Subject: Re:Start of the FITS Euro3D Public Comment Period
From:"Lucio Chiappetti" <nospam@lambrate.inaf.it>
Date: Thu, 15 Nov 2007 12:25:14 +0100
To: fitsbits@donar.cv.nrao.edu

On Fri, 9 Nov 2007, William Pence wrote:

> > Comments may be posted here on the FITSBITS mail exploder or the

While the value of this format is undoubted, and the document is rather 
clear, there are a number of little inconsistencies or clarifications 
required here and there.


*) section 3.1 (primary header) requires the use of EXTEND and EPOCH
   (alongside with EQUINOX). The former is no longer mandatory in the
   draft standard 3.0, but the latter is deprecated already in the
   current standard (in favour of EQUINOX)

*) the list of columns in section 3.1 gives data types in verbose
   form (e.g. long integer). It would help to have HERE also the TFORMn
   code (e.g. 1J)

*) pag. 10 column 2 (flag). In cases where this column is identically
   TRUE could it be collapsed in a keyword according to the Greenbank
   convention ?

*) pag. 10 columns 4-5. The "decision" described to "avoid heaps"
   rules out TFORMn='nP'. Is this mandatory or just a recommendation ?

*) pag. 11 columns 10-11-12. If the spectrum is shorter than the max
   length, what is the content of the unused part of the array ?
   This is not defined contrary to columns 6,7,8,9 which use NaN.

*) is there any keyword which gives the max length, other than reading
   the "r" in TFORM10='r?' TFORM11='rJ' TFORM12='r?'  ?

*) section 4.2 4.3 etc. are given in descriptive (example) rather than
   prescriptive form

*) why are XPOS and YPOS rD ? I'd expect rJ 

*) 4.5 CUNITS is "unique and applies to all spectra including science,
   data quality and statistical error spectra". While it is obvious to
   me that the real spectra and its errors are in the same physical flux
   unit (some erg/cm2/s/A or whatever), I fail to see how the data
   quality can be expressed in flux units (in fact see also 4.6, and
   also the last bullet in 4 at pag.8). 

   In general it is not understood why CUNITS is not just replaced
   by the more standard TUNIT10 and TUNIT12 (or even just TUNIT10 with
   the implied provision that column 12 has the same units)

*) 5.2 TFORM1='1B' is described as signed 8 bit integer, but according
   to FITS standard B is unsigned !

*) 5.2 TFORM1='1A'. Does this imply only the first character of
   Square Rectang Hexagon Circle is used ? Or should it be nA, n TBD ?

*) 5.3 example : NAXIS1=34 does not match the sum of the TFORMn !
*) appendix A : here too NAXIS1 does not match the sum of the TFORMn !
*) 4.4.1 :  here too NAXIS1 does not match the sum of the TFORMn !
   I suggest the author checks *ALL* values of NAXIS1 !

Lucio Chiappetti