[fitsbits] Primary & Alternate WCS Keyword Order

[Lucio Chiappetti]

> I fully agree on those statements. Usual FITS handling libraries allows 
> random access to headers, so order is not an issue.

There are not only libraries, but also humans reading FITS headers !
And in that case having SOME understandable order helps !

>> A major instrument being commissioned has randomly ordered keywords in 
>> each header, has randomly ordered HDUs, and ...

This seems something that will be no problem for a smart software 
accessing things by name, but will cause hiccups to any human operator 
trying to use the data "manually".

I believe a PRACTICAL COMPROMISE would be that the standard itself does 
not dictate too strict positioning requirements (well, the mandatory kwds 
at beginning are there, and EVEN the fixed form for some of them, and 
"once FITS forever FITS"), i.e. mean ON THE READING S/W ... but reasonable 
practice groups keywords in an order (any order) which makes life easy to 
an human operator.

This could be for instance the mandatory kwds first, then for a BINTABLE 
things like TFIELD and then all TFORMs, all TTYPs, all TUNITs in numeric 
order (but might also be, the TFORM, TTYP, TUNIT etc. of column 1, then 
those of column 2 etc.). Or for any file with multiple WCSs, all kwds of 
WCS 1 (blank) first, then all those of WCSa, then those of WCSb.

In this respect the requirement on WCSAXES coming first (the issue which 
started this thread) could sensibly be interpreted as applying internally 
to each WCS (i.e. making Bill's 3a/3b legal)

>> There may be higher order implications such as the inheritance issues 
>> you mentioned.

Well, I understood there was NO inheritance issue among different WCSs 
which are all independent.

Inheritance among HDUs is subject of specific conventions, isn't it ?

-- 
Lucio Chiappetti - INAF/IASF - via Bassini 15 - I-20133 Milano (Italy)