Interaction between 3rd column for smooth and boxes/filledcurves

[Alex van der Spek]

There appears to be an (unintended?) interaction between the 3rd column
arguments in using clauses when smooth is in effect.

Let "rand.txt " be a one column data file with 1000 random deviates

Compare;

gnuplot>plot 'rand.txt' using 1:(1/1000.):(0.5) smooth kdensity with lines,
\
'rand.txt' using 1:(1/1000.):(0.5) smooth kdensity with
boxes

In the first the 3rd argument is taken as the width of the gaussian kernel.

In the second the 3rd argument is taken as the width of the boxes.

Similar interaction occurs when plot style is filledcurves, which also uses
an optional 3rd column.

I am not sure if this would classify as a bug. It certainly is confusing. I
suspect it is unintended.

Why can the width of the kernel be taken from a 3rd column? It appears to me
one would always want to take the width to be a constant. Nice to have it as
a parameter one can set but why not as a separate parameter for smooth
kdensity rather than from the using clause?

Regards,
Alex van der Spek