[iaufwg] Call for vote on the new FITS Standard

[Eric W Greisen]

Mark Calabretta wrote:
> On Tue 2008/05/06 18:03:44 -0400, William Thompson wrote
> in a message to: IAU-FWG <iaufwg at nrao.edu>
>
> Dear Bill,
>
>   
>> Nowhere does it state that CRPIXj runs from 1 to N, instead of from
>> 0 to N-1.  I've found that the distinction between these two cases
>> trips up many novice FITS users, and is the most common mistake made.
>>     
>
> I agree with the sentiment but not the proposed change to the definition
> of CRPIXja.
>
> In the language of Eq. 8.1, CRPIXja simply records r_j (so it wouldn't
> be a bad idea for the definition of CRPIXja to refer back to that).  In
> turn, r_j is just a specific value of p_j.
>
> The problem is that p_j, i.e. "pixel coordinates", are never properly
> defined.  They are actually a generalization of storage coordinates
> i.e. the integer array indices used to address the data array - refer
> to http://www.atnf.csiro.au/people/mcalabre/WCS/Intro/WCS04.html.
>
> Basically there needs to be a statement somewhere near Eq. 8.1 referring
> to the "index count" defined in Sect. 3.3.2 with a statement that
> p_j = 1.0 coincides with I_j = 1, i.e. the index of the first datum on
> axis j, and that p_j increments in step with I_j.  Non-integral p_j may
> be understood to indicate interpolation between two data values and is
> disallowed for certain "conventional" axis types.
>
> As for visualising FITS image data (a completely separate issue), there
> should also be a statement somewhere that when a datum is represented
> as a "pixel", I_j, and hence integral p_j, coincides with the centre of
> the pixel.
>
> Also, the reference to the HEALPix paper on p120 should be MNRAS, 381,
> 865.
>
>   
I have been rather silent but must agree that this wording needs to be 
corrected and clarified.  If someone
as knowledgeable as Bill Thompson is confused, then we have a serious 
problem.  I thought that the j of CRPIXja is restricted to 1 to WCSAXES 
and the value of it may be anything but is a count wrt to the first 
datum on that axis - or voxel if you like.
Someone should suggest a corrected wording.

Eric Greisen