[iaufwg] FITS panel recommendations; ready for review?

[Robert Hanisch]

On 7/5/07 5:43 PM, "Steve Allen" <sla at ucolick.org> wrote:

> On Thu 2007-07-05T16:46:22 -0400, William Pence hath writ:
> 
>>     Are the technical panel's recommendations ready for wider
>>     public review and comment?  If not, what deficiencies need
>>     to be fixed before starting the public review process?
> 
> In conformance with RFC 2119, most instances of "must not"
> should be italic for both words.
> 
> Point 7 of the summary refers to section 4.1.2.3 and the added rule
> that keywords with a value shall not be repeated in a HDU.  Almost
> every FITS file from the optical spectrographs and imagers at Keck has
> always violated this rule.  I don't defend that practice, I simply
> report it.  This new rule would ex post facto declare all those Keck
> files to be non-conforming.

I wouldn't defend the practice either, but I think it is better to provide
guidance on what should be done when duplicate keywords are found rather
than to prohibit their existence.   I suppose change #5 to Section 3.7 would
get around this, but it seems like making three left turns just to make one
right turn.

Several of the changes aim to remove duplication, but #10 in Section 4.4.1.2
inserts a duplication.

I don't understand #14, as EXTNAME / EXTVER / EXTLEVEL would imply that they
refer to extensions and not the primary HDU.  I would guess that HDUNAME
etc. have been introduced in order to construct a referencing system for the
components of a FITS file, and I would think the distinction would in fact
be useful rather than pedantic.

In Other Individual Recommendations...

14, re/ DATE, the statement is obsolete but is relevant to reading FITS
files written prior to 1 Jan 2000.  Shouldn't it remain in the document so
that people know explicitly when this rule went into effect?

In Other Global Recommendations...

4, re/ twos-complement, this is very curious.  Indeed, most documents I
scanned on the web write "two's complement", but sometimes it is
"two's-complement" and sometimes it is "twos complement".  Obviously this is
intended to be a possessive form, as "two is complement" makes no sense.
Even more curious, 
http://www.answers.com/topic/signed-number-representations talks about
"ones' complement".   So, "two's complement" implies the complement is of a
single two, and "ones' complement" implies the complement is of plural ones.
Who knows?  Personally I think the possessive does not really make sense
here, and "twos" is functioning more like an adjective (which would make the
original "twos-complement" more appropriate), but the prevailing usage seems
to be what the panel is suggesting.  (Sorry - that was a grammar fanatic's
digression!  Or perhaps a grammar fanatics' digression.)

9, re/ logical record... I don't think "FITS block" is particularly
illuminating, nor does it seem quite right in terms of a hierarchy.  That
is, one or more FITS blocks comprise a header block, and a header block and
data block comprise a header-data unit.   "FITS block" sounds to me like a
larger scale aggregation.


All comments above are simply on the summary of changes, not on the full
text.

Bob