[Fwd: VOTimeSeries]

[Doug Tody]

On Mon, 2 Jul 2007, Rob Seaman wrote:

> On Jul 2, 2007, at 3:13 PM, Roy Williams wrote:
> 
> > (1) The message from Enrique Solano shows an existing, compliant, SSA
> > server, together with an example of a time series:
> > http://www.ivoa.net/forum/dm/0706/1344.htm
> > This is built with the internationally accepted SSA data model, with the
> > internationally accepted table format,
> 
> Ah!  Thanks.  I missed that somehow when the discussion immediately moved
> back to DM considerations.  It's not quite as concise as the napkin model,
> but not too bad.

I was about to mention this as well, but Roy beat me to it.

Of course Enrique's data is a real example and not just a napkin sketch.
Nonetheless it is not much more complex than some of the napkin sketches
we have seen.

A simple time series instance can look very simple even if it is based
on a more complex formal data model.  This is why the napkin sketches
always look much simpler than a formal Spectrum data model specification.

> This example undoubtedly doesn't cover the full range of
> VOTable flavors that could be returned in response to different queries.
> To support this we would likely be obliged to add full VOTable support.

Why?  Why not just embed the serialized timeSeries instance in the
VOEvent, and then extract it and feed to a standard VOTable library?
(or just point to it if it is external).  If VOEvent is a real container,
rather than an uber-model which will ultimately grow to encompass
all astronomy data, then you should be able to do this sort of thing.
You should decide what is core to VOEvent, and rely upon interoperability
with other standards for the stuff on the edges.

> A productive avenue might be to combine a simple native VOEvent time
> series element to cover the proverbial 90% of the cases, and encourage
> using <References> to external SSA compliant VOTables for the more
> demanding 10%.

It depends entirely upon what it is being used for, and whether or not
"real data" is required.  In an extreme case you might merely compute
some standard metrics or moments for the time series, and include those
in the event, accessing some external dataset for more complex analysis.

	- Doug