Comparison of pgf plotting packages

I know three ways to do function plots based on the TikZ/pgf package:

Which of those options is the most powerful? Why? Can you give typical cases where one package is better/easier to use than the other?

-

Well, I can't speak to the features of tkz-fct, but pgfplots is in every way superior to plain tikz. I am quite new at it, but here are the things I've already enjoyed:

• It draws the axes automatically. You probably understand how wonderful that is: I don't have to worry about the range (which can change as I develop the picture); I don't have to worry about where to put the arrows; I don't have to worry about how much the axes will "overhang" into the quadrants I'm not using (e.g. if I draw a first-quadrant picture, how the tails of the axes will look as they cross the origin).

• It has logical, "graph" coordinates that are independent of picture coordinates or page dimensions; these coordinates scale automatically to fit the picture and you can specify their extent, too! No more guessing how high to draw an asymptote!

• The syntax for writing a function in algebraic notation is a little nicer: you can write x instead of \x.

• You can filter coordinates: no more guessing how close to the asymptote to make the domain in order to get the picture to look decent! Also, no more splitting the domain around singularities.

This alone makes it my package of choice as a casual user: if I want to draw a random picture, it's as simple as

\begin{tikzpicture}
\begin{axis}

I do not know what would constitute a power of a package, and what would make one more powerful then other, but it is my impression that tkz-fct package is aimed mainly to creating plots of the kind that you would find in a calculus textbook, while pgfplots excels at creating more of a "scientific" or "applied" plots, something you would use in an article about physics or biology.
Funny coincidence: I have just submitted my article to IEEE-TRO crawling with pgfplots stuff :) – percusse Jan 2 '12 at 19:58