Registry usecases

[Roy Williams]

Niall

Thank you for these excellent use cases. Maybe I can present some
components and their relationshios that might make solutions.

First, I am assuming a single virtual repository that actually
consists of a collection of individually published repositories. Each
major data center, and others, would present their datasets by a
standard, harvestable protocol, and these can be collected by regular
robotic harvesting into one or more collector repositories. Think of
the collector as being run by NVO, AVO, etc, and a repository as being
run by Vizier, MAST, CfA, etc. From the repository, we find out about
a dataset, eg 2MASS, Rosat, etc.

> 1> A SN goes off but isn't detected until well after its peak  I
want to
> get all IR, Optical, and UV data (derived photometry or images or
> spectra) from any data source that may have been observing it during
a
> specified time period that has a chanse of detecting/resolving it.
This
> needs to reject all sky surveys that didnt observe during this
period of
> time (e.g. DSS), but should include any pointed observations that
might
> have done this.

Probably a good strategy for this case is to ask the collector for
datasets with the word "synoptic" in the descriptive text of each
dataset -- surely these are the only ones that will have data from 2
weeks ago? If the repository chose to exhibit the space/time metadata
description, then we can ask it (or the collector) to check if we have
the right overlap before even contacting the candidate dataset.

We hope that the datasets themselves have some NVO services running.
We can imagine that the Cone Search version 2 will include a time
window, so that a single cone/time serach can be sent automatically to
many candidate datasets. For the others, we do what we do now -- spend
human time figuring out how their idiosyncratic interface works, and
then pose the query we want to pose.

> 2> I have discovered a subclass of dwarf galaxies that fall into a
> particular color, color, absolute V mangeude space.  I need to get
> colors U-K and R-H (either colors, magnetudes, or data from which
> mangetudes can be extracted in that order) for dwarf galaxies with
known
> V mangetudes and redshift or absolute V mangetudes to make up a
> candidate list for followup observations. So for this I would like
to
> know which services that provide the colors I need are based on
other
> data in the VO, so I don't redo the color measurement when I don't
have
> to (i.e. allow me to discover the region where a the Sloan - 2Mass
cross
> corelation color service and not Sloan and 2Mass data services where
for
> areas covered by Sloan, and then to fall back onto images from 2Mass
and
> a different optical survey for which no cross corelation exists
where
> Sloan didn't observe.  Also it shouldn't tell me to try the South
Pole
> observatory which cannot observe during the time the SN went off).

I'm not sure what are the dependancies you are talking about. Perhaps
you could define more closely this "Sloan - 2Mass cross correlation
color service"?

But the part about catalogs that have specific datatypes is pretty
good in Vizier -- you select the UCDs for the attributes you want, and
ask Vizier which catalogs have those attributes.

> 3> I want to make the highest S/N image of a region of the sky in
the V
> band from all images available to study a very distant galaxy
cluster's
> structure.  For this I need all independant observations (e.g. not
based
> on the same raw data) at V of this area of the sky that have a
> resolution and pixel size on the sky of at least 1 and 0.5
arcseconds
> respectivly so I can resample and combine them.

This wants metadata about the survey itself (sky coverage), plus
metadata about an image survey (resolution, wavebands). Probably the
"independent observation" part you would have to check carefully
yourself, I wouldn't want to trust a computer with that.

Roy