[Illustration: FIG. 13.--Bastion at Troyes.]
The earliest bastions were modified bulwarks with straight faces and flanks, attached to the main wall, for which the old towers often acted as keeps; and at first the terms bulwark and bastion were more or less interchangeable. Fig. 13, taken from a contemporary MS. by Viollet-le-Duc, shows a bastion added to the old wall of Troyes about 1528. On the other hand, in fig. 14 (taken from an English MS. of 1559, which again is based on the Italian work of Zanchi published in 1554), we find _a a_ spoken of as "bulwarks" and _b b_ as "bastilions." The triangular works between the bastilions are described as "ramparts," intended to protect the curtains from breaching fire. (We may also notice in this design the broad ditch, the counterscarp with narrow covered way, and loopholes indicating counterscarp galleries.)
Towards the end of the 16th century the term "bulwark" began to be reserved for banks of earth thrown up a little distance in front of the main wall to protect it from breaching fire, and it thus reverted to its original defensive intention. The term "bastion" henceforth denoted an artillery position connected by flanks to the main wall; and the question of the arrangement of these flanks was one of the main preoccupations of engineers. Flanks retired, casemated or open, or sometimes in several tiers were proposed in infinite variety.
Thus, while in the early part of the 16th century the actual modification of existing defences was proceeding very slowly on account of the expense involved, the era of theoretical "systems" had begun, based on the mutual relations of flank and face. These can be grouped under three heads as follows:--
1. The _crémaillère_ or indented trace: Faces and flanks succeeding each other in regular order (fig. 15).
2. The _tenaille_ trace: Flanks back to back between the faces (fig. 16). The development of the flanks in this case gives us the _star_ trace (fig. 17).
3. The _bastioned_ trace: Flanks facing each other and connected by curtains (fig. 18).
In comparing these three traces it will be observed that unless casemates are used the flanking in the first two is incomplete. Guns on the ramparts of the faces cannot defend the flanks, and therefore there are "dead" angles in the ditch. In the bastioned trace there is no "dead" ground, provided the flanks are so far apart that a shot from the rampart of a flank can reach the ditch at the centre of the curtain.