registry-updatable data

[Ray Plante]

On Thu, 7 Apr 2005, KevinBenson wrote:
> Okay I don't have any real objection to this.
> If IVOA curators (or group of people from projects) are willing to sit down
> and all agree on this verificationLevel and the original registry sets this
> attribute up with that value.  But earlier e-mails were applying that any
> registry can have there own verification level for the same record.  

The problem is that I'm happy to set the verificationLevel of records that 
*I* create to the highest value, because I do everything perfectly always.  

The idea is that every registry can set its own standards for what is a 
quality record.  Application developers will naturally migrate toward 
registries that enforce a high standard.  Underlying this is the fact that 
it takes human effort to attach the highest rating.  So if the originator 
is not the one to rate the record (because they are biased), who does.  Do 
we appoint a quality czar to ensure uniform ratings?  That's not really 
necessary (if feasible).  Let every full registry curator care for its 
collection.  

The fact that applications may behave slightly differently depending on 
which registry is used is going to be a fact of life, if only for latency 
reasons.  This is not something we need to worry about.  Where differences 
will be more important are services that produce scientific results 
directly, like mosaicing or sextractor services.  Even here differences 
are a good thing if it causes us to understand how the method affects the 
answer.  

All this said, we have discussed in the NVO the idea of the IVOA 
publishing recommended standards for quality to help promote more 
uniformity.  However, a lot of it would be subjective (are there enough 
sentences in this <description> tag?).  Are we really served by a single 
view of quality in this case?


> Hence
> our search interface will not work because it will bring back records
> inconsistently.  This only would work if Datascope is ever tied to lets say
> a Carnivore registry, if you change it to stsci or Astrogrid which might
> have put it at "0" because they have not got around to setting
> verificationLevels.  Then Datascope is broken and inconsistent between our
> registries.

Note: you will only get inconsistant results if you include this 
stamping/recordCuration information in the query.  DataScope will prefer 
to use the registry that is best maintained.  

cheers,
Ray