ACTION-2815: Draft a complete response to ISSUE-2338 and research the answers to the questions asked

Regarding Olaf's question (below), I have created the following response. It is to the best of my knowledge correct and actionable.

C and C’
In reading the specification it appears that C and C’ do correspond to the interpolation range for channel, and that as noted, it appears that the functions for other values types imply that 0 <= {C|C’} <= 1 (and more notes below on C=1). What is not clear to me is whether or not C, for types identity linear, and gamma are assumed to be 1, while others are calculated.

My first glance at the table looks as though your issue with C=1 for table and discrete could be a problem with the text.

Table:
It seems that this:
which provides a list of n+1 values (i.e., v0 to vn)

should read:
which provides a list of n values (i.e., v0 to vn-1)

Discrete:
It seems that this:
which provides a list of n+1 values (i.e., v0 to vn-1)

should read:
which provides a list of n values (i.e., v0 to vn-1)


N vs. n
N appears indeed to be n.

These should be clarified in the specification.

==============================
(orginal mail)
Hello www-svg,

within the definition of the attribute type for the elements feFunc*
http://www.w3.org/TR/SVG11/filters.html#feComponentTransferTypeAttribute
there seem to be some variables in the formulas undefined.

Is my interpretation correct, that C and C' correspond to the relative
channel value the subelement represents?
Between 0 and 1? Colors have to be transformed to this range?
C before, C' after transformation?
For example for feFuncR this means the relative value of the red channel?

And what exactly is N?
My assumption is, that it is N = n. Is this correct?
At least the formulas for table and discrete look useful for
C from the interval [0;1)
For C = 1 the formulas have the problem,
that the v(n+1) for table does not exist and v(n) not for discrete.
k is one too large, there is a need to provide another formula for C=1
- something like: use formulas for [0;1) and v(n) for C=1 for table and
v(n-1) for C=1 for discrete.
For the table type this is only a formal problem, because (C-k/N) = 0
anyway, whatever v(n+1) might be; the result is C' = v(n).
Nevertheless the formula should not contain the undefined v(n+1).
For the discrete type C'=v(n) for C=1 resulting in k=N is a problem,
because the v(i) are defined in such a way, that the last value is v(n-1).


Best wishes

Olaf