Drop SkyService?

[Thomas McGlynn]

While I like the idea of having a DataCollection object
with coverage elements, I don't think its quite enough...

E.g., suppose we register the ROSAT data collection at
the HEASARC, with nice numbers for spatial, spectral, and
temporal resolution appropriate to the collection.
Now someone else builds a service, say an object detection
algorithm, which can use our collection.  Now if the service
simply inherits the characteristics of the linked to collections,
it may indicate that this service is capable of temporal
reolution of 1 ms -- since that's what ROSAT data could
be analyzed at in some contexts.  However, that's almost
certainly wrong. The service very likely combines entire
ROSAT observations and then looks at data, so if we were
to give it a temporal resolution it might be 10 Ksec.

Of course a service specifically designed to look for short term temporal
variability in X-ray data would split up the obsevation.  Perhaps
it could be characterized as having 10 second resolution...
Where do I put in the metadata this service allows one to look
for variability on a given timescale.

Similarly, there may be some change in the characteristics of the
data midway duing the mission.  The entire mission lifetime is
given as the temporal coverage of the dataset.  However some service
might work only with data early in the mission, e.g., before the
particle background got too high.  So the temporal coverage for the
service is different from the temporal coverage of the dataset.

Or again there might be gamma ray source detection algorithm that
works nicely outside the galactic plane but doesn't work well
in that high background region.

So services can have coverage limits that are independent of the datasets
they use. It may be that providing this is just not feasible in general, and for
the very simple services we are looking at right now -- which
deliver existing datasets with little or no processing -- this problem
may not be significant.

I guess I don't have any problem with the suggestion for right now, but
I'd hope that approach taken wouldn't foreclose on services having coverage
data eventually.


	Regards,
	Tom