NVO Metadata Working Group Meeting
Minutes of Aug 15, 2002
(taken by Steve Kent)

Attendance:  Tim Cornwell, John Good, Bob Hanisch, Steve Kent, 
	     Jonathan McDowell, Tom McGlynn, Joe Masserella, 
	     Ray Plante, Arnold Rots, Doug Tody

Proposal: Emails for discussion should NOT be sent around just before the
phonecon [laughter].

NEWS:

Jonathan, Doug, Arnold, and Ray met to talk about data model concepts.  The
process will be ongoing.

DISCUSSION: Simple Image Access Specification
(http://aoc.nrao.edu/~dtody/sim.html)

Tom:
   1. We need to get away from cryptic abbreviations in the URL keywords.
   2. Need a more general coordinate specifier than RA, DEC.  Can we
      use a single keyword with, say, two numbers?

There was much discussion on these points.

Doug: Need to get something going fast

Ray: We might make staged decisions based on what we need for the demo and
   what we need in the longer term.

Bob: We should use FITS keywords when appropriate because they exist and
   have a wide usage.

Bruce: URSA's experience is to prefer a single keyword

What about images with different types of axes? (no one wanted to pursue this).

Should the cone search switch to using one keyword?  NO, too close to
   November

SUMMARY of this discussion:

1. We agree that it is OK to use longer names for keyword.  We don't need to
   be beholden to any decisions made right now for conventions in the long
   term.

2. We will use FITS where applicable.  A discussion ensued about CTYPE,
   and it was agreed that this keyword embodies too much complexity;
   instead, we will use PROJECTION.

3. We will use a single keyword for position.  Right now only RA, DEC, J2000
   is allowed for the system.  I think there was a suggestion to use comma
   as a separator.

The remaining discussion focused on returning other info.

Roy: Can VOTable return other stuff besides 1 table?  Yes, but not for now.
   If we loosen requirements later, we can expect the current implementation
   to break.  Note that additional information can be returned in the image
   file itself.

Arnold: Can we add "resolution" (PSF) to the returned info?  Consensus was
   no because it can be complicated, and some archives don't have that info.

Arnold & Doug: Region and/or mask information will be needed in teh future to
   define which parts of an image are valid, since the image returned is
   often only an approximation to that requested.  Note that one can
   flag invalid pixels in the image right now (kind of clumsy, though).

What about errors?  Are they even in WCS yet?  Bob: Only in the rough
   draft of paper IV.

Arnold: Why not strings?
Ray: They are in XML but are not allowed in VOTable (which adheres to fixed-
  length rows in FITS tables).  (Arnold not happy).

Arnold:  Can we return exposure times?  This is necessary for the galaxy
   morphology demo.
Others: This is tough.  Info not available or even meaningful for some data
   sets.  Also, it is less useful as one moves away from the telescope.

Doug: Calibrations are another issue.  Bob: This might be useful to end users
   if made available as an optional parameter.  You now need to specify
   central wavelength, limiting flux, ...

There was a feeling that we were succumbing to mission creep.

Roy: Virtual images can be produced with arbitrary sizes and resolutions, and
   so one could in principle return an infinite number in response to a
   query.
Doug: The intent is that the user specifies a default image size, and leave
   it up to the service implementer to do the right thing.

Ray: Can we finish off the document soon?  Consensus was yes, although
   Arnold has more comments on the image staging section.

[Ajourn]