[SpectralDM] - UTypes changes

[Douglas Tody]

> For this update of the 2013 Spectral working draft,  the goal was to keep
> the footprint fairly small.. making the Dataset level changes which were
> pretty stable at that time.

Right, as I said earlier, when we started this 3 or so years back
to goal was to minimize changes and fast-track a Spectrum -> Spectral2.0
so that we could use it for TimeSeries and the like.  However since
then we have gotten SSA 1.1 and Spectrum 1.1 which are stable Recs
in use now, based upon the older versions of the data models, and
ImageDM is coming along, extending ObsCore.

I really don't see any reason for an intermediate SpectralDM 2.0 at
this point. If ImageDM is to be based upon ObsCore, then SDM 2.0 should
be as well, and we will have a consistent set of modern data models.

> BTW, that would have to be Char-1.13 since Char2 is not a
> recommendation.

It is not necessary to move Char2 to Rec status to do this.  All we
need to do are make the minor changes to ObsTap/ObsCore as some of
us proposed last fall.  We already defined the slight modifications
to the Utypes at that time.

FWIW, we have already implemented (prototyped) all this in the current
DALServer framework.  The SSA implementation provides both 1.1 and 2.0
interfaces, with the 2.0 version extending ObsCore the same way the
"stable" version of the ImageDM that I put out last fall does.  Not
being able to wait longer for final standards, we have already started
to generate metadata in the NRAO and ALMA archives based upon this
approach.  It allows ObsCore/ObsTAP to be used for 90% of the metadata,
and ImageDM and SDM2.0 merely extend the core model.  There is a
proto-SIAV2 based upon this approach as well of course, and it exposes a
SIA1.0 interface to support legacy applications.

So we already have at least one working prototype of all this.  My
suggestion is to fast-track an updated ObsTAP spec with the proposed
Char2 tweaks, and base SpectralDM 2.0 and ImageDM upon this.  An
intermediate version of SpectralDM2.0 is not needed.  SSA/Spectrum 1.1
can be used instead until a real 2.0 becomes available.

 	- Doug



On Sat, 3 May 2014, CresitelloDittmar, Mark wrote:

> On Fri, May 2, 2014 at 5:45 PM, Douglas Tody <dtody at nrao.edu> wrote:
>
>> So for example, in the latest (and earlier) SDM draft we have:
>>
>>     Char.SpatialAxis.Coverage.Location.Value
>>
>> Whereas in ObsCore and the ImageDM (and Char2) as of last fall:
>>
>>     Char.SpatialAxis.Coverage.Location.Coord.Position2D.Value2.C1
>
>    Char.SpatialAxis.Coverage.Location.Coord.Position2D.Value2.C2
>>
>
>
> Ah.. Those are Char utypes, and the Spectral model has a simplified Char
> model as per the original requirements.
> You're suggesting that this revision should remove the simplifications and
> reference the complete model. BTW, that would have to be Char-1.13 since
> Char2 is not a recommendation.  Either way, Char uses STC Coordinates for
> these.
>
> For this update of the 2013 Spectral working draft,  the goal was to keep
> the footprint fairly small.. making the Dataset level changes which were
> pretty stable at that time.  We wanted to allow this doc (which has been
> sitting idle) to move forward, as there are people anxious to use it.
>
> Once the Cube work progressed further, and the Data sections became more
> stable, another update would be needed, to bring it in line.
>
>
>
>>
>> Others differ as well, although I have not done a careful inventory.
>>
>> I have not followed all the discussions, but as this has nothing to do
>> with Dataset, and is inconsistent with ObsCore/ObsTAP, I suspect there
>> is an inconsistency.
>>
>>         - Doug
>>
>