Revised ImageDM-ObsCore architecture

[Douglas Tody]

On Tue, 12 Nov 2013, CresitelloDittmar, Mark wrote:

> Doug,
>
> Good to see the revision.. thanks.. I haven't had a chance to read the doc
> yet, but
> spent today working the model (UML) between Observation/Image/Spectral..
> going
> from the previous diagram I sent out.
>
> One thing that I'm trying to do with the diagram, is show very
> specifically, how the
> ObsCore model and Image/Spectral are connected.  The document should start
> with the reference to ObsCore.  The ObservationDM in your diagram modifies
> ObsCore V1.0 with the changes desired for ImageDM.  This would require an
> update
> to ObsCore, which I don't think is in the works.

The consensus earlier is that an update to ObsCore *is* required, as
there are issues with a number of the Utypes in that version of the
model.  The revised Utypes Francois and Mireille have been generating
reflect the work done on Char2 to make things internally consistent and
better integrated with STC.  To use these consistently we have no choice
but to have a new version of ObsCore/ObsTAP.  Then the ImageDM
architecture is much simpler, as most of it is just ObsCore and Char.

This should be more clear in the revised DM spreadsheet; I hope to post
that later today.

> I'm including a revision of my earlier diagram.
> ObsCore Package:
>  On the right, is the ObsCore model as in that document.. except that I've
> extended it from
>  an abstract ObservationDataset class.   (I have trouble here, because the
> Image/Spectral
>  classes need DataID,Target,Curation objects which are Bigger than what is
> in ObsCore.)
>  I'm pretty sure this matches very closely to the ObsCore document.
>
> ObservationDM Package:
>  Is what is needed by Image/Spectral.  That is in pretty close agreement
> with Spectral..
>  there are name differences for things like dataProductType.
>  Note:
>   - got rid of "Dataset" class.. the ObservationDataset IS a dataset.
>   - this does NOT include the Provenance objects.. they are not in
> Spectral.
>     They could be.. (since 0 is allowed).

I need to review your new figures but I think the obvious class names
are Observation, Characterisation, Image, etc.  Mireille suggested
earlier that the "Obs.x" element in Observation (containing
dataProductType etc.) could be replaced by the more general Dataset.x.
It would be beneficial to ObsCore to add, e.g., Dataset.Datamodel to
explicitly document the Datamodel used.  An alternative to Observation
adopting Dataset might be to replace Dataset with Obs with the same
semantics as the current Dataset.  I will review your new figures next
however.

> ImageDM package
>  - adds a Provenance container (from Doug's diagram).  Note that ObsCore
> does not have
>    a Provenance class, but associates ObsConfig and Proposal directly.
>  - adds CoodSys object.

I don't understand this - ObsCore does include Provenance.  See for
example Figure 2, and Table 5 on pg 37.

 	- Doug