results victoria meeting and roadmap [SNAP] parameter data model

Paul Harrison pharriso at eso.org
Fri Jun 2 01:50:08 PDT 2006


Hi,

I did not attend any of the theory sessions in Victoria, but Gerard's  
summary has prompted me to draw your attention to some possible  
crossover with the work of the Registry and Grid WGs.

The most obvious overlap is in the area described by Frédéric Boone's  
presentation on a parameter data model, where the Registry WG is  
defining a general model of an "application" including its parameters  
http://www.ivoa.net/twiki/bin/view/IVOA/RegDMApplications, which is  
based on the Common Execution Architecture (CEA) work within  
Astrogrid. CEA was originally designed to "wrap" legacy code as  
asynchronously callable web services (following the UWS pattern of  
the Grid WG http://www.ivoa.net/twiki/bin/view/IVOA/ 
IvoaGridAndWebServices#Asynchronous_activities_proposal)  with  
parameter lists that were described in the registry so that user  
interfaces could be dynamically generated. Astrogrid has a mature  
implementation of CEA which includes

* servers that can quickly be used to wrap a legacy commandline  
application - http://software.astrogrid.org/components/cea/index.html
* a workflow component that can be used to run CEA applications -  
http://software.astrogrid.org/components/jes/index.html
* a desktop client that can call the (remote) applications via a  
dynamic GUI or via your favourite scripting language - http:// 
software.astrogrid.org/userdocs/workbench.html

The CEA parameter model was intended to be as general as possible and  
I think that it would already be suitable for use as part of SNAP,  
though I would be interested in working with you to add any features  
that you thought necessary. In addition, whilst the access to pre- 
computed theoretical data sets might be considered to be "Data  
Access", I can see that there are use cases where the theoretician  
wants to instruct a service to run a program to generate new data  
with new parameters, which would fit better with the "Execution"  
pattern being developed as UWS from the Grid WG.

Paul Harrison



More information about the theory mailing list