David Colander makes a point that helps to clarify some of the disagreement. He is right that "the classicals knew of supply and demand": they used the terms and connected them in a perfectly sensible way to the movement of MARKET prices. But that does not mean that they understood the terms to coincide with the well-defined price-elastic functions of modern S&D theory. 
 
The difference can't just be about whether the cost side or the demand side is emphasized. As David notes, the classicals knew that "demand" influences price in some critical sense: Smith's discussion of increasing returns owing to the division of labor clearly implies that the size of the market for, say, pins will have a bearing on the price of pins. Ricardo's discussion of differential rent implies that the size of the market for grain will have a bearing on the price of grain. Marshall knew that long-period equilibrium price would coincide with long-period cost of production. They all agreed that in the long run, the markets for produced goods will tend to clear, since firms won't persistently produce outputs in excess of what can be sold. 
 
In what respect then do Ricardo & Smith differ from Marshall?  I would say the key differences are these:
 
1. Smith & Ricardo do not explain the distribution of income in terms of price-elastic factor demand curves (see my earlier comment that labor-market clearing is an essential feature of modern S&D theory).
 
2. Smith & Ricardo don't conceptualize demand behavior in terms of price-elastic functions. They didn't say much about this, but from what they do say, their understanding of demand determination appears to be more institutional and more sociologically nuanced than what we find in standard S&D textbooks. 
 
3. Nor do they seem to conceive of seller behavior in terms of price-elastic supply functions. Intersectoral capital flows adjust the output of a good to what Smith calls the "effctual demand" for the good, the amount that can be sold at the natural price.
 
Smith & Ricardo did not outline a mechanism that reconciles demand and output when costs vary with the level of production, and one might well argue that this is a defect of their theory. Marshall seems to have thought this way. One might also dispute, as I would, whether modern S&D theory is a useful way to rectify the deficiency. But I don't think you can really say that the differences between classicals & neoclassicals is essntially a difference of emphasis. Differences (1)-(3) maen, to me at least, that their theories are different theories.
 
Gary Mongiovi