philosophical question about VOEvent

[Rob Seaman]

I was attending the Eventful Universe symposium last week and having  
similarly philosophically "surreal" (quote from A. Becker)  
discussions.  The last slide of the final summary presentation  
compared yours truly to Parmenides, the ancient philosopher most tied  
to metaphysical ontology.

In any event, the time domain remains a hot ticket and several major  
players received a (hopefully non-fatal) dose of VOEvent exposure.

Roy said:

> the schema is ostensibly divided into "philosophical" sections: Who,  
> What, Wherewhen, Why etc, but there are also "structural" sections

I see them all as structural and pragmatic, but that is just my own  
personal philosophy.

On Mar 19, 2010, at 11:15 AM, Matthew Graham wrote:

> On Mar 19, 2010, at 11:09 AM, Roy Williams wrote:
>
>>> So use "probability" in <Why> to report your "percentage measure  
>>> of fit" and put the proper details of how likelihood was  
>>> "factually measured" in the <What> section.
>>
>> But my imaginary event author does not have a "probability"  
>> measure, that is the point! All he has is the chi-squared X.
>
> So they just don't report a probability.

VOEvent is a pragmatic standard, not a Platonic ideal.  We are working  
on v2.0.  Perhaps we might discuss adding an (optional) chi-square  
attribute to an <Inference>?  This seems like a very useful feature!

>> I wonder if we can change the definition of the <What> section  
>> slightly, replacing this:
>> <What> was factually measured or observed to occur
>> by this:
>> <What> was factually measured or observed to occur, or other  
>> information relevant to the event,
>>
>> It would make life much easier for the implementors of event  
>> services!
>
> But then they can put stuff in the <What> section that properly  
> belongs in the <WhereWhen> or <How> or <Why>.
>
> Why not just scrap the whole v1.0 data model for v2.0 since it  
> clearly is a bad idea.

As the king of hyperbole, I think I recognize such when I see it.

VOEvent v1.1 has been extremely successful.  We were, however,  
discussing the desirable (if delayed) introduction of v2.0 features at  
the very first VOEvent workshop.  v2.0 is an evolutionary improvement  
to the standard.  The most basic requirement for evolution is a  
species that has proven itself viable, only then can new variations  
arise and be selected.

The structural architecture of VOEvent is similar to any other  
standard - a tug-of-war between constraint and flexibility.  All  
information in the parent <VOEvent> element is pertinent to  
characterizing the event.  If we were to permit anything and  
everything to be dumped into <What>, then we would be simply shifting  
the problem to a different element and it would turn into the <What>  
standard.

	<Who> is for author's provenance
	<What> is for empirical measurements (including time series of same)
	<WhereWhen> is for targeting in spacetime (including ephemerides)
	<How> is for instrumental signature
	<Why> is for scientific characterization

(Roughly, <What> is the dependent variables, and <WhereWhen> is the  
independent variables.)

It is certainly useful to consider how best to get these five fingers  
of the same hand to work together productively - but part of that is  
to avoid placing undue emphasis on any single finger.  (Strangely  
enough, I have a perfect graphic in mind for that :-)

<What> was granted a flexibility of representation through ad hoc  
grouped params that the other elements don't enjoy.  It is natural for  
someone contributing the heavy lifting of designing specific packets  
for new VOEvent customers to find this attractive.  However, folks who  
have designed and handled lots of FITS data might recognize the sirens  
sweetly singing.

What new requirements are being revealed by these new data sets?  How  
best should we meet them in a coherently revised VOEvent standard/ 
schema?

Rob