[fitsbits] CRPIX clarification

[Jonathan McDowell]

Mark,
 I was a bit unclear earlier, sorry. I shouldn't try and
reply to email while on vacation!

 However, I think the statement that I was making is nothing to do with
your very valid concerns. My point is that we have integer pixel numbers
(P1,P2,...) like  (241,32,81) and fractional (real-valued) pixels (X1,X2,...)
like (241.3, 32.1, 81.9). Different software systems use different
schemes to map between these; I have seen both
 [A]    Xi = Pi              Pi = (int)Xi
and
 [B]    Xi = Pi + 0.5        Pi = (int)(Xi-0.5)
and of course this is an independent choice from the schemes
 [F]    i = 1,....N
 [C]    i = 0,....N-1

Some people prefer the latter  because then, if you
are [B],[C] then for an NxN image
the 'lower left corner of the lower left pixel' is (0.0,0.0)
and the 'upper right corner of the upper right pixel' is (N,N),
In contrast, with our scheme [A],[F], our images
run from (0.5,0.5) to (N+0.5,N+0.5) when considered as a real-valued
coordinate system, and this is seen as ugly.

 You may consider that [A] is implied by the very concept of fractional
pixel coordinates and that no sane person would interpret the standard
to mean [B],  but history indicates that this is a fallible assumption.

I guess in your discussion [B]  is equivalent to putting the square area
such that the delta function is at one corner of it.
I think you get a wrong interpretation of the data.

On your issues (1) and (2) about the interpolation, I basically agree.
It is true that for the case with both equatorial and galactic coords
on the same image, the world coord loci of a pixel boundary are different
in the the different coord systems so the naive idea that a pixel
is a light bucket on the sky is only valid per-single-WCS etc.
Nevertheless it's a useful concept.

 - Jonathan