Steve Allen wrote:

We also do not have a set of altered words for Bill Pence to propose
which make it clear that traverse along array indices isn't really
pixels and does not really have units until and unless the WCS says
they do, and that in the case where the coordinate along the array
axis can reasonably be interpreted as a real-valued entity the data
value is intended to correspond to the measured quantity at the
integer values which run from 1 to NAXISj.

This discussion reminds me of the innumerable "which way is up?"
questions in the IRAF mail over the years. The answer goes something
like "the question is meaningless until you display the image".

A couple of points for semantic musing. What does binning of pixels
do to those integer values? (Or decimation, subsampling or block
averaging.) Coordinates (center of the pixel or otherwise) may start
as integer values 0, 1, 2, 3. Bin by two and this turns into 0.5,
2.5, 4.5, ... Also, "measured quantity" isn't really the extent of it
in a world of theory and simulations. It's more like the distinction
between independent and dependent variables.

And many times the WCS is implicit, as with an image display causing
the up/down/left/right wave function to collapse. Call this the
astronomy.net effect. Even in an image completely devoid of metadata,
celestial fiducial marks express an implicit WCS.

Rob