates. 



The orientation of these plans is a matter of interest. 

 When forts lay along a frontier, of course the lie of the 

 fort would be determined by the lie of the frontier. In 

 the majority of other cases, so far as I can find, the 

 diagonals, roughly speaking, are directed towards the 

 cardinal points. Of course this may be purely a matter 

 of chance, due to the lie of the ground. ^^ 



The exact length of Melandra, measuring to the outer 



32. See p. 67. 



33. Vegetius {De Re Milit. , 23), is explicit on this matter : " Pwta 

 autem quae appellatur praetoria aut orientem spectare debet, aut ilium 

 locum qui ad hostes resjnciet." 'Whj orie7item, I wonder ? 



