Question: harvesting managed vs. all resource records

[Ray Plante]

Hi Aurelien,

Thanks for your comments!

On Mon, 4 Apr 2005, Aurelien Stebe wrote:
> I don't understand why we would need to know where a resource is coming 
> from ?

I take it from your referenence to managed resources below that you get 
the understand the need for harvesting only resources records managed by 
the harvestee.  (For others, see short explanation at 
http://www.ivoa.net/twiki/bin/view/IVOA/HarvestAll.)

In general, we need to know which registry a resource record originates 
from so that we can trace back and correct problems with individual 
records.  For example, if a record is not compliant in its encoding (a 
responsibility of the publishing registry), we can contact the curator of 
that registry and let him/her know.  Tracking down encoding and content 
problems is something we're in the thick of here in the NVO.  

> Except for cases when a resource is updated twice the same day.
> For this, the "updated" attribute could be made a "datetime" instead of 
> "date",
> or we could force all registry to support the "datetime" granularity for 
> the OAI interface.

In practice, I would worry about clock skews when comparing dates too 
finely, but perhaps the "xs:date" type imposes arbitrarily too coarse a 
granularity.

OAI requires the use of UTC in all its times (I actually haven't check to 
see if we comply with this).  If full seconds granularity is given, it 
enforces (through the schema) that the special UTC designator (a trailing 
Z) be used.  If we change the VOResource updated attribute type to all 
finer granularity, we should probably impose this same restriction.  

cheers,
Ray