UN Science Workshop in Tokyo, June, 2007

[Rob Seaman]

Hi Kirk,

> Thank you Rick, what you say is right on.  Yes, CDF does
> provide a framework for delivering (and storing) time series
> data.  So, let us not re-invent things that already exist.

I'll take another look at CDF, but the web page (http:// 
cdf.gsfc.nasa.gov) reminds one of the opaqueness of the standard - it  
practically smacks you in the nose with its statement that the format  
is defined through a library implementation.  This is almost the  
opposite of an XML standard like VOEvent, and is one reason that  
astronomers remain wedded to FITS, a standard defined through the  
refereed literature.

I was not trying to suggest that VOEvent stick its nose in other  
people's bidness, and I'm also not sure that was the point Rick was  
trying to make.  Of all the VO protocols, VOEvent is the most  
egalitarian.  We rely on STC, on tables, on URIs, on UCDs, on RTML,  
etc. and so forth.  There is no reason a VOEvent packet might not  
reference a CDF file.  A VOEvent packet is a publication relying on  
several underlying standards.  The publish/subscribe paradigm is  
specifically a way to gather data/metadata/information from diverse  
sources for distribution to an interoperating set of applications  
supporting one or more common communities.  A CDF file has no  
inherent notion of citations, for instance.

>  And of course VOEvent is obviously not the same as STC,
> nor is VOEvent supposed to be a data format for time series.

VOEvent is "supposed to be" whatever compliant projects use it for.   
Expressing a time series or other complex data structure in XML will  
often not be the "best" choice according to various metrics.  It will  
almost always be a viable choice except for applications that are  
strongly resource bounded.

> I think someone in this thread asked about our level
> of interaction with the heliophysics research community.
> I know that there are a lot of interactions, especially in
> my group at NASA-Goddard, where we interact with several of
> the VxO's (Virtual 'anything' Observatories): VSO, VITMO,
> VIRBO, VSPO, VHO, and more (including the SPASE data dictionary
> project).  So, the interaction and synergy is taking place,
> and hopefully will grow in the months and years ahead.

It our job to make sure it does grow.  That was my most fundamental  
point :–)

>  In particular, our CDF group is using NetCDF and OPeNDAP
> (from Peter Fox's UCAR group).  I believe that some of these
> technologies would be quite useful to IVOA.

Indeed.  It is often better to adopt a pre-existing standard than to  
attempt to supplant it.  As I said, VOEvent has been more open to  
this than most VO technologies.  There was immediate consensus to  
adopt RTML.  STC generated a lot of chatter - but really, there was  
no alternative available.  Folks may or may not like the <Params>,  
but we never sought to do anything other than adopt VOTable or raw  
XML technology for the purpose.  The discussions regarding UCDs have  
been raucous, but as long as namespaces are permitted by the  
standard, I suspect we'll attempt to make this work for VOEvent, too.

Rob