I reproduce below part of what Allyn Young (1876-1929) had to say about this issue, in his
LSE lectures, 1927-29 as transcribed by Nicholas Kaldor (and published in the Journal of
Economic Studies, 17, 3/4, 1990).
   
Young made these comments as a prelude to a long discourse on   
(i) the limitations of Marshall's one-thing-at-a-time approach to supply and demand
curves;
(ii) why a demand curve is not a utility curve, or, at least, why Marshallian consumer
supluses cannot be added to give a measure of total welfare; and
(iii) why, over time and in a theory of growth, the industry demand curve is not
independent of the industry supply curve and vice versa (still less the aggregate demand
curve from the aggregate supply curve). Earlier (p.26), he had written: "Marshall's supply
and demand curves hold ceteris paribus and cannot be integrated to give the whole social
structure; cff. Say's 'supply is demand' with Marshall's curves." From all this he
proceeds (pp.44-48) to an outline of his 1928 BA presidential address on his macroeconomic
conception of "increasing returns and economics progress".
   
Young writes (pp.35-36):  
   
    "Demand price" is the price at which a specified quantity of goods will be taken off
the market. The older economists in saying that "prices are determined by supply and
demand" used common concepts which they themselves did not scrutinise carefully.
    They had in mind that p = f(s.d), more particulalry, one aspect of it -- i.e., changes
in demand and supply and what effect they have on price. J.S. Mill first attempted to give
precision to the notion. "Price will always be, or tend to be, at that point where supply
eqauls demand." There he considers price not so much as f(s.d.), but rather supply and
demand as functions of price (s = f(p); d = f(p)). And this is more than changing the
dependent and independent variable; _it is not the same statement_.
    Taking Mill's tendency statement prima facie, it is a tautology; but it really implies
a generalisation about the character of supply and demand functions. Marshall then used
curves to give Mill's statement greater precision. (Cournot and Jenkin [first of all] had
already used curves.) Cairnes, however, favoured the older conception of supply and demand
varying and so affecting price. Regarding these curves, the negatively sloping demand
curve is really a generalisation from the market. It is implicit, not explicit in Mill.
Notice that Marshall violated mathematical convention by using price as the independent
variable and yet measured price vertically. Contrast Cournot's curves.
    On Marshall's curve, demand equals the amount that will be purchased at a given price,
and really means hypothetical purchases. Movement up and down the curve does not mean a
change in demand. While the older economists were talking about changes in demand, with
Marshall this can only be shown by shifting the whole curve...
   
Roger Sandilands