Review how mask HDUs are compressed

[weaverba137]

In reviewing the data model for release 18.3, I've found several files that have FITS-compressed HDUs. These HDUs are all masks:

• fiberflat files (EXTNAME = MASK)
• frame files (EXTNAME = MASK)
• spectra-NSIDE files (EXTNAME = [BRZ]_MASK)

The compression does not appear to be applied consistently on these files. In particular, the spectra-NSIDE files may be adding extraneous header keywords that may cause astropy.io.fits to interpret the compressed HDUs as compressed primary HDUs, instead of compressed extension HDUs (see astropy/astropy#7312).

In addition, fiberflat and frame files use RICE_1, while spectra-NSIDE files use GZIP_1. Is there a reason for this choice? All of the masks are 2D images, so the dimensionality is not the reason.

Note, I don't think this issue needs to be resolved for release 18.3, but it is something that could bite us in the future.

[weaverba137]

@julienguy, I think I saw another issue related to this, and I think the conclusion was that these mask HDUs should not be compressed at all. Could you please comment?

[weaverba137]

@julienguy, never mind, it's #603 I was thinking of.