Reminder>Poll on the 'DATExxxx' year 2000 issue

[Preben Grosbol]

Dear Friends of FITS,

 The next millennium is approaching fast which raises the ambiguity
issue for the FITS 'DATExxxx' keywords, as Peter Bunclark of RGO noted
in his posting "DATE-OBS='31/12/99'" to sci.astro.fits/fitsbits on
1996-06-24.  Don Wells, Chair of IAU-FWG, outlined the procedure to
identify a solution in his posting of 1996-07-26 and has asked me to
conduct it.

 Two types of solutions were proposed during the discussion in the
sci.astro.fits News group, namely:
 1) a new data value string definition for the 'DATExxxx' keywords, or
 2) a new set of keywords to replace the 'DATExxxx' ones.

 The main arguments for and against these proposals can be summarized
as follows:

 1) New format for 'date value string':
    Pro: The semantic of all 'DATExxxx' keywords is maintained. 
         It is easy for human readers of FITS headers to understand.
    Con: Direct conflict with definition of format in basic FITS paper.
         It may result in wrong values or even failure of readers.
         Readers must be modified to decode the new format correctly.

 2) New set of 'date' keywords:
    Pro: Any format for the 'value string' can be defined.
         The semantics can be clearly defined e.g. use of time system.
         Old readers will not fail or give wrong date.
    Con: Readers must be modified to use the new keyword correctly.
         Human readers must learn to interpret the new keyword.

 For either option, all FITS writers which write DATExxxx keywords
will need to be changed, and all FITS readers which interpret DATExxxx
keywords will need to be changed.

 Most comments favored the extended ISO 8601 style format for a new
date value string, possibly including its time definition.  Thus, the
main issue at this point is which of the two options is the better
rather than the actual new format.

 The members of the IAU FITS WG were asked to express their opinion
on the two options. The main conclusions were:

  1) Both options have a chance to be approved by the IAU FITS WG
     but option 2 only with great difficulty.

  2) There is a preference for maintaining the 'DATExxxx' keywords
     but with an ISO style data value string.

  3) There is some preference to allow the date/time ISO format.
     The main issue is the less precise definition of many
     'DATExxxx' keywords (e.g. DATE-OBS), making it meaningless
     to specify an exact time.

 As stated above, any solution will require some code changes.  Thus,
it is important to know the opinion of the general FITS user community
before a final proposal is made. I therefore ask you to answer the
following questions:

 a) Which of the two types of solution would you prefer?
     1) a new date value string definition for the 'DATExxxx' keywords
     2) a new set of keywords to replace the 'DATExxxx' ones

 b) Do you have any preference for the new 'date value string'?
     1) 'CCYY-MM-DD'                e.g.  '2123-10-21'
     2) 'ISO_STYLE CCYY-MM-DD'      e.g.  'ISO_STYLE 2123-10-21'
     3) a non-ISO format            e.g.  '21/10/2123'

 c) Should a full date/time format (e.g. ISO 8601) be allowed?

 This is an informal questionnaire which is intended to indicate which
options will have the higher chance to succeed.  The result will be
forwarded to the FITS groups and will thereby help them in their
decision process.

 Please send your answers to these three questions to me
<pgrosbol at eso.org> with copy to Don <dwells at nrao.edu> before
1996-09-18, so that the results of this poll can be discussed at the
ADASS'96 meeting.

Best regards,
Preben Grosbol