A question on keywords for 3-D images

[Frank Valdes]

Hello Arnold,

I appreciate your concern.  However, the usage of this is larger than you
may think.  The inheritance approach began, I believe, with the needs at
HST for the new instruments (NICMOS/STIS).  Years of software development
has gone into this and it has been discussed and status papers have been
presented at ADASS for many years.  The IRAF FITS kernel support for
inheritance began at ST.  I expect HST to continue to use this
approach.  NOAO adopted this because it made sense for mosaic instruments
but it is a smaller effort.  One thing I did not say in my posting is that
for the NOAO Mosaic the header design is such that critical keywords are
included in all extensions even though they are the same in all extensions.
For example, exposure time appears in all extensions.  Thus any software
which may not merge the global keywords will still have a useful minimal
header for interpreting the data.

I purposely phrased my comment that this was a "convention" since, as
you say, it is not a FITS standard.  However the idea of conventions is
a known practice in the community.  There is no "assumption" that this
is supported by all FITS readers (though all readers should be able to
ingest both the primary (global) and extension headers.  For the
specific formats, NICMOS, STIS, NOAO/MOSAIC, other similar mosaics,
people will be advised of the format.  It is not very hard to merge
headers if needed.

Frank Valdes

My role in this has been in the design of the NOAO Mosaic format and I had
little to do with the development of the convention.

> From arots at head-cfa.harvard.edu Thu Nov  5 08:25:03 1998
> 
> I would strongly warn against the use of this convention.  If all you
> want is to use your images in IRAF and if you know others will do the
> same, in perpetuity, it's fine.  But most other systems explicitly
> assume the opposite: each HDU stands by itself and there is no
> inheritance.  Previous discussions on the subject concluded that the
> notion of inheritance is not supported by the FITS standard and may
> not be assumed.
> 
>   - Arnold Rots