Day 2, at Wellington Park hotel

Use case: in the data stream from a gravitational-wave 
detector, find transients due to stellar collapses, 
merging neutron stars etc.

1.  There is a library of theoretical shapes for the 
    signals of interesting events. Use this to make matched 
    filter-kernels to search the data stream.

2.  Find and record promising events.

2a. Possibly alert other observatories for follow-up, 
    if events are found in real time.

3.  Identify source in catalogues of other observatory.

4.  Get masses, spins of stars involved.

5.  Correlate with other wavebands.

5a. Correlate with other known recent events, e.g. 
    gamma-ray bursts.

Notes:

 - There are many models to be used as the matched filter, 
   so many seraches need to be run, possibly in parallel.  
   This could be done on a compute grid, but is a continuous 
   process so is not suited to a shared grid.

 - The search area for identifying the source is around a 
   square degree.

 - Possible cross-identifications can be galaxies out to ~1Gpc, 
   so there is a vast amount of searching to do.  This may need 
   the data grid.

 - Parameters of the event come from the best-matching model.

 - The best chance of matches is from gamma-ray bursts, posibly 
   from SWIFT data.

 - Chance of reliable, serendipitous matches is tiny; a hopeful 
   search of all catalogues may not be useful.


Interesting point: RGM doesn't want any compute grid.  But many 
use cases may reveal a real need for one.



Data policy and curation
------------------------
PPARC policy is that Astrogrid doesn't fund data centres.  
Archives do not have guaranteed long-term funding. ESA archives 
ditto.  Publishers of electronic journal don't want long-term 
maintence load of their archives.

Data are fragile.  Lifetime ~ 5years.

PPARC has no data policy, no legal obligation to curation.

Problem with readying old archives for transparent usage on 
grid: tension between safe access for average users (who need 
not to be tricked by subtle problems of quality) and greater 
data volume for power users who can deal with tricks and traps.

David Giaretta: first need is for a policy on curation.


Metadata
--------
Francoise Genova: data centres have crucial role in 
_collaborating_ to define metadata. It would be bad to let 
single institutions to define their own metadata. Talks 
between data centres are more important than community 
consultation. BUT commnity input is also important to 
define requirements. Validation: needs to be done "close 
to the knowledge".  That may mean at data source, not at 
data centre.


Other
-----
PPARC will have an AO for grad students for grid work 
around the end of February.

Irish computational grid: 3 nodes in Eire running by end of 
March; may add a node in Armagh.