New version of the SIA V2.0 PR

[Walter Landry]

Hello Everyone,

I have discussed my issues with SIA v2 in other forums, but I was told
I should express them here for the TCG.

As background, I tried to implement an SIA v2 service for a virtual
image service for the Planck satellite.  This is a use case that SIA
v1.0 covered well with its support for Image Mosaicing Service, but
now seems to have been split off into AccessData.  I gave a talk about
my experience at the Banff Interop [1], so this email is a
distillation of the problems I found.

1) RANGE

   Section 2.1.1 includes a RANGE shape.  A RANGE is not a square on
   the sky, but users could easily think that it is.  This makes it
   error-prone.

   Also, a RANGE shape is significantly different from the other two
   types, so it could be a non-trivial amount of work to implement.
   In my particular case, it was enough work such that I did not
   implement it, though I did implement CIRCLE and POLYGON.  This is
   in contrast to something like a BOX shape, which only requires
   transforming the BOX into an equivalent POLYGON.

   Finally, the only use case that I have heard for RANGE is for
   tiling the sky.  However, tiling the sky is fairly easy with a
   POLYGON.

2) Syntax

   SIA v2 invents yet another syntax.  This creates an unnecessary
   burden on both implementors and users.  The syntax that was chosen
   is also error-prone.  We should reuse an existing syntax.  In my
   presentation [1], I presented my work on using Javascript syntax.
   It preserves the simplicity of key-value parameters while scaling
   to extremely complex inputs.  But whatever is decided, please do
   not invent a new syntax.

Cheers,
Walter Landry


[1] http://wiki.ivoa.net/internal/IVOA/InterOpOct2014DAL/Wlandry_DAL_I.pdf