Kessener's comprehensive and scholarly contribution, which bravely tackles the basic science in some detail and which will surely be an enduring reference, nevertheless perpetuates some technical misunderstandings that recur in the literature. I begin with the term “siphon”. It is unfortunate that this term continues to be used. The term applies to the situation when, at some point in a closed conduit, the hydrostatic pressure is less than atmospheric. It is not just that the overwhelming proportion of the innumerable pipe systems in the world never know sub-atmospheric pressure and should not be confused with this unusual arrangement, which warrants a technical term of its own (indeed, many modern textbooks do not mention siphons); but to introduce the term inappropriately also leads to misleading reporting and misconceptions. When we reach the discussion of the hydraulics of this system, real siphons find their way into the argument (p. 125), though the evidence shows that there was no siphon at Aspendos. The presence of air pockets and so on is almost irrelevant to operations if no point in the system lies above the Hydraulic Grade Line; were it not so, no garden hose would work reliably.

It is quite true that the problems of getting a siphon to start and to keep it running are significant: even without leaks, air tends to come out of solution and will eventually destroy the liquid column. But in a pipe system which does not go sub-atmospheric, the inverse (if we might dare to use the word) applies: air trapped at a high point will tend to be taken into solution, and the confused internal flow at such a point will be very efficient at entraining and so removing air. Just as a garden hose may splutter when first turned on, very rapidly all air is expelled from the system and there is no question that flow will eventually be established without further intervention.