add xtype multishape, remove multipolygon; resolves issue #39

[pdowler]

explicitly allow multi-interval using arraysize="*" (or the like)

The actual name xtype="multishape" is open to debate. xtype="region" has some mind share and is also appropriate.

The "union" interpretation satisfies the practical use cases (e.g. observation footprints).

aside: in my opinion, there is no implied query-ability to this (or any other xtype) here in DALI: that is up to the relevant standards (ADQL, TAP, DAP, etc)

[mbtaylor]

Since we have shape and multishape, I think we should have interval and multiinterval rather than putting them both in the same xtype. Single interval will probably be the more common usage pattern, and knowing up front that you're dealing with a single interval would simplify implementation in many scenarios. I feel like Markus is going to agree with me on this.

The semantics of the multiple interval is not made explicit here - should we say it's union, or leave that to whatever semantics is imposed by the usage context?

[msdemlei]

On Thu, May 08, 2025 at 12:58:10AM -0700, Mark Taylor wrote: simplify implementation in many scenarios. I feel like Markus is going to agree with me on this.

He does, wholeheartedly...

[pdowler]

I don't feel strongly that multi-interval benefits from specifying union, but it is the common meaning in the use cases that is driving this (spectral and temporal coverage).... but I'll put it in and I think we can see what happens during RFC.

OK, I'm ambivalent about subtle arraysize vs new xtype so will change that to xtype="multiinterval" (despite also not liking ii in a word, ugh).

[mbtaylor]

OK apart from the one typo noted.