Call for proposals for SIAP Version 2

[John Good]

Arnold -

> Though I in general agree with this, the comments below are too
> restrictive.  A convex polygon may be sufficient for queries, but
> certainly not for general applications.  And one has to be careful
> when dealing with projections.  Great circles are straight lines in
> TAN projections, but not in any other.  Concave polygons and regions
> with holes will need to be dealt with.

I agree that complex regions will have to be dealt with in general.
However, convex polygons are so simple to deal with and cover
such a large fraction of the query instances that we are likely to
encounter that I think it would be short-sighted not to include
them in the baseline request syntax (both for image access and for
catalogs).

Besides, a general complex region to a service that does not have
that ability to provide complex filtering would benefit from that
service at least being able to return the data from a bounding
polygon, which is easy to create from the complex region definition.

While it is true that TAN is the only projection where image corners
are precisely connected by straight lines,  many real-world images
cover small enough regions of the sky to make the polygon approximation
good enough for almost any projection (SIN, STG, etc.)  I'll think you'll
find that for most of the images coming out of archives, we're talking
about a fraction of a pixel maximum deviation from the a great circle.

Also, for most common projection, the great circle connected area
contains the true area, so queries return a few more candidates than
an exact filter would produce, which can then be filtered.

- John