Re: [Plastic-devs] spectrum messaging

[Mark T. <m.b...@br...>]

On Wed, 16 Aug 2006, Thomas Boch wrote:

> Mark,
> 
> 
> > (...)
> > 
> > So maybe
> > 
> >    1D:  ivo://votech.org/fits/line/loadFromURL
> >    2D:  ivo://votech.org/fits/image/loadFromURL
> >   3+D:  ivo://votech.org/fits/cube/loadFromURL
> 
> I think I like your suggestion. 
> 2 questions :
> 
> - How would FITS-extensions files fit into this scheme ?

Good question.  Possible answers:

  1. The array must be the primary HDU
  2. The array must be the first array-type HDU (may be primary, but
       if that's empty the first extension of type IMAGE)
  3. 1 and/or 2 unless the FITS URL has some kind of appendage to
       mark the extension ("#n" or "[n]" to denote extention number n?)
  4. 1 and/or 2 unless there's an optional message argument giving
       extension number (0 is primary, 1 is first extension)
  5. Mandatory message argument giving extension number 
       (0 is primary, 1 is first extension)

I probably favour 4.

Incidentally, we should in any case have an additional argument for
all these messages giving an identifier for the array, analogous
to the one sent in ivo://votech.org/votable/loadFromURL


> - If we want to send spectrum or timeseries serialized in VOTable,
> should we define a ivo://votech.org/votable/line/loadFromURL ? I am a
> bit worried by the multiplication of messages this solution would imply
> (though I don't see any better way).

My feeling is that you should have different message IDs for 
different data formats, but address the semantic issues using the
fragment mechanism that John has suggested (and until that's 
in place, just leave them to the user to work out).  The thinking 
behind this is that applications need to know what the data format
is in order to load it (and to know whether they will be able to),
but the question of whether it's sensible to load it is better
answered by a human.  So you don't have the message you mention 
above, but applications would advertise messages like

   ivo://votech.org/votable/loadFromURL#timeseries
   ivo://votech.org/votable/loadFromURL#spectrum

as appropriate.  The data is still a VOTable, and you might still 
want to load it into a generic table viewer such as TOPCAT.
The situation is different for 1,2,3-dimensional FITS arrays,
because a program that knows how to display a 1-d array
(typically a spectrum or time-series viewer) is not especially
likely to know how to display a 2-d array (typically a sky image).

There is room for argument in specific cases (including the one above)
about where the distinction between data format description and 
semantic hint lies, but (until anyone persuades me differently) 
I'd say that this is a good rule to answer questions about what 
functions require new messages to be defined.

Mark

-- 
Mark Taylor   Astronomical Programmer   Physics, Bristol University, UK
m.b...@br... +44-117-928-8776 http://www.star.bris.ac.uk/~mbt/