[fitsbits] Draft WCS Paper V: Time

[Rob Seaman]

On Apr 6, 2009, at 6:41 PM, Robert Hanisch wrote:

> This reference is a tribute to Peter Bunclark, I think, and in his  
> memory it would be nice to let it stand.


That is certainly sufficient justification.  FITS and ADASS and IVOA  
are certainly not overflowing with whimsy - a little poetry (Latin or  
otherwise) can't hurt.

> As a standards document I don't quite see how it helps anything,

That said, Astronomy is a field where the actions of the Vatican from  
the 16th century are still in force.  A normative standard should cite  
normative references.  It seems likely that nothing has superseded the  
1582 reference.  In addition to whimsy and poetry, this is likely  
pedantically correct as a fundamental reference.

> unless we had a FITS file somehow documenting the supernova of 1054,  
> or something like that.

It is not unlikely that such a file, for instance a table or  
simulation, would be created and used within a research or outreach  
context.  The standard needs to support such cases.  There is no irony  
here.

> But for such a thing, we do not need to understand TAI or whatever.

I'm not quite sure what you are saying.  Astronomy is broad in space  
and deep in time.  It may seem quaint for FITS to wrestle with complex  
issues of timekeeping in a nominally historical context, but FITS  
should be applicable for a variety of proleptic use cases.  The reach  
of FITS should exceed our grasp.

> The usual thing is to let the authors publish their version, and  
> then the
> FITS commandos come in and distill it down into the formal language  
> of the
> standard.  It would be nice if this translation could be an identity  
> matrix.

For these issues the authors are the FITS commandos.  If you are  
suggesting that papers I-V are not normative, I'm not sure I agree  
with that interpretation.  Surely for a number of WCS issues the  
"formal language of the standard" still resides in the published papers?

> Arnold has already chastised me for finding the keywords to be  
> obtuse and
> inconsistent, and he is relying on history in his rebuttal. OK, we  
> messed up
> before, and to be consistent, we will mess up again.  Doesn't  
> matter, it is
> just a list of keywords.  I just find it strange to represent "time"  
> as  T,
> TIM, or TIME.  Or "center" as CEN, CENT, or CENTE.  We accepted this  
> before,
> so I guess we go with it.

If usage governs then the biggest issue is DATE-OBS.  Many millions of  
files (and more every day) contain a DATE-OBS that expresses only a  
date.

Any standard is a trade-off between the descriptive (usage) and the  
proscriptive (newly conceived architecture).  I would suggest that the  
paper should express the minimal set of keywords needed to address the  
full set of astronomical use cases.  These will prescribe the desired  
new usage.  In addition, the paper should describe the range of usage  
known to exist in the community.  Ideally there will be (some)  
overlap, but historical precedence shouldn't overrule improving  
timekeeping in the standard.

Whether keyword names are truncated is only a minor aspect of this.   
Rather, tell people what their institutions ought to be doing, and  
also how to interpret what other institutions ought not to be doing.

Rob