1. Properly speaking the term 'replica' should be reserved for objects or processes achieving the highest levels of accuracy and authenticity in all respects. Lesser standards are realised in 'repro-ductions', 'reconstructions', and 'representations', in that descending order. Really first class replication is probably extremely difficult to achieve, and it is, therefore, unfortunate that the term 'replica' is used generically. To avoid this confusion I prefer, and will try to use, either copy or re-creation as the general term.
2. An example from ancient water-supply is instructive. With rare exceptions, Roman aqueducts contain so many changes of slope, direction, and cross-section at very short intervals that mod-ern open-channel flow analysis is very difficult indeed to apply. But for the Romans such varia-tions did not matter; if an aqueduct had a downward slope it would flow. If it did not flow satisfactorily then there was always the option of putting right everything that had gone wrong. What we might call 'design by successive modification' is by no means unknown even today. The Nimes aqueduct is one that performed badly — significant overflow was experienced — and so the aqueduct's sides were raised at strategic points along the route.
3. The Engineering of Medieval Cathedrals, ed. L.T. Courtenay (AshgateNariorum, Aldershot,1997).